Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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#include <linux/static_call.h>
31#include <linux/memcontrol.h>
32#include <linux/cfi.h>
33#include <asm/rqspinlock.h>
34
35struct bpf_verifier_env;
36struct bpf_verifier_log;
37struct perf_event;
38struct bpf_prog;
39struct bpf_prog_aux;
40struct bpf_map;
41struct bpf_arena;
42struct sock;
43struct seq_file;
44struct btf;
45struct btf_type;
46struct exception_table_entry;
47struct seq_operations;
48struct bpf_iter_aux_info;
49struct bpf_local_storage;
50struct bpf_local_storage_map;
51struct kobject;
52struct mem_cgroup;
53struct module;
54struct bpf_func_state;
55struct ftrace_ops;
56struct cgroup;
57struct bpf_token;
58struct user_namespace;
59struct super_block;
60struct inode;
61
62extern struct idr btf_idr;
63extern spinlock_t btf_idr_lock;
64extern struct kobject *btf_kobj;
65extern struct bpf_mem_alloc bpf_global_ma, bpf_global_percpu_ma;
66extern bool bpf_global_ma_set;
67
68typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64);
69typedef int (*bpf_iter_init_seq_priv_t)(void *private_data,
70 struct bpf_iter_aux_info *aux);
71typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
72typedef unsigned int (*bpf_func_t)(const void *,
73 const struct bpf_insn *);
74struct bpf_iter_seq_info {
75 const struct seq_operations *seq_ops;
76 bpf_iter_init_seq_priv_t init_seq_private;
77 bpf_iter_fini_seq_priv_t fini_seq_private;
78 u32 seq_priv_size;
79};
80
81/* map is generic key/value storage optionally accessible by eBPF programs */
82struct bpf_map_ops {
83 /* funcs callable from userspace (via syscall) */
84 int (*map_alloc_check)(union bpf_attr *attr);
85 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
86 void (*map_release)(struct bpf_map *map, struct file *map_file);
87 void (*map_free)(struct bpf_map *map);
88 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
89 void (*map_release_uref)(struct bpf_map *map);
90 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
91 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr,
92 union bpf_attr __user *uattr);
93 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key,
94 void *value, u64 flags);
95 int (*map_lookup_and_delete_batch)(struct bpf_map *map,
96 const union bpf_attr *attr,
97 union bpf_attr __user *uattr);
98 int (*map_update_batch)(struct bpf_map *map, struct file *map_file,
99 const union bpf_attr *attr,
100 union bpf_attr __user *uattr);
101 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr,
102 union bpf_attr __user *uattr);
103
104 /* funcs callable from userspace and from eBPF programs */
105 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
106 long (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
107 long (*map_delete_elem)(struct bpf_map *map, void *key);
108 long (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
109 long (*map_pop_elem)(struct bpf_map *map, void *value);
110 long (*map_peek_elem)(struct bpf_map *map, void *value);
111 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu);
112
113 /* funcs called by prog_array and perf_event_array map */
114 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
115 int fd);
116 /* If need_defer is true, the implementation should guarantee that
117 * the to-be-put element is still alive before the bpf program, which
118 * may manipulate it, exists.
119 */
120 void (*map_fd_put_ptr)(struct bpf_map *map, void *ptr, bool need_defer);
121 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
122 u32 (*map_fd_sys_lookup_elem)(void *ptr);
123 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
124 struct seq_file *m);
125 int (*map_check_btf)(const struct bpf_map *map,
126 const struct btf *btf,
127 const struct btf_type *key_type,
128 const struct btf_type *value_type);
129
130 /* Prog poke tracking helpers. */
131 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
132 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
133 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
134 struct bpf_prog *new);
135
136 /* Direct value access helpers. */
137 int (*map_direct_value_addr)(const struct bpf_map *map,
138 u64 *imm, u32 off);
139 int (*map_direct_value_meta)(const struct bpf_map *map,
140 u64 imm, u32 *off);
141 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
142 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp,
143 struct poll_table_struct *pts);
144 unsigned long (*map_get_unmapped_area)(struct file *filep, unsigned long addr,
145 unsigned long len, unsigned long pgoff,
146 unsigned long flags);
147
148 /* Functions called by bpf_local_storage maps */
149 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap,
150 void *owner, u32 size);
151 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap,
152 void *owner, u32 size);
153 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner);
154
155 /* Misc helpers.*/
156 long (*map_redirect)(struct bpf_map *map, u64 key, u64 flags);
157
158 /* map_meta_equal must be implemented for maps that can be
159 * used as an inner map. It is a runtime check to ensure
160 * an inner map can be inserted to an outer map.
161 *
162 * Some properties of the inner map has been used during the
163 * verification time. When inserting an inner map at the runtime,
164 * map_meta_equal has to ensure the inserting map has the same
165 * properties that the verifier has used earlier.
166 */
167 bool (*map_meta_equal)(const struct bpf_map *meta0,
168 const struct bpf_map *meta1);
169
170
171 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env,
172 struct bpf_func_state *caller,
173 struct bpf_func_state *callee);
174 long (*map_for_each_callback)(struct bpf_map *map,
175 bpf_callback_t callback_fn,
176 void *callback_ctx, u64 flags);
177
178 u64 (*map_mem_usage)(const struct bpf_map *map);
179
180 /* BTF id of struct allocated by map_alloc */
181 int *map_btf_id;
182
183 /* bpf_iter info used to open a seq_file */
184 const struct bpf_iter_seq_info *iter_seq_info;
185};
186
187enum {
188 /* Support at most 11 fields in a BTF type */
189 BTF_FIELDS_MAX = 11,
190};
191
192enum btf_field_type {
193 BPF_SPIN_LOCK = (1 << 0),
194 BPF_TIMER = (1 << 1),
195 BPF_KPTR_UNREF = (1 << 2),
196 BPF_KPTR_REF = (1 << 3),
197 BPF_KPTR_PERCPU = (1 << 4),
198 BPF_KPTR = BPF_KPTR_UNREF | BPF_KPTR_REF | BPF_KPTR_PERCPU,
199 BPF_LIST_HEAD = (1 << 5),
200 BPF_LIST_NODE = (1 << 6),
201 BPF_RB_ROOT = (1 << 7),
202 BPF_RB_NODE = (1 << 8),
203 BPF_GRAPH_NODE = BPF_RB_NODE | BPF_LIST_NODE,
204 BPF_GRAPH_ROOT = BPF_RB_ROOT | BPF_LIST_HEAD,
205 BPF_REFCOUNT = (1 << 9),
206 BPF_WORKQUEUE = (1 << 10),
207 BPF_UPTR = (1 << 11),
208 BPF_RES_SPIN_LOCK = (1 << 12),
209};
210
211typedef void (*btf_dtor_kfunc_t)(void *);
212
213struct btf_field_kptr {
214 struct btf *btf;
215 struct module *module;
216 /* dtor used if btf_is_kernel(btf), otherwise the type is
217 * program-allocated, dtor is NULL, and __bpf_obj_drop_impl is used
218 */
219 btf_dtor_kfunc_t dtor;
220 u32 btf_id;
221};
222
223struct btf_field_graph_root {
224 struct btf *btf;
225 u32 value_btf_id;
226 u32 node_offset;
227 struct btf_record *value_rec;
228};
229
230struct btf_field {
231 u32 offset;
232 u32 size;
233 enum btf_field_type type;
234 union {
235 struct btf_field_kptr kptr;
236 struct btf_field_graph_root graph_root;
237 };
238};
239
240struct btf_record {
241 u32 cnt;
242 u32 field_mask;
243 int spin_lock_off;
244 int res_spin_lock_off;
245 int timer_off;
246 int wq_off;
247 int refcount_off;
248 struct btf_field fields[];
249};
250
251/* Non-opaque version of bpf_rb_node in uapi/linux/bpf.h */
252struct bpf_rb_node_kern {
253 struct rb_node rb_node;
254 void *owner;
255} __attribute__((aligned(8)));
256
257/* Non-opaque version of bpf_list_node in uapi/linux/bpf.h */
258struct bpf_list_node_kern {
259 struct list_head list_head;
260 void *owner;
261} __attribute__((aligned(8)));
262
263struct bpf_map {
264 const struct bpf_map_ops *ops;
265 struct bpf_map *inner_map_meta;
266#ifdef CONFIG_SECURITY
267 void *security;
268#endif
269 enum bpf_map_type map_type;
270 u32 key_size;
271 u32 value_size;
272 u32 max_entries;
273 u64 map_extra; /* any per-map-type extra fields */
274 u32 map_flags;
275 u32 id;
276 struct btf_record *record;
277 int numa_node;
278 u32 btf_key_type_id;
279 u32 btf_value_type_id;
280 u32 btf_vmlinux_value_type_id;
281 struct btf *btf;
282#ifdef CONFIG_MEMCG
283 struct obj_cgroup *objcg;
284#endif
285 char name[BPF_OBJ_NAME_LEN];
286 struct mutex freeze_mutex;
287 atomic64_t refcnt;
288 atomic64_t usercnt;
289 /* rcu is used before freeing and work is only used during freeing */
290 union {
291 struct work_struct work;
292 struct rcu_head rcu;
293 };
294 atomic64_t writecnt;
295 /* 'Ownership' of program-containing map is claimed by the first program
296 * that is going to use this map or by the first program which FD is
297 * stored in the map to make sure that all callers and callees have the
298 * same prog type, JITed flag and xdp_has_frags flag.
299 */
300 struct {
301 const struct btf_type *attach_func_proto;
302 spinlock_t lock;
303 enum bpf_prog_type type;
304 bool jited;
305 bool xdp_has_frags;
306 } owner;
307 bool bypass_spec_v1;
308 bool frozen; /* write-once; write-protected by freeze_mutex */
309 bool free_after_mult_rcu_gp;
310 bool free_after_rcu_gp;
311 atomic64_t sleepable_refcnt;
312 s64 __percpu *elem_count;
313};
314
315static inline const char *btf_field_type_name(enum btf_field_type type)
316{
317 switch (type) {
318 case BPF_SPIN_LOCK:
319 return "bpf_spin_lock";
320 case BPF_RES_SPIN_LOCK:
321 return "bpf_res_spin_lock";
322 case BPF_TIMER:
323 return "bpf_timer";
324 case BPF_WORKQUEUE:
325 return "bpf_wq";
326 case BPF_KPTR_UNREF:
327 case BPF_KPTR_REF:
328 return "kptr";
329 case BPF_KPTR_PERCPU:
330 return "percpu_kptr";
331 case BPF_UPTR:
332 return "uptr";
333 case BPF_LIST_HEAD:
334 return "bpf_list_head";
335 case BPF_LIST_NODE:
336 return "bpf_list_node";
337 case BPF_RB_ROOT:
338 return "bpf_rb_root";
339 case BPF_RB_NODE:
340 return "bpf_rb_node";
341 case BPF_REFCOUNT:
342 return "bpf_refcount";
343 default:
344 WARN_ON_ONCE(1);
345 return "unknown";
346 }
347}
348
349#if IS_ENABLED(CONFIG_DEBUG_KERNEL)
350#define BPF_WARN_ONCE(cond, format...) WARN_ONCE(cond, format)
351#else
352#define BPF_WARN_ONCE(cond, format...) BUILD_BUG_ON_INVALID(cond)
353#endif
354
355static inline u32 btf_field_type_size(enum btf_field_type type)
356{
357 switch (type) {
358 case BPF_SPIN_LOCK:
359 return sizeof(struct bpf_spin_lock);
360 case BPF_RES_SPIN_LOCK:
361 return sizeof(struct bpf_res_spin_lock);
362 case BPF_TIMER:
363 return sizeof(struct bpf_timer);
364 case BPF_WORKQUEUE:
365 return sizeof(struct bpf_wq);
366 case BPF_KPTR_UNREF:
367 case BPF_KPTR_REF:
368 case BPF_KPTR_PERCPU:
369 case BPF_UPTR:
370 return sizeof(u64);
371 case BPF_LIST_HEAD:
372 return sizeof(struct bpf_list_head);
373 case BPF_LIST_NODE:
374 return sizeof(struct bpf_list_node);
375 case BPF_RB_ROOT:
376 return sizeof(struct bpf_rb_root);
377 case BPF_RB_NODE:
378 return sizeof(struct bpf_rb_node);
379 case BPF_REFCOUNT:
380 return sizeof(struct bpf_refcount);
381 default:
382 WARN_ON_ONCE(1);
383 return 0;
384 }
385}
386
387static inline u32 btf_field_type_align(enum btf_field_type type)
388{
389 switch (type) {
390 case BPF_SPIN_LOCK:
391 return __alignof__(struct bpf_spin_lock);
392 case BPF_RES_SPIN_LOCK:
393 return __alignof__(struct bpf_res_spin_lock);
394 case BPF_TIMER:
395 return __alignof__(struct bpf_timer);
396 case BPF_WORKQUEUE:
397 return __alignof__(struct bpf_wq);
398 case BPF_KPTR_UNREF:
399 case BPF_KPTR_REF:
400 case BPF_KPTR_PERCPU:
401 case BPF_UPTR:
402 return __alignof__(u64);
403 case BPF_LIST_HEAD:
404 return __alignof__(struct bpf_list_head);
405 case BPF_LIST_NODE:
406 return __alignof__(struct bpf_list_node);
407 case BPF_RB_ROOT:
408 return __alignof__(struct bpf_rb_root);
409 case BPF_RB_NODE:
410 return __alignof__(struct bpf_rb_node);
411 case BPF_REFCOUNT:
412 return __alignof__(struct bpf_refcount);
413 default:
414 WARN_ON_ONCE(1);
415 return 0;
416 }
417}
418
419static inline void bpf_obj_init_field(const struct btf_field *field, void *addr)
420{
421 memset(addr, 0, field->size);
422
423 switch (field->type) {
424 case BPF_REFCOUNT:
425 refcount_set((refcount_t *)addr, 1);
426 break;
427 case BPF_RB_NODE:
428 RB_CLEAR_NODE((struct rb_node *)addr);
429 break;
430 case BPF_LIST_HEAD:
431 case BPF_LIST_NODE:
432 INIT_LIST_HEAD((struct list_head *)addr);
433 break;
434 case BPF_RB_ROOT:
435 /* RB_ROOT_CACHED 0-inits, no need to do anything after memset */
436 case BPF_SPIN_LOCK:
437 case BPF_RES_SPIN_LOCK:
438 case BPF_TIMER:
439 case BPF_WORKQUEUE:
440 case BPF_KPTR_UNREF:
441 case BPF_KPTR_REF:
442 case BPF_KPTR_PERCPU:
443 case BPF_UPTR:
444 break;
445 default:
446 WARN_ON_ONCE(1);
447 return;
448 }
449}
450
451static inline bool btf_record_has_field(const struct btf_record *rec, enum btf_field_type type)
452{
453 if (IS_ERR_OR_NULL(rec))
454 return false;
455 return rec->field_mask & type;
456}
457
458static inline void bpf_obj_init(const struct btf_record *rec, void *obj)
459{
460 int i;
461
462 if (IS_ERR_OR_NULL(rec))
463 return;
464 for (i = 0; i < rec->cnt; i++)
465 bpf_obj_init_field(&rec->fields[i], obj + rec->fields[i].offset);
466}
467
468/* 'dst' must be a temporary buffer and should not point to memory that is being
469 * used in parallel by a bpf program or bpf syscall, otherwise the access from
470 * the bpf program or bpf syscall may be corrupted by the reinitialization,
471 * leading to weird problems. Even 'dst' is newly-allocated from bpf memory
472 * allocator, it is still possible for 'dst' to be used in parallel by a bpf
473 * program or bpf syscall.
