xref: /linux/include/linux/bpf.h (revision 3e1c6f35)
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 
34 struct bpf_verifier_env;
35 struct bpf_verifier_log;
36 struct perf_event;
37 struct bpf_prog;
38 struct bpf_prog_aux;
39 struct bpf_map;
40 struct bpf_arena;
41 struct sock;
42 struct seq_file;
43 struct btf;
44 struct btf_type;
45 struct exception_table_entry;
46 struct seq_operations;
47 struct bpf_iter_aux_info;
48 struct bpf_local_storage;
49 struct bpf_local_storage_map;
50 struct kobject;
51 struct mem_cgroup;
52 struct module;
53 struct bpf_func_state;
54 struct ftrace_ops;
55 struct cgroup;
56 struct bpf_token;
57 struct user_namespace;
58 struct super_block;
59 struct inode;
60 
61 extern struct idr btf_idr;
62 extern spinlock_t btf_idr_lock;
63 extern struct kobject *btf_kobj;
64 extern struct bpf_mem_alloc bpf_global_ma, bpf_global_percpu_ma;
65 extern bool bpf_global_ma_set;
66 
67 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64);
68 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data,
69 					struct bpf_iter_aux_info *aux);
70 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
71 typedef unsigned int (*bpf_func_t)(const void *,
72 				   const struct bpf_insn *);
73 struct bpf_iter_seq_info {
74 	const struct seq_operations *seq_ops;
75 	bpf_iter_init_seq_priv_t init_seq_private;
76 	bpf_iter_fini_seq_priv_t fini_seq_private;
77 	u32 seq_priv_size;
78 };
79 
80 /* map is generic key/value storage optionally accessible by eBPF programs */
81 struct bpf_map_ops {
82 	/* funcs callable from userspace (via syscall) */
83 	int (*map_alloc_check)(union bpf_attr *attr);
84 	struct bpf_map *(*map_alloc)(union bpf_attr *attr);
85 	void (*map_release)(struct bpf_map *map, struct file *map_file);
86 	void (*map_free)(struct bpf_map *map);
87 	int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
88 	void (*map_release_uref)(struct bpf_map *map);
89 	void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
90 	int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr,
91 				union bpf_attr __user *uattr);
92 	int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key,
93 					  void *value, u64 flags);
94 	int (*map_lookup_and_delete_batch)(struct bpf_map *map,
95 					   const union bpf_attr *attr,
96 					   union bpf_attr __user *uattr);
97 	int (*map_update_batch)(struct bpf_map *map, struct file *map_file,
98 				const union bpf_attr *attr,
99 				union bpf_attr __user *uattr);
100 	int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr,
101 				union bpf_attr __user *uattr);
102 
103 	/* funcs callable from userspace and from eBPF programs */
104 	void *(*map_lookup_elem)(struct bpf_map *map, void *key);
105 	long (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
106 	long (*map_delete_elem)(struct bpf_map *map, void *key);
107 	long (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
108 	long (*map_pop_elem)(struct bpf_map *map, void *value);
109 	long (*map_peek_elem)(struct bpf_map *map, void *value);
110 	void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu);
111 
112 	/* funcs called by prog_array and perf_event_array map */
113 	void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
114 				int fd);
115 	/* If need_defer is true, the implementation should guarantee that
116 	 * the to-be-put element is still alive before the bpf program, which
117 	 * may manipulate it, exists.
118 	 */
119 	void (*map_fd_put_ptr)(struct bpf_map *map, void *ptr, bool need_defer);
120 	int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
121 	u32 (*map_fd_sys_lookup_elem)(void *ptr);
122 	void (*map_seq_show_elem)(struct bpf_map *map, void *key,
123 				  struct seq_file *m);
124 	int (*map_check_btf)(const struct bpf_map *map,
125 			     const struct btf *btf,
126 			     const struct btf_type *key_type,
127 			     const struct btf_type *value_type);
128 
129 	/* Prog poke tracking helpers. */
130 	int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
131 	void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
132 	void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
133 			     struct bpf_prog *new);
134 
135 	/* Direct value access helpers. */
136 	int (*map_direct_value_addr)(const struct bpf_map *map,
137 				     u64 *imm, u32 off);
138 	int (*map_direct_value_meta)(const struct bpf_map *map,
139 				     u64 imm, u32 *off);
140 	int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
141 	__poll_t (*map_poll)(struct bpf_map *map, struct file *filp,
142 			     struct poll_table_struct *pts);
143 	unsigned long (*map_get_unmapped_area)(struct file *filep, unsigned long addr,
144 					       unsigned long len, unsigned long pgoff,
145 					       unsigned long flags);
146 
147 	/* Functions called by bpf_local_storage maps */
148 	int (*map_local_storage_charge)(struct bpf_local_storage_map *smap,
149 					void *owner, u32 size);
150 	void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap,
151 					   void *owner, u32 size);
152 	struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner);
153 
154 	/* Misc helpers.*/
155 	long (*map_redirect)(struct bpf_map *map, u64 key, u64 flags);
156 
157 	/* map_meta_equal must be implemented for maps that can be
158 	 * used as an inner map.  It is a runtime check to ensure
159 	 * an inner map can be inserted to an outer map.
160 	 *
161 	 * Some properties of the inner map has been used during the
162 	 * verification time.  When inserting an inner map at the runtime,
163 	 * map_meta_equal has to ensure the inserting map has the same
164 	 * properties that the verifier has used earlier.
165 	 */
166 	bool (*map_meta_equal)(const struct bpf_map *meta0,
167 			       const struct bpf_map *meta1);
168 
169 
170 	int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env,
171 					      struct bpf_func_state *caller,
172 					      struct bpf_func_state *callee);
173 	long (*map_for_each_callback)(struct bpf_map *map,
174 				     bpf_callback_t callback_fn,
175 				     void *callback_ctx, u64 flags);
176 
177 	u64 (*map_mem_usage)(const struct bpf_map *map);
178 
179 	/* BTF id of struct allocated by map_alloc */
180 	int *map_btf_id;
181 
182 	/* bpf_iter info used to open a seq_file */
183 	const struct bpf_iter_seq_info *iter_seq_info;
184 };
185 
186 enum {
187 	/* Support at most 11 fields in a BTF type */
188 	BTF_FIELDS_MAX	   = 11,
189 };
190 
191 enum btf_field_type {
192 	BPF_SPIN_LOCK  = (1 << 0),
193 	BPF_TIMER      = (1 << 1),
194 	BPF_KPTR_UNREF = (1 << 2),
195 	BPF_KPTR_REF   = (1 << 3),
196 	BPF_KPTR_PERCPU = (1 << 4),
197 	BPF_KPTR       = BPF_KPTR_UNREF | BPF_KPTR_REF | BPF_KPTR_PERCPU,
198 	BPF_LIST_HEAD  = (1 << 5),
199 	BPF_LIST_NODE  = (1 << 6),
200 	BPF_RB_ROOT    = (1 << 7),
201 	BPF_RB_NODE    = (1 << 8),
202 	BPF_GRAPH_NODE = BPF_RB_NODE | BPF_LIST_NODE,
203 	BPF_GRAPH_ROOT = BPF_RB_ROOT | BPF_LIST_HEAD,
204 	BPF_REFCOUNT   = (1 << 9),
205 	BPF_WORKQUEUE  = (1 << 10),
206 };
207 
208 typedef void (*btf_dtor_kfunc_t)(void *);
209 
210 struct btf_field_kptr {
211 	struct btf *btf;
212 	struct module *module;
213 	/* dtor used if btf_is_kernel(btf), otherwise the type is
214 	 * program-allocated, dtor is NULL,  and __bpf_obj_drop_impl is used
215 	 */
216 	btf_dtor_kfunc_t dtor;
217 	u32 btf_id;
218 };
219 
220 struct btf_field_graph_root {
221 	struct btf *btf;
222 	u32 value_btf_id;
223 	u32 node_offset;
224 	struct btf_record *value_rec;
225 };
226 
227 struct btf_field {
228 	u32 offset;
229 	u32 size;
230 	enum btf_field_type type;
231 	union {
232 		struct btf_field_kptr kptr;
233 		struct btf_field_graph_root graph_root;
234 	};
235 };
236 
237 struct btf_record {
238 	u32 cnt;
239 	u32 field_mask;
240 	int spin_lock_off;
241 	int timer_off;
242 	int wq_off;
243 	int refcount_off;
244 	struct btf_field fields[];
245 };
246 
247 /* Non-opaque version of bpf_rb_node in uapi/linux/bpf.h */
248 struct bpf_rb_node_kern {
249 	struct rb_node rb_node;
250 	void *owner;
251 } __attribute__((aligned(8)));
252 
253 /* Non-opaque version of bpf_list_node in uapi/linux/bpf.h */
254 struct bpf_list_node_kern {
255 	struct list_head list_head;
256 	void *owner;
257 } __attribute__((aligned(8)));
258 
259 struct bpf_map {
260 	const struct bpf_map_ops *ops;
261 	struct bpf_map *inner_map_meta;
262 #ifdef CONFIG_SECURITY
263 	void *security;
264 #endif
265 	enum bpf_map_type map_type;
266 	u32 key_size;
267 	u32 value_size;
268 	u32 max_entries;
269 	u64 map_extra; /* any per-map-type extra fields */
270 	u32 map_flags;
271 	u32 id;
272 	struct btf_record *record;
273 	int numa_node;
274 	u32 btf_key_type_id;
275 	u32 btf_value_type_id;
276 	u32 btf_vmlinux_value_type_id;
277 	struct btf *btf;
278 #ifdef CONFIG_MEMCG_KMEM
279 	struct obj_cgroup *objcg;
280 #endif
281 	char name[BPF_OBJ_NAME_LEN];
282 	struct mutex freeze_mutex;
283 	atomic64_t refcnt;
284 	atomic64_t usercnt;
285 	/* rcu is used before freeing and work is only used during freeing */
286 	union {
287 		struct work_struct work;
288 		struct rcu_head rcu;
289 	};
290 	atomic64_t writecnt;
291 	/* 'Ownership' of program-containing map is claimed by the first program
292 	 * that is going to use this map or by the first program which FD is
293 	 * stored in the map to make sure that all callers and callees have the
294 	 * same prog type, JITed flag and xdp_has_frags flag.
295 	 */
296 	struct {
297 		spinlock_t lock;
298 		enum bpf_prog_type type;
299 		bool jited;
300 		bool xdp_has_frags;
301 	} owner;
302 	bool bypass_spec_v1;
303 	bool frozen; /* write-once; write-protected by freeze_mutex */
304 	bool free_after_mult_rcu_gp;
305 	bool free_after_rcu_gp;
306 	atomic64_t sleepable_refcnt;
307 	s64 __percpu *elem_count;
308 };
309 
btf_field_type_name(enum btf_field_type type)310 static inline const char *btf_field_type_name(enum btf_field_type type)
311 {
312 	switch (type) {
313 	case BPF_SPIN_LOCK:
314 		return "bpf_spin_lock";
315 	case BPF_TIMER:
316 		return "bpf_timer";
317 	case BPF_WORKQUEUE:
318 		return "bpf_wq";
319 	case BPF_KPTR_UNREF:
320 	case BPF_KPTR_REF:
321 		return "kptr";
322 	case BPF_KPTR_PERCPU:
323 		return "percpu_kptr";
324 	case BPF_LIST_HEAD:
325 		return "bpf_list_head";
326 	case BPF_LIST_NODE:
327 		return "bpf_list_node";
328 	case BPF_RB_ROOT:
329 		return "bpf_rb_root";
330 	case BPF_RB_NODE:
331 		return "bpf_rb_node";
332 	case BPF_REFCOUNT:
333 		return "bpf_refcount";
334 	default:
335 		WARN_ON_ONCE(1);
336 		return "unknown";
337 	}
338 }
339 
btf_field_type_size(enum btf_field_type type)340 static inline u32 btf_field_type_size(enum btf_field_type type)
341 {
342 	switch (type) {
343 	case BPF_SPIN_LOCK:
344 		return sizeof(struct bpf_spin_lock);
345 	case BPF_TIMER:
346 		return sizeof(struct bpf_timer);
347 	case BPF_WORKQUEUE:
348 		return sizeof(struct bpf_wq);
349 	case BPF_KPTR_UNREF:
350 	case BPF_KPTR_REF:
351 	case BPF_KPTR_PERCPU:
352 		return sizeof(u64);
353 	case BPF_LIST_HEAD:
354 		return sizeof(struct bpf_list_head);
355 	case BPF_LIST_NODE:
356 		return sizeof(struct bpf_list_node);
357 	case BPF_RB_ROOT:
358 		return sizeof(struct bpf_rb_root);
359 	case BPF_RB_NODE:
360 		return sizeof(struct bpf_rb_node);
361 	case BPF_REFCOUNT:
362 		return sizeof(struct bpf_refcount);
363 	default:
364 		WARN_ON_ONCE(1);
365 		return 0;
366 	}
367 }
368 
btf_field_type_align(enum btf_field_type type)369 static inline u32 btf_field_type_align(enum btf_field_type type)
370 {
371 	switch (type) {
372 	case BPF_SPIN_LOCK:
373 		return __alignof__(struct bpf_spin_lock);
374 	case BPF_TIMER:
375 		return __alignof__(struct bpf_timer);
376 	case BPF_WORKQUEUE:
377 		return __alignof__(struct bpf_wq);
378 	case BPF_KPTR_UNREF:
379 	case BPF_KPTR_REF:
380 	case BPF_KPTR_PERCPU:
381 		return __alignof__(u64);
382 	case BPF_LIST_HEAD:
383 		return __alignof__(struct bpf_list_head);
384 	case BPF_LIST_NODE:
385 		return __alignof__(struct bpf_list_node);
386 	case BPF_RB_ROOT:
387 		return __alignof__(struct bpf_rb_root);
388 	case BPF_RB_NODE:
389 		return __alignof__(struct bpf_rb_node);
390 	case BPF_REFCOUNT:
391 		return __alignof__(struct bpf_refcount);
392 	default:
393 		WARN_ON_ONCE(1);
394 		return 0;
395 	}
396 }
397 
bpf_obj_init_field(const struct btf_field * field,void * addr)398 static inline void bpf_obj_init_field(const struct btf_field *field, void *addr)
399 {
400 	memset(addr, 0, field->size);
401 
402 	switch (field->type) {
403 	case BPF_REFCOUNT:
404 		refcount_set((refcount_t *)addr, 1);
405 		break;
406 	case BPF_RB_NODE:
407 		RB_CLEAR_NODE((struct rb_node *)addr);
408 		break;
409 	case BPF_LIST_HEAD:
410 	case BPF_LIST_NODE:
411 		INIT_LIST_HEAD((struct list_head *)addr);
412 		break;
413 	case BPF_RB_ROOT:
414 		/* RB_ROOT_CACHED 0-inits, no need to do anything after memset */
415 	case BPF_SPIN_LOCK:
416 	case BPF_TIMER:
417 	case BPF_WORKQUEUE:
418 	case BPF_KPTR_UNREF:
419 	case BPF_KPTR_REF:
420 	case BPF_KPTR_PERCPU:
421 		break;
422 	default:
423 		WARN_ON_ONCE(1);
424 		return;
425 	}
426 }
427 
btf_record_has_field(const struct btf_record * rec,enum btf_field_type type)428 static inline bool btf_record_has_field(const struct btf_record *rec, enum btf_field_type type)
429 {
430 	if (IS_ERR_OR_NULL(rec))
431 		return false;
432 	return rec->field_mask & type;
433 }
434 
bpf_obj_init(const struct btf_record * rec,void * obj)435 static inline void bpf_obj_init(const struct btf_record *rec, void *obj)
436 {
437 	int i;
438 
439 	if (IS_ERR_OR_NULL(rec))
440 		return;
441 	for (i = 0; i < rec->cnt; i++)
442 		bpf_obj_init_field(&rec->fields[i], obj + rec->fields[i].offset);
443 }
444 
445 /* 'dst' must be a temporary buffer and should not point to memory that is being
446  * used in parallel by a bpf program or bpf syscall, otherwise the access from
447  * the bpf program or bpf syscall may be corrupted by the reinitialization,
448  * leading to weird problems. Even 'dst' is newly-allocated from bpf memory
449  * allocator, it is still possible for 'dst' to be used in parallel by a bpf
450  * program or bpf syscall.
451  */
check_and_init_map_value(struct bpf_map * map,void * dst)452 static inline void check_and_init_map_value(struct bpf_map *map, void *dst)
453 {
454 	bpf_obj_init(map->record, dst);
455 }
456 
457 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
458  * forced to use 'long' read/writes to try to atomically copy long counters.
459  * Best-effort only.  No barriers here, since it _will_ race with concurrent
460  * updates from BPF programs. Called from bpf syscall and mostly used with
461  * size 8 or 16 bytes, so ask compiler to inline it.
