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> | |
a67edbf4 | 13 | #include <linux/bpf_trace.h> |
99c55f7d AS |
14 | #include <linux/syscalls.h> |
15 | #include <linux/slab.h> | |
3f07c014 | 16 | #include <linux/sched/signal.h> |
d407bd25 DB |
17 | #include <linux/vmalloc.h> |
18 | #include <linux/mmzone.h> | |
99c55f7d | 19 | #include <linux/anon_inodes.h> |
db20fd2b | 20 | #include <linux/file.h> |
09756af4 AS |
21 | #include <linux/license.h> |
22 | #include <linux/filter.h> | |
2541517c | 23 | #include <linux/version.h> |
535e7b4b | 24 | #include <linux/kernel.h> |
dc4bb0e2 | 25 | #include <linux/idr.h> |
cb4d2b3f MKL |
26 | #include <linux/cred.h> |
27 | #include <linux/timekeeping.h> | |
28 | #include <linux/ctype.h> | |
99c55f7d | 29 | |
14dc6f04 MKL |
30 | #define IS_FD_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PROG_ARRAY || \ |
31 | (map)->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || \ | |
32 | (map)->map_type == BPF_MAP_TYPE_CGROUP_ARRAY || \ | |
33 | (map)->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS) | |
34 | #define IS_FD_HASH(map) ((map)->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) | |
35 | #define IS_FD_MAP(map) (IS_FD_ARRAY(map) || IS_FD_HASH(map)) | |
36 | ||
6e71b04a CF |
37 | #define BPF_OBJ_FLAG_MASK (BPF_F_RDONLY | BPF_F_WRONLY) |
38 | ||
b121d1e7 | 39 | DEFINE_PER_CPU(int, bpf_prog_active); |
dc4bb0e2 MKL |
40 | static DEFINE_IDR(prog_idr); |
41 | static DEFINE_SPINLOCK(prog_idr_lock); | |
f3f1c054 MKL |
42 | static DEFINE_IDR(map_idr); |
43 | static DEFINE_SPINLOCK(map_idr_lock); | |
b121d1e7 | 44 | |
1be7f75d AS |
45 | int sysctl_unprivileged_bpf_disabled __read_mostly; |
46 | ||
40077e0c JB |
47 | static const struct bpf_map_ops * const bpf_map_types[] = { |
48 | #define BPF_PROG_TYPE(_id, _ops) | |
49 | #define BPF_MAP_TYPE(_id, _ops) \ | |
50 | [_id] = &_ops, | |
51 | #include <linux/bpf_types.h> | |
52 | #undef BPF_PROG_TYPE | |
53 | #undef BPF_MAP_TYPE | |
54 | }; | |
99c55f7d | 55 | |
752ba56f MS |
56 | /* |
57 | * If we're handed a bigger struct than we know of, ensure all the unknown bits | |
58 | * are 0 - i.e. new user-space does not rely on any kernel feature extensions | |
59 | * we don't know about yet. | |
60 | * | |
61 | * There is a ToCToU between this function call and the following | |
62 | * copy_from_user() call. However, this is not a concern since this function is | |
63 | * meant to be a future-proofing of bits. | |
64 | */ | |
58291a74 MS |
65 | static int check_uarg_tail_zero(void __user *uaddr, |
66 | size_t expected_size, | |
67 | size_t actual_size) | |
68 | { | |
69 | unsigned char __user *addr; | |
70 | unsigned char __user *end; | |
71 | unsigned char val; | |
72 | int err; | |
73 | ||
752ba56f MS |
74 | if (unlikely(actual_size > PAGE_SIZE)) /* silly large */ |
75 | return -E2BIG; | |
76 | ||
77 | if (unlikely(!access_ok(VERIFY_READ, uaddr, actual_size))) | |
78 | return -EFAULT; | |
79 | ||
58291a74 MS |
80 | if (actual_size <= expected_size) |
81 | return 0; | |
82 | ||
83 | addr = uaddr + expected_size; | |
84 | end = uaddr + actual_size; | |
85 | ||
86 | for (; addr < end; addr++) { | |
87 | err = get_user(val, addr); | |
88 | if (err) | |
89 | return err; | |
90 | if (val) | |
91 | return -E2BIG; | |
92 | } | |
93 | ||
94 | return 0; | |
95 | } | |
96 | ||
99c55f7d AS |
97 | static struct bpf_map *find_and_alloc_map(union bpf_attr *attr) |
98 | { | |
99c55f7d AS |
99 | struct bpf_map *map; |
100 | ||
40077e0c JB |
101 | if (attr->map_type >= ARRAY_SIZE(bpf_map_types) || |
102 | !bpf_map_types[attr->map_type]) | |
103 | return ERR_PTR(-EINVAL); | |
99c55f7d | 104 | |
40077e0c JB |
105 | map = bpf_map_types[attr->map_type]->map_alloc(attr); |
106 | if (IS_ERR(map)) | |
107 | return map; | |
108 | map->ops = bpf_map_types[attr->map_type]; | |
109 | map->map_type = attr->map_type; | |
110 | return map; | |
99c55f7d AS |
111 | } |
112 | ||
96eabe7a | 113 | void *bpf_map_area_alloc(size_t size, int numa_node) |
d407bd25 DB |
114 | { |
115 | /* We definitely need __GFP_NORETRY, so OOM killer doesn't | |
116 | * trigger under memory pressure as we really just want to | |
117 | * fail instead. | |
118 | */ | |
119 | const gfp_t flags = __GFP_NOWARN | __GFP_NORETRY | __GFP_ZERO; | |
120 | void *area; | |
121 | ||
122 | if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) { | |
96eabe7a | 123 | area = kmalloc_node(size, GFP_USER | flags, numa_node); |
d407bd25 DB |
124 | if (area != NULL) |
125 | return area; | |
126 | } | |
127 | ||
96eabe7a MKL |
128 | return __vmalloc_node_flags_caller(size, numa_node, GFP_KERNEL | flags, |
129 | __builtin_return_address(0)); | |
d407bd25 DB |
130 | } |
131 | ||
132 | void bpf_map_area_free(void *area) | |
133 | { | |
134 | kvfree(area); | |
135 | } | |
136 | ||
6c905981 AS |
137 | int bpf_map_precharge_memlock(u32 pages) |
138 | { | |
139 | struct user_struct *user = get_current_user(); | |
140 | unsigned long memlock_limit, cur; | |
141 | ||
142 | memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
143 | cur = atomic_long_read(&user->locked_vm); | |
144 | free_uid(user); | |
145 | if (cur + pages > memlock_limit) | |
146 | return -EPERM; | |
147 | return 0; | |
148 | } | |
149 | ||
aaac3ba9 AS |
150 | static int bpf_map_charge_memlock(struct bpf_map *map) |
151 | { | |
152 | struct user_struct *user = get_current_user(); | |
153 | unsigned long memlock_limit; | |
154 | ||
155 | memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
156 | ||
157 | atomic_long_add(map->pages, &user->locked_vm); | |
158 | ||
159 | if (atomic_long_read(&user->locked_vm) > memlock_limit) { | |
160 | atomic_long_sub(map->pages, &user->locked_vm); | |
161 | free_uid(user); | |
162 | return -EPERM; | |
163 | } | |
164 | map->user = user; | |
165 | return 0; | |
166 | } | |
167 | ||
168 | static void bpf_map_uncharge_memlock(struct bpf_map *map) | |
169 | { | |
170 | struct user_struct *user = map->user; | |
171 | ||
172 | atomic_long_sub(map->pages, &user->locked_vm); | |
173 | free_uid(user); | |
174 | } | |
175 | ||
f3f1c054 MKL |
176 | static int bpf_map_alloc_id(struct bpf_map *map) |
177 | { | |
178 | int id; | |
179 | ||
180 | spin_lock_bh(&map_idr_lock); | |
181 | id = idr_alloc_cyclic(&map_idr, map, 1, INT_MAX, GFP_ATOMIC); | |
182 | if (id > 0) | |
183 | map->id = id; | |
184 | spin_unlock_bh(&map_idr_lock); | |
185 | ||
186 | if (WARN_ON_ONCE(!id)) | |
187 | return -ENOSPC; | |
188 | ||
189 | return id > 0 ? 0 : id; | |
190 | } | |
191 | ||
bd5f5f4e | 192 | static void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock) |
f3f1c054 | 193 | { |
930651a7 ED |
194 | unsigned long flags; |
195 | ||
bd5f5f4e | 196 | if (do_idr_lock) |
930651a7 | 197 | spin_lock_irqsave(&map_idr_lock, flags); |
bd5f5f4e MKL |
198 | else |
199 | __acquire(&map_idr_lock); | |
200 | ||
f3f1c054 | 201 | idr_remove(&map_idr, map->id); |
bd5f5f4e MKL |
202 | |
203 | if (do_idr_lock) | |
930651a7 | 204 | spin_unlock_irqrestore(&map_idr_lock, flags); |
bd5f5f4e MKL |
205 | else |
206 | __release(&map_idr_lock); | |
f3f1c054 MKL |
207 | } |
208 | ||
99c55f7d AS |
209 | /* called from workqueue */ |
210 | static void bpf_map_free_deferred(struct work_struct *work) | |
211 | { | |
212 | struct bpf_map *map = container_of(work, struct bpf_map, work); | |
213 | ||
aaac3ba9 | 214 | bpf_map_uncharge_memlock(map); |
afdb09c7 | 215 | security_bpf_map_free(map); |
99c55f7d AS |
216 | /* implementation dependent freeing */ |
217 | map->ops->map_free(map); | |
218 | } | |
219 | ||
c9da161c DB |
220 | static void bpf_map_put_uref(struct bpf_map *map) |
221 | { | |
222 | if (atomic_dec_and_test(&map->usercnt)) { | |
223 | if (map->map_type == BPF_MAP_TYPE_PROG_ARRAY) | |
224 | bpf_fd_array_map_clear(map); | |
225 | } | |
226 | } | |
227 | ||
99c55f7d AS |
228 | /* decrement map refcnt and schedule it for freeing via workqueue |
229 | * (unrelying map implementation ops->map_free() might sleep) | |
230 | */ | |
