xref: /linux-6.15/include/linux/bpf.h (revision 97eb35f3)
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 
35 struct bpf_verifier_env;
36 struct bpf_verifier_log;
37 struct perf_event;
38 struct bpf_prog;
39 struct bpf_prog_aux;
40 struct bpf_map;
41 struct bpf_arena;
42 struct sock;
43 struct seq_file;
44 struct btf;
45 struct btf_type;
46 struct exception_table_entry;
47 struct seq_operations;
48 struct bpf_iter_aux_info;
49 struct bpf_local_storage;
50 struct bpf_local_storage_map;
51 struct kobject;
52 struct mem_cgroup;
53 struct module;
54 struct bpf_func_state;
55 struct ftrace_ops;
56 struct cgroup;
57 struct bpf_token;
58 struct user_namespace;
59 struct super_block;
60 struct inode;
61 
62 extern struct idr btf_idr;
63 extern spinlock_t btf_idr_lock;
64 extern struct kobject *btf_kobj;
65 extern struct bpf_mem_alloc bpf_global_ma, bpf_global_percpu_ma;
66 extern bool bpf_global_ma_set;
67 
68 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64);
69 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data,
70 					struct bpf_iter_aux_info *aux);
71 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
72 typedef unsigned int (*bpf_func_t)(const void *,
73 				   const struct bpf_insn *);
74 struct 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 */
82 struct 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 
187 enum {
188 	/* Support at most 11 fields in a BTF type */
189 	BTF_FIELDS_MAX	   = 11,
190 };
191 
192 enum 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 };
209 
210 typedef void (*btf_dtor_kfunc_t)(void *);
211 
212 struct btf_field_kptr {
213 	struct btf *btf;
214 	struct module *module;
215 	/* dtor used if btf_is_kernel(btf), otherwise the type is
216 	 * program-allocated, dtor is NULL,  and __bpf_obj_drop_impl is used
217 	 */
218 	btf_dtor_kfunc_t dtor;
219 	u32 btf_id;
220 };
221 
222 struct btf_field_graph_root {
223 	struct btf *btf;
224 	u32 value_btf_id;
225 	u32 node_offset;
226 	struct btf_record *value_rec;
227 };
228 
229 struct btf_field {
230 	u32 offset;
231 	u32 size;
232 	enum btf_field_type type;
233 	union {
234 		struct btf_field_kptr kptr;
235 		struct btf_field_graph_root graph_root;
236 	};
237 };
238 
239 struct btf_record {
240 	u32 cnt;
241 	u32 field_mask;
242 	int spin_lock_off;
243 	int timer_off;
244 	int wq_off;
245 	int refcount_off;
246 	struct btf_field fields[];
247 };
248 
249 /* Non-opaque version of bpf_rb_node in uapi/linux/bpf.h */
250 struct bpf_rb_node_kern {
251 	struct rb_node rb_node;
252 	void *owner;
253 } __attribute__((aligned(8)));
254 
255 /* Non-opaque version of bpf_list_node in uapi/linux/bpf.h */
256 struct bpf_list_node_kern {
257 	struct list_head list_head;
258 	void *owner;
259 } __attribute__((aligned(8)));
260 
261 struct bpf_map {
262 	const struct bpf_map_ops *ops;
263 	struct bpf_map *inner_map_meta;
264 #ifdef CONFIG_SECURITY
265 	void *security;
266 #endif
267 	enum bpf_map_type map_type;
268 	u32 key_size;
269 	u32 value_size;
270 	u32 max_entries;
271 	u64 map_extra; /* any per-map-type extra fields */
272 	u32 map_flags;
273 	u32 id;
274 	struct btf_record *record;
275 	int numa_node;
276 	u32 btf_key_type_id;
277 	u32 btf_value_type_id;
278 	u32 btf_vmlinux_value_type_id;
279 	struct btf *btf;
280 #ifdef CONFIG_MEMCG
281 	struct obj_cgroup *objcg;
282 #endif
283 	char name[BPF_OBJ_NAME_LEN];
284 	struct mutex freeze_mutex;
285 	atomic64_t refcnt;
286 	atomic64_t usercnt;
287 	/* rcu is used before freeing and work is only used during freeing */
288 	union {
289 		struct work_struct work;
290 		struct rcu_head rcu;
291 	};
292 	atomic64_t writecnt;
293 	/* 'Ownership' of program-containing map is claimed by the first program
294 	 * that is going to use this map or by the first program which FD is
295 	 * stored in the map to make sure that all callers and callees have the
296 	 * same prog type, JITed flag and xdp_has_frags flag.
297 	 */
298 	struct {
299 		const struct btf_type *attach_func_proto;
300 		spinlock_t lock;
301 		enum bpf_prog_type type;
302 		bool jited;
303 		bool xdp_has_frags;
304 	} owner;
305 	bool bypass_spec_v1;
306 	bool frozen; /* write-once; write-protected by freeze_mutex */
307 	bool free_after_mult_rcu_gp;
308 	bool free_after_rcu_gp;
309 	atomic64_t sleepable_refcnt;
310 	s64 __percpu *elem_count;
311 };
312 
313 static inline const char *btf_field_type_name(enum btf_field_type type)
314 {
315 	switch (type) {
316 	case BPF_SPIN_LOCK:
317 		return "bpf_spin_lock";
318 	case BPF_TIMER:
319 		return "bpf_timer";
320 	case BPF_WORKQUEUE:
321 		return "bpf_wq";
322 	case BPF_KPTR_UNREF:
323 	case BPF_KPTR_REF:
324 		return "kptr";
325 	case BPF_KPTR_PERCPU:
326 		return "percpu_kptr";
327 	case BPF_UPTR:
328 		return "uptr";
329 	case BPF_LIST_HEAD:
330 		return "bpf_list_head";
331 	case BPF_LIST_NODE:
332 		return "bpf_list_node";
333 	case BPF_RB_ROOT:
334 		return "bpf_rb_root";
335 	case BPF_RB_NODE:
336 		return "bpf_rb_node";
337 	case BPF_REFCOUNT:
338 		return "bpf_refcount";
339 	default:
340 		WARN_ON_ONCE(1);
341 		return "unknown";
342 	}
343 }
344 
345 static inline u32 btf_field_type_size(enum btf_field_type type)
346 {
347 	switch (type) {
348 	case BPF_SPIN_LOCK:
349 		return sizeof(struct bpf_spin_lock);
350 	case BPF_TIMER:
351 		return sizeof(struct bpf_timer);
352 	case BPF_WORKQUEUE:
353 		return sizeof(struct bpf_wq);
354 	case BPF_KPTR_UNREF:
355 	case BPF_KPTR_REF:
356 	case BPF_KPTR_PERCPU:
357 	case BPF_UPTR:
358 		return sizeof(u64);
359 	case BPF_LIST_HEAD:
360 		return sizeof(struct bpf_list_head);
361 	case BPF_LIST_NODE:
362 		return sizeof(struct bpf_list_node);
363 	case BPF_RB_ROOT:
364 		return sizeof(struct bpf_rb_root);
365 	case BPF_RB_NODE:
366 		return sizeof(struct bpf_rb_node);
367 	case BPF_REFCOUNT:
368 		return sizeof(struct bpf_refcount);
369 	default:
370 		WARN_ON_ONCE(1);
371 		return 0;
372 	}
373 }
374 
375 static inline u32 btf_field_type_align(enum btf_field_type type)
376 {
377 	switch (type) {
378 	case BPF_SPIN_LOCK:
379 		return __alignof__(struct bpf_spin_lock);
380 	case BPF_TIMER:
381 		return __alignof__(struct bpf_timer);
382 	case BPF_WORKQUEUE:
383 		return __alignof__(struct bpf_wq);
384 	case BPF_KPTR_UNREF:
385 	case BPF_KPTR_REF:
386 	case BPF_KPTR_PERCPU:
387 	case BPF_UPTR:
388 		return __alignof__(u64);
389 	case BPF_LIST_HEAD:
390 		return __alignof__(struct bpf_list_head);
391 	case BPF_LIST_NODE:
392 		return __alignof__(struct bpf_list_node);
393 	case BPF_RB_ROOT:
394 		return __alignof__(struct bpf_rb_root);
395 	case BPF_RB_NODE:
396 		return __alignof__(struct bpf_rb_node);
397 	case BPF_REFCOUNT:
398 		return __alignof__(struct bpf_refcount);
399 	default:
400 		WARN_ON_ONCE(1);
401 		return 0;
402 	}
403 }
404 
405 static inline void bpf_obj_init_field(const struct btf_field *field, void *addr)
406 {
407 	memset(addr, 0, field->size);
408 
409 	switch (field->type) {
410 	case BPF_REFCOUNT:
411 		refcount_set((refcount_t *)addr, 1);
412 		break;
413 	case BPF_RB_NODE:
414 		RB_CLEAR_NODE((struct rb_node *)addr);
415 		break;
416 	case BPF_LIST_HEAD:
417 	case BPF_LIST_NODE:
418 		INIT_LIST_HEAD((struct list_head *)addr);
419 		break;
420 	case BPF_RB_ROOT:
421 		/* RB_ROOT_CACHED 0-inits, no need to do anything after memset */
422 	case BPF_SPIN_LOCK:
423 	case BPF_TIMER:
424 	case BPF_WORKQUEUE:
425 	case BPF_KPTR_UNREF:
426 	case BPF_KPTR_REF:
427 	case BPF_KPTR_PERCPU:
428 	case BPF_UPTR:
429 		break;
430 	default:
431 		WARN_ON_ONCE(1);
432 		return;
433 	}
434 }
435 
436 static inline bool btf_record_has_field(const struct btf_record *rec, enum btf_field_type type)
437 {
438 	if (IS_ERR_OR_NULL(rec))
439 		return false;
440 	return rec->field_mask & type;
441 }
442 
443 static inline void bpf_obj_init(const struct btf_record *rec, void *obj)
444 {
445 	int i;
446 
447 	if (IS_ERR_OR_NULL(rec))
448 		return;
449 	for (i = 0; i < rec->cnt; i++)
450 		bpf_obj_init_field(&rec->fields[i], obj + rec->fields[i].offset);
451 }
452 
453 /* 'dst' must be a temporary buffer and should not point to memory that is being
454  * used in parallel by a bpf program or bpf syscall, otherwise the access from
455  * the bpf program or bpf syscall may be corrupted by the reinitialization,
456  * leading to weird problems. Even 'dst' is newly-allocated from bpf memory
457  * allocator, it is still possible for 'dst' to be used in parallel by a bpf
458  * program or bpf syscall.
459  */
460 static inline void check_and_init_map_value(struct bpf_map *map, void *dst)
461 {
462 	bpf_obj_init(map->record, dst);
463 }
464 
465 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
466  * forced to use 'long' read/writes to try to atomically copy long counters.
467  * Best-effort only.  No barriers here, since it _will_ race with concurrent
468  * updates from BPF programs. Called from bpf syscall and mostly used with
469  * size 8 or 16 bytes, so ask compiler to inline it.
470  */
471 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
472 {
473 	const long *lsrc = src;
474 	long *ldst = dst;
475 
476 	size /= sizeof(long);
477 	while (size--)
478 		data_race(*ldst++ = *lsrc++);
479 }
480 
481 /* copy everything but bpf_spin_lock, bpf_timer, and kptrs. There could be one of each. */
482 static inline void bpf_obj_memcpy(struct btf_record *rec,
483 				  void *dst, void *src, u32 size,
484 				  bool long_memcpy)
485 {
486 	u32 curr_off = 0;
487 	int i;
488 
489 	if (IS_ERR_OR_NULL(rec)) {
490 		if (long_memcpy)
491 			bpf_long_memcpy(dst, src, round_up(size, 8));
492 		else
493 			memcpy(dst, src, size);
494 		return;
495 	}
496 
497 	for (i = 0; i < rec->cnt; i++) {
498 		u32 next_off = rec->fields[i].offset;
499 		u32 sz = next_off - curr_off;
500 
501 		memcpy(dst + curr_off, src + curr_off, sz);
502 		curr_off += rec->fields[i].size + sz;
503 	}
504 	memcpy(dst + curr_off, src + curr_off, size - curr_off);
505 }
506 
507 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
508 {
509 	bpf_obj_memcpy(map->record, dst, src, map->value_size, false);
510 }
511 
512 static inline void copy_map_value_long(struct bpf_map *map, void *dst, void *src)
513 {
514 	bpf_obj_memcpy(map->record, dst, src, map->value_size, true);
515 }
516 
517 static inline void bpf_obj_swap_uptrs(const struct btf_record *rec, void *dst, void *src)
518 {
519 	unsigned long *src_uptr, *dst_uptr;
520 	const struct btf_field *field;
521 	int i;
522 
523 	if (!btf_record_has_field(rec, BPF_UPTR))
524 		return;
525 
526 	for (i = 0, field = rec->fields; i < rec->cnt; i++, field++) {
527 		if (field->type != BPF_UPTR)
528 			continue;
529 
530 		src_uptr = src + field->offset;
531 		dst_uptr = dst + field->offset;
532 		swap(*src_uptr, *dst_uptr);
533 	}
534 }
535 
536 static inline void bpf_obj_memzero(struct btf_record *rec, void *dst, u32 size)
537 {
538 	u32 curr_off = 0;
539 	int i;
540 
541 	if (IS_ERR_OR_NULL(rec)) {
542 		memset(dst, 0, size);
543 		return;
544 	}
545 
546 	for (i = 0; i < rec->cnt; i++) {
547 		u32 next_off = rec->fields[i].offset;
548 		u32 sz = next_off - curr_off;
549 
550 		memset(dst + curr_off, 0, sz);
551 		curr_off += rec->fields[i].size + sz;
552 	}
553 	memset(dst + curr_off, 0, size - curr_off);
554 }
555 
556 static inline void zero_map_value(struct bpf_map *map, void *dst)
557 {
558 	bpf_obj_memzero(map->record, dst, map->value_size);
559 }
560 
561 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
562 			   bool lock_src);
563 void bpf_timer_cancel_and_free(void *timer);
564 void bpf_wq_cancel_and_free(void *timer);
565 void bpf_list_head_free(const struct btf_field *field, void *list_head,
566 			struct bpf_spin_lock *spin_lock);
567 void bpf_rb_root_free(const struct btf_field *field, void *rb_root,
568 		      struct bpf_spin_lock *spin_lock);
569 u64 bpf_arena_get_kern_vm_start(struct bpf_arena *arena);
570 u64 bpf_arena_get_user_vm_start(struct bpf_arena *arena);
571 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
572 
573 struct bpf_offload_dev;
574 struct bpf_offloaded_map;
575 
576 struct bpf_map_dev_ops {
577 	int (*map_get_next_key)(struct bpf_offloaded_map *map,
578 				void *key, void *next_key);
579 	int (*map_lookup_elem)(struct bpf_offloaded_map *map,
580 			       void *key, void *value);
581 	int (*map_update_elem)(struct bpf_offloaded_map *map,
582 			       void *key, void *value, u64 flags);
583 	int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
584 };
585 
586 struct bpf_offloaded_map {
587 	struct bpf_map map;
588 	struct net_device *netdev;
589 	const struct bpf_map_dev_ops *dev_ops;
590 	void *dev_priv;
591 	struct list_head offloads;
592 };
593 
594 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
595 {
596 	return container_of(map, struct bpf_offloaded_map, map);
597 }
598 
599 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
600 {
601 	return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
602 }
603 
604 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
605 {
606 	return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
607 		map->ops->map_seq_show_elem;
608 }
609 
610 int map_check_no_btf(const struct bpf_map *map,
611 		     const struct btf *btf,
612 		     const struct btf_type *key_type,
613 		     const struct btf_type *value_type);
614 
615 bool bpf_map_meta_equal(const struct bpf_map *meta0,
616 			const struct bpf_map *meta1);
617 
618 extern const struct bpf_map_ops bpf_map_offload_ops;
619 
620 /* bpf_type_flag contains a set of flags that are applicable to the values of
621  * arg_type, ret_type and reg_type. For example, a pointer value may be null,
622  * or a memory is read-only. We classify types into two categories: base types
623  * and extended types. Extended types are base types combined with a type flag.
