]>
Commit | Line | Data |
---|---|---|
99c55f7d AS |
1 | /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com |
2 | * | |
3 | * This program is free software; you can redistribute it and/or | |
4 | * modify it under the terms of version 2 of the GNU General Public | |
5 | * License as published by the Free Software Foundation. | |
6 | * | |
7 | * This program is distributed in the hope that it will be useful, but | |
8 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
9 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
10 | * General Public License for more details. | |
11 | */ | |
12 | #include <linux/bpf.h> | |
13 | #include <linux/syscalls.h> | |
14 | #include <linux/slab.h> | |
15 | #include <linux/anon_inodes.h> | |
db20fd2b | 16 | #include <linux/file.h> |
09756af4 AS |
17 | #include <linux/license.h> |
18 | #include <linux/filter.h> | |
2541517c | 19 | #include <linux/version.h> |
99c55f7d AS |
20 | |
21 | static LIST_HEAD(bpf_map_types); | |
22 | ||
23 | static struct bpf_map *find_and_alloc_map(union bpf_attr *attr) | |
24 | { | |
25 | struct bpf_map_type_list *tl; | |
26 | struct bpf_map *map; | |
27 | ||
28 | list_for_each_entry(tl, &bpf_map_types, list_node) { | |
29 | if (tl->type == attr->map_type) { | |
30 | map = tl->ops->map_alloc(attr); | |
31 | if (IS_ERR(map)) | |
32 | return map; | |
33 | map->ops = tl->ops; | |
34 | map->map_type = attr->map_type; | |
35 | return map; | |
36 | } | |
37 | } | |
38 | return ERR_PTR(-EINVAL); | |
39 | } | |
40 | ||
41 | /* boot time registration of different map implementations */ | |
42 | void bpf_register_map_type(struct bpf_map_type_list *tl) | |
43 | { | |
44 | list_add(&tl->list_node, &bpf_map_types); | |
45 | } | |
46 | ||
47 | /* called from workqueue */ | |
48 | static void bpf_map_free_deferred(struct work_struct *work) | |
49 | { | |
50 | struct bpf_map *map = container_of(work, struct bpf_map, work); | |
51 | ||
52 | /* implementation dependent freeing */ | |
53 | map->ops->map_free(map); | |
54 | } | |
55 | ||
56 | /* decrement map refcnt and schedule it for freeing via workqueue | |
57 | * (unrelying map implementation ops->map_free() might sleep) | |
58 | */ | |
59 | void bpf_map_put(struct bpf_map *map) | |
60 | { | |
61 | if (atomic_dec_and_test(&map->refcnt)) { | |
62 | INIT_WORK(&map->work, bpf_map_free_deferred); | |
63 | schedule_work(&map->work); | |
64 | } | |
65 | } | |
66 | ||
67 | static int bpf_map_release(struct inode *inode, struct file *filp) | |
68 | { | |
69 | struct bpf_map *map = filp->private_data; | |
70 | ||
04fd61ab AS |
71 | if (map->map_type == BPF_MAP_TYPE_PROG_ARRAY) |
72 | /* prog_array stores refcnt-ed bpf_prog pointers | |
73 | * release them all when user space closes prog_array_fd | |
74 | */ | |
2a36f0b9 | 75 | bpf_fd_array_map_clear(map); |
04fd61ab | 76 | |
99c55f7d AS |
77 | bpf_map_put(map); |
78 | return 0; | |
79 | } | |
80 | ||
81 | static const struct file_operations bpf_map_fops = { | |
82 | .release = bpf_map_release, | |
83 | }; | |
84 | ||