bd5f5f4e | 231 | static void __bpf_map_put(struct bpf_map *map, bool do_idr_lock) |
99c55f7d AS |
232 | { |
233 | if (atomic_dec_and_test(&map->refcnt)) { | |
34ad5580 | 234 | /* bpf_map_free_id() must be called first */ |
bd5f5f4e | 235 | bpf_map_free_id(map, do_idr_lock); |
99c55f7d AS |
236 | INIT_WORK(&map->work, bpf_map_free_deferred); |
237 | schedule_work(&map->work); | |
238 | } | |
239 | } | |
240 | ||
bd5f5f4e MKL |
241 | void bpf_map_put(struct bpf_map *map) |
242 | { | |
243 | __bpf_map_put(map, true); | |
244 | } | |
245 | ||
c9da161c | 246 | void bpf_map_put_with_uref(struct bpf_map *map) |
99c55f7d | 247 | { |
c9da161c | 248 | bpf_map_put_uref(map); |
99c55f7d | 249 | bpf_map_put(map); |
c9da161c DB |
250 | } |
251 | ||
252 | static int bpf_map_release(struct inode *inode, struct file *filp) | |
253 | { | |
61d1b6a4 DB |
254 | struct bpf_map *map = filp->private_data; |
255 | ||
256 | if (map->ops->map_release) | |
257 | map->ops->map_release(map, filp); | |
258 | ||
259 | bpf_map_put_with_uref(map); | |
99c55f7d AS |
260 | return 0; |
261 | } | |
262 | ||
f99bf205 DB |
263 | #ifdef CONFIG_PROC_FS |
264 | static void bpf_map_show_fdinfo(struct seq_file *m, struct file *filp) | |
265 | { | |
266 | const struct bpf_map *map = filp->private_data; | |
21116b70 DB |
267 | const struct bpf_array *array; |
268 | u32 owner_prog_type = 0; | |
9780c0ab | 269 | u32 owner_jited = 0; |
21116b70 DB |
270 | |
271 | if (map->map_type == BPF_MAP_TYPE_PROG_ARRAY) { | |
272 | array = container_of(map, struct bpf_array, map); | |
273 | owner_prog_type = array->owner_prog_type; | |
9780c0ab | 274 | owner_jited = array->owner_jited; |
21116b70 | 275 | } |
f99bf205 DB |
276 | |
277 | seq_printf(m, | |
278 | "map_type:\t%u\n" | |
279 | "key_size:\t%u\n" | |
280 | "value_size:\t%u\n" | |
322cea2f | 281 | "max_entries:\t%u\n" |
21116b70 DB |
282 | "map_flags:\t%#x\n" |
283 | "memlock:\t%llu\n", | |
f99bf205 DB |
284 | map->map_type, |
285 | map->key_size, | |
286 | map->value_size, | |
322cea2f | 287 | map->max_entries, |
21116b70 DB |
288 | map->map_flags, |
289 | map->pages * 1ULL << PAGE_SHIFT); | |
290 | ||
9780c0ab | 291 | if (owner_prog_type) { |
21116b70 DB |
292 | seq_printf(m, "owner_prog_type:\t%u\n", |
293 | owner_prog_type); | |
9780c0ab DB |
294 | seq_printf(m, "owner_jited:\t%u\n", |
295 | owner_jited); | |
296 | } | |
f99bf205 DB |
297 | } |
298 | #endif | |
299 | ||
6e71b04a CF |
300 | static ssize_t bpf_dummy_read(struct file *filp, char __user *buf, size_t siz, |
301 | loff_t *ppos) | |
302 | { | |
303 | /* We need this handler such that alloc_file() enables | |
304 | * f_mode with FMODE_CAN_READ. | |
305 | */ | |
306 | return -EINVAL; | |
307 | } | |
308 | ||
309 | static ssize_t bpf_dummy_write(struct file *filp, const char __user *buf, | |
310 | size_t siz, loff_t *ppos) | |
311 | { | |
312 | /* We need this handler such that alloc_file() enables | |
313 | * f_mode with FMODE_CAN_WRITE. | |
314 | */ | |
315 | return -EINVAL; | |
316 | } | |
317 | ||
f66e448c | 318 | const struct file_operations bpf_map_fops = { |
f99bf205 DB |
319 | #ifdef CONFIG_PROC_FS |
320 | .show_fdinfo = bpf_map_show_fdinfo, | |
321 | #endif | |
322 | .release = bpf_map_release, | |
6e71b04a CF |
323 | .read = bpf_dummy_read, |
324 | .write = bpf_dummy_write, | |
99c55f7d AS |
325 | }; |
326 | ||
6e71b04a | 327 | int bpf_map_new_fd(struct bpf_map *map, int flags) |
aa79781b | 328 | { |
afdb09c7 CF |
329 | int ret; |
330 | ||
331 | ret = security_bpf_map(map, OPEN_FMODE(flags)); | |
332 | if (ret < 0) | |
333 | return ret; | |
334 | ||
aa79781b | 335 | return anon_inode_getfd("bpf-map", &bpf_map_fops, map, |
6e71b04a CF |
336 | flags | O_CLOEXEC); |
337 | } | |
338 | ||
339 | int bpf_get_file_flag(int flags) | |
340 | { | |
341 | if ((flags & BPF_F_RDONLY) && (flags & BPF_F_WRONLY)) | |
342 | return -EINVAL; | |
343 | if (flags & BPF_F_RDONLY) | |
344 | return O_RDONLY; | |
345 | if (flags & BPF_F_WRONLY) | |
346 | return O_WRONLY; | |
347 | return O_RDWR; | |
aa79781b DB |
348 | } |
349 | ||
99c55f7d AS |
350 | /* helper macro to check that unused fields 'union bpf_attr' are zero */ |
351 | #define CHECK_ATTR(CMD) \ | |
352 | memchr_inv((void *) &attr->CMD##_LAST_FIELD + \ | |
353 | sizeof(attr->CMD##_LAST_FIELD), 0, \ | |
354 | sizeof(*attr) - \ | |
355 | offsetof(union bpf_attr, CMD##_LAST_FIELD) - \ | |
356 | sizeof(attr->CMD##_LAST_FIELD)) != NULL | |
357 | ||
cb4d2b3f MKL |
358 | /* dst and src must have at least BPF_OBJ_NAME_LEN number of bytes. |
359 | * Return 0 on success and < 0 on error. | |
360 | */ | |
361 | static int bpf_obj_name_cpy(char *dst, const char *src) | |
362 | { | |
363 | const char *end = src + BPF_OBJ_NAME_LEN; | |
364 | ||
473d9734 MKL |
365 | memset(dst, 0, BPF_OBJ_NAME_LEN); |
366 | ||
cb4d2b3f MKL |
367 | /* Copy all isalnum() and '_' char */ |
368 | while (src < end && *src) { | |
369 | if (!isalnum(*src) && *src != '_') | |
370 | return -EINVAL; | |
371 | *dst++ = *src++; | |
372 | } | |
373 | ||
374 | /* No '\0' found in BPF_OBJ_NAME_LEN number of bytes */ | |
375 | if (src == end) | |
376 | return -EINVAL; | |
377 | ||
cb4d2b3f MKL |
378 | return 0; |
379 | } | |
380 | ||
ad5b177b | 381 | #define BPF_MAP_CREATE_LAST_FIELD map_name |
99c55f7d AS |
382 | /* called via syscall */ |
383 | static int map_create(union bpf_attr *attr) | |
384 | { | |
96eabe7a | 385 | int numa_node = bpf_map_attr_numa_node(attr); |
99c55f7d | 386 | struct bpf_map *map; |
6e71b04a | 387 | int f_flags; |
99c55f7d AS |
388 | int err; |
389 | ||
390 | err = CHECK_ATTR(BPF_MAP_CREATE); | |
391 | if (err) | |
392 | return -EINVAL; | |
393 | ||
6e71b04a CF |
394 | f_flags = bpf_get_file_flag(attr->map_flags); |
395 | if (f_flags < 0) | |
396 | return f_flags; | |
397 | ||
96eabe7a | 398 | if (numa_node != NUMA_NO_NODE && |
96e5ae4e ED |
399 | ((unsigned int)numa_node >= nr_node_ids || |
400 | !node_online(numa_node))) | |
96eabe7a MKL |
401 | return -EINVAL; |
402 | ||
99c55f7d AS |
403 | /* find map type and init map: hashtable vs rbtree vs bloom vs ... */ |
404 | map = find_and_alloc_map(attr); | |
405 | if (IS_ERR(map)) | |
406 | return PTR_ERR(map); | |
407 | ||
ad5b177b MKL |
408 | err = bpf_obj_name_cpy(map->name, attr->map_name); |
409 | if (err) | |
410 | goto free_map_nouncharge; | |
411 | ||
99c55f7d | 412 | atomic_set(&map->refcnt, 1); |
c9da161c | 413 | atomic_set(&map->usercnt, 1); |
99c55f7d | 414 | |
afdb09c7 | 415 | err = security_bpf_map_alloc(map); |
aaac3ba9 | 416 | if (err) |
20b2b24f | 417 | goto free_map_nouncharge; |
aaac3ba9 | 418 | |
afdb09c7 CF |
419 | err = bpf_map_charge_memlock(map); |
420 | if (err) | |
421 | goto free_map_sec; | |
422 | ||
f3f1c054 MKL |
423 | err = bpf_map_alloc_id(map); |
424 | if (err) | |
425 | goto free_map; | |
426 | ||
6e71b04a | 427 | err = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
428 | if (err < 0) { |
429 | /* failed to allocate fd. | |
430 | * bpf_map_put() is needed because the above | |
431 | * bpf_map_alloc_id() has published the map | |
432 | * to the userspace and the userspace may | |
433 | * have refcnt-ed it through BPF_MAP_GET_FD_BY_ID. | |
434 | */ | |
435 | bpf_map_put(map); | |
436 | return err; | |
437 | } | |
99c55f7d | 438 | |
a67edbf4 | 439 | trace_bpf_map_create(map, err); |
99c55f7d AS |
440 | return err; |
441 | ||
442 | free_map: | |
20b2b24f | 443 | bpf_map_uncharge_memlock(map); |
afdb09c7 CF |
444 | free_map_sec: |
445 | security_bpf_map_free(map); | |
20b2b24f | 446 | free_map_nouncharge: |
99c55f7d AS |
447 | map->ops->map_free(map); |
448 | return err; | |
449 | } | |
450 | ||
db20fd2b AS |
451 | /* if error is returned, fd is released. |
452 | * On success caller should complete fd access with matching fdput() | |
453 | */ | |
c2101297 | 454 | struct bpf_map *__bpf_map_get(struct fd f) |
db20fd2b | 455 | { |
db20fd2b AS |
456 | if (!f.file) |
457 | return ERR_PTR(-EBADF); | |
db20fd2b AS |
458 | if (f.file->f_op != &bpf_map_fops) { |
459 | fdput(f); | |