474 */
475static inline void check_and_init_map_value(struct bpf_map *map, void *dst)
476{
477 bpf_obj_init(map->record, dst);
478}
479
480/* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
481 * forced to use 'long' read/writes to try to atomically copy long counters.
482 * Best-effort only. No barriers here, since it _will_ race with concurrent
483 * updates from BPF programs. Called from bpf syscall and mostly used with
484 * size 8 or 16 bytes, so ask compiler to inline it.
485 */
486static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
487{
488 const long *lsrc = src;
489 long *ldst = dst;
490
491 size /= sizeof(long);
492 while (size--)
493 data_race(*ldst++ = *lsrc++);
494}
495
496/* copy everything but bpf_spin_lock, bpf_timer, and kptrs. There could be one of each. */
497static inline void bpf_obj_memcpy(struct btf_record *rec,
498 void *dst, void *src, u32 size,
499 bool long_memcpy)
500{
501 u32 curr_off = 0;
502 int i;
503
504 if (IS_ERR_OR_NULL(rec)) {
505 if (long_memcpy)
506 bpf_long_memcpy(dst, src, round_up(size, 8));
507 else
508 memcpy(dst, src, size);
509 return;
510 }
511
512 for (i = 0; i < rec->cnt; i++) {
513 u32 next_off = rec->fields[i].offset;
514 u32 sz = next_off - curr_off;
515
516 memcpy(dst + curr_off, src + curr_off, sz);
517 curr_off += rec->fields[i].size + sz;
518 }
519 memcpy(dst + curr_off, src + curr_off, size - curr_off);
520}
521
522static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
523{
524 bpf_obj_memcpy(map->record, dst, src, map->value_size, false);
525}
526
527static inline void copy_map_value_long(struct bpf_map *map, void *dst, void *src)
528{
529 bpf_obj_memcpy(map->record, dst, src, map->value_size, true);
530}
531
532static inline void bpf_obj_swap_uptrs(const struct btf_record *rec, void *dst, void *src)
533{
534 unsigned long *src_uptr, *dst_uptr;
535 const struct btf_field *field;
536 int i;
537
538 if (!btf_record_has_field(rec, BPF_UPTR))
539 return;
540
541 for (i = 0, field = rec->fields; i < rec->cnt; i++, field++) {
542 if (field->type != BPF_UPTR)
543 continue;
544
545 src_uptr = src + field->offset;
546 dst_uptr = dst + field->offset;
547 swap(*src_uptr, *dst_uptr);
548 }
549}
550
551static inline void bpf_obj_memzero(struct btf_record *rec, void *dst, u32 size)
552{
553 u32 curr_off = 0;
554 int i;
555
556 if (IS_ERR_OR_NULL(rec)) {
557 memset(dst, 0, size);
558 return;
559 }
560
561 for (i = 0; i < rec->cnt; i++) {
562 u32 next_off = rec->fields[i].offset;
563 u32 sz = next_off - curr_off;
564
565 memset(dst + curr_off, 0, sz);
566 curr_off += rec->fields[i].size + sz;
567 }
568 memset(dst + curr_off, 0, size - curr_off);
569}
570
571static inline void zero_map_value(struct bpf_map *map, void *dst)
572{
573 bpf_obj_memzero(map->record, dst, map->value_size);
574}
575
576void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
577 bool lock_src);
578void bpf_timer_cancel_and_free(void *timer);
579void bpf_wq_cancel_and_free(void *timer);
580void bpf_list_head_free(const struct btf_field *field, void *list_head,
581 struct bpf_spin_lock *spin_lock);
582void bpf_rb_root_free(const struct btf_field *field, void *rb_root,
583 struct bpf_spin_lock *spin_lock);
584u64 bpf_arena_get_kern_vm_start(struct bpf_arena *arena);
585u64 bpf_arena_get_user_vm_start(struct bpf_arena *arena);
586int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
587
588struct bpf_offload_dev;
589struct bpf_offloaded_map;
590
591struct bpf_map_dev_ops {
592 int (*map_get_next_key)(struct bpf_offloaded_map *map,
593 void *key, void *next_key);
594 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
595 void *key, void *value);
596 int (*map_update_elem)(struct bpf_offloaded_map *map,
597 void *key, void *value, u64 flags);
598 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
599};
600
601struct bpf_offloaded_map {
602 struct bpf_map map;
603 struct net_device *netdev;
604 const struct bpf_map_dev_ops *dev_ops;
605 void *dev_priv;
606 struct list_head offloads;
607};
608
609static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
610{
611 return container_of(map, struct bpf_offloaded_map, map);
612}
613
614static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
615{
616 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
617}
618
619static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
620{
621 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
622 map->ops->map_seq_show_elem;
623}
624
625int map_check_no_btf(const struct bpf_map *map,
626 const struct btf *btf,
627 const struct btf_type *key_type,
628 const struct btf_type *value_type);
629
630bool bpf_map_meta_equal(const struct bpf_map *meta0,
631 const struct bpf_map *meta1);
632
633extern const struct bpf_map_ops bpf_map_offload_ops;
634
635/* bpf_type_flag contains a set of flags that are applicable to the values of
636 * arg_type, ret_type and reg_type. For example, a pointer value may be null,
637 * or a memory is read-only. We classify types into two categories: base types
638 * and extended types. Extended types are base types combined with a type flag.
639 *
640 * Currently there are no more than 32 base types in arg_type, ret_type and
641 * reg_types.
642 */
643#define BPF_BASE_TYPE_BITS 8
644
645enum bpf_type_flag {
646 /* PTR may be NULL. */
647 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS),
648
649 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is
650 * compatible with both mutable and immutable memory.
651 */
652 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS),
653
654 /* MEM points to BPF ring buffer reservation. */
655 MEM_RINGBUF = BIT(2 + BPF_BASE_TYPE_BITS),
656
657 /* MEM is in user address space. */
658 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS),
659
660 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged
661 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In
662 * order to drop this tag, it must be passed into bpf_per_cpu_ptr()
663 * or bpf_this_cpu_ptr(), which will return the pointer corresponding
664 * to the specified cpu.
665 */
666 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS),
667
668 /* Indicates that the argument will be released. */
669 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS),
670
671 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark
672 * unreferenced and referenced kptr loaded from map value using a load
673 * instruction, so that they can only be dereferenced but not escape the
674 * BPF program into the kernel (i.e. cannot be passed as arguments to
675 * kfunc or bpf helpers).
676 */
677 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS),
678
679 /* MEM can be uninitialized. */
680 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS),
681
682 /* DYNPTR points to memory local to the bpf program. */
683 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS),
684
685 /* DYNPTR points to a kernel-produced ringbuf record. */
686 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS),
687
688 /* Size is known at compile time. */
689 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS),
690
691 /* MEM is of an allocated object of type in program BTF. This is used to
692 * tag PTR_TO_BTF_ID allocated using bpf_obj_new.
693 */
694 MEM_ALLOC = BIT(11 + BPF_BASE_TYPE_BITS),
695
696 /* PTR was passed from the kernel in a trusted context, and may be
697 * passed to KF_TRUSTED_ARGS kfuncs or BPF helper functions.
698 * Confusingly, this is _not_ the opposite of PTR_UNTRUSTED above.
699 * PTR_UNTRUSTED refers to a kptr that was read directly from a map
700 * without invoking bpf_kptr_xchg(). What we really need to know is
701 * whether a pointer is safe to pass to a kfunc or BPF helper function.
702 * While PTR_UNTRUSTED pointers are unsafe to pass to kfuncs and BPF
703 * helpers, they do not cover all possible instances of unsafe
704 * pointers. For example, a pointer that was obtained from walking a
705 * struct will _not_ get the PTR_UNTRUSTED type modifier, despite the
706 * fact that it may be NULL, invalid, etc. This is due to backwards
707 * compatibility requirements, as this was the behavior that was first
708 * introduced when kptrs were added. The behavior is now considered
709 * deprecated, and PTR_UNTRUSTED will eventually be removed.
710 *
711 * PTR_TRUSTED, on the other hand, is a pointer that the kernel
712 * guarantees to be valid and safe to pass to kfuncs and BPF helpers.
713 * For example, pointers passed to tracepoint arguments are considered
714 * PTR_TRUSTED, as are pointers that are passed to struct_ops
715 * callbacks. As alluded to above, pointers that are obtained from
716 * walking PTR_TRUSTED pointers are _not_ trusted. For example, if a
717 * struct task_struct *task is PTR_TRUSTED, then accessing
718 * task->last_wakee will lose the PTR_TRUSTED modifier when it's stored
719 * in a BPF register. Similarly, pointers passed to certain programs
720 * types such as kretprobes are not guaranteed to be valid, as they may
721 * for example contain an object that was recently freed.
722 */
723 PTR_TRUSTED = BIT(12 + BPF_BASE_TYPE_BITS),
724
725 /* MEM is tagged with rcu and memory access needs rcu_read_lock protection. */
726 MEM_RCU = BIT(13 + BPF_BASE_TYPE_BITS),
727
728 /* Used to tag PTR_TO_BTF_ID | MEM_ALLOC references which are non-owning.
729 * Currently only valid for linked-list and rbtree nodes. If the nodes
730 * have a bpf_refcount_field, they must be tagged MEM_RCU as well.
731 */
732 NON_OWN_REF = BIT(14 + BPF_BASE_TYPE_BITS),
733
734 /* DYNPTR points to sk_buff */
735 DYNPTR_TYPE_SKB = BIT(15 + BPF_BASE_TYPE_BITS),
736
737 /* DYNPTR points to xdp_buff */
738 DYNPTR_TYPE_XDP = BIT(16 + BPF_BASE_TYPE_BITS),
739
740 /* Memory must be aligned on some architectures, used in combination with
741 * MEM_FIXED_SIZE.
742 */
743 MEM_ALIGNED = BIT(17 + BPF_BASE_TYPE_BITS),
744
745 /* MEM is being written to, often combined with MEM_UNINIT. Non-presence
746 * of MEM_WRITE means that MEM is only being read. MEM_WRITE without the
747 * MEM_UNINIT means that memory needs to be initialized since it is also
748 * read.
749 */
750 MEM_WRITE = BIT(18 + BPF_BASE_TYPE_BITS),
751
752 __BPF_TYPE_FLAG_MAX,
753 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1,
754};
755
756#define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF | DYNPTR_TYPE_SKB \
757 | DYNPTR_TYPE_XDP)
758
759/* Max number of base types. */
760#define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS)
761
762/* Max number of all types. */
763#define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1))
764
765/* function argument constraints */
766enum bpf_arg_type {
767 ARG_DONTCARE = 0, /* unused argument in helper function */
768
769 /* the following constraints used to prototype
770 * bpf_map_lookup/update/delete_elem() functions
771 */
772 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
773 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
774 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
775
776 /* Used to prototype bpf_memcmp() and other functions that access data
777 * on eBPF program stack
778 */
779 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
780 ARG_PTR_TO_ARENA,
781
782 ARG_CONST_SIZE, /* number of bytes accessed from memory */
783 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
784
785 ARG_PTR_TO_CTX, /* pointer to context */
786 ARG_ANYTHING, /* any (initialized) argument is ok */
787 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
788 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
789 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
790 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
791 ARG_PTR_TO_RINGBUF_MEM, /* pointer to dynamically reserved ringbuf memory */
792 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
793 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
794 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */
795 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */
796 ARG_PTR_TO_STACK, /* pointer to stack */
797 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */
798 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */
799 ARG_KPTR_XCHG_DEST, /* pointer to destination that kptrs are bpf_kptr_xchg'd into */
800 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */
801 __BPF_ARG_TYPE_MAX,
802
803 /* Extended arg_types. */
804 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE,
805 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM,
806 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX,
807 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET,
808 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK,
809 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID,
810 /* Pointer to memory does not need to be initialized, since helper function
811 * fills all bytes or clears them in error case.
812 */
813 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | MEM_WRITE | ARG_PTR_TO_MEM,
814 /* Pointer to valid memory of size known at compile time. */
815 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM,
816
817 /* This must be the last entry. Its purpose is to ensure the enum is
818 * wide enough to hold the higher bits reserved for bpf_type_flag.
819 */
820 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT,
821};
822static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
823
824/* type of values returned from helper functions */
825enum bpf_return_type {
826 RET_INTEGER, /* function returns integer */
827 RET_VOID, /* function doesn't return anything */
828 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
829 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */
830 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */
831 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */
832 RET_PTR_TO_MEM, /* returns a pointer to memory */
833 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */
834 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */
835 __BPF_RET_TYPE_MAX,
836
837 /* Extended ret_types. */
838 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE,
839 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET,
840 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK,
841 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON,
842 RET_PTR_TO_RINGBUF_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_RINGBUF | RET_PTR_TO_MEM,
843 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MEM,
844 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID,
845 RET_PTR_TO_BTF_ID_TRUSTED = PTR_TRUSTED | RET_PTR_TO_BTF_ID,
846
847 /* This must be the last entry. Its purpose is to ensure the enum is
848 * wide enough to hold the higher bits reserved for bpf_type_flag.
849 */
850 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT,
851};
852static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
853
854/* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
855 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
856 * instructions after verifying
857 */
858struct bpf_func_proto {
859 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
860 bool gpl_only;
861 bool pkt_access;
862 bool might_sleep;
863 /* set to true if helper follows contract for llvm
864 * attribute bpf_fastcall:
865 * - void functions do not scratch r0
866 * - functions taking N arguments scratch only registers r1-rN
867 */
868 bool allow_fastcall;
869 enum bpf_return_type ret_type;
870 union {
871 struct {
872 enum bpf_arg_type arg1_type;
873 enum bpf_arg_type arg2_type;
874 enum bpf_arg_type arg3_type;
875 enum bpf_arg_type arg4_type;
876 enum bpf_arg_type arg5_type;
877 };
878 enum bpf_arg_type arg_type[5];
879 };
880 union {
881 struct {
882 u32 *arg1_btf_id;
883 u32 *arg2_btf_id;
884 u32 *arg3_btf_id;
885 u32 *arg4_btf_id;
886 u32 *arg5_btf_id;
887 };
888 u32 *arg_btf_id[5];
889 struct {
890 size_t arg1_size;
891 size_t arg2_size;
892 size_t arg3_size;
893 size_t arg4_size;
894 size_t arg5_size;
895 };
896 size_t arg_size[5];
897 };
898 int *ret_btf_id; /* return value btf_id */
899 bool (*allowed)(const struct bpf_prog *prog);
900};
901
902/* bpf_context is intentionally undefined structure. Pointer to bpf_context is
903 * the first argument to eBPF programs.