85 | /* helper macro to check that unused fields 'union bpf_attr' are zero */ | |
86 | #define CHECK_ATTR(CMD) \ | |
87 | memchr_inv((void *) &attr->CMD##_LAST_FIELD + \ | |
88 | sizeof(attr->CMD##_LAST_FIELD), 0, \ | |
89 | sizeof(*attr) - \ | |
90 | offsetof(union bpf_attr, CMD##_LAST_FIELD) - \ | |
91 | sizeof(attr->CMD##_LAST_FIELD)) != NULL | |
92 | ||
93 | #define BPF_MAP_CREATE_LAST_FIELD max_entries | |
94 | /* called via syscall */ | |
95 | static int map_create(union bpf_attr *attr) | |
96 | { | |
97 | struct bpf_map *map; | |
98 | int err; | |
99 | ||
100 | err = CHECK_ATTR(BPF_MAP_CREATE); | |
101 | if (err) | |
102 | return -EINVAL; | |
103 | ||
104 | /* find map type and init map: hashtable vs rbtree vs bloom vs ... */ | |
105 | map = find_and_alloc_map(attr); | |
106 | if (IS_ERR(map)) | |
107 | return PTR_ERR(map); | |
108 | ||
109 | atomic_set(&map->refcnt, 1); | |
110 | ||
111 | err = anon_inode_getfd("bpf-map", &bpf_map_fops, map, O_RDWR | O_CLOEXEC); | |
112 | ||
113 | if (err < 0) | |
114 | /* failed to allocate fd */ | |
115 | goto free_map; | |
116 | ||
117 | return err; | |
118 | ||
119 | free_map: | |
120 | map->ops->map_free(map); | |
121 | return err; | |
122 | } | |
123 | ||
db20fd2b AS |
124 | /* if error is returned, fd is released. |
125 | * On success caller should complete fd access with matching fdput() | |
126 | */ | |
127 | struct bpf_map *bpf_map_get(struct fd f) | |
128 | { | |
129 | struct bpf_map *map; | |
130 | ||
131 | if (!f.file) | |
132 | return ERR_PTR(-EBADF); | |
133 | ||
134 | if (f.file->f_op != &bpf_map_fops) { | |
135 | fdput(f); | |
136 | return ERR_PTR(-EINVAL); | |
137 | } | |
138 | ||
139 | map = f.file->private_data; | |
140 | ||
141 | return map; | |
142 | } | |
143 | ||
144 | /* helper to convert user pointers passed inside __aligned_u64 fields */ | |
145 | static void __user *u64_to_ptr(__u64 val) | |
146 | { | |
147 | return (void __user *) (unsigned long) val; | |
148 | } | |
149 | ||
150 | /* last field in 'union bpf_attr' used by this command */ | |
151 | #define BPF_MAP_LOOKUP_ELEM_LAST_FIELD value | |
152 | ||
153 | static int map_lookup_elem(union bpf_attr *attr) | |
154 | { | |
155 | void __user *ukey = u64_to_ptr(attr->key); | |
156 | void __user *uvalue = u64_to_ptr(attr->value); | |
157 | int ufd = attr->map_fd; | |
db20fd2b | 158 | struct bpf_map *map; |
8ebe667c | 159 | void *key, *value, *ptr; |
592867bf | 160 | struct fd f; |
db20fd2b AS |
161 | int err; |
162 | ||
163 | if (CHECK_ATTR(BPF_MAP_LOOKUP_ELEM)) | |
164 | return -EINVAL; | |
165 | ||
592867bf | 166 | f = fdget(ufd); |
db20fd2b AS |
167 | map = bpf_map_get(f); |
168 | if (IS_ERR(map)) | |
169 | return PTR_ERR(map); | |
170 | ||
171 | err = -ENOMEM; | |
172 | key = kmalloc(map->key_size, GFP_USER); | |
173 | if (!key) | |
174 | goto err_put; | |
175 | ||
176 | err = -EFAULT; | |
177 | if (copy_from_user(key, ukey, map->key_size) != 0) | |
178 | goto free_key; | |
179 | ||
8ebe667c AS |
180 | err = -ENOMEM; |
181 | value = kmalloc(map->value_size, GFP_USER); | |
db20fd2b | 182 | if (!value) |
