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> |
f56a653c | 14 | #include <linux/btf.h> |
99c55f7d AS |
15 | #include <linux/syscalls.h> |
16 | #include <linux/slab.h> | |
3f07c014 | 17 | #include <linux/sched/signal.h> |
d407bd25 DB |
18 | #include <linux/vmalloc.h> |
19 | #include <linux/mmzone.h> | |
99c55f7d | 20 | #include <linux/anon_inodes.h> |
db20fd2b | 21 | #include <linux/file.h> |
09756af4 AS |
22 | #include <linux/license.h> |
23 | #include <linux/filter.h> | |
2541517c | 24 | #include <linux/version.h> |
535e7b4b | 25 | #include <linux/kernel.h> |
dc4bb0e2 | 26 | #include <linux/idr.h> |
cb4d2b3f MKL |
27 | #include <linux/cred.h> |
28 | #include <linux/timekeeping.h> | |
29 | #include <linux/ctype.h> | |
a26ca7c9 | 30 | #include <linux/btf.h> |
9ef09e35 | 31 | #include <linux/nospec.h> |
99c55f7d | 32 | |
14dc6f04 MKL |
33 | #define IS_FD_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PROG_ARRAY || \ |
34 | (map)->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || \ | |
35 | (map)->map_type == BPF_MAP_TYPE_CGROUP_ARRAY || \ | |
36 | (map)->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS) | |
37 | #define IS_FD_HASH(map) ((map)->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) | |
38 | #define IS_FD_MAP(map) (IS_FD_ARRAY(map) || IS_FD_HASH(map)) | |
39 | ||
6e71b04a CF |
40 | #define BPF_OBJ_FLAG_MASK (BPF_F_RDONLY | BPF_F_WRONLY) |
41 | ||
b121d1e7 | 42 | DEFINE_PER_CPU(int, bpf_prog_active); |
dc4bb0e2 MKL |
43 | static DEFINE_IDR(prog_idr); |
44 | static DEFINE_SPINLOCK(prog_idr_lock); | |
f3f1c054 MKL |
45 | static DEFINE_IDR(map_idr); |
46 | static DEFINE_SPINLOCK(map_idr_lock); | |
b121d1e7 | 47 | |
1be7f75d AS |
48 | int sysctl_unprivileged_bpf_disabled __read_mostly; |
49 | ||
40077e0c JB |
50 | static const struct bpf_map_ops * const bpf_map_types[] = { |
51 | #define BPF_PROG_TYPE(_id, _ops) | |
52 | #define BPF_MAP_TYPE(_id, _ops) \ | |
53 | [_id] = &_ops, | |
54 | #include <linux/bpf_types.h> | |
55 | #undef BPF_PROG_TYPE | |
56 | #undef BPF_MAP_TYPE | |
57 | }; | |
99c55f7d | 58 | |
752ba56f MS |
59 | /* |
60 | * If we're handed a bigger struct than we know of, ensure all the unknown bits | |
61 | * are 0 - i.e. new user-space does not rely on any kernel feature extensions | |
62 | * we don't know about yet. | |
63 | * | |
64 | * There is a ToCToU between this function call and the following | |
65 | * copy_from_user() call. However, this is not a concern since this function is | |
66 | * meant to be a future-proofing of bits. | |
67 | */ | |
dcab51f1 MKL |
68 | int bpf_check_uarg_tail_zero(void __user *uaddr, |
69 | size_t expected_size, | |
70 | size_t actual_size) | |
58291a74 MS |
71 | { |
72 | unsigned char __user *addr; | |
73 | unsigned char __user *end; | |
74 | unsigned char val; | |
75 | int err; | |
76 | ||
752ba56f MS |
77 | if (unlikely(actual_size > PAGE_SIZE)) /* silly large */ |
78 | return -E2BIG; | |
79 | ||
80 | if (unlikely(!access_ok(VERIFY_READ, uaddr, actual_size))) | |
81 | return -EFAULT; | |
82 | ||
58291a74 MS |
83 | if (actual_size <= expected_size) |
84 | return 0; | |
85 | ||
86 | addr = uaddr + expected_size; | |
87 | end = uaddr + actual_size; | |
88 | ||
89 | for (; addr < end; addr++) { | |
90 | err = get_user(val, addr); | |
91 | if (err) | |
92 | return err; | |
93 | if (val) | |
94 | return -E2BIG; | |
95 | } | |
96 | ||
97 | return 0; | |
98 | } | |
99 | ||
a3884572 JK |
100 | const struct bpf_map_ops bpf_map_offload_ops = { |
101 | .map_alloc = bpf_map_offload_map_alloc, | |
102 | .map_free = bpf_map_offload_map_free, | |
103 | }; | |
104 | ||
99c55f7d AS |
105 | static struct bpf_map *find_and_alloc_map(union bpf_attr *attr) |
106 | { | |
1110f3a9 | 107 | const struct bpf_map_ops *ops; |
9ef09e35 | 108 | u32 type = attr->map_type; |
99c55f7d | 109 | struct bpf_map *map; |
1110f3a9 | 110 | int err; |
99c55f7d | 111 | |
9ef09e35 | 112 | if (type >= ARRAY_SIZE(bpf_map_types)) |
1110f3a9 | 113 | return ERR_PTR(-EINVAL); |
9ef09e35 MR |
114 | type = array_index_nospec(type, ARRAY_SIZE(bpf_map_types)); |
115 | ops = bpf_map_types[type]; | |
1110f3a9 | 116 | if (!ops) |
40077e0c | 117 | return ERR_PTR(-EINVAL); |
99c55f7d | 118 | |
1110f3a9 JK |
119 | if (ops->map_alloc_check) { |
120 | err = ops->map_alloc_check(attr); | |
121 | if (err) | |
122 | return ERR_PTR(err); | |
123 | } | |
a3884572 JK |
124 | if (attr->map_ifindex) |
125 | ops = &bpf_map_offload_ops; | |
1110f3a9 | 126 | map = ops->map_alloc(attr); |
40077e0c JB |
127 | if (IS_ERR(map)) |
128 | return map; | |
1110f3a9 | 129 | map->ops = ops; |
9ef09e35 | 130 | map->map_type = type; |
40077e0c | 131 | return map; |
99c55f7d AS |
132 | } |
133 | ||
96eabe7a | 134 | void *bpf_map_area_alloc(size_t size, int numa_node) |
d407bd25 DB |
135 | { |
136 | /* We definitely need __GFP_NORETRY, so OOM killer doesn't | |
137 | * trigger under memory pressure as we really just want to | |
138 | * fail instead. | |
139 | */ | |
140 | const gfp_t flags = __GFP_NOWARN | __GFP_NORETRY | __GFP_ZERO; | |
141 | void *area; | |
142 | ||
143 | if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) { | |
96eabe7a | 144 | area = kmalloc_node(size, GFP_USER | flags, numa_node); |
d407bd25 DB |
145 | if (area != NULL) |
146 | return area; | |
147 | } | |
148 | ||
96eabe7a MKL |
149 | return __vmalloc_node_flags_caller(size, numa_node, GFP_KERNEL | flags, |
150 | __builtin_return_address(0)); | |
d407bd25 DB |
151 | } |
152 | ||
153 | void bpf_map_area_free(void *area) | |
154 | { | |
155 | kvfree(area); | |
156 | } | |
157 | ||
bd475643 JK |
158 | void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr) |
159 | { | |
160 | map->map_type = attr->map_type; | |
161 | map->key_size = attr->key_size; | |
162 | map->value_size = attr->value_size; | |
163 | map->max_entries = attr->max_entries; | |
164 | map->map_flags = attr->map_flags; | |
165 | map->numa_node = bpf_map_attr_numa_node(attr); | |
166 | } | |
167 | ||
6c905981 AS |
168 | int bpf_map_precharge_memlock(u32 pages) |
169 | { | |
170 | struct user_struct *user = get_current_user(); | |
171 | unsigned long memlock_limit, cur; | |