904 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
905 */
906struct bpf_context;
907
908enum bpf_access_type {
909 BPF_READ = 1,
910 BPF_WRITE = 2
911};
912
913/* types of values stored in eBPF registers */
914/* Pointer types represent:
915 * pointer
916 * pointer + imm
917 * pointer + (u16) var
918 * pointer + (u16) var + imm
919 * if (range > 0) then [ptr, ptr + range - off) is safe to access
920 * if (id > 0) means that some 'var' was added
921 * if (off > 0) means that 'imm' was added
922 */
923enum bpf_reg_type {
924 NOT_INIT = 0, /* nothing was written into register */
925 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
926 PTR_TO_CTX, /* reg points to bpf_context */
927 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
928 PTR_TO_MAP_VALUE, /* reg points to map element value */
929 PTR_TO_MAP_KEY, /* reg points to a map element key */
930 PTR_TO_STACK, /* reg == frame_pointer + offset */
931 PTR_TO_PACKET_META, /* skb->data - meta_len */
932 PTR_TO_PACKET, /* reg points to skb->data */
933 PTR_TO_PACKET_END, /* skb->data + headlen */
934 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
935 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
936 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
937 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
938 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
939 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
940 /* PTR_TO_BTF_ID points to a kernel struct that does not need
941 * to be null checked by the BPF program. This does not imply the
942 * pointer is _not_ null and in practice this can easily be a null
943 * pointer when reading pointer chains. The assumption is program
944 * context will handle null pointer dereference typically via fault
945 * handling. The verifier must keep this in mind and can make no
946 * assumptions about null or non-null when doing branch analysis.
947 * Further, when passed into helpers the helpers can not, without
948 * additional context, assume the value is non-null.
949 */
950 PTR_TO_BTF_ID,
951 PTR_TO_MEM, /* reg points to valid memory region */
952 PTR_TO_ARENA,
953 PTR_TO_BUF, /* reg points to a read/write buffer */
954 PTR_TO_FUNC, /* reg points to a bpf program function */
955 CONST_PTR_TO_DYNPTR, /* reg points to a const struct bpf_dynptr */
956 __BPF_REG_TYPE_MAX,
957
958 /* Extended reg_types. */
959 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE,
960 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET,
961 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON,
962 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK,
963 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
964 * been checked for null. Used primarily to inform the verifier
965 * an explicit null check is required for this struct.
966 */
967 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID,
968
969 /* This must be the last entry. Its purpose is to ensure the enum is
970 * wide enough to hold the higher bits reserved for bpf_type_flag.
971 */
972 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT,
973};
974static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
975
976/* The information passed from prog-specific *_is_valid_access
977 * back to the verifier.
978 */
979struct bpf_insn_access_aux {
980 enum bpf_reg_type reg_type;
981 bool is_ldsx;
982 union {
983 int ctx_field_size;
984 struct {
985 struct btf *btf;
986 u32 btf_id;
987 u32 ref_obj_id;
988 };
989 };
990 struct bpf_verifier_log *log; /* for verbose logs */
991 bool is_retval; /* is accessing function return value ? */
992};
993
994static inline void
995bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
996{
997 aux->ctx_field_size = size;
998}
999
1000static bool bpf_is_ldimm64(const struct bpf_insn *insn)
1001{
1002 return insn->code == (BPF_LD | BPF_IMM | BPF_DW);
1003}
1004
1005static inline bool bpf_pseudo_func(const struct bpf_insn *insn)
1006{
1007 return bpf_is_ldimm64(insn) && insn->src_reg == BPF_PSEUDO_FUNC;
1008}
1009
1010/* Given a BPF_ATOMIC instruction @atomic_insn, return true if it is an
1011 * atomic load or store, and false if it is a read-modify-write instruction.
1012 */
1013static inline bool
1014bpf_atomic_is_load_store(const struct bpf_insn *atomic_insn)
1015{
1016 switch (atomic_insn->imm) {
1017 case BPF_LOAD_ACQ:
1018 case BPF_STORE_REL:
1019 return true;
1020 default:
1021 return false;
1022 }
1023}
1024
1025struct bpf_prog_ops {
1026 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
1027 union bpf_attr __user *uattr);
1028};
1029
1030struct bpf_reg_state;
1031struct bpf_verifier_ops {
1032 /* return eBPF function prototype for verification */
1033 const struct bpf_func_proto *
1034 (*get_func_proto)(enum bpf_func_id func_id,
1035 const struct bpf_prog *prog);
1036
1037 /* return true if 'size' wide access at offset 'off' within bpf_context
1038 * with 'type' (read or write) is allowed
1039 */
1040 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
1041 const struct bpf_prog *prog,
1042 struct bpf_insn_access_aux *info);
1043 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
1044 const struct bpf_prog *prog);
1045 int (*gen_epilogue)(struct bpf_insn *insn, const struct bpf_prog *prog,
1046 s16 ctx_stack_off);
1047 int (*gen_ld_abs)(const struct bpf_insn *orig,
1048 struct bpf_insn *insn_buf);
1049 u32 (*convert_ctx_access)(enum bpf_access_type type,
1050 const struct bpf_insn *src,
1051 struct bpf_insn *dst,
1052 struct bpf_prog *prog, u32 *target_size);
1053 int (*btf_struct_access)(struct bpf_verifier_log *log,
1054 const struct bpf_reg_state *reg,
1055 int off, int size);
1056};
1057
1058struct bpf_prog_offload_ops {
1059 /* verifier basic callbacks */
1060 int (*insn_hook)(struct bpf_verifier_env *env,
1061 int insn_idx, int prev_insn_idx);
1062 int (*finalize)(struct bpf_verifier_env *env);
1063 /* verifier optimization callbacks (called after .finalize) */
1064 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
1065 struct bpf_insn *insn);
1066 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
1067 /* program management callbacks */
1068 int (*prepare)(struct bpf_prog *prog);
1069 int (*translate)(struct bpf_prog *prog);
1070 void (*destroy)(struct bpf_prog *prog);
1071};
1072
1073struct bpf_prog_offload {
1074 struct bpf_prog *prog;
1075 struct net_device *netdev;
1076 struct bpf_offload_dev *offdev;
1077 void *dev_priv;
1078 struct list_head offloads;
1079 bool dev_state;
1080 bool opt_failed;
1081 void *jited_image;
1082 u32 jited_len;
1083};
1084
1085enum bpf_cgroup_storage_type {
1086 BPF_CGROUP_STORAGE_SHARED,
1087 BPF_CGROUP_STORAGE_PERCPU,
1088 __BPF_CGROUP_STORAGE_MAX
1089};
1090
1091#define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
1092
1093/* The longest tracepoint has 12 args.
1094 * See include/trace/bpf_probe.h
1095 */
1096#define MAX_BPF_FUNC_ARGS 12
1097
1098/* The maximum number of arguments passed through registers
1099 * a single function may have.
1100 */
1101#define MAX_BPF_FUNC_REG_ARGS 5
1102
1103/* The argument is a structure. */
1104#define BTF_FMODEL_STRUCT_ARG BIT(0)
1105
1106/* The argument is signed. */
1107#define BTF_FMODEL_SIGNED_ARG BIT(1)
1108
1109struct btf_func_model {
1110 u8 ret_size;
1111 u8 ret_flags;
1112 u8 nr_args;
1113 u8 arg_size[MAX_BPF_FUNC_ARGS];
1114 u8 arg_flags[MAX_BPF_FUNC_ARGS];
1115};
1116
1117/* Restore arguments before returning from trampoline to let original function
1118 * continue executing. This flag is used for fentry progs when there are no
1119 * fexit progs.
1120 */
1121#define BPF_TRAMP_F_RESTORE_REGS BIT(0)
1122/* Call original function after fentry progs, but before fexit progs.
1123 * Makes sense for fentry/fexit, normal calls and indirect calls.
1124 */
1125#define BPF_TRAMP_F_CALL_ORIG BIT(1)
1126/* Skip current frame and return to parent. Makes sense for fentry/fexit
1127 * programs only. Should not be used with normal calls and indirect calls.
1128 */
1129#define BPF_TRAMP_F_SKIP_FRAME BIT(2)
1130/* Store IP address of the caller on the trampoline stack,
1131 * so it's available for trampoline's programs.
1132 */
1133#define BPF_TRAMP_F_IP_ARG BIT(3)
1134/* Return the return value of fentry prog. Only used by bpf_struct_ops. */
1135#define BPF_TRAMP_F_RET_FENTRY_RET BIT(4)
1136
1137/* Get original function from stack instead of from provided direct address.
1138 * Makes sense for trampolines with fexit or fmod_ret programs.
1139 */
1140#define BPF_TRAMP_F_ORIG_STACK BIT(5)
1141
1142/* This trampoline is on a function with another ftrace_ops with IPMODIFY,
1143 * e.g., a live patch. This flag is set and cleared by ftrace call backs,
1144 */
1145#define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6)
1146
1147/* Indicate that current trampoline is in a tail call context. Then, it has to
1148 * cache and restore tail_call_cnt to avoid infinite tail call loop.
1149 */
1150#define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7)
1151
1152/*
1153 * Indicate the trampoline should be suitable to receive indirect calls;
1154 * without this indirectly calling the generated code can result in #UD/#CP,
1155 * depending on the CFI options.
1156 *
1157 * Used by bpf_struct_ops.
1158 *
1159 * Incompatible with FENTRY usage, overloads @func_addr argument.
1160 */
1161#define BPF_TRAMP_F_INDIRECT BIT(8)
1162
1163/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
1164 * bytes on x86.
1165 */
1166enum {
1167#if defined(__s390x__)
1168 BPF_MAX_TRAMP_LINKS = 27,
1169#else
1170 BPF_MAX_TRAMP_LINKS = 38,
1171#endif
1172};
1173
1174struct bpf_tramp_links {
1175 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS];
1176 int nr_links;
1177};
1178
1179struct bpf_tramp_run_ctx;
1180
1181/* Different use cases for BPF trampoline:
1182 * 1. replace nop at the function entry (kprobe equivalent)
1183 * flags = BPF_TRAMP_F_RESTORE_REGS
1184 * fentry = a set of programs to run before returning from trampoline
1185 *
1186 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
1187 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
1188 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
1189 * fentry = a set of program to run before calling original function
1190 * fexit = a set of program to run after original function
1191 *
1192 * 3. replace direct call instruction anywhere in the function body
1193 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
1194 * With flags = 0
1195 * fentry = a set of programs to run before returning from trampoline
1196 * With flags = BPF_TRAMP_F_CALL_ORIG
1197 * orig_call = original callback addr or direct function addr
1198 * fentry = a set of program to run before calling original function
1199 * fexit = a set of program to run after original function
1200 */
1201struct bpf_tramp_image;
1202int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *image_end,
1203 const struct btf_func_model *m, u32 flags,
1204 struct bpf_tramp_links *tlinks,
1205 void *func_addr);
1206void *arch_alloc_bpf_trampoline(unsigned int size);
1207void arch_free_bpf_trampoline(void *image, unsigned int size);
1208int __must_check arch_protect_bpf_trampoline(void *image, unsigned int size);
1209int arch_bpf_trampoline_size(const struct btf_func_model *m, u32 flags,
1210 struct bpf_tramp_links *tlinks, void *func_addr);
1211
1212u64 notrace __bpf_prog_enter_sleepable_recur(struct bpf_prog *prog,
1213 struct bpf_tramp_run_ctx *run_ctx);
1214void notrace __bpf_prog_exit_sleepable_recur(struct bpf_prog *prog, u64 start,
1215 struct bpf_tramp_run_ctx *run_ctx);
1216void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
1217void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
1218typedef u64 (*bpf_trampoline_enter_t)(struct bpf_prog *prog,
1219 struct bpf_tramp_run_ctx *run_ctx);
1220typedef void (*bpf_trampoline_exit_t)(struct bpf_prog *prog, u64 start,
1221 struct bpf_tramp_run_ctx *run_ctx);
1222bpf_trampoline_enter_t bpf_trampoline_enter(const struct bpf_prog *prog);
1223bpf_trampoline_exit_t bpf_trampoline_exit(const struct bpf_prog *prog);
1224
1225struct bpf_ksym {
1226 unsigned long start;
1227 unsigned long end;
1228 char name[KSYM_NAME_LEN];
1229 struct list_head lnode;
1230 struct latch_tree_node tnode;
1231 bool prog;
1232};
1233
1234enum bpf_tramp_prog_type {
1235 BPF_TRAMP_FENTRY,
1236 BPF_TRAMP_FEXIT,
1237 BPF_TRAMP_MODIFY_RETURN,
1238 BPF_TRAMP_MAX,
1239 BPF_TRAMP_REPLACE, /* more than MAX */
1240};
1241
1242struct bpf_tramp_image {
1243 void *image;
1244 int size;
1245 struct bpf_ksym ksym;
1246 struct percpu_ref pcref;
1247 void *ip_after_call;
1248 void *ip_epilogue;
1249 union {
1250 struct rcu_head rcu;
1251 struct work_struct work;
1252 };
1253};
1254
1255struct bpf_trampoline {
1256 /* hlist for trampoline_table */
1257 struct hlist_node hlist;
1258 struct ftrace_ops *fops;
1259 /* serializes access to fields of this trampoline */
1260 struct mutex mutex;
1261 refcount_t refcnt;
1262 u32 flags;
1263 u64 key;
1264 struct {
1265 struct btf_func_model model;
1266 void *addr;
1267 bool ftrace_managed;
1268 } func;
1269 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
1270 * program by replacing one of its functions. func.addr is the address
1271 * of the function it replaced.
1272 */
1273 struct bpf_prog *extension_prog;
1274 /* list of BPF programs using this trampoline */
1275 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
1276 /* Number of attached programs. A counter per kind. */
1277 int progs_cnt[BPF_TRAMP_MAX];
1278 /* Executable image of trampoline */
1279 struct bpf_tramp_image *cur_image;
1280};
1281
1282struct bpf_attach_target_info {
1283 struct btf_func_model fmodel;
1284 long tgt_addr;
1285 struct module *tgt_mod;
1286 const char *tgt_name;
1287 const struct btf_type *tgt_type;
1288};
1289
1290#define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
1291
1292struct bpf_dispatcher_prog {
1293 struct bpf_prog *prog;
1294 refcount_t users;
1295};
1296
1297struct bpf_dispatcher {
1298 /* dispatcher mutex */
1299 struct mutex mutex;
1300 void *func;
1301 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
1302 int num_progs;
1303 void *image;
1304 void *rw_image;
1305 u32 image_off;
1306 struct bpf_ksym ksym;
1307#ifdef CONFIG_HAVE_STATIC_CALL
1308 struct static_call_key *sc_key;
1309 void *sc_tramp;
1310#endif
1311};
1312
1313#ifndef __bpfcall
1314#define __bpfcall __nocfi
1315#endif
1316
1317static __always_inline __bpfcall unsigned int bpf_dispatcher_nop_func(
1318 const void *ctx,
1319 const struct bpf_insn *insnsi,
1320 bpf_func_t bpf_func)
1321{
1322 return bpf_func(ctx, insnsi);
1323}
1324
1325/* the implementation of the opaque uapi struct bpf_dynptr */
1326struct bpf_dynptr_kern {
1327 void *data;
1328 /* Size represents the number of usable bytes of dynptr data.