624  *
625  * Currently there are no more than 32 base types in arg_type, ret_type and
626  * reg_types.
627  */
628 #define BPF_BASE_TYPE_BITS	8
629 
630 enum bpf_type_flag {
631 	/* PTR may be NULL. */
632 	PTR_MAYBE_NULL		= BIT(0 + BPF_BASE_TYPE_BITS),
633 
634 	/* MEM is read-only. When applied on bpf_arg, it indicates the arg is
635 	 * compatible with both mutable and immutable memory.
636 	 */
637 	MEM_RDONLY		= BIT(1 + BPF_BASE_TYPE_BITS),
638 
639 	/* MEM points to BPF ring buffer reservation. */
640 	MEM_RINGBUF		= BIT(2 + BPF_BASE_TYPE_BITS),
641 
642 	/* MEM is in user address space. */
643 	MEM_USER		= BIT(3 + BPF_BASE_TYPE_BITS),
644 
645 	/* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged
646 	 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In
647 	 * order to drop this tag, it must be passed into bpf_per_cpu_ptr()
648 	 * or bpf_this_cpu_ptr(), which will return the pointer corresponding
649 	 * to the specified cpu.
650 	 */
651 	MEM_PERCPU		= BIT(4 + BPF_BASE_TYPE_BITS),
652 
653 	/* Indicates that the argument will be released. */
654 	OBJ_RELEASE		= BIT(5 + BPF_BASE_TYPE_BITS),
655 
656 	/* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark
657 	 * unreferenced and referenced kptr loaded from map value using a load
658 	 * instruction, so that they can only be dereferenced but not escape the
659 	 * BPF program into the kernel (i.e. cannot be passed as arguments to
660 	 * kfunc or bpf helpers).
661 	 */
662 	PTR_UNTRUSTED		= BIT(6 + BPF_BASE_TYPE_BITS),
663 
664 	/* MEM can be uninitialized. */
665 	MEM_UNINIT		= BIT(7 + BPF_BASE_TYPE_BITS),
666 
667 	/* DYNPTR points to memory local to the bpf program. */
668 	DYNPTR_TYPE_LOCAL	= BIT(8 + BPF_BASE_TYPE_BITS),
669 
670 	/* DYNPTR points to a kernel-produced ringbuf record. */
671 	DYNPTR_TYPE_RINGBUF	= BIT(9 + BPF_BASE_TYPE_BITS),
672 
673 	/* Size is known at compile time. */
674 	MEM_FIXED_SIZE		= BIT(10 + BPF_BASE_TYPE_BITS),
675 
676 	/* MEM is of an allocated object of type in program BTF. This is used to
677 	 * tag PTR_TO_BTF_ID allocated using bpf_obj_new.
678 	 */
679 	MEM_ALLOC		= BIT(11 + BPF_BASE_TYPE_BITS),
680 
681 	/* PTR was passed from the kernel in a trusted context, and may be
682 	 * passed to KF_TRUSTED_ARGS kfuncs or BPF helper functions.
683 	 * Confusingly, this is _not_ the opposite of PTR_UNTRUSTED above.
684 	 * PTR_UNTRUSTED refers to a kptr that was read directly from a map
685 	 * without invoking bpf_kptr_xchg(). What we really need to know is
686 	 * whether a pointer is safe to pass to a kfunc or BPF helper function.
687 	 * While PTR_UNTRUSTED pointers are unsafe to pass to kfuncs and BPF
688 	 * helpers, they do not cover all possible instances of unsafe
689 	 * pointers. For example, a pointer that was obtained from walking a
690 	 * struct will _not_ get the PTR_UNTRUSTED type modifier, despite the
691 	 * fact that it may be NULL, invalid, etc. This is due to backwards
692 	 * compatibility requirements, as this was the behavior that was first
693 	 * introduced when kptrs were added. The behavior is now considered
694 	 * deprecated, and PTR_UNTRUSTED will eventually be removed.
695 	 *
696 	 * PTR_TRUSTED, on the other hand, is a pointer that the kernel
697 	 * guarantees to be valid and safe to pass to kfuncs and BPF helpers.
698 	 * For example, pointers passed to tracepoint arguments are considered
699 	 * PTR_TRUSTED, as are pointers that are passed to struct_ops
700 	 * callbacks. As alluded to above, pointers that are obtained from
701 	 * walking PTR_TRUSTED pointers are _not_ trusted. For example, if a
702 	 * struct task_struct *task is PTR_TRUSTED, then accessing
703 	 * task->last_wakee will lose the PTR_TRUSTED modifier when it's stored
704 	 * in a BPF register. Similarly, pointers passed to certain programs
705 	 * types such as kretprobes are not guaranteed to be valid, as they may
706 	 * for example contain an object that was recently freed.
707 	 */
708 	PTR_TRUSTED		= BIT(12 + BPF_BASE_TYPE_BITS),
709 
710 	/* MEM is tagged with rcu and memory access needs rcu_read_lock protection. */
711 	MEM_RCU			= BIT(13 + BPF_BASE_TYPE_BITS),
712 
713 	/* Used to tag PTR_TO_BTF_ID | MEM_ALLOC references which are non-owning.
714 	 * Currently only valid for linked-list and rbtree nodes. If the nodes
715 	 * have a bpf_refcount_field, they must be tagged MEM_RCU as well.
716 	 */
717 	NON_OWN_REF		= BIT(14 + BPF_BASE_TYPE_BITS),
718 
719 	/* DYNPTR points to sk_buff */
720 	DYNPTR_TYPE_SKB		= BIT(15 + BPF_BASE_TYPE_BITS),
721 
722 	/* DYNPTR points to xdp_buff */
723 	DYNPTR_TYPE_XDP		= BIT(16 + BPF_BASE_TYPE_BITS),
724 
725 	/* Memory must be aligned on some architectures, used in combination with
726 	 * MEM_FIXED_SIZE.
727 	 */
728 	MEM_ALIGNED		= BIT(17 + BPF_BASE_TYPE_BITS),
729 
730 	/* MEM is being written to, often combined with MEM_UNINIT. Non-presence
731 	 * of MEM_WRITE means that MEM is only being read. MEM_WRITE without the
732 	 * MEM_UNINIT means that memory needs to be initialized since it is also
733 	 * read.
734 	 */
735 	MEM_WRITE		= BIT(18 + BPF_BASE_TYPE_BITS),
736 
737 	__BPF_TYPE_FLAG_MAX,
738 	__BPF_TYPE_LAST_FLAG	= __BPF_TYPE_FLAG_MAX - 1,
739 };
740 
741 #define DYNPTR_TYPE_FLAG_MASK	(DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF | DYNPTR_TYPE_SKB \
742 				 | DYNPTR_TYPE_XDP)
743 
744 /* Max number of base types. */
745 #define BPF_BASE_TYPE_LIMIT	(1UL << BPF_BASE_TYPE_BITS)
746 
747 /* Max number of all types. */
748 #define BPF_TYPE_LIMIT		(__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1))
749 
750 /* function argument constraints */
751 enum bpf_arg_type {
752 	ARG_DONTCARE = 0,	/* unused argument in helper function */
753 
754 	/* the following constraints used to prototype
755 	 * bpf_map_lookup/update/delete_elem() functions
756 	 */
757 	ARG_CONST_MAP_PTR,	/* const argument used as pointer to bpf_map */
758 	ARG_PTR_TO_MAP_KEY,	/* pointer to stack used as map key */
759 	ARG_PTR_TO_MAP_VALUE,	/* pointer to stack used as map value */
760 
761 	/* Used to prototype bpf_memcmp() and other functions that access data
762 	 * on eBPF program stack
763 	 */
764 	ARG_PTR_TO_MEM,		/* pointer to valid memory (stack, packet, map value) */
765 	ARG_PTR_TO_ARENA,
766 
767 	ARG_CONST_SIZE,		/* number of bytes accessed from memory */
768 	ARG_CONST_SIZE_OR_ZERO,	/* number of bytes accessed from memory or 0 */
769 
770 	ARG_PTR_TO_CTX,		/* pointer to context */
771 	ARG_ANYTHING,		/* any (initialized) argument is ok */
772 	ARG_PTR_TO_SPIN_LOCK,	/* pointer to bpf_spin_lock */
773 	ARG_PTR_TO_SOCK_COMMON,	/* pointer to sock_common */
774 	ARG_PTR_TO_SOCKET,	/* pointer to bpf_sock (fullsock) */
775 	ARG_PTR_TO_BTF_ID,	/* pointer to in-kernel struct */
776 	ARG_PTR_TO_RINGBUF_MEM,	/* pointer to dynamically reserved ringbuf memory */
777 	ARG_CONST_ALLOC_SIZE_OR_ZERO,	/* number of allocated bytes requested */
778 	ARG_PTR_TO_BTF_ID_SOCK_COMMON,	/* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
779 	ARG_PTR_TO_PERCPU_BTF_ID,	/* pointer to in-kernel percpu type */
780 	ARG_PTR_TO_FUNC,	/* pointer to a bpf program function */
781 	ARG_PTR_TO_STACK,	/* pointer to stack */
782 	ARG_PTR_TO_CONST_STR,	/* pointer to a null terminated read-only string */
783 	ARG_PTR_TO_TIMER,	/* pointer to bpf_timer */
784 	ARG_KPTR_XCHG_DEST,	/* pointer to destination that kptrs are bpf_kptr_xchg'd into */
785 	ARG_PTR_TO_DYNPTR,      /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */
786 	__BPF_ARG_TYPE_MAX,
787 
788 	/* Extended arg_types. */
789 	ARG_PTR_TO_MAP_VALUE_OR_NULL	= PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE,
790 	ARG_PTR_TO_MEM_OR_NULL		= PTR_MAYBE_NULL | ARG_PTR_TO_MEM,
791 	ARG_PTR_TO_CTX_OR_NULL		= PTR_MAYBE_NULL | ARG_PTR_TO_CTX,
792 	ARG_PTR_TO_SOCKET_OR_NULL	= PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET,
793 	ARG_PTR_TO_STACK_OR_NULL	= PTR_MAYBE_NULL | ARG_PTR_TO_STACK,
794 	ARG_PTR_TO_BTF_ID_OR_NULL	= PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID,
795 	/* Pointer to memory does not need to be initialized, since helper function
796 	 * fills all bytes or clears them in error case.
797 	 */
798 	ARG_PTR_TO_UNINIT_MEM		= MEM_UNINIT | MEM_WRITE | ARG_PTR_TO_MEM,
799 	/* Pointer to valid memory of size known at compile time. */
800 	ARG_PTR_TO_FIXED_SIZE_MEM	= MEM_FIXED_SIZE | ARG_PTR_TO_MEM,
801 
802 	/* This must be the last entry. Its purpose is to ensure the enum is
803 	 * wide enough to hold the higher bits reserved for bpf_type_flag.
804 	 */
805 	__BPF_ARG_TYPE_LIMIT	= BPF_TYPE_LIMIT,
806 };
807 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
808 
809 /* type of values returned from helper functions */
810 enum bpf_return_type {
811 	RET_INTEGER,			/* function returns integer */
812 	RET_VOID,			/* function doesn't return anything */
813 	RET_PTR_TO_MAP_VALUE,		/* returns a pointer to map elem value */
814 	RET_PTR_TO_SOCKET,		/* returns a pointer to a socket */
815 	RET_PTR_TO_TCP_SOCK,		/* returns a pointer to a tcp_sock */
816 	RET_PTR_TO_SOCK_COMMON,		/* returns a pointer to a sock_common */
817 	RET_PTR_TO_MEM,			/* returns a pointer to memory */
818 	RET_PTR_TO_MEM_OR_BTF_ID,	/* returns a pointer to a valid memory or a btf_id */
819 	RET_PTR_TO_BTF_ID,		/* returns a pointer to a btf_id */
820 	__BPF_RET_TYPE_MAX,
821 
822 	/* Extended ret_types. */
823 	RET_PTR_TO_MAP_VALUE_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE,
824 	RET_PTR_TO_SOCKET_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_SOCKET,
825 	RET_PTR_TO_TCP_SOCK_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK,
826 	RET_PTR_TO_SOCK_COMMON_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON,
827 	RET_PTR_TO_RINGBUF_MEM_OR_NULL	= PTR_MAYBE_NULL | MEM_RINGBUF | RET_PTR_TO_MEM,
828 	RET_PTR_TO_DYNPTR_MEM_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_MEM,
829 	RET_PTR_TO_BTF_ID_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID,
830 	RET_PTR_TO_BTF_ID_TRUSTED	= PTR_TRUSTED	 | RET_PTR_TO_BTF_ID,
831 
832 	/* This must be the last entry. Its purpose is to ensure the enum is
833 	 * wide enough to hold the higher bits reserved for bpf_type_flag.
834 	 */
835 	__BPF_RET_TYPE_LIMIT	= BPF_TYPE_LIMIT,
836 };
837 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
838 
839 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
840  * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
841  * instructions after verifying
842  */
843 struct bpf_func_proto {
844 	u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
845 	bool gpl_only;
846 	bool pkt_access;
847 	bool might_sleep;
848 	/* set to true if helper follows contract for llvm
849 	 * attribute bpf_fastcall:
850 	 * - void functions do not scratch r0
851 	 * - functions taking N arguments scratch only registers r1-rN
852 	 */
853 	bool allow_fastcall;
854 	enum bpf_return_type ret_type;
855 	union {
856 		struct {
857 			enum bpf_arg_type arg1_type;
858 			enum bpf_arg_type arg2_type;
859 			enum bpf_arg_type arg3_type;
860 			enum bpf_arg_type arg4_type;
861 			enum bpf_arg_type arg5_type;
862 		};
863 		enum bpf_arg_type arg_type[5];
864 	};
865 	union {
866 		struct {
867 			u32 *arg1_btf_id;
868 			u32 *arg2_btf_id;
869 			u32 *arg3_btf_id;
870 			u32 *arg4_btf_id;
871 			u32 *arg5_btf_id;
872 		};
873 		u32 *arg_btf_id[5];
874 		struct {
875 			size_t arg1_size;
876 			size_t arg2_size;
877 			size_t arg3_size;
878 			size_t arg4_size;
879 			size_t arg5_size;
880 		};
881 		size_t arg_size[5];
882 	};
883 	int *ret_btf_id; /* return value btf_id */
884 	bool (*allowed)(const struct bpf_prog *prog);
885 };
886 
887 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
888  * the first argument to eBPF programs.
889  * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
890  */
891 struct bpf_context;
892 
893 enum bpf_access_type {
894 	BPF_READ = 1,
895 	BPF_WRITE = 2
896 };
897 
898 /* types of values stored in eBPF registers */
899 /* Pointer types represent:
900  * pointer
901  * pointer + imm
902  * pointer + (u16) var
903  * pointer + (u16) var + imm
904  * if (range > 0) then [ptr, ptr + range - off) is safe to access
905  * if (id > 0) means that some 'var' was added
906  * if (off > 0) means that 'imm' was added
907  */
908 enum bpf_reg_type {
909 	NOT_INIT = 0,		 /* nothing was written into register */
910 	SCALAR_VALUE,		 /* reg doesn't contain a valid pointer */
911 	PTR_TO_CTX,		 /* reg points to bpf_context */
912 	CONST_PTR_TO_MAP,	 /* reg points to struct bpf_map */
913 	PTR_TO_MAP_VALUE,	 /* reg points to map element value */
914 	PTR_TO_MAP_KEY,		 /* reg points to a map element key */
915 	PTR_TO_STACK,		 /* reg == frame_pointer + offset */
916 	PTR_TO_PACKET_META,	 /* skb->data - meta_len */
917 	PTR_TO_PACKET,		 /* reg points to skb->data */
918 	PTR_TO_PACKET_END,	 /* skb->data + headlen */
919 	PTR_TO_FLOW_KEYS,	 /* reg points to bpf_flow_keys */
920 	PTR_TO_SOCKET,		 /* reg points to struct bpf_sock */
921 	PTR_TO_SOCK_COMMON,	 /* reg points to sock_common */
922 	PTR_TO_TCP_SOCK,	 /* reg points to struct tcp_sock */
923 	PTR_TO_TP_BUFFER,	 /* reg points to a writable raw tp's buffer */
924 	PTR_TO_XDP_SOCK,	 /* reg points to struct xdp_sock */
925 	/* PTR_TO_BTF_ID points to a kernel struct that does not need
926 	 * to be null checked by the BPF program. This does not imply the
927 	 * pointer is _not_ null and in practice this can easily be a null
928 	 * pointer when reading pointer chains. The assumption is program
929 	 * context will handle null pointer dereference typically via fault
930 	 * handling. The verifier must keep this in mind and can make no
931 	 * assumptions about null or non-null when doing branch analysis.