462  */
bpf_long_memcpy(void * dst,const void * src,u32 size)463 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
464 {
465 	const long *lsrc = src;
466 	long *ldst = dst;
467 
468 	size /= sizeof(long);
469 	while (size--)
470 		data_race(*ldst++ = *lsrc++);
471 }
472 
473 /* copy everything but bpf_spin_lock, bpf_timer, and kptrs. There could be one of each. */
bpf_obj_memcpy(struct btf_record * rec,void * dst,void * src,u32 size,bool long_memcpy)474 static inline void bpf_obj_memcpy(struct btf_record *rec,
475 				  void *dst, void *src, u32 size,
476 				  bool long_memcpy)
477 {
478 	u32 curr_off = 0;
479 	int i;
480 
481 	if (IS_ERR_OR_NULL(rec)) {
482 		if (long_memcpy)
483 			bpf_long_memcpy(dst, src, round_up(size, 8));
484 		else
485 			memcpy(dst, src, size);
486 		return;
487 	}
488 
489 	for (i = 0; i < rec->cnt; i++) {
490 		u32 next_off = rec->fields[i].offset;
491 		u32 sz = next_off - curr_off;
492 
493 		memcpy(dst + curr_off, src + curr_off, sz);
494 		curr_off += rec->fields[i].size + sz;
495 	}
496 	memcpy(dst + curr_off, src + curr_off, size - curr_off);
497 }
498 
copy_map_value(struct bpf_map * map,void * dst,void * src)499 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
500 {
501 	bpf_obj_memcpy(map->record, dst, src, map->value_size, false);
502 }
503 
copy_map_value_long(struct bpf_map * map,void * dst,void * src)504 static inline void copy_map_value_long(struct bpf_map *map, void *dst, void *src)
505 {
506 	bpf_obj_memcpy(map->record, dst, src, map->value_size, true);
507 }
508 
bpf_obj_memzero(struct btf_record * rec,void * dst,u32 size)509 static inline void bpf_obj_memzero(struct btf_record *rec, void *dst, u32 size)
510 {
511 	u32 curr_off = 0;
512 	int i;
513 
514 	if (IS_ERR_OR_NULL(rec)) {
515 		memset(dst, 0, size);
516 		return;
517 	}
518 
519 	for (i = 0; i < rec->cnt; i++) {
520 		u32 next_off = rec->fields[i].offset;
521 		u32 sz = next_off - curr_off;
522 
523 		memset(dst + curr_off, 0, sz);
524 		curr_off += rec->fields[i].size + sz;
525 	}
526 	memset(dst + curr_off, 0, size - curr_off);
527 }
528 
zero_map_value(struct bpf_map * map,void * dst)529 static inline void zero_map_value(struct bpf_map *map, void *dst)
530 {
531 	bpf_obj_memzero(map->record, dst, map->value_size);
532 }
533 
534 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
535 			   bool lock_src);
536 void bpf_timer_cancel_and_free(void *timer);
537 void bpf_wq_cancel_and_free(void *timer);
538 void bpf_list_head_free(const struct btf_field *field, void *list_head,
539 			struct bpf_spin_lock *spin_lock);
540 void bpf_rb_root_free(const struct btf_field *field, void *rb_root,
541 		      struct bpf_spin_lock *spin_lock);
542 u64 bpf_arena_get_kern_vm_start(struct bpf_arena *arena);
543 u64 bpf_arena_get_user_vm_start(struct bpf_arena *arena);
544 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
545 
546 struct bpf_offload_dev;
547 struct bpf_offloaded_map;
548 
549 struct bpf_map_dev_ops {
550 	int (*map_get_next_key)(struct bpf_offloaded_map *map,
551 				void *key, void *next_key);
552 	int (*map_lookup_elem)(struct bpf_offloaded_map *map,
553 			       void *key, void *value);
554 	int (*map_update_elem)(struct bpf_offloaded_map *map,
555 			       void *key, void *value, u64 flags);
556 	int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
557 };
558 
559 struct bpf_offloaded_map {
560 	struct bpf_map map;
561 	struct net_device *netdev;
562 	const struct bpf_map_dev_ops *dev_ops;
563 	void *dev_priv;
564 	struct list_head offloads;
565 };
566 
map_to_offmap(struct bpf_map * map)567 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
568 {
569 	return container_of(map, struct bpf_offloaded_map, map);
570 }
571 
bpf_map_offload_neutral(const struct bpf_map * map)572 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
573 {
574 	return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
575 }
576 
bpf_map_support_seq_show(const struct bpf_map * map)577 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
578 {
579 	return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
580 		map->ops->map_seq_show_elem;
581 }
582 
583 int map_check_no_btf(const struct bpf_map *map,
584 		     const struct btf *btf,
585 		     const struct btf_type *key_type,
586 		     const struct btf_type *value_type);
587 
588 bool bpf_map_meta_equal(const struct bpf_map *meta0,
589 			const struct bpf_map *meta1);
590 
591 extern const struct bpf_map_ops bpf_map_offload_ops;
592 
593 /* bpf_type_flag contains a set of flags that are applicable to the values of
594  * arg_type, ret_type and reg_type. For example, a pointer value may be null,
595  * or a memory is read-only. We classify types into two categories: base types
596  * and extended types. Extended types are base types combined with a type flag.
597  *
598  * Currently there are no more than 32 base types in arg_type, ret_type and
599  * reg_types.
600  */
601 #define BPF_BASE_TYPE_BITS	8
602 
603 enum bpf_type_flag {
604 	/* PTR may be NULL. */
605 	PTR_MAYBE_NULL		= BIT(0 + BPF_BASE_TYPE_BITS),
606 
607 	/* MEM is read-only. When applied on bpf_arg, it indicates the arg is
608 	 * compatible with both mutable and immutable memory.
609 	 */
610 	MEM_RDONLY		= BIT(1 + BPF_BASE_TYPE_BITS),
611 
612 	/* MEM points to BPF ring buffer reservation. */
613 	MEM_RINGBUF		= BIT(2 + BPF_BASE_TYPE_BITS),
614 
615 	/* MEM is in user address space. */
616 	MEM_USER		= BIT(3 + BPF_BASE_TYPE_BITS),
617 
618 	/* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged
619 	 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In
620 	 * order to drop this tag, it must be passed into bpf_per_cpu_ptr()
621 	 * or bpf_this_cpu_ptr(), which will return the pointer corresponding
622 	 * to the specified cpu.
623 	 */
624 	MEM_PERCPU		= BIT(4 + BPF_BASE_TYPE_BITS),
625 
626 	/* Indicates that the argument will be released. */
627 	OBJ_RELEASE		= BIT(5 + BPF_BASE_TYPE_BITS),
628 
629 	/* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark
630 	 * unreferenced and referenced kptr loaded from map value using a load
631 	 * instruction, so that they can only be dereferenced but not escape the
632 	 * BPF program into the kernel (i.e. cannot be passed as arguments to
633 	 * kfunc or bpf helpers).
634 	 */
635 	PTR_UNTRUSTED		= BIT(6 + BPF_BASE_TYPE_BITS),
636 
637 	MEM_UNINIT		= BIT(7 + BPF_BASE_TYPE_BITS),
638 
639 	/* DYNPTR points to memory local to the bpf program. */
640 	DYNPTR_TYPE_LOCAL	= BIT(8 + BPF_BASE_TYPE_BITS),
641 
642 	/* DYNPTR points to a kernel-produced ringbuf record. */
643 	DYNPTR_TYPE_RINGBUF	= BIT(9 + BPF_BASE_TYPE_BITS),
644 
645 	/* Size is known at compile time. */
646 	MEM_FIXED_SIZE		= BIT(10 + BPF_BASE_TYPE_BITS),
647 
648 	/* MEM is of an allocated object of type in program BTF. This is used to
649 	 * tag PTR_TO_BTF_ID allocated using bpf_obj_new.
650 	 */
651 	MEM_ALLOC		= BIT(11 + BPF_BASE_TYPE_BITS),
652 
653 	/* PTR was passed from the kernel in a trusted context, and may be
654 	 * passed to KF_TRUSTED_ARGS kfuncs or BPF helper functions.
655 	 * Confusingly, this is _not_ the opposite of PTR_UNTRUSTED above.
656 	 * PTR_UNTRUSTED refers to a kptr that was read directly from a map
657 	 * without invoking bpf_kptr_xchg(). What we really need to know is
658 	 * whether a pointer is safe to pass to a kfunc or BPF helper function.
659 	 * While PTR_UNTRUSTED pointers are unsafe to pass to kfuncs and BPF
660 	 * helpers, they do not cover all possible instances of unsafe
661 	 * pointers. For example, a pointer that was obtained from walking a
662 	 * struct will _not_ get the PTR_UNTRUSTED type modifier, despite the
663 	 * fact that it may be NULL, invalid, etc. This is due to backwards
664 	 * compatibility requirements, as this was the behavior that was first
665 	 * introduced when kptrs were added. The behavior is now considered
666 	 * deprecated, and PTR_UNTRUSTED will eventually be removed.
667 	 *
668 	 * PTR_TRUSTED, on the other hand, is a pointer that the kernel
669 	 * guarantees to be valid and safe to pass to kfuncs and BPF helpers.
670 	 * For example, pointers passed to tracepoint arguments are considered
671 	 * PTR_TRUSTED, as are pointers that are passed to struct_ops
672 	 * callbacks. As alluded to above, pointers that are obtained from
673 	 * walking PTR_TRUSTED pointers are _not_ trusted. For example, if a
674 	 * struct task_struct *task is PTR_TRUSTED, then accessing
675 	 * task->last_wakee will lose the PTR_TRUSTED modifier when it's stored
676 	 * in a BPF register. Similarly, pointers passed to certain programs
677 	 * types such as kretprobes are not guaranteed to be valid, as they may
678 	 * for example contain an object that was recently freed.
679 	 */
680 	PTR_TRUSTED		= BIT(12 + BPF_BASE_TYPE_BITS),
681 
682 	/* MEM is tagged with rcu and memory access needs rcu_read_lock protection. */
683 	MEM_RCU			= BIT(13 + BPF_BASE_TYPE_BITS),
684 
685 	/* Used to tag PTR_TO_BTF_ID | MEM_ALLOC references which are non-owning.
686 	 * Currently only valid for linked-list and rbtree nodes. If the nodes
687 	 * have a bpf_refcount_field, they must be tagged MEM_RCU as well.
688 	 */
689 	NON_OWN_REF		= BIT(14 + BPF_BASE_TYPE_BITS),
690 
691 	/* DYNPTR points to sk_buff */
692 	DYNPTR_TYPE_SKB		= BIT(15 + BPF_BASE_TYPE_BITS),
693 
694 	/* DYNPTR points to xdp_buff */
695 	DYNPTR_TYPE_XDP		= BIT(16 + BPF_BASE_TYPE_BITS),
696 
697 	__BPF_TYPE_FLAG_MAX,
698 	__BPF_TYPE_LAST_FLAG	= __BPF_TYPE_FLAG_MAX - 1,
699 };
700 
701 #define DYNPTR_TYPE_FLAG_MASK	(DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF | DYNPTR_TYPE_SKB \
702 				 | DYNPTR_TYPE_XDP)
703 
704 /* Max number of base types. */
705 #define BPF_BASE_TYPE_LIMIT	(1UL << BPF_BASE_TYPE_BITS)
706 
707 /* Max number of all types. */
708 #define BPF_TYPE_LIMIT		(__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1))
709 
710 /* function argument constraints */
711 enum bpf_arg_type {
712 	ARG_DONTCARE = 0,	/* unused argument in helper function */
713 
714 	/* the following constraints used to prototype
715 	 * bpf_map_lookup/update/delete_elem() functions
716 	 */
717 	ARG_CONST_MAP_PTR,	/* const argument used as pointer to bpf_map */
718 	ARG_PTR_TO_MAP_KEY,	/* pointer to stack used as map key */
719 	ARG_PTR_TO_MAP_VALUE,	/* pointer to stack used as map value */
720 
721 	/* Used to prototype bpf_memcmp() and other functions that access data
722 	 * on eBPF program stack
723 	 */
724 	ARG_PTR_TO_MEM,		/* pointer to valid memory (stack, packet, map value) */
725 	ARG_PTR_TO_ARENA,
726 
727 	ARG_CONST_SIZE,		/* number of bytes accessed from memory */
728 	ARG_CONST_SIZE_OR_ZERO,	/* number of bytes accessed from memory or 0 */
729 
730 	ARG_PTR_TO_CTX,		/* pointer to context */
731 	ARG_ANYTHING,		/* any (initialized) argument is ok */
732 	ARG_PTR_TO_SPIN_LOCK,	/* pointer to bpf_spin_lock */
733 	ARG_PTR_TO_SOCK_COMMON,	/* pointer to sock_common */
734 	ARG_PTR_TO_INT,		/* pointer to int */
735 	ARG_PTR_TO_LONG,	/* pointer to long */
736 	ARG_PTR_TO_SOCKET,	/* pointer to bpf_sock (fullsock) */
737 	ARG_PTR_TO_BTF_ID,	/* pointer to in-kernel struct */
738 	ARG_PTR_TO_RINGBUF_MEM,	/* pointer to dynamically reserved ringbuf memory */
739 	ARG_CONST_ALLOC_SIZE_OR_ZERO,	/* number of allocated bytes requested */
740 	ARG_PTR_TO_BTF_ID_SOCK_COMMON,	/* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
741 	ARG_PTR_TO_PERCPU_BTF_ID,	/* pointer to in-kernel percpu type */
742 	ARG_PTR_TO_FUNC,	/* pointer to a bpf program function */
743 	ARG_PTR_TO_STACK,	/* pointer to stack */
744 	ARG_PTR_TO_CONST_STR,	/* pointer to a null terminated read-only string */
745 	ARG_PTR_TO_TIMER,	/* pointer to bpf_timer */
746 	ARG_PTR_TO_KPTR,	/* pointer to referenced kptr */
747 	ARG_PTR_TO_DYNPTR,      /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */
748 	__BPF_ARG_TYPE_MAX,
749 
750 	/* Extended arg_types. */
751 	ARG_PTR_TO_MAP_VALUE_OR_NULL	= PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE,
752 	ARG_PTR_TO_MEM_OR_NULL		= PTR_MAYBE_NULL | ARG_PTR_TO_MEM,
753 	ARG_PTR_TO_CTX_OR_NULL		= PTR_MAYBE_NULL | ARG_PTR_TO_CTX,
754 	ARG_PTR_TO_SOCKET_OR_NULL	= PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET,
755 	ARG_PTR_TO_STACK_OR_NULL	= PTR_MAYBE_NULL | ARG_PTR_TO_STACK,
756 	ARG_PTR_TO_BTF_ID_OR_NULL	= PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID,
757 	/* pointer to memory does not need to be initialized, helper function must fill
758 	 * all bytes or clear them in error case.
759 	 */
760 	ARG_PTR_TO_UNINIT_MEM		= MEM_UNINIT | ARG_PTR_TO_MEM,
761 	/* Pointer to valid memory of size known at compile time. */
762 	ARG_PTR_TO_FIXED_SIZE_MEM	= MEM_FIXED_SIZE | ARG_PTR_TO_MEM,
763 
764 	/* This must be the last entry. Its purpose is to ensure the enum is
765 	 * wide enough to hold the higher bits reserved for bpf_type_flag.
766 	 */
767 	__BPF_ARG_TYPE_LIMIT	= BPF_TYPE_LIMIT,
768 };
769 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
770 
771 /* type of values returned from helper functions */
772 enum bpf_return_type {
773 	RET_INTEGER,			/* function returns integer */
774 	RET_VOID,			/* function doesn't return anything */
775 	RET_PTR_TO_MAP_VALUE,		/* returns a pointer to map elem value */
776 	RET_PTR_TO_SOCKET,		/* returns a pointer to a socket */
777 	RET_PTR_TO_TCP_SOCK,		/* returns a pointer to a tcp_sock */
778 	RET_PTR_TO_SOCK_COMMON,		/* returns a pointer to a sock_common */
779 	RET_PTR_TO_MEM,			/* returns a pointer to memory */
780 	RET_PTR_TO_MEM_OR_BTF_ID,	/* returns a pointer to a valid memory or a btf_id */
781 	RET_PTR_TO_BTF_ID,		/* returns a pointer to a btf_id */
782 	__BPF_RET_TYPE_MAX,
783 
784 	/* Extended ret_types. */
785 	RET_PTR_TO_MAP_VALUE_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE,
786 	RET_PTR_TO_SOCKET_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_SOCKET,
787 	RET_PTR_TO_TCP_SOCK_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK,
788 	RET_PTR_TO_SOCK_COMMON_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON,
789 	RET_PTR_TO_RINGBUF_MEM_OR_NULL	= PTR_MAYBE_NULL | MEM_RINGBUF | RET_PTR_TO_MEM,
790 	RET_PTR_TO_DYNPTR_MEM_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_MEM,
791 	RET_PTR_TO_BTF_ID_OR_NULL	= PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID,
792 	RET_PTR_TO_BTF_ID_TRUSTED	= PTR_TRUSTED	 | RET_PTR_TO_BTF_ID,
793 
794 	/* This must be the last entry. Its purpose is to ensure the enum is
795 	 * wide enough to hold the higher bits reserved for bpf_type_flag.
796 	 */
797 	__BPF_RET_TYPE_LIMIT	= BPF_TYPE_LIMIT,
798 };
799 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
800 
801 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
802  * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
803  * instructions after verifying
804  */
805 struct bpf_func_proto {
806 	u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
807 	bool gpl_only;
808 	bool pkt_access;
809 	bool might_sleep;
810 	enum bpf_return_type ret_type;
811 	union {
812 		struct {
813 			enum bpf_arg_type arg1_type;
814 			enum bpf_arg_type arg2_type;
815 			enum bpf_arg_type arg3_type;
816 			enum bpf_arg_type arg4_type;
817 			enum bpf_arg_type arg5_type;
818 		};
819 		enum bpf_arg_type arg_type[5];
820 	};
821 	union {
822 		struct {
823 			u32 *arg1_btf_id;
824 			u32 *arg2_btf_id;
825 			u32 *arg3_btf_id;
826 			u32 *arg4_btf_id;
827 			u32 *arg5_btf_id;
828 		};
829 		u32 *arg_btf_id[5];
830 		struct {
831 			size_t arg1_size;
832 			size_t arg2_size;
833 			size_t arg3_size;
834 			size_t arg4_size;
835 			size_t arg5_size;
836 		};
837 		size_t arg_size[5];
838 	};
839 	int *ret_btf_id; /* return value btf_id */
840 	bool (*allowed)(const struct bpf_prog *prog);
841 };
842 
843 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
844  * the first argument to eBPF programs.