1329 * If for example the offset is at 4 for a local dynptr whose data is
1330 * of type u64, the number of usable bytes is 4.
1331 *
1332 * The upper 8 bits are reserved. It is as follows:
1333 * Bits 0 - 23 = size
1334 * Bits 24 - 30 = dynptr type
1335 * Bit 31 = whether dynptr is read-only
1336 */
1337 u32 size;
1338 u32 offset;
1339} __aligned(8);
1340
1341enum bpf_dynptr_type {
1342 BPF_DYNPTR_TYPE_INVALID,
1343 /* Points to memory that is local to the bpf program */
1344 BPF_DYNPTR_TYPE_LOCAL,
1345 /* Underlying data is a ringbuf record */
1346 BPF_DYNPTR_TYPE_RINGBUF,
1347 /* Underlying data is a sk_buff */
1348 BPF_DYNPTR_TYPE_SKB,
1349 /* Underlying data is a xdp_buff */
1350 BPF_DYNPTR_TYPE_XDP,
1351};
1352
1353int bpf_dynptr_check_size(u32 size);
1354u32 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr);
1355const void *__bpf_dynptr_data(const struct bpf_dynptr_kern *ptr, u32 len);
1356void *__bpf_dynptr_data_rw(const struct bpf_dynptr_kern *ptr, u32 len);
1357bool __bpf_dynptr_is_rdonly(const struct bpf_dynptr_kern *ptr);
1358int __bpf_dynptr_write(const struct bpf_dynptr_kern *dst, u32 offset,
1359 void *src, u32 len, u64 flags);
1360void *bpf_dynptr_slice_rdwr(const struct bpf_dynptr *p, u32 offset,
1361 void *buffer__opt, u32 buffer__szk);
1362
1363static inline int bpf_dynptr_check_off_len(const struct bpf_dynptr_kern *ptr, u32 offset, u32 len)
1364{
1365 u32 size = __bpf_dynptr_size(ptr);
1366
1367 if (len > size || offset > size - len)
1368 return -E2BIG;
1369
1370 return 0;
1371}
1372
1373#ifdef CONFIG_BPF_JIT
1374int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
1375 struct bpf_trampoline *tr,
1376 struct bpf_prog *tgt_prog);
1377int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
1378 struct bpf_trampoline *tr,
1379 struct bpf_prog *tgt_prog);
1380struct bpf_trampoline *bpf_trampoline_get(u64 key,
1381 struct bpf_attach_target_info *tgt_info);
1382void bpf_trampoline_put(struct bpf_trampoline *tr);
1383int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_funcs);
1384
1385/*
1386 * When the architecture supports STATIC_CALL replace the bpf_dispatcher_fn
1387 * indirection with a direct call to the bpf program. If the architecture does
1388 * not have STATIC_CALL, avoid a double-indirection.
1389 */
1390#ifdef CONFIG_HAVE_STATIC_CALL
1391
1392#define __BPF_DISPATCHER_SC_INIT(_name) \
1393 .sc_key = &STATIC_CALL_KEY(_name), \
1394 .sc_tramp = STATIC_CALL_TRAMP_ADDR(_name),
1395
1396#define __BPF_DISPATCHER_SC(name) \
1397 DEFINE_STATIC_CALL(bpf_dispatcher_##name##_call, bpf_dispatcher_nop_func)
1398
1399#define __BPF_DISPATCHER_CALL(name) \
1400 static_call(bpf_dispatcher_##name##_call)(ctx, insnsi, bpf_func)
1401
1402#define __BPF_DISPATCHER_UPDATE(_d, _new) \
1403 __static_call_update((_d)->sc_key, (_d)->sc_tramp, (_new))
1404
1405#else
1406#define __BPF_DISPATCHER_SC_INIT(name)
1407#define __BPF_DISPATCHER_SC(name)
1408#define __BPF_DISPATCHER_CALL(name) bpf_func(ctx, insnsi)
1409#define __BPF_DISPATCHER_UPDATE(_d, _new)
1410#endif
1411
1412#define BPF_DISPATCHER_INIT(_name) { \
1413 .mutex = __MUTEX_INITIALIZER(_name.mutex), \
1414 .func = &_name##_func, \
1415 .progs = {}, \
1416 .num_progs = 0, \
1417 .image = NULL, \
1418 .image_off = 0, \
1419 .ksym = { \
1420 .name = #_name, \
1421 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
1422 }, \
1423 __BPF_DISPATCHER_SC_INIT(_name##_call) \
1424}
1425
1426#define DEFINE_BPF_DISPATCHER(name) \
1427 __BPF_DISPATCHER_SC(name); \
1428 noinline __bpfcall unsigned int bpf_dispatcher_##name##_func( \
1429 const void *ctx, \
1430 const struct bpf_insn *insnsi, \
1431 bpf_func_t bpf_func) \
1432 { \
1433 return __BPF_DISPATCHER_CALL(name); \
1434 } \
1435 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
1436 struct bpf_dispatcher bpf_dispatcher_##name = \
1437 BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
1438
1439#define DECLARE_BPF_DISPATCHER(name) \
1440 unsigned int bpf_dispatcher_##name##_func( \
1441 const void *ctx, \
1442 const struct bpf_insn *insnsi, \
1443 bpf_func_t bpf_func); \
1444 extern struct bpf_dispatcher bpf_dispatcher_##name;
1445
1446#define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
1447#define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
1448void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
1449 struct bpf_prog *to);
1450/* Called only from JIT-enabled code, so there's no need for stubs. */
1451void bpf_image_ksym_init(void *data, unsigned int size, struct bpf_ksym *ksym);
1452void bpf_image_ksym_add(struct bpf_ksym *ksym);
1453void bpf_image_ksym_del(struct bpf_ksym *ksym);
1454void bpf_ksym_add(struct bpf_ksym *ksym);
1455void bpf_ksym_del(struct bpf_ksym *ksym);
1456int bpf_jit_charge_modmem(u32 size);
1457void bpf_jit_uncharge_modmem(u32 size);
1458bool bpf_prog_has_trampoline(const struct bpf_prog *prog);
1459#else
1460static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
1461 struct bpf_trampoline *tr,
1462 struct bpf_prog *tgt_prog)
1463{
1464 return -ENOTSUPP;
1465}
1466static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
1467 struct bpf_trampoline *tr,
1468 struct bpf_prog *tgt_prog)
1469{
1470 return -ENOTSUPP;
1471}
1472static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
1473 struct bpf_attach_target_info *tgt_info)
1474{
1475 return NULL;
1476}
1477static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
1478#define DEFINE_BPF_DISPATCHER(name)
1479#define DECLARE_BPF_DISPATCHER(name)
1480#define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
1481#define BPF_DISPATCHER_PTR(name) NULL
1482static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
1483 struct bpf_prog *from,
1484 struct bpf_prog *to) {}
1485static inline bool is_bpf_image_address(unsigned long address)
1486{
1487 return false;
1488}
1489static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog)
1490{
1491 return false;
1492}
1493#endif
1494
1495struct bpf_func_info_aux {
1496 u16 linkage;
1497 bool unreliable;
1498 bool called : 1;
1499 bool verified : 1;
1500};
1501
1502enum bpf_jit_poke_reason {
1503 BPF_POKE_REASON_TAIL_CALL,
1504};
1505
1506/* Descriptor of pokes pointing /into/ the JITed image. */
1507struct bpf_jit_poke_descriptor {
1508 void *tailcall_target;
1509 void *tailcall_bypass;
1510 void *bypass_addr;
1511 void *aux;
1512 union {
1513 struct {
1514 struct bpf_map *map;
1515 u32 key;
1516 } tail_call;
1517 };
1518 bool tailcall_target_stable;
1519 u8 adj_off;
1520 u16 reason;
1521 u32 insn_idx;
1522};
1523
1524/* reg_type info for ctx arguments */
1525struct bpf_ctx_arg_aux {
1526 u32 offset;
1527 enum bpf_reg_type reg_type;
1528 struct btf *btf;
1529 u32 btf_id;
1530 u32 ref_obj_id;
1531 bool refcounted;
1532};
1533
1534struct btf_mod_pair {
1535 struct btf *btf;
1536 struct module *module;
1537};
1538
1539struct bpf_kfunc_desc_tab;
1540
1541struct bpf_prog_aux {
1542 atomic64_t refcnt;
1543 u32 used_map_cnt;
1544 u32 used_btf_cnt;
1545 u32 max_ctx_offset;
1546 u32 max_pkt_offset;
1547 u32 max_tp_access;
1548 u32 stack_depth;
1549 u32 id;
1550 u32 func_cnt; /* used by non-func prog as the number of func progs */
1551 u32 real_func_cnt; /* includes hidden progs, only used for JIT and freeing progs */
1552 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
1553 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1554 u32 attach_st_ops_member_off;
1555 u32 ctx_arg_info_size;
1556 u32 max_rdonly_access;
1557 u32 max_rdwr_access;
1558 struct btf *attach_btf;
1559 struct bpf_ctx_arg_aux *ctx_arg_info;
1560 void __percpu *priv_stack_ptr;
1561 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
1562 struct bpf_prog *dst_prog;
1563 struct bpf_trampoline *dst_trampoline;
1564 enum bpf_prog_type saved_dst_prog_type;
1565 enum bpf_attach_type saved_dst_attach_type;
1566 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
1567 bool dev_bound; /* Program is bound to the netdev. */
1568 bool offload_requested; /* Program is bound and offloaded to the netdev. */
1569 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
1570 bool attach_tracing_prog; /* true if tracing another tracing program */
1571 bool func_proto_unreliable;
1572 bool tail_call_reachable;
1573 bool xdp_has_frags;
1574 bool exception_cb;
1575 bool exception_boundary;
1576 bool is_extended; /* true if extended by freplace program */
1577 bool jits_use_priv_stack;
1578 bool priv_stack_requested;
1579 bool changes_pkt_data;
1580 bool might_sleep;
1581 u64 prog_array_member_cnt; /* counts how many times as member of prog_array */
1582 struct mutex ext_mutex; /* mutex for is_extended and prog_array_member_cnt */
1583 struct bpf_arena *arena;
1584 void (*recursion_detected)(struct bpf_prog *prog); /* callback if recursion is detected */
1585 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
1586 const struct btf_type *attach_func_proto;
1587 /* function name for valid attach_btf_id */
1588 const char *attach_func_name;
1589 struct bpf_prog **func;
1590 void *jit_data; /* JIT specific data. arch dependent */
1591 struct bpf_jit_poke_descriptor *poke_tab;
1592 struct bpf_kfunc_desc_tab *kfunc_tab;
1593 struct bpf_kfunc_btf_tab *kfunc_btf_tab;
1594 u32 size_poke_tab;
1595#ifdef CONFIG_FINEIBT
1596 struct bpf_ksym ksym_prefix;
1597#endif
1598 struct bpf_ksym ksym;
1599 const struct bpf_prog_ops *ops;
1600 const struct bpf_struct_ops *st_ops;
1601 struct bpf_map **used_maps;
1602 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
1603 struct btf_mod_pair *used_btfs;
1604 struct bpf_prog *prog;
1605 struct user_struct *user;
1606 u64 load_time; /* ns since boottime */
1607 u32 verified_insns;
1608 int cgroup_atype; /* enum cgroup_bpf_attach_type */
1609 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1610 char name[BPF_OBJ_NAME_LEN];
1611 u64 (*bpf_exception_cb)(u64 cookie, u64 sp, u64 bp, u64, u64);
1612#ifdef CONFIG_SECURITY
1613 void *security;
1614#endif
1615 struct bpf_token *token;
1616 struct bpf_prog_offload *offload;
1617 struct btf *btf;
1618 struct bpf_func_info *func_info;
1619 struct bpf_func_info_aux *func_info_aux;
1620 /* bpf_line_info loaded from userspace. linfo->insn_off
1621 * has the xlated insn offset.
1622 * Both the main and sub prog share the same linfo.
1623 * The subprog can access its first linfo by
1624 * using the linfo_idx.
1625 */
1626 struct bpf_line_info *linfo;
1627 /* jited_linfo is the jited addr of the linfo. It has a
1628 * one to one mapping to linfo:
1629 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
1630 * Both the main and sub prog share the same jited_linfo.
1631 * The subprog can access its first jited_linfo by
1632 * using the linfo_idx.
1633 */
1634 void **jited_linfo;
1635 u32 func_info_cnt;
1636 u32 nr_linfo;
1637 /* subprog can use linfo_idx to access its first linfo and
1638 * jited_linfo.