932 	 * Further, when passed into helpers the helpers can not, without
933 	 * additional context, assume the value is non-null.
934 	 */
935 	PTR_TO_BTF_ID,
936 	PTR_TO_MEM,		 /* reg points to valid memory region */
937 	PTR_TO_ARENA,
938 	PTR_TO_BUF,		 /* reg points to a read/write buffer */
939 	PTR_TO_FUNC,		 /* reg points to a bpf program function */
940 	CONST_PTR_TO_DYNPTR,	 /* reg points to a const struct bpf_dynptr */
941 	__BPF_REG_TYPE_MAX,
942 
943 	/* Extended reg_types. */
944 	PTR_TO_MAP_VALUE_OR_NULL	= PTR_MAYBE_NULL | PTR_TO_MAP_VALUE,
945 	PTR_TO_SOCKET_OR_NULL		= PTR_MAYBE_NULL | PTR_TO_SOCKET,
946 	PTR_TO_SOCK_COMMON_OR_NULL	= PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON,
947 	PTR_TO_TCP_SOCK_OR_NULL		= PTR_MAYBE_NULL | PTR_TO_TCP_SOCK,
948 	/* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
949 	 * been checked for null. Used primarily to inform the verifier
950 	 * an explicit null check is required for this struct.
951 	 */
952 	PTR_TO_BTF_ID_OR_NULL		= PTR_MAYBE_NULL | PTR_TO_BTF_ID,
953 
954 	/* This must be the last entry. Its purpose is to ensure the enum is
955 	 * wide enough to hold the higher bits reserved for bpf_type_flag.
956 	 */
957 	__BPF_REG_TYPE_LIMIT	= BPF_TYPE_LIMIT,
958 };
959 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
960 
961 /* The information passed from prog-specific *_is_valid_access
962  * back to the verifier.
963  */
964 struct bpf_insn_access_aux {
965 	enum bpf_reg_type reg_type;
966 	bool is_ldsx;
967 	union {
968 		int ctx_field_size;
969 		struct {
970 			struct btf *btf;
971 			u32 btf_id;
972 			u32 ref_obj_id;
973 		};
974 	};
975 	struct bpf_verifier_log *log; /* for verbose logs */
976 	bool is_retval; /* is accessing function return value ? */
977 };
978 
979 static inline void
980 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
981 {
982 	aux->ctx_field_size = size;
983 }
984 
985 static bool bpf_is_ldimm64(const struct bpf_insn *insn)
986 {
987 	return insn->code == (BPF_LD | BPF_IMM | BPF_DW);
988 }
989 
990 static inline bool bpf_pseudo_func(const struct bpf_insn *insn)
991 {
992 	return bpf_is_ldimm64(insn) && insn->src_reg == BPF_PSEUDO_FUNC;
993 }
994 
995 /* Given a BPF_ATOMIC instruction @atomic_insn, return true if it is an
996  * atomic load or store, and false if it is a read-modify-write instruction.
997  */
998 static inline bool
999 bpf_atomic_is_load_store(const struct bpf_insn *atomic_insn)
1000 {
1001 	switch (atomic_insn->imm) {
1002 	case BPF_LOAD_ACQ:
1003 	case BPF_STORE_REL:
1004 		return true;
1005 	default:
1006 		return false;
1007 	}
1008 }
1009 
1010 struct bpf_prog_ops {
1011 	int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
1012 			union bpf_attr __user *uattr);
1013 };
1014 
1015 struct bpf_reg_state;
1016 struct bpf_verifier_ops {
1017 	/* return eBPF function prototype for verification */
1018 	const struct bpf_func_proto *
1019 	(*get_func_proto)(enum bpf_func_id func_id,
1020 			  const struct bpf_prog *prog);
1021 
1022 	/* return true if 'size' wide access at offset 'off' within bpf_context
1023 	 * with 'type' (read or write) is allowed
1024 	 */
1025 	bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
1026 				const struct bpf_prog *prog,
1027 				struct bpf_insn_access_aux *info);
1028 	int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
1029 			    const struct bpf_prog *prog);
1030 	int (*gen_epilogue)(struct bpf_insn *insn, const struct bpf_prog *prog,
1031 			    s16 ctx_stack_off);
1032 	int (*gen_ld_abs)(const struct bpf_insn *orig,
1033 			  struct bpf_insn *insn_buf);
1034 	u32 (*convert_ctx_access)(enum bpf_access_type type,
1035 				  const struct bpf_insn *src,
1036 				  struct bpf_insn *dst,
1037 				  struct bpf_prog *prog, u32 *target_size);
1038 	int (*btf_struct_access)(struct bpf_verifier_log *log,
1039 				 const struct bpf_reg_state *reg,
1040 				 int off, int size);
1041 };
1042 
1043 struct bpf_prog_offload_ops {
1044 	/* verifier basic callbacks */
1045 	int (*insn_hook)(struct bpf_verifier_env *env,
1046 			 int insn_idx, int prev_insn_idx);
1047 	int (*finalize)(struct bpf_verifier_env *env);
1048 	/* verifier optimization callbacks (called after .finalize) */
1049 	int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
1050 			    struct bpf_insn *insn);
1051 	int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
1052 	/* program management callbacks */
1053 	int (*prepare)(struct bpf_prog *prog);
1054 	int (*translate)(struct bpf_prog *prog);
1055 	void (*destroy)(struct bpf_prog *prog);
1056 };
1057 
1058 struct bpf_prog_offload {
1059 	struct bpf_prog		*prog;
1060 	struct net_device	*netdev;
1061 	struct bpf_offload_dev	*offdev;
1062 	void			*dev_priv;
1063 	struct list_head	offloads;
1064 	bool			dev_state;
1065 	bool			opt_failed;
1066 	void			*jited_image;
1067 	u32			jited_len;
1068 };
1069 
1070 enum bpf_cgroup_storage_type {
1071 	BPF_CGROUP_STORAGE_SHARED,
1072 	BPF_CGROUP_STORAGE_PERCPU,
1073 	__BPF_CGROUP_STORAGE_MAX
1074 };
1075 
1076 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
1077 
1078 /* The longest tracepoint has 12 args.
1079  * See include/trace/bpf_probe.h
1080  */
1081 #define MAX_BPF_FUNC_ARGS 12
1082 
1083 /* The maximum number of arguments passed through registers
1084  * a single function may have.
1085  */
1086 #define MAX_BPF_FUNC_REG_ARGS 5
1087 
1088 /* The argument is a structure. */
1089 #define BTF_FMODEL_STRUCT_ARG		BIT(0)
1090 
1091 /* The argument is signed. */
1092 #define BTF_FMODEL_SIGNED_ARG		BIT(1)
1093 
1094 struct btf_func_model {
1095 	u8 ret_size;
1096 	u8 ret_flags;
1097 	u8 nr_args;
1098 	u8 arg_size[MAX_BPF_FUNC_ARGS];
1099 	u8 arg_flags[MAX_BPF_FUNC_ARGS];
1100 };
1101 
1102 /* Restore arguments before returning from trampoline to let original function
1103  * continue executing. This flag is used for fentry progs when there are no
1104  * fexit progs.
1105  */
1106 #define BPF_TRAMP_F_RESTORE_REGS	BIT(0)
1107 /* Call original function after fentry progs, but before fexit progs.
1108  * Makes sense for fentry/fexit, normal calls and indirect calls.
1109  */
1110 #define BPF_TRAMP_F_CALL_ORIG		BIT(1)
1111 /* Skip current frame and return to parent.  Makes sense for fentry/fexit
1112  * programs only. Should not be used with normal calls and indirect calls.
1113  */
1114 #define BPF_TRAMP_F_SKIP_FRAME		BIT(2)
1115 /* Store IP address of the caller on the trampoline stack,
1116  * so it's available for trampoline's programs.
1117  */
1118 #define BPF_TRAMP_F_IP_ARG		BIT(3)
1119 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */
1120 #define BPF_TRAMP_F_RET_FENTRY_RET	BIT(4)
1121 
1122 /* Get original function from stack instead of from provided direct address.
1123  * Makes sense for trampolines with fexit or fmod_ret programs.
1124  */
1125 #define BPF_TRAMP_F_ORIG_STACK		BIT(5)
1126 
1127 /* This trampoline is on a function with another ftrace_ops with IPMODIFY,
1128  * e.g., a live patch. This flag is set and cleared by ftrace call backs,
1129  */
1130 #define BPF_TRAMP_F_SHARE_IPMODIFY	BIT(6)
1131 
1132 /* Indicate that current trampoline is in a tail call context. Then, it has to
1133  * cache and restore tail_call_cnt to avoid infinite tail call loop.
1134  */
1135 #define BPF_TRAMP_F_TAIL_CALL_CTX	BIT(7)
1136 
1137 /*
1138  * Indicate the trampoline should be suitable to receive indirect calls;
1139  * without this indirectly calling the generated code can result in #UD/#CP,
1140  * depending on the CFI options.
1141  *
1142  * Used by bpf_struct_ops.
1143  *
1144  * Incompatible with FENTRY usage, overloads @func_addr argument.
1145  */
1146 #define BPF_TRAMP_F_INDIRECT		BIT(8)
1147 
1148 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
1149  * bytes on x86.
1150  */
1151 enum {
1152 #if defined(__s390x__)
1153 	BPF_MAX_TRAMP_LINKS = 27,
1154 #else
1155 	BPF_MAX_TRAMP_LINKS = 38,
1156 #endif
1157 };
1158 
1159 struct bpf_tramp_links {
1160 	struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS];
1161 	int nr_links;
1162 };
1163 
1164 struct bpf_tramp_run_ctx;
1165 
1166 /* Different use cases for BPF trampoline:
1167  * 1. replace nop at the function entry (kprobe equivalent)
1168  *    flags = BPF_TRAMP_F_RESTORE_REGS
1169  *    fentry = a set of programs to run before returning from trampoline
1170  *
1171  * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
1172  *    flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
1173  *    orig_call = fentry_ip + MCOUNT_INSN_SIZE
1174  *    fentry = a set of program to run before calling original function
1175  *    fexit = a set of program to run after original function
1176  *
1177  * 3. replace direct call instruction anywhere in the function body
1178  *    or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
1179  *    With flags = 0
1180  *      fentry = a set of programs to run before returning from trampoline
1181  *    With flags = BPF_TRAMP_F_CALL_ORIG
1182  *      orig_call = original callback addr or direct function addr
1183  *      fentry = a set of program to run before calling original function
1184  *      fexit = a set of program to run after original function
1185  */
1186 struct bpf_tramp_image;
1187 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *image_end,
1188 				const struct btf_func_model *m, u32 flags,
1189 				struct bpf_tramp_links *tlinks,
1190 				void *func_addr);
1191 void *arch_alloc_bpf_trampoline(unsigned int size);
1192 void arch_free_bpf_trampoline(void *image, unsigned int size);
1193 int __must_check arch_protect_bpf_trampoline(void *image, unsigned int size);
1194 int arch_bpf_trampoline_size(const struct btf_func_model *m, u32 flags,
1195 			     struct bpf_tramp_links *tlinks, void *func_addr);
1196 
1197 u64 notrace __bpf_prog_enter_sleepable_recur(struct bpf_prog *prog,
1198 					     struct bpf_tramp_run_ctx *run_ctx);
1199 void notrace __bpf_prog_exit_sleepable_recur(struct bpf_prog *prog, u64 start,
1200 					     struct bpf_tramp_run_ctx *run_ctx);
1201 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
1202 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
1203 typedef u64 (*bpf_trampoline_enter_t)(struct bpf_prog *prog,
1204 				      struct bpf_tramp_run_ctx *run_ctx);
1205 typedef void (*bpf_trampoline_exit_t)(struct bpf_prog *prog, u64 start,
1206 				      struct bpf_tramp_run_ctx *run_ctx);
1207 bpf_trampoline_enter_t bpf_trampoline_enter(const struct bpf_prog *prog);
1208 bpf_trampoline_exit_t bpf_trampoline_exit(const struct bpf_prog *prog);
1209 
1210 struct bpf_ksym {
1211 	unsigned long		 start;
1212 	unsigned long		 end;
1213 	char			 name[KSYM_NAME_LEN];
1214 	struct list_head	 lnode;
1215 	struct latch_tree_node	 tnode;
1216 	bool			 prog;
1217 };
1218 
1219 enum bpf_tramp_prog_type {
1220 	BPF_TRAMP_FENTRY,
1221 	BPF_TRAMP_FEXIT,
1222 	BPF_TRAMP_MODIFY_RETURN,
1223 	BPF_TRAMP_MAX,
1224 	BPF_TRAMP_REPLACE, /* more than MAX */
1225 };
1226 
1227 struct bpf_tramp_image {
1228 	void *image;
1229 	int size;
1230 	struct bpf_ksym ksym;
1231 	struct percpu_ref pcref;
1232 	void *ip_after_call;
1233 	void *ip_epilogue;
1234 	union {
1235 		struct rcu_head rcu;
1236 		struct work_struct work;
1237 	};
1238 };
1239 
1240 struct bpf_trampoline {
1241 	/* hlist for trampoline_table */
1242 	struct hlist_node hlist;
1243 	struct ftrace_ops *fops;
1244 	/* serializes access to fields of this trampoline */
1245 	struct mutex mutex;
1246 	refcount_t refcnt;
1247 	u32 flags;
1248 	u64 key;
1249 	struct {
1250 		struct btf_func_model model;
1251 		void *addr;
1252 		bool ftrace_managed;
1253 	} func;
1254 	/* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
1255 	 * program by replacing one of its functions. func.addr is the address
1256 	 * of the function it replaced.
1257 	 */
1258 	struct bpf_prog *extension_prog;
1259 	/* list of BPF programs using this trampoline */
1260 	struct hlist_head progs_hlist[BPF_TRAMP_MAX];
1261 	/* Number of attached programs. A counter per kind. */
1262 	int progs_cnt[BPF_TRAMP_MAX];
1263 	/* Executable image of trampoline */
1264 	struct bpf_tramp_image *cur_image;
1265 };
1266 
1267 struct bpf_attach_target_info {
1268 	struct btf_func_model fmodel;
1269 	long tgt_addr;
1270 	struct module *tgt_mod;
1271 	const char *tgt_name;
1272 	const struct btf_type *tgt_type;
1273 };
1274 
1275 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
1276 
1277 struct bpf_dispatcher_prog {
1278 	struct bpf_prog *prog;
1279 	refcount_t users;
1280 };
1281 
1282 struct bpf_dispatcher {
1283 	/* dispatcher mutex */
1284 	struct mutex mutex;
1285 	void *func;
1286 	struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
1287 	int num_progs;
1288 	void *image;
1289 	void *rw_image;
1290 	u32 image_off;
1291 	struct bpf_ksym ksym;
1292 #ifdef CONFIG_HAVE_STATIC_CALL
1293 	struct static_call_key *sc_key;
1294 	void *sc_tramp;
1295 #endif
1296 };
1297 
1298 #ifndef __bpfcall
1299 #define __bpfcall __nocfi
1300 #endif
1301 
1302 static __always_inline __bpfcall unsigned int bpf_dispatcher_nop_func(
1303 	const void *ctx,
1304 	const struct bpf_insn *insnsi,
1305 	bpf_func_t bpf_func)
1306 {
1307 	return bpf_func(ctx, insnsi);
1308 }
1309 
1310 /* the implementation of the opaque uapi struct bpf_dynptr */
1311 struct bpf_dynptr_kern {
1312 	void *data;
1313 	/* Size represents the number of usable bytes of dynptr data.