172 | ||
173 | memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
174 | cur = atomic_long_read(&user->locked_vm); | |
175 | free_uid(user); | |
176 | if (cur + pages > memlock_limit) | |
177 | return -EPERM; | |
178 | return 0; | |
179 | } | |
180 | ||
aaac3ba9 AS |
181 | static int bpf_map_charge_memlock(struct bpf_map *map) |
182 | { | |
183 | struct user_struct *user = get_current_user(); | |
184 | unsigned long memlock_limit; | |
185 | ||
186 | memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
187 | ||
188 | atomic_long_add(map->pages, &user->locked_vm); | |
189 | ||
190 | if (atomic_long_read(&user->locked_vm) > memlock_limit) { | |
191 | atomic_long_sub(map->pages, &user->locked_vm); | |
192 | free_uid(user); | |
193 | return -EPERM; | |
194 | } | |
195 | map->user = user; | |
196 | return 0; | |
197 | } | |
198 | ||
199 | static void bpf_map_uncharge_memlock(struct bpf_map *map) | |
200 | { | |
201 | struct user_struct *user = map->user; | |
202 | ||
203 | atomic_long_sub(map->pages, &user->locked_vm); | |
204 | free_uid(user); | |
205 | } | |
206 | ||
f3f1c054 MKL |
207 | static int bpf_map_alloc_id(struct bpf_map *map) |
208 | { | |
209 | int id; | |
210 | ||
b76354cd | 211 | idr_preload(GFP_KERNEL); |
f3f1c054 MKL |
212 | spin_lock_bh(&map_idr_lock); |
213 | id = idr_alloc_cyclic(&map_idr, map, 1, INT_MAX, GFP_ATOMIC); | |
214 | if (id > 0) | |
215 | map->id = id; | |
216 | spin_unlock_bh(&map_idr_lock); | |
b76354cd | 217 | idr_preload_end(); |
f3f1c054 MKL |
218 | |
219 | if (WARN_ON_ONCE(!id)) | |
220 | return -ENOSPC; | |
221 | ||
222 | return id > 0 ? 0 : id; | |
223 | } | |
224 | ||
a3884572 | 225 | void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock) |
f3f1c054 | 226 | { |
930651a7 ED |
227 | unsigned long flags; |
228 | ||
a3884572 JK |
229 | /* Offloaded maps are removed from the IDR store when their device |
230 | * disappears - even if someone holds an fd to them they are unusable, | |
231 | * the memory is gone, all ops will fail; they are simply waiting for | |
232 | * refcnt to drop to be freed. | |
233 | */ | |
234 | if (!map->id) | |
235 | return; | |
236 | ||
bd5f5f4e | 237 | if (do_idr_lock) |
930651a7 | 238 | spin_lock_irqsave(&map_idr_lock, flags); |
bd5f5f4e MKL |
239 | else |
240 | __acquire(&map_idr_lock); | |
241 | ||
f3f1c054 | 242 | idr_remove(&map_idr, map->id); |
a3884572 | 243 | map->id = 0; |
bd5f5f4e MKL |
244 | |
245 | if (do_idr_lock) | |
930651a7 | 246 | spin_unlock_irqrestore(&map_idr_lock, flags); |
bd5f5f4e MKL |
247 | else |
248 | __release(&map_idr_lock); | |
f3f1c054 MKL |
249 | } |
250 | ||
99c55f7d AS |
251 | /* called from workqueue */ |
252 | static void bpf_map_free_deferred(struct work_struct *work) | |
253 | { | |
254 | struct bpf_map *map = container_of(work, struct bpf_map, work); | |
255 | ||
aaac3ba9 | 256 | bpf_map_uncharge_memlock(map); |
afdb09c7 | 257 | security_bpf_map_free(map); |
99c55f7d AS |
258 | /* implementation dependent freeing */ |
259 | map->ops->map_free(map); | |
260 | } | |
261 | ||
c9da161c DB |
262 | static void bpf_map_put_uref(struct bpf_map *map) |
263 | { | |
264 | if (atomic_dec_and_test(&map->usercnt)) { | |
ba6b8de4 JF |
265 | if (map->ops->map_release_uref) |
266 | map->ops->map_release_uref(map); | |
c9da161c DB |
267 | } |
268 | } | |
269 | ||
99c55f7d AS |
270 | /* decrement map refcnt and schedule it for freeing via workqueue |
271 | * (unrelying map implementation ops->map_free() might sleep) | |
272 | */ | |
bd5f5f4e | 273 | static void __bpf_map_put(struct bpf_map *map, bool do_idr_lock) |
99c55f7d AS |
274 | { |
275 | if (atomic_dec_and_test(&map->refcnt)) { | |
34ad5580 | 276 | /* bpf_map_free_id() must be called first */ |
bd5f5f4e | 277 | bpf_map_free_id(map, do_idr_lock); |
78958fca | 278 | btf_put(map->btf); |
99c55f7d AS |
279 | INIT_WORK(&map->work, bpf_map_free_deferred); |
280 | schedule_work(&map->work); | |
281 | } | |
282 | } | |
283 | ||
bd5f5f4e MKL |
284 | void bpf_map_put(struct bpf_map *map) |
285 | { | |
286 | __bpf_map_put(map, true); | |
287 | } | |
630a4d38 | 288 | EXPORT_SYMBOL_GPL(bpf_map_put); |
bd5f5f4e | 289 | |
c9da161c | 290 | void bpf_map_put_with_uref(struct bpf_map *map) |
99c55f7d | 291 | { |
c9da161c | 292 | bpf_map_put_uref(map); |
99c55f7d | 293 | bpf_map_put(map); |
c9da161c DB |
294 | } |
295 | ||
296 | static int bpf_map_release(struct inode *inode, struct file *filp) | |
297 | { | |
61d1b6a4 DB |
298 | struct bpf_map *map = filp->private_data; |
299 | ||
300 | if (map->ops->map_release) | |
301 | map->ops->map_release(map, filp); | |
302 | ||
303 | bpf_map_put_with_uref(map); | |
99c55f7d AS |
304 | return 0; |
305 | } | |
306 | ||
f99bf205 DB |
307 | #ifdef CONFIG_PROC_FS |
308 | static void bpf_map_show_fdinfo(struct seq_file *m, struct file *filp) | |
309 | { | |
310 | const struct bpf_map *map = filp->private_data; | |
21116b70 DB |
311 | const struct bpf_array *array; |
312 | u32 owner_prog_type = 0; | |
9780c0ab | 313 | u32 owner_jited = 0; |
21116b70 DB |
314 | |
315 | if (map->map_type == BPF_MAP_TYPE_PROG_ARRAY) { | |
316 | array = container_of(map, struct bpf_array, map); | |
317 | owner_prog_type = array->owner_prog_type; | |
9780c0ab | 318 | owner_jited = array->owner_jited; |
21116b70 | 319 | } |
f99bf205 DB |
320 | |
321 | seq_printf(m, | |
322 | "map_type:\t%u\n" | |
323 | "key_size:\t%u\n" | |
324 | "value_size:\t%u\n" | |
322cea2f | 325 | "max_entries:\t%u\n" |
21116b70 DB |
326 | "map_flags:\t%#x\n" |
327 | "memlock:\t%llu\n", | |
f99bf205 DB |
328 | map->map_type, |
329 | map->key_size, | |
330 | map->value_size, | |
322cea2f | 331 | map->max_entries, |
21116b70 DB |
332 | map->map_flags, |
333 | map->pages * 1ULL << PAGE_SHIFT); | |
334 | ||
9780c0ab | 335 | if (owner_prog_type) { |
21116b70 DB |
336 | seq_printf(m, "owner_prog_type:\t%u\n", |
337 | owner_prog_type); | |
9780c0ab DB |
338 | seq_printf(m, "owner_jited:\t%u\n", |
339 | owner_jited); | |
340 | } | |
f99bf205 DB |
341 | } |
342 | #endif | |
343 | ||
6e71b04a CF |
344 | static ssize_t bpf_dummy_read(struct file *filp, char __user *buf, size_t siz, |