845  * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
846  */
847 struct bpf_context;
848 
849 enum bpf_access_type {
850 	BPF_READ = 1,
851 	BPF_WRITE = 2
852 };
853 
854 /* types of values stored in eBPF registers */
855 /* Pointer types represent:
856  * pointer
857  * pointer + imm
858  * pointer + (u16) var
859  * pointer + (u16) var + imm
860  * if (range > 0) then [ptr, ptr + range - off) is safe to access
861  * if (id > 0) means that some 'var' was added
862  * if (off > 0) means that 'imm' was added
863  */
864 enum bpf_reg_type {
865 	NOT_INIT = 0,		 /* nothing was written into register */
866 	SCALAR_VALUE,		 /* reg doesn't contain a valid pointer */
867 	PTR_TO_CTX,		 /* reg points to bpf_context */
868 	CONST_PTR_TO_MAP,	 /* reg points to struct bpf_map */
869 	PTR_TO_MAP_VALUE,	 /* reg points to map element value */
870 	PTR_TO_MAP_KEY,		 /* reg points to a map element key */
871 	PTR_TO_STACK,		 /* reg == frame_pointer + offset */
872 	PTR_TO_PACKET_META,	 /* skb->data - meta_len */
873 	PTR_TO_PACKET,		 /* reg points to skb->data */
874 	PTR_TO_PACKET_END,	 /* skb->data + headlen */
875 	PTR_TO_FLOW_KEYS,	 /* reg points to bpf_flow_keys */
876 	PTR_TO_SOCKET,		 /* reg points to struct bpf_sock */
877 	PTR_TO_SOCK_COMMON,	 /* reg points to sock_common */
878 	PTR_TO_TCP_SOCK,	 /* reg points to struct tcp_sock */
879 	PTR_TO_TP_BUFFER,	 /* reg points to a writable raw tp's buffer */
880 	PTR_TO_XDP_SOCK,	 /* reg points to struct xdp_sock */
881 	/* PTR_TO_BTF_ID points to a kernel struct that does not need
882 	 * to be null checked by the BPF program. This does not imply the
883 	 * pointer is _not_ null and in practice this can easily be a null
884 	 * pointer when reading pointer chains. The assumption is program
885 	 * context will handle null pointer dereference typically via fault
886 	 * handling. The verifier must keep this in mind and can make no
887 	 * assumptions about null or non-null when doing branch analysis.
888 	 * Further, when passed into helpers the helpers can not, without
889 	 * additional context, assume the value is non-null.
890 	 */
891 	PTR_TO_BTF_ID,
892 	/* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
893 	 * been checked for null. Used primarily to inform the verifier
894 	 * an explicit null check is required for this struct.
895 	 */
896 	PTR_TO_MEM,		 /* reg points to valid memory region */
897 	PTR_TO_ARENA,
898 	PTR_TO_BUF,		 /* reg points to a read/write buffer */
899 	PTR_TO_FUNC,		 /* reg points to a bpf program function */
900 	CONST_PTR_TO_DYNPTR,	 /* reg points to a const struct bpf_dynptr */
901 	__BPF_REG_TYPE_MAX,
902 
903 	/* Extended reg_types. */
904 	PTR_TO_MAP_VALUE_OR_NULL	= PTR_MAYBE_NULL | PTR_TO_MAP_VALUE,
905 	PTR_TO_SOCKET_OR_NULL		= PTR_MAYBE_NULL | PTR_TO_SOCKET,
906 	PTR_TO_SOCK_COMMON_OR_NULL	= PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON,
907 	PTR_TO_TCP_SOCK_OR_NULL		= PTR_MAYBE_NULL | PTR_TO_TCP_SOCK,
908 	PTR_TO_BTF_ID_OR_NULL		= PTR_MAYBE_NULL | PTR_TO_BTF_ID,
909 
910 	/* This must be the last entry. Its purpose is to ensure the enum is
911 	 * wide enough to hold the higher bits reserved for bpf_type_flag.
912 	 */
913 	__BPF_REG_TYPE_LIMIT	= BPF_TYPE_LIMIT,
914 };
915 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
916 
917 /* The information passed from prog-specific *_is_valid_access
918  * back to the verifier.
919  */
920 struct bpf_insn_access_aux {
921 	enum bpf_reg_type reg_type;
922 	union {
923 		int ctx_field_size;
924 		struct {
925 			struct btf *btf;
926 			u32 btf_id;
927 		};
928 	};
929 	struct bpf_verifier_log *log; /* for verbose logs */
930 };
931 
932 static inline void
bpf_ctx_record_field_size(struct bpf_insn_access_aux * aux,u32 size)933 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
934 {
935 	aux->ctx_field_size = size;
936 }
937 
bpf_is_ldimm64(const struct bpf_insn * insn)938 static bool bpf_is_ldimm64(const struct bpf_insn *insn)
939 {
940 	return insn->code == (BPF_LD | BPF_IMM | BPF_DW);
941 }
942 
bpf_pseudo_func(const struct bpf_insn * insn)943 static inline bool bpf_pseudo_func(const struct bpf_insn *insn)
944 {
945 	return bpf_is_ldimm64(insn) && insn->src_reg == BPF_PSEUDO_FUNC;
946 }
947 
948 struct bpf_prog_ops {
949 	int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
950 			union bpf_attr __user *uattr);
951 };
952 
953 struct bpf_reg_state;
954 struct bpf_verifier_ops {
955 	/* return eBPF function prototype for verification */
956 	const struct bpf_func_proto *
957 	(*get_func_proto)(enum bpf_func_id func_id,
958 			  const struct bpf_prog *prog);
959 
960 	/* return true if 'size' wide access at offset 'off' within bpf_context
961 	 * with 'type' (read or write) is allowed
962 	 */
963 	bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
964 				const struct bpf_prog *prog,
965 				struct bpf_insn_access_aux *info);
966 	int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
967 			    const struct bpf_prog *prog);
968 	int (*gen_ld_abs)(const struct bpf_insn *orig,
969 			  struct bpf_insn *insn_buf);
970 	u32 (*convert_ctx_access)(enum bpf_access_type type,
971 				  const struct bpf_insn *src,
972 				  struct bpf_insn *dst,
973 				  struct bpf_prog *prog, u32 *target_size);
974 	int (*btf_struct_access)(struct bpf_verifier_log *log,
975 				 const struct bpf_reg_state *reg,
976 				 int off, int size);
977 };
978 
979 struct bpf_prog_offload_ops {
980 	/* verifier basic callbacks */
981 	int (*insn_hook)(struct bpf_verifier_env *env,
982 			 int insn_idx, int prev_insn_idx);
983 	int (*finalize)(struct bpf_verifier_env *env);
984 	/* verifier optimization callbacks (called after .finalize) */
985 	int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
986 			    struct bpf_insn *insn);
987 	int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
988 	/* program management callbacks */
989 	int (*prepare)(struct bpf_prog *prog);
990 	int (*translate)(struct bpf_prog *prog);
991 	void (*destroy)(struct bpf_prog *prog);
992 };
993 
994 struct bpf_prog_offload {
995 	struct bpf_prog		*prog;
996 	struct net_device	*netdev;
997 	struct bpf_offload_dev	*offdev;
998 	void			*dev_priv;
999 	struct list_head	offloads;
1000 	bool			dev_state;
1001 	bool			opt_failed;
1002 	void			*jited_image;
1003 	u32			jited_len;
1004 };
1005 
1006 enum bpf_cgroup_storage_type {
1007 	BPF_CGROUP_STORAGE_SHARED,
1008 	BPF_CGROUP_STORAGE_PERCPU,
1009 	__BPF_CGROUP_STORAGE_MAX
1010 };
1011 
1012 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
1013 
1014 /* The longest tracepoint has 12 args.
1015  * See include/trace/bpf_probe.h
1016  */
1017 #define MAX_BPF_FUNC_ARGS 12
1018 
1019 /* The maximum number of arguments passed through registers
1020  * a single function may have.
1021  */
1022 #define MAX_BPF_FUNC_REG_ARGS 5
1023 
1024 /* The argument is a structure. */
1025 #define BTF_FMODEL_STRUCT_ARG		BIT(0)
1026 
1027 /* The argument is signed. */
1028 #define BTF_FMODEL_SIGNED_ARG		BIT(1)
1029 
1030 struct btf_func_model {
1031 	u8 ret_size;
1032 	u8 ret_flags;
1033 	u8 nr_args;
1034 	u8 arg_size[MAX_BPF_FUNC_ARGS];
1035 	u8 arg_flags[MAX_BPF_FUNC_ARGS];
1036 };
1037 
1038 /* Restore arguments before returning from trampoline to let original function
1039  * continue executing. This flag is used for fentry progs when there are no
1040  * fexit progs.
1041  */
1042 #define BPF_TRAMP_F_RESTORE_REGS	BIT(0)
1043 /* Call original function after fentry progs, but before fexit progs.
1044  * Makes sense for fentry/fexit, normal calls and indirect calls.
1045  */
1046 #define BPF_TRAMP_F_CALL_ORIG		BIT(1)
1047 /* Skip current frame and return to parent.  Makes sense for fentry/fexit
1048  * programs only. Should not be used with normal calls and indirect calls.
1049  */
1050 #define BPF_TRAMP_F_SKIP_FRAME		BIT(2)
1051 /* Store IP address of the caller on the trampoline stack,
1052  * so it's available for trampoline's programs.
1053  */
1054 #define BPF_TRAMP_F_IP_ARG		BIT(3)
1055 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */
1056 #define BPF_TRAMP_F_RET_FENTRY_RET	BIT(4)
1057 
1058 /* Get original function from stack instead of from provided direct address.
1059  * Makes sense for trampolines with fexit or fmod_ret programs.
1060  */
1061 #define BPF_TRAMP_F_ORIG_STACK		BIT(5)
1062 
1063 /* This trampoline is on a function with another ftrace_ops with IPMODIFY,
1064  * e.g., a live patch. This flag is set and cleared by ftrace call backs,
1065  */
1066 #define BPF_TRAMP_F_SHARE_IPMODIFY	BIT(6)
1067 
1068 /* Indicate that current trampoline is in a tail call context. Then, it has to
1069  * cache and restore tail_call_cnt to avoid infinite tail call loop.
1070  */
1071 #define BPF_TRAMP_F_TAIL_CALL_CTX	BIT(7)
1072 
1073 /*
1074  * Indicate the trampoline should be suitable to receive indirect calls;
1075  * without this indirectly calling the generated code can result in #UD/#CP,
1076  * depending on the CFI options.
1077  *
1078  * Used by bpf_struct_ops.
1079  *
1080  * Incompatible with FENTRY usage, overloads @func_addr argument.
1081  */
1082 #define BPF_TRAMP_F_INDIRECT		BIT(8)
1083 
1084 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
1085  * bytes on x86.
1086  */
1087 enum {
1088 #if defined(__s390x__)
1089 	BPF_MAX_TRAMP_LINKS = 27,
1090 #else
1091 	BPF_MAX_TRAMP_LINKS = 38,
1092 #endif
1093 };
1094 
1095 struct bpf_tramp_links {
1096 	struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS];
1097 	int nr_links;
1098 };
1099 
1100 struct bpf_tramp_run_ctx;
1101 
1102 /* Different use cases for BPF trampoline:
1103  * 1. replace nop at the function entry (kprobe equivalent)
1104  *    flags = BPF_TRAMP_F_RESTORE_REGS
1105  *    fentry = a set of programs to run before returning from trampoline
1106  *
1107  * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
1108  *    flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
1109  *    orig_call = fentry_ip + MCOUNT_INSN_SIZE
1110  *    fentry = a set of program to run before calling original function
1111  *    fexit = a set of program to run after original function
1112  *
1113  * 3. replace direct call instruction anywhere in the function body
1114  *    or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
1115  *    With flags = 0
1116  *      fentry = a set of programs to run before returning from trampoline
1117  *    With flags = BPF_TRAMP_F_CALL_ORIG
1118  *      orig_call = original callback addr or direct function addr
1119  *      fentry = a set of program to run before calling original function
1120  *      fexit = a set of program to run after original function
1121  */
1122 struct bpf_tramp_image;
1123 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *image_end,
1124 				const struct btf_func_model *m, u32 flags,
1125 				struct bpf_tramp_links *tlinks,
1126 				void *func_addr);
1127 void *arch_alloc_bpf_trampoline(unsigned int size);
1128 void arch_free_bpf_trampoline(void *image, unsigned int size);
1129 int __must_check arch_protect_bpf_trampoline(void *image, unsigned int size);
1130 int arch_bpf_trampoline_size(const struct btf_func_model *m, u32 flags,
1131 			     struct bpf_tramp_links *tlinks, void *func_addr);
1132 
1133 u64 notrace __bpf_prog_enter_sleepable_recur(struct bpf_prog *prog,
1134 					     struct bpf_tramp_run_ctx *run_ctx);
1135 void notrace __bpf_prog_exit_sleepable_recur(struct bpf_prog *prog, u64 start,
1136 					     struct bpf_tramp_run_ctx *run_ctx);
1137 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
1138 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
1139 typedef u64 (*bpf_trampoline_enter_t)(struct bpf_prog *prog,
1140 				      struct bpf_tramp_run_ctx *run_ctx);
1141 typedef void (*bpf_trampoline_exit_t)(struct bpf_prog *prog, u64 start,
1142 				      struct bpf_tramp_run_ctx *run_ctx);
1143 bpf_trampoline_enter_t bpf_trampoline_enter(const struct bpf_prog *prog);
1144 bpf_trampoline_exit_t bpf_trampoline_exit(const struct bpf_prog *prog);
1145 
1146 struct bpf_ksym {
1147 	unsigned long		 start;
1148 	unsigned long		 end;
1149 	char			 name[KSYM_NAME_LEN];
1150 	struct list_head	 lnode;
1151 	struct latch_tree_node	 tnode;
1152 	bool			 prog;
1153 };
1154 
1155 enum bpf_tramp_prog_type {
1156 	BPF_TRAMP_FENTRY,
1157 	BPF_TRAMP_FEXIT,
1158 	BPF_TRAMP_MODIFY_RETURN,
1159 	BPF_TRAMP_MAX,
1160 	BPF_TRAMP_REPLACE, /* more than MAX */
1161 };
1162 
1163 struct bpf_tramp_image {
1164 	void *image;
1165 	int size;
1166 	struct bpf_ksym ksym;
1167 	struct percpu_ref pcref;
1168 	void *ip_after_call;
1169 	void *ip_epilogue;
1170 	union {
1171 		struct rcu_head rcu;
1172 		struct work_struct work;
1173 	};
1174 };
1175 
1176 struct bpf_trampoline {
1177 	/* hlist for trampoline_table */
1178 	struct hlist_node hlist;
1179 	struct ftrace_ops *fops;
1180 	/* serializes access to fields of this trampoline */
1181 	struct mutex mutex;
1182 	refcount_t refcnt;
1183 	u32 flags;
1184 	u64 key;
1185 	struct {
1186 		struct btf_func_model model;
1187 		void *addr;
1188 		bool ftrace_managed;
1189 	} func;
1190 	/* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
1191 	 * program by replacing one of its functions. func.addr is the address
1192 	 * of the function it replaced.
1193 	 */
1194 	struct bpf_prog *extension_prog;
1195 	/* list of BPF programs using this trampoline */
1196 	struct hlist_head progs_hlist[BPF_TRAMP_MAX];
1197 	/* Number of attached programs. A counter per kind. */
1198 	int progs_cnt[BPF_TRAMP_MAX];
1199 	/* Executable image of trampoline */
1200 	struct bpf_tramp_image *cur_image;
1201 };
1202 
1203 struct bpf_attach_target_info {
1204 	struct btf_func_model fmodel;
1205 	long tgt_addr;
1206 	struct module *tgt_mod;
1207 	const char *tgt_name;
1208 	const struct btf_type *tgt_type;
1209 };
1210 
1211 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
1212 
1213 struct bpf_dispatcher_prog {
1214 	struct bpf_prog *prog;
1215 	refcount_t users;
1216 };
1217 
1218 struct bpf_dispatcher {
1219 	/* dispatcher mutex */
1220 	struct mutex mutex;
1221 	void *func;
1222 	struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
1223 	int num_progs;
1224 	void *image;
1225 	void *rw_image;
1226 	u32 image_off;
1227 	struct bpf_ksym ksym;
1228 #ifdef CONFIG_HAVE_STATIC_CALL
1229 	struct static_call_key *sc_key;
1230 	void *sc_tramp;
1231 #endif
1232 };
1233 
1234 #ifndef __bpfcall
1235 #define __bpfcall __nocfi
1236 #endif
1237 
bpf_dispatcher_nop_func(const void * ctx,const struct bpf_insn * insnsi,bpf_func_t bpf_func)1238 static __always_inline __bpfcall unsigned int bpf_dispatcher_nop_func(
1239 	const void *ctx,
1240 	const struct bpf_insn *insnsi,
1241 	bpf_func_t bpf_func)
1242 {
1243 	return bpf_func(ctx, insnsi);
1244 }
1245 
1246 /* the implementation of the opaque uapi struct bpf_dynptr */
1247 struct bpf_dynptr_kern {
1248 	void *data;
1249 	/* Size represents the number of usable bytes of dynptr data.