8ebe667c AS |
183 | goto free_key; |
184 | ||
185 | rcu_read_lock(); | |
186 | ptr = map->ops->map_lookup_elem(map, key); | |
187 | if (ptr) | |
188 | memcpy(value, ptr, map->value_size); | |
189 | rcu_read_unlock(); | |
190 | ||
191 | err = -ENOENT; | |
192 | if (!ptr) | |
193 | goto free_value; | |
db20fd2b AS |
194 | |
195 | err = -EFAULT; | |
196 | if (copy_to_user(uvalue, value, map->value_size) != 0) | |
8ebe667c | 197 | goto free_value; |
db20fd2b AS |
198 | |
199 | err = 0; | |
200 | ||
8ebe667c AS |
201 | free_value: |
202 | kfree(value); | |
db20fd2b AS |
203 | free_key: |
204 | kfree(key); | |
205 | err_put: | |
206 | fdput(f); | |
207 | return err; | |
208 | } | |
209 | ||
3274f520 | 210 | #define BPF_MAP_UPDATE_ELEM_LAST_FIELD flags |
db20fd2b AS |
211 | |
212 | static int map_update_elem(union bpf_attr *attr) | |
213 | { | |
214 | void __user *ukey = u64_to_ptr(attr->key); | |
215 | void __user *uvalue = u64_to_ptr(attr->value); | |
216 | int ufd = attr->map_fd; | |
db20fd2b AS |
217 | struct bpf_map *map; |
218 | void *key, *value; | |
592867bf | 219 | struct fd f; |
db20fd2b AS |
220 | int err; |
221 | ||
222 | if (CHECK_ATTR(BPF_MAP_UPDATE_ELEM)) | |
223 | return -EINVAL; | |
224 | ||
592867bf | 225 | f = fdget(ufd); |
db20fd2b AS |
226 | map = bpf_map_get(f); |
227 | if (IS_ERR(map)) | |
228 | return PTR_ERR(map); | |
229 | ||
230 | err = -ENOMEM; | |
231 | key = kmalloc(map->key_size, GFP_USER); | |
232 | if (!key) | |
233 | goto err_put; | |
234 | ||
235 | err = -EFAULT; | |
236 | if (copy_from_user(key, ukey, map->key_size) != 0) | |
237 | goto free_key; | |
238 | ||
239 | err = -ENOMEM; | |
240 | value = kmalloc(map->value_size, GFP_USER); | |
241 | if (!value) | |
242 | goto free_key; | |
243 | ||
244 | err = -EFAULT; | |
245 | if (copy_from_user(value, uvalue, map->value_size) != 0) | |
246 | goto free_value; | |
247 | ||
248 | /* eBPF program that use maps are running under rcu_read_lock(), | |
249 | * therefore all map accessors rely on this fact, so do the same here | |
250 | */ | |
251 | rcu_read_lock(); | |
3274f520 | 252 | err = map->ops->map_update_elem(map, key, value, attr->flags); |
db20fd2b AS |
253 | rcu_read_unlock(); |
254 | ||
255 | free_value: | |
256 | kfree(value); | |
257 | free_key: | |
258 | kfree(key); | |
259 | err_put: | |
260 | fdput(f); | |
261 | return err; | |
262 | } | |
263 | ||
264 | #define BPF_MAP_DELETE_ELEM_LAST_FIELD key | |
265 | ||
266 | static int map_delete_elem(union bpf_attr *attr) | |
267 | { | |
268 | void __user *ukey = u64_to_ptr(attr->key); | |
269 | int ufd = attr->map_fd; | |
db20fd2b | 270 | struct bpf_map *map; |
592867bf | 271 | struct fd f; |
db20fd2b AS |
272 | void *key; |
273 | int err; | |
274 | ||
275 | if (CHECK_ATTR(BPF_MAP_DELETE_ELEM)) | |
276 | return -EINVAL; | |
277 | ||
592867bf | 278 | f = fdget(ufd); |
db20fd2b AS |
279 | map = bpf_map_get(f); |
280 | if (IS_ERR(map)) | |
281 | return PTR_ERR(map); | |
282 | ||
283 | err = -ENOMEM; | |
284 | key = kmalloc(map->key_size, GFP_USER); | |
285 | if (!key) | |