345 | loff_t *ppos) | |
346 | { | |
347 | /* We need this handler such that alloc_file() enables | |
348 | * f_mode with FMODE_CAN_READ. | |
349 | */ | |
350 | return -EINVAL; | |
351 | } | |
352 | ||
353 | static ssize_t bpf_dummy_write(struct file *filp, const char __user *buf, | |
354 | size_t siz, loff_t *ppos) | |
355 | { | |
356 | /* We need this handler such that alloc_file() enables | |
357 | * f_mode with FMODE_CAN_WRITE. | |
358 | */ | |
359 | return -EINVAL; | |
360 | } | |
361 | ||
f66e448c | 362 | const struct file_operations bpf_map_fops = { |
f99bf205 DB |
363 | #ifdef CONFIG_PROC_FS |
364 | .show_fdinfo = bpf_map_show_fdinfo, | |
365 | #endif | |
366 | .release = bpf_map_release, | |
6e71b04a CF |
367 | .read = bpf_dummy_read, |
368 | .write = bpf_dummy_write, | |
99c55f7d AS |
369 | }; |
370 | ||
6e71b04a | 371 | int bpf_map_new_fd(struct bpf_map *map, int flags) |
aa79781b | 372 | { |
afdb09c7 CF |
373 | int ret; |
374 | ||
375 | ret = security_bpf_map(map, OPEN_FMODE(flags)); | |
376 | if (ret < 0) | |
377 | return ret; | |
378 | ||
aa79781b | 379 | return anon_inode_getfd("bpf-map", &bpf_map_fops, map, |
6e71b04a CF |
380 | flags | O_CLOEXEC); |
381 | } | |
382 | ||
383 | int bpf_get_file_flag(int flags) | |
384 | { | |
385 | if ((flags & BPF_F_RDONLY) && (flags & BPF_F_WRONLY)) | |
386 | return -EINVAL; | |
387 | if (flags & BPF_F_RDONLY) | |
388 | return O_RDONLY; | |
389 | if (flags & BPF_F_WRONLY) | |
390 | return O_WRONLY; | |
391 | return O_RDWR; | |
aa79781b DB |
392 | } |
393 | ||
99c55f7d AS |
394 | /* helper macro to check that unused fields 'union bpf_attr' are zero */ |
395 | #define CHECK_ATTR(CMD) \ | |
396 | memchr_inv((void *) &attr->CMD##_LAST_FIELD + \ | |
397 | sizeof(attr->CMD##_LAST_FIELD), 0, \ | |
398 | sizeof(*attr) - \ | |
399 | offsetof(union bpf_attr, CMD##_LAST_FIELD) - \ | |
400 | sizeof(attr->CMD##_LAST_FIELD)) != NULL | |
401 | ||
cb4d2b3f MKL |
402 | /* dst and src must have at least BPF_OBJ_NAME_LEN number of bytes. |
403 | * Return 0 on success and < 0 on error. | |
404 | */ | |
405 | static int bpf_obj_name_cpy(char *dst, const char *src) | |
406 | { | |
407 | const char *end = src + BPF_OBJ_NAME_LEN; | |
408 | ||
473d9734 MKL |
409 | memset(dst, 0, BPF_OBJ_NAME_LEN); |
410 | ||
cb4d2b3f MKL |
411 | /* Copy all isalnum() and '_' char */ |
412 | while (src < end && *src) { | |
413 | if (!isalnum(*src) && *src != '_') | |
414 | return -EINVAL; | |
415 | *dst++ = *src++; | |
416 | } | |
417 | ||
418 | /* No '\0' found in BPF_OBJ_NAME_LEN number of bytes */ | |
419 | if (src == end) | |
420 | return -EINVAL; | |
421 | ||
cb4d2b3f MKL |
422 | return 0; |
423 | } | |
424 | ||
9b2cf328 | 425 | #define BPF_MAP_CREATE_LAST_FIELD btf_value_type_id |
99c55f7d AS |
426 | /* called via syscall */ |
427 | static int map_create(union bpf_attr *attr) | |
428 | { | |
96eabe7a | 429 | int numa_node = bpf_map_attr_numa_node(attr); |
99c55f7d | 430 | struct bpf_map *map; |
6e71b04a | 431 | int f_flags; |
99c55f7d AS |
432 | int err; |
433 | ||
434 | err = CHECK_ATTR(BPF_MAP_CREATE); | |
435 | if (err) | |
436 | return -EINVAL; | |
437 | ||
6e71b04a CF |
438 | f_flags = bpf_get_file_flag(attr->map_flags); |
439 | if (f_flags < 0) | |
440 | return f_flags; | |
441 | ||
96eabe7a | 442 | if (numa_node != NUMA_NO_NODE && |
96e5ae4e ED |
443 | ((unsigned int)numa_node >= nr_node_ids || |
444 | !node_online(numa_node))) | |
96eabe7a MKL |
445 | return -EINVAL; |
446 | ||
99c55f7d AS |
447 | /* find map type and init map: hashtable vs rbtree vs bloom vs ... */ |
448 | map = find_and_alloc_map(attr); | |
449 | if (IS_ERR(map)) | |
450 | return PTR_ERR(map); | |
451 | ||
ad5b177b MKL |
452 | err = bpf_obj_name_cpy(map->name, attr->map_name); |
453 | if (err) | |
454 | goto free_map_nouncharge; | |
455 | ||
99c55f7d | 456 | atomic_set(&map->refcnt, 1); |
c9da161c | 457 | atomic_set(&map->usercnt, 1); |
99c55f7d | 458 | |
a26ca7c9 | 459 | if (bpf_map_support_seq_show(map) && |
9b2cf328 | 460 | (attr->btf_key_type_id || attr->btf_value_type_id)) { |
a26ca7c9 MKL |
461 | struct btf *btf; |
462 | ||
9b2cf328 | 463 | if (!attr->btf_key_type_id || !attr->btf_value_type_id) { |
a26ca7c9 MKL |
464 | err = -EINVAL; |
465 | goto free_map_nouncharge; | |
466 | } | |
467 | ||
468 | btf = btf_get_by_fd(attr->btf_fd); | |
469 | if (IS_ERR(btf)) { | |
470 | err = PTR_ERR(btf); | |
471 | goto free_map_nouncharge; | |
472 | } | |
473 | ||
9b2cf328 MKL |
474 | err = map->ops->map_check_btf(map, btf, attr->btf_key_type_id, |
475 | attr->btf_value_type_id); | |
a26ca7c9 MKL |
476 | if (err) { |
477 | btf_put(btf); | |
478 | goto free_map_nouncharge; | |
479 | } | |
480 | ||
481 | map->btf = btf; | |
9b2cf328 MKL |
482 | map->btf_key_type_id = attr->btf_key_type_id; |
483 | map->btf_value_type_id = attr->btf_value_type_id; | |
a26ca7c9 MKL |
484 | } |
485 | ||
afdb09c7 | 486 | err = security_bpf_map_alloc(map); |
aaac3ba9 | 487 | if (err) |
20b2b24f | 488 | goto free_map_nouncharge; |
aaac3ba9 | 489 | |
afdb09c7 CF |
490 | err = bpf_map_charge_memlock(map); |
491 | if (err) | |
492 | goto free_map_sec; | |
493 | ||
f3f1c054 MKL |
494 | err = bpf_map_alloc_id(map); |
495 | if (err) | |
496 | goto free_map; | |
497 | ||
6e71b04a | 498 | err = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
499 | if (err < 0) { |
500 | /* failed to allocate fd. | |
501 | * bpf_map_put() is needed because the above | |
502 | * bpf_map_alloc_id() has published the map | |
503 | * to the userspace and the userspace may | |
504 | * have refcnt-ed it through BPF_MAP_GET_FD_BY_ID. | |
505 | */ | |
506 | bpf_map_put(map); | |
507 | return err; | |
508 | } | |
99c55f7d AS |
509 | |
510 | return err; | |
511 | ||
512 | free_map: | |
20b2b24f | 513 | bpf_map_uncharge_memlock(map); |
afdb09c7 CF |
514 | free_map_sec: |
515 | security_bpf_map_free(map); | |
20b2b24f | 516 | free_map_nouncharge: |
a26ca7c9 | 517 | btf_put(map->btf); |
99c55f7d AS |
518 | map->ops->map_free(map); |
519 | return err; | |
520 | } | |
521 | ||
db20fd2b AS |
522 | /* if error is returned, fd is released. |
523 | * On success caller should complete fd access with matching fdput() | |