1314 	 * If for example the offset is at 4 for a local dynptr whose data is
1315 	 * of type u64, the number of usable bytes is 4.
1316 	 *
1317 	 * The upper 8 bits are reserved. It is as follows:
1318 	 * Bits 0 - 23 = size
1319 	 * Bits 24 - 30 = dynptr type
1320 	 * Bit 31 = whether dynptr is read-only
1321 	 */
1322 	u32 size;
1323 	u32 offset;
1324 } __aligned(8);
1325 
1326 enum bpf_dynptr_type {
1327 	BPF_DYNPTR_TYPE_INVALID,
1328 	/* Points to memory that is local to the bpf program */
1329 	BPF_DYNPTR_TYPE_LOCAL,
1330 	/* Underlying data is a ringbuf record */
1331 	BPF_DYNPTR_TYPE_RINGBUF,
1332 	/* Underlying data is a sk_buff */
1333 	BPF_DYNPTR_TYPE_SKB,
1334 	/* Underlying data is a xdp_buff */
1335 	BPF_DYNPTR_TYPE_XDP,
1336 };
1337 
1338 int bpf_dynptr_check_size(u32 size);
1339 u32 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr);
1340 const void *__bpf_dynptr_data(const struct bpf_dynptr_kern *ptr, u32 len);
1341 void *__bpf_dynptr_data_rw(const struct bpf_dynptr_kern *ptr, u32 len);
1342 bool __bpf_dynptr_is_rdonly(const struct bpf_dynptr_kern *ptr);
1343 
1344 #ifdef CONFIG_BPF_JIT
1345 int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
1346 			     struct bpf_trampoline *tr,
1347 			     struct bpf_prog *tgt_prog);
1348 int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
1349 			       struct bpf_trampoline *tr,
1350 			       struct bpf_prog *tgt_prog);
1351 struct bpf_trampoline *bpf_trampoline_get(u64 key,
1352 					  struct bpf_attach_target_info *tgt_info);
1353 void bpf_trampoline_put(struct bpf_trampoline *tr);
1354 int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_funcs);
1355 
1356 /*
1357  * When the architecture supports STATIC_CALL replace the bpf_dispatcher_fn
1358  * indirection with a direct call to the bpf program. If the architecture does
1359  * not have STATIC_CALL, avoid a double-indirection.
1360  */
1361 #ifdef CONFIG_HAVE_STATIC_CALL
1362 
1363 #define __BPF_DISPATCHER_SC_INIT(_name)				\
1364 	.sc_key = &STATIC_CALL_KEY(_name),			\
1365 	.sc_tramp = STATIC_CALL_TRAMP_ADDR(_name),
1366 
1367 #define __BPF_DISPATCHER_SC(name)				\
1368 	DEFINE_STATIC_CALL(bpf_dispatcher_##name##_call, bpf_dispatcher_nop_func)
1369 
1370 #define __BPF_DISPATCHER_CALL(name)				\
1371 	static_call(bpf_dispatcher_##name##_call)(ctx, insnsi, bpf_func)
1372 
1373 #define __BPF_DISPATCHER_UPDATE(_d, _new)			\
1374 	__static_call_update((_d)->sc_key, (_d)->sc_tramp, (_new))
1375 
1376 #else
1377 #define __BPF_DISPATCHER_SC_INIT(name)
1378 #define __BPF_DISPATCHER_SC(name)
1379 #define __BPF_DISPATCHER_CALL(name)		bpf_func(ctx, insnsi)
1380 #define __BPF_DISPATCHER_UPDATE(_d, _new)
1381 #endif
1382 
1383 #define BPF_DISPATCHER_INIT(_name) {				\
1384 	.mutex = __MUTEX_INITIALIZER(_name.mutex),		\
1385 	.func = &_name##_func,					\
1386 	.progs = {},						\
1387 	.num_progs = 0,						\
1388 	.image = NULL,						\
1389 	.image_off = 0,						\
1390 	.ksym = {						\
1391 		.name  = #_name,				\
1392 		.lnode = LIST_HEAD_INIT(_name.ksym.lnode),	\
1393 	},							\
1394 	__BPF_DISPATCHER_SC_INIT(_name##_call)			\
1395 }
1396 
1397 #define DEFINE_BPF_DISPATCHER(name)					\
1398 	__BPF_DISPATCHER_SC(name);					\
1399 	noinline __bpfcall unsigned int bpf_dispatcher_##name##_func(	\
1400 		const void *ctx,					\
1401 		const struct bpf_insn *insnsi,				\
1402 		bpf_func_t bpf_func)					\
1403 	{								\
1404 		return __BPF_DISPATCHER_CALL(name);			\
1405 	}								\
1406 	EXPORT_SYMBOL(bpf_dispatcher_##name##_func);			\
1407 	struct bpf_dispatcher bpf_dispatcher_##name =			\
1408 		BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
1409 
1410 #define DECLARE_BPF_DISPATCHER(name)					\
1411 	unsigned int bpf_dispatcher_##name##_func(			\
1412 		const void *ctx,					\
1413 		const struct bpf_insn *insnsi,				\
1414 		bpf_func_t bpf_func);					\
1415 	extern struct bpf_dispatcher bpf_dispatcher_##name;
1416 
1417 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
1418 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
1419 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
1420 				struct bpf_prog *to);
1421 /* Called only from JIT-enabled code, so there's no need for stubs. */
1422 void bpf_image_ksym_init(void *data, unsigned int size, struct bpf_ksym *ksym);
1423 void bpf_image_ksym_add(struct bpf_ksym *ksym);
1424 void bpf_image_ksym_del(struct bpf_ksym *ksym);
1425 void bpf_ksym_add(struct bpf_ksym *ksym);
1426 void bpf_ksym_del(struct bpf_ksym *ksym);
1427 int bpf_jit_charge_modmem(u32 size);
1428 void bpf_jit_uncharge_modmem(u32 size);
1429 bool bpf_prog_has_trampoline(const struct bpf_prog *prog);
1430 #else
1431 static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
1432 					   struct bpf_trampoline *tr,
1433 					   struct bpf_prog *tgt_prog)
1434 {
1435 	return -ENOTSUPP;
1436 }
1437 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
1438 					     struct bpf_trampoline *tr,
1439 					     struct bpf_prog *tgt_prog)
1440 {
1441 	return -ENOTSUPP;
1442 }
1443 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
1444 							struct bpf_attach_target_info *tgt_info)
1445 {
1446 	return NULL;
1447 }
1448 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
1449 #define DEFINE_BPF_DISPATCHER(name)
1450 #define DECLARE_BPF_DISPATCHER(name)
1451 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
1452 #define BPF_DISPATCHER_PTR(name) NULL
1453 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
1454 					      struct bpf_prog *from,
1455 					      struct bpf_prog *to) {}
1456 static inline bool is_bpf_image_address(unsigned long address)
1457 {
1458 	return false;
1459 }
1460 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog)
1461 {
1462 	return false;
1463 }
1464 #endif
1465 
1466 struct bpf_func_info_aux {
1467 	u16 linkage;
1468 	bool unreliable;
1469 	bool called : 1;
1470 	bool verified : 1;
1471 };
1472 
1473 enum bpf_jit_poke_reason {
1474 	BPF_POKE_REASON_TAIL_CALL,
1475 };
1476 
1477 /* Descriptor of pokes pointing /into/ the JITed image. */
1478 struct bpf_jit_poke_descriptor {
1479 	void *tailcall_target;
1480 	void *tailcall_bypass;
1481 	void *bypass_addr;
1482 	void *aux;
1483 	union {
1484 		struct {
1485 			struct bpf_map *map;
1486 			u32 key;
1487 		} tail_call;
1488 	};
1489 	bool tailcall_target_stable;
1490 	u8 adj_off;
1491 	u16 reason;
1492 	u32 insn_idx;
1493 };
1494 
1495 /* reg_type info for ctx arguments */
1496 struct bpf_ctx_arg_aux {
1497 	u32 offset;
1498 	enum bpf_reg_type reg_type;
1499 	struct btf *btf;
1500 	u32 btf_id;
1501 	u32 ref_obj_id;
1502 	bool refcounted;
1503 };
1504 
1505 struct btf_mod_pair {
1506 	struct btf *btf;
1507 	struct module *module;
1508 };
1509 
1510 struct bpf_kfunc_desc_tab;
1511 
1512 struct bpf_prog_aux {
1513 	atomic64_t refcnt;
1514 	u32 used_map_cnt;
1515 	u32 used_btf_cnt;
1516 	u32 max_ctx_offset;
1517 	u32 max_pkt_offset;
1518 	u32 max_tp_access;
1519 	u32 stack_depth;
1520 	u32 id;
1521 	u32 func_cnt; /* used by non-func prog as the number of func progs */
1522 	u32 real_func_cnt; /* includes hidden progs, only used for JIT and freeing progs */
1523 	u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
1524 	u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1525 	u32 ctx_arg_info_size;
1526 	u32 max_rdonly_access;
1527 	u32 max_rdwr_access;
1528 	struct btf *attach_btf;
1529 	struct bpf_ctx_arg_aux *ctx_arg_info;
1530 	void __percpu *priv_stack_ptr;
1531 	struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
1532 	struct bpf_prog *dst_prog;
1533 	struct bpf_trampoline *dst_trampoline;
1534 	enum bpf_prog_type saved_dst_prog_type;
1535 	enum bpf_attach_type saved_dst_attach_type;
1536 	bool verifier_zext; /* Zero extensions has been inserted by verifier. */
1537 	bool dev_bound; /* Program is bound to the netdev. */
1538 	bool offload_requested; /* Program is bound and offloaded to the netdev. */
1539 	bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
1540 	bool attach_tracing_prog; /* true if tracing another tracing program */
1541 	bool func_proto_unreliable;
1542 	bool tail_call_reachable;
1543 	bool xdp_has_frags;
1544 	bool exception_cb;
1545 	bool exception_boundary;
1546 	bool is_extended; /* true if extended by freplace program */
1547 	bool jits_use_priv_stack;
1548 	bool priv_stack_requested;
1549 	bool changes_pkt_data;
1550 	bool might_sleep;
1551 	u64 prog_array_member_cnt; /* counts how many times as member of prog_array */
1552 	struct mutex ext_mutex; /* mutex for is_extended and prog_array_member_cnt */
1553 	struct bpf_arena *arena;
1554 	void (*recursion_detected)(struct bpf_prog *prog); /* callback if recursion is detected */
1555 	/* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
1556 	const struct btf_type *attach_func_proto;
1557 	/* function name for valid attach_btf_id */
1558 	const char *attach_func_name;
1559 	struct bpf_prog **func;
1560 	void *jit_data; /* JIT specific data. arch dependent */
1561 	struct bpf_jit_poke_descriptor *poke_tab;
1562 	struct bpf_kfunc_desc_tab *kfunc_tab;
1563 	struct bpf_kfunc_btf_tab *kfunc_btf_tab;
1564 	u32 size_poke_tab;
1565 #ifdef CONFIG_FINEIBT
1566 	struct bpf_ksym ksym_prefix;
1567 #endif
1568 	struct bpf_ksym ksym;
1569 	const struct bpf_prog_ops *ops;
1570 	struct bpf_map **used_maps;
1571 	struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
1572 	struct btf_mod_pair *used_btfs;
1573 	struct bpf_prog *prog;
1574 	struct user_struct *user;
1575 	u64 load_time; /* ns since boottime */
1576 	u32 verified_insns;
1577 	int cgroup_atype; /* enum cgroup_bpf_attach_type */
1578 	struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1579 	char name[BPF_OBJ_NAME_LEN];
1580 	u64 (*bpf_exception_cb)(u64 cookie, u64 sp, u64 bp, u64, u64);
1581 #ifdef CONFIG_SECURITY
1582 	void *security;
1583 #endif
1584 	struct bpf_token *token;
1585 	struct bpf_prog_offload *offload;
1586 	struct btf *btf;
1587 	struct bpf_func_info *func_info;
1588 	struct bpf_func_info_aux *func_info_aux;
1589 	/* bpf_line_info loaded from userspace.  linfo->insn_off
1590 	 * has the xlated insn offset.
1591 	 * Both the main and sub prog share the same linfo.
1592 	 * The subprog can access its first linfo by
1593 	 * using the linfo_idx.
1594 	 */
1595 	struct bpf_line_info *linfo;
1596 	/* jited_linfo is the jited addr of the linfo.  It has a
1597 	 * one to one mapping to linfo:
1598 	 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
1599 	 * Both the main and sub prog share the same jited_linfo.
1600 	 * The subprog can access its first jited_linfo by
1601 	 * using the linfo_idx.
1602 	 */
1603 	void **jited_linfo;
1604 	u32 func_info_cnt;
1605 	u32 nr_linfo;
1606 	/* subprog can use linfo_idx to access its first linfo and
1607 	 * jited_linfo.