286 | goto err_put; | |
287 | ||
288 | err = -EFAULT; | |
289 | if (copy_from_user(key, ukey, map->key_size) != 0) | |
290 | goto free_key; | |
291 | ||
292 | rcu_read_lock(); | |
293 | err = map->ops->map_delete_elem(map, key); | |
294 | rcu_read_unlock(); | |
295 | ||
296 | free_key: | |
297 | kfree(key); | |
298 | err_put: | |
299 | fdput(f); | |
300 | return err; | |
301 | } | |
302 | ||
303 | /* last field in 'union bpf_attr' used by this command */ | |
304 | #define BPF_MAP_GET_NEXT_KEY_LAST_FIELD next_key | |
305 | ||
306 | static int map_get_next_key(union bpf_attr *attr) | |
307 | { | |
308 | void __user *ukey = u64_to_ptr(attr->key); | |
309 | void __user *unext_key = u64_to_ptr(attr->next_key); | |
310 | int ufd = attr->map_fd; | |
db20fd2b AS |
311 | struct bpf_map *map; |
312 | void *key, *next_key; | |
592867bf | 313 | struct fd f; |
db20fd2b AS |
314 | int err; |
315 | ||
316 | if (CHECK_ATTR(BPF_MAP_GET_NEXT_KEY)) | |
317 | return -EINVAL; | |
318 | ||
592867bf | 319 | f = fdget(ufd); |
db20fd2b AS |
320 | map = bpf_map_get(f); |
321 | if (IS_ERR(map)) | |
322 | return PTR_ERR(map); | |
323 | ||
324 | err = -ENOMEM; | |
325 | key = kmalloc(map->key_size, GFP_USER); | |
326 | if (!key) | |
327 | goto err_put; | |
328 | ||
329 | err = -EFAULT; | |
330 | if (copy_from_user(key, ukey, map->key_size) != 0) | |
331 | goto free_key; | |
332 | ||
333 | err = -ENOMEM; | |
334 | next_key = kmalloc(map->key_size, GFP_USER); | |
335 | if (!next_key) | |
336 | goto free_key; | |
337 | ||
338 | rcu_read_lock(); | |
339 | err = map->ops->map_get_next_key(map, key, next_key); | |
340 | rcu_read_unlock(); | |
341 | if (err) | |
342 | goto free_next_key; | |
343 | ||
344 | err = -EFAULT; | |
345 | if (copy_to_user(unext_key, next_key, map->key_size) != 0) | |
346 | goto free_next_key; | |
347 | ||
348 | err = 0; | |
349 | ||
350 | free_next_key: | |
351 | kfree(next_key); | |
352 | free_key: | |
353 | kfree(key); | |
354 | err_put: | |
355 | fdput(f); | |
356 | return err; | |
357 | } | |
358 | ||
09756af4 AS |
359 | static LIST_HEAD(bpf_prog_types); |
360 | ||
361 | static int find_prog_type(enum bpf_prog_type type, struct bpf_prog *prog) | |
362 | { | |
363 | struct bpf_prog_type_list *tl; | |
364 | ||
365 | list_for_each_entry(tl, &bpf_prog_types, list_node) { | |
366 | if (tl->type == type) { | |
367 | prog->aux->ops = tl->ops; | |
24701ece | 368 | prog->type = type; |
09756af4 AS |
369 | return 0; |
370 | } | |
371 | } | |
24701ece | 372 | |
09756af4 AS |
373 | return -EINVAL; |
374 | } | |
375 | ||
376 | void bpf_register_prog_type(struct bpf_prog_type_list *tl) | |
377 | { | |
378 | list_add(&tl->list_node, &bpf_prog_types); | |
379 | } | |
380 | ||
0a542a86 AS |
381 | /* fixup insn->imm field of bpf_call instructions: |
382 | * if (insn->imm == BPF_FUNC_map_lookup_elem) | |
383 | * insn->imm = bpf_map_lookup_elem - __bpf_call_base; | |
384 | * else if (insn->imm == BPF_FUNC_map_update_elem) | |
385 | * insn->imm = bpf_map_update_elem - __bpf_call_base; | |
386 | * else ... | |