524 | */ | |
c2101297 | 525 | struct bpf_map *__bpf_map_get(struct fd f) |
db20fd2b | 526 | { |
db20fd2b AS |
527 | if (!f.file) |
528 | return ERR_PTR(-EBADF); | |
db20fd2b AS |
529 | if (f.file->f_op != &bpf_map_fops) { |
530 | fdput(f); | |
531 | return ERR_PTR(-EINVAL); | |
532 | } | |
533 | ||
c2101297 DB |
534 | return f.file->private_data; |
535 | } | |
536 | ||
92117d84 AS |
537 | /* prog's and map's refcnt limit */ |
538 | #define BPF_MAX_REFCNT 32768 | |
539 | ||
540 | struct bpf_map *bpf_map_inc(struct bpf_map *map, bool uref) | |
c9da161c | 541 | { |
92117d84 AS |
542 | if (atomic_inc_return(&map->refcnt) > BPF_MAX_REFCNT) { |
543 | atomic_dec(&map->refcnt); | |
544 | return ERR_PTR(-EBUSY); | |
545 | } | |
c9da161c DB |
546 | if (uref) |
547 | atomic_inc(&map->usercnt); | |
92117d84 | 548 | return map; |
c9da161c | 549 | } |
630a4d38 | 550 | EXPORT_SYMBOL_GPL(bpf_map_inc); |
c9da161c DB |
551 | |
552 | struct bpf_map *bpf_map_get_with_uref(u32 ufd) | |
c2101297 DB |
553 | { |
554 | struct fd f = fdget(ufd); | |
555 | struct bpf_map *map; | |
556 | ||
557 | map = __bpf_map_get(f); | |
558 | if (IS_ERR(map)) | |
559 | return map; | |
560 | ||
92117d84 | 561 | map = bpf_map_inc(map, true); |
c2101297 | 562 | fdput(f); |
db20fd2b AS |
563 | |
564 | return map; | |
565 | } | |
566 | ||
bd5f5f4e MKL |
567 | /* map_idr_lock should have been held */ |
568 | static struct bpf_map *bpf_map_inc_not_zero(struct bpf_map *map, | |
569 | bool uref) | |
570 | { | |
571 | int refold; | |
572 | ||
573 | refold = __atomic_add_unless(&map->refcnt, 1, 0); | |
574 | ||
575 | if (refold >= BPF_MAX_REFCNT) { | |
576 | __bpf_map_put(map, false); | |
577 | return ERR_PTR(-EBUSY); | |
578 | } | |
579 | ||
580 | if (!refold) | |
581 | return ERR_PTR(-ENOENT); | |
582 | ||
583 | if (uref) | |
584 | atomic_inc(&map->usercnt); | |
585 | ||
586 | return map; | |
587 | } | |
588 | ||
b8cdc051 AS |
589 | int __weak bpf_stackmap_copy(struct bpf_map *map, void *key, void *value) |
590 | { | |
591 | return -ENOTSUPP; | |
592 | } | |
593 | ||
db20fd2b AS |
594 | /* last field in 'union bpf_attr' used by this command */ |
595 | #define BPF_MAP_LOOKUP_ELEM_LAST_FIELD value | |
596 | ||
597 | static int map_lookup_elem(union bpf_attr *attr) | |
598 | { | |
535e7b4b MS |
599 | void __user *ukey = u64_to_user_ptr(attr->key); |
600 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 601 | int ufd = attr->map_fd; |
db20fd2b | 602 | struct bpf_map *map; |
8ebe667c | 603 | void *key, *value, *ptr; |
15a07b33 | 604 | u32 value_size; |
592867bf | 605 | struct fd f; |
db20fd2b AS |
606 | int err; |
607 | ||
608 | if (CHECK_ATTR(BPF_MAP_LOOKUP_ELEM)) | |
609 | return -EINVAL; | |
610 | ||
592867bf | 611 | f = fdget(ufd); |
c2101297 | 612 | map = __bpf_map_get(f); |
db20fd2b AS |
613 | if (IS_ERR(map)) |
614 | return PTR_ERR(map); | |
615 | ||
6e71b04a CF |
616 | if (!(f.file->f_mode & FMODE_CAN_READ)) { |
617 | err = -EPERM; | |
618 | goto err_put; | |
619 | } | |
620 | ||
e4448ed8 AV |
621 | key = memdup_user(ukey, map->key_size); |
622 | if (IS_ERR(key)) { | |
623 | err = PTR_ERR(key); | |
db20fd2b | 624 | goto err_put; |
e4448ed8 | 625 | } |
db20fd2b | 626 | |
15a07b33 | 627 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
8f844938 | 628 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || |
15a07b33 AS |
629 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) |
630 | value_size = round_up(map->value_size, 8) * num_possible_cpus(); | |
14dc6f04 MKL |
631 | else if (IS_FD_MAP(map)) |
632 | value_size = sizeof(u32); | |
15a07b33 AS |
633 | else |
634 | value_size = map->value_size; | |
635 | ||
8ebe667c | 636 | err = -ENOMEM; |
15a07b33 | 637 | value = kmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b | 638 | if (!value) |
8ebe667c AS |
639 | goto free_key; |
640 | ||
a3884572 JK |
641 | if (bpf_map_is_dev_bound(map)) { |
642 | err = bpf_map_offload_lookup_elem(map, key, value); | |
643 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || | |
644 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
15a07b33 AS |
645 | err = bpf_percpu_hash_copy(map, key, value); |
646 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
647 | err = bpf_percpu_array_copy(map, key, value); | |
557c0c6e AS |
648 | } else if (map->map_type == BPF_MAP_TYPE_STACK_TRACE) { |
649 | err = bpf_stackmap_copy(map, key, value); | |
14dc6f04 MKL |
650 | } else if (IS_FD_ARRAY(map)) { |
651 | err = bpf_fd_array_map_lookup_elem(map, key, value); | |
652 | } else if (IS_FD_HASH(map)) { | |
653 | err = bpf_fd_htab_map_lookup_elem(map, key, value); | |
15a07b33 AS |
654 | } else { |
655 | rcu_read_lock(); | |
656 | ptr = map->ops->map_lookup_elem(map, key); | |
657 | if (ptr) | |
658 | memcpy(value, ptr, value_size); | |
659 | rcu_read_unlock(); | |
660 | err = ptr ? 0 : -ENOENT; | |
661 | } | |
8ebe667c | 662 | |
15a07b33 | 663 | if (err) |
8ebe667c | 664 | goto free_value; |
db20fd2b AS |
665 | |
666 | err = -EFAULT; | |
15a07b33 | 667 | if (copy_to_user(uvalue, value, value_size) != 0) |
8ebe667c | 668 | goto free_value; |
db20fd2b AS |
669 | |
670 | err = 0; | |
671 | ||
8ebe667c AS |
672 | free_value: |
673 | kfree(value); | |
db20fd2b AS |
674 | free_key: |
675 | kfree(key); | |
676 | err_put: | |
677 | fdput(f); | |
678 | return err; | |
679 | } | |
680 | ||
3274f520 | 681 | #define BPF_MAP_UPDATE_ELEM_LAST_FIELD flags |
db20fd2b AS |
682 | |
683 | static int map_update_elem(union bpf_attr *attr) | |
684 | { | |
535e7b4b MS |
685 | void __user *ukey = u64_to_user_ptr(attr->key); |
686 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 687 | int ufd = attr->map_fd; |
db20fd2b AS |
688 | struct bpf_map *map; |
689 | void *key, *value; | |
15a07b33 | 690 | u32 value_size; |
592867bf | 691 | struct fd f; |
db20fd2b AS |
692 | int err; |
693 | ||
694 | if (CHECK_ATTR(BPF_MAP_UPDATE_ELEM)) | |
695 | return -EINVAL; | |
696 | ||
592867bf | 697 | f = fdget(ufd); |
c2101297 | 698 | map = __bpf_map_get(f); |
db20fd2b AS |
699 | if (IS_ERR(map)) |
700 | return PTR_ERR(map); | |
701 | ||
6e71b04a CF |