460 | return ERR_PTR(-EINVAL); | |
461 | } | |
462 | ||
c2101297 DB |
463 | return f.file->private_data; |
464 | } | |
465 | ||
92117d84 AS |
466 | /* prog's and map's refcnt limit */ |
467 | #define BPF_MAX_REFCNT 32768 | |
468 | ||
469 | struct bpf_map *bpf_map_inc(struct bpf_map *map, bool uref) | |
c9da161c | 470 | { |
92117d84 AS |
471 | if (atomic_inc_return(&map->refcnt) > BPF_MAX_REFCNT) { |
472 | atomic_dec(&map->refcnt); | |
473 | return ERR_PTR(-EBUSY); | |
474 | } | |
c9da161c DB |
475 | if (uref) |
476 | atomic_inc(&map->usercnt); | |
92117d84 | 477 | return map; |
c9da161c DB |
478 | } |
479 | ||
480 | struct bpf_map *bpf_map_get_with_uref(u32 ufd) | |
c2101297 DB |
481 | { |
482 | struct fd f = fdget(ufd); | |
483 | struct bpf_map *map; | |
484 | ||
485 | map = __bpf_map_get(f); | |
486 | if (IS_ERR(map)) | |
487 | return map; | |
488 | ||
92117d84 | 489 | map = bpf_map_inc(map, true); |
c2101297 | 490 | fdput(f); |
db20fd2b AS |
491 | |
492 | return map; | |
493 | } | |
494 | ||
bd5f5f4e MKL |
495 | /* map_idr_lock should have been held */ |
496 | static struct bpf_map *bpf_map_inc_not_zero(struct bpf_map *map, | |
497 | bool uref) | |
498 | { | |
499 | int refold; | |
500 | ||
501 | refold = __atomic_add_unless(&map->refcnt, 1, 0); | |
502 | ||
503 | if (refold >= BPF_MAX_REFCNT) { | |
504 | __bpf_map_put(map, false); | |
505 | return ERR_PTR(-EBUSY); | |
506 | } | |
507 | ||
508 | if (!refold) | |
509 | return ERR_PTR(-ENOENT); | |
510 | ||
511 | if (uref) | |
512 | atomic_inc(&map->usercnt); | |
513 | ||
514 | return map; | |
515 | } | |
516 | ||
b8cdc051 AS |
517 | int __weak bpf_stackmap_copy(struct bpf_map *map, void *key, void *value) |
518 | { | |
519 | return -ENOTSUPP; | |
520 | } | |
521 | ||
db20fd2b AS |
522 | /* last field in 'union bpf_attr' used by this command */ |
523 | #define BPF_MAP_LOOKUP_ELEM_LAST_FIELD value | |
524 | ||
525 | static int map_lookup_elem(union bpf_attr *attr) | |
526 | { | |
535e7b4b MS |
527 | void __user *ukey = u64_to_user_ptr(attr->key); |
528 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 529 | int ufd = attr->map_fd; |
db20fd2b | 530 | struct bpf_map *map; |
8ebe667c | 531 | void *key, *value, *ptr; |
15a07b33 | 532 | u32 value_size; |
592867bf | 533 | struct fd f; |
db20fd2b AS |
534 | int err; |
535 | ||
536 | if (CHECK_ATTR(BPF_MAP_LOOKUP_ELEM)) | |
537 | return -EINVAL; | |
538 | ||
592867bf | 539 | f = fdget(ufd); |
c2101297 | 540 | map = __bpf_map_get(f); |
db20fd2b AS |
541 | if (IS_ERR(map)) |
542 | return PTR_ERR(map); | |
543 | ||
6e71b04a CF |
544 | if (!(f.file->f_mode & FMODE_CAN_READ)) { |
545 | err = -EPERM; | |
546 | goto err_put; | |
547 | } | |
548 | ||
e4448ed8 AV |
549 | key = memdup_user(ukey, map->key_size); |
550 | if (IS_ERR(key)) { | |
551 | err = PTR_ERR(key); | |
db20fd2b | 552 | goto err_put; |
e4448ed8 | 553 | } |
db20fd2b | 554 | |
15a07b33 | 555 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
8f844938 | 556 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || |
15a07b33 AS |
557 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) |
558 | value_size = round_up(map->value_size, 8) * num_possible_cpus(); | |
14dc6f04 MKL |
559 | else if (IS_FD_MAP(map)) |
560 | value_size = sizeof(u32); | |
15a07b33 AS |
561 | else |
562 | value_size = map->value_size; | |
563 | ||
8ebe667c | 564 | err = -ENOMEM; |
15a07b33 | 565 | value = kmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b | 566 | if (!value) |
8ebe667c AS |
567 | goto free_key; |
568 | ||
8f844938 MKL |
569 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
570 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
15a07b33 AS |
571 | err = bpf_percpu_hash_copy(map, key, value); |
572 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
573 | err = bpf_percpu_array_copy(map, key, value); | |
557c0c6e AS |
574 | } else if (map->map_type == BPF_MAP_TYPE_STACK_TRACE) { |
575 | err = bpf_stackmap_copy(map, key, value); | |
14dc6f04 MKL |
576 | } else if (IS_FD_ARRAY(map)) { |
577 | err = bpf_fd_array_map_lookup_elem(map, key, value); | |
578 | } else if (IS_FD_HASH(map)) { | |
579 | err = bpf_fd_htab_map_lookup_elem(map, key, value); | |
15a07b33 AS |
580 | } else { |
581 | rcu_read_lock(); | |
582 | ptr = map->ops->map_lookup_elem(map, key); | |
583 | if (ptr) | |
584 | memcpy(value, ptr, value_size); | |
585 | rcu_read_unlock(); | |
586 | err = ptr ? 0 : -ENOENT; | |
587 | } | |
8ebe667c | 588 | |
15a07b33 | 589 | if (err) |
8ebe667c | 590 | goto free_value; |
db20fd2b AS |
591 | |
592 | err = -EFAULT; | |
15a07b33 | 593 | if (copy_to_user(uvalue, value, value_size) != 0) |
8ebe667c | 594 | goto free_value; |
db20fd2b | 595 | |
a67edbf4 | 596 | trace_bpf_map_lookup_elem(map, ufd, key, value); |
db20fd2b AS |
597 | err = 0; |
598 | ||
8ebe667c AS |
599 | free_value: |
600 | kfree(value); | |
db20fd2b AS |
601 | free_key: |
602 | kfree(key); | |
603 | err_put: | |
604 | fdput(f); | |
605 | return err; | |
606 | } | |
607 | ||
3274f520 | 608 | #define BPF_MAP_UPDATE_ELEM_LAST_FIELD flags |
db20fd2b AS |
609 | |
610 | static int map_update_elem(union bpf_attr *attr) | |
611 | { | |
535e7b4b MS |
612 | void __user *ukey = u64_to_user_ptr(attr->key); |
613 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 614 | int ufd = attr->map_fd; |
db20fd2b AS |
615 | struct bpf_map *map; |
616 | void *key, *value; | |
15a07b33 | 617 | u32 value_size; |
592867bf | 618 | struct fd f; |
db20fd2b AS |
619 | int err; |
620 | ||
621 | if (CHECK_ATTR(BPF_MAP_UPDATE_ELEM)) | |
622 | return -EINVAL; | |
623 | ||
592867bf | 624 | f = fdget(ufd); |
c2101297 | 625 | map = __bpf_map_get(f); |
db20fd2b AS |
626 | if (IS_ERR(map)) |
627 | return PTR_ERR(map); | |
628 | ||
6e71b04a CF |
629 | if (!(f.file->f_mode & FMODE_CAN_WRITE)) { |
630 | err = -EPERM; | |
631 | goto err_put; | |
632 | } | |
633 | ||
e4448ed8 AV |
634 | key = memdup_user(ukey, map->key_size); |
635 | if (IS_ERR(key)) { | |
636 | err = PTR_ERR(key); | |
db20fd2b | 637 | goto err_put; |
e4448ed8 | 638 | } |
db20fd2b | 639 | |
15a07b33 | 640 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
8f844938 | 641 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || |
15a07b33 AS |
642 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) |
643 | value_size = round_up(map->value_size, 8) * num_possible_cpus(); | |
644 | else | |
645 | value_size = map->value_size; | |
646 | ||
db20fd2b | 647 | err = -ENOMEM; |
15a07b33 | 648 | value = kmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b AS |
649 | if (!value) |
650 | goto free_key; | |
651 | ||
652 | err = -EFAULT; | |
15a07b33 | 653 | if (copy_from_user(value, uvalue, value_size) != 0) |
db20fd2b AS |
654 | goto free_value; |
655 | ||
6710e112 JDB |
656 | /* Need to create a kthread, thus must support schedule */ |
657 | if (map->map_type == BPF_MAP_TYPE_CPUMAP) { | |
658 | err = map->ops->map_update_elem(map, key, value, attr->flags); | |
659 | goto out; | |
660 | } | |
661 | ||
b121d1e7 AS |
662 | /* must increment bpf_prog_active to avoid kprobe+bpf triggering from |
663 | * inside bpf map update or delete otherwise deadlocks are possible | |
664 | */ | |
665 | preempt_disable(); | |
666 | __this_cpu_inc(bpf_prog_active); | |
8f844938 MKL |
667 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
668 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
15a07b33 AS |
669 | err = bpf_percpu_hash_update(map, key, value, attr->flags); |
670 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
671 | err = bpf_percpu_array_update(map, key, value, attr->flags); | |
d056a788 | 672 | } else if (map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || |
4ed8ec52 | 673 | map->map_type == BPF_MAP_TYPE_PROG_ARRAY || |
56f668df MKL |
674 | map->map_type == BPF_MAP_TYPE_CGROUP_ARRAY || |
675 | map->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS) { | |
d056a788 DB |
676 | rcu_read_lock(); |
677 | err = bpf_fd_array_map_update_elem(map, f.file, key, value, | |