387 | * | |
388 | * this function is called after eBPF program passed verification | |
389 | */ | |
390 | static void fixup_bpf_calls(struct bpf_prog *prog) | |
391 | { | |
392 | const struct bpf_func_proto *fn; | |
393 | int i; | |
394 | ||
395 | for (i = 0; i < prog->len; i++) { | |
396 | struct bpf_insn *insn = &prog->insnsi[i]; | |
397 | ||
398 | if (insn->code == (BPF_JMP | BPF_CALL)) { | |
399 | /* we reach here when program has bpf_call instructions | |
400 | * and it passed bpf_check(), means that | |
401 | * ops->get_func_proto must have been supplied, check it | |
402 | */ | |
403 | BUG_ON(!prog->aux->ops->get_func_proto); | |
404 | ||
c46646d0 DB |
405 | if (insn->imm == BPF_FUNC_get_route_realm) |
406 | prog->dst_needed = 1; | |
3ad00405 DB |
407 | if (insn->imm == BPF_FUNC_get_prandom_u32) |
408 | bpf_user_rnd_init_once(); | |
04fd61ab AS |
409 | if (insn->imm == BPF_FUNC_tail_call) { |
410 | /* mark bpf_tail_call as different opcode | |
411 | * to avoid conditional branch in | |
412 | * interpeter for every normal call | |
413 | * and to prevent accidental JITing by | |
414 | * JIT compiler that doesn't support | |
415 | * bpf_tail_call yet | |
416 | */ | |
417 | insn->imm = 0; | |
418 | insn->code |= BPF_X; | |
419 | continue; | |
420 | } | |
421 | ||
0a542a86 AS |
422 | fn = prog->aux->ops->get_func_proto(insn->imm); |
423 | /* all functions that have prototype and verifier allowed | |
424 | * programs to call them, must be real in-kernel functions | |
425 | */ | |
426 | BUG_ON(!fn->func); | |
427 | insn->imm = fn->func - __bpf_call_base; | |
428 | } | |
429 | } | |
430 | } | |
431 | ||
09756af4 AS |
432 | /* drop refcnt on maps used by eBPF program and free auxilary data */ |
433 | static void free_used_maps(struct bpf_prog_aux *aux) | |
434 | { | |
435 | int i; | |
436 | ||
437 | for (i = 0; i < aux->used_map_cnt; i++) | |
438 | bpf_map_put(aux->used_maps[i]); | |
439 | ||
440 | kfree(aux->used_maps); | |
441 | } | |
442 | ||
abf2e7d6 AS |
443 | static void __prog_put_rcu(struct rcu_head *rcu) |
444 | { | |
445 | struct bpf_prog_aux *aux = container_of(rcu, struct bpf_prog_aux, rcu); | |
446 | ||
447 | free_used_maps(aux); | |
448 | bpf_prog_free(aux->prog); | |
449 | } | |
450 | ||
451 | /* version of bpf_prog_put() that is called after a grace period */ | |
452 | void bpf_prog_put_rcu(struct bpf_prog *prog) | |
453 | { | |
454 | if (atomic_dec_and_test(&prog->aux->refcnt)) { | |
455 | prog->aux->prog = prog; | |
456 | call_rcu(&prog->aux->rcu, __prog_put_rcu); | |
457 | } | |
458 | } | |
459 | ||
09756af4 AS |
460 | void bpf_prog_put(struct bpf_prog *prog) |
461 | { | |
462 | if (atomic_dec_and_test(&prog->aux->refcnt)) { | |
463 | free_used_maps(prog->aux); | |
464 | bpf_prog_free(prog); | |
465 | } | |
466 | } | |
e2e9b654 | 467 | EXPORT_SYMBOL_GPL(bpf_prog_put); |
09756af4 AS |
468 | |
469 | static int bpf_prog_release(struct inode *inode, struct file *filp) | |
470 | { | |
471 | struct bpf_prog *prog = filp->private_data; | |
472 | ||
abf2e7d6 | 473 | bpf_prog_put_rcu(prog); |