1250 	 * If for example the offset is at 4 for a local dynptr whose data is
1251 	 * of type u64, the number of usable bytes is 4.
1252 	 *
1253 	 * The upper 8 bits are reserved. It is as follows:
1254 	 * Bits 0 - 23 = size
1255 	 * Bits 24 - 30 = dynptr type
1256 	 * Bit 31 = whether dynptr is read-only
1257 	 */
1258 	u32 size;
1259 	u32 offset;
1260 } __aligned(8);
1261 
1262 enum bpf_dynptr_type {
1263 	BPF_DYNPTR_TYPE_INVALID,
1264 	/* Points to memory that is local to the bpf program */
1265 	BPF_DYNPTR_TYPE_LOCAL,
1266 	/* Underlying data is a ringbuf record */
1267 	BPF_DYNPTR_TYPE_RINGBUF,
1268 	/* Underlying data is a sk_buff */
1269 	BPF_DYNPTR_TYPE_SKB,
1270 	/* Underlying data is a xdp_buff */
1271 	BPF_DYNPTR_TYPE_XDP,
1272 };
1273 
1274 int bpf_dynptr_check_size(u32 size);
1275 u32 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr);
1276 const void *__bpf_dynptr_data(const struct bpf_dynptr_kern *ptr, u32 len);
1277 void *__bpf_dynptr_data_rw(const struct bpf_dynptr_kern *ptr, u32 len);
1278 bool __bpf_dynptr_is_rdonly(const struct bpf_dynptr_kern *ptr);
1279 
1280 #ifdef CONFIG_BPF_JIT
1281 int bpf_trampoline_link_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
1282 int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
1283 struct bpf_trampoline *bpf_trampoline_get(u64 key,
1284 					  struct bpf_attach_target_info *tgt_info);
1285 void bpf_trampoline_put(struct bpf_trampoline *tr);
1286 int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_funcs);
1287 
1288 /*
1289  * When the architecture supports STATIC_CALL replace the bpf_dispatcher_fn
1290  * indirection with a direct call to the bpf program. If the architecture does
1291  * not have STATIC_CALL, avoid a double-indirection.
1292  */
1293 #ifdef CONFIG_HAVE_STATIC_CALL
1294 
1295 #define __BPF_DISPATCHER_SC_INIT(_name)				\
1296 	.sc_key = &STATIC_CALL_KEY(_name),			\
1297 	.sc_tramp = STATIC_CALL_TRAMP_ADDR(_name),
1298 
1299 #define __BPF_DISPATCHER_SC(name)				\
1300 	DEFINE_STATIC_CALL(bpf_dispatcher_##name##_call, bpf_dispatcher_nop_func)
1301 
1302 #define __BPF_DISPATCHER_CALL(name)				\
1303 	static_call(bpf_dispatcher_##name##_call)(ctx, insnsi, bpf_func)
1304 
1305 #define __BPF_DISPATCHER_UPDATE(_d, _new)			\
1306 	__static_call_update((_d)->sc_key, (_d)->sc_tramp, (_new))
1307 
1308 #else
1309 #define __BPF_DISPATCHER_SC_INIT(name)
1310 #define __BPF_DISPATCHER_SC(name)
1311 #define __BPF_DISPATCHER_CALL(name)		bpf_func(ctx, insnsi)
1312 #define __BPF_DISPATCHER_UPDATE(_d, _new)
1313 #endif
1314 
1315 #define BPF_DISPATCHER_INIT(_name) {				\
1316 	.mutex = __MUTEX_INITIALIZER(_name.mutex),		\
1317 	.func = &_name##_func,					\
1318 	.progs = {},						\
1319 	.num_progs = 0,						\
1320 	.image = NULL,						\
1321 	.image_off = 0,						\
1322 	.ksym = {						\
1323 		.name  = #_name,				\
1324 		.lnode = LIST_HEAD_INIT(_name.ksym.lnode),	\
1325 	},							\
1326 	__BPF_DISPATCHER_SC_INIT(_name##_call)			\
1327 }
1328 
1329 #define DEFINE_BPF_DISPATCHER(name)					\
1330 	__BPF_DISPATCHER_SC(name);					\
1331 	noinline __bpfcall unsigned int bpf_dispatcher_##name##_func(	\
1332 		const void *ctx,					\
1333 		const struct bpf_insn *insnsi,				\
1334 		bpf_func_t bpf_func)					\
1335 	{								\
1336 		return __BPF_DISPATCHER_CALL(name);			\
1337 	}								\
1338 	EXPORT_SYMBOL(bpf_dispatcher_##name##_func);			\
1339 	struct bpf_dispatcher bpf_dispatcher_##name =			\
1340 		BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
1341 
1342 #define DECLARE_BPF_DISPATCHER(name)					\
1343 	unsigned int bpf_dispatcher_##name##_func(			\
1344 		const void *ctx,					\
1345 		const struct bpf_insn *insnsi,				\
1346 		bpf_func_t bpf_func);					\
1347 	extern struct bpf_dispatcher bpf_dispatcher_##name;
1348 
1349 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
1350 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
1351 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
1352 				struct bpf_prog *to);
1353 /* Called only from JIT-enabled code, so there's no need for stubs. */
1354 void bpf_image_ksym_add(void *data, unsigned int size, struct bpf_ksym *ksym);
1355 void bpf_image_ksym_del(struct bpf_ksym *ksym);
1356 void bpf_ksym_add(struct bpf_ksym *ksym);
1357 void bpf_ksym_del(struct bpf_ksym *ksym);
1358 int bpf_jit_charge_modmem(u32 size);
1359 void bpf_jit_uncharge_modmem(u32 size);
1360 bool bpf_prog_has_trampoline(const struct bpf_prog *prog);
1361 #else
bpf_trampoline_link_prog(struct bpf_tramp_link * link,struct bpf_trampoline * tr)1362 static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
1363 					   struct bpf_trampoline *tr)
1364 {
1365 	return -ENOTSUPP;
1366 }
bpf_trampoline_unlink_prog(struct bpf_tramp_link * link,struct bpf_trampoline * tr)1367 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
1368 					     struct bpf_trampoline *tr)
1369 {
1370 	return -ENOTSUPP;
1371 }
bpf_trampoline_get(u64 key,struct bpf_attach_target_info * tgt_info)1372 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
1373 							struct bpf_attach_target_info *tgt_info)
1374 {
1375 	return NULL;
1376 }
bpf_trampoline_put(struct bpf_trampoline * tr)1377 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
1378 #define DEFINE_BPF_DISPATCHER(name)
1379 #define DECLARE_BPF_DISPATCHER(name)
1380 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
1381 #define BPF_DISPATCHER_PTR(name) NULL
bpf_dispatcher_change_prog(struct bpf_dispatcher * d,struct bpf_prog * from,struct bpf_prog * to)1382 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
1383 					      struct bpf_prog *from,
1384 					      struct bpf_prog *to) {}
is_bpf_image_address(unsigned long address)1385 static inline bool is_bpf_image_address(unsigned long address)
1386 {
1387 	return false;
1388 }
bpf_prog_has_trampoline(const struct bpf_prog * prog)1389 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog)
1390 {
1391 	return false;
1392 }
1393 #endif
1394 
1395 struct bpf_func_info_aux {
1396 	u16 linkage;
1397 	bool unreliable;
1398 	bool called : 1;
1399 	bool verified : 1;
1400 };
1401 
1402 enum bpf_jit_poke_reason {
1403 	BPF_POKE_REASON_TAIL_CALL,
1404 };
1405 
1406 /* Descriptor of pokes pointing /into/ the JITed image. */
1407 struct bpf_jit_poke_descriptor {
1408 	void *tailcall_target;
1409 	void *tailcall_bypass;
1410 	void *bypass_addr;
1411 	void *aux;
1412 	union {
1413 		struct {
1414 			struct bpf_map *map;
1415 			u32 key;
1416 		} tail_call;
1417 	};
1418 	bool tailcall_target_stable;
1419 	u8 adj_off;
1420 	u16 reason;
1421 	u32 insn_idx;
1422 };
1423 
1424 /* reg_type info for ctx arguments */
1425 struct bpf_ctx_arg_aux {
1426 	u32 offset;
1427 	enum bpf_reg_type reg_type;
1428 	struct btf *btf;
1429 	u32 btf_id;
1430 };
1431 
1432 struct btf_mod_pair {
1433 	struct btf *btf;
1434 	struct module *module;
1435 };
1436 
1437 struct bpf_kfunc_desc_tab;
1438 
1439 struct bpf_prog_aux {
1440 	atomic64_t refcnt;
1441 	u32 used_map_cnt;
1442 	u32 used_btf_cnt;
1443 	u32 max_ctx_offset;
1444 	u32 max_pkt_offset;
1445 	u32 max_tp_access;
1446 	u32 stack_depth;
1447 	u32 id;
1448 	u32 func_cnt; /* used by non-func prog as the number of func progs */
1449 	u32 real_func_cnt; /* includes hidden progs, only used for JIT and freeing progs */
1450 	u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
1451 	u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1452 	u32 ctx_arg_info_size;
1453 	u32 max_rdonly_access;
1454 	u32 max_rdwr_access;
1455 	struct btf *attach_btf;
1456 	const struct bpf_ctx_arg_aux *ctx_arg_info;
1457 	struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
1458 	struct bpf_prog *dst_prog;
1459 	struct bpf_trampoline *dst_trampoline;
1460 	enum bpf_prog_type saved_dst_prog_type;
1461 	enum bpf_attach_type saved_dst_attach_type;
1462 	bool verifier_zext; /* Zero extensions has been inserted by verifier. */
1463 	bool dev_bound; /* Program is bound to the netdev. */
1464 	bool offload_requested; /* Program is bound and offloaded to the netdev. */
1465 	bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
1466 	bool attach_tracing_prog; /* true if tracing another tracing program */
1467 	bool func_proto_unreliable;
1468 	bool tail_call_reachable;
1469 	bool xdp_has_frags;
1470 	bool exception_cb;
1471 	bool exception_boundary;
1472 	struct bpf_arena *arena;
1473 	/* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
1474 	const struct btf_type *attach_func_proto;
1475 	/* function name for valid attach_btf_id */
1476 	const char *attach_func_name;
1477 	struct bpf_prog **func;
1478 	void *jit_data; /* JIT specific data. arch dependent */
1479 	struct bpf_jit_poke_descriptor *poke_tab;
1480 	struct bpf_kfunc_desc_tab *kfunc_tab;
1481 	struct bpf_kfunc_btf_tab *kfunc_btf_tab;
1482 	u32 size_poke_tab;
1483 #ifdef CONFIG_FINEIBT
1484 	struct bpf_ksym ksym_prefix;
1485 #endif
1486 	struct bpf_ksym ksym;
1487 	const struct bpf_prog_ops *ops;
1488 	struct bpf_map **used_maps;
1489 	struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
1490 	struct btf_mod_pair *used_btfs;
1491 	struct bpf_prog *prog;
1492 	struct user_struct *user;
1493 	u64 load_time; /* ns since boottime */
1494 	u32 verified_insns;
1495 	int cgroup_atype; /* enum cgroup_bpf_attach_type */
1496 	struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1497 	char name[BPF_OBJ_NAME_LEN];
1498 	u64 (*bpf_exception_cb)(u64 cookie, u64 sp, u64 bp, u64, u64);
1499 #ifdef CONFIG_SECURITY
1500 	void *security;
1501 #endif
1502 	struct bpf_token *token;
1503 	struct bpf_prog_offload *offload;
1504 	struct btf *btf;
1505 	struct bpf_func_info *func_info;
1506 	struct bpf_func_info_aux *func_info_aux;
1507 	/* bpf_line_info loaded from userspace.  linfo->insn_off
1508 	 * has the xlated insn offset.
1509 	 * Both the main and sub prog share the same linfo.
1510 	 * The subprog can access its first linfo by
1511 	 * using the linfo_idx.
1512 	 */
1513 	struct bpf_line_info *linfo;
1514 	/* jited_linfo is the jited addr of the linfo.  It has a
1515 	 * one to one mapping to linfo:
1516 	 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
1517 	 * Both the main and sub prog share the same jited_linfo.
1518 	 * The subprog can access its first jited_linfo by
1519 	 * using the linfo_idx.
1520 	 */
1521 	void **jited_linfo;
1522 	u32 func_info_cnt;
1523 	u32 nr_linfo;
1524 	/* subprog can use linfo_idx to access its first linfo and
1525 	 * jited_linfo.
1526 	 * main prog always has linfo_idx == 0
1527 	 */
1528 	u32 linfo_idx;
1529 	struct module *mod;
1530 	u32 num_exentries;
1531 	struct exception_table_entry *extable;
1532 	union {
1533 		struct work_struct work;
1534 		struct rcu_head	rcu;
1535 	};
1536 };
1537 
1538 struct bpf_prog {
1539 	u16			pages;		/* Number of allocated pages */
1540 	u16			jited:1,	/* Is our filter JIT'ed? */
1541 				jit_requested:1,/* archs need to JIT the prog */
1542 				gpl_compatible:1, /* Is filter GPL compatible? */
1543 				cb_access:1,	/* Is control block accessed? */
1544 				dst_needed:1,	/* Do we need dst entry? */
1545 				blinding_requested:1, /* needs constant blinding */
1546 				blinded:1,	/* Was blinded */
1547 				is_func:1,	/* program is a bpf function */
1548 				kprobe_override:1, /* Do we override a kprobe? */
1549 				has_callchain_buf:1, /* callchain buffer allocated? */
1550 				enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */
1551 				call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */
1552 				call_get_func_ip:1, /* Do we call get_func_ip() */
1553 				tstamp_type_access:1, /* Accessed __sk_buff->tstamp_type */
1554 				sleepable:1;	/* BPF program is sleepable */
1555 	enum bpf_prog_type	type;		/* Type of BPF program */
1556 	enum bpf_attach_type	expected_attach_type; /* For some prog types */
1557 	u32			len;		/* Number of filter blocks */
1558 	u32			jited_len;	/* Size of jited insns in bytes */
1559 	u8			tag[BPF_TAG_SIZE];
1560 	struct bpf_prog_stats __percpu *stats;
1561 	int __percpu		*active;
1562 	unsigned int		(*bpf_func)(const void *ctx,
1563 					    const struct bpf_insn *insn);
1564 	struct bpf_prog_aux	*aux;		/* Auxiliary fields */
1565 	struct sock_fprog_kern	*orig_prog;	/* Original BPF program */
1566 	/* Instructions for interpreter */
1567 	union {
1568 		DECLARE_FLEX_ARRAY(struct sock_filter, insns);
1569 		DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi);
1570 	};
1571 };
1572 
1573 struct bpf_array_aux {
1574 	/* Programs with direct jumps into programs part of this array. */
1575 	struct list_head poke_progs;
1576 	struct bpf_map *map;
1577 	struct mutex poke_mutex;
1578 	struct work_struct work;
1579 };
1580 
1581 struct bpf_link {
1582 	atomic64_t refcnt;
1583 	u32 id;
1584 	enum bpf_link_type type;
1585 	const struct bpf_link_ops *ops;
1586 	struct bpf_prog *prog;
1587 	/* rcu is used before freeing, work can be used to schedule that
1588 	 * RCU-based freeing before that, so they never overlap
1589 	 */
1590 	union {
1591 		struct rcu_head rcu;
1592 		struct work_struct work;
1593 	};
1594 };
1595 
1596 struct bpf_link_ops {
1597 	void (*release)(struct bpf_link *link);
1598 	/* deallocate link resources callback, called without RCU grace period
1599 	 * waiting
1600 	 */
1601 	void (*dealloc)(struct bpf_link *link);
1602 	/* deallocate link resources callback, called after RCU grace period;
1603 	 * if underlying BPF program is sleepable we go through tasks trace
1604 	 * RCU GP and then "classic" RCU GP
1605 	 */
1606 	void (*dealloc_deferred)(struct bpf_link *link);
1607 	int (*detach)(struct bpf_link *link);
1608 	int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
1609 			   struct bpf_prog *old_prog);
1610 	void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
1611 	int (*fill_link_info)(const struct bpf_link *link,
1612 			      struct bpf_link_info *info);
1613 	int (*update_map)(struct bpf_link *link, struct bpf_map *new_map,
1614 			  struct bpf_map *old_map);
1615 };
1616 
1617 struct bpf_tramp_link {
1618 	struct bpf_link link;
1619 	struct hlist_node tramp_hlist;
1620 	u64 cookie;
1621 };
1622 
1623 struct bpf_shim_tramp_link {
1624 	struct bpf_tramp_link link;
1625 	struct bpf_trampoline *trampoline;
1626 };
1627 
1628 struct bpf_tracing_link {
1629 	struct bpf_tramp_link link;
1630 	enum bpf_attach_type attach_type;
1631 	struct bpf_trampoline *trampoline;
1632 	struct bpf_prog *tgt_prog;
1633 };
1634 
1635 struct bpf_raw_tp_link {
1636 	struct bpf_link link;
1637 	struct bpf_raw_event_map *btp;
1638 	u64 cookie;
1639 };
1640 
1641 struct bpf_link_primer {
1642 	struct bpf_link *link;
1643 	struct file *file;
1644 	int fd;
1645 	u32 id;
1646 };
1647 
1648 struct bpf_mount_opts {
1649 	kuid_t uid;
1650 	kgid_t gid;
1651 	umode_t mode;
1652 
1653 	/* BPF token-related delegation options */
1654 	u64 delegate_cmds;
1655 	u64 delegate_maps;
1656 	u64 delegate_progs;
1657 	u64 delegate_attachs;
1658 };
1659 
1660 struct bpf_token {
1661 	struct work_struct work;
1662 	atomic64_t refcnt;
1663 	struct user_namespace *userns;
1664 	u64 allowed_cmds;
1665 	u64 allowed_maps;
1666 	u64 allowed_progs;
1667 	u64 allowed_attachs;
1668 #ifdef CONFIG_SECURITY
1669 	void *security;
1670 #endif
1671 };
1672 
1673 struct bpf_struct_ops_value;
1674 struct btf_member;
1675 
1676 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
1677 /**
1678  * struct bpf_struct_ops - A structure of callbacks allowing a subsystem to
1679  *			   define a BPF_MAP_TYPE_STRUCT_OPS map type composed
1680  *			   of BPF_PROG_TYPE_STRUCT_OPS progs.