678 | attr->flags); | |
679 | rcu_read_unlock(); | |
bcc6b1b7 MKL |
680 | } else if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) { |
681 | rcu_read_lock(); | |
682 | err = bpf_fd_htab_map_update_elem(map, f.file, key, value, | |
683 | attr->flags); | |
684 | rcu_read_unlock(); | |
15a07b33 AS |
685 | } else { |
686 | rcu_read_lock(); | |
687 | err = map->ops->map_update_elem(map, key, value, attr->flags); | |
688 | rcu_read_unlock(); | |
689 | } | |
b121d1e7 AS |
690 | __this_cpu_dec(bpf_prog_active); |
691 | preempt_enable(); | |
6710e112 | 692 | out: |
a67edbf4 DB |
693 | if (!err) |
694 | trace_bpf_map_update_elem(map, ufd, key, value); | |
db20fd2b AS |
695 | free_value: |
696 | kfree(value); | |
697 | free_key: | |
698 | kfree(key); | |
699 | err_put: | |
700 | fdput(f); | |
701 | return err; | |
702 | } | |
703 | ||
704 | #define BPF_MAP_DELETE_ELEM_LAST_FIELD key | |
705 | ||
706 | static int map_delete_elem(union bpf_attr *attr) | |
707 | { | |
535e7b4b | 708 | void __user *ukey = u64_to_user_ptr(attr->key); |
db20fd2b | 709 | int ufd = attr->map_fd; |
db20fd2b | 710 | struct bpf_map *map; |
592867bf | 711 | struct fd f; |
db20fd2b AS |
712 | void *key; |
713 | int err; | |
714 | ||
715 | if (CHECK_ATTR(BPF_MAP_DELETE_ELEM)) | |
716 | return -EINVAL; | |
717 | ||
592867bf | 718 | f = fdget(ufd); |
c2101297 | 719 | map = __bpf_map_get(f); |
db20fd2b AS |
720 | if (IS_ERR(map)) |
721 | return PTR_ERR(map); | |
722 | ||
6e71b04a CF |
723 | if (!(f.file->f_mode & FMODE_CAN_WRITE)) { |
724 | err = -EPERM; | |
725 | goto err_put; | |
726 | } | |
727 | ||
e4448ed8 AV |
728 | key = memdup_user(ukey, map->key_size); |
729 | if (IS_ERR(key)) { | |
730 | err = PTR_ERR(key); | |
db20fd2b | 731 | goto err_put; |
e4448ed8 | 732 | } |
db20fd2b | 733 | |
b121d1e7 AS |
734 | preempt_disable(); |
735 | __this_cpu_inc(bpf_prog_active); | |
db20fd2b AS |
736 | rcu_read_lock(); |
737 | err = map->ops->map_delete_elem(map, key); | |
738 | rcu_read_unlock(); | |
b121d1e7 AS |
739 | __this_cpu_dec(bpf_prog_active); |
740 | preempt_enable(); | |
db20fd2b | 741 | |
a67edbf4 DB |
742 | if (!err) |
743 | trace_bpf_map_delete_elem(map, ufd, key); | |
db20fd2b AS |
744 | kfree(key); |
745 | err_put: | |
746 | fdput(f); | |
747 | return err; | |
748 | } | |
749 | ||
750 | /* last field in 'union bpf_attr' used by this command */ | |
751 | #define BPF_MAP_GET_NEXT_KEY_LAST_FIELD next_key | |
752 | ||
753 | static int map_get_next_key(union bpf_attr *attr) | |
754 | { | |
535e7b4b MS |
755 | void __user *ukey = u64_to_user_ptr(attr->key); |
756 | void __user *unext_key = u64_to_user_ptr(attr->next_key); | |
db20fd2b | 757 | int ufd = attr->map_fd; |
db20fd2b AS |
758 | struct bpf_map *map; |
759 | void *key, *next_key; | |
592867bf | 760 | struct fd f; |
db20fd2b AS |
761 | int err; |
762 | ||
763 | if (CHECK_ATTR(BPF_MAP_GET_NEXT_KEY)) | |
764 | return -EINVAL; | |
765 | ||
592867bf | 766 | f = fdget(ufd); |
c2101297 | 767 | map = __bpf_map_get(f); |
db20fd2b AS |
768 | if (IS_ERR(map)) |
769 | return PTR_ERR(map); | |
770 | ||
6e71b04a CF |
771 | if (!(f.file->f_mode & FMODE_CAN_READ)) { |
772 | err = -EPERM; | |
773 | goto err_put; | |
774 | } | |
775 | ||
8fe45924 | 776 | if (ukey) { |
e4448ed8 AV |
777 | key = memdup_user(ukey, map->key_size); |
778 | if (IS_ERR(key)) { | |
779 | err = PTR_ERR(key); | |
8fe45924 | 780 | goto err_put; |
e4448ed8 | 781 | } |
8fe45924 TQ |
782 | } else { |
783 | key = NULL; | |
784 | } | |
db20fd2b AS |
785 | |
786 | err = -ENOMEM; | |
787 | next_key = kmalloc(map->key_size, GFP_USER); | |
788 | if (!next_key) | |
789 | goto free_key; | |
790 | ||
791 | rcu_read_lock(); | |
792 | err = map->ops->map_get_next_key(map, key, next_key); | |
793 | rcu_read_unlock(); | |
794 | if (err) | |
795 | goto free_next_key; | |
796 | ||
797 | err = -EFAULT; | |
798 | if (copy_to_user(unext_key, next_key, map->key_size) != 0) | |
799 | goto free_next_key; | |
800 | ||
a67edbf4 | 801 | trace_bpf_map_next_key(map, ufd, key, next_key); |
db20fd2b AS |
802 | err = 0; |
803 | ||
804 | free_next_key: | |
805 | kfree(next_key); | |
806 | free_key: | |
807 | kfree(key); | |
808 | err_put: | |
809 | fdput(f); | |
810 | return err; | |
811 | } | |
812 | ||
7de16e3a JK |
813 | static const struct bpf_prog_ops * const bpf_prog_types[] = { |
814 | #define BPF_PROG_TYPE(_id, _name) \ | |
815 | [_id] = & _name ## _prog_ops, | |
816 | #define BPF_MAP_TYPE(_id, _ops) | |
817 | #include <linux/bpf_types.h> | |
818 | #undef BPF_PROG_TYPE | |
819 | #undef BPF_MAP_TYPE | |
820 | }; | |
821 | ||
09756af4 AS |
822 | static int find_prog_type(enum bpf_prog_type type, struct bpf_prog *prog) |
823 | { | |
be9370a7 JB |
824 | if (type >= ARRAY_SIZE(bpf_prog_types) || !bpf_prog_types[type]) |
825 | return -EINVAL; | |
09756af4 | 826 | |
ab3f0063 JK |
827 | if (!bpf_prog_is_dev_bound(prog->aux)) |
828 | prog->aux->ops = bpf_prog_types[type]; | |
829 | else | |
830 | prog->aux->ops = &bpf_offload_prog_ops; | |
be9370a7 JB |
831 | prog->type = type; |
832 | return 0; | |
09756af4 AS |
833 | } |
834 | ||
835 | /* drop refcnt on maps used by eBPF program and free auxilary data */ | |
836 | static void free_used_maps(struct bpf_prog_aux *aux) | |
837 | { | |
838 | int i; | |
839 | ||
840 | for (i = 0; i < aux->used_map_cnt; i++) | |
841 | bpf_map_put(aux->used_maps[i]); | |
842 | ||
843 | kfree(aux->used_maps); | |
844 | } | |
845 | ||
5ccb071e DB |
846 | int __bpf_prog_charge(struct user_struct *user, u32 pages) |
847 | { | |
848 | unsigned long memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
849 | unsigned long user_bufs; | |
850 | ||
851 | if (user) { | |
852 | user_bufs = atomic_long_add_return(pages, &user->locked_vm); | |
853 | if (user_bufs > memlock_limit) { | |
854 | atomic_long_sub(pages, &user->locked_vm); | |
855 | return -EPERM; | |
856 | } | |
857 | } | |
858 | ||
859 | return 0; | |
860 | } | |
861 | ||
862 | void __bpf_prog_uncharge(struct user_struct *user, u32 pages) | |
863 | { | |
864 | if (user) | |
865 | atomic_long_sub(pages, &user->locked_vm); | |
866 | } | |
867 | ||
aaac3ba9 AS |
868 | static int bpf_prog_charge_memlock(struct bpf_prog *prog) |
869 | { | |
870 | struct user_struct *user = get_current_user(); | |
5ccb071e | 871 | int ret; |
aaac3ba9 | 872 | |
5ccb071e DB |
873 | ret = __bpf_prog_charge(user, prog->pages); |
874 | if (ret) { | |
aaac3ba9 | 875 | free_uid(user); |
5ccb071e | 876 | return ret; |
aaac3ba9 | 877 | } |
5ccb071e | 878 | |
aaac3ba9 AS |
879 | prog->aux->user = user; |
880 | return 0; | |
881 | } | |
882 | ||
883 | static void bpf_prog_uncharge_memlock(struct bpf_prog *prog) | |
884 | { | |
885 | struct user_struct *user = prog->aux->user; | |
886 | ||
5ccb071e | 887 | __bpf_prog_uncharge(user, prog->pages); |
aaac3ba9 AS |
888 | free_uid(user); |
889 | } | |
890 | ||
dc4bb0e2 MKL |
891 | static int bpf_prog_alloc_id(struct bpf_prog *prog) |
892 | { | |
893 | int id; | |
894 | ||
895 | spin_lock_bh(&prog_idr_lock); | |
896 | id = idr_alloc_cyclic(&prog_idr, prog, 1, INT_MAX, GFP_ATOMIC); | |
897 | if (id > 0) | |
898 | prog->aux->id = id; | |
899 | spin_unlock_bh(&prog_idr_lock); | |
900 | ||
901 | /* id is in [1, INT_MAX) */ | |
902 | if (WARN_ON_ONCE(!id)) | |
903 | return -ENOSPC; | |
904 | ||
905 | return id > 0 ? 0 : id; | |
906 | } | |
907 | ||
b16d9aa4 | 908 | static void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock) |
dc4bb0e2 MKL |
909 | { |
910 | /* cBPF to eBPF migrations are currently not in the idr store. */ | |
911 | if (!prog->aux->id) | |
912 | return; | |
913 | ||
b16d9aa4 MKL |
914 | if (do_idr_lock) |
915 | spin_lock_bh(&prog_idr_lock); | |
916 | else | |
917 | __acquire(&prog_idr_lock); | |
918 | ||
dc4bb0e2 | 919 | idr_remove(&prog_idr, prog->aux->id); |
b16d9aa4 MKL |
920 | |
921 | if (do_idr_lock) | |
922 | spin_unlock_bh(&prog_idr_lock); | |
923 | else | |
924 | __release(&prog_idr_lock); | |
dc4bb0e2 MKL |
925 | } |
926 | ||
1aacde3d | 927 | static void __bpf_prog_put_rcu(struct rcu_head *rcu) |