1639 * main prog always has linfo_idx == 0
1640 */
1641 u32 linfo_idx;
1642 struct module *mod;
1643 u32 num_exentries;
1644 struct exception_table_entry *extable;
1645 union {
1646 struct work_struct work;
1647 struct rcu_head rcu;
1648 };
1649};
1650
1651struct bpf_prog {
1652 u16 pages; /* Number of allocated pages */
1653 u16 jited:1, /* Is our filter JIT'ed? */
1654 jit_requested:1,/* archs need to JIT the prog */
1655 gpl_compatible:1, /* Is filter GPL compatible? */
1656 cb_access:1, /* Is control block accessed? */
1657 dst_needed:1, /* Do we need dst entry? */
1658 blinding_requested:1, /* needs constant blinding */
1659 blinded:1, /* Was blinded */
1660 is_func:1, /* program is a bpf function */
1661 kprobe_override:1, /* Do we override a kprobe? */
1662 has_callchain_buf:1, /* callchain buffer allocated? */
1663 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */
1664 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */
1665 call_get_func_ip:1, /* Do we call get_func_ip() */
1666 tstamp_type_access:1, /* Accessed __sk_buff->tstamp_type */
1667 sleepable:1; /* BPF program is sleepable */
1668 enum bpf_prog_type type; /* Type of BPF program */
1669 enum bpf_attach_type expected_attach_type; /* For some prog types */
1670 u32 len; /* Number of filter blocks */
1671 u32 jited_len; /* Size of jited insns in bytes */
1672 u8 tag[BPF_TAG_SIZE];
1673 struct bpf_prog_stats __percpu *stats;
1674 int __percpu *active;
1675 unsigned int (*bpf_func)(const void *ctx,
1676 const struct bpf_insn *insn);
1677 struct bpf_prog_aux *aux; /* Auxiliary fields */
1678 struct sock_fprog_kern *orig_prog; /* Original BPF program */
1679 /* Instructions for interpreter */
1680 union {
1681 DECLARE_FLEX_ARRAY(struct sock_filter, insns);
1682 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi);
1683 };
1684};
1685
1686struct bpf_array_aux {
1687 /* Programs with direct jumps into programs part of this array. */
1688 struct list_head poke_progs;
1689 struct bpf_map *map;
1690 struct mutex poke_mutex;
1691 struct work_struct work;
1692};
1693
1694struct bpf_link {
1695 atomic64_t refcnt;
1696 u32 id;
1697 enum bpf_link_type type;
1698 const struct bpf_link_ops *ops;
1699 struct bpf_prog *prog;
1700 /* whether BPF link itself has "sleepable" semantics, which can differ
1701 * from underlying BPF program having a "sleepable" semantics, as BPF
1702 * link's semantics is determined by target attach hook
1703 */
1704 bool sleepable;
1705 /* rcu is used before freeing, work can be used to schedule that
1706 * RCU-based freeing before that, so they never overlap
1707 */
1708 union {
1709 struct rcu_head rcu;
1710 struct work_struct work;
1711 };
1712};
1713
1714struct bpf_link_ops {
1715 void (*release)(struct bpf_link *link);
1716 /* deallocate link resources callback, called without RCU grace period
1717 * waiting
1718 */
1719 void (*dealloc)(struct bpf_link *link);
1720 /* deallocate link resources callback, called after RCU grace period;
1721 * if either the underlying BPF program is sleepable or BPF link's
1722 * target hook is sleepable, we'll go through tasks trace RCU GP and
1723 * then "classic" RCU GP; this need for chaining tasks trace and
1724 * classic RCU GPs is designated by setting bpf_link->sleepable flag
1725 */
1726 void (*dealloc_deferred)(struct bpf_link *link);
1727 int (*detach)(struct bpf_link *link);
1728 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
1729 struct bpf_prog *old_prog);
1730 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
1731 int (*fill_link_info)(const struct bpf_link *link,
1732 struct bpf_link_info *info);
1733 int (*update_map)(struct bpf_link *link, struct bpf_map *new_map,
1734 struct bpf_map *old_map);
1735 __poll_t (*poll)(struct file *file, struct poll_table_struct *pts);
1736};
1737
1738struct bpf_tramp_link {
1739 struct bpf_link link;
1740 struct hlist_node tramp_hlist;
1741 u64 cookie;
1742};
1743
1744struct bpf_shim_tramp_link {
1745 struct bpf_tramp_link link;
1746 struct bpf_trampoline *trampoline;
1747};
1748
1749struct bpf_tracing_link {
1750 struct bpf_tramp_link link;
1751 enum bpf_attach_type attach_type;
1752 struct bpf_trampoline *trampoline;
1753 struct bpf_prog *tgt_prog;
1754};
1755
1756struct bpf_raw_tp_link {
1757 struct bpf_link link;
1758 struct bpf_raw_event_map *btp;
1759 u64 cookie;
1760};
1761
1762struct bpf_link_primer {
1763 struct bpf_link *link;
1764 struct file *file;
1765 int fd;
1766 u32 id;
1767};
1768
1769struct bpf_mount_opts {
1770 kuid_t uid;
1771 kgid_t gid;
1772 umode_t mode;
1773
1774 /* BPF token-related delegation options */
1775 u64 delegate_cmds;
1776 u64 delegate_maps;
1777 u64 delegate_progs;
1778 u64 delegate_attachs;
1779};
1780
1781struct bpf_token {
1782 struct work_struct work;
1783 atomic64_t refcnt;
1784 struct user_namespace *userns;
1785 u64 allowed_cmds;
1786 u64 allowed_maps;
1787 u64 allowed_progs;
1788 u64 allowed_attachs;
1789#ifdef CONFIG_SECURITY
1790 void *security;
1791#endif
1792};
1793
1794struct bpf_struct_ops_value;
1795struct btf_member;
1796
1797#define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
1798/**
1799 * struct bpf_struct_ops - A structure of callbacks allowing a subsystem to
1800 * define a BPF_MAP_TYPE_STRUCT_OPS map type composed
1801 * of BPF_PROG_TYPE_STRUCT_OPS progs.
1802 * @verifier_ops: A structure of callbacks that are invoked by the verifier
1803 * when determining whether the struct_ops progs in the
1804 * struct_ops map are valid.
1805 * @init: A callback that is invoked a single time, and before any other
1806 * callback, to initialize the structure. A nonzero return value means
1807 * the subsystem could not be initialized.
1808 * @check_member: When defined, a callback invoked by the verifier to allow
1809 * the subsystem to determine if an entry in the struct_ops map
1810 * is valid. A nonzero return value means that the map is
1811 * invalid and should be rejected by the verifier.
1812 * @init_member: A callback that is invoked for each member of the struct_ops
1813 * map to allow the subsystem to initialize the member. A nonzero
1814 * value means the member could not be initialized. This callback
1815 * is exclusive with the @type, @type_id, @value_type, and
1816 * @value_id fields.
1817 * @reg: A callback that is invoked when the struct_ops map has been
1818 * initialized and is being attached to. Zero means the struct_ops map
1819 * has been successfully registered and is live. A nonzero return value
1820 * means the struct_ops map could not be registered.
1821 * @unreg: A callback that is invoked when the struct_ops map should be
1822 * unregistered.
1823 * @update: A callback that is invoked when the live struct_ops map is being
1824 * updated to contain new values. This callback is only invoked when
1825 * the struct_ops map is loaded with BPF_F_LINK. If not defined, the
1826 * it is assumed that the struct_ops map cannot be updated.
1827 * @validate: A callback that is invoked after all of the members have been
1828 * initialized. This callback should perform static checks on the
1829 * map, meaning that it should either fail or succeed
1830 * deterministically. A struct_ops map that has been validated may
1831 * not necessarily succeed in being registered if the call to @reg
1832 * fails. For example, a valid struct_ops map may be loaded, but
1833 * then fail to be registered due to there being another active
1834 * struct_ops map on the system in the subsystem already. For this
1835 * reason, if this callback is not defined, the check is skipped as
1836 * the struct_ops map will have final verification performed in
1837 * @reg.
1838 * @type: BTF type.
1839 * @value_type: Value type.
1840 * @name: The name of the struct bpf_struct_ops object.
1841 * @func_models: Func models
1842 * @type_id: BTF type id.
1843 * @value_id: BTF value id.
1844 */
1845struct bpf_struct_ops {
1846 const struct bpf_verifier_ops *verifier_ops;
1847 int (*init)(struct btf *btf);
1848 int (*check_member)(const struct btf_type *t,
1849 const struct btf_member *member,
1850 const struct bpf_prog *prog);
1851 int (*init_member)(const struct btf_type *t,
1852 const struct btf_member *member,
1853 void *kdata, const void *udata);
1854 int (*reg)(void *kdata, struct bpf_link *link);
1855 void (*unreg)(void *kdata, struct bpf_link *link);
1856 int (*update)(void *kdata, void *old_kdata, struct bpf_link *link);
1857 int (*validate)(void *kdata);
1858 void *cfi_stubs;
1859 struct module *owner;
1860 const char *name;
1861 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
1862};
1863
1864/* Every member of a struct_ops type has an instance even a member is not
1865 * an operator (function pointer). The "info" field will be assigned to
1866 * prog->aux->ctx_arg_info of BPF struct_ops programs to provide the
1867 * argument information required by the verifier to verify the program.
1868 *
1869 * btf_ctx_access() will lookup prog->aux->ctx_arg_info to find the
1870 * corresponding entry for an given argument.
1871 */
1872struct bpf_struct_ops_arg_info {
1873 struct bpf_ctx_arg_aux *info;
1874 u32 cnt;
1875};
1876
1877struct bpf_struct_ops_desc {
1878 struct bpf_struct_ops *st_ops;
1879
1880 const struct btf_type *type;
1881 const struct btf_type *value_type;
1882 u32 type_id;
1883 u32 value_id;
1884
1885 /* Collection of argument information for each member */
1886 struct bpf_struct_ops_arg_info *arg_info;
1887};
1888
1889enum bpf_struct_ops_state {
1890 BPF_STRUCT_OPS_STATE_INIT,
1891 BPF_STRUCT_OPS_STATE_INUSE,
1892 BPF_STRUCT_OPS_STATE_TOBEFREE,
1893 BPF_STRUCT_OPS_STATE_READY,
1894};
1895
1896struct bpf_struct_ops_common_value {
1897 refcount_t refcnt;
1898 enum bpf_struct_ops_state state;
1899};
1900
1901#if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
1902/* This macro helps developer to register a struct_ops type and generate
1903 * type information correctly. Developers should use this macro to register
1904 * a struct_ops type instead of calling __register_bpf_struct_ops() directly.
1905 */
1906#define register_bpf_struct_ops(st_ops, type) \
1907 ({ \
1908 struct bpf_struct_ops_##type { \
1909 struct bpf_struct_ops_common_value common; \
1910 struct type data ____cacheline_aligned_in_smp; \
1911 }; \
1912 BTF_TYPE_EMIT(struct bpf_struct_ops_##type); \
1913 __register_bpf_struct_ops(st_ops); \
1914 })
1915#define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
1916bool bpf_struct_ops_get(const void *kdata);
1917void bpf_struct_ops_put(const void *kdata);
1918int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff);
1919int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
1920 void *value);
1921int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks,
1922 struct bpf_tramp_link *link,
1923 const struct btf_func_model *model,
1924 void *stub_func,
1925 void **image, u32 *image_off,
1926 bool allow_alloc);
1927void bpf_struct_ops_image_free(void *image);
1928static inline bool bpf_try_module_get(const void *data, struct module *owner)
1929{
1930 if (owner == BPF_MODULE_OWNER)
1931 return bpf_struct_ops_get(data);
1932 else
1933 return try_module_get(owner);
1934}
1935static inline void bpf_module_put(const void *data, struct module *owner)
1936{
1937 if (owner == BPF_MODULE_OWNER)
1938 bpf_struct_ops_put(data);
1939 else
1940 module_put(owner);
1941}
1942int bpf_struct_ops_link_create(union bpf_attr *attr);
1943
1944#ifdef CONFIG_NET
1945/* Define it here to avoid the use of forward declaration */
1946struct bpf_dummy_ops_state {
1947 int val;
1948};
1949
1950struct bpf_dummy_ops {
1951 int (*test_1)(struct bpf_dummy_ops_state *cb);
1952 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2,
1953 char a3, unsigned long a4);
1954 int (*test_sleepable)(struct bpf_dummy_ops_state *cb);
1955};
1956
1957int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr,
1958 union bpf_attr __user *uattr);
1959#endif
1960int bpf_struct_ops_desc_init(struct bpf_struct_ops_desc *st_ops_desc,
1961 struct btf *btf,
1962 struct bpf_verifier_log *log);
1963void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map);
1964void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc);
1965#else
1966#define register_bpf_struct_ops(st_ops, type) ({ (void *)(st_ops); 0; })
1967static inline bool bpf_try_module_get(const void *data, struct module *owner)
1968{
1969 return try_module_get(owner);
1970}
1971static inline void bpf_module_put(const void *data, struct module *owner)
1972{
1973 module_put(owner);
1974}
1975static inline int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff)
1976{
1977 return -ENOTSUPP;
1978}
1979static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
1980 void *key,
1981 void *value)
1982{
1983 return -EINVAL;
1984}
1985static inline int bpf_struct_ops_link_create(union bpf_attr *attr)
1986{
1987 return -EOPNOTSUPP;
1988}
1989static inline void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map)
1990{
1991}
1992
1993static inline void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc)
1994{
1995}
1996
1997#endif
1998
1999int bpf_prog_ctx_arg_info_init(struct bpf_prog *prog,
2000 const struct bpf_ctx_arg_aux *info, u32 cnt);
2001
2002#if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM)
2003int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
2004 int cgroup_atype);
2005void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog);
2006#else
2007static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
2008 int cgroup_atype)
2009{
2010 return -EOPNOTSUPP;
2011}
2012static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog)
2013{
2014}
2015#endif
2016
2017struct bpf_array {
2018 struct bpf_map map;
2019 u32 elem_size;
2020 u32 index_mask;
2021 struct bpf_array_aux *aux;
2022 union {
2023 DECLARE_FLEX_ARRAY(char, value) __aligned(8);
2024 DECLARE_FLEX_ARRAY(void *, ptrs) __aligned(8);
2025 DECLARE_FLEX_ARRAY(void __percpu *, pptrs) __aligned(8);
2026 };
2027};
2028
2029#define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
2030#define MAX_TAIL_CALL_CNT 33
2031
2032/* Maximum number of loops for bpf_loop and bpf_iter_num.
2033 * It's enum to expose it (and thus make it discoverable) through BTF.
2034 */
2035enum {
2036 BPF_MAX_LOOPS = 8 * 1024 * 1024,
2037 BPF_MAX_TIMED_LOOPS = 0xffff,
2038};
2039
2040#define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
2041 BPF_F_RDONLY_PROG | \
2042 BPF_F_WRONLY | \
2043 BPF_F_WRONLY_PROG)
2044
2045#define BPF_MAP_CAN_READ BIT(0)
2046#define BPF_MAP_CAN_WRITE BIT(1)
2047
2048/* Maximum number of user-producer ring buffer samples that can be drained in
2049 * a call to bpf_user_ringbuf_drain().
2050 */
2051#define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024)
2052
2053static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
2054{
2055 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
2056
2057 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
2058 * not possible.
2059 */
2060 if (access_flags & BPF_F_RDONLY_PROG)
2061 return BPF_MAP_CAN_READ;
2062 else if (access_flags & BPF_F_WRONLY_PROG)
2063 return BPF_MAP_CAN_WRITE;
2064 else
2065 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
2066}
2067
2068static inline bool bpf_map_flags_access_ok(u32 access_flags)
2069{
2070 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
2071 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
2072}
2073
2074struct bpf_event_entry {
2075 struct perf_event *event;
2076 struct file *perf_file;
2077 struct file *map_file;
2078 struct rcu_head rcu;
2079};
2080
2081static inline bool map_type_contains_progs(struct bpf_map *map)
2082{
2083 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY ||
2084 map->map_type == BPF_MAP_TYPE_DEVMAP ||
2085 map->map_type == BPF_MAP_TYPE_CPUMAP;
2086}
2087
2088bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp);
2089int bpf_prog_calc_tag(struct bpf_prog *fp);
2090
2091const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
2092const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void);
2093
2094const struct bpf_func_proto *bpf_get_perf_event_read_value_proto(void);
2095
2096typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
2097 unsigned long off, unsigned long len);
2098typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
2099 const struct bpf_insn *src,
2100 struct bpf_insn *dst,
2101 struct bpf_prog *prog,
2102 u32 *target_size);
2103
2104u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
2105 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
2106
2107/* an array of programs to be executed under rcu_lock.
2108 *
2109 * Typical usage:
2110 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run);
2111 *
2112 * the structure returned by bpf_prog_array_alloc() should be populated
2113 * with program pointers and the last pointer must be NULL.
2114 * The user has to keep refcnt on the program and make sure the program
2115 * is removed from the array before bpf_prog_put().
2116 * The 'struct bpf_prog_array *' should only be replaced with xchg()
2117 * since other cpus are walking the array of pointers in parallel.
