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