abf2e7d6 AS |
928 | { |
929 | struct bpf_prog_aux *aux = container_of(rcu, struct bpf_prog_aux, rcu); | |
930 | ||
931 | free_used_maps(aux); | |
aaac3ba9 | 932 | bpf_prog_uncharge_memlock(aux->prog); |
afdb09c7 | 933 | security_bpf_prog_free(aux); |
abf2e7d6 AS |
934 | bpf_prog_free(aux->prog); |
935 | } | |
936 | ||
b16d9aa4 | 937 | static void __bpf_prog_put(struct bpf_prog *prog, bool do_idr_lock) |
09756af4 | 938 | { |
a67edbf4 | 939 | if (atomic_dec_and_test(&prog->aux->refcnt)) { |
4f74d809 DB |
940 | int i; |
941 | ||
a67edbf4 | 942 | trace_bpf_prog_put_rcu(prog); |
34ad5580 | 943 | /* bpf_prog_free_id() must be called first */ |
b16d9aa4 | 944 | bpf_prog_free_id(prog, do_idr_lock); |
4f74d809 DB |
945 | |
946 | for (i = 0; i < prog->aux->func_cnt; i++) | |
947 | bpf_prog_kallsyms_del(prog->aux->func[i]); | |
74451e66 | 948 | bpf_prog_kallsyms_del(prog); |
4f74d809 | 949 | |
1aacde3d | 950 | call_rcu(&prog->aux->rcu, __bpf_prog_put_rcu); |
a67edbf4 | 951 | } |
09756af4 | 952 | } |
b16d9aa4 MKL |
953 | |
954 | void bpf_prog_put(struct bpf_prog *prog) | |
955 | { | |
956 | __bpf_prog_put(prog, true); | |
957 | } | |
e2e9b654 | 958 | EXPORT_SYMBOL_GPL(bpf_prog_put); |
09756af4 AS |
959 | |
960 | static int bpf_prog_release(struct inode *inode, struct file *filp) | |
961 | { | |
962 | struct bpf_prog *prog = filp->private_data; | |
963 | ||
1aacde3d | 964 | bpf_prog_put(prog); |
09756af4 AS |
965 | return 0; |
966 | } | |
967 | ||
7bd509e3 DB |
968 | #ifdef CONFIG_PROC_FS |
969 | static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp) | |
970 | { | |
971 | const struct bpf_prog *prog = filp->private_data; | |
f1f7714e | 972 | char prog_tag[sizeof(prog->tag) * 2 + 1] = { }; |
7bd509e3 | 973 | |
f1f7714e | 974 | bin2hex(prog_tag, prog->tag, sizeof(prog->tag)); |
7bd509e3 DB |
975 | seq_printf(m, |
976 | "prog_type:\t%u\n" | |
977 | "prog_jited:\t%u\n" | |
f1f7714e | 978 | "prog_tag:\t%s\n" |
7bd509e3 DB |
979 | "memlock:\t%llu\n", |
980 | prog->type, | |
981 | prog->jited, | |
f1f7714e | 982 | prog_tag, |
7bd509e3 DB |
983 | prog->pages * 1ULL << PAGE_SHIFT); |
984 | } | |
985 | #endif | |
986 | ||
f66e448c | 987 | const struct file_operations bpf_prog_fops = { |
7bd509e3 DB |
988 | #ifdef CONFIG_PROC_FS |
989 | .show_fdinfo = bpf_prog_show_fdinfo, | |
990 | #endif | |
991 | .release = bpf_prog_release, | |
6e71b04a CF |
992 | .read = bpf_dummy_read, |
993 | .write = bpf_dummy_write, | |
09756af4 AS |
994 | }; |
995 | ||
b2197755 | 996 | int bpf_prog_new_fd(struct bpf_prog *prog) |
aa79781b | 997 | { |
afdb09c7 CF |
998 | int ret; |
999 | ||
1000 | ret = security_bpf_prog(prog); | |
1001 | if (ret < 0) | |
1002 | return ret; | |
1003 | ||
aa79781b DB |
1004 | return anon_inode_getfd("bpf-prog", &bpf_prog_fops, prog, |
1005 | O_RDWR | O_CLOEXEC); | |
1006 | } | |
1007 | ||
113214be | 1008 | static struct bpf_prog *____bpf_prog_get(struct fd f) |
09756af4 | 1009 | { |
09756af4 AS |
1010 | if (!f.file) |
1011 | return ERR_PTR(-EBADF); | |
09756af4 AS |
1012 | if (f.file->f_op != &bpf_prog_fops) { |
1013 | fdput(f); | |
1014 | return ERR_PTR(-EINVAL); | |
1015 | } | |
1016 | ||
c2101297 | 1017 | return f.file->private_data; |
09756af4 AS |
1018 | } |
1019 | ||
59d3656d | 1020 | struct bpf_prog *bpf_prog_add(struct bpf_prog *prog, int i) |
92117d84 | 1021 | { |
59d3656d BB |
1022 | if (atomic_add_return(i, &prog->aux->refcnt) > BPF_MAX_REFCNT) { |
1023 | atomic_sub(i, &prog->aux->refcnt); | |
92117d84 AS |
1024 | return ERR_PTR(-EBUSY); |
1025 | } | |
1026 | return prog; | |
1027 | } | |
59d3656d BB |
1028 | EXPORT_SYMBOL_GPL(bpf_prog_add); |
1029 | ||
c540594f DB |
1030 | void bpf_prog_sub(struct bpf_prog *prog, int i) |
1031 | { | |
1032 | /* Only to be used for undoing previous bpf_prog_add() in some | |
1033 | * error path. We still know that another entity in our call | |
1034 | * path holds a reference to the program, thus atomic_sub() can | |
1035 | * be safely used in such cases! | |
1036 | */ | |
1037 | WARN_ON(atomic_sub_return(i, &prog->aux->refcnt) == 0); | |
1038 | } | |
1039 | EXPORT_SYMBOL_GPL(bpf_prog_sub); | |
1040 | ||
59d3656d BB |
1041 | struct bpf_prog *bpf_prog_inc(struct bpf_prog *prog) |
1042 | { | |
1043 | return bpf_prog_add(prog, 1); | |
1044 | } | |
97bc402d | 1045 | EXPORT_SYMBOL_GPL(bpf_prog_inc); |
92117d84 | 1046 | |
b16d9aa4 | 1047 | /* prog_idr_lock should have been held */ |
a6f6df69 | 1048 | struct bpf_prog *bpf_prog_inc_not_zero(struct bpf_prog *prog) |
b16d9aa4 MKL |
1049 | { |
1050 | int refold; | |
1051 | ||
1052 | refold = __atomic_add_unless(&prog->aux->refcnt, 1, 0); | |
1053 | ||
1054 | if (refold >= BPF_MAX_REFCNT) { | |
1055 | __bpf_prog_put(prog, false); | |
1056 | return ERR_PTR(-EBUSY); | |
1057 | } | |
1058 | ||
1059 | if (!refold) | |
1060 | return ERR_PTR(-ENOENT); | |
1061 | ||
1062 | return prog; | |
1063 | } | |
a6f6df69 | 1064 | EXPORT_SYMBOL_GPL(bpf_prog_inc_not_zero); |
b16d9aa4 | 1065 | |
288b3de5 JK |
1066 | static bool bpf_prog_get_ok(struct bpf_prog *prog, |
1067 | enum bpf_prog_type *attach_type, bool attach_drv) | |
248f346f | 1068 | { |
288b3de5 JK |
1069 | /* not an attachment, just a refcount inc, always allow */ |
1070 | if (!attach_type) | |
1071 | return true; | |
248f346f JK |
1072 | |
1073 | if (prog->type != *attach_type) | |
1074 | return false; | |
288b3de5 | 1075 | if (bpf_prog_is_dev_bound(prog->aux) && !attach_drv) |
248f346f JK |
1076 | return false; |
1077 | ||
1078 | return true; | |
1079 | } | |
1080 | ||
1081 | static struct bpf_prog *__bpf_prog_get(u32 ufd, enum bpf_prog_type *attach_type, | |
288b3de5 | 1082 | bool attach_drv) |
09756af4 AS |
1083 | { |
1084 | struct fd f = fdget(ufd); | |
1085 | struct bpf_prog *prog; | |
1086 | ||
113214be | 1087 | prog = ____bpf_prog_get(f); |
09756af4 AS |
1088 | if (IS_ERR(prog)) |
1089 | return prog; | |
288b3de5 | 1090 | if (!bpf_prog_get_ok(prog, attach_type, attach_drv)) { |
113214be DB |
1091 | prog = ERR_PTR(-EINVAL); |
1092 | goto out; | |
1093 | } | |
09756af4 | 1094 | |
92117d84 | 1095 | prog = bpf_prog_inc(prog); |
113214be | 1096 | out: |
09756af4 AS |
1097 | fdput(f); |
1098 | return prog; | |
1099 | } | |
113214be DB |
1100 | |
1101 | struct bpf_prog *bpf_prog_get(u32 ufd) | |
1102 | { | |
288b3de5 | 1103 | return __bpf_prog_get(ufd, NULL, false); |
113214be DB |
1104 | } |
1105 | ||
248f346f | 1106 | struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, |
288b3de5 | 1107 | bool attach_drv) |
248f346f | 1108 | { |
288b3de5 | 1109 | struct bpf_prog *prog = __bpf_prog_get(ufd, &type, attach_drv); |
248f346f JK |
1110 | |
1111 | if (!IS_ERR(prog)) | |
1112 | trace_bpf_prog_get_type(prog); | |
1113 | return prog; | |
1114 | } | |
6c8dfe21 | 1115 | EXPORT_SYMBOL_GPL(bpf_prog_get_type_dev); |
248f346f | 1116 | |
09756af4 | 1117 | /* last field in 'union bpf_attr' used by this command */ |
1f6f4cb7 | 1118 | #define BPF_PROG_LOAD_LAST_FIELD prog_ifindex |
09756af4 AS |
1119 | |
1120 | static int bpf_prog_load(union bpf_attr *attr) | |
1121 | { | |
1122 | enum bpf_prog_type type = attr->prog_type; | |
1123 | struct bpf_prog *prog; | |
1124 | int err; | |
1125 | char license[128]; | |
1126 | bool is_gpl; | |
1127 | ||
1128 | if (CHECK_ATTR(BPF_PROG_LOAD)) | |
1129 | return -EINVAL; | |
1130 | ||
e07b98d9 DM |
1131 | if (attr->prog_flags & ~BPF_F_STRICT_ALIGNMENT) |
1132 | return -EINVAL; | |
1133 | ||
09756af4 | 1134 | /* copy eBPF program license from user space */ |
535e7b4b | 1135 | if (strncpy_from_user(license, u64_to_user_ptr(attr->license), |
09756af4 AS |
1136 | sizeof(license) - 1) < 0) |
1137 | return -EFAULT; | |
1138 | license[sizeof(license) - 1] = 0; | |
1139 | ||
1140 | /* eBPF programs must be GPL compatible to use GPL-ed functions */ | |
1141 | is_gpl = license_is_gpl_compatible(license); | |
1142 | ||
ef0915ca DB |
1143 | if (attr->insn_cnt == 0 || attr->insn_cnt > BPF_MAXINSNS) |
1144 | return -E2BIG; | |
09756af4 | 1145 | |
2541517c AS |