1681  * @verifier_ops: A structure of callbacks that are invoked by the verifier
1682  *		  when determining whether the struct_ops progs in the
1683  *		  struct_ops map are valid.
1684  * @init: A callback that is invoked a single time, and before any other
1685  *	  callback, to initialize the structure. A nonzero return value means
1686  *	  the subsystem could not be initialized.
1687  * @check_member: When defined, a callback invoked by the verifier to allow
1688  *		  the subsystem to determine if an entry in the struct_ops map
1689  *		  is valid. A nonzero return value means that the map is
1690  *		  invalid and should be rejected by the verifier.
1691  * @init_member: A callback that is invoked for each member of the struct_ops
1692  *		 map to allow the subsystem to initialize the member. A nonzero
1693  *		 value means the member could not be initialized. This callback
1694  *		 is exclusive with the @type, @type_id, @value_type, and
1695  *		 @value_id fields.
1696  * @reg: A callback that is invoked when the struct_ops map has been
1697  *	 initialized and is being attached to. Zero means the struct_ops map
1698  *	 has been successfully registered and is live. A nonzero return value
1699  *	 means the struct_ops map could not be registered.
1700  * @unreg: A callback that is invoked when the struct_ops map should be
1701  *	   unregistered.
1702  * @update: A callback that is invoked when the live struct_ops map is being
1703  *	    updated to contain new values. This callback is only invoked when
1704  *	    the struct_ops map is loaded with BPF_F_LINK. If not defined, the
1705  *	    it is assumed that the struct_ops map cannot be updated.
1706  * @validate: A callback that is invoked after all of the members have been
1707  *	      initialized. This callback should perform static checks on the
1708  *	      map, meaning that it should either fail or succeed
1709  *	      deterministically. A struct_ops map that has been validated may
1710  *	      not necessarily succeed in being registered if the call to @reg
1711  *	      fails. For example, a valid struct_ops map may be loaded, but
1712  *	      then fail to be registered due to there being another active
1713  *	      struct_ops map on the system in the subsystem already. For this
1714  *	      reason, if this callback is not defined, the check is skipped as
1715  *	      the struct_ops map will have final verification performed in
1716  *	      @reg.
1717  * @type: BTF type.
1718  * @value_type: Value type.
1719  * @name: The name of the struct bpf_struct_ops object.
1720  * @func_models: Func models
1721  * @type_id: BTF type id.
1722  * @value_id: BTF value id.
1723  */
1724 struct bpf_struct_ops {
1725 	const struct bpf_verifier_ops *verifier_ops;
1726 	int (*init)(struct btf *btf);
1727 	int (*check_member)(const struct btf_type *t,
1728 			    const struct btf_member *member,
1729 			    const struct bpf_prog *prog);
1730 	int (*init_member)(const struct btf_type *t,
1731 			   const struct btf_member *member,
1732 			   void *kdata, const void *udata);
1733 	int (*reg)(void *kdata);
1734 	void (*unreg)(void *kdata);
1735 	int (*update)(void *kdata, void *old_kdata);
1736 	int (*validate)(void *kdata);
1737 	void *cfi_stubs;
1738 	struct module *owner;
1739 	const char *name;
1740 	struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
1741 };
1742 
1743 /* Every member of a struct_ops type has an instance even a member is not
1744  * an operator (function pointer). The "info" field will be assigned to
1745  * prog->aux->ctx_arg_info of BPF struct_ops programs to provide the
1746  * argument information required by the verifier to verify the program.
1747  *
1748  * btf_ctx_access() will lookup prog->aux->ctx_arg_info to find the
1749  * corresponding entry for an given argument.
1750  */
1751 struct bpf_struct_ops_arg_info {
1752 	struct bpf_ctx_arg_aux *info;
1753 	u32 cnt;
1754 };
1755 
1756 struct bpf_struct_ops_desc {
1757 	struct bpf_struct_ops *st_ops;
1758 
1759 	const struct btf_type *type;
1760 	const struct btf_type *value_type;
1761 	u32 type_id;
1762 	u32 value_id;
1763 
1764 	/* Collection of argument information for each member */
1765 	struct bpf_struct_ops_arg_info *arg_info;
1766 };
1767 
1768 enum bpf_struct_ops_state {
1769 	BPF_STRUCT_OPS_STATE_INIT,
1770 	BPF_STRUCT_OPS_STATE_INUSE,
1771 	BPF_STRUCT_OPS_STATE_TOBEFREE,
1772 	BPF_STRUCT_OPS_STATE_READY,
1773 };
1774 
1775 struct bpf_struct_ops_common_value {
1776 	refcount_t refcnt;
1777 	enum bpf_struct_ops_state state;
1778 };
1779 
1780 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
1781 /* This macro helps developer to register a struct_ops type and generate
1782  * type information correctly. Developers should use this macro to register
1783  * a struct_ops type instead of calling __register_bpf_struct_ops() directly.
1784  */
1785 #define register_bpf_struct_ops(st_ops, type)				\
1786 	({								\
1787 		struct bpf_struct_ops_##type {				\
1788 			struct bpf_struct_ops_common_value common;	\
1789 			struct type data ____cacheline_aligned_in_smp;	\
1790 		};							\
1791 		BTF_TYPE_EMIT(struct bpf_struct_ops_##type);		\
1792 		__register_bpf_struct_ops(st_ops);			\
1793 	})
1794 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
1795 bool bpf_struct_ops_get(const void *kdata);
1796 void bpf_struct_ops_put(const void *kdata);
1797 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
1798 				       void *value);
1799 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks,
1800 				      struct bpf_tramp_link *link,
1801 				      const struct btf_func_model *model,
1802 				      void *stub_func,
1803 				      void **image, u32 *image_off,
1804 				      bool allow_alloc);
1805 void bpf_struct_ops_image_free(void *image);
bpf_try_module_get(const void * data,struct module * owner)1806 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1807 {
1808 	if (owner == BPF_MODULE_OWNER)
1809 		return bpf_struct_ops_get(data);
1810 	else
1811 		return try_module_get(owner);
1812 }
bpf_module_put(const void * data,struct module * owner)1813 static inline void bpf_module_put(const void *data, struct module *owner)
1814 {
1815 	if (owner == BPF_MODULE_OWNER)
1816 		bpf_struct_ops_put(data);
1817 	else
1818 		module_put(owner);
1819 }
1820 int bpf_struct_ops_link_create(union bpf_attr *attr);
1821 
1822 #ifdef CONFIG_NET
1823 /* Define it here to avoid the use of forward declaration */
1824 struct bpf_dummy_ops_state {
1825 	int val;
1826 };
1827 
1828 struct bpf_dummy_ops {
1829 	int (*test_1)(struct bpf_dummy_ops_state *cb);
1830 	int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2,
1831 		      char a3, unsigned long a4);
1832 	int (*test_sleepable)(struct bpf_dummy_ops_state *cb);
1833 };
1834 
1835 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr,
1836 			    union bpf_attr __user *uattr);
1837 #endif
1838 int bpf_struct_ops_desc_init(struct bpf_struct_ops_desc *st_ops_desc,
1839 			     struct btf *btf,
1840 			     struct bpf_verifier_log *log);
1841 void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map);
1842 void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc);
1843 #else
1844 #define register_bpf_struct_ops(st_ops, type) ({ (void *)(st_ops); 0; })
bpf_try_module_get(const void * data,struct module * owner)1845 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1846 {
1847 	return try_module_get(owner);
1848 }
bpf_module_put(const void * data,struct module * owner)1849 static inline void bpf_module_put(const void *data, struct module *owner)
1850 {
1851 	module_put(owner);
1852 }
bpf_struct_ops_map_sys_lookup_elem(struct bpf_map * map,void * key,void * value)1853 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
1854 						     void *key,
1855 						     void *value)
1856 {
1857 	return -EINVAL;
1858 }
bpf_struct_ops_link_create(union bpf_attr * attr)1859 static inline int bpf_struct_ops_link_create(union bpf_attr *attr)
1860 {
1861 	return -EOPNOTSUPP;
1862 }
bpf_map_struct_ops_info_fill(struct bpf_map_info * info,struct bpf_map * map)1863 static inline void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map)
1864 {
1865 }
1866 
bpf_struct_ops_desc_release(struct bpf_struct_ops_desc * st_ops_desc)1867 static inline void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc)
1868 {
1869 }
1870 
1871 #endif
1872 
1873 #if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM)
1874 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1875 				    int cgroup_atype);
1876 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog);
1877 #else
bpf_trampoline_link_cgroup_shim(struct bpf_prog * prog,int cgroup_atype)1878 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1879 						  int cgroup_atype)
1880 {
1881 	return -EOPNOTSUPP;
1882 }
bpf_trampoline_unlink_cgroup_shim(struct bpf_prog * prog)1883 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog)
1884 {
1885 }
1886 #endif
1887 
1888 struct bpf_array {
1889 	struct bpf_map map;
1890 	u32 elem_size;
1891 	u32 index_mask;
1892 	struct bpf_array_aux *aux;
1893 	union {
1894 		DECLARE_FLEX_ARRAY(char, value) __aligned(8);
1895 		DECLARE_FLEX_ARRAY(void *, ptrs) __aligned(8);
1896 		DECLARE_FLEX_ARRAY(void __percpu *, pptrs) __aligned(8);
1897 	};
1898 };
1899 
1900 #define BPF_COMPLEXITY_LIMIT_INSNS      1000000 /* yes. 1M insns */
1901 #define MAX_TAIL_CALL_CNT 33
1902 
1903 /* Maximum number of loops for bpf_loop and bpf_iter_num.
1904  * It's enum to expose it (and thus make it discoverable) through BTF.
1905  */
1906 enum {
1907 	BPF_MAX_LOOPS = 8 * 1024 * 1024,
1908 };
1909 
1910 #define BPF_F_ACCESS_MASK	(BPF_F_RDONLY |		\
1911 				 BPF_F_RDONLY_PROG |	\
1912 				 BPF_F_WRONLY |		\
1913 				 BPF_F_WRONLY_PROG)
1914 
1915 #define BPF_MAP_CAN_READ	BIT(0)
1916 #define BPF_MAP_CAN_WRITE	BIT(1)
1917 
1918 /* Maximum number of user-producer ring buffer samples that can be drained in
1919  * a call to bpf_user_ringbuf_drain().
1920  */
1921 #define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024)
1922 
bpf_map_flags_to_cap(struct bpf_map * map)1923 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
1924 {
1925 	u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1926 
1927 	/* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
1928 	 * not possible.
1929 	 */
1930 	if (access_flags & BPF_F_RDONLY_PROG)
1931 		return BPF_MAP_CAN_READ;
1932 	else if (access_flags & BPF_F_WRONLY_PROG)
1933 		return BPF_MAP_CAN_WRITE;
1934 	else
1935 		return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
1936 }
1937 
bpf_map_flags_access_ok(u32 access_flags)1938 static inline bool bpf_map_flags_access_ok(u32 access_flags)
1939 {
1940 	return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
1941 	       (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1942 }
1943 
1944 struct bpf_event_entry {
1945 	struct perf_event *event;
1946 	struct file *perf_file;
1947 	struct file *map_file;
1948 	struct rcu_head rcu;
1949 };
1950 
map_type_contains_progs(struct bpf_map * map)1951 static inline bool map_type_contains_progs(struct bpf_map *map)
1952 {
1953 	return map->map_type == BPF_MAP_TYPE_PROG_ARRAY ||
1954 	       map->map_type == BPF_MAP_TYPE_DEVMAP ||
1955 	       map->map_type == BPF_MAP_TYPE_CPUMAP;
1956 }
1957 
1958 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp);
1959 int bpf_prog_calc_tag(struct bpf_prog *fp);
1960 
1961 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
1962 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void);
1963 
1964 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
1965 					unsigned long off, unsigned long len);
1966 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
1967 					const struct bpf_insn *src,
1968 					struct bpf_insn *dst,
1969 					struct bpf_prog *prog,
1970 					u32 *target_size);
1971 
1972 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
1973 		     void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
1974 
1975 /* an array of programs to be executed under rcu_lock.
1976  *
1977  * Typical usage:
1978  * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run);
1979  *
1980  * the structure returned by bpf_prog_array_alloc() should be populated
1981  * with program pointers and the last pointer must be NULL.
1982  * The user has to keep refcnt on the program and make sure the program
1983  * is removed from the array before bpf_prog_put().
1984  * The 'struct bpf_prog_array *' should only be replaced with xchg()
1985  * since other cpus are walking the array of pointers in parallel.
1986  */
1987 struct bpf_prog_array_item {
1988 	struct bpf_prog *prog;
1989 	union {
1990 		struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1991 		u64 bpf_cookie;
1992 	};
1993 };
1994 
1995 struct bpf_prog_array {
1996 	struct rcu_head rcu;
1997 	struct bpf_prog_array_item items[];
1998 };
1999 
2000 struct bpf_empty_prog_array {
2001 	struct bpf_prog_array hdr;
2002 	struct bpf_prog *null_prog;
2003 };
2004 
2005 /* to avoid allocating empty bpf_prog_array for cgroups that
2006  * don't have bpf program attached use one global 'bpf_empty_prog_array'
2007  * It will not be modified the caller of bpf_prog_array_alloc()
2008  * (since caller requested prog_cnt == 0)
2009  * that pointer should be 'freed' by bpf_prog_array_free()
2010  */
2011 extern struct bpf_empty_prog_array bpf_empty_prog_array;
2012 
2013 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
2014 void bpf_prog_array_free(struct bpf_prog_array *progs);
2015 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */
2016 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs);
2017 int bpf_prog_array_length(struct bpf_prog_array *progs);
2018 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
2019 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
2020 				__u32 __user *prog_ids, u32 cnt);
2021 
2022 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
2023 				struct bpf_prog *old_prog);
2024 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
2025 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
2026 			     struct bpf_prog *prog);
2027 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
2028 			     u32 *prog_ids, u32 request_cnt,
2029 			     u32 *prog_cnt);
2030 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
2031 			struct bpf_prog *exclude_prog,
2032 			struct bpf_prog *include_prog,
2033 			u64 bpf_cookie,
2034 			struct bpf_prog_array **new_array);
2035 
2036 struct bpf_run_ctx {};
2037 
2038 struct bpf_cg_run_ctx {
2039 	struct bpf_run_ctx run_ctx;
2040 	const struct bpf_prog_array_item *prog_item;
2041 	int retval;
2042 };
2043 
2044 struct bpf_trace_run_ctx {
2045 	struct bpf_run_ctx run_ctx;
2046 	u64 bpf_cookie;
2047 	bool is_uprobe;
2048 };
2049 
2050 struct bpf_tramp_run_ctx {
2051 	struct bpf_run_ctx run_ctx;
2052 	u64 bpf_cookie;
2053 	struct bpf_run_ctx *saved_run_ctx;
2054 };
2055 
bpf_set_run_ctx(struct bpf_run_ctx * new_ctx)2056 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)
2057 {
2058 	struct bpf_run_ctx *old_ctx = NULL;
2059 
2060 #ifdef CONFIG_BPF_SYSCALL
2061 	old_ctx = current->bpf_ctx;
2062 	current->bpf_ctx = new_ctx;
2063 #endif
2064 	return old_ctx;
2065 }
2066 
bpf_reset_run_ctx(struct bpf_run_ctx * old_ctx)2067 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
2068 {
2069 #ifdef CONFIG_BPF_SYSCALL
2070 	current->bpf_ctx = old_ctx;
2071 #endif
2072 }
2073 
2074 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
2075 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE			(1 << 0)
2076 /* BPF program asks to set CN on the packet. */
2077 #define BPF_RET_SET_CN						(1 << 0)
2078 
2079 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx);
2080 
2081 static __always_inline u32
bpf_prog_run_array(const struct bpf_prog_array * array,const void * ctx,bpf_prog_run_fn run_prog)2082 bpf_prog_run_array(const struct bpf_prog_array *array,
2083 		   const void *ctx, bpf_prog_run_fn run_prog)
2084 {
2085 	const struct bpf_prog_array_item *item;
2086 	const struct bpf_prog *prog;
2087 	struct bpf_run_ctx *old_run_ctx;
2088 	struct bpf_trace_run_ctx run_ctx;
2089 	u32 ret = 1;
2090 
2091 	RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held");
2092 
2093 	if (unlikely(!array))
2094 		return ret;
2095 
2096 	run_ctx.is_uprobe = false;
2097 
2098 	migrate_disable();
2099 	old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
2100 	item = &array->items[0];
2101 	while ((prog = READ_ONCE(item->prog))) {
2102 		run_ctx.bpf_cookie = item->bpf_cookie;
2103 		ret &= run_prog(prog, ctx);
2104 		item++;
2105 	}
2106 	bpf_reset_run_ctx(old_run_ctx);
2107 	migrate_enable();
2108 	return ret;
2109 }
2110 
2111 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs:
2112  *
2113  * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array
2114  * overall. As a result, we must use the bpf_prog_array_free_sleepable
2115  * in order to use the tasks_trace rcu grace period.