1608 	 * main prog always has linfo_idx == 0
1609 	 */
1610 	u32 linfo_idx;
1611 	struct module *mod;
1612 	u32 num_exentries;
1613 	struct exception_table_entry *extable;
1614 	union {
1615 		struct work_struct work;
1616 		struct rcu_head	rcu;
1617 	};
1618 };
1619 
1620 struct bpf_prog {
1621 	u16			pages;		/* Number of allocated pages */
1622 	u16			jited:1,	/* Is our filter JIT'ed? */
1623 				jit_requested:1,/* archs need to JIT the prog */
1624 				gpl_compatible:1, /* Is filter GPL compatible? */
1625 				cb_access:1,	/* Is control block accessed? */
1626 				dst_needed:1,	/* Do we need dst entry? */
1627 				blinding_requested:1, /* needs constant blinding */
1628 				blinded:1,	/* Was blinded */
1629 				is_func:1,	/* program is a bpf function */
1630 				kprobe_override:1, /* Do we override a kprobe? */
1631 				has_callchain_buf:1, /* callchain buffer allocated? */
1632 				enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */
1633 				call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */
1634 				call_get_func_ip:1, /* Do we call get_func_ip() */
1635 				tstamp_type_access:1, /* Accessed __sk_buff->tstamp_type */
1636 				sleepable:1;	/* BPF program is sleepable */
1637 	enum bpf_prog_type	type;		/* Type of BPF program */
1638 	enum bpf_attach_type	expected_attach_type; /* For some prog types */
1639 	u32			len;		/* Number of filter blocks */
1640 	u32			jited_len;	/* Size of jited insns in bytes */
1641 	u8			tag[BPF_TAG_SIZE];
1642 	struct bpf_prog_stats __percpu *stats;
1643 	int __percpu		*active;
1644 	unsigned int		(*bpf_func)(const void *ctx,
1645 					    const struct bpf_insn *insn);
1646 	struct bpf_prog_aux	*aux;		/* Auxiliary fields */
1647 	struct sock_fprog_kern	*orig_prog;	/* Original BPF program */
1648 	/* Instructions for interpreter */
1649 	union {
1650 		DECLARE_FLEX_ARRAY(struct sock_filter, insns);
1651 		DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi);
1652 	};
1653 };
1654 
1655 struct bpf_array_aux {
1656 	/* Programs with direct jumps into programs part of this array. */
1657 	struct list_head poke_progs;
1658 	struct bpf_map *map;
1659 	struct mutex poke_mutex;
1660 	struct work_struct work;
1661 };
1662 
1663 struct bpf_link {
1664 	atomic64_t refcnt;
1665 	u32 id;
1666 	enum bpf_link_type type;
1667 	const struct bpf_link_ops *ops;
1668 	struct bpf_prog *prog;
1669 	/* whether BPF link itself has "sleepable" semantics, which can differ
1670 	 * from underlying BPF program having a "sleepable" semantics, as BPF
1671 	 * link's semantics is determined by target attach hook
1672 	 */
1673 	bool sleepable;
1674 	/* rcu is used before freeing, work can be used to schedule that
1675 	 * RCU-based freeing before that, so they never overlap
1676 	 */
1677 	union {
1678 		struct rcu_head rcu;
1679 		struct work_struct work;
1680 	};
1681 };
1682 
1683 struct bpf_link_ops {
1684 	void (*release)(struct bpf_link *link);
1685 	/* deallocate link resources callback, called without RCU grace period
1686 	 * waiting
1687 	 */
1688 	void (*dealloc)(struct bpf_link *link);
1689 	/* deallocate link resources callback, called after RCU grace period;
1690 	 * if either the underlying BPF program is sleepable or BPF link's
1691 	 * target hook is sleepable, we'll go through tasks trace RCU GP and
1692 	 * then "classic" RCU GP; this need for chaining tasks trace and
1693 	 * classic RCU GPs is designated by setting bpf_link->sleepable flag
1694 	 */
1695 	void (*dealloc_deferred)(struct bpf_link *link);
1696 	int (*detach)(struct bpf_link *link);
1697 	int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
1698 			   struct bpf_prog *old_prog);
1699 	void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
1700 	int (*fill_link_info)(const struct bpf_link *link,
1701 			      struct bpf_link_info *info);
1702 	int (*update_map)(struct bpf_link *link, struct bpf_map *new_map,
1703 			  struct bpf_map *old_map);
1704 	__poll_t (*poll)(struct file *file, struct poll_table_struct *pts);
1705 };
1706 
1707 struct bpf_tramp_link {
1708 	struct bpf_link link;
1709 	struct hlist_node tramp_hlist;
1710 	u64 cookie;
1711 };
1712 
1713 struct bpf_shim_tramp_link {
1714 	struct bpf_tramp_link link;
1715 	struct bpf_trampoline *trampoline;
1716 };
1717 
1718 struct bpf_tracing_link {
1719 	struct bpf_tramp_link link;
1720 	enum bpf_attach_type attach_type;
1721 	struct bpf_trampoline *trampoline;
1722 	struct bpf_prog *tgt_prog;
1723 };
1724 
1725 struct bpf_raw_tp_link {
1726 	struct bpf_link link;
1727 	struct bpf_raw_event_map *btp;
1728 	u64 cookie;
1729 };
1730 
1731 struct bpf_link_primer {
1732 	struct bpf_link *link;
1733 	struct file *file;
1734 	int fd;
1735 	u32 id;
1736 };
1737 
1738 struct bpf_mount_opts {
1739 	kuid_t uid;
1740 	kgid_t gid;
1741 	umode_t mode;
1742 
1743 	/* BPF token-related delegation options */
1744 	u64 delegate_cmds;
1745 	u64 delegate_maps;
1746 	u64 delegate_progs;
1747 	u64 delegate_attachs;
1748 };
1749 
1750 struct bpf_token {
1751 	struct work_struct work;
1752 	atomic64_t refcnt;
1753 	struct user_namespace *userns;
1754 	u64 allowed_cmds;
1755 	u64 allowed_maps;
1756 	u64 allowed_progs;
1757 	u64 allowed_attachs;
1758 #ifdef CONFIG_SECURITY
1759 	void *security;
1760 #endif
1761 };
1762 
1763 struct bpf_struct_ops_value;
1764 struct btf_member;
1765 
1766 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
1767 /**
1768  * struct bpf_struct_ops - A structure of callbacks allowing a subsystem to
1769  *			   define a BPF_MAP_TYPE_STRUCT_OPS map type composed
1770  *			   of BPF_PROG_TYPE_STRUCT_OPS progs.
1771  * @verifier_ops: A structure of callbacks that are invoked by the verifier
1772  *		  when determining whether the struct_ops progs in the
1773  *		  struct_ops map are valid.
1774  * @init: A callback that is invoked a single time, and before any other
1775  *	  callback, to initialize the structure. A nonzero return value means
1776  *	  the subsystem could not be initialized.
1777  * @check_member: When defined, a callback invoked by the verifier to allow
1778  *		  the subsystem to determine if an entry in the struct_ops map
1779  *		  is valid. A nonzero return value means that the map is
1780  *		  invalid and should be rejected by the verifier.
1781  * @init_member: A callback that is invoked for each member of the struct_ops
1782  *		 map to allow the subsystem to initialize the member. A nonzero
1783  *		 value means the member could not be initialized. This callback
1784  *		 is exclusive with the @type, @type_id, @value_type, and
1785  *		 @value_id fields.
1786  * @reg: A callback that is invoked when the struct_ops map has been
1787  *	 initialized and is being attached to. Zero means the struct_ops map
1788  *	 has been successfully registered and is live. A nonzero return value
1789  *	 means the struct_ops map could not be registered.
1790  * @unreg: A callback that is invoked when the struct_ops map should be
1791  *	   unregistered.
1792  * @update: A callback that is invoked when the live struct_ops map is being
1793  *	    updated to contain new values. This callback is only invoked when
1794  *	    the struct_ops map is loaded with BPF_F_LINK. If not defined, the
1795  *	    it is assumed that the struct_ops map cannot be updated.
1796  * @validate: A callback that is invoked after all of the members have been
1797  *	      initialized. This callback should perform static checks on the
1798  *	      map, meaning that it should either fail or succeed
1799  *	      deterministically. A struct_ops map that has been validated may
1800  *	      not necessarily succeed in being registered if the call to @reg
1801  *	      fails. For example, a valid struct_ops map may be loaded, but
1802  *	      then fail to be registered due to there being another active
1803  *	      struct_ops map on the system in the subsystem already. For this
1804  *	      reason, if this callback is not defined, the check is skipped as
1805  *	      the struct_ops map will have final verification performed in
1806  *	      @reg.
1807  * @type: BTF type.
1808  * @value_type: Value type.
1809  * @name: The name of the struct bpf_struct_ops object.
1810  * @func_models: Func models
1811  * @type_id: BTF type id.
1812  * @value_id: BTF value id.
1813  */
1814 struct bpf_struct_ops {
1815 	const struct bpf_verifier_ops *verifier_ops;
1816 	int (*init)(struct btf *btf);
1817 	int (*check_member)(const struct btf_type *t,
1818 			    const struct btf_member *member,
1819 			    const struct bpf_prog *prog);
1820 	int (*init_member)(const struct btf_type *t,
1821 			   const struct btf_member *member,
1822 			   void *kdata, const void *udata);
1823 	int (*reg)(void *kdata, struct bpf_link *link);
1824 	void (*unreg)(void *kdata, struct bpf_link *link);
1825 	int (*update)(void *kdata, void *old_kdata, struct bpf_link *link);
1826 	int (*validate)(void *kdata);
1827 	void *cfi_stubs;
1828 	struct module *owner;
1829 	const char *name;
1830 	struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
1831 };
1832 
1833 /* Every member of a struct_ops type has an instance even a member is not
1834  * an operator (function pointer). The "info" field will be assigned to
1835  * prog->aux->ctx_arg_info of BPF struct_ops programs to provide the
1836  * argument information required by the verifier to verify the program.
1837  *
1838  * btf_ctx_access() will lookup prog->aux->ctx_arg_info to find the
1839  * corresponding entry for an given argument.
1840  */
1841 struct bpf_struct_ops_arg_info {
1842 	struct bpf_ctx_arg_aux *info;
1843 	u32 cnt;
1844 };
1845 
1846 struct bpf_struct_ops_desc {
1847 	struct bpf_struct_ops *st_ops;
1848 
1849 	const struct btf_type *type;
1850 	const struct btf_type *value_type;
1851 	u32 type_id;
1852 	u32 value_id;
1853 
1854 	/* Collection of argument information for each member */
1855 	struct bpf_struct_ops_arg_info *arg_info;
1856 };
1857 
1858 enum bpf_struct_ops_state {
1859 	BPF_STRUCT_OPS_STATE_INIT,
1860 	BPF_STRUCT_OPS_STATE_INUSE,
1861 	BPF_STRUCT_OPS_STATE_TOBEFREE,
1862 	BPF_STRUCT_OPS_STATE_READY,
1863 };
1864 
1865 struct bpf_struct_ops_common_value {
1866 	refcount_t refcnt;
1867 	enum bpf_struct_ops_state state;
1868 };
1869 
1870 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
1871 /* This macro helps developer to register a struct_ops type and generate
1872  * type information correctly. Developers should use this macro to register
1873  * a struct_ops type instead of calling __register_bpf_struct_ops() directly.
1874  */
1875 #define register_bpf_struct_ops(st_ops, type)				\
1876 	({								\
1877 		struct bpf_struct_ops_##type {				\
1878 			struct bpf_struct_ops_common_value common;	\
1879 			struct type data ____cacheline_aligned_in_smp;	\
1880 		};							\
1881 		BTF_TYPE_EMIT(struct bpf_struct_ops_##type);		\
1882 		__register_bpf_struct_ops(st_ops);			\
1883 	})
1884 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
1885 bool bpf_struct_ops_get(const void *kdata);
1886 void bpf_struct_ops_put(const void *kdata);
1887 int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff);
1888 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
1889 				       void *value);
1890 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks,
1891 				      struct bpf_tramp_link *link,
1892 				      const struct btf_func_model *model,
1893 				      void *stub_func,
1894 				      void **image, u32 *image_off,
1895 				      bool allow_alloc);
1896 void bpf_struct_ops_image_free(void *image);
1897 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1898 {
1899 	if (owner == BPF_MODULE_OWNER)
1900 		return bpf_struct_ops_get(data);
1901 	else
1902 		return try_module_get(owner);
1903 }
1904 static inline void bpf_module_put(const void *data, struct module *owner)
1905 {
1906 	if (owner == BPF_MODULE_OWNER)
1907 		bpf_struct_ops_put(data);
1908 	else
1909 		module_put(owner);
1910 }
1911 int bpf_struct_ops_link_create(union bpf_attr *attr);
1912 
1913 #ifdef CONFIG_NET
1914 /* Define it here to avoid the use of forward declaration */
1915 struct bpf_dummy_ops_state {
1916 	int val;
1917 };
1918 
1919 struct bpf_dummy_ops {
1920 	int (*test_1)(struct bpf_dummy_ops_state *cb);
1921 	int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2,
1922 		      char a3, unsigned long a4);
1923 	int (*test_sleepable)(struct bpf_dummy_ops_state *cb);
1924 };
1925 
1926 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr,
1927 			    union bpf_attr __user *uattr);
1928 #endif
1929 int bpf_struct_ops_desc_init(struct bpf_struct_ops_desc *st_ops_desc,
1930 			     struct btf *btf,
1931 			     struct bpf_verifier_log *log);
1932 void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map);
1933 void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc);
1934 #else
1935 #define register_bpf_struct_ops(st_ops, type) ({ (void *)(st_ops); 0; })
1936 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1937 {
1938 	return try_module_get(owner);
1939 }
1940 static inline void bpf_module_put(const void *data, struct module *owner)
1941 {
1942 	module_put(owner);
1943 }
1944 static inline int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff)
1945 {
1946 	return -ENOTSUPP;
1947 }
1948 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
1949 						     void *key,
1950 						     void *value)
1951 {
1952 	return -EINVAL;
1953 }
1954 static inline int bpf_struct_ops_link_create(union bpf_attr *attr)
1955 {
1956 	return -EOPNOTSUPP;
1957 }
1958 static inline void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map)
1959 {
1960 }
1961 
1962 static inline void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc)
1963 {
1964 }
1965 
1966 #endif
1967 
1968 int bpf_prog_ctx_arg_info_init(struct bpf_prog *prog,
1969 			       const struct bpf_ctx_arg_aux *info, u32 cnt);
1970 
1971 #if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM)
1972 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1973 				    int cgroup_atype);
1974 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog);
1975 #else
1976 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1977 						  int cgroup_atype)
1978 {
1979 	return -EOPNOTSUPP;
1980 }
1981 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog)
1982 {
1983 }
1984 #endif
1985 
1986 struct bpf_array {
1987 	struct bpf_map map;
1988 	u32 elem_size;
1989 	u32 index_mask;
1990 	struct bpf_array_aux *aux;
1991 	union {
1992 		DECLARE_FLEX_ARRAY(char, value) __aligned(8);
1993 		DECLARE_FLEX_ARRAY(void *, ptrs) __aligned(8);
1994 		DECLARE_FLEX_ARRAY(void __percpu *, pptrs) __aligned(8);
1995 	};
1996 };
1997 
1998 #define BPF_COMPLEXITY_LIMIT_INSNS      1000000 /* yes. 1M insns */
1999 #define MAX_TAIL_CALL_CNT 33
2000 
2001 /* Maximum number of loops for bpf_loop and bpf_iter_num.
2002  * It's enum to expose it (and thus make it discoverable) through BTF.
2003  */
2004 enum {
2005 	BPF_MAX_LOOPS = 8 * 1024 * 1024,
2006 	BPF_MAX_TIMED_LOOPS = 0xffff,
2007 };
2008 
2009 #define BPF_F_ACCESS_MASK	(BPF_F_RDONLY |		\
2010 				 BPF_F_RDONLY_PROG |	\
2011 				 BPF_F_WRONLY |		\
2012 				 BPF_F_WRONLY_PROG)
2013 
2014 #define BPF_MAP_CAN_READ	BIT(0)
2015 #define BPF_MAP_CAN_WRITE	BIT(1)
2016 
2017 /* Maximum number of user-producer ring buffer samples that can be drained in
2018  * a call to bpf_user_ringbuf_drain().
2019  */
2020 #define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024)
2021 
2022 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
2023 {
2024 	u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
2025 
2026 	/* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
2027 	 * not possible.
2028 	 */
2029 	if (access_flags & BPF_F_RDONLY_PROG)
2030 		return BPF_MAP_CAN_READ;
2031 	else if (access_flags & BPF_F_WRONLY_PROG)
2032 		return BPF_MAP_CAN_WRITE;
2033 	else
2034 		return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
2035 }
2036 
2037 static inline bool bpf_map_flags_access_ok(u32 access_flags)
2038 {
2039 	return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
2040 	       (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
2041 }
2042 
2043 struct bpf_event_entry {
2044 	struct perf_event *event;
2045 	struct file *perf_file;
2046 	struct file *map_file;
2047 	struct rcu_head rcu;
2048 };
2049 
2050 static inline bool map_type_contains_progs(struct bpf_map *map)
2051 {
2052 	return map->map_type == BPF_MAP_TYPE_PROG_ARRAY ||
2053 	       map->map_type == BPF_MAP_TYPE_DEVMAP ||
2054 	       map->map_type == BPF_MAP_TYPE_CPUMAP;
2055 }
2056 
2057 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp);
2058 int bpf_prog_calc_tag(struct bpf_prog *fp);
2059 
2060 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
2061 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void);
2062 
2063 const struct bpf_func_proto *bpf_get_perf_event_read_value_proto(void);
2064 
2065 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
2066 					unsigned long off, unsigned long len);
2067 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
2068 					const struct bpf_insn *src,
2069 					struct bpf_insn *dst,
2070 					struct bpf_prog *prog,
2071 					u32 *target_size);
2072 
2073 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
2074 		     void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
2075 
2076 /* an array of programs to be executed under rcu_lock.
2077  *
2078  * Typical usage:
2079  * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run);
2080  *
2081  * the structure returned by bpf_prog_array_alloc() should be populated
2082  * with program pointers and the last pointer must be NULL.
2083  * The user has to keep refcnt on the program and make sure the program
2084  * is removed from the array before bpf_prog_put().
2085  * The 'struct bpf_prog_array *' should only be replaced with xchg()
2086  * since other cpus are walking the array of pointers in parallel.