09756af4 AS |
474 | return 0; |
475 | } | |
476 | ||
477 | static const struct file_operations bpf_prog_fops = { | |
478 | .release = bpf_prog_release, | |
479 | }; | |
480 | ||
481 | static struct bpf_prog *get_prog(struct fd f) | |
482 | { | |
483 | struct bpf_prog *prog; | |
484 | ||
485 | if (!f.file) | |
486 | return ERR_PTR(-EBADF); | |
487 | ||
488 | if (f.file->f_op != &bpf_prog_fops) { | |
489 | fdput(f); | |
490 | return ERR_PTR(-EINVAL); | |
491 | } | |
492 | ||
493 | prog = f.file->private_data; | |
494 | ||
495 | return prog; | |
496 | } | |
497 | ||
498 | /* called by sockets/tracing/seccomp before attaching program to an event | |
499 | * pairs with bpf_prog_put() | |
500 | */ | |
501 | struct bpf_prog *bpf_prog_get(u32 ufd) | |
502 | { | |
503 | struct fd f = fdget(ufd); | |
504 | struct bpf_prog *prog; | |
505 | ||
506 | prog = get_prog(f); | |
507 | ||
508 | if (IS_ERR(prog)) | |
509 | return prog; | |
510 | ||
511 | atomic_inc(&prog->aux->refcnt); | |
512 | fdput(f); | |
513 | return prog; | |
514 | } | |
e2e9b654 | 515 | EXPORT_SYMBOL_GPL(bpf_prog_get); |
09756af4 AS |
516 | |
517 | /* last field in 'union bpf_attr' used by this command */ | |
2541517c | 518 | #define BPF_PROG_LOAD_LAST_FIELD kern_version |
09756af4 AS |
519 | |
520 | static int bpf_prog_load(union bpf_attr *attr) | |
521 | { | |
522 | enum bpf_prog_type type = attr->prog_type; | |
523 | struct bpf_prog *prog; | |
524 | int err; | |
525 | char license[128]; | |
526 | bool is_gpl; | |
527 | ||
528 | if (CHECK_ATTR(BPF_PROG_LOAD)) | |
529 | return -EINVAL; | |
530 | ||
531 | /* copy eBPF program license from user space */ | |
532 | if (strncpy_from_user(license, u64_to_ptr(attr->license), | |
533 | sizeof(license) - 1) < 0) | |
534 | return -EFAULT; | |
535 | license[sizeof(license) - 1] = 0; | |
536 | ||
537 | /* eBPF programs must be GPL compatible to use GPL-ed functions */ | |
538 | is_gpl = license_is_gpl_compatible(license); | |
539 | ||
540 | if (attr->insn_cnt >= BPF_MAXINSNS) | |
541 | return -EINVAL; | |
542 | ||
2541517c AS |
543 | if (type == BPF_PROG_TYPE_KPROBE && |
544 | attr->kern_version != LINUX_VERSION_CODE) | |
545 | return -EINVAL; | |
546 | ||
09756af4 AS |
547 | /* plain bpf_prog allocation */ |
548 | prog = bpf_prog_alloc(bpf_prog_size(attr->insn_cnt), GFP_USER); | |
549 | if (!prog) | |
550 | return -ENOMEM; | |
551 | ||
552 | prog->len = attr->insn_cnt; | |
553 | ||
554 | err = -EFAULT; | |
555 | if (copy_from_user(prog->insns, u64_to_ptr(attr->insns), | |
556 | prog->len * sizeof(struct bpf_insn)) != 0) | |
557 | goto free_prog; | |
558 | ||
559 | prog->orig_prog = NULL; | |
a91263d5 | 560 | prog->jited = 0; |
09756af4 AS |
561 | |
562 | atomic_set(&prog->aux->refcnt, 1); | |
a91263d5 | 563 | prog->gpl_compatible = is_gpl ? 1 : 0; |
09756af4 AS |
564 | |
565 | /* find program type: socket_filter vs tracing_filter */ | |
566 | err = find_prog_type(type, prog); | |
567 | if (err < 0) | |
568 | goto free_prog; | |
569 | ||
570 | /* run eBPF verifier */ | |