2116  *
2117  * When a non-sleepable program is inside the array, we take the rcu read
2118  * section and disable preemption for that program alone, so it can access
2119  * rcu-protected dynamically sized maps.
2120  */
2121 static __always_inline u32
bpf_prog_run_array_uprobe(const struct bpf_prog_array __rcu * array_rcu,const void * ctx,bpf_prog_run_fn run_prog)2122 bpf_prog_run_array_uprobe(const struct bpf_prog_array __rcu *array_rcu,
2123 			  const void *ctx, bpf_prog_run_fn run_prog)
2124 {
2125 	const struct bpf_prog_array_item *item;
2126 	const struct bpf_prog *prog;
2127 	const struct bpf_prog_array *array;
2128 	struct bpf_run_ctx *old_run_ctx;
2129 	struct bpf_trace_run_ctx run_ctx;
2130 	u32 ret = 1;
2131 
2132 	might_fault();
2133 
2134 	rcu_read_lock_trace();
2135 	migrate_disable();
2136 
2137 	run_ctx.is_uprobe = true;
2138 
2139 	array = rcu_dereference_check(array_rcu, rcu_read_lock_trace_held());
2140 	if (unlikely(!array))
2141 		goto out;
2142 	old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
2143 	item = &array->items[0];
2144 	while ((prog = READ_ONCE(item->prog))) {
2145 		if (!prog->sleepable)
2146 			rcu_read_lock();
2147 
2148 		run_ctx.bpf_cookie = item->bpf_cookie;
2149 		ret &= run_prog(prog, ctx);
2150 		item++;
2151 
2152 		if (!prog->sleepable)
2153 			rcu_read_unlock();
2154 	}
2155 	bpf_reset_run_ctx(old_run_ctx);
2156 out:
2157 	migrate_enable();
2158 	rcu_read_unlock_trace();
2159 	return ret;
2160 }
2161 
2162 #ifdef CONFIG_BPF_SYSCALL
2163 DECLARE_PER_CPU(int, bpf_prog_active);
2164 extern struct mutex bpf_stats_enabled_mutex;
2165 
2166 /*
2167  * Block execution of BPF programs attached to instrumentation (perf,
2168  * kprobes, tracepoints) to prevent deadlocks on map operations as any of
2169  * these events can happen inside a region which holds a map bucket lock
2170  * and can deadlock on it.
2171  */
bpf_disable_instrumentation(void)2172 static inline void bpf_disable_instrumentation(void)
2173 {
2174 	migrate_disable();
2175 	this_cpu_inc(bpf_prog_active);
2176 }
2177 
bpf_enable_instrumentation(void)2178 static inline void bpf_enable_instrumentation(void)
2179 {
2180 	this_cpu_dec(bpf_prog_active);
2181 	migrate_enable();
2182 }
2183 
2184 extern const struct super_operations bpf_super_ops;
2185 extern const struct file_operations bpf_map_fops;
2186 extern const struct file_operations bpf_prog_fops;
2187 extern const struct file_operations bpf_iter_fops;
2188 
2189 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
2190 	extern const struct bpf_prog_ops _name ## _prog_ops; \
2191 	extern const struct bpf_verifier_ops _name ## _verifier_ops;
2192 #define BPF_MAP_TYPE(_id, _ops) \
2193 	extern const struct bpf_map_ops _ops;
2194 #define BPF_LINK_TYPE(_id, _name)
2195 #include <linux/bpf_types.h>
2196 #undef BPF_PROG_TYPE
2197 #undef BPF_MAP_TYPE
2198 #undef BPF_LINK_TYPE
2199 
2200 extern const struct bpf_prog_ops bpf_offload_prog_ops;
2201 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
2202 extern const struct bpf_verifier_ops xdp_analyzer_ops;
2203 
2204 struct bpf_prog *bpf_prog_get(u32 ufd);
2205 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
2206 				       bool attach_drv);
2207 void bpf_prog_add(struct bpf_prog *prog, int i);
2208 void bpf_prog_sub(struct bpf_prog *prog, int i);
2209 void bpf_prog_inc(struct bpf_prog *prog);
2210 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
2211 void bpf_prog_put(struct bpf_prog *prog);
2212 
2213 void bpf_prog_free_id(struct bpf_prog *prog);
2214 void bpf_map_free_id(struct bpf_map *map);
2215 
2216 struct btf_field *btf_record_find(const struct btf_record *rec,
2217 				  u32 offset, u32 field_mask);
2218 void btf_record_free(struct btf_record *rec);
2219 void bpf_map_free_record(struct bpf_map *map);
2220 struct btf_record *btf_record_dup(const struct btf_record *rec);
2221 bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b);
2222 void bpf_obj_free_timer(const struct btf_record *rec, void *obj);
2223 void bpf_obj_free_workqueue(const struct btf_record *rec, void *obj);
2224 void bpf_obj_free_fields(const struct btf_record *rec, void *obj);
2225 void __bpf_obj_drop_impl(void *p, const struct btf_record *rec, bool percpu);
2226 
2227 struct bpf_map *bpf_map_get(u32 ufd);
2228 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
2229 struct bpf_map *__bpf_map_get(struct fd f);
2230 void bpf_map_inc(struct bpf_map *map);
2231 void bpf_map_inc_with_uref(struct bpf_map *map);
2232 struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref);
2233 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
2234 void bpf_map_put_with_uref(struct bpf_map *map);
2235 void bpf_map_put(struct bpf_map *map);
2236 void *bpf_map_area_alloc(u64 size, int numa_node);
2237 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
2238 void bpf_map_area_free(void *base);
2239 bool bpf_map_write_active(const struct bpf_map *map);
2240 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
2241 int  generic_map_lookup_batch(struct bpf_map *map,
2242 			      const union bpf_attr *attr,
2243 			      union bpf_attr __user *uattr);
2244 int  generic_map_update_batch(struct bpf_map *map, struct file *map_file,
2245 			      const union bpf_attr *attr,
2246 			      union bpf_attr __user *uattr);
2247 int  generic_map_delete_batch(struct bpf_map *map,
2248 			      const union bpf_attr *attr,
2249 			      union bpf_attr __user *uattr);
2250 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
2251 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
2252 
2253 int bpf_map_alloc_pages(const struct bpf_map *map, gfp_t gfp, int nid,
2254 			unsigned long nr_pages, struct page **page_array);
2255 #ifdef CONFIG_MEMCG_KMEM
2256 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
2257 			   int node);
2258 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags);
2259 void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size,
2260 		       gfp_t flags);
2261 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
2262 				    size_t align, gfp_t flags);
2263 #else
2264 static inline void *
bpf_map_kmalloc_node(const struct bpf_map * map,size_t size,gfp_t flags,int node)2265 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
2266 		     int node)
2267 {
2268 	return kmalloc_node(size, flags, node);
2269 }
2270 
2271 static inline void *
bpf_map_kzalloc(const struct bpf_map * map,size_t size,gfp_t flags)2272 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags)
2273 {
2274 	return kzalloc(size, flags);
2275 }
2276 
2277 static inline void *
bpf_map_kvcalloc(struct bpf_map * map,size_t n,size_t size,gfp_t flags)2278 bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size, gfp_t flags)
2279 {
2280 	return kvcalloc(n, size, flags);
2281 }
2282 
2283 static inline void __percpu *
bpf_map_alloc_percpu(const struct bpf_map * map,size_t size,size_t align,gfp_t flags)2284 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align,
2285 		     gfp_t flags)
2286 {
2287 	return __alloc_percpu_gfp(size, align, flags);
2288 }
2289 #endif
2290 
2291 static inline int
bpf_map_init_elem_count(struct bpf_map * map)2292 bpf_map_init_elem_count(struct bpf_map *map)
2293 {
2294 	size_t size = sizeof(*map->elem_count), align = size;
2295 	gfp_t flags = GFP_USER | __GFP_NOWARN;
2296 
2297 	map->elem_count = bpf_map_alloc_percpu(map, size, align, flags);
2298 	if (!map->elem_count)
2299 		return -ENOMEM;
2300 
2301 	return 0;
2302 }
2303 
2304 static inline void
bpf_map_free_elem_count(struct bpf_map * map)2305 bpf_map_free_elem_count(struct bpf_map *map)
2306 {
2307 	free_percpu(map->elem_count);
2308 }
2309 
bpf_map_inc_elem_count(struct bpf_map * map)2310 static inline void bpf_map_inc_elem_count(struct bpf_map *map)
2311 {
2312 	this_cpu_inc(*map->elem_count);
2313 }
2314 
bpf_map_dec_elem_count(struct bpf_map * map)2315 static inline void bpf_map_dec_elem_count(struct bpf_map *map)
2316 {
2317 	this_cpu_dec(*map->elem_count);
2318 }
2319 
2320 extern int sysctl_unprivileged_bpf_disabled;
2321 
2322 bool bpf_token_capable(const struct bpf_token *token, int cap);
2323 
bpf_allow_ptr_leaks(const struct bpf_token * token)2324 static inline bool bpf_allow_ptr_leaks(const struct bpf_token *token)
2325 {
2326 	return bpf_token_capable(token, CAP_PERFMON);
2327 }
2328 
bpf_allow_uninit_stack(const struct bpf_token * token)2329 static inline bool bpf_allow_uninit_stack(const struct bpf_token *token)
2330 {
2331 	return bpf_token_capable(token, CAP_PERFMON);
2332 }
2333 
bpf_bypass_spec_v1(const struct bpf_token * token)2334 static inline bool bpf_bypass_spec_v1(const struct bpf_token *token)
2335 {
2336 	return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON);
2337 }
2338 
bpf_bypass_spec_v4(const struct bpf_token * token)2339 static inline bool bpf_bypass_spec_v4(const struct bpf_token *token)
2340 {
2341 	return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON);
2342 }
2343 
2344 int bpf_map_new_fd(struct bpf_map *map, int flags);
2345 int bpf_prog_new_fd(struct bpf_prog *prog);
2346 
2347 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2348 		   const struct bpf_link_ops *ops, struct bpf_prog *prog);
2349 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
2350 int bpf_link_settle(struct bpf_link_primer *primer);
2351 void bpf_link_cleanup(struct bpf_link_primer *primer);
2352 void bpf_link_inc(struct bpf_link *link);
2353 void bpf_link_put(struct bpf_link *link);
2354 int bpf_link_new_fd(struct bpf_link *link);
2355 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
2356 struct bpf_link *bpf_link_get_curr_or_next(u32 *id);
2357 
2358 void bpf_token_inc(struct bpf_token *token);
2359 void bpf_token_put(struct bpf_token *token);
2360 int bpf_token_create(union bpf_attr *attr);
2361 struct bpf_token *bpf_token_get_from_fd(u32 ufd);
2362 
2363 bool bpf_token_allow_cmd(const struct bpf_token *token, enum bpf_cmd cmd);
2364 bool bpf_token_allow_map_type(const struct bpf_token *token, enum bpf_map_type type);
2365 bool bpf_token_allow_prog_type(const struct bpf_token *token,
2366 			       enum bpf_prog_type prog_type,
2367 			       enum bpf_attach_type attach_type);
2368 
2369 int bpf_obj_pin_user(u32 ufd, int path_fd, const char __user *pathname);
2370 int bpf_obj_get_user(int path_fd, const char __user *pathname, int flags);
2371 struct inode *bpf_get_inode(struct super_block *sb, const struct inode *dir,
2372 			    umode_t mode);
2373 
2374 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
2375 #define DEFINE_BPF_ITER_FUNC(target, args...)			\
2376 	extern int bpf_iter_ ## target(args);			\
2377 	int __init bpf_iter_ ## target(args) { return 0; }
2378 
2379 /*
2380  * The task type of iterators.
2381  *
2382  * For BPF task iterators, they can be parameterized with various
2383  * parameters to visit only some of tasks.
2384  *
2385  * BPF_TASK_ITER_ALL (default)
2386  *	Iterate over resources of every task.
2387  *
2388  * BPF_TASK_ITER_TID
2389  *	Iterate over resources of a task/tid.
2390  *
2391  * BPF_TASK_ITER_TGID
2392  *	Iterate over resources of every task of a process / task group.
2393  */
2394 enum bpf_iter_task_type {
2395 	BPF_TASK_ITER_ALL = 0,
2396 	BPF_TASK_ITER_TID,
2397 	BPF_TASK_ITER_TGID,
2398 };
2399 
2400 struct bpf_iter_aux_info {
2401 	/* for map_elem iter */
2402 	struct bpf_map *map;
2403 
2404 	/* for cgroup iter */
2405 	struct {
2406 		struct cgroup *start; /* starting cgroup */
2407 		enum bpf_cgroup_iter_order order;
2408 	} cgroup;
2409 	struct {
2410 		enum bpf_iter_task_type	type;
2411 		u32 pid;
2412 	} task;
2413 };
2414 
2415 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
2416 					union bpf_iter_link_info *linfo,
2417 					struct bpf_iter_aux_info *aux);
2418 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
2419 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
2420 					struct seq_file *seq);
2421 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
2422 					 struct bpf_link_info *info);
2423 typedef const struct bpf_func_proto *
2424 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id,
2425 			     const struct bpf_prog *prog);
2426 
2427 enum bpf_iter_feature {
2428 	BPF_ITER_RESCHED	= BIT(0),
2429 };
2430 
2431 #define BPF_ITER_CTX_ARG_MAX 2
2432 struct bpf_iter_reg {
2433 	const char *target;
2434 	bpf_iter_attach_target_t attach_target;
2435 	bpf_iter_detach_target_t detach_target;
2436 	bpf_iter_show_fdinfo_t show_fdinfo;
2437 	bpf_iter_fill_link_info_t fill_link_info;
2438 	bpf_iter_get_func_proto_t get_func_proto;
2439 	u32 ctx_arg_info_size;
2440 	u32 feature;
2441 	struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
2442 	const struct bpf_iter_seq_info *seq_info;
2443 };
2444 
2445 struct bpf_iter_meta {
2446 	__bpf_md_ptr(struct seq_file *, seq);
2447 	u64 session_id;
2448 	u64 seq_num;
2449 };
2450 
2451 struct bpf_iter__bpf_map_elem {
2452 	__bpf_md_ptr(struct bpf_iter_meta *, meta);
2453 	__bpf_md_ptr(struct bpf_map *, map);
2454 	__bpf_md_ptr(void *, key);
2455 	__bpf_md_ptr(void *, value);
2456 };
2457 
2458 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
2459 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
2460 bool bpf_iter_prog_supported(struct bpf_prog *prog);
2461 const struct bpf_func_proto *
2462 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
2463 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog);
2464 int bpf_iter_new_fd(struct bpf_link *link);
2465 bool bpf_link_is_iter(struct bpf_link *link);
2466 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
2467 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
2468 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
2469 			      struct seq_file *seq);
2470 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
2471 				struct bpf_link_info *info);
2472 
2473 int map_set_for_each_callback_args(struct bpf_verifier_env *env,
2474 				   struct bpf_func_state *caller,
2475 				   struct bpf_func_state *callee);
2476 
2477 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
2478 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
2479 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
2480 			   u64 flags);
2481 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
2482 			    u64 flags);
2483 
2484 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
2485 
2486 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
2487 				 void *key, void *value, u64 map_flags);
2488 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2489 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
2490 				void *key, void *value, u64 map_flags);
2491 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2492 
2493 int bpf_get_file_flag(int flags);
2494 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size,
2495 			     size_t actual_size);
2496 
2497 /* verify correctness of eBPF program */
2498 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size);
2499 
2500 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
2501 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
2502 #endif
2503 
2504 struct btf *bpf_get_btf_vmlinux(void);
2505 
2506 /* Map specifics */
2507 struct xdp_frame;
2508 struct sk_buff;
2509 struct bpf_dtab_netdev;
2510 struct bpf_cpu_map_entry;
2511 
2512 void __dev_flush(void);
2513 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2514 		    struct net_device *dev_rx);
2515 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2516 		    struct net_device *dev_rx);
2517 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2518 			  struct bpf_map *map, bool exclude_ingress);
2519 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
2520 			     struct bpf_prog *xdp_prog);
2521 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2522 			   struct bpf_prog *xdp_prog, struct bpf_map *map,
2523 			   bool exclude_ingress);
2524 
2525 void __cpu_map_flush(void);
2526 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf,
2527 		    struct net_device *dev_rx);
2528 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2529 			     struct sk_buff *skb);
2530 
2531 /* Return map's numa specified by userspace */
bpf_map_attr_numa_node(const union bpf_attr * attr)2532 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
2533 {
2534 	return (attr->map_flags & BPF_F_NUMA_NODE) ?