2087  */
2088 struct bpf_prog_array_item {
2089 	struct bpf_prog *prog;
2090 	union {
2091 		struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
2092 		u64 bpf_cookie;
2093 	};
2094 };
2095 
2096 struct bpf_prog_array {
2097 	struct rcu_head rcu;
2098 	struct bpf_prog_array_item items[];
2099 };
2100 
2101 struct bpf_empty_prog_array {
2102 	struct bpf_prog_array hdr;
2103 	struct bpf_prog *null_prog;
2104 };
2105 
2106 /* to avoid allocating empty bpf_prog_array for cgroups that
2107  * don't have bpf program attached use one global 'bpf_empty_prog_array'
2108  * It will not be modified the caller of bpf_prog_array_alloc()
2109  * (since caller requested prog_cnt == 0)
2110  * that pointer should be 'freed' by bpf_prog_array_free()
2111  */
2112 extern struct bpf_empty_prog_array bpf_empty_prog_array;
2113 
2114 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
2115 void bpf_prog_array_free(struct bpf_prog_array *progs);
2116 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */
2117 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs);
2118 int bpf_prog_array_length(struct bpf_prog_array *progs);
2119 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
2120 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
2121 				__u32 __user *prog_ids, u32 cnt);
2122 
2123 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
2124 				struct bpf_prog *old_prog);
2125 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
2126 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
2127 			     struct bpf_prog *prog);
2128 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
2129 			     u32 *prog_ids, u32 request_cnt,
2130 			     u32 *prog_cnt);
2131 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
2132 			struct bpf_prog *exclude_prog,
2133 			struct bpf_prog *include_prog,
2134 			u64 bpf_cookie,
2135 			struct bpf_prog_array **new_array);
2136 
2137 struct bpf_run_ctx {};
2138 
2139 struct bpf_cg_run_ctx {
2140 	struct bpf_run_ctx run_ctx;
2141 	const struct bpf_prog_array_item *prog_item;
2142 	int retval;
2143 };
2144 
2145 struct bpf_trace_run_ctx {
2146 	struct bpf_run_ctx run_ctx;
2147 	u64 bpf_cookie;
2148 	bool is_uprobe;
2149 };
2150 
2151 struct bpf_tramp_run_ctx {
2152 	struct bpf_run_ctx run_ctx;
2153 	u64 bpf_cookie;
2154 	struct bpf_run_ctx *saved_run_ctx;
2155 };
2156 
2157 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)
2158 {
2159 	struct bpf_run_ctx *old_ctx = NULL;
2160 
2161 #ifdef CONFIG_BPF_SYSCALL
2162 	old_ctx = current->bpf_ctx;
2163 	current->bpf_ctx = new_ctx;
2164 #endif
2165 	return old_ctx;
2166 }
2167 
2168 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
2169 {
2170 #ifdef CONFIG_BPF_SYSCALL
2171 	current->bpf_ctx = old_ctx;
2172 #endif
2173 }
2174 
2175 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
2176 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE			(1 << 0)
2177 /* BPF program asks to set CN on the packet. */
2178 #define BPF_RET_SET_CN						(1 << 0)
2179 
2180 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx);
2181 
2182 static __always_inline u32
2183 bpf_prog_run_array(const struct bpf_prog_array *array,
2184 		   const void *ctx, bpf_prog_run_fn run_prog)
2185 {
2186 	const struct bpf_prog_array_item *item;
2187 	const struct bpf_prog *prog;
2188 	struct bpf_run_ctx *old_run_ctx;
2189 	struct bpf_trace_run_ctx run_ctx;
2190 	u32 ret = 1;
2191 
2192 	RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held");
2193 
2194 	if (unlikely(!array))
2195 		return ret;
2196 
2197 	run_ctx.is_uprobe = false;
2198 
2199 	migrate_disable();
2200 	old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
2201 	item = &array->items[0];
2202 	while ((prog = READ_ONCE(item->prog))) {
2203 		run_ctx.bpf_cookie = item->bpf_cookie;
2204 		ret &= run_prog(prog, ctx);
2205 		item++;
2206 	}
2207 	bpf_reset_run_ctx(old_run_ctx);
2208 	migrate_enable();
2209 	return ret;
2210 }
2211 
2212 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs:
2213  *
2214  * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array
2215  * overall. As a result, we must use the bpf_prog_array_free_sleepable
2216  * in order to use the tasks_trace rcu grace period.
2217  *
2218  * When a non-sleepable program is inside the array, we take the rcu read
2219  * section and disable preemption for that program alone, so it can access
2220  * rcu-protected dynamically sized maps.
2221  */
2222 static __always_inline u32
2223 bpf_prog_run_array_uprobe(const struct bpf_prog_array *array,
2224 			  const void *ctx, bpf_prog_run_fn run_prog)
2225 {
2226 	const struct bpf_prog_array_item *item;
2227 	const struct bpf_prog *prog;
2228 	struct bpf_run_ctx *old_run_ctx;
2229 	struct bpf_trace_run_ctx run_ctx;
2230 	u32 ret = 1;
2231 
2232 	might_fault();
2233 	RCU_LOCKDEP_WARN(!rcu_read_lock_trace_held(), "no rcu lock held");
2234 
2235 	if (unlikely(!array))
2236 		return ret;
2237 
2238 	migrate_disable();
2239 
2240 	run_ctx.is_uprobe = true;
2241 
2242 	old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
2243 	item = &array->items[0];
2244 	while ((prog = READ_ONCE(item->prog))) {
2245 		if (!prog->sleepable)
2246 			rcu_read_lock();
2247 
2248 		run_ctx.bpf_cookie = item->bpf_cookie;
2249 		ret &= run_prog(prog, ctx);
2250 		item++;
2251 
2252 		if (!prog->sleepable)
2253 			rcu_read_unlock();
2254 	}
2255 	bpf_reset_run_ctx(old_run_ctx);
2256 	migrate_enable();
2257 	return ret;
2258 }
2259 
2260 #ifdef CONFIG_BPF_SYSCALL
2261 DECLARE_PER_CPU(int, bpf_prog_active);
2262 extern struct mutex bpf_stats_enabled_mutex;
2263 
2264 /*
2265  * Block execution of BPF programs attached to instrumentation (perf,
2266  * kprobes, tracepoints) to prevent deadlocks on map operations as any of
2267  * these events can happen inside a region which holds a map bucket lock
2268  * and can deadlock on it.
2269  */
2270 static inline void bpf_disable_instrumentation(void)
2271 {
2272 	migrate_disable();
2273 	this_cpu_inc(bpf_prog_active);
2274 }
2275 
2276 static inline void bpf_enable_instrumentation(void)
2277 {
2278 	this_cpu_dec(bpf_prog_active);
2279 	migrate_enable();
2280 }
2281 
2282 extern const struct super_operations bpf_super_ops;
2283 extern const struct file_operations bpf_map_fops;
2284 extern const struct file_operations bpf_prog_fops;
2285 extern const struct file_operations bpf_iter_fops;
2286 
2287 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
2288 	extern const struct bpf_prog_ops _name ## _prog_ops; \
2289 	extern const struct bpf_verifier_ops _name ## _verifier_ops;
2290 #define BPF_MAP_TYPE(_id, _ops) \
2291 	extern const struct bpf_map_ops _ops;
2292 #define BPF_LINK_TYPE(_id, _name)
2293 #include <linux/bpf_types.h>
2294 #undef BPF_PROG_TYPE
2295 #undef BPF_MAP_TYPE
2296 #undef BPF_LINK_TYPE
2297 
2298 extern const struct bpf_prog_ops bpf_offload_prog_ops;
2299 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
2300 extern const struct bpf_verifier_ops xdp_analyzer_ops;
2301 
2302 struct bpf_prog *bpf_prog_get(u32 ufd);
2303 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
2304 				       bool attach_drv);
2305 void bpf_prog_add(struct bpf_prog *prog, int i);
2306 void bpf_prog_sub(struct bpf_prog *prog, int i);
2307 void bpf_prog_inc(struct bpf_prog *prog);
2308 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
2309 void bpf_prog_put(struct bpf_prog *prog);
2310 
2311 void bpf_prog_free_id(struct bpf_prog *prog);
2312 void bpf_map_free_id(struct bpf_map *map);
2313 
2314 struct btf_field *btf_record_find(const struct btf_record *rec,
2315 				  u32 offset, u32 field_mask);
2316 void btf_record_free(struct btf_record *rec);
2317 void bpf_map_free_record(struct bpf_map *map);
2318 struct btf_record *btf_record_dup(const struct btf_record *rec);
2319 bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b);
2320 void bpf_obj_free_timer(const struct btf_record *rec, void *obj);
2321 void bpf_obj_free_workqueue(const struct btf_record *rec, void *obj);
2322 void bpf_obj_free_fields(const struct btf_record *rec, void *obj);
2323 void __bpf_obj_drop_impl(void *p, const struct btf_record *rec, bool percpu);
2324 
2325 struct bpf_map *bpf_map_get(u32 ufd);
2326 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
2327 
2328 /*
2329  * The __bpf_map_get() and __btf_get_by_fd() functions parse a file
2330  * descriptor and return a corresponding map or btf object.
2331  * Their names are double underscored to emphasize the fact that they
2332  * do not increase refcnt. To also increase refcnt use corresponding
2333  * bpf_map_get() and btf_get_by_fd() functions.
2334  */
2335 
2336 static inline struct bpf_map *__bpf_map_get(struct fd f)
2337 {
2338 	if (fd_empty(f))
2339 		return ERR_PTR(-EBADF);
2340 	if (unlikely(fd_file(f)->f_op != &bpf_map_fops))
2341 		return ERR_PTR(-EINVAL);
2342 	return fd_file(f)->private_data;
2343 }
2344 
2345 static inline struct btf *__btf_get_by_fd(struct fd f)
2346 {
2347 	if (fd_empty(f))
2348 		return ERR_PTR(-EBADF);
2349 	if (unlikely(fd_file(f)->f_op != &btf_fops))
2350 		return ERR_PTR(-EINVAL);
2351 	return fd_file(f)->private_data;
2352 }
2353 
2354 void bpf_map_inc(struct bpf_map *map);
2355 void bpf_map_inc_with_uref(struct bpf_map *map);
2356 struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref);
2357 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
2358 void bpf_map_put_with_uref(struct bpf_map *map);
2359 void bpf_map_put(struct bpf_map *map);
2360 void *bpf_map_area_alloc(u64 size, int numa_node);
2361 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
2362 void bpf_map_area_free(void *base);
2363 bool bpf_map_write_active(const struct bpf_map *map);
2364 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
2365 int  generic_map_lookup_batch(struct bpf_map *map,
2366 			      const union bpf_attr *attr,
2367 			      union bpf_attr __user *uattr);
2368 int  generic_map_update_batch(struct bpf_map *map, struct file *map_file,
2369 			      const union bpf_attr *attr,
2370 			      union bpf_attr __user *uattr);
2371 int  generic_map_delete_batch(struct bpf_map *map,
2372 			      const union bpf_attr *attr,
2373 			      union bpf_attr __user *uattr);
2374 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
2375 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
2376 
2377 int bpf_map_alloc_pages(const struct bpf_map *map, gfp_t gfp, int nid,
2378 			unsigned long nr_pages, struct page **page_array);
2379 #ifdef CONFIG_MEMCG
2380 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
2381 			   int node);
2382 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags);
2383 void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size,
2384 		       gfp_t flags);
2385 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
2386 				    size_t align, gfp_t flags);
2387 #else
2388 /*
2389  * These specialized allocators have to be macros for their allocations to be
2390  * accounted separately (to have separate alloc_tag).
2391  */
2392 #define bpf_map_kmalloc_node(_map, _size, _flags, _node)	\
2393 		kmalloc_node(_size, _flags, _node)
2394 #define bpf_map_kzalloc(_map, _size, _flags)			\
2395 		kzalloc(_size, _flags)
2396 #define bpf_map_kvcalloc(_map, _n, _size, _flags)		\
2397 		kvcalloc(_n, _size, _flags)
2398 #define bpf_map_alloc_percpu(_map, _size, _align, _flags)	\
2399 		__alloc_percpu_gfp(_size, _align, _flags)
2400 #endif
2401 
2402 static inline int
2403 bpf_map_init_elem_count(struct bpf_map *map)
2404 {
2405 	size_t size = sizeof(*map->elem_count), align = size;
2406 	gfp_t flags = GFP_USER | __GFP_NOWARN;
2407 
2408 	map->elem_count = bpf_map_alloc_percpu(map, size, align, flags);
2409 	if (!map->elem_count)
2410 		return -ENOMEM;
2411 
2412 	return 0;
2413 }
2414 
2415 static inline void
2416 bpf_map_free_elem_count(struct bpf_map *map)
2417 {
2418 	free_percpu(map->elem_count);
2419 }
2420 
2421 static inline void bpf_map_inc_elem_count(struct bpf_map *map)
2422 {
2423 	this_cpu_inc(*map->elem_count);
2424 }
2425 
2426 static inline void bpf_map_dec_elem_count(struct bpf_map *map)
2427 {
2428 	this_cpu_dec(*map->elem_count);
2429 }
2430 
2431 extern int sysctl_unprivileged_bpf_disabled;
2432 
2433 bool bpf_token_capable(const struct bpf_token *token, int cap);
2434 
2435 static inline bool bpf_allow_ptr_leaks(const struct bpf_token *token)
2436 {
2437 	return bpf_token_capable(token, CAP_PERFMON);
2438 }
2439 
2440 static inline bool bpf_allow_uninit_stack(const struct bpf_token *token)
2441 {
2442 	return bpf_token_capable(token, CAP_PERFMON);
2443 }
2444 
2445 static inline bool bpf_bypass_spec_v1(const struct bpf_token *token)
2446 {
2447 	return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON);
2448 }
2449 
2450 static inline bool bpf_bypass_spec_v4(const struct bpf_token *token)
2451 {
2452 	return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON);
2453 }
2454 
2455 int bpf_map_new_fd(struct bpf_map *map, int flags);
2456 int bpf_prog_new_fd(struct bpf_prog *prog);
2457 
2458 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2459 		   const struct bpf_link_ops *ops, struct bpf_prog *prog);
2460 void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type,
2461 			     const struct bpf_link_ops *ops, struct bpf_prog *prog,
2462 			     bool sleepable);
2463 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
2464 int bpf_link_settle(struct bpf_link_primer *primer);
2465 void bpf_link_cleanup(struct bpf_link_primer *primer);
2466 void bpf_link_inc(struct bpf_link *link);
2467 struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link);
2468 void bpf_link_put(struct bpf_link *link);
2469 int bpf_link_new_fd(struct bpf_link *link);
2470 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
2471 struct bpf_link *bpf_link_get_curr_or_next(u32 *id);
2472 
2473 void bpf_token_inc(struct bpf_token *token);
2474 void bpf_token_put(struct bpf_token *token);
2475 int bpf_token_create(union bpf_attr *attr);
2476 struct bpf_token *bpf_token_get_from_fd(u32 ufd);
2477 
2478 bool bpf_token_allow_cmd(const struct bpf_token *token, enum bpf_cmd cmd);
2479 bool bpf_token_allow_map_type(const struct bpf_token *token, enum bpf_map_type type);
2480 bool bpf_token_allow_prog_type(const struct bpf_token *token,
2481 			       enum bpf_prog_type prog_type,
2482 			       enum bpf_attach_type attach_type);
2483 
2484 int bpf_obj_pin_user(u32 ufd, int path_fd, const char __user *pathname);
2485 int bpf_obj_get_user(int path_fd, const char __user *pathname, int flags);
2486 struct inode *bpf_get_inode(struct super_block *sb, const struct inode *dir,
2487 			    umode_t mode);
2488 
2489 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
2490 #define DEFINE_BPF_ITER_FUNC(target, args...)			\
2491 	extern int bpf_iter_ ## target(args);			\
2492 	int __init bpf_iter_ ## target(args) { return 0; }
2493 
2494 /*
2495  * The task type of iterators.
2496  *
2497  * For BPF task iterators, they can be parameterized with various
2498  * parameters to visit only some of tasks.