702 | if (!(f.file->f_mode & FMODE_CAN_WRITE)) { |
703 | err = -EPERM; | |
704 | goto err_put; | |
705 | } | |
706 | ||
e4448ed8 AV |
707 | key = memdup_user(ukey, map->key_size); |
708 | if (IS_ERR(key)) { | |
709 | err = PTR_ERR(key); | |
db20fd2b | 710 | goto err_put; |
e4448ed8 | 711 | } |
db20fd2b | 712 | |
15a07b33 | 713 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
8f844938 | 714 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || |
15a07b33 AS |
715 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) |
716 | value_size = round_up(map->value_size, 8) * num_possible_cpus(); | |
717 | else | |
718 | value_size = map->value_size; | |
719 | ||
db20fd2b | 720 | err = -ENOMEM; |
15a07b33 | 721 | value = kmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b AS |
722 | if (!value) |
723 | goto free_key; | |
724 | ||
725 | err = -EFAULT; | |
15a07b33 | 726 | if (copy_from_user(value, uvalue, value_size) != 0) |
db20fd2b AS |
727 | goto free_value; |
728 | ||
6710e112 | 729 | /* Need to create a kthread, thus must support schedule */ |
a3884572 JK |
730 | if (bpf_map_is_dev_bound(map)) { |
731 | err = bpf_map_offload_update_elem(map, key, value, attr->flags); | |
732 | goto out; | |
733 | } else if (map->map_type == BPF_MAP_TYPE_CPUMAP) { | |
6710e112 JDB |
734 | err = map->ops->map_update_elem(map, key, value, attr->flags); |
735 | goto out; | |
736 | } | |
737 | ||
b121d1e7 AS |
738 | /* must increment bpf_prog_active to avoid kprobe+bpf triggering from |
739 | * inside bpf map update or delete otherwise deadlocks are possible | |
740 | */ | |
741 | preempt_disable(); | |
742 | __this_cpu_inc(bpf_prog_active); | |
8f844938 MKL |
743 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
744 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
15a07b33 AS |
745 | err = bpf_percpu_hash_update(map, key, value, attr->flags); |
746 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
747 | err = bpf_percpu_array_update(map, key, value, attr->flags); | |
9c147b56 | 748 | } else if (IS_FD_ARRAY(map)) { |
d056a788 DB |
749 | rcu_read_lock(); |
750 | err = bpf_fd_array_map_update_elem(map, f.file, key, value, | |
751 | attr->flags); | |
752 | rcu_read_unlock(); | |
bcc6b1b7 MKL |
753 | } else if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) { |
754 | rcu_read_lock(); | |
755 | err = bpf_fd_htab_map_update_elem(map, f.file, key, value, | |
756 | attr->flags); | |
757 | rcu_read_unlock(); | |
15a07b33 AS |
758 | } else { |
759 | rcu_read_lock(); | |
760 | err = map->ops->map_update_elem(map, key, value, attr->flags); | |
761 | rcu_read_unlock(); | |
762 | } | |
b121d1e7 AS |
763 | __this_cpu_dec(bpf_prog_active); |
764 | preempt_enable(); | |
6710e112 | 765 | out: |
db20fd2b AS |
766 | free_value: |
767 | kfree(value); | |
768 | free_key: | |
769 | kfree(key); | |
770 | err_put: | |
771 | fdput(f); | |
772 | return err; | |
773 | } | |
774 | ||
775 | #define BPF_MAP_DELETE_ELEM_LAST_FIELD key | |
776 | ||
777 | static int map_delete_elem(union bpf_attr *attr) | |
778 | { | |
535e7b4b | 779 | void __user *ukey = u64_to_user_ptr(attr->key); |
db20fd2b | 780 | int ufd = attr->map_fd; |
db20fd2b | 781 | struct bpf_map *map; |
592867bf | 782 | struct fd f; |
db20fd2b AS |
783 | void *key; |
784 | int err; | |
785 | ||
786 | if (CHECK_ATTR(BPF_MAP_DELETE_ELEM)) | |
787 | return -EINVAL; | |
788 | ||
592867bf | 789 | f = fdget(ufd); |
c2101297 | 790 | map = __bpf_map_get(f); |
db20fd2b AS |
791 | if (IS_ERR(map)) |
792 | return PTR_ERR(map); | |
793 | ||
6e71b04a CF |
794 | if (!(f.file->f_mode & FMODE_CAN_WRITE)) { |
795 | err = -EPERM; | |
796 | goto err_put; | |
797 | } | |
798 | ||
e4448ed8 AV |
799 | key = memdup_user(ukey, map->key_size); |
800 | if (IS_ERR(key)) { | |
801 | err = PTR_ERR(key); | |
db20fd2b | 802 | goto err_put; |
e4448ed8 | 803 | } |
db20fd2b | 804 | |
a3884572 JK |
805 | if (bpf_map_is_dev_bound(map)) { |
806 | err = bpf_map_offload_delete_elem(map, key); | |
807 | goto out; | |
808 | } | |
809 | ||
b121d1e7 AS |
810 | preempt_disable(); |
811 | __this_cpu_inc(bpf_prog_active); | |
db20fd2b AS |
812 | rcu_read_lock(); |
813 | err = map->ops->map_delete_elem(map, key); | |
814 | rcu_read_unlock(); | |
b121d1e7 AS |
815 | __this_cpu_dec(bpf_prog_active); |
816 | preempt_enable(); | |
a3884572 | 817 | out: |
db20fd2b AS |
818 | kfree(key); |
819 | err_put: | |
820 | fdput(f); | |
821 | return err; | |
822 | } | |
823 | ||
824 | /* last field in 'union bpf_attr' used by this command */ | |
825 | #define BPF_MAP_GET_NEXT_KEY_LAST_FIELD next_key | |
826 | ||
827 | static int map_get_next_key(union bpf_attr *attr) | |
828 | { | |
535e7b4b MS |
829 | void __user *ukey = u64_to_user_ptr(attr->key); |
830 | void __user *unext_key = u64_to_user_ptr(attr->next_key); | |
db20fd2b | 831 | int ufd = attr->map_fd; |
db20fd2b AS |
832 | struct bpf_map *map; |
833 | void *key, *next_key; | |
592867bf | 834 | struct fd f; |
db20fd2b AS |
835 | int err; |
836 | ||
837 | if (CHECK_ATTR(BPF_MAP_GET_NEXT_KEY)) | |
838 | return -EINVAL; | |
839 | ||
592867bf | 840 | f = fdget(ufd); |
c2101297 | 841 | map = __bpf_map_get(f); |
db20fd2b AS |
842 | if (IS_ERR(map)) |
843 | return PTR_ERR(map); | |
844 | ||
6e71b04a CF |
845 | if (!(f.file->f_mode & FMODE_CAN_READ)) { |
846 | err = -EPERM; | |
847 | goto err_put; | |
848 | } | |
849 | ||
8fe45924 | 850 | if (ukey) { |
e4448ed8 AV |
851 | key = memdup_user(ukey, map->key_size); |
852 | if (IS_ERR(key)) { | |
853 | err = PTR_ERR(key); | |
8fe45924 | 854 | goto err_put; |
e4448ed8 | 855 | } |
8fe45924 TQ |
856 | } else { |
857 | key = NULL; | |
858 | } | |
db20fd2b AS |
859 | |
860 | err = -ENOMEM; | |
861 | next_key = kmalloc(map->key_size, GFP_USER); | |
862 | if (!next_key) | |
863 | goto free_key; | |
864 | ||
a3884572 JK |
865 | if (bpf_map_is_dev_bound(map)) { |
866 | err = bpf_map_offload_get_next_key(map, key, next_key); | |
867 | goto out; | |
868 | } | |
869 | ||
db20fd2b AS |
870 | rcu_read_lock(); |
871 | err = map->ops->map_get_next_key(map, key, next_key); | |
872 | rcu_read_unlock(); | |
a3884572 | 873 | out: |
db20fd2b AS |
874 | if (err) |