1146 | if (type == BPF_PROG_TYPE_KPROBE && |
1147 | attr->kern_version != LINUX_VERSION_CODE) | |
1148 | return -EINVAL; | |
1149 | ||
80b7d819 CF |
1150 | if (type != BPF_PROG_TYPE_SOCKET_FILTER && |
1151 | type != BPF_PROG_TYPE_CGROUP_SKB && | |
1152 | !capable(CAP_SYS_ADMIN)) | |
1be7f75d AS |
1153 | return -EPERM; |
1154 | ||
09756af4 AS |
1155 | /* plain bpf_prog allocation */ |
1156 | prog = bpf_prog_alloc(bpf_prog_size(attr->insn_cnt), GFP_USER); | |
1157 | if (!prog) | |
1158 | return -ENOMEM; | |
1159 | ||
afdb09c7 | 1160 | err = security_bpf_prog_alloc(prog->aux); |
aaac3ba9 AS |
1161 | if (err) |
1162 | goto free_prog_nouncharge; | |
1163 | ||
afdb09c7 CF |
1164 | err = bpf_prog_charge_memlock(prog); |
1165 | if (err) | |
1166 | goto free_prog_sec; | |
1167 | ||
09756af4 AS |
1168 | prog->len = attr->insn_cnt; |
1169 | ||
1170 | err = -EFAULT; | |
535e7b4b | 1171 | if (copy_from_user(prog->insns, u64_to_user_ptr(attr->insns), |
aafe6ae9 | 1172 | bpf_prog_insn_size(prog)) != 0) |
09756af4 AS |
1173 | goto free_prog; |
1174 | ||
1175 | prog->orig_prog = NULL; | |
a91263d5 | 1176 | prog->jited = 0; |
09756af4 AS |
1177 | |
1178 | atomic_set(&prog->aux->refcnt, 1); | |
a91263d5 | 1179 | prog->gpl_compatible = is_gpl ? 1 : 0; |
09756af4 | 1180 | |
1f6f4cb7 | 1181 | if (attr->prog_ifindex) { |
ab3f0063 JK |
1182 | err = bpf_prog_offload_init(prog, attr); |
1183 | if (err) | |
1184 | goto free_prog; | |
1185 | } | |
1186 | ||
09756af4 AS |
1187 | /* find program type: socket_filter vs tracing_filter */ |
1188 | err = find_prog_type(type, prog); | |
1189 | if (err < 0) | |
1190 | goto free_prog; | |
1191 | ||
cb4d2b3f MKL |
1192 | prog->aux->load_time = ktime_get_boot_ns(); |
1193 | err = bpf_obj_name_cpy(prog->aux->name, attr->prog_name); | |
1194 | if (err) | |
1195 | goto free_prog; | |
1196 | ||
09756af4 | 1197 | /* run eBPF verifier */ |
9bac3d6d | 1198 | err = bpf_check(&prog, attr); |
09756af4 AS |
1199 | if (err < 0) |
1200 | goto free_used_maps; | |
1201 | ||
1202 | /* eBPF program is ready to be JITed */ | |
1c2a088a AS |
1203 | if (!prog->bpf_func) |
1204 | prog = bpf_prog_select_runtime(prog, &err); | |
04fd61ab AS |
1205 | if (err < 0) |
1206 | goto free_used_maps; | |
09756af4 | 1207 | |
dc4bb0e2 MKL |
1208 | err = bpf_prog_alloc_id(prog); |
1209 | if (err) | |
1210 | goto free_used_maps; | |
1211 | ||
aa79781b | 1212 | err = bpf_prog_new_fd(prog); |
b16d9aa4 MKL |
1213 | if (err < 0) { |
1214 | /* failed to allocate fd. | |
1215 | * bpf_prog_put() is needed because the above | |
1216 | * bpf_prog_alloc_id() has published the prog | |
1217 | * to the userspace and the userspace may | |
1218 | * have refcnt-ed it through BPF_PROG_GET_FD_BY_ID. | |
1219 | */ | |
1220 | bpf_prog_put(prog); | |
1221 | return err; | |
1222 | } | |
09756af4 | 1223 | |
74451e66 | 1224 | bpf_prog_kallsyms_add(prog); |
a67edbf4 | 1225 | trace_bpf_prog_load(prog, err); |
09756af4 AS |
1226 | return err; |
1227 | ||
1228 | free_used_maps: | |
1229 | free_used_maps(prog->aux); | |
1230 | free_prog: | |
aaac3ba9 | 1231 | bpf_prog_uncharge_memlock(prog); |
afdb09c7 CF |
1232 | free_prog_sec: |
1233 | security_bpf_prog_free(prog->aux); | |
aaac3ba9 | 1234 | free_prog_nouncharge: |
09756af4 AS |
1235 | bpf_prog_free(prog); |
1236 | return err; | |
1237 | } | |
1238 | ||
6e71b04a | 1239 | #define BPF_OBJ_LAST_FIELD file_flags |
b2197755 DB |
1240 | |
1241 | static int bpf_obj_pin(const union bpf_attr *attr) | |
1242 | { | |
6e71b04a | 1243 | if (CHECK_ATTR(BPF_OBJ) || attr->file_flags != 0) |
b2197755 DB |
1244 | return -EINVAL; |
1245 | ||
535e7b4b | 1246 | return bpf_obj_pin_user(attr->bpf_fd, u64_to_user_ptr(attr->pathname)); |
b2197755 DB |
1247 | } |
1248 | ||
1249 | static int bpf_obj_get(const union bpf_attr *attr) | |
1250 | { | |
6e71b04a CF |
1251 | if (CHECK_ATTR(BPF_OBJ) || attr->bpf_fd != 0 || |
1252 | attr->file_flags & ~BPF_OBJ_FLAG_MASK) | |
b2197755 DB |
1253 | return -EINVAL; |
1254 | ||
6e71b04a CF |
1255 | return bpf_obj_get_user(u64_to_user_ptr(attr->pathname), |
1256 | attr->file_flags); | |
b2197755 DB |
1257 | } |
1258 | ||
f4324551 DM |
1259 | #ifdef CONFIG_CGROUP_BPF |
1260 | ||
464bc0fd | 1261 | #define BPF_PROG_ATTACH_LAST_FIELD attach_flags |
174a79ff | 1262 | |
5a67da2a | 1263 | static int sockmap_get_from_fd(const union bpf_attr *attr, bool attach) |
174a79ff | 1264 | { |
5a67da2a | 1265 | struct bpf_prog *prog = NULL; |
174a79ff JF |
1266 | int ufd = attr->target_fd; |
1267 | struct bpf_map *map; | |
1268 | struct fd f; | |
1269 | int err; | |
1270 | ||
1271 | f = fdget(ufd); | |
1272 | map = __bpf_map_get(f); | |
1273 | if (IS_ERR(map)) | |
1274 | return PTR_ERR(map); | |
1275 | ||
5a67da2a JF |
1276 | if (attach) { |
1277 | prog = bpf_prog_get_type(attr->attach_bpf_fd, | |
1278 | BPF_PROG_TYPE_SK_SKB); | |
1279 | if (IS_ERR(prog)) { | |
1280 | fdput(f); | |
1281 | return PTR_ERR(prog); | |
1282 | } | |
174a79ff JF |
1283 | } |
1284 | ||
5a67da2a | 1285 | err = sock_map_prog(map, prog, attr->attach_type); |
174a79ff JF |
1286 | if (err) { |
1287 | fdput(f); | |
5a67da2a JF |
1288 | if (prog) |
1289 | bpf_prog_put(prog); | |
ae2b27b8 | 1290 | return err; |
174a79ff JF |
1291 | } |
1292 | ||
1293 | fdput(f); | |
ae2b27b8 | 1294 | return 0; |
174a79ff | 1295 | } |
f4324551 | 1296 | |
324bda9e AS |
1297 | #define BPF_F_ATTACH_MASK \ |
1298 | (BPF_F_ALLOW_OVERRIDE | BPF_F_ALLOW_MULTI) | |
1299 | ||
f4324551 DM |
1300 | static int bpf_prog_attach(const union bpf_attr *attr) |
1301 | { | |
7f677633 | 1302 | enum bpf_prog_type ptype; |
f4324551 DM |
1303 | struct bpf_prog *prog; |
1304 | struct cgroup *cgrp; | |
7f677633 | 1305 | int ret; |
f4324551 DM |
1306 | |
1307 | if (!capable(CAP_NET_ADMIN)) | |
1308 | return -EPERM; | |
1309 | ||
1310 | if (CHECK_ATTR(BPF_PROG_ATTACH)) | |
1311 | return -EINVAL; | |
1312 | ||
324bda9e | 1313 | if (attr->attach_flags & ~BPF_F_ATTACH_MASK) |
7f677633 AS |
1314 | return -EINVAL; |
1315 | ||
f4324551 DM |
1316 | switch (attr->attach_type) { |
1317 | case BPF_CGROUP_INET_INGRESS: | |
1318 | case BPF_CGROUP_INET_EGRESS: | |
b2cd1257 | 1319 | ptype = BPF_PROG_TYPE_CGROUP_SKB; |
f4324551 | 1320 | break; |
61023658 DA |
1321 | case BPF_CGROUP_INET_SOCK_CREATE: |
1322 | ptype = BPF_PROG_TYPE_CGROUP_SOCK; | |
1323 | break; | |
40304b2a LB |
1324 | case BPF_CGROUP_SOCK_OPS: |
1325 | ptype = BPF_PROG_TYPE_SOCK_OPS; | |
1326 | break; | |
ebc614f6 RG |
1327 | case BPF_CGROUP_DEVICE: |
1328 | ptype = BPF_PROG_TYPE_CGROUP_DEVICE; | |
1329 | break; | |
464bc0fd JF |
1330 | case BPF_SK_SKB_STREAM_PARSER: |
1331 | case BPF_SK_SKB_STREAM_VERDICT: | |
5a67da2a | 1332 | return sockmap_get_from_fd(attr, true); |
f4324551 DM |
1333 | default: |
1334 | return -EINVAL; | |
1335 | } | |
1336 | ||
b2cd1257 DA |
1337 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); |
1338 | if (IS_ERR(prog)) | |
1339 | return PTR_ERR(prog); | |
1340 | ||
1341 | cgrp = cgroup_get_from_fd(attr->target_fd); | |
1342 | if (IS_ERR(cgrp)) { | |
1343 | bpf_prog_put(prog); | |
1344 | return PTR_ERR(cgrp); | |
1345 | } | |
1346 | ||
324bda9e AS |
1347 | ret = cgroup_bpf_attach(cgrp, prog, attr->attach_type, |
1348 | attr->attach_flags); | |
7f677633 AS |
1349 | if (ret) |
1350 | bpf_prog_put(prog); | |
b2cd1257 DA |
1351 | cgroup_put(cgrp); |
1352 | ||
7f677633 | 1353 | return ret; |
f4324551 DM |
1354 | } |
1355 | ||
1356 | #define BPF_PROG_DETACH_LAST_FIELD attach_type | |
1357 | ||
1358 | static int bpf_prog_detach(const union bpf_attr *attr) | |
1359 | { | |
324bda9e AS |
1360 | enum bpf_prog_type ptype; |
1361 | struct bpf_prog *prog; | |
f4324551 | 1362 | struct cgroup *cgrp; |
7f677633 | 1363 | int ret; |
f4324551 DM |
1364 | |
1365 | if (!capable(CAP_NET_ADMIN)) | |
1366 | return -EPERM; | |
1367 | ||
1368 | if (CHECK_ATTR(BPF_PROG_DETACH)) | |
1369 | return -EINVAL; | |
1370 | ||
1371 | switch (attr->attach_type) { | |
1372 | case BPF_CGROUP_INET_INGRESS: | |
1373 | case BPF_CGROUP_INET_EGRESS: | |
324bda9e AS |