2499  *
2500  * BPF_TASK_ITER_ALL (default)
2501  *	Iterate over resources of every task.
2502  *
2503  * BPF_TASK_ITER_TID
2504  *	Iterate over resources of a task/tid.
2505  *
2506  * BPF_TASK_ITER_TGID
2507  *	Iterate over resources of every task of a process / task group.
2508  */
2509 enum bpf_iter_task_type {
2510 	BPF_TASK_ITER_ALL = 0,
2511 	BPF_TASK_ITER_TID,
2512 	BPF_TASK_ITER_TGID,
2513 };
2514 
2515 struct bpf_iter_aux_info {
2516 	/* for map_elem iter */
2517 	struct bpf_map *map;
2518 
2519 	/* for cgroup iter */
2520 	struct {
2521 		struct cgroup *start; /* starting cgroup */
2522 		enum bpf_cgroup_iter_order order;
2523 	} cgroup;
2524 	struct {
2525 		enum bpf_iter_task_type	type;
2526 		u32 pid;
2527 	} task;
2528 };
2529 
2530 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
2531 					union bpf_iter_link_info *linfo,
2532 					struct bpf_iter_aux_info *aux);
2533 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
2534 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
2535 					struct seq_file *seq);
2536 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
2537 					 struct bpf_link_info *info);
2538 typedef const struct bpf_func_proto *
2539 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id,
2540 			     const struct bpf_prog *prog);
2541 
2542 enum bpf_iter_feature {
2543 	BPF_ITER_RESCHED	= BIT(0),
2544 };
2545 
2546 #define BPF_ITER_CTX_ARG_MAX 2
2547 struct bpf_iter_reg {
2548 	const char *target;
2549 	bpf_iter_attach_target_t attach_target;
2550 	bpf_iter_detach_target_t detach_target;
2551 	bpf_iter_show_fdinfo_t show_fdinfo;
2552 	bpf_iter_fill_link_info_t fill_link_info;
2553 	bpf_iter_get_func_proto_t get_func_proto;
2554 	u32 ctx_arg_info_size;
2555 	u32 feature;
2556 	struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
2557 	const struct bpf_iter_seq_info *seq_info;
2558 };
2559 
2560 struct bpf_iter_meta {
2561 	__bpf_md_ptr(struct seq_file *, seq);
2562 	u64 session_id;
2563 	u64 seq_num;
2564 };
2565 
2566 struct bpf_iter__bpf_map_elem {
2567 	__bpf_md_ptr(struct bpf_iter_meta *, meta);
2568 	__bpf_md_ptr(struct bpf_map *, map);
2569 	__bpf_md_ptr(void *, key);
2570 	__bpf_md_ptr(void *, value);
2571 };
2572 
2573 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
2574 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
2575 int bpf_iter_prog_supported(struct bpf_prog *prog);
2576 const struct bpf_func_proto *
2577 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
2578 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog);
2579 int bpf_iter_new_fd(struct bpf_link *link);
2580 bool bpf_link_is_iter(struct bpf_link *link);
2581 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
2582 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
2583 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
2584 			      struct seq_file *seq);
2585 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
2586 				struct bpf_link_info *info);
2587 
2588 int map_set_for_each_callback_args(struct bpf_verifier_env *env,
2589 				   struct bpf_func_state *caller,
2590 				   struct bpf_func_state *callee);
2591 
2592 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
2593 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
2594 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
2595 			   u64 flags);
2596 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
2597 			    u64 flags);
2598 
2599 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
2600 
2601 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
2602 				 void *key, void *value, u64 map_flags);
2603 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2604 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
2605 				void *key, void *value, u64 map_flags);
2606 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2607 
2608 int bpf_get_file_flag(int flags);
2609 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size,
2610 			     size_t actual_size);
2611 
2612 /* verify correctness of eBPF program */
2613 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size);
2614 
2615 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
2616 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
2617 #endif
2618 
2619 struct btf *bpf_get_btf_vmlinux(void);
2620 
2621 /* Map specifics */
2622 struct xdp_frame;
2623 struct sk_buff;
2624 struct bpf_dtab_netdev;
2625 struct bpf_cpu_map_entry;
2626 
2627 void __dev_flush(struct list_head *flush_list);
2628 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2629 		    struct net_device *dev_rx);
2630 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2631 		    struct net_device *dev_rx);
2632 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2633 			  struct bpf_map *map, bool exclude_ingress);
2634 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
2635 			     const struct bpf_prog *xdp_prog);
2636 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2637 			   const struct bpf_prog *xdp_prog,
2638 			   struct bpf_map *map, bool exclude_ingress);
2639 
2640 void __cpu_map_flush(struct list_head *flush_list);
2641 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf,
2642 		    struct net_device *dev_rx);
2643 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2644 			     struct sk_buff *skb);
2645 
2646 /* Return map's numa specified by userspace */
2647 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
2648 {
2649 	return (attr->map_flags & BPF_F_NUMA_NODE) ?
2650 		attr->numa_node : NUMA_NO_NODE;
2651 }
2652 
2653 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
2654 int array_map_alloc_check(union bpf_attr *attr);
2655 
2656 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
2657 			  union bpf_attr __user *uattr);
2658 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
2659 			  union bpf_attr __user *uattr);
2660 int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2661 			      const union bpf_attr *kattr,
2662 			      union bpf_attr __user *uattr);
2663 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2664 				     const union bpf_attr *kattr,
2665 				     union bpf_attr __user *uattr);
2666 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
2667 			     const union bpf_attr *kattr,
2668 			     union bpf_attr __user *uattr);
2669 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2670 				const union bpf_attr *kattr,
2671 				union bpf_attr __user *uattr);
2672 int bpf_prog_test_run_nf(struct bpf_prog *prog,
2673 			 const union bpf_attr *kattr,
2674 			 union bpf_attr __user *uattr);
2675 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
2676 		    const struct bpf_prog *prog,
2677 		    struct bpf_insn_access_aux *info);
2678 
2679 static inline bool bpf_tracing_ctx_access(int off, int size,
2680 					  enum bpf_access_type type)
2681 {
2682 	if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
2683 		return false;
2684 	if (type != BPF_READ)
2685 		return false;
2686 	if (off % size != 0)
2687 		return false;
2688 	return true;
2689 }
2690 
2691 static inline bool bpf_tracing_btf_ctx_access(int off, int size,
2692 					      enum bpf_access_type type,
2693 					      const struct bpf_prog *prog,
2694 					      struct bpf_insn_access_aux *info)
2695 {
2696 	if (!bpf_tracing_ctx_access(off, size, type))
2697 		return false;
2698 	return btf_ctx_access(off, size, type, prog, info);
2699 }
2700 
2701 int btf_struct_access(struct bpf_verifier_log *log,
2702 		      const struct bpf_reg_state *reg,
2703 		      int off, int size, enum bpf_access_type atype,
2704 		      u32 *next_btf_id, enum bpf_type_flag *flag, const char **field_name);
2705 bool btf_struct_ids_match(struct bpf_verifier_log *log,
2706 			  const struct btf *btf, u32 id, int off,
2707 			  const struct btf *need_btf, u32 need_type_id,
2708 			  bool strict);
2709 
2710 int btf_distill_func_proto(struct bpf_verifier_log *log,
2711 			   struct btf *btf,
2712 			   const struct btf_type *func_proto,
2713 			   const char *func_name,
2714 			   struct btf_func_model *m);
2715 
2716 struct bpf_reg_state;
2717 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog);
2718 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
2719 			 struct btf *btf, const struct btf_type *t);
2720 const char *btf_find_decl_tag_value(const struct btf *btf, const struct btf_type *pt,
2721 				    int comp_idx, const char *tag_key);
2722 int btf_find_next_decl_tag(const struct btf *btf, const struct btf_type *pt,
2723 			   int comp_idx, const char *tag_key, int last_id);
2724 
2725 struct bpf_prog *bpf_prog_by_id(u32 id);
2726 struct bpf_link *bpf_link_by_id(u32 id);
2727 
2728 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id,
2729 						 const struct bpf_prog *prog);
2730 void bpf_task_storage_free(struct task_struct *task);
2731 void bpf_cgrp_storage_free(struct cgroup *cgroup);
2732 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
2733 const struct btf_func_model *
2734 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2735 			 const struct bpf_insn *insn);
2736 int bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
2737 		       u16 btf_fd_idx, u8 **func_addr);
2738 
2739 struct bpf_core_ctx {
2740 	struct bpf_verifier_log *log;
2741 	const struct btf *btf;
2742 };
2743 
2744 bool btf_nested_type_is_trusted(struct bpf_verifier_log *log,
2745 				const struct bpf_reg_state *reg,
2746 				const char *field_name, u32 btf_id, const char *suffix);
2747 
2748 bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log,
2749 			       const struct btf *reg_btf, u32 reg_id,
2750 			       const struct btf *arg_btf, u32 arg_id);
2751 
2752 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo,
2753 		   int relo_idx, void *insn);
2754 
2755 static inline bool unprivileged_ebpf_enabled(void)
2756 {
2757 	return !sysctl_unprivileged_bpf_disabled;
2758 }
2759 
2760 /* Not all bpf prog type has the bpf_ctx.
2761  * For the bpf prog type that has initialized the bpf_ctx,
2762  * this function can be used to decide if a kernel function
2763  * is called by a bpf program.
2764  */
2765 static inline bool has_current_bpf_ctx(void)
2766 {
2767 	return !!current->bpf_ctx;
2768 }
2769 
2770 void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog);
2771 
2772 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2773 		     enum bpf_dynptr_type type, u32 offset, u32 size);
2774 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr);
2775 void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr);
2776 
2777 #else /* !CONFIG_BPF_SYSCALL */
2778 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
2779 {
2780 	return ERR_PTR(-EOPNOTSUPP);
2781 }
2782 
2783 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
2784 						     enum bpf_prog_type type,
2785 						     bool attach_drv)
2786 {
2787 	return ERR_PTR(-EOPNOTSUPP);
2788 }
2789 
2790 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
2791 {
2792 }
2793 
2794 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
2795 {
2796 }
2797 
2798 static inline void bpf_prog_put(struct bpf_prog *prog)
2799 {
2800 }
2801 
2802 static inline void bpf_prog_inc(struct bpf_prog *prog)
2803 {
2804 }
2805 
2806 static inline struct bpf_prog *__must_check
2807 bpf_prog_inc_not_zero(struct bpf_prog *prog)
2808 {
2809 	return ERR_PTR(-EOPNOTSUPP);
2810 }
2811 
2812 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2813 				 const struct bpf_link_ops *ops,
2814 				 struct bpf_prog *prog)
2815 {
2816 }
2817 
2818 static inline void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type,
2819 					   const struct bpf_link_ops *ops, struct bpf_prog *prog,
2820 					   bool sleepable)
2821 {
2822 }
2823 
2824 static inline int bpf_link_prime(struct bpf_link *link,
2825 				 struct bpf_link_primer *primer)
2826 {
2827 	return -EOPNOTSUPP;
2828 }
2829 
2830 static inline int bpf_link_settle(struct bpf_link_primer *primer)
2831 {
2832 	return -EOPNOTSUPP;
2833 }
2834 
2835 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
2836 {
2837 }
2838 
2839 static inline void bpf_link_inc(struct bpf_link *link)
2840 {
2841 }
2842 
2843 static inline struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link)
2844 {
2845 	return NULL;
2846 }
2847 
2848 static inline void bpf_link_put(struct bpf_link *link)
2849 {
2850 }
2851 
2852 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
2853 {
2854 	return -EOPNOTSUPP;
2855 }
2856 
2857 static inline bool bpf_token_capable(const struct bpf_token *token, int cap)
2858 {
2859 	return capable(cap) || (cap != CAP_SYS_ADMIN && capable(CAP_SYS_ADMIN));
2860 }
2861 
2862 static inline void bpf_token_inc(struct bpf_token *token)
2863 {
2864 }
2865 
2866 static inline void bpf_token_put(struct bpf_token *token)
2867 {
2868 }
2869 
2870 static inline struct bpf_token *bpf_token_get_from_fd(u32 ufd)
2871 {
2872 	return ERR_PTR(-EOPNOTSUPP);
2873 }
2874 
2875 static inline void __dev_flush(struct list_head *flush_list)
2876 {
2877 }
2878 
2879 struct xdp_frame;
2880 struct bpf_dtab_netdev;
2881 struct bpf_cpu_map_entry;
2882 
2883 static inline
2884 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2885 		    struct net_device *dev_rx)
2886 {
2887 	return 0;
2888 }
2889 
2890 static inline
2891 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2892 		    struct net_device *dev_rx)
2893 {
2894 	return 0;
2895 }
2896 
2897 static inline
2898 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2899 			  struct bpf_map *map, bool exclude_ingress)
2900 {
2901 	return 0;
2902 }
2903 
2904 struct sk_buff;
2905 
2906 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
2907 					   struct sk_buff *skb,
2908 					   const struct bpf_prog *xdp_prog)
2909 {
2910 	return 0;
2911 }
2912 
2913 static inline
2914 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2915 			   const struct bpf_prog *xdp_prog,
2916 			   struct bpf_map *map, bool exclude_ingress)
2917 {
2918 	return 0;
2919 }
2920 
2921 static inline void __cpu_map_flush(struct list_head *flush_list)
2922 {
2923 }
2924 
2925 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
2926 				  struct xdp_frame *xdpf,
2927 				  struct net_device *dev_rx)
2928 {
2929 	return 0;
2930 }
2931 
2932 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2933 					   struct sk_buff *skb)
2934 {
2935 	return -EOPNOTSUPP;
2936 }
2937 
2938 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
2939 				enum bpf_prog_type type)
2940 {
2941 	return ERR_PTR(-EOPNOTSUPP);
2942 }
2943 
2944 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
2945 					const union bpf_attr *kattr,
2946 					union bpf_attr __user *uattr)
2947 {
2948 	return -ENOTSUPP;
2949 }
2950 
2951 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
2952 					const union bpf_attr *kattr,
2953 					union bpf_attr __user *uattr)
2954 {
2955 	return -ENOTSUPP;
2956 }
2957 
2958 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2959 					    const union bpf_attr *kattr,
2960 					    union bpf_attr __user *uattr)
2961 {
2962 	return -ENOTSUPP;
2963 }
2964 
2965 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2966 						   const union bpf_attr *kattr,
2967 						   union bpf_attr __user *uattr)
2968 {
2969 	return -ENOTSUPP;
2970 }
2971 
2972 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2973 					      const union bpf_attr *kattr,
2974 					      union bpf_attr __user *uattr)
2975 {
2976 	return -ENOTSUPP;
2977 }
2978 
2979 static inline void bpf_map_put(struct bpf_map *map)
2980 {
2981 }
2982 
2983 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
2984 {
2985 	return ERR_PTR(-ENOTSUPP);
2986 }
2987 
2988 static inline int btf_struct_access(struct bpf_verifier_log *log,
2989 				    const struct bpf_reg_state *reg,
2990 				    int off, int size, enum bpf_access_type atype,
2991 				    u32 *next_btf_id, enum bpf_type_flag *flag,
2992 				    const char **field_name)
2993 {
2994 	return -EACCES;
2995 }
2996 
2997 static inline const struct bpf_func_proto *
2998 bpf_base_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
2999 {
3000 	return NULL;
3001 }
3002 
3003 static inline void bpf_task_storage_free(struct task_struct *task)
3004 {
3005 }
3006 
3007 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog)
3008 {
3009 	return false;
3010 }
3011 
3012 static inline const struct btf_func_model *
3013 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
3014 			 const struct bpf_insn *insn)
3015 {
3016 	return NULL;
3017 }
3018 
3019 static inline int
3020 bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
3021 		   u16 btf_fd_idx, u8 **func_addr)
3022 {
3023 	return -ENOTSUPP;
3024 }
3025 
3026 static inline bool unprivileged_ebpf_enabled(void)
3027 {
3028 	return false;
3029 }
3030 
3031 static inline bool has_current_bpf_ctx(void)
3032 {
3033 	return false;
3034 }
3035 
3036 static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog)
3037 {
3038 }
3039 
3040 static inline void bpf_cgrp_storage_free(struct cgroup *cgroup)
3041 {
3042 }
3043 
3044 static inline void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
3045 				   enum bpf_dynptr_type type, u32 offset, u32 size)
3046 {
3047 }
3048 
3049 static inline void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr)
3050 {
3051 }
3052 
3053 static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr)
3054 {
3055 }
3056 #endif /* CONFIG_BPF_SYSCALL */
3057 
3058 static __always_inline int
3059 bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
3060 {
3061 	int ret = -EFAULT;
3062 
3063 	if (IS_ENABLED(CONFIG_BPF_EVENTS))
3064 		ret = copy_from_kernel_nofault(dst, unsafe_ptr, size);
3065 	if (unlikely(ret < 0))
3066 		memset(dst, 0, size);
3067 	return ret;
3068 }
3069 
3070 void __bpf_free_used_btfs(struct btf_mod_pair *used_btfs, u32 len);
3071 