875 | goto free_next_key; | |
876 | ||
877 | err = -EFAULT; | |
878 | if (copy_to_user(unext_key, next_key, map->key_size) != 0) | |
879 | goto free_next_key; | |
880 | ||
881 | err = 0; | |
882 | ||
883 | free_next_key: | |
884 | kfree(next_key); | |
885 | free_key: | |
886 | kfree(key); | |
887 | err_put: | |
888 | fdput(f); | |
889 | return err; | |
890 | } | |
891 | ||
7de16e3a JK |
892 | static const struct bpf_prog_ops * const bpf_prog_types[] = { |
893 | #define BPF_PROG_TYPE(_id, _name) \ | |
894 | [_id] = & _name ## _prog_ops, | |
895 | #define BPF_MAP_TYPE(_id, _ops) | |
896 | #include <linux/bpf_types.h> | |
897 | #undef BPF_PROG_TYPE | |
898 | #undef BPF_MAP_TYPE | |
899 | }; | |
900 | ||
09756af4 AS |
901 | static int find_prog_type(enum bpf_prog_type type, struct bpf_prog *prog) |
902 | { | |
d0f1a451 DB |
903 | const struct bpf_prog_ops *ops; |
904 | ||
905 | if (type >= ARRAY_SIZE(bpf_prog_types)) | |
906 | return -EINVAL; | |
907 | type = array_index_nospec(type, ARRAY_SIZE(bpf_prog_types)); | |
908 | ops = bpf_prog_types[type]; | |
909 | if (!ops) | |
be9370a7 | 910 | return -EINVAL; |
09756af4 | 911 | |
ab3f0063 | 912 | if (!bpf_prog_is_dev_bound(prog->aux)) |
d0f1a451 | 913 | prog->aux->ops = ops; |
ab3f0063 JK |
914 | else |
915 | prog->aux->ops = &bpf_offload_prog_ops; | |
be9370a7 JB |
916 | prog->type = type; |
917 | return 0; | |
09756af4 AS |
918 | } |
919 | ||
920 | /* drop refcnt on maps used by eBPF program and free auxilary data */ | |
921 | static void free_used_maps(struct bpf_prog_aux *aux) | |
922 | { | |
923 | int i; | |
924 | ||
925 | for (i = 0; i < aux->used_map_cnt; i++) | |
926 | bpf_map_put(aux->used_maps[i]); | |
927 | ||
928 | kfree(aux->used_maps); | |
929 | } | |
930 | ||
5ccb071e DB |
931 | int __bpf_prog_charge(struct user_struct *user, u32 pages) |
932 | { | |
933 | unsigned long memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
934 | unsigned long user_bufs; | |
935 | ||
936 | if (user) { | |
937 | user_bufs = atomic_long_add_return(pages, &user->locked_vm); | |
938 | if (user_bufs > memlock_limit) { | |
939 | atomic_long_sub(pages, &user->locked_vm); | |
940 | return -EPERM; | |
941 | } | |
942 | } | |
943 | ||
944 | return 0; | |
945 | } | |
946 | ||
947 | void __bpf_prog_uncharge(struct user_struct *user, u32 pages) | |
948 | { | |
949 | if (user) | |
950 | atomic_long_sub(pages, &user->locked_vm); | |
951 | } | |
952 | ||
aaac3ba9 AS |
953 | static int bpf_prog_charge_memlock(struct bpf_prog *prog) |
954 | { | |
955 | struct user_struct *user = get_current_user(); | |
5ccb071e | 956 | int ret; |
aaac3ba9 | 957 | |
5ccb071e DB |
958 | ret = __bpf_prog_charge(user, prog->pages); |
959 | if (ret) { | |
aaac3ba9 | 960 | free_uid(user); |
5ccb071e | 961 | return ret; |
aaac3ba9 | 962 | } |
5ccb071e | 963 | |
aaac3ba9 AS |
964 | prog->aux->user = user; |
965 | return 0; | |
966 | } | |
967 | ||
968 | static void bpf_prog_uncharge_memlock(struct bpf_prog *prog) | |
969 | { | |
970 | struct user_struct *user = prog->aux->user; | |
971 | ||
5ccb071e | 972 | __bpf_prog_uncharge(user, prog->pages); |
aaac3ba9 AS |
973 | free_uid(user); |
974 | } | |
975 | ||
dc4bb0e2 MKL |
976 | static int bpf_prog_alloc_id(struct bpf_prog *prog) |
977 | { | |
978 | int id; | |
979 | ||
b76354cd | 980 | idr_preload(GFP_KERNEL); |
dc4bb0e2 MKL |
981 | spin_lock_bh(&prog_idr_lock); |
982 | id = idr_alloc_cyclic(&prog_idr, prog, 1, INT_MAX, GFP_ATOMIC); | |
983 | if (id > 0) | |
984 | prog->aux->id = id; | |
985 | spin_unlock_bh(&prog_idr_lock); | |
b76354cd | 986 | idr_preload_end(); |
dc4bb0e2 MKL |
987 | |
988 | /* id is in [1, INT_MAX) */ | |
989 | if (WARN_ON_ONCE(!id)) | |
990 | return -ENOSPC; | |
991 | ||
992 | return id > 0 ? 0 : id; | |
993 | } | |
994 | ||
ad8ad79f | 995 | void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock) |
dc4bb0e2 | 996 | { |
ad8ad79f JK |
997 | /* cBPF to eBPF migrations are currently not in the idr store. |
998 | * Offloaded programs are removed from the store when their device | |
999 | * disappears - even if someone grabs an fd to them they are unusable, | |
1000 | * simply waiting for refcnt to drop to be freed. | |
1001 | */ | |
dc4bb0e2 MKL |
1002 | if (!prog->aux->id) |
1003 | return; | |
1004 | ||
b16d9aa4 MKL |
1005 | if (do_idr_lock) |
1006 | spin_lock_bh(&prog_idr_lock); | |
1007 | else | |
1008 | __acquire(&prog_idr_lock); | |
1009 | ||
dc4bb0e2 | 1010 | idr_remove(&prog_idr, prog->aux->id); |
ad8ad79f | 1011 | prog->aux->id = 0; |
b16d9aa4 MKL |
1012 | |
1013 | if (do_idr_lock) | |
1014 | spin_unlock_bh(&prog_idr_lock); | |
1015 | else | |
1016 | __release(&prog_idr_lock); | |
dc4bb0e2 MKL |
1017 | } |
1018 | ||
1aacde3d | 1019 | static void __bpf_prog_put_rcu(struct rcu_head *rcu) |
abf2e7d6 AS |
1020 | { |
1021 | struct bpf_prog_aux *aux = container_of(rcu, struct bpf_prog_aux, rcu); | |
1022 | ||
1023 | free_used_maps(aux); | |
aaac3ba9 | 1024 | bpf_prog_uncharge_memlock(aux->prog); |
afdb09c7 | 1025 | security_bpf_prog_free(aux); |
abf2e7d6 AS |
1026 | bpf_prog_free(aux->prog); |
1027 | } | |
1028 | ||
b16d9aa4 | 1029 | static void __bpf_prog_put(struct bpf_prog *prog, bool do_idr_lock) |
09756af4 | 1030 | { |
a67edbf4 | 1031 | if (atomic_dec_and_test(&prog->aux->refcnt)) { |
4f74d809 DB |
1032 | int i; |
1033 | ||
34ad5580 | 1034 | /* bpf_prog_free_id() must be called first */ |
b16d9aa4 | 1035 | bpf_prog_free_id(prog, do_idr_lock); |
4f74d809 DB |
1036 | |
1037 | for (i = 0; i < prog->aux->func_cnt; i++) | |
1038 | bpf_prog_kallsyms_del(prog->aux->func[i]); | |
74451e66 | 1039 | bpf_prog_kallsyms_del(prog); |
4f74d809 | 1040 | |
1aacde3d | 1041 | call_rcu(&prog->aux->rcu, __bpf_prog_put_rcu); |
a67edbf4 | 1042 | } |
09756af4 | 1043 | } |
b16d9aa4 MKL |
1044 | |
1045 | void bpf_prog_put(struct bpf_prog *prog) | |
1046 | { | |
1047 | __bpf_prog_put(prog, true); | |
1048 | } | |
e2e9b654 | 1049 | EXPORT_SYMBOL_GPL(bpf_prog_put); |
09756af4 AS |
1050 | |
1051 | static int bpf_prog_release(struct inode *inode, struct file *filp) | |
1052 | { | |
1053 | struct bpf_prog *prog = filp->private_data; | |
1054 | ||
1aacde3d | 1055 | bpf_prog_put(prog); |
