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