3072 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
3073 						 enum bpf_prog_type type)
3074 {
3075 	return bpf_prog_get_type_dev(ufd, type, false);
3076 }
3077 
3078 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
3079 			  struct bpf_map **used_maps, u32 len);
3080 
3081 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
3082 
3083 int bpf_prog_offload_compile(struct bpf_prog *prog);
3084 void bpf_prog_dev_bound_destroy(struct bpf_prog *prog);
3085 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
3086 			       struct bpf_prog *prog);
3087 
3088 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
3089 
3090 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
3091 int bpf_map_offload_update_elem(struct bpf_map *map,
3092 				void *key, void *value, u64 flags);
3093 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
3094 int bpf_map_offload_get_next_key(struct bpf_map *map,
3095 				 void *key, void *next_key);
3096 
3097 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
3098 
3099 struct bpf_offload_dev *
3100 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
3101 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
3102 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
3103 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
3104 				    struct net_device *netdev);
3105 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
3106 				       struct net_device *netdev);
3107 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
3108 
3109 void unpriv_ebpf_notify(int new_state);
3110 
3111 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
3112 int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
3113 			      struct bpf_prog_aux *prog_aux);
3114 void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, u32 func_id);
3115 int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr);
3116 int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, struct bpf_prog *old_prog);
3117 void bpf_dev_bound_netdev_unregister(struct net_device *dev);
3118 
3119 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
3120 {
3121 	return aux->dev_bound;
3122 }
3123 
3124 static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux)
3125 {
3126 	return aux->offload_requested;
3127 }
3128 
3129 bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs);
3130 
3131 static inline bool bpf_map_is_offloaded(struct bpf_map *map)
3132 {
3133 	return unlikely(map->ops == &bpf_map_offload_ops);
3134 }
3135 
3136 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
3137 void bpf_map_offload_map_free(struct bpf_map *map);
3138 u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map);
3139 int bpf_prog_test_run_syscall(struct bpf_prog *prog,
3140 			      const union bpf_attr *kattr,
3141 			      union bpf_attr __user *uattr);
3142 
3143 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
3144 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
3145 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
3146 int sock_map_bpf_prog_query(const union bpf_attr *attr,
3147 			    union bpf_attr __user *uattr);
3148 int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog);
3149 
3150 void sock_map_unhash(struct sock *sk);
3151 void sock_map_destroy(struct sock *sk);
3152 void sock_map_close(struct sock *sk, long timeout);
3153 #else
3154 static inline int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
3155 					    struct bpf_prog_aux *prog_aux)
3156 {
3157 	return -EOPNOTSUPP;
3158 }
3159 
3160 static inline void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog,
3161 						u32 func_id)
3162 {
3163 	return NULL;
3164 }
3165 
3166 static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog,
3167 					  union bpf_attr *attr)
3168 {
3169 	return -EOPNOTSUPP;
3170 }
3171 
3172 static inline int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog,
3173 					     struct bpf_prog *old_prog)
3174 {
3175 	return -EOPNOTSUPP;
3176 }
3177 
3178 static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev)
3179 {
3180 }
3181 
3182 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
3183 {
3184 	return false;
3185 }
3186 
3187 static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux)
3188 {
3189 	return false;
3190 }
3191 
3192 static inline bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs)
3193 {
3194 	return false;
3195 }
3196 
3197 static inline bool bpf_map_is_offloaded(struct bpf_map *map)
3198 {
3199 	return false;
3200 }
3201 
3202 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
3203 {
3204 	return ERR_PTR(-EOPNOTSUPP);
3205 }
3206 
3207 static inline void bpf_map_offload_map_free(struct bpf_map *map)
3208 {
3209 }
3210 
3211 static inline u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map)
3212 {
3213 	return 0;
3214 }
3215 
3216 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog,
3217 					    const union bpf_attr *kattr,
3218 					    union bpf_attr __user *uattr)
3219 {
3220 	return -ENOTSUPP;
3221 }
3222 
3223 #ifdef CONFIG_BPF_SYSCALL
3224 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
3225 				       struct bpf_prog *prog)
3226 {
3227 	return -EINVAL;
3228 }
3229 
3230 static inline int sock_map_prog_detach(const union bpf_attr *attr,
3231 				       enum bpf_prog_type ptype)
3232 {
3233 	return -EOPNOTSUPP;
3234 }
3235 
3236 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
3237 					   u64 flags)
3238 {
3239 	return -EOPNOTSUPP;
3240 }
3241 
3242 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr,
3243 					  union bpf_attr __user *uattr)
3244 {
3245 	return -EINVAL;
3246 }
3247 
3248 static inline int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog)
3249 {
3250 	return -EOPNOTSUPP;
3251 }
3252 #endif /* CONFIG_BPF_SYSCALL */
3253 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
3254 
3255 static __always_inline void
3256 bpf_prog_inc_misses_counters(const struct bpf_prog_array *array)
3257 {
3258 	const struct bpf_prog_array_item *item;
3259 	struct bpf_prog *prog;
3260 
3261 	if (unlikely(!array))
3262 		return;
3263 
3264 	item = &array->items[0];
3265 	while ((prog = READ_ONCE(item->prog))) {
3266 		bpf_prog_inc_misses_counter(prog);
3267 		item++;
3268 	}
3269 }
3270 
3271 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
3272 void bpf_sk_reuseport_detach(struct sock *sk);
3273 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
3274 				       void *value);
3275 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
3276 				       void *value, u64 map_flags);
3277 #else
3278 static inline void bpf_sk_reuseport_detach(struct sock *sk)
3279 {
3280 }
3281 
3282 #ifdef CONFIG_BPF_SYSCALL
3283 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
3284 						     void *key, void *value)
3285 {
3286 	return -EOPNOTSUPP;
3287 }
3288 
3289 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
3290 						     void *key, void *value,
3291 						     u64 map_flags)
3292 {
3293 	return -EOPNOTSUPP;
3294 }
3295 #endif /* CONFIG_BPF_SYSCALL */
3296 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
3297 
3298 /* verifier prototypes for helper functions called from eBPF programs */
3299 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
3300 extern const struct bpf_func_proto bpf_map_update_elem_proto;
3301 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
3302 extern const struct bpf_func_proto bpf_map_push_elem_proto;
3303 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
3304 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
3305 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto;
3306 
3307 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
3308 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
3309 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
3310 extern const struct bpf_func_proto bpf_tail_call_proto;
3311 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
3312 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
3313 extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto;
3314 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
3315 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
3316 extern const struct bpf_func_proto bpf_get_current_comm_proto;
3317 extern const struct bpf_func_proto bpf_get_stackid_proto;
3318 extern const struct bpf_func_proto bpf_get_stack_proto;
3319 extern const struct bpf_func_proto bpf_get_stack_sleepable_proto;
3320 extern const struct bpf_func_proto bpf_get_task_stack_proto;
3321 extern const struct bpf_func_proto bpf_get_task_stack_sleepable_proto;
3322 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
3323 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
3324 extern const struct bpf_func_proto bpf_sock_map_update_proto;
3325 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
3326 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
3327 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
3328 extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto;
3329 extern const struct bpf_func_proto bpf_current_task_under_cgroup_proto;
3330 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
3331 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
3332 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
3333 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
3334 extern const struct bpf_func_proto bpf_spin_lock_proto;
3335 extern const struct bpf_func_proto bpf_spin_unlock_proto;
3336 extern const struct bpf_func_proto bpf_get_local_storage_proto;
3337 extern const struct bpf_func_proto bpf_strtol_proto;
3338 extern const struct bpf_func_proto bpf_strtoul_proto;
3339 extern const struct bpf_func_proto bpf_tcp_sock_proto;
3340 extern const struct bpf_func_proto bpf_jiffies64_proto;
3341 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
3342 extern const struct bpf_func_proto bpf_event_output_data_proto;
3343 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
3344 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
3345 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
3346 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
3347 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
3348 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto;
3349 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto;
3350 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto;
3351 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
3352 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
3353 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
3354 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
3355 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
3356 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto;
3357 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto;
3358 extern const struct bpf_func_proto bpf_copy_from_user_proto;
3359 extern const struct bpf_func_proto bpf_snprintf_btf_proto;
3360 extern const struct bpf_func_proto bpf_snprintf_proto;
3361 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
3362 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
3363 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto;
3364 extern const struct bpf_func_proto bpf_sock_from_file_proto;
3365 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto;
3366 extern const struct bpf_func_proto bpf_task_storage_get_recur_proto;
3367 extern const struct bpf_func_proto bpf_task_storage_get_proto;
3368 extern const struct bpf_func_proto bpf_task_storage_delete_recur_proto;
3369 extern const struct bpf_func_proto bpf_task_storage_delete_proto;
3370 extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
3371 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto;
3372 extern const struct bpf_func_proto bpf_sk_setsockopt_proto;
3373 extern const struct bpf_func_proto bpf_sk_getsockopt_proto;
3374 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto;
3375 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto;
3376 extern const struct bpf_func_proto bpf_find_vma_proto;
3377 extern const struct bpf_func_proto bpf_loop_proto;
3378 extern const struct bpf_func_proto bpf_copy_from_user_task_proto;
3379 extern const struct bpf_func_proto bpf_set_retval_proto;
3380 extern const struct bpf_func_proto bpf_get_retval_proto;
3381 extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto;
3382 extern const struct bpf_func_proto bpf_cgrp_storage_get_proto;
3383 extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto;
3384 
3385 const struct bpf_func_proto *tracing_prog_func_proto(
3386   enum bpf_func_id func_id, const struct bpf_prog *prog);
3387 
3388 /* Shared helpers among cBPF and eBPF. */
3389 void bpf_user_rnd_init_once(void);
3390 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3391 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3392 
3393 #if defined(CONFIG_NET)
3394 bool bpf_sock_common_is_valid_access(int off, int size,
3395 				     enum bpf_access_type type,
3396 				     struct bpf_insn_access_aux *info);
3397 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3398 			      struct bpf_insn_access_aux *info);
3399 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3400 				const struct bpf_insn *si,
3401 				struct bpf_insn *insn_buf,
3402 				struct bpf_prog *prog,
3403 				u32 *target_size);
3404 int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags,
3405 			       struct bpf_dynptr *ptr);
3406 #else
3407 static inline bool bpf_sock_common_is_valid_access(int off, int size,
3408 						   enum bpf_access_type type,
3409 						   struct bpf_insn_access_aux *info)
3410 {
3411 	return false;
3412 }
3413 static inline bool bpf_sock_is_valid_access(int off, int size,
3414 					    enum bpf_access_type type,
3415 					    struct bpf_insn_access_aux *info)
3416 {
3417 	return false;
3418 }
3419 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3420 					      const struct bpf_insn *si,
3421 					      struct bpf_insn *insn_buf,
3422 					      struct bpf_prog *prog,
3423 					      u32 *target_size)
3424 {
3425 	return 0;
3426 }
3427 static inline int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags,
3428 					     struct bpf_dynptr *ptr)
3429 {
3430 	return -EOPNOTSUPP;
3431 }
3432 #endif
3433 
3434 #ifdef CONFIG_INET
3435 struct sk_reuseport_kern {
3436 	struct sk_buff *skb;
3437 	struct sock *sk;
3438 	struct sock *selected_sk;
3439 	struct sock *migrating_sk;
3440 	void *data_end;
3441 	u32 hash;
3442 	u32 reuseport_id;
3443 	bool bind_inany;
3444 };
3445 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3446 				  struct bpf_insn_access_aux *info);
3447 
3448 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3449 				    const struct bpf_insn *si,
3450 				    struct bpf_insn *insn_buf,
3451 				    struct bpf_prog *prog,
3452 				    u32 *target_size);
3453 
3454 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3455 				  struct bpf_insn_access_aux *info);
3456 
3457 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3458 				    const struct bpf_insn *si,
3459 				    struct bpf_insn *insn_buf,
3460 				    struct bpf_prog *prog,
3461 				    u32 *target_size);
3462 #else
3463 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
3464 						enum bpf_access_type type,
3465 						struct bpf_insn_access_aux *info)
3466 {
3467 	return false;
3468 }
3469 
3470 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3471 						  const struct bpf_insn *si,
3472 						  struct bpf_insn *insn_buf,
3473 						  struct bpf_prog *prog,
3474 						  u32 *target_size)
3475 {
3476 	return 0;
3477 }
3478 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
3479 						enum bpf_access_type type,
3480 						struct bpf_insn_access_aux *info)
3481 {
3482 	return false;
3483 }
3484 
3485 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3486 						  const struct bpf_insn *si,
3487 						  struct bpf_insn *insn_buf,
3488 						  struct bpf_prog *prog,
3489 						  u32 *target_size)
3490 {
3491 	return 0;
3492 }
3493 #endif /* CONFIG_INET */
3494 
3495 enum bpf_text_poke_type {
3496 	BPF_MOD_CALL,
3497 	BPF_MOD_JUMP,
3498 };
3499 
3500 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
3501 		       void *addr1, void *addr2);
3502 
3503 void bpf_arch_poke_desc_update(struct bpf_jit_poke_descriptor *poke,
3504 			       struct bpf_prog *new, struct bpf_prog *old);
3505 
3506 void *bpf_arch_text_copy(void *dst, void *src, size_t len);
3507 int bpf_arch_text_invalidate(void *dst, size_t len);
3508 
3509 struct btf_id_set;
3510 bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
3511 
3512 #define MAX_BPRINTF_VARARGS		12
3513 #define MAX_BPRINTF_BUF			1024
3514 
3515 struct bpf_bprintf_data {
3516 	u32 *bin_args;
3517 	char *buf;
3518 	bool get_bin_args;
3519 	bool get_buf;
3520 };
3521 
3522 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
3523 			u32 num_args, struct bpf_bprintf_data *data);
3524 void bpf_bprintf_cleanup(struct bpf_bprintf_data *data);
3525 
3526 #ifdef CONFIG_BPF_LSM
3527 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype);
3528 void bpf_cgroup_atype_put(int cgroup_atype);
3529 #else
3530 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {}
3531 static inline void bpf_cgroup_atype_put(int cgroup_atype) {}
3532 #endif /* CONFIG_BPF_LSM */
3533 
3534 struct key;
3535 
3536 #ifdef CONFIG_KEYS
3537 struct bpf_key {
3538 	struct key *key;
3539 	bool has_ref;
3540 };
3541 #endif /* CONFIG_KEYS */
3542 
3543 static inline bool type_is_alloc(u32 type)
3544 {
3545 	return type & MEM_ALLOC;
3546 }
3547 
3548 static inline gfp_t bpf_memcg_flags(gfp_t flags)
3549 {
3550 	if (memcg_bpf_enabled())
3551 		return flags | __GFP_ACCOUNT;
3552 	return flags;
3553 }
3554 
3555 static inline bool bpf_is_subprog(const struct bpf_prog *prog)
3556 {
3557 	return prog->aux->func_idx != 0;
3558 }
3559 
3560 #endif /* _LINUX_BPF_H */
3561