09756af4 AS |
1056 | return 0; |
1057 | } | |
1058 | ||
7bd509e3 DB |
1059 | #ifdef CONFIG_PROC_FS |
1060 | static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp) | |
1061 | { | |
1062 | const struct bpf_prog *prog = filp->private_data; | |
f1f7714e | 1063 | char prog_tag[sizeof(prog->tag) * 2 + 1] = { }; |
7bd509e3 | 1064 | |
f1f7714e | 1065 | bin2hex(prog_tag, prog->tag, sizeof(prog->tag)); |
7bd509e3 DB |
1066 | seq_printf(m, |
1067 | "prog_type:\t%u\n" | |
1068 | "prog_jited:\t%u\n" | |
f1f7714e | 1069 | "prog_tag:\t%s\n" |
7bd509e3 DB |
1070 | "memlock:\t%llu\n", |
1071 | prog->type, | |
1072 | prog->jited, | |
f1f7714e | 1073 | prog_tag, |
7bd509e3 DB |
1074 | prog->pages * 1ULL << PAGE_SHIFT); |
1075 | } | |
1076 | #endif | |
1077 | ||
f66e448c | 1078 | const struct file_operations bpf_prog_fops = { |
7bd509e3 DB |
1079 | #ifdef CONFIG_PROC_FS |
1080 | .show_fdinfo = bpf_prog_show_fdinfo, | |
1081 | #endif | |
1082 | .release = bpf_prog_release, | |
6e71b04a CF |
1083 | .read = bpf_dummy_read, |
1084 | .write = bpf_dummy_write, | |
09756af4 AS |
1085 | }; |
1086 | ||
b2197755 | 1087 | int bpf_prog_new_fd(struct bpf_prog *prog) |
aa79781b | 1088 | { |
afdb09c7 CF |
1089 | int ret; |
1090 | ||
1091 | ret = security_bpf_prog(prog); | |
1092 | if (ret < 0) | |
1093 | return ret; | |
1094 | ||
aa79781b DB |
1095 | return anon_inode_getfd("bpf-prog", &bpf_prog_fops, prog, |
1096 | O_RDWR | O_CLOEXEC); | |
1097 | } | |
1098 | ||
113214be | 1099 | static struct bpf_prog *____bpf_prog_get(struct fd f) |
09756af4 | 1100 | { |
09756af4 AS |
1101 | if (!f.file) |
1102 | return ERR_PTR(-EBADF); | |
09756af4 AS |
1103 | if (f.file->f_op != &bpf_prog_fops) { |
1104 | fdput(f); | |
1105 | return ERR_PTR(-EINVAL); | |
1106 | } | |
1107 | ||
c2101297 | 1108 | return f.file->private_data; |
09756af4 AS |
1109 | } |
1110 | ||
59d3656d | 1111 | struct bpf_prog *bpf_prog_add(struct bpf_prog *prog, int i) |
92117d84 | 1112 | { |
59d3656d BB |
1113 | if (atomic_add_return(i, &prog->aux->refcnt) > BPF_MAX_REFCNT) { |
1114 | atomic_sub(i, &prog->aux->refcnt); | |
92117d84 AS |
1115 | return ERR_PTR(-EBUSY); |
1116 | } | |
1117 | return prog; | |
1118 | } | |
59d3656d BB |
1119 | EXPORT_SYMBOL_GPL(bpf_prog_add); |
1120 | ||
c540594f DB |
1121 | void bpf_prog_sub(struct bpf_prog *prog, int i) |
1122 | { | |
1123 | /* Only to be used for undoing previous bpf_prog_add() in some | |
1124 | * error path. We still know that another entity in our call | |
1125 | * path holds a reference to the program, thus atomic_sub() can | |
1126 | * be safely used in such cases! | |
1127 | */ | |
1128 | WARN_ON(atomic_sub_return(i, &prog->aux->refcnt) == 0); | |
1129 | } | |
1130 | EXPORT_SYMBOL_GPL(bpf_prog_sub); | |
1131 | ||
59d3656d BB |
1132 | struct bpf_prog *bpf_prog_inc(struct bpf_prog *prog) |
1133 | { | |
1134 | return bpf_prog_add(prog, 1); | |
1135 | } | |
97bc402d | 1136 | EXPORT_SYMBOL_GPL(bpf_prog_inc); |
92117d84 | 1137 | |
b16d9aa4 | 1138 | /* prog_idr_lock should have been held */ |
a6f6df69 | 1139 | struct bpf_prog *bpf_prog_inc_not_zero(struct bpf_prog *prog) |
b16d9aa4 MKL |
1140 | { |
1141 | int refold; | |
1142 | ||
1143 | refold = __atomic_add_unless(&prog->aux->refcnt, 1, 0); | |
1144 | ||
1145 | if (refold >= BPF_MAX_REFCNT) { | |
1146 | __bpf_prog_put(prog, false); | |
1147 | return ERR_PTR(-EBUSY); | |
1148 | } | |
1149 | ||
1150 | if (!refold) | |
1151 | return ERR_PTR(-ENOENT); | |
1152 | ||
1153 | return prog; | |
1154 | } | |
a6f6df69 | 1155 | EXPORT_SYMBOL_GPL(bpf_prog_inc_not_zero); |
b16d9aa4 | 1156 | |
040ee692 | 1157 | bool bpf_prog_get_ok(struct bpf_prog *prog, |
288b3de5 | 1158 | enum bpf_prog_type *attach_type, bool attach_drv) |
248f346f | 1159 | { |
288b3de5 JK |
1160 | /* not an attachment, just a refcount inc, always allow */ |
1161 | if (!attach_type) | |
1162 | return true; | |
248f346f JK |
1163 | |
1164 | if (prog->type != *attach_type) | |
1165 | return false; | |
288b3de5 | 1166 | if (bpf_prog_is_dev_bound(prog->aux) && !attach_drv) |
248f346f JK |
1167 | return false; |
1168 | ||
1169 | return true; | |
1170 | } | |
1171 | ||
1172 | static struct bpf_prog *__bpf_prog_get(u32 ufd, enum bpf_prog_type *attach_type, | |
288b3de5 | 1173 | bool attach_drv) |
09756af4 AS |
1174 | { |
1175 | struct fd f = fdget(ufd); | |
1176 | struct bpf_prog *prog; | |
1177 | ||
113214be | 1178 | prog = ____bpf_prog_get(f); |
09756af4 AS |
1179 | if (IS_ERR(prog)) |
1180 | return prog; | |
288b3de5 | 1181 | if (!bpf_prog_get_ok(prog, attach_type, attach_drv)) { |
113214be DB |
1182 | prog = ERR_PTR(-EINVAL); |
1183 | goto out; | |
1184 | } | |
09756af4 | 1185 | |
92117d84 | 1186 | prog = bpf_prog_inc(prog); |
113214be | 1187 | out: |
09756af4 AS |
1188 | fdput(f); |
1189 | return prog; | |
1190 | } | |
113214be DB |
1191 | |
1192 | struct bpf_prog *bpf_prog_get(u32 ufd) | |
1193 | { | |
288b3de5 | 1194 | return __bpf_prog_get(ufd, NULL, false); |
113214be DB |
1195 | } |
1196 | ||
248f346f | 1197 | struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, |
288b3de5 | 1198 | bool attach_drv) |
248f346f | 1199 | { |
4d220ed0 | 1200 | return __bpf_prog_get(ufd, &type, attach_drv); |
248f346f | 1201 | } |
6c8dfe21 | 1202 | EXPORT_SYMBOL_GPL(bpf_prog_get_type_dev); |
248f346f | 1203 | |
aac3fc32 AI |
1204 | /* Initially all BPF programs could be loaded w/o specifying |
1205 | * expected_attach_type. Later for some of them specifying expected_attach_type | |
1206 | * at load time became required so that program could be validated properly. | |
1207 | * Programs of types that are allowed to be loaded both w/ and w/o (for | |
1208 | * backward compatibility) expected_attach_type, should have the default attach | |
1209 | * type assigned to expected_attach_type for the latter case, so that it can be | |
1210 | * validated later at attach time. | |
1211 | * | |
1212 | * bpf_prog_load_fixup_attach_type() sets expected_attach_type in @attr if | |
1213 | * prog type requires it but has some attach types that have to be backward | |
1214 | * compatible. | |
1215 | */ | |