2118 */
2119struct bpf_prog_array_item {
2120 struct bpf_prog *prog;
2121 union {
2122 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
2123 u64 bpf_cookie;
2124 };
2125};
2126
2127struct bpf_prog_array {
2128 struct rcu_head rcu;
2129 struct bpf_prog_array_item items[];
2130};
2131
2132struct bpf_empty_prog_array {
2133 struct bpf_prog_array hdr;
2134 struct bpf_prog *null_prog;
2135};
2136
2137/* to avoid allocating empty bpf_prog_array for cgroups that
2138 * don't have bpf program attached use one global 'bpf_empty_prog_array'
2139 * It will not be modified the caller of bpf_prog_array_alloc()
2140 * (since caller requested prog_cnt == 0)
2141 * that pointer should be 'freed' by bpf_prog_array_free()
2142 */
2143extern struct bpf_empty_prog_array bpf_empty_prog_array;
2144
2145struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
2146void bpf_prog_array_free(struct bpf_prog_array *progs);
2147/* Use when traversal over the bpf_prog_array uses tasks_trace rcu */
2148void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs);
2149int bpf_prog_array_length(struct bpf_prog_array *progs);
2150bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
2151int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
2152 __u32 __user *prog_ids, u32 cnt);
2153
2154void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
2155 struct bpf_prog *old_prog);
2156int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
2157int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
2158 struct bpf_prog *prog);
2159int bpf_prog_array_copy_info(struct bpf_prog_array *array,
2160 u32 *prog_ids, u32 request_cnt,
2161 u32 *prog_cnt);
2162int bpf_prog_array_copy(struct bpf_prog_array *old_array,
2163 struct bpf_prog *exclude_prog,
2164 struct bpf_prog *include_prog,
2165 u64 bpf_cookie,
2166 struct bpf_prog_array **new_array);
2167
2168struct bpf_run_ctx {};
2169
2170struct bpf_cg_run_ctx {
2171 struct bpf_run_ctx run_ctx;
2172 const struct bpf_prog_array_item *prog_item;
2173 int retval;
2174};
2175
2176struct bpf_trace_run_ctx {
2177 struct bpf_run_ctx run_ctx;
2178 u64 bpf_cookie;
2179 bool is_uprobe;
2180};
2181
2182struct bpf_tramp_run_ctx {
2183 struct bpf_run_ctx run_ctx;
2184 u64 bpf_cookie;
2185 struct bpf_run_ctx *saved_run_ctx;
2186};
2187
2188static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)
2189{
2190 struct bpf_run_ctx *old_ctx = NULL;
2191
2192#ifdef CONFIG_BPF_SYSCALL
2193 old_ctx = current->bpf_ctx;
2194 current->bpf_ctx = new_ctx;
2195#endif
2196 return old_ctx;
2197}
2198
2199static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
2200{
2201#ifdef CONFIG_BPF_SYSCALL
2202 current->bpf_ctx = old_ctx;
2203#endif
2204}
2205
2206/* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
2207#define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0)
2208/* BPF program asks to set CN on the packet. */
2209#define BPF_RET_SET_CN (1 << 0)
2210
2211typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx);
2212
2213static __always_inline u32
2214bpf_prog_run_array(const struct bpf_prog_array *array,
2215 const void *ctx, bpf_prog_run_fn run_prog)
2216{
2217 const struct bpf_prog_array_item *item;
2218 const struct bpf_prog *prog;
2219 struct bpf_run_ctx *old_run_ctx;
2220 struct bpf_trace_run_ctx run_ctx;
2221 u32 ret = 1;
2222
2223 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held");
2224
2225 if (unlikely(!array))
2226 return ret;
2227
2228 run_ctx.is_uprobe = false;
2229
2230 migrate_disable();
2231 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
2232 item = &array->items[0];
2233 while ((prog = READ_ONCE(item->prog))) {
2234 run_ctx.bpf_cookie = item->bpf_cookie;
2235 ret &= run_prog(prog, ctx);
2236 item++;
2237 }
2238 bpf_reset_run_ctx(old_run_ctx);
2239 migrate_enable();
2240 return ret;
2241}
2242
2243/* Notes on RCU design for bpf_prog_arrays containing sleepable programs:
2244 *
2245 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array
2246 * overall. As a result, we must use the bpf_prog_array_free_sleepable
2247 * in order to use the tasks_trace rcu grace period.
2248 *
2249 * When a non-sleepable program is inside the array, we take the rcu read
2250 * section and disable preemption for that program alone, so it can access
2251 * rcu-protected dynamically sized maps.
2252 */
2253static __always_inline u32
2254bpf_prog_run_array_uprobe(const struct bpf_prog_array *array,
2255 const void *ctx, bpf_prog_run_fn run_prog)
2256{
2257 const struct bpf_prog_array_item *item;
2258 const struct bpf_prog *prog;
2259 struct bpf_run_ctx *old_run_ctx;
2260 struct bpf_trace_run_ctx run_ctx;
2261 u32 ret = 1;
2262
2263 might_fault();
2264 RCU_LOCKDEP_WARN(!rcu_read_lock_trace_held(), "no rcu lock held");
2265
2266 if (unlikely(!array))
2267 return ret;
2268
2269 migrate_disable();
2270
2271 run_ctx.is_uprobe = true;
2272
2273 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
2274 item = &array->items[0];
2275 while ((prog = READ_ONCE(item->prog))) {
2276 if (!prog->sleepable)
2277 rcu_read_lock();
2278
2279 run_ctx.bpf_cookie = item->bpf_cookie;
2280 ret &= run_prog(prog, ctx);
2281 item++;
2282
2283 if (!prog->sleepable)
2284 rcu_read_unlock();
2285 }
2286 bpf_reset_run_ctx(old_run_ctx);
2287 migrate_enable();
2288 return ret;
2289}
2290
2291#ifdef CONFIG_BPF_SYSCALL
2292DECLARE_PER_CPU(int, bpf_prog_active);
2293extern struct mutex bpf_stats_enabled_mutex;
2294
2295/*
2296 * Block execution of BPF programs attached to instrumentation (perf,
2297 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
2298 * these events can happen inside a region which holds a map bucket lock
2299 * and can deadlock on it.
2300 */
2301static inline void bpf_disable_instrumentation(void)
2302{
2303 migrate_disable();
2304 this_cpu_inc(bpf_prog_active);
2305}
2306
2307static inline void bpf_enable_instrumentation(void)
2308{
2309 this_cpu_dec(bpf_prog_active);
2310 migrate_enable();
2311}
2312
2313extern const struct super_operations bpf_super_ops;
2314extern const struct file_operations bpf_map_fops;
2315extern const struct file_operations bpf_prog_fops;
2316extern const struct file_operations bpf_iter_fops;
2317
2318#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
2319 extern const struct bpf_prog_ops _name ## _prog_ops; \
2320 extern const struct bpf_verifier_ops _name ## _verifier_ops;
2321#define BPF_MAP_TYPE(_id, _ops) \
2322 extern const struct bpf_map_ops _ops;
2323#define BPF_LINK_TYPE(_id, _name)
2324#include <linux/bpf_types.h>
2325#undef BPF_PROG_TYPE
2326#undef BPF_MAP_TYPE
2327#undef BPF_LINK_TYPE
2328
2329extern const struct bpf_prog_ops bpf_offload_prog_ops;
2330extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
2331extern const struct bpf_verifier_ops xdp_analyzer_ops;
2332
2333struct bpf_prog *bpf_prog_get(u32 ufd);
2334struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
2335 bool attach_drv);
2336void bpf_prog_add(struct bpf_prog *prog, int i);
2337void bpf_prog_sub(struct bpf_prog *prog, int i);
2338void bpf_prog_inc(struct bpf_prog *prog);
2339struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
2340void bpf_prog_put(struct bpf_prog *prog);
2341
2342void bpf_prog_free_id(struct bpf_prog *prog);
2343void bpf_map_free_id(struct bpf_map *map);
2344
2345struct btf_field *btf_record_find(const struct btf_record *rec,
2346 u32 offset, u32 field_mask);
2347void btf_record_free(struct btf_record *rec);
2348void bpf_map_free_record(struct bpf_map *map);
2349struct btf_record *btf_record_dup(const struct btf_record *rec);
2350bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b);
2351void bpf_obj_free_timer(const struct btf_record *rec, void *obj);
2352void bpf_obj_free_workqueue(const struct btf_record *rec, void *obj);
2353void bpf_obj_free_fields(const struct btf_record *rec, void *obj);
2354void __bpf_obj_drop_impl(void *p, const struct btf_record *rec, bool percpu);
2355
2356struct bpf_map *bpf_map_get(u32 ufd);
2357struct bpf_map *bpf_map_get_with_uref(u32 ufd);
2358
2359/*
2360 * The __bpf_map_get() and __btf_get_by_fd() functions parse a file
2361 * descriptor and return a corresponding map or btf object.
2362 * Their names are double underscored to emphasize the fact that they
2363 * do not increase refcnt. To also increase refcnt use corresponding
2364 * bpf_map_get() and btf_get_by_fd() functions.
2365 */
2366
2367static inline struct bpf_map *__bpf_map_get(struct fd f)
2368{
2369 if (fd_empty(f))
2370 return ERR_PTR(-EBADF);
2371 if (unlikely(fd_file(f)->f_op != &bpf_map_fops))
2372 return ERR_PTR(-EINVAL);
2373 return fd_file(f)->private_data;
2374}
2375
2376static inline struct btf *__btf_get_by_fd(struct fd f)
2377{
2378 if (fd_empty(f))
2379 return ERR_PTR(-EBADF);
2380 if (unlikely(fd_file(f)->f_op != &btf_fops))
2381 return ERR_PTR(-EINVAL);
2382 return fd_file(f)->private_data;
2383}
2384
2385void bpf_map_inc(struct bpf_map *map);
2386void bpf_map_inc_with_uref(struct bpf_map *map);
2387struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref);
2388struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
2389void bpf_map_put_with_uref(struct bpf_map *map);
2390void bpf_map_put(struct bpf_map *map);
2391void *bpf_map_area_alloc(u64 size, int numa_node);
2392void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
2393void bpf_map_area_free(void *base);
2394bool bpf_map_write_active(const struct bpf_map *map);
2395void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
2396int generic_map_lookup_batch(struct bpf_map *map,
2397 const union bpf_attr *attr,
2398 union bpf_attr __user *uattr);
2399int generic_map_update_batch(struct bpf_map *map, struct file *map_file,
2400 const union bpf_attr *attr,
2401 union bpf_attr __user *uattr);
2402int generic_map_delete_batch(struct bpf_map *map,
2403 const union bpf_attr *attr,
2404 union bpf_attr __user *uattr);
2405struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
2406struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
2407
2408int bpf_map_alloc_pages(const struct bpf_map *map, int nid,
2409 unsigned long nr_pages, struct page **page_array);
2410#ifdef CONFIG_MEMCG
2411void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
2412 int node);
2413void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags);
2414void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size,
2415 gfp_t flags);
2416void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
2417 size_t align, gfp_t flags);
2418#else
2419/*
2420 * These specialized allocators have to be macros for their allocations to be
2421 * accounted separately (to have separate alloc_tag).
2422 */
2423#define bpf_map_kmalloc_node(_map, _size, _flags, _node) \
2424 kmalloc_node(_size, _flags, _node)
2425#define bpf_map_kzalloc(_map, _size, _flags) \
2426 kzalloc(_size, _flags)
2427#define bpf_map_kvcalloc(_map, _n, _size, _flags) \
2428 kvcalloc(_n, _size, _flags)
2429#define bpf_map_alloc_percpu(_map, _size, _align, _flags) \
2430 __alloc_percpu_gfp(_size, _align, _flags)
2431#endif
2432
2433static inline int
2434bpf_map_init_elem_count(struct bpf_map *map)
2435{
2436 size_t size = sizeof(*map->elem_count), align = size;
2437 gfp_t flags = GFP_USER | __GFP_NOWARN;
2438
2439 map->elem_count = bpf_map_alloc_percpu(map, size, align, flags);
2440 if (!map->elem_count)
2441 return -ENOMEM;
2442
2443 return 0;
2444}
2445
2446static inline void
2447bpf_map_free_elem_count(struct bpf_map *map)
2448{
2449 free_percpu(map->elem_count);
2450}
2451
2452static inline void bpf_map_inc_elem_count(struct bpf_map *map)
2453{
2454 this_cpu_inc(*map->elem_count);
2455}
2456
2457static inline void bpf_map_dec_elem_count(struct bpf_map *map)
2458{
2459 this_cpu_dec(*map->elem_count);
2460}
2461
2462extern int sysctl_unprivileged_bpf_disabled;
2463
2464bool bpf_token_capable(const struct bpf_token *token, int cap);
2465
2466static inline bool bpf_allow_ptr_leaks(const struct bpf_token *token)
2467{
2468 return bpf_token_capable(token, CAP_PERFMON);
2469}
2470
2471static inline bool bpf_allow_uninit_stack(const struct bpf_token *token)
2472{
2473 return bpf_token_capable(token, CAP_PERFMON);
2474}
2475
2476static inline bool bpf_bypass_spec_v1(const struct bpf_token *token)
2477{
2478 return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON);
2479}
2480
2481static inline bool bpf_bypass_spec_v4(const struct bpf_token *token)
2482{
2483 return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON);
2484}
2485
2486int bpf_map_new_fd(struct bpf_map *map, int flags);
2487int bpf_prog_new_fd(struct bpf_prog *prog);
2488
2489void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2490 const struct bpf_link_ops *ops, struct bpf_prog *prog);
2491void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type,
2492 const struct bpf_link_ops *ops, struct bpf_prog *prog,
2493 bool sleepable);
2494int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
2495int bpf_link_settle(struct bpf_link_primer *primer);
2496void bpf_link_cleanup(struct bpf_link_primer *primer);
2497void bpf_link_inc(struct bpf_link *link);
2498struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link);
2499void bpf_link_put(struct bpf_link *link);
2500int bpf_link_new_fd(struct bpf_link *link);
2501struct bpf_link *bpf_link_get_from_fd(u32 ufd);
2502struct bpf_link *bpf_link_get_curr_or_next(u32 *id);
2503
2504void bpf_token_inc(struct bpf_token *token);
2505void bpf_token_put(struct bpf_token *token);
2506int bpf_token_create(union bpf_attr *attr);
2507struct bpf_token *bpf_token_get_from_fd(u32 ufd);
2508
2509bool bpf_token_allow_cmd(const struct bpf_token *token, enum bpf_cmd cmd);
2510bool bpf_token_allow_map_type(const struct bpf_token *token, enum bpf_map_type type);
2511bool bpf_token_allow_prog_type(const struct bpf_token *token,
2512 enum bpf_prog_type prog_type,
2513 enum bpf_attach_type attach_type);
2514
2515int bpf_obj_pin_user(u32 ufd, int path_fd, const char __user *pathname);
2516int bpf_obj_get_user(int path_fd, const char __user *pathname, int flags);
2517struct inode *bpf_get_inode(struct super_block *sb, const struct inode *dir,
2518 umode_t mode);
2519
2520#define BPF_ITER_FUNC_PREFIX "bpf_iter_"
2521#define DEFINE_BPF_ITER_FUNC(target, args...) \
2522 extern int bpf_iter_ ## target(args); \
2523 int __init bpf_iter_ ## target(args) { return 0; }
2524
2525/*
2526 * The task type of iterators.
2527 *
2528 * For BPF task iterators, they can be parameterized with various
2529 * parameters to visit only some of tasks.
2530 *
2531 * BPF_TASK_ITER_ALL (default)
2532 * Iterate over resources of every task.
2533 *
2534 * BPF_TASK_ITER_TID
2535 * Iterate over resources of a task/tid.
2536 *
2537 * BPF_TASK_ITER_TGID
2538 * Iterate over resources of every task of a process / task group.