2535 		attr->numa_node : NUMA_NO_NODE;
2536 }
2537 
2538 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
2539 int array_map_alloc_check(union bpf_attr *attr);
2540 
2541 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
2542 			  union bpf_attr __user *uattr);
2543 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
2544 			  union bpf_attr __user *uattr);
2545 int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2546 			      const union bpf_attr *kattr,
2547 			      union bpf_attr __user *uattr);
2548 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2549 				     const union bpf_attr *kattr,
2550 				     union bpf_attr __user *uattr);
2551 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
2552 			     const union bpf_attr *kattr,
2553 			     union bpf_attr __user *uattr);
2554 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2555 				const union bpf_attr *kattr,
2556 				union bpf_attr __user *uattr);
2557 int bpf_prog_test_run_nf(struct bpf_prog *prog,
2558 			 const union bpf_attr *kattr,
2559 			 union bpf_attr __user *uattr);
2560 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
2561 		    const struct bpf_prog *prog,
2562 		    struct bpf_insn_access_aux *info);
2563 
bpf_tracing_ctx_access(int off,int size,enum bpf_access_type type)2564 static inline bool bpf_tracing_ctx_access(int off, int size,
2565 					  enum bpf_access_type type)
2566 {
2567 	if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
2568 		return false;
2569 	if (type != BPF_READ)
2570 		return false;
2571 	if (off % size != 0)
2572 		return false;
2573 	return true;
2574 }
2575 
bpf_tracing_btf_ctx_access(int off,int size,enum bpf_access_type type,const struct bpf_prog * prog,struct bpf_insn_access_aux * info)2576 static inline bool bpf_tracing_btf_ctx_access(int off, int size,
2577 					      enum bpf_access_type type,
2578 					      const struct bpf_prog *prog,
2579 					      struct bpf_insn_access_aux *info)
2580 {
2581 	if (!bpf_tracing_ctx_access(off, size, type))
2582 		return false;
2583 	return btf_ctx_access(off, size, type, prog, info);
2584 }
2585 
2586 int btf_struct_access(struct bpf_verifier_log *log,
2587 		      const struct bpf_reg_state *reg,
2588 		      int off, int size, enum bpf_access_type atype,
2589 		      u32 *next_btf_id, enum bpf_type_flag *flag, const char **field_name);
2590 bool btf_struct_ids_match(struct bpf_verifier_log *log,
2591 			  const struct btf *btf, u32 id, int off,
2592 			  const struct btf *need_btf, u32 need_type_id,
2593 			  bool strict);
2594 
2595 int btf_distill_func_proto(struct bpf_verifier_log *log,
2596 			   struct btf *btf,
2597 			   const struct btf_type *func_proto,
2598 			   const char *func_name,
2599 			   struct btf_func_model *m);
2600 
2601 struct bpf_reg_state;
2602 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog);
2603 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
2604 			 struct btf *btf, const struct btf_type *t);
2605 const char *btf_find_decl_tag_value(const struct btf *btf, const struct btf_type *pt,
2606 				    int comp_idx, const char *tag_key);
2607 int btf_find_next_decl_tag(const struct btf *btf, const struct btf_type *pt,
2608 			   int comp_idx, const char *tag_key, int last_id);
2609 
2610 struct bpf_prog *bpf_prog_by_id(u32 id);
2611 struct bpf_link *bpf_link_by_id(u32 id);
2612 
2613 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id,
2614 						 const struct bpf_prog *prog);
2615 void bpf_task_storage_free(struct task_struct *task);
2616 void bpf_cgrp_storage_free(struct cgroup *cgroup);
2617 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
2618 const struct btf_func_model *
2619 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2620 			 const struct bpf_insn *insn);
2621 int bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
2622 		       u16 btf_fd_idx, u8 **func_addr);
2623 
2624 struct bpf_core_ctx {
2625 	struct bpf_verifier_log *log;
2626 	const struct btf *btf;
2627 };
2628 
2629 bool btf_nested_type_is_trusted(struct bpf_verifier_log *log,
2630 				const struct bpf_reg_state *reg,
2631 				const char *field_name, u32 btf_id, const char *suffix);
2632 
2633 bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log,
2634 			       const struct btf *reg_btf, u32 reg_id,
2635 			       const struct btf *arg_btf, u32 arg_id);
2636 
2637 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo,
2638 		   int relo_idx, void *insn);
2639 
unprivileged_ebpf_enabled(void)2640 static inline bool unprivileged_ebpf_enabled(void)
2641 {
2642 	return !sysctl_unprivileged_bpf_disabled;
2643 }
2644 
2645 /* Not all bpf prog type has the bpf_ctx.
2646  * For the bpf prog type that has initialized the bpf_ctx,
2647  * this function can be used to decide if a kernel function
2648  * is called by a bpf program.
2649  */
has_current_bpf_ctx(void)2650 static inline bool has_current_bpf_ctx(void)
2651 {
2652 	return !!current->bpf_ctx;
2653 }
2654 
2655 void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog);
2656 
2657 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2658 		     enum bpf_dynptr_type type, u32 offset, u32 size);
2659 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr);
2660 void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr);
2661 
2662 bool dev_check_flush(void);
2663 bool cpu_map_check_flush(void);
2664 #else /* !CONFIG_BPF_SYSCALL */
bpf_prog_get(u32 ufd)2665 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
2666 {
2667 	return ERR_PTR(-EOPNOTSUPP);
2668 }
2669 
bpf_prog_get_type_dev(u32 ufd,enum bpf_prog_type type,bool attach_drv)2670 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
2671 						     enum bpf_prog_type type,
2672 						     bool attach_drv)
2673 {
2674 	return ERR_PTR(-EOPNOTSUPP);
2675 }
2676 
bpf_prog_add(struct bpf_prog * prog,int i)2677 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
2678 {
2679 }
2680 
bpf_prog_sub(struct bpf_prog * prog,int i)2681 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
2682 {
2683 }
2684 
bpf_prog_put(struct bpf_prog * prog)2685 static inline void bpf_prog_put(struct bpf_prog *prog)
2686 {
2687 }
2688 
bpf_prog_inc(struct bpf_prog * prog)2689 static inline void bpf_prog_inc(struct bpf_prog *prog)
2690 {
2691 }
2692 
2693 static inline struct bpf_prog *__must_check
bpf_prog_inc_not_zero(struct bpf_prog * prog)2694 bpf_prog_inc_not_zero(struct bpf_prog *prog)
2695 {
2696 	return ERR_PTR(-EOPNOTSUPP);
2697 }
2698 
bpf_link_init(struct bpf_link * link,enum bpf_link_type type,const struct bpf_link_ops * ops,struct bpf_prog * prog)2699 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2700 				 const struct bpf_link_ops *ops,
2701 				 struct bpf_prog *prog)
2702 {
2703 }
2704 
bpf_link_prime(struct bpf_link * link,struct bpf_link_primer * primer)2705 static inline int bpf_link_prime(struct bpf_link *link,
2706 				 struct bpf_link_primer *primer)
2707 {
2708 	return -EOPNOTSUPP;
2709 }
2710 
bpf_link_settle(struct bpf_link_primer * primer)2711 static inline int bpf_link_settle(struct bpf_link_primer *primer)
2712 {
2713 	return -EOPNOTSUPP;
2714 }
2715 
bpf_link_cleanup(struct bpf_link_primer * primer)2716 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
2717 {
2718 }
2719 
bpf_link_inc(struct bpf_link * link)2720 static inline void bpf_link_inc(struct bpf_link *link)
2721 {
2722 }
2723 
bpf_link_put(struct bpf_link * link)2724 static inline void bpf_link_put(struct bpf_link *link)
2725 {
2726 }
2727 
bpf_obj_get_user(const char __user * pathname,int flags)2728 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
2729 {
2730 	return -EOPNOTSUPP;
2731 }
2732 
bpf_token_capable(const struct bpf_token * token,int cap)2733 static inline bool bpf_token_capable(const struct bpf_token *token, int cap)
2734 {
2735 	return capable(cap) || (cap != CAP_SYS_ADMIN && capable(CAP_SYS_ADMIN));
2736 }
2737 
bpf_token_inc(struct bpf_token * token)2738 static inline void bpf_token_inc(struct bpf_token *token)
2739 {
2740 }
2741 
bpf_token_put(struct bpf_token * token)2742 static inline void bpf_token_put(struct bpf_token *token)
2743 {
2744 }
2745 
bpf_token_get_from_fd(u32 ufd)2746 static inline struct bpf_token *bpf_token_get_from_fd(u32 ufd)
2747 {
2748 	return ERR_PTR(-EOPNOTSUPP);
2749 }
2750 
__dev_flush(void)2751 static inline void __dev_flush(void)
2752 {
2753 }
2754 
2755 struct xdp_frame;
2756 struct bpf_dtab_netdev;
2757 struct bpf_cpu_map_entry;
2758 
2759 static inline
dev_xdp_enqueue(struct net_device * dev,struct xdp_frame * xdpf,struct net_device * dev_rx)2760 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2761 		    struct net_device *dev_rx)
2762 {
2763 	return 0;
2764 }
2765 
2766 static inline
dev_map_enqueue(struct bpf_dtab_netdev * dst,struct xdp_frame * xdpf,struct net_device * dev_rx)2767 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2768 		    struct net_device *dev_rx)
2769 {
2770 	return 0;
2771 }
2772 
2773 static inline
dev_map_enqueue_multi(struct xdp_frame * xdpf,struct net_device * dev_rx,struct bpf_map * map,bool exclude_ingress)2774 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2775 			  struct bpf_map *map, bool exclude_ingress)
2776 {
2777 	return 0;
2778 }
2779 
2780 struct sk_buff;
2781 
dev_map_generic_redirect(struct bpf_dtab_netdev * dst,struct sk_buff * skb,struct bpf_prog * xdp_prog)2782 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
2783 					   struct sk_buff *skb,
2784 					   struct bpf_prog *xdp_prog)
2785 {
2786 	return 0;
2787 }
2788 
2789 static inline
dev_map_redirect_multi(struct net_device * dev,struct sk_buff * skb,struct bpf_prog * xdp_prog,struct bpf_map * map,bool exclude_ingress)2790 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2791 			   struct bpf_prog *xdp_prog, struct bpf_map *map,
2792 			   bool exclude_ingress)
2793 {
2794 	return 0;
2795 }
2796 
__cpu_map_flush(void)2797 static inline void __cpu_map_flush(void)
2798 {
2799 }
2800 
cpu_map_enqueue(struct bpf_cpu_map_entry * rcpu,struct xdp_frame * xdpf,struct net_device * dev_rx)2801 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
2802 				  struct xdp_frame *xdpf,
2803 				  struct net_device *dev_rx)
2804 {
2805 	return 0;
2806 }
2807 
cpu_map_generic_redirect(struct bpf_cpu_map_entry * rcpu,struct sk_buff * skb)2808 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2809 					   struct sk_buff *skb)
2810 {
2811 	return -EOPNOTSUPP;
2812 }
2813 
bpf_prog_get_type_path(const char * name,enum bpf_prog_type type)2814 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
2815 				enum bpf_prog_type type)
2816 {
2817 	return ERR_PTR(-EOPNOTSUPP);
2818 }
2819 
bpf_prog_test_run_xdp(struct bpf_prog * prog,const union bpf_attr * kattr,union bpf_attr __user * uattr)2820 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
2821 					const union bpf_attr *kattr,
2822 					union bpf_attr __user *uattr)
2823 {
2824 	return -ENOTSUPP;
2825 }
2826 
bpf_prog_test_run_skb(struct bpf_prog * prog,const union bpf_attr * kattr,union bpf_attr __user * uattr)2827 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
2828 					const union bpf_attr *kattr,
2829 					union bpf_attr __user *uattr)
2830 {
2831 	return -ENOTSUPP;
2832 }
2833 
bpf_prog_test_run_tracing(struct bpf_prog * prog,const union bpf_attr * kattr,union bpf_attr __user * uattr)2834 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2835 					    const union bpf_attr *kattr,
2836 					    union bpf_attr __user *uattr)
2837 {
2838 	return -ENOTSUPP;
2839 }
2840 
bpf_prog_test_run_flow_dissector(struct bpf_prog * prog,const union bpf_attr * kattr,union bpf_attr __user * uattr)2841 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2842 						   const union bpf_attr *kattr,
2843 						   union bpf_attr __user *uattr)
2844 {
2845 	return -ENOTSUPP;
2846 }
2847 
bpf_prog_test_run_sk_lookup(struct bpf_prog * prog,const union bpf_attr * kattr,union bpf_attr __user * uattr)2848 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2849 					      const union bpf_attr *kattr,
2850 					      union bpf_attr __user *uattr)
2851 {
2852 	return -ENOTSUPP;
2853 }
2854 
bpf_map_put(struct bpf_map * map)2855 static inline void bpf_map_put(struct bpf_map *map)
2856 {
2857 }
2858 
bpf_prog_by_id(u32 id)2859 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
2860 {
2861 	return ERR_PTR(-ENOTSUPP);
2862 }
2863 
btf_struct_access(struct bpf_verifier_log * log,const struct bpf_reg_state * reg,int off,int size,enum bpf_access_type atype,u32 * next_btf_id,enum bpf_type_flag * flag,const char ** field_name)2864 static inline int btf_struct_access(struct bpf_verifier_log *log,
2865 				    const struct bpf_reg_state *reg,
2866 				    int off, int size, enum bpf_access_type atype,
2867 				    u32 *next_btf_id, enum bpf_type_flag *flag,
2868 				    const char **field_name)
2869 {
2870 	return -EACCES;
2871 }
2872 
2873 static inline const struct bpf_func_proto *
bpf_base_func_proto(enum bpf_func_id func_id,const struct bpf_prog * prog)2874 bpf_base_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
2875 {
2876 	return NULL;
2877 }
2878 
bpf_task_storage_free(struct task_struct * task)2879 static inline void bpf_task_storage_free(struct task_struct *task)
2880 {
2881 }
2882 
bpf_prog_has_kfunc_call(const struct bpf_prog * prog)2883 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog)
2884 {
2885 	return false;
2886 }
2887 
2888 static inline const struct btf_func_model *
bpf_jit_find_kfunc_model(const struct bpf_prog * prog,const struct bpf_insn * insn)2889 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2890 			 const struct bpf_insn *insn)
2891 {
2892 	return NULL;
2893 }
2894 
2895 static inline int
bpf_get_kfunc_addr(const struct bpf_prog * prog,u32 func_id,u16 btf_fd_idx,u8 ** func_addr)2896 bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
2897 		   u16 btf_fd_idx, u8 **func_addr)
2898 {
2899 	return -ENOTSUPP;
2900 }
2901 
unprivileged_ebpf_enabled(void)2902 static inline bool unprivileged_ebpf_enabled(void)
2903 {
2904 	return false;
2905 }
2906 
has_current_bpf_ctx(void)2907 static inline bool has_current_bpf_ctx(void)
2908 {
2909 	return false;
2910 }
2911 
bpf_prog_inc_misses_counter(struct bpf_prog * prog)2912 static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog)
2913 {
2914 }
2915 
bpf_cgrp_storage_free(struct cgroup * cgroup)2916 static inline void bpf_cgrp_storage_free(struct cgroup *cgroup)
2917 {
2918 }
2919 
bpf_dynptr_init(struct bpf_dynptr_kern * ptr,void * data,enum bpf_dynptr_type type,u32 offset,u32 size)2920 static inline void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2921 				   enum bpf_dynptr_type type, u32 offset, u32 size)
2922 {
2923 }
2924 
bpf_dynptr_set_null(struct bpf_dynptr_kern * ptr)2925 static inline void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr)
2926 {
2927 }
2928 
bpf_dynptr_set_rdonly(struct bpf_dynptr_kern * ptr)2929 static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr)
2930 {
2931 }
2932 #endif /* CONFIG_BPF_SYSCALL */
2933 
2934 static __always_inline int
bpf_probe_read_kernel_common(void * dst,u32 size,const void * unsafe_ptr)2935 bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
2936 {
2937 	int ret = -EFAULT;
2938 
2939 	if (IS_ENABLED(CONFIG_BPF_EVENTS))
2940 		ret = copy_from_kernel_nofault(dst, unsafe_ptr, size);
2941 	if (unlikely(ret < 0))
2942 		memset(dst, 0, size);
2943 	return ret;
2944 }
2945 
2946 void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
2947 			  struct btf_mod_pair *used_btfs, u32 len);
2948 
bpf_prog_get_type(u32 ufd,enum bpf_prog_type type)2949 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
2950 						 enum bpf_prog_type type)
2951 {
2952 	return bpf_prog_get_type_dev(ufd, type, false);
2953 }
2954 
2955 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
2956 			  struct bpf_map **used_maps, u32 len);
2957 
2958 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
2959 
2960 int bpf_prog_offload_compile(struct bpf_prog *prog);
2961 void bpf_prog_dev_bound_destroy(struct bpf_prog *prog);
2962 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
2963 			       struct bpf_prog *prog);
2964 
2965 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
2966 
2967 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
2968 int bpf_map_offload_update_elem(struct bpf_map *map,
2969 				void *key, void *value, u64 flags);
2970 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
2971 int bpf_map_offload_get_next_key(struct bpf_map *map,
2972 				 void *key, void *next_key);
2973 
2974 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
2975 
2976 struct bpf_offload_dev *
2977 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
2978 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
2979 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
2980 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
2981 				    struct net_device *netdev);
2982 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
2983 				       struct net_device *netdev);
2984 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
2985 
2986 void unpriv_ebpf_notify(int new_state);
2987 
2988 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
2989 int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
2990 			      struct bpf_prog_aux *prog_aux);
2991 void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, u32 func_id);
2992 int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr);
2993 int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, struct bpf_prog *old_prog);
2994 void bpf_dev_bound_netdev_unregister(struct net_device *dev);
2995 
bpf_prog_is_dev_bound(const struct bpf_prog_aux * aux)2996 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
2997 {
2998 	return aux->dev_bound;
2999 }
3000 
bpf_prog_is_offloaded(const struct bpf_prog_aux * aux)3001 static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux)
3002 {
3003 	return aux->offload_requested;
3004 }
3005 