1374 | ptype = BPF_PROG_TYPE_CGROUP_SKB; |
1375 | break; | |
61023658 | 1376 | case BPF_CGROUP_INET_SOCK_CREATE: |
324bda9e AS |
1377 | ptype = BPF_PROG_TYPE_CGROUP_SOCK; |
1378 | break; | |
40304b2a | 1379 | case BPF_CGROUP_SOCK_OPS: |
324bda9e | 1380 | ptype = BPF_PROG_TYPE_SOCK_OPS; |
f4324551 | 1381 | break; |
ebc614f6 RG |
1382 | case BPF_CGROUP_DEVICE: |
1383 | ptype = BPF_PROG_TYPE_CGROUP_DEVICE; | |
1384 | break; | |
5a67da2a JF |
1385 | case BPF_SK_SKB_STREAM_PARSER: |
1386 | case BPF_SK_SKB_STREAM_VERDICT: | |
324bda9e | 1387 | return sockmap_get_from_fd(attr, false); |
f4324551 DM |
1388 | default: |
1389 | return -EINVAL; | |
1390 | } | |
1391 | ||
324bda9e AS |
1392 | cgrp = cgroup_get_from_fd(attr->target_fd); |
1393 | if (IS_ERR(cgrp)) | |
1394 | return PTR_ERR(cgrp); | |
1395 | ||
1396 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); | |
1397 | if (IS_ERR(prog)) | |
1398 | prog = NULL; | |
1399 | ||
1400 | ret = cgroup_bpf_detach(cgrp, prog, attr->attach_type, 0); | |
1401 | if (prog) | |
1402 | bpf_prog_put(prog); | |
1403 | cgroup_put(cgrp); | |
7f677633 | 1404 | return ret; |
f4324551 | 1405 | } |
40304b2a | 1406 | |
468e2f64 AS |
1407 | #define BPF_PROG_QUERY_LAST_FIELD query.prog_cnt |
1408 | ||
1409 | static int bpf_prog_query(const union bpf_attr *attr, | |
1410 | union bpf_attr __user *uattr) | |
1411 | { | |
1412 | struct cgroup *cgrp; | |
1413 | int ret; | |
1414 | ||
1415 | if (!capable(CAP_NET_ADMIN)) | |
1416 | return -EPERM; | |
1417 | if (CHECK_ATTR(BPF_PROG_QUERY)) | |
1418 | return -EINVAL; | |
1419 | if (attr->query.query_flags & ~BPF_F_QUERY_EFFECTIVE) | |
1420 | return -EINVAL; | |
1421 | ||
1422 | switch (attr->query.attach_type) { | |
1423 | case BPF_CGROUP_INET_INGRESS: | |
1424 | case BPF_CGROUP_INET_EGRESS: | |
1425 | case BPF_CGROUP_INET_SOCK_CREATE: | |
1426 | case BPF_CGROUP_SOCK_OPS: | |
ebc614f6 | 1427 | case BPF_CGROUP_DEVICE: |
468e2f64 AS |
1428 | break; |
1429 | default: | |
1430 | return -EINVAL; | |
1431 | } | |
1432 | cgrp = cgroup_get_from_fd(attr->query.target_fd); | |
1433 | if (IS_ERR(cgrp)) | |
1434 | return PTR_ERR(cgrp); | |
1435 | ret = cgroup_bpf_query(cgrp, attr, uattr); | |
1436 | cgroup_put(cgrp); | |
1437 | return ret; | |
1438 | } | |
f4324551 DM |
1439 | #endif /* CONFIG_CGROUP_BPF */ |
1440 | ||
1cf1cae9 AS |
1441 | #define BPF_PROG_TEST_RUN_LAST_FIELD test.duration |
1442 | ||
1443 | static int bpf_prog_test_run(const union bpf_attr *attr, | |
1444 | union bpf_attr __user *uattr) | |
1445 | { | |
1446 | struct bpf_prog *prog; | |
1447 | int ret = -ENOTSUPP; | |
1448 | ||
1449 | if (CHECK_ATTR(BPF_PROG_TEST_RUN)) | |
1450 | return -EINVAL; | |
1451 | ||
1452 | prog = bpf_prog_get(attr->test.prog_fd); | |
1453 | if (IS_ERR(prog)) | |
1454 | return PTR_ERR(prog); | |
1455 | ||
1456 | if (prog->aux->ops->test_run) | |
1457 | ret = prog->aux->ops->test_run(prog, attr, uattr); | |
1458 | ||
1459 | bpf_prog_put(prog); | |
1460 | return ret; | |
1461 | } | |
1462 | ||
34ad5580 MKL |
1463 | #define BPF_OBJ_GET_NEXT_ID_LAST_FIELD next_id |
1464 | ||
1465 | static int bpf_obj_get_next_id(const union bpf_attr *attr, | |
1466 | union bpf_attr __user *uattr, | |
1467 | struct idr *idr, | |
1468 | spinlock_t *lock) | |
1469 | { | |
1470 | u32 next_id = attr->start_id; | |
1471 | int err = 0; | |
1472 | ||
1473 | if (CHECK_ATTR(BPF_OBJ_GET_NEXT_ID) || next_id >= INT_MAX) | |
1474 | return -EINVAL; | |
1475 | ||
1476 | if (!capable(CAP_SYS_ADMIN)) | |
1477 | return -EPERM; | |
1478 | ||
1479 | next_id++; | |
1480 | spin_lock_bh(lock); | |
1481 | if (!idr_get_next(idr, &next_id)) | |
1482 | err = -ENOENT; | |
1483 | spin_unlock_bh(lock); | |
1484 | ||
1485 | if (!err) | |
1486 | err = put_user(next_id, &uattr->next_id); | |
1487 | ||
1488 | return err; | |
1489 | } | |
1490 | ||
b16d9aa4 MKL |
1491 | #define BPF_PROG_GET_FD_BY_ID_LAST_FIELD prog_id |
1492 | ||
1493 | static int bpf_prog_get_fd_by_id(const union bpf_attr *attr) | |
1494 | { | |
1495 | struct bpf_prog *prog; | |
1496 | u32 id = attr->prog_id; | |
1497 | int fd; | |
1498 | ||
1499 | if (CHECK_ATTR(BPF_PROG_GET_FD_BY_ID)) | |
1500 | return -EINVAL; | |
1501 | ||
1502 | if (!capable(CAP_SYS_ADMIN)) | |
1503 | return -EPERM; | |
1504 | ||
1505 | spin_lock_bh(&prog_idr_lock); | |
1506 | prog = idr_find(&prog_idr, id); | |
1507 | if (prog) | |
1508 | prog = bpf_prog_inc_not_zero(prog); | |
1509 | else | |
1510 | prog = ERR_PTR(-ENOENT); | |
1511 | spin_unlock_bh(&prog_idr_lock); | |
1512 | ||
1513 | if (IS_ERR(prog)) | |
1514 | return PTR_ERR(prog); | |
1515 | ||
1516 | fd = bpf_prog_new_fd(prog); | |
1517 | if (fd < 0) | |
1518 | bpf_prog_put(prog); | |
1519 | ||
1520 | return fd; | |
1521 | } | |
1522 | ||
6e71b04a | 1523 | #define BPF_MAP_GET_FD_BY_ID_LAST_FIELD open_flags |
bd5f5f4e MKL |
1524 | |
1525 | static int bpf_map_get_fd_by_id(const union bpf_attr *attr) | |
1526 | { | |
1527 | struct bpf_map *map; | |
1528 | u32 id = attr->map_id; | |
6e71b04a | 1529 | int f_flags; |
bd5f5f4e MKL |
1530 | int fd; |
1531 | ||
6e71b04a CF |
1532 | if (CHECK_ATTR(BPF_MAP_GET_FD_BY_ID) || |
1533 | attr->open_flags & ~BPF_OBJ_FLAG_MASK) | |
bd5f5f4e MKL |
1534 | return -EINVAL; |
1535 | ||
1536 | if (!capable(CAP_SYS_ADMIN)) | |
1537 | return -EPERM; | |
1538 | ||
6e71b04a CF |
1539 | f_flags = bpf_get_file_flag(attr->open_flags); |
1540 | if (f_flags < 0) | |
1541 | return f_flags; | |
1542 | ||
bd5f5f4e MKL |
1543 | spin_lock_bh(&map_idr_lock); |
1544 | map = idr_find(&map_idr, id); | |
1545 | if (map) | |
1546 | map = bpf_map_inc_not_zero(map, true); | |
1547 | else | |
1548 | map = ERR_PTR(-ENOENT); | |
1549 | spin_unlock_bh(&map_idr_lock); | |
1550 | ||
1551 | if (IS_ERR(map)) | |
1552 | return PTR_ERR(map); | |
1553 | ||
6e71b04a | 1554 | fd = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
1555 | if (fd < 0) |
1556 | bpf_map_put(map); | |
1557 | ||
1558 | return fd; | |
1559 | } | |
1560 | ||
7105e828 DB |
1561 | static const struct bpf_map *bpf_map_from_imm(const struct bpf_prog *prog, |
1562 | unsigned long addr) | |
1563 | { | |
1564 | int i; | |
1565 | ||
1566 | for (i = 0; i < prog->aux->used_map_cnt; i++) | |
1567 | if (prog->aux->used_maps[i] == (void *)addr) | |
1568 | return prog->aux->used_maps[i]; | |
1569 | return NULL; | |
1570 | } | |
1571 | ||
1572 | static struct bpf_insn *bpf_insn_prepare_dump(const struct bpf_prog *prog) | |
1573 | { | |
1574 | const struct bpf_map *map; | |
1575 | struct bpf_insn *insns; | |
1576 | u64 imm; | |
1577 | int i; | |
1578 | ||
1579 | insns = kmemdup(prog->insnsi, bpf_prog_insn_size(prog), | |
1580 | GFP_USER); | |
1581 | if (!insns) | |
1582 | return insns; | |
1583 | ||
1584 | for (i = 0; i < prog->len; i++) { | |
1585 | if (insns[i].code == (BPF_JMP | BPF_TAIL_CALL)) { | |
1586 | insns[i].code = BPF_JMP | BPF_CALL; | |
1587 | insns[i].imm = BPF_FUNC_tail_call; | |
1588 | /* fall-through */ | |
1589 | } | |
1590 | if (insns[i].code == (BPF_JMP | BPF_CALL) || | |
1591 | insns[i].code == (BPF_JMP | BPF_CALL_ARGS)) { | |
1592 | if (insns[i].code == (BPF_JMP | BPF_CALL_ARGS)) | |
1593 | insns[i].code = BPF_JMP | BPF_CALL; | |
1594 | if (!bpf_dump_raw_ok()) | |
1595 | insns[i].imm = 0; | |
1596 | continue; | |
1597 | } | |
1598 | ||
1599 | if (insns[i].code != (BPF_LD | BPF_IMM | BPF_DW)) | |
1600 | continue; | |
1601 | ||
1602 | imm = ((u64)insns[i + 1].imm << 32) | (u32)insns[i].imm; | |
1603 | map = bpf_map_from_imm(prog, imm); | |
1604 | if (map) { | |
1605 | insns[i].src_reg = BPF_PSEUDO_MAP_FD; | |
1606 | insns[i].imm = map->id; | |
1607 | insns[i + 1].imm = 0; | |
1608 | continue; | |
1609 | } | |
1610 | ||
1611 | if (!bpf_dump_raw_ok() && | |
1612 | imm == (unsigned long)prog->aux) { | |
1613 | insns[i].imm = 0; | |
1614 | insns[i + 1].imm = 0; | |
1615 | continue; | |
1616 | } | |
1617 | } | |
1618 | ||
1619 | return insns; | |
1620 | } | |
1621 | ||
1e270976 MKL |
1622 | static int bpf_prog_get_info_by_fd(struct bpf_prog *prog, |
1623 | const union bpf_attr *attr, | |