2539 */
2540enum bpf_iter_task_type {
2541 BPF_TASK_ITER_ALL = 0,
2542 BPF_TASK_ITER_TID,
2543 BPF_TASK_ITER_TGID,
2544};
2545
2546struct bpf_iter_aux_info {
2547 /* for map_elem iter */
2548 struct bpf_map *map;
2549
2550 /* for cgroup iter */
2551 struct {
2552 struct cgroup *start; /* starting cgroup */
2553 enum bpf_cgroup_iter_order order;
2554 } cgroup;
2555 struct {
2556 enum bpf_iter_task_type type;
2557 u32 pid;
2558 } task;
2559};
2560
2561typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
2562 union bpf_iter_link_info *linfo,
2563 struct bpf_iter_aux_info *aux);
2564typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
2565typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
2566 struct seq_file *seq);
2567typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
2568 struct bpf_link_info *info);
2569typedef const struct bpf_func_proto *
2570(*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id,
2571 const struct bpf_prog *prog);
2572
2573enum bpf_iter_feature {
2574 BPF_ITER_RESCHED = BIT(0),
2575};
2576
2577#define BPF_ITER_CTX_ARG_MAX 2
2578struct bpf_iter_reg {
2579 const char *target;
2580 bpf_iter_attach_target_t attach_target;
2581 bpf_iter_detach_target_t detach_target;
2582 bpf_iter_show_fdinfo_t show_fdinfo;
2583 bpf_iter_fill_link_info_t fill_link_info;
2584 bpf_iter_get_func_proto_t get_func_proto;
2585 u32 ctx_arg_info_size;
2586 u32 feature;
2587 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
2588 const struct bpf_iter_seq_info *seq_info;
2589};
2590
2591struct bpf_iter_meta {
2592 __bpf_md_ptr(struct seq_file *, seq);
2593 u64 session_id;
2594 u64 seq_num;
2595};
2596
2597struct bpf_iter__bpf_map_elem {
2598 __bpf_md_ptr(struct bpf_iter_meta *, meta);
2599 __bpf_md_ptr(struct bpf_map *, map);
2600 __bpf_md_ptr(void *, key);
2601 __bpf_md_ptr(void *, value);
2602};
2603
2604int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
2605void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
2606int bpf_iter_prog_supported(struct bpf_prog *prog);
2607const struct bpf_func_proto *
2608bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
2609int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog);
2610int bpf_iter_new_fd(struct bpf_link *link);
2611bool bpf_link_is_iter(struct bpf_link *link);
2612struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
2613int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
2614void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
2615 struct seq_file *seq);
2616int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
2617 struct bpf_link_info *info);
2618
2619int map_set_for_each_callback_args(struct bpf_verifier_env *env,
2620 struct bpf_func_state *caller,
2621 struct bpf_func_state *callee);
2622
2623int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
2624int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
2625int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
2626 u64 flags);
2627int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
2628 u64 flags);
2629
2630int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
2631
2632int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
2633 void *key, void *value, u64 map_flags);
2634int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2635int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
2636 void *key, void *value, u64 map_flags);
2637int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2638
2639int bpf_get_file_flag(int flags);
2640int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size,
2641 size_t actual_size);
2642
2643/* verify correctness of eBPF program */
2644int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size);
2645
2646#ifndef CONFIG_BPF_JIT_ALWAYS_ON
2647void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
2648#endif
2649
2650struct btf *bpf_get_btf_vmlinux(void);
2651
2652/* Map specifics */
2653struct xdp_frame;
2654struct sk_buff;
2655struct bpf_dtab_netdev;
2656struct bpf_cpu_map_entry;
2657
2658void __dev_flush(struct list_head *flush_list);
2659int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2660 struct net_device *dev_rx);
2661int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2662 struct net_device *dev_rx);
2663int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2664 struct bpf_map *map, bool exclude_ingress);
2665int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
2666 const struct bpf_prog *xdp_prog);
2667int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2668 const struct bpf_prog *xdp_prog,
2669 struct bpf_map *map, bool exclude_ingress);
2670
2671void __cpu_map_flush(struct list_head *flush_list);
2672int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf,
2673 struct net_device *dev_rx);
2674int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2675 struct sk_buff *skb);
2676
2677/* Return map's numa specified by userspace */
2678static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
2679{
2680 return (attr->map_flags & BPF_F_NUMA_NODE) ?
2681 attr->numa_node : NUMA_NO_NODE;
2682}
2683
2684struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
2685int array_map_alloc_check(union bpf_attr *attr);
2686
2687int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
2688 union bpf_attr __user *uattr);
2689int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
2690 union bpf_attr __user *uattr);
2691int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2692 const union bpf_attr *kattr,
2693 union bpf_attr __user *uattr);
2694int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2695 const union bpf_attr *kattr,
2696 union bpf_attr __user *uattr);
2697int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
2698 const union bpf_attr *kattr,
2699 union bpf_attr __user *uattr);
2700int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2701 const union bpf_attr *kattr,
2702 union bpf_attr __user *uattr);
2703int bpf_prog_test_run_nf(struct bpf_prog *prog,
2704 const union bpf_attr *kattr,
2705 union bpf_attr __user *uattr);
2706bool btf_ctx_access(int off, int size, enum bpf_access_type type,
2707 const struct bpf_prog *prog,
2708 struct bpf_insn_access_aux *info);
2709
2710static inline bool bpf_tracing_ctx_access(int off, int size,
2711 enum bpf_access_type type)
2712{
2713 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
2714 return false;
2715 if (type != BPF_READ)
2716 return false;
2717 if (off % size != 0)
2718 return false;
2719 return true;
2720}
2721
2722static inline bool bpf_tracing_btf_ctx_access(int off, int size,
2723 enum bpf_access_type type,
2724 const struct bpf_prog *prog,
2725 struct bpf_insn_access_aux *info)
2726{
2727 if (!bpf_tracing_ctx_access(off, size, type))
2728 return false;
2729 return btf_ctx_access(off, size, type, prog, info);
2730}
2731
2732int btf_struct_access(struct bpf_verifier_log *log,
2733 const struct bpf_reg_state *reg,
2734 int off, int size, enum bpf_access_type atype,
2735 u32 *next_btf_id, enum bpf_type_flag *flag, const char **field_name);
2736bool btf_struct_ids_match(struct bpf_verifier_log *log,
2737 const struct btf *btf, u32 id, int off,
2738 const struct btf *need_btf, u32 need_type_id,
2739 bool strict);
2740
2741int btf_distill_func_proto(struct bpf_verifier_log *log,
2742 struct btf *btf,
2743 const struct btf_type *func_proto,
2744 const char *func_name,
2745 struct btf_func_model *m);
2746
2747struct bpf_reg_state;
2748int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog);
2749int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
2750 struct btf *btf, const struct btf_type *t);
2751const char *btf_find_decl_tag_value(const struct btf *btf, const struct btf_type *pt,
2752 int comp_idx, const char *tag_key);
2753int btf_find_next_decl_tag(const struct btf *btf, const struct btf_type *pt,
2754 int comp_idx, const char *tag_key, int last_id);
2755
2756struct bpf_prog *bpf_prog_by_id(u32 id);
2757struct bpf_link *bpf_link_by_id(u32 id);
2758
2759const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id,
2760 const struct bpf_prog *prog);
2761void bpf_task_storage_free(struct task_struct *task);
2762void bpf_cgrp_storage_free(struct cgroup *cgroup);
2763bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
2764const struct btf_func_model *
2765bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2766 const struct bpf_insn *insn);
2767int bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
2768 u16 btf_fd_idx, u8 **func_addr);
2769
2770struct bpf_core_ctx {
2771 struct bpf_verifier_log *log;
2772 const struct btf *btf;
2773};
2774
2775bool btf_nested_type_is_trusted(struct bpf_verifier_log *log,
2776 const struct bpf_reg_state *reg,
2777 const char *field_name, u32 btf_id, const char *suffix);
2778
2779bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log,
2780 const struct btf *reg_btf, u32 reg_id,
2781 const struct btf *arg_btf, u32 arg_id);
2782
2783int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo,
2784 int relo_idx, void *insn);
2785
2786static inline bool unprivileged_ebpf_enabled(void)
2787{
2788 return !sysctl_unprivileged_bpf_disabled;
2789}
2790
2791/* Not all bpf prog type has the bpf_ctx.
2792 * For the bpf prog type that has initialized the bpf_ctx,
2793 * this function can be used to decide if a kernel function
2794 * is called by a bpf program.
2795 */
2796static inline bool has_current_bpf_ctx(void)
2797{
2798 return !!current->bpf_ctx;
2799}
2800
2801void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog);
2802
2803void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2804 enum bpf_dynptr_type type, u32 offset, u32 size);
2805void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr);
2806void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr);
2807
2808#else /* !CONFIG_BPF_SYSCALL */
2809static inline struct bpf_prog *bpf_prog_get(u32 ufd)
2810{
2811 return ERR_PTR(-EOPNOTSUPP);
2812}
2813
2814static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
2815 enum bpf_prog_type type,
2816 bool attach_drv)
2817{
2818 return ERR_PTR(-EOPNOTSUPP);
2819}
2820
2821static inline void bpf_prog_add(struct bpf_prog *prog, int i)
2822{
2823}
2824
2825static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
2826{
2827}
2828
2829static inline void bpf_prog_put(struct bpf_prog *prog)
2830{
2831}
2832
2833static inline void bpf_prog_inc(struct bpf_prog *prog)
2834{
2835}
2836
2837static inline struct bpf_prog *__must_check
2838bpf_prog_inc_not_zero(struct bpf_prog *prog)
2839{
2840 return ERR_PTR(-EOPNOTSUPP);
2841}
2842
2843static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2844 const struct bpf_link_ops *ops,
2845 struct bpf_prog *prog)
2846{
2847}
2848
2849static inline void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type,
2850 const struct bpf_link_ops *ops, struct bpf_prog *prog,
2851 bool sleepable)
2852{
2853}
2854
2855static inline int bpf_link_prime(struct bpf_link *link,
2856 struct bpf_link_primer *primer)
2857{
2858 return -EOPNOTSUPP;
2859}
2860
2861static inline int bpf_link_settle(struct bpf_link_primer *primer)
2862{
2863 return -EOPNOTSUPP;
2864}
2865
2866static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
2867{
2868}
2869
2870static inline void bpf_link_inc(struct bpf_link *link)
2871{
2872}
2873
2874static inline struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link)
2875{
2876 return NULL;
2877}
2878
2879static inline void bpf_link_put(struct bpf_link *link)
2880{
2881}
2882
2883static inline int bpf_obj_get_user(const char __user *pathname, int flags)
2884{
2885 return -EOPNOTSUPP;
2886}
2887
2888static inline bool bpf_token_capable(const struct bpf_token *token, int cap)
2889{
2890 return capable(cap) || (cap != CAP_SYS_ADMIN && capable(CAP_SYS_ADMIN));
2891}
2892
2893static inline void bpf_token_inc(struct bpf_token *token)
2894{
2895}
2896
2897static inline void bpf_token_put(struct bpf_token *token)
2898{
2899}
2900
2901static inline struct bpf_token *bpf_token_get_from_fd(u32 ufd)
2902{
2903 return ERR_PTR(-EOPNOTSUPP);
2904}
2905
2906static inline void __dev_flush(struct list_head *flush_list)
2907{
2908}
2909
2910struct xdp_frame;
2911struct bpf_dtab_netdev;
2912struct bpf_cpu_map_entry;
2913
2914static inline
2915int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2916 struct net_device *dev_rx)
2917{
2918 return 0;
2919}
2920
2921static inline
2922int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2923 struct net_device *dev_rx)
2924{
2925 return 0;
2926}
2927
2928static inline
2929int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2930 struct bpf_map *map, bool exclude_ingress)
2931{
2932 return 0;
2933}
2934
2935struct sk_buff;
2936
2937static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
2938 struct sk_buff *skb,
2939 const struct bpf_prog *xdp_prog)
2940{
2941 return 0;
2942}
2943
2944static inline
2945int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2946 const struct bpf_prog *xdp_prog,
2947 struct bpf_map *map, bool exclude_ingress)
2948{
2949 return 0;
2950}
2951
2952static inline void __cpu_map_flush(struct list_head *flush_list)
2953{
2954}
2955
2956static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
2957 struct xdp_frame *xdpf,
2958 struct net_device *dev_rx)
2959{
2960 return 0;
2961}
2962
2963static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2964 struct sk_buff *skb)
2965{
2966 return -EOPNOTSUPP;
2967}
2968
2969static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
2970 enum bpf_prog_type type)
2971{
2972 return ERR_PTR(-EOPNOTSUPP);
2973}
2974
2975static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
2976 const union bpf_attr *kattr,
2977 union bpf_attr __user *uattr)
2978{
2979 return -ENOTSUPP;
2980}
2981
2982static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
2983 const union bpf_attr *kattr,
2984 union bpf_attr __user *uattr)
2985{
2986 return -ENOTSUPP;
2987}
2988
2989static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2990 const union bpf_attr *kattr,
2991 union bpf_attr __user *uattr)
2992{
2993 return -ENOTSUPP;
2994}
2995
2996static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2997 const union bpf_attr *kattr,
2998 union bpf_attr __user *uattr)
2999{
3000 return -ENOTSUPP;
3001}
3002
3003static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
3004 const union bpf_attr *kattr,
3005 union bpf_attr __user *uattr)
3006{
3007 return -ENOTSUPP;
3008}
3009
3010static inline void bpf_map_put(struct bpf_map *map)
3011{
3012}
3013
3014static inline struct bpf_prog *bpf_prog_by_id(u32 id)
3015{
3016 return ERR_PTR(-ENOTSUPP);
3017}
3018
3019static inline int btf_struct_access(struct bpf_verifier_log *log,
3020 const struct bpf_reg_state *reg,
3021 int off, int size, enum bpf_access_type atype,
3022 u32 *next_btf_id, enum bpf_type_flag *flag,
3023 const char **field_name)
3024{
3025 return -EACCES;
3026}
3027
3028static inline const struct bpf_func_proto *
3029bpf_base_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
3030{
3031 return NULL;
3032}
3033
3034static inline void bpf_task_storage_free(struct task_struct *task)
3035{
3036}
3037
3038static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog)
3039{
3040 return false;
3041}
3042
3043static inline const struct btf_func_model *
3044bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
3045 const struct bpf_insn *insn)
3046{
3047 return NULL;
3048}
3049
3050static inline int
3051bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
3052 u16 btf_fd_idx, u8 **func_addr)
3053{
3054 return -ENOTSUPP;
3055}
3056
3057static inline bool unprivileged_ebpf_enabled(void)
3058{
3059 return false;
3060}
3061
3062static inline bool has_current_bpf_ctx(void)
3063{
3064 return false;
3065}
3066
3067static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog)
3068{
3069}
3070
3071static inline void bpf_cgrp_storage_free(struct cgroup *cgroup)
3072{
3073}
3074
3075static inline void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
3076 enum bpf_dynptr_type type, u32 offset, u32 size)
3077{
3078}
3079
3080static inline void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr)
3081{
3082}
3083
3084static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr)
3085{
3086}
3087#endif /* CONFIG_BPF_SYSCALL */
3088
3089static __always_inline int
3090bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
3091{
3092 int ret = -EFAULT;
3093
3094 if (IS_ENABLED(CONFIG_BPF_EVENTS))
3095 ret = copy_from_kernel_nofault(dst, unsafe_ptr, size);
3096 if (unlikely(ret < 0))
3097 memset(dst, 0, size);
3098 return ret;
3099}
3100
3101void __bpf_free_used_btfs(struct btf_mod_pair *used_btfs, u32 len);
3102
3103static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
3104 enum bpf_prog_type type)
3105{
3106 return bpf_prog_get_type_dev(ufd, type, false);
3107}
3108
3109void __bpf_free_used_maps(struct bpf_prog_aux *aux,
3110 struct bpf_map **used_maps, u32 len);