3006 bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs);
3007 
bpf_map_is_offloaded(struct bpf_map * map)3008 static inline bool bpf_map_is_offloaded(struct bpf_map *map)
3009 {
3010 	return unlikely(map->ops == &bpf_map_offload_ops);
3011 }
3012 
3013 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
3014 void bpf_map_offload_map_free(struct bpf_map *map);
3015 u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map);
3016 int bpf_prog_test_run_syscall(struct bpf_prog *prog,
3017 			      const union bpf_attr *kattr,
3018 			      union bpf_attr __user *uattr);
3019 
3020 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
3021 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
3022 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
3023 int sock_map_bpf_prog_query(const union bpf_attr *attr,
3024 			    union bpf_attr __user *uattr);
3025 int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog);
3026 
3027 void sock_map_unhash(struct sock *sk);
3028 void sock_map_destroy(struct sock *sk);
3029 void sock_map_close(struct sock *sk, long timeout);
3030 #else
bpf_dev_bound_kfunc_check(struct bpf_verifier_log * log,struct bpf_prog_aux * prog_aux)3031 static inline int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
3032 					    struct bpf_prog_aux *prog_aux)
3033 {
3034 	return -EOPNOTSUPP;
3035 }
3036 
bpf_dev_bound_resolve_kfunc(struct bpf_prog * prog,u32 func_id)3037 static inline void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog,
3038 						u32 func_id)
3039 {
3040 	return NULL;
3041 }
3042 
bpf_prog_dev_bound_init(struct bpf_prog * prog,union bpf_attr * attr)3043 static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog,
3044 					  union bpf_attr *attr)
3045 {
3046 	return -EOPNOTSUPP;
3047 }
3048 
bpf_prog_dev_bound_inherit(struct bpf_prog * new_prog,struct bpf_prog * old_prog)3049 static inline int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog,
3050 					     struct bpf_prog *old_prog)
3051 {
3052 	return -EOPNOTSUPP;
3053 }
3054 
bpf_dev_bound_netdev_unregister(struct net_device * dev)3055 static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev)
3056 {
3057 }
3058 
bpf_prog_is_dev_bound(const struct bpf_prog_aux * aux)3059 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
3060 {
3061 	return false;
3062 }
3063 
bpf_prog_is_offloaded(struct bpf_prog_aux * aux)3064 static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux)
3065 {
3066 	return false;
3067 }
3068 
bpf_prog_dev_bound_match(const struct bpf_prog * lhs,const struct bpf_prog * rhs)3069 static inline bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs)
3070 {
3071 	return false;
3072 }
3073 
bpf_map_is_offloaded(struct bpf_map * map)3074 static inline bool bpf_map_is_offloaded(struct bpf_map *map)
3075 {
3076 	return false;
3077 }
3078 
bpf_map_offload_map_alloc(union bpf_attr * attr)3079 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
3080 {
3081 	return ERR_PTR(-EOPNOTSUPP);
3082 }
3083 
bpf_map_offload_map_free(struct bpf_map * map)3084 static inline void bpf_map_offload_map_free(struct bpf_map *map)
3085 {
3086 }
3087 
bpf_map_offload_map_mem_usage(const struct bpf_map * map)3088 static inline u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map)
3089 {
3090 	return 0;
3091 }
3092 
bpf_prog_test_run_syscall(struct bpf_prog * prog,const union bpf_attr * kattr,union bpf_attr __user * uattr)3093 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog,
3094 					    const union bpf_attr *kattr,
3095 					    union bpf_attr __user *uattr)
3096 {
3097 	return -ENOTSUPP;
3098 }
3099 
3100 #ifdef CONFIG_BPF_SYSCALL
sock_map_get_from_fd(const union bpf_attr * attr,struct bpf_prog * prog)3101 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
3102 				       struct bpf_prog *prog)
3103 {
3104 	return -EINVAL;
3105 }
3106 
sock_map_prog_detach(const union bpf_attr * attr,enum bpf_prog_type ptype)3107 static inline int sock_map_prog_detach(const union bpf_attr *attr,
3108 				       enum bpf_prog_type ptype)
3109 {
3110 	return -EOPNOTSUPP;
3111 }
3112 
sock_map_update_elem_sys(struct bpf_map * map,void * key,void * value,u64 flags)3113 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
3114 					   u64 flags)
3115 {
3116 	return -EOPNOTSUPP;
3117 }
3118 
sock_map_bpf_prog_query(const union bpf_attr * attr,union bpf_attr __user * uattr)3119 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr,
3120 					  union bpf_attr __user *uattr)
3121 {
3122 	return -EINVAL;
3123 }
3124 
sock_map_link_create(const union bpf_attr * attr,struct bpf_prog * prog)3125 static inline int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog)
3126 {
3127 	return -EOPNOTSUPP;
3128 }
3129 #endif /* CONFIG_BPF_SYSCALL */
3130 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
3131 
3132 static __always_inline void
bpf_prog_inc_misses_counters(const struct bpf_prog_array * array)3133 bpf_prog_inc_misses_counters(const struct bpf_prog_array *array)
3134 {
3135 	const struct bpf_prog_array_item *item;
3136 	struct bpf_prog *prog;
3137 
3138 	if (unlikely(!array))
3139 		return;
3140 
3141 	item = &array->items[0];
3142 	while ((prog = READ_ONCE(item->prog))) {
3143 		bpf_prog_inc_misses_counter(prog);
3144 		item++;
3145 	}
3146 }
3147 
3148 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
3149 void bpf_sk_reuseport_detach(struct sock *sk);
3150 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
3151 				       void *value);
3152 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
3153 				       void *value, u64 map_flags);
3154 #else
bpf_sk_reuseport_detach(struct sock * sk)3155 static inline void bpf_sk_reuseport_detach(struct sock *sk)
3156 {
3157 }
3158 
3159 #ifdef CONFIG_BPF_SYSCALL
bpf_fd_reuseport_array_lookup_elem(struct bpf_map * map,void * key,void * value)3160 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
3161 						     void *key, void *value)
3162 {
3163 	return -EOPNOTSUPP;
3164 }
3165 
bpf_fd_reuseport_array_update_elem(struct bpf_map * map,void * key,void * value,u64 map_flags)3166 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
3167 						     void *key, void *value,
3168 						     u64 map_flags)
3169 {
3170 	return -EOPNOTSUPP;
3171 }
3172 #endif /* CONFIG_BPF_SYSCALL */
3173 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
3174 
3175 /* verifier prototypes for helper functions called from eBPF programs */
3176 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
3177 extern const struct bpf_func_proto bpf_map_update_elem_proto;
3178 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
3179 extern const struct bpf_func_proto bpf_map_push_elem_proto;
3180 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
3181 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
3182 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto;
3183 
3184 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
3185 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
3186 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
3187 extern const struct bpf_func_proto bpf_tail_call_proto;
3188 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
3189 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
3190 extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto;
3191 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
3192 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
3193 extern const struct bpf_func_proto bpf_get_current_comm_proto;
3194 extern const struct bpf_func_proto bpf_get_stackid_proto;
3195 extern const struct bpf_func_proto bpf_get_stack_proto;
3196 extern const struct bpf_func_proto bpf_get_task_stack_proto;
3197 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
3198 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
3199 extern const struct bpf_func_proto bpf_sock_map_update_proto;
3200 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
3201 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
3202 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
3203 extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto;
3204 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
3205 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
3206 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
3207 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
3208 extern const struct bpf_func_proto bpf_spin_lock_proto;
3209 extern const struct bpf_func_proto bpf_spin_unlock_proto;
3210 extern const struct bpf_func_proto bpf_get_local_storage_proto;
3211 extern const struct bpf_func_proto bpf_strtol_proto;
3212 extern const struct bpf_func_proto bpf_strtoul_proto;
3213 extern const struct bpf_func_proto bpf_tcp_sock_proto;
3214 extern const struct bpf_func_proto bpf_jiffies64_proto;
3215 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
3216 extern const struct bpf_func_proto bpf_event_output_data_proto;
3217 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
3218 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
3219 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
3220 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
3221 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
3222 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto;
3223 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto;
3224 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto;
3225 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
3226 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
3227 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
3228 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
3229 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
3230 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto;
3231 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto;
3232 extern const struct bpf_func_proto bpf_copy_from_user_proto;
3233 extern const struct bpf_func_proto bpf_snprintf_btf_proto;
3234 extern const struct bpf_func_proto bpf_snprintf_proto;
3235 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
3236 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
3237 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto;
3238 extern const struct bpf_func_proto bpf_sock_from_file_proto;
3239 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto;
3240 extern const struct bpf_func_proto bpf_task_storage_get_recur_proto;
3241 extern const struct bpf_func_proto bpf_task_storage_get_proto;
3242 extern const struct bpf_func_proto bpf_task_storage_delete_recur_proto;
3243 extern const struct bpf_func_proto bpf_task_storage_delete_proto;
3244 extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
3245 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto;
3246 extern const struct bpf_func_proto bpf_sk_setsockopt_proto;
3247 extern const struct bpf_func_proto bpf_sk_getsockopt_proto;
3248 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto;
3249 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto;
3250 extern const struct bpf_func_proto bpf_find_vma_proto;
3251 extern const struct bpf_func_proto bpf_loop_proto;
3252 extern const struct bpf_func_proto bpf_copy_from_user_task_proto;
3253 extern const struct bpf_func_proto bpf_set_retval_proto;
3254 extern const struct bpf_func_proto bpf_get_retval_proto;
3255 extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto;
3256 extern const struct bpf_func_proto bpf_cgrp_storage_get_proto;
3257 extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto;
3258 
3259 const struct bpf_func_proto *tracing_prog_func_proto(
3260   enum bpf_func_id func_id, const struct bpf_prog *prog);
3261 
3262 /* Shared helpers among cBPF and eBPF. */
3263 void bpf_user_rnd_init_once(void);
3264 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3265 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3266 
3267 #if defined(CONFIG_NET)
3268 bool bpf_sock_common_is_valid_access(int off, int size,
3269 				     enum bpf_access_type type,
3270 				     struct bpf_insn_access_aux *info);
3271 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3272 			      struct bpf_insn_access_aux *info);
3273 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3274 				const struct bpf_insn *si,
3275 				struct bpf_insn *insn_buf,
3276 				struct bpf_prog *prog,
3277 				u32 *target_size);
3278 int bpf_dynptr_from_skb_rdonly(struct sk_buff *skb, u64 flags,
3279 			       struct bpf_dynptr_kern *ptr);
3280 #else
bpf_sock_common_is_valid_access(int off,int size,enum bpf_access_type type,struct bpf_insn_access_aux * info)3281 static inline bool bpf_sock_common_is_valid_access(int off, int size,
3282 						   enum bpf_access_type type,
3283 						   struct bpf_insn_access_aux *info)
3284 {
3285 	return false;
3286 }
bpf_sock_is_valid_access(int off,int size,enum bpf_access_type type,struct bpf_insn_access_aux * info)3287 static inline bool bpf_sock_is_valid_access(int off, int size,
3288 					    enum bpf_access_type type,
3289 					    struct bpf_insn_access_aux *info)
3290 {
3291 	return false;
3292 }
bpf_sock_convert_ctx_access(enum bpf_access_type type,const struct bpf_insn * si,struct bpf_insn * insn_buf,struct bpf_prog * prog,u32 * target_size)3293 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3294 					      const struct bpf_insn *si,
3295 					      struct bpf_insn *insn_buf,
3296 					      struct bpf_prog *prog,
3297 					      u32 *target_size)
3298 {
3299 	return 0;
3300 }
bpf_dynptr_from_skb_rdonly(struct sk_buff * skb,u64 flags,struct bpf_dynptr_kern * ptr)3301 static inline int bpf_dynptr_from_skb_rdonly(struct sk_buff *skb, u64 flags,
3302 					     struct bpf_dynptr_kern *ptr)
3303 {
3304 	return -EOPNOTSUPP;
3305 }
3306 #endif
3307 
3308 #ifdef CONFIG_INET
3309 struct sk_reuseport_kern {
3310 	struct sk_buff *skb;
3311 	struct sock *sk;
3312 	struct sock *selected_sk;
3313 	struct sock *migrating_sk;
3314 	void *data_end;
3315 	u32 hash;
3316 	u32 reuseport_id;
3317 	bool bind_inany;
3318 };
3319 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3320 				  struct bpf_insn_access_aux *info);
3321 
3322 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3323 				    const struct bpf_insn *si,
3324 				    struct bpf_insn *insn_buf,
3325 				    struct bpf_prog *prog,
3326 				    u32 *target_size);
3327 
3328 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3329 				  struct bpf_insn_access_aux *info);
3330 
3331 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3332 				    const struct bpf_insn *si,
3333 				    struct bpf_insn *insn_buf,
3334 				    struct bpf_prog *prog,
3335 				    u32 *target_size);
3336 #else
bpf_tcp_sock_is_valid_access(int off,int size,enum bpf_access_type type,struct bpf_insn_access_aux * info)3337 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
3338 						enum bpf_access_type type,
3339 						struct bpf_insn_access_aux *info)
3340 {
3341 	return false;
3342 }
3343 
bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,const struct bpf_insn * si,struct bpf_insn * insn_buf,struct bpf_prog * prog,u32 * target_size)3344 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3345 						  const struct bpf_insn *si,
3346 						  struct bpf_insn *insn_buf,
3347 						  struct bpf_prog *prog,
3348 						  u32 *target_size)
3349 {
3350 	return 0;
3351 }
bpf_xdp_sock_is_valid_access(int off,int size,enum bpf_access_type type,struct bpf_insn_access_aux * info)3352 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
3353 						enum bpf_access_type type,
3354 						struct bpf_insn_access_aux *info)
3355 {
3356 	return false;
3357 }
3358 
bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,const struct bpf_insn * si,struct bpf_insn * insn_buf,struct bpf_prog * prog,u32 * target_size)3359 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3360 						  const struct bpf_insn *si,
3361 						  struct bpf_insn *insn_buf,
3362 						  struct bpf_prog *prog,
3363 						  u32 *target_size)
3364 {
3365 	return 0;
3366 }
3367 #endif /* CONFIG_INET */
3368 
3369 enum bpf_text_poke_type {
3370 	BPF_MOD_CALL,
3371 	BPF_MOD_JUMP,
3372 };
3373 
3374 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
3375 		       void *addr1, void *addr2);
3376 
3377 void bpf_arch_poke_desc_update(struct bpf_jit_poke_descriptor *poke,
3378 			       struct bpf_prog *new, struct bpf_prog *old);
3379 
3380 void *bpf_arch_text_copy(void *dst, void *src, size_t len);
3381 int bpf_arch_text_invalidate(void *dst, size_t len);
3382 
3383 struct btf_id_set;
3384 bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
3385 
3386 #define MAX_BPRINTF_VARARGS		12
3387 #define MAX_BPRINTF_BUF			1024
3388 
3389 struct bpf_bprintf_data {
3390 	u32 *bin_args;
3391 	char *buf;
3392 	bool get_bin_args;
3393 	bool get_buf;
3394 };
3395 
3396 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
3397 			u32 num_args, struct bpf_bprintf_data *data);
3398 void bpf_bprintf_cleanup(struct bpf_bprintf_data *data);
3399 
3400 #ifdef CONFIG_BPF_LSM
3401 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype);
3402 void bpf_cgroup_atype_put(int cgroup_atype);
3403 #else
bpf_cgroup_atype_get(u32 attach_btf_id,int cgroup_atype)3404 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {}
bpf_cgroup_atype_put(int cgroup_atype)3405 static inline void bpf_cgroup_atype_put(int cgroup_atype) {}
3406 #endif /* CONFIG_BPF_LSM */
3407 
3408 struct key;
3409 
3410 #ifdef CONFIG_KEYS
3411 struct bpf_key {
3412 	struct key *key;
3413 	bool has_ref;
3414 };
3415 #endif /* CONFIG_KEYS */
3416 
type_is_alloc(u32 type)3417 static inline bool type_is_alloc(u32 type)
3418 {
3419 	return type & MEM_ALLOC;
3420 }
3421 
bpf_memcg_flags(gfp_t flags)3422 static inline gfp_t bpf_memcg_flags(gfp_t flags)
3423 {
3424 	if (memcg_bpf_enabled())
3425 		return flags | __GFP_ACCOUNT;
3426 	return flags;
3427 }
3428 
bpf_is_subprog(const struct bpf_prog * prog)3429 static inline bool bpf_is_subprog(const struct bpf_prog *prog)
3430 {
3431 	return prog->aux->func_idx != 0;
3432 }
3433 
3434 #endif /* _LINUX_BPF_H */
3435