9bac3d6d | 571 | err = bpf_check(&prog, attr); |
09756af4 AS |
572 | if (err < 0) |
573 | goto free_used_maps; | |
574 | ||
0a542a86 AS |
575 | /* fixup BPF_CALL->imm field */ |
576 | fixup_bpf_calls(prog); | |
577 | ||
09756af4 | 578 | /* eBPF program is ready to be JITed */ |
04fd61ab AS |
579 | err = bpf_prog_select_runtime(prog); |
580 | if (err < 0) | |
581 | goto free_used_maps; | |
09756af4 AS |
582 | |
583 | err = anon_inode_getfd("bpf-prog", &bpf_prog_fops, prog, O_RDWR | O_CLOEXEC); | |
09756af4 AS |
584 | if (err < 0) |
585 | /* failed to allocate fd */ | |
586 | goto free_used_maps; | |
587 | ||
588 | return err; | |
589 | ||
590 | free_used_maps: | |
591 | free_used_maps(prog->aux); | |
592 | free_prog: | |
593 | bpf_prog_free(prog); | |
594 | return err; | |
595 | } | |
596 | ||
99c55f7d AS |
597 | SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size) |
598 | { | |
599 | union bpf_attr attr = {}; | |
600 | int err; | |
601 | ||
602 | /* the syscall is limited to root temporarily. This restriction will be | |
603 | * lifted when security audit is clean. Note that eBPF+tracing must have | |
604 | * this restriction, since it may pass kernel data to user space | |
605 | */ | |
606 | if (!capable(CAP_SYS_ADMIN)) | |
607 | return -EPERM; | |
608 | ||
609 | if (!access_ok(VERIFY_READ, uattr, 1)) | |
610 | return -EFAULT; | |
611 | ||
612 | if (size > PAGE_SIZE) /* silly large */ | |
613 | return -E2BIG; | |
614 | ||
615 | /* If we're handed a bigger struct than we know of, | |
616 | * ensure all the unknown bits are 0 - i.e. new | |
617 | * user-space does not rely on any kernel feature | |
618 | * extensions we dont know about yet. | |
619 | */ | |
620 | if (size > sizeof(attr)) { | |
621 | unsigned char __user *addr; | |
622 | unsigned char __user *end; | |
623 | unsigned char val; | |
624 | ||
625 | addr = (void __user *)uattr + sizeof(attr); | |
626 | end = (void __user *)uattr + size; | |
627 | ||
628 | for (; addr < end; addr++) { | |
629 | err = get_user(val, addr); | |
630 | if (err) | |
631 | return err; | |
632 | if (val) | |
633 | return -E2BIG; | |
634 | } | |
635 | size = sizeof(attr); | |
636 | } | |
637 | ||
638 | /* copy attributes from user space, may be less than sizeof(bpf_attr) */ | |
639 | if (copy_from_user(&attr, uattr, size) != 0) | |
640 | return -EFAULT; | |
641 | ||
642 | switch (cmd) { | |
643 | case BPF_MAP_CREATE: | |
644 | err = map_create(&attr); | |
645 | break; | |
db20fd2b AS |
646 | case BPF_MAP_LOOKUP_ELEM: |
647 | err = map_lookup_elem(&attr); | |
648 | break; | |
649 | case BPF_MAP_UPDATE_ELEM: | |
650 | err = map_update_elem(&attr); | |
651 | break; | |
652 | case BPF_MAP_DELETE_ELEM: | |
653 | err = map_delete_elem(&attr); | |
654 | break; | |
655 | case BPF_MAP_GET_NEXT_KEY: | |
656 | err = map_get_next_key(&attr); | |
657 | break; | |
09756af4 AS |
658 | case BPF_PROG_LOAD: |
659 | err = bpf_prog_load(&attr); | |
660 | break; | |
99c55f7d AS |
661 | default: |
662 | err = -EINVAL; | |
663 | break; | |
664 | } | |
665 | ||
666 | return err; | |
667 | } |