3111
3112bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
3113
3114int bpf_prog_offload_compile(struct bpf_prog *prog);
3115void bpf_prog_dev_bound_destroy(struct bpf_prog *prog);
3116int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
3117 struct bpf_prog *prog);
3118
3119int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
3120
3121int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
3122int bpf_map_offload_update_elem(struct bpf_map *map,
3123 void *key, void *value, u64 flags);
3124int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
3125int bpf_map_offload_get_next_key(struct bpf_map *map,
3126 void *key, void *next_key);
3127
3128bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
3129
3130struct bpf_offload_dev *
3131bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
3132void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
3133void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
3134int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
3135 struct net_device *netdev);
3136void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
3137 struct net_device *netdev);
3138bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
3139
3140void unpriv_ebpf_notify(int new_state);
3141
3142#if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
3143int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
3144 struct bpf_prog_aux *prog_aux);
3145void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, u32 func_id);
3146int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr);
3147int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, struct bpf_prog *old_prog);
3148void bpf_dev_bound_netdev_unregister(struct net_device *dev);
3149
3150static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
3151{
3152 return aux->dev_bound;
3153}
3154
3155static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux)
3156{
3157 return aux->offload_requested;
3158}
3159
3160bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs);
3161
3162static inline bool bpf_map_is_offloaded(struct bpf_map *map)
3163{
3164 return unlikely(map->ops == &bpf_map_offload_ops);
3165}
3166
3167struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
3168void bpf_map_offload_map_free(struct bpf_map *map);
3169u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map);
3170int bpf_prog_test_run_syscall(struct bpf_prog *prog,
3171 const union bpf_attr *kattr,
3172 union bpf_attr __user *uattr);
3173
3174int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
3175int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
3176int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
3177int sock_map_bpf_prog_query(const union bpf_attr *attr,
3178 union bpf_attr __user *uattr);
3179int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog);
3180
3181void sock_map_unhash(struct sock *sk);
3182void sock_map_destroy(struct sock *sk);
3183void sock_map_close(struct sock *sk, long timeout);
3184#else
3185static inline int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
3186 struct bpf_prog_aux *prog_aux)
3187{
3188 return -EOPNOTSUPP;
3189}
3190
3191static inline void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog,
3192 u32 func_id)
3193{
3194 return NULL;
3195}
3196
3197static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog,
3198 union bpf_attr *attr)
3199{
3200 return -EOPNOTSUPP;
3201}
3202
3203static inline int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog,
3204 struct bpf_prog *old_prog)
3205{
3206 return -EOPNOTSUPP;
3207}
3208
3209static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev)
3210{
3211}
3212
3213static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
3214{
3215 return false;
3216}
3217
3218static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux)
3219{
3220 return false;
3221}
3222
3223static inline bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs)
3224{
3225 return false;
3226}
3227
3228static inline bool bpf_map_is_offloaded(struct bpf_map *map)
3229{
3230 return false;
3231}
3232
3233static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
3234{
3235 return ERR_PTR(-EOPNOTSUPP);
3236}
3237
3238static inline void bpf_map_offload_map_free(struct bpf_map *map)
3239{
3240}
3241
3242static inline u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map)
3243{
3244 return 0;
3245}
3246
3247static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog,
3248 const union bpf_attr *kattr,
3249 union bpf_attr __user *uattr)
3250{
3251 return -ENOTSUPP;
3252}
3253
3254#ifdef CONFIG_BPF_SYSCALL
3255static inline int sock_map_get_from_fd(const union bpf_attr *attr,
3256 struct bpf_prog *prog)
3257{
3258 return -EINVAL;
3259}
3260
3261static inline int sock_map_prog_detach(const union bpf_attr *attr,
3262 enum bpf_prog_type ptype)
3263{
3264 return -EOPNOTSUPP;
3265}
3266
3267static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
3268 u64 flags)
3269{
3270 return -EOPNOTSUPP;
3271}
3272
3273static inline int sock_map_bpf_prog_query(const union bpf_attr *attr,
3274 union bpf_attr __user *uattr)
3275{
3276 return -EINVAL;
3277}
3278
3279static inline int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog)
3280{
3281 return -EOPNOTSUPP;
3282}
3283#endif /* CONFIG_BPF_SYSCALL */
3284#endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
3285
3286static __always_inline void
3287bpf_prog_inc_misses_counters(const struct bpf_prog_array *array)
3288{
3289 const struct bpf_prog_array_item *item;
3290 struct bpf_prog *prog;
3291
3292 if (unlikely(!array))
3293 return;
3294
3295 item = &array->items[0];
3296 while ((prog = READ_ONCE(item->prog))) {
3297 bpf_prog_inc_misses_counter(prog);
3298 item++;
3299 }
3300}
3301
3302#if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
3303void bpf_sk_reuseport_detach(struct sock *sk);
3304int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
3305 void *value);
3306int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
3307 void *value, u64 map_flags);
3308#else
3309static inline void bpf_sk_reuseport_detach(struct sock *sk)
3310{
3311}
3312
3313#ifdef CONFIG_BPF_SYSCALL
3314static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
3315 void *key, void *value)
3316{
3317 return -EOPNOTSUPP;
3318}
3319
3320static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
3321 void *key, void *value,
3322 u64 map_flags)
3323{
3324 return -EOPNOTSUPP;
3325}
3326#endif /* CONFIG_BPF_SYSCALL */
3327#endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
3328
3329/* verifier prototypes for helper functions called from eBPF programs */
3330extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
3331extern const struct bpf_func_proto bpf_map_update_elem_proto;
3332extern const struct bpf_func_proto bpf_map_delete_elem_proto;
3333extern const struct bpf_func_proto bpf_map_push_elem_proto;
3334extern const struct bpf_func_proto bpf_map_pop_elem_proto;
3335extern const struct bpf_func_proto bpf_map_peek_elem_proto;
3336extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto;
3337
3338extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
3339extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
3340extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
3341extern const struct bpf_func_proto bpf_tail_call_proto;
3342extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
3343extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
3344extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto;
3345extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
3346extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
3347extern const struct bpf_func_proto bpf_get_current_comm_proto;
3348extern const struct bpf_func_proto bpf_get_stackid_proto;
3349extern const struct bpf_func_proto bpf_get_stack_proto;
3350extern const struct bpf_func_proto bpf_get_stack_sleepable_proto;
3351extern const struct bpf_func_proto bpf_get_task_stack_proto;
3352extern const struct bpf_func_proto bpf_get_task_stack_sleepable_proto;
3353extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
3354extern const struct bpf_func_proto bpf_get_stack_proto_pe;
3355extern const struct bpf_func_proto bpf_sock_map_update_proto;
3356extern const struct bpf_func_proto bpf_sock_hash_update_proto;
3357extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
3358extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
3359extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto;
3360extern const struct bpf_func_proto bpf_current_task_under_cgroup_proto;
3361extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
3362extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
3363extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
3364extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
3365extern const struct bpf_func_proto bpf_spin_lock_proto;
3366extern const struct bpf_func_proto bpf_spin_unlock_proto;
3367extern const struct bpf_func_proto bpf_get_local_storage_proto;
3368extern const struct bpf_func_proto bpf_strtol_proto;
3369extern const struct bpf_func_proto bpf_strtoul_proto;
3370extern const struct bpf_func_proto bpf_tcp_sock_proto;
3371extern const struct bpf_func_proto bpf_jiffies64_proto;
3372extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
3373extern const struct bpf_func_proto bpf_event_output_data_proto;
3374extern const struct bpf_func_proto bpf_ringbuf_output_proto;
3375extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
3376extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
3377extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
3378extern const struct bpf_func_proto bpf_ringbuf_query_proto;
3379extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto;
3380extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto;
3381extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto;
3382extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
3383extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
3384extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
3385extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
3386extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
3387extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto;
3388extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto;
3389extern const struct bpf_func_proto bpf_copy_from_user_proto;
3390extern const struct bpf_func_proto bpf_snprintf_btf_proto;
3391extern const struct bpf_func_proto bpf_snprintf_proto;
3392extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
3393extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
3394extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto;
3395extern const struct bpf_func_proto bpf_sock_from_file_proto;
3396extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto;
3397extern const struct bpf_func_proto bpf_task_storage_get_recur_proto;
3398extern const struct bpf_func_proto bpf_task_storage_get_proto;
3399extern const struct bpf_func_proto bpf_task_storage_delete_recur_proto;
3400extern const struct bpf_func_proto bpf_task_storage_delete_proto;
3401extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
3402extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto;
3403extern const struct bpf_func_proto bpf_sk_setsockopt_proto;
3404extern const struct bpf_func_proto bpf_sk_getsockopt_proto;
3405extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto;
3406extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto;
3407extern const struct bpf_func_proto bpf_find_vma_proto;
3408extern const struct bpf_func_proto bpf_loop_proto;
3409extern const struct bpf_func_proto bpf_copy_from_user_task_proto;
3410extern const struct bpf_func_proto bpf_set_retval_proto;
3411extern const struct bpf_func_proto bpf_get_retval_proto;
3412extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto;
3413extern const struct bpf_func_proto bpf_cgrp_storage_get_proto;
3414extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto;
3415
3416const struct bpf_func_proto *tracing_prog_func_proto(
3417 enum bpf_func_id func_id, const struct bpf_prog *prog);
3418
3419/* Shared helpers among cBPF and eBPF. */
3420void bpf_user_rnd_init_once(void);
3421u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3422u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3423
3424#if defined(CONFIG_NET)
3425bool bpf_sock_common_is_valid_access(int off, int size,
3426 enum bpf_access_type type,
3427 struct bpf_insn_access_aux *info);
3428bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3429 struct bpf_insn_access_aux *info);
3430u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3431 const struct bpf_insn *si,
3432 struct bpf_insn *insn_buf,
3433 struct bpf_prog *prog,
3434 u32 *target_size);
3435int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags,
3436 struct bpf_dynptr *ptr);
3437#else
3438static inline bool bpf_sock_common_is_valid_access(int off, int size,
3439 enum bpf_access_type type,
3440 struct bpf_insn_access_aux *info)
3441{
3442 return false;
3443}
3444static inline bool bpf_sock_is_valid_access(int off, int size,
3445 enum bpf_access_type type,
3446 struct bpf_insn_access_aux *info)
3447{
3448 return false;
3449}
3450static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3451 const struct bpf_insn *si,
3452 struct bpf_insn *insn_buf,
3453 struct bpf_prog *prog,
3454 u32 *target_size)
3455{
3456 return 0;
3457}
3458static inline int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags,
3459 struct bpf_dynptr *ptr)
3460{
3461 return -EOPNOTSUPP;
3462}
3463#endif
3464
3465#ifdef CONFIG_INET
3466struct sk_reuseport_kern {
3467 struct sk_buff *skb;
3468 struct sock *sk;
3469 struct sock *selected_sk;
3470 struct sock *migrating_sk;
3471 void *data_end;
3472 u32 hash;
3473 u32 reuseport_id;
3474 bool bind_inany;
3475};
3476bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3477 struct bpf_insn_access_aux *info);
3478
3479u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3480 const struct bpf_insn *si,
3481 struct bpf_insn *insn_buf,
3482 struct bpf_prog *prog,
3483 u32 *target_size);
3484
3485bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3486 struct bpf_insn_access_aux *info);
3487
3488u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3489 const struct bpf_insn *si,
3490 struct bpf_insn *insn_buf,
3491 struct bpf_prog *prog,
3492 u32 *target_size);
3493#else
3494static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
3495 enum bpf_access_type type,
3496 struct bpf_insn_access_aux *info)
3497{
3498 return false;
3499}
3500
3501static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3502 const struct bpf_insn *si,
3503 struct bpf_insn *insn_buf,
3504 struct bpf_prog *prog,
3505 u32 *target_size)
3506{
3507 return 0;
3508}
3509static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
3510 enum bpf_access_type type,
3511 struct bpf_insn_access_aux *info)
3512{
3513 return false;
3514}
3515
3516static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3517 const struct bpf_insn *si,
3518 struct bpf_insn *insn_buf,
3519 struct bpf_prog *prog,
3520 u32 *target_size)
3521{
3522 return 0;
3523}
3524#endif /* CONFIG_INET */
3525
3526enum bpf_text_poke_type {
3527 BPF_MOD_CALL,
3528 BPF_MOD_JUMP,
3529};
3530
3531int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
3532 void *addr1, void *addr2);
3533
3534void bpf_arch_poke_desc_update(struct bpf_jit_poke_descriptor *poke,
3535 struct bpf_prog *new, struct bpf_prog *old);
3536
3537void *bpf_arch_text_copy(void *dst, void *src, size_t len);
3538int bpf_arch_text_invalidate(void *dst, size_t len);
3539
3540struct btf_id_set;
3541bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
3542
3543#define MAX_BPRINTF_VARARGS 12
3544#define MAX_BPRINTF_BUF 1024
3545
3546struct bpf_bprintf_data {
3547 u32 *bin_args;
3548 char *buf;
3549 bool get_bin_args;
3550 bool get_buf;
3551};
3552
3553int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
3554 u32 num_args, struct bpf_bprintf_data *data);
3555void bpf_bprintf_cleanup(struct bpf_bprintf_data *data);
3556
3557#ifdef CONFIG_BPF_LSM
3558void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype);
3559void bpf_cgroup_atype_put(int cgroup_atype);
3560#else
3561static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {}
3562static inline void bpf_cgroup_atype_put(int cgroup_atype) {}
3563#endif /* CONFIG_BPF_LSM */
3564
3565struct key;
3566
3567#ifdef CONFIG_KEYS
3568struct bpf_key {
3569 struct key *key;
3570 bool has_ref;
3571};
3572#endif /* CONFIG_KEYS */
3573
3574static inline bool type_is_alloc(u32 type)
3575{
3576 return type & MEM_ALLOC;
3577}
3578
3579static inline gfp_t bpf_memcg_flags(gfp_t flags)
3580{
3581 if (memcg_bpf_enabled())
3582 return flags | __GFP_ACCOUNT;
3583 return flags;
3584}
3585
3586static inline bool bpf_is_subprog(const struct bpf_prog *prog)
3587{
3588 return prog->aux->func_idx != 0;
3589}
3590
3591#endif /* _LINUX_BPF_H */