1624 | union bpf_attr __user *uattr) | |
1625 | { | |
1626 | struct bpf_prog_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
1627 | struct bpf_prog_info info = {}; | |
1628 | u32 info_len = attr->info.info_len; | |
1629 | char __user *uinsns; | |
1630 | u32 ulen; | |
1631 | int err; | |
1632 | ||
1633 | err = check_uarg_tail_zero(uinfo, sizeof(info), info_len); | |
1634 | if (err) | |
1635 | return err; | |
1636 | info_len = min_t(u32, sizeof(info), info_len); | |
1637 | ||
1638 | if (copy_from_user(&info, uinfo, info_len)) | |
89b09689 | 1639 | return -EFAULT; |
1e270976 MKL |
1640 | |
1641 | info.type = prog->type; | |
1642 | info.id = prog->aux->id; | |
cb4d2b3f MKL |
1643 | info.load_time = prog->aux->load_time; |
1644 | info.created_by_uid = from_kuid_munged(current_user_ns(), | |
1645 | prog->aux->user->uid); | |
1e270976 MKL |
1646 | |
1647 | memcpy(info.tag, prog->tag, sizeof(prog->tag)); | |
cb4d2b3f MKL |
1648 | memcpy(info.name, prog->aux->name, sizeof(prog->aux->name)); |
1649 | ||
1650 | ulen = info.nr_map_ids; | |
1651 | info.nr_map_ids = prog->aux->used_map_cnt; | |
1652 | ulen = min_t(u32, info.nr_map_ids, ulen); | |
1653 | if (ulen) { | |
721e08da | 1654 | u32 __user *user_map_ids = u64_to_user_ptr(info.map_ids); |
cb4d2b3f MKL |
1655 | u32 i; |
1656 | ||
1657 | for (i = 0; i < ulen; i++) | |
1658 | if (put_user(prog->aux->used_maps[i]->id, | |
1659 | &user_map_ids[i])) | |
1660 | return -EFAULT; | |
1661 | } | |
1e270976 MKL |
1662 | |
1663 | if (!capable(CAP_SYS_ADMIN)) { | |
1664 | info.jited_prog_len = 0; | |
1665 | info.xlated_prog_len = 0; | |
1666 | goto done; | |
1667 | } | |
1668 | ||
1669 | ulen = info.jited_prog_len; | |
1670 | info.jited_prog_len = prog->jited_len; | |
1671 | if (info.jited_prog_len && ulen) { | |
7105e828 DB |
1672 | if (bpf_dump_raw_ok()) { |
1673 | uinsns = u64_to_user_ptr(info.jited_prog_insns); | |
1674 | ulen = min_t(u32, info.jited_prog_len, ulen); | |
1675 | if (copy_to_user(uinsns, prog->bpf_func, ulen)) | |
1676 | return -EFAULT; | |
1677 | } else { | |
1678 | info.jited_prog_insns = 0; | |
1679 | } | |
1e270976 MKL |
1680 | } |
1681 | ||
1682 | ulen = info.xlated_prog_len; | |
9975a54b | 1683 | info.xlated_prog_len = bpf_prog_insn_size(prog); |
1e270976 | 1684 | if (info.xlated_prog_len && ulen) { |
7105e828 DB |
1685 | struct bpf_insn *insns_sanitized; |
1686 | bool fault; | |
1687 | ||
1688 | if (prog->blinded && !bpf_dump_raw_ok()) { | |
1689 | info.xlated_prog_insns = 0; | |
1690 | goto done; | |
1691 | } | |
1692 | insns_sanitized = bpf_insn_prepare_dump(prog); | |
1693 | if (!insns_sanitized) | |
1694 | return -ENOMEM; | |
1e270976 MKL |
1695 | uinsns = u64_to_user_ptr(info.xlated_prog_insns); |
1696 | ulen = min_t(u32, info.xlated_prog_len, ulen); | |
7105e828 DB |
1697 | fault = copy_to_user(uinsns, insns_sanitized, ulen); |
1698 | kfree(insns_sanitized); | |
1699 | if (fault) | |
1e270976 MKL |
1700 | return -EFAULT; |
1701 | } | |
1702 | ||
1703 | done: | |
1704 | if (copy_to_user(uinfo, &info, info_len) || | |
1705 | put_user(info_len, &uattr->info.info_len)) | |
1706 | return -EFAULT; | |
1707 | ||
1708 | return 0; | |
1709 | } | |
1710 | ||
1711 | static int bpf_map_get_info_by_fd(struct bpf_map *map, | |
1712 | const union bpf_attr *attr, | |
1713 | union bpf_attr __user *uattr) | |
1714 | { | |
1715 | struct bpf_map_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
1716 | struct bpf_map_info info = {}; | |
1717 | u32 info_len = attr->info.info_len; | |
1718 | int err; | |
1719 | ||
1720 | err = check_uarg_tail_zero(uinfo, sizeof(info), info_len); | |
1721 | if (err) | |
1722 | return err; | |
1723 | info_len = min_t(u32, sizeof(info), info_len); | |
1724 | ||
1725 | info.type = map->map_type; | |
1726 | info.id = map->id; | |
1727 | info.key_size = map->key_size; | |
1728 | info.value_size = map->value_size; | |
1729 | info.max_entries = map->max_entries; | |
1730 | info.map_flags = map->map_flags; | |
ad5b177b | 1731 | memcpy(info.name, map->name, sizeof(map->name)); |
1e270976 MKL |
1732 | |
1733 | if (copy_to_user(uinfo, &info, info_len) || | |
1734 | put_user(info_len, &uattr->info.info_len)) | |
1735 | return -EFAULT; | |
1736 | ||
1737 | return 0; | |
1738 | } | |
1739 | ||
1740 | #define BPF_OBJ_GET_INFO_BY_FD_LAST_FIELD info.info | |
1741 | ||
1742 | static int bpf_obj_get_info_by_fd(const union bpf_attr *attr, | |
1743 | union bpf_attr __user *uattr) | |
1744 | { | |
1745 | int ufd = attr->info.bpf_fd; | |
1746 | struct fd f; | |
1747 | int err; | |
1748 | ||
1749 | if (CHECK_ATTR(BPF_OBJ_GET_INFO_BY_FD)) | |
1750 | return -EINVAL; | |
1751 | ||
1752 | f = fdget(ufd); | |
1753 | if (!f.file) | |
1754 | return -EBADFD; | |
1755 | ||
1756 | if (f.file->f_op == &bpf_prog_fops) | |
1757 | err = bpf_prog_get_info_by_fd(f.file->private_data, attr, | |
1758 | uattr); | |
1759 | else if (f.file->f_op == &bpf_map_fops) | |
1760 | err = bpf_map_get_info_by_fd(f.file->private_data, attr, | |
1761 | uattr); | |
1762 | else | |
1763 | err = -EINVAL; | |
1764 | ||
1765 | fdput(f); | |
1766 | return err; | |
1767 | } | |
1768 | ||
99c55f7d AS |
1769 | SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size) |
1770 | { | |
1771 | union bpf_attr attr = {}; | |
1772 | int err; | |
1773 | ||
1be7f75d | 1774 | if (!capable(CAP_SYS_ADMIN) && sysctl_unprivileged_bpf_disabled) |
99c55f7d AS |
1775 | return -EPERM; |
1776 | ||
1e270976 MKL |
1777 | err = check_uarg_tail_zero(uattr, sizeof(attr), size); |
1778 | if (err) | |
1779 | return err; | |
1780 | size = min_t(u32, size, sizeof(attr)); | |
99c55f7d AS |
1781 | |
1782 | /* copy attributes from user space, may be less than sizeof(bpf_attr) */ | |
1783 | if (copy_from_user(&attr, uattr, size) != 0) | |
1784 | return -EFAULT; | |
1785 | ||
afdb09c7 CF |
1786 | err = security_bpf(cmd, &attr, size); |
1787 | if (err < 0) | |
1788 | return err; | |
1789 | ||
99c55f7d AS |
1790 | switch (cmd) { |
1791 | case BPF_MAP_CREATE: | |
1792 | err = map_create(&attr); | |
1793 | break; | |
db20fd2b AS |
1794 | case BPF_MAP_LOOKUP_ELEM: |
1795 | err = map_lookup_elem(&attr); | |
1796 | break; | |
1797 | case BPF_MAP_UPDATE_ELEM: | |
1798 | err = map_update_elem(&attr); | |
1799 | break; | |
1800 | case BPF_MAP_DELETE_ELEM: | |
1801 | err = map_delete_elem(&attr); | |
1802 | break; | |
1803 | case BPF_MAP_GET_NEXT_KEY: | |
1804 | err = map_get_next_key(&attr); | |
1805 | break; | |
09756af4 AS |
1806 | case BPF_PROG_LOAD: |
1807 | err = bpf_prog_load(&attr); | |
1808 | break; | |
b2197755 DB |
1809 | case BPF_OBJ_PIN: |
1810 | err = bpf_obj_pin(&attr); | |
1811 | break; | |
1812 | case BPF_OBJ_GET: | |
1813 | err = bpf_obj_get(&attr); | |
1814 | break; | |
f4324551 DM |
1815 | #ifdef CONFIG_CGROUP_BPF |
1816 | case BPF_PROG_ATTACH: | |
1817 | err = bpf_prog_attach(&attr); | |
1818 | break; | |
1819 | case BPF_PROG_DETACH: | |
1820 | err = bpf_prog_detach(&attr); | |
1821 | break; | |
468e2f64 AS |
1822 | case BPF_PROG_QUERY: |
1823 | err = bpf_prog_query(&attr, uattr); | |
1824 | break; | |
f4324551 | 1825 | #endif |
1cf1cae9 AS |
1826 | case BPF_PROG_TEST_RUN: |
1827 | err = bpf_prog_test_run(&attr, uattr); | |
1828 | break; | |
34ad5580 MKL |
1829 | case BPF_PROG_GET_NEXT_ID: |
1830 | err = bpf_obj_get_next_id(&attr, uattr, | |
1831 | &prog_idr, &prog_idr_lock); | |
1832 | break; | |
1833 | case BPF_MAP_GET_NEXT_ID: | |
1834 | err = bpf_obj_get_next_id(&attr, uattr, | |
1835 | &map_idr, &map_idr_lock); | |
1836 | break; | |
b16d9aa4 MKL |
1837 | case BPF_PROG_GET_FD_BY_ID: |
1838 | err = bpf_prog_get_fd_by_id(&attr); | |
1839 | break; | |
bd5f5f4e MKL |
1840 | case BPF_MAP_GET_FD_BY_ID: |
1841 | err = bpf_map_get_fd_by_id(&attr); | |
1842 | break; | |
1e270976 MKL |
1843 | case BPF_OBJ_GET_INFO_BY_FD: |
1844 | err = bpf_obj_get_info_by_fd(&attr, uattr); | |
1845 | break; | |
99c55f7d AS |
1846 | default: |
1847 | err = -EINVAL; | |
1848 | break; | |
1849 | } | |
1850 | ||
1851 | return err; | |
1852 | } |