1216 | static void bpf_prog_load_fixup_attach_type(union bpf_attr *attr) | |
1217 | { | |
1218 | switch (attr->prog_type) { | |
1219 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
1220 | /* Unfortunately BPF_ATTACH_TYPE_UNSPEC enumeration doesn't | |
1221 | * exist so checking for non-zero is the way to go here. | |
1222 | */ | |
1223 | if (!attr->expected_attach_type) | |
1224 | attr->expected_attach_type = | |
1225 | BPF_CGROUP_INET_SOCK_CREATE; | |
1226 | break; | |
1227 | } | |
1228 | } | |
1229 | ||
5e43f899 AI |
1230 | static int |
1231 | bpf_prog_load_check_attach_type(enum bpf_prog_type prog_type, | |
1232 | enum bpf_attach_type expected_attach_type) | |
1233 | { | |
4fbac77d | 1234 | switch (prog_type) { |
aac3fc32 AI |
1235 | case BPF_PROG_TYPE_CGROUP_SOCK: |
1236 | switch (expected_attach_type) { | |
1237 | case BPF_CGROUP_INET_SOCK_CREATE: | |
1238 | case BPF_CGROUP_INET4_POST_BIND: | |
1239 | case BPF_CGROUP_INET6_POST_BIND: | |
1240 | return 0; | |
1241 | default: | |
1242 | return -EINVAL; | |
1243 | } | |
4fbac77d AI |
1244 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: |
1245 | switch (expected_attach_type) { | |
1246 | case BPF_CGROUP_INET4_BIND: | |
1247 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1248 | case BPF_CGROUP_INET4_CONNECT: |
1249 | case BPF_CGROUP_INET6_CONNECT: | |
4fbac77d AI |
1250 | return 0; |
1251 | default: | |
1252 | return -EINVAL; | |
1253 | } | |
1254 | default: | |
1255 | return 0; | |
1256 | } | |
5e43f899 AI |
1257 | } |
1258 | ||
09756af4 | 1259 | /* last field in 'union bpf_attr' used by this command */ |
5e43f899 | 1260 | #define BPF_PROG_LOAD_LAST_FIELD expected_attach_type |
09756af4 AS |
1261 | |
1262 | static int bpf_prog_load(union bpf_attr *attr) | |
1263 | { | |
1264 | enum bpf_prog_type type = attr->prog_type; | |
1265 | struct bpf_prog *prog; | |
1266 | int err; | |
1267 | char license[128]; | |
1268 | bool is_gpl; | |
1269 | ||
1270 | if (CHECK_ATTR(BPF_PROG_LOAD)) | |
1271 | return -EINVAL; | |
1272 | ||
e07b98d9 DM |
1273 | if (attr->prog_flags & ~BPF_F_STRICT_ALIGNMENT) |
1274 | return -EINVAL; | |
1275 | ||
09756af4 | 1276 | /* copy eBPF program license from user space */ |
535e7b4b | 1277 | if (strncpy_from_user(license, u64_to_user_ptr(attr->license), |
09756af4 AS |
1278 | sizeof(license) - 1) < 0) |
1279 | return -EFAULT; | |
1280 | license[sizeof(license) - 1] = 0; | |
1281 | ||
1282 | /* eBPF programs must be GPL compatible to use GPL-ed functions */ | |
1283 | is_gpl = license_is_gpl_compatible(license); | |
1284 | ||
ef0915ca DB |
1285 | if (attr->insn_cnt == 0 || attr->insn_cnt > BPF_MAXINSNS) |
1286 | return -E2BIG; | |
09756af4 | 1287 | |
2541517c AS |
1288 | if (type == BPF_PROG_TYPE_KPROBE && |
1289 | attr->kern_version != LINUX_VERSION_CODE) | |
1290 | return -EINVAL; | |
1291 | ||
80b7d819 CF |
1292 | if (type != BPF_PROG_TYPE_SOCKET_FILTER && |
1293 | type != BPF_PROG_TYPE_CGROUP_SKB && | |
1294 | !capable(CAP_SYS_ADMIN)) | |
1be7f75d AS |
1295 | return -EPERM; |
1296 | ||
aac3fc32 | 1297 | bpf_prog_load_fixup_attach_type(attr); |
5e43f899 AI |
1298 | if (bpf_prog_load_check_attach_type(type, attr->expected_attach_type)) |
1299 | return -EINVAL; | |
1300 | ||
09756af4 AS |
1301 | /* plain bpf_prog allocation */ |
1302 | prog = bpf_prog_alloc(bpf_prog_size(attr->insn_cnt), GFP_USER); | |
1303 | if (!prog) | |
1304 | return -ENOMEM; | |
1305 | ||
5e43f899 AI |
1306 | prog->expected_attach_type = attr->expected_attach_type; |
1307 | ||
9a18eedb JK |
1308 | prog->aux->offload_requested = !!attr->prog_ifindex; |
1309 | ||
afdb09c7 | 1310 | err = security_bpf_prog_alloc(prog->aux); |
aaac3ba9 AS |
1311 | if (err) |
1312 | goto free_prog_nouncharge; | |
1313 | ||
afdb09c7 CF |
1314 | err = bpf_prog_charge_memlock(prog); |
1315 | if (err) | |
1316 | goto free_prog_sec; | |
1317 | ||
09756af4 AS |
1318 | prog->len = attr->insn_cnt; |
1319 | ||
1320 | err = -EFAULT; | |
535e7b4b | 1321 | if (copy_from_user(prog->insns, u64_to_user_ptr(attr->insns), |
aafe6ae9 | 1322 | bpf_prog_insn_size(prog)) != 0) |
09756af4 AS |
1323 | goto free_prog; |
1324 | ||
1325 | prog->orig_prog = NULL; | |
a91263d5 | 1326 | prog->jited = 0; |
09756af4 AS |
1327 | |
1328 | atomic_set(&prog->aux->refcnt, 1); | |
a91263d5 | 1329 | prog->gpl_compatible = is_gpl ? 1 : 0; |
09756af4 | 1330 | |
9a18eedb | 1331 | if (bpf_prog_is_dev_bound(prog->aux)) { |
ab3f0063 JK |
1332 | err = bpf_prog_offload_init(prog, attr); |
1333 | if (err) | |
1334 | goto free_prog; | |
1335 | } | |
1336 | ||
09756af4 AS |
1337 | /* find program type: socket_filter vs tracing_filter */ |
1338 | err = find_prog_type(type, prog); | |
1339 | if (err < 0) | |
1340 | goto free_prog; | |
1341 | ||
cb4d2b3f MKL |
1342 | prog->aux->load_time = ktime_get_boot_ns(); |
1343 | err = bpf_obj_name_cpy(prog->aux->name, attr->prog_name); | |
1344 | if (err) | |
1345 | goto free_prog; | |
1346 | ||
09756af4 | 1347 | /* run eBPF verifier */ |
9bac3d6d | 1348 | err = bpf_check(&prog, attr); |
09756af4 AS |
1349 | if (err < 0) |
1350 | goto free_used_maps; | |
1351 | ||
1352 | /* eBPF program is ready to be JITed */ | |
1c2a088a AS |
1353 | if (!prog->bpf_func) |
1354 | prog = bpf_prog_select_runtime(prog, &err); | |
04fd61ab AS |
1355 | if (err < 0) |
1356 | goto free_used_maps; | |
09756af4 | 1357 | |
dc4bb0e2 MKL |
1358 | err = bpf_prog_alloc_id(prog); |
1359 | if (err) | |
1360 | goto free_used_maps; | |
1361 | ||
aa79781b | 1362 | err = bpf_prog_new_fd(prog); |
b16d9aa4 MKL |
1363 | if (err < 0) { |
1364 | /* failed to allocate fd. | |
1365 | * bpf_prog_put() is needed because the above | |
1366 | * bpf_prog_alloc_id() has published the prog | |
1367 | * to the userspace and the userspace may | |
1368 | * have refcnt-ed it through BPF_PROG_GET_FD_BY_ID. | |
1369 | */ | |
1370 | bpf_prog_put(prog); | |
1371 | return err; | |
1372 | } | |
09756af4 | 1373 | |
74451e66 | 1374 | bpf_prog_kallsyms_add(prog); |
09756af4 AS |
1375 | return err; |
1376 | ||
1377 | free_used_maps: | |
1378 | free_used_maps(prog->aux); | |
1379 | free_prog: | |
aaac3ba9 | 1380 | bpf_prog_uncharge_memlock(prog); |
afdb09c7 CF |
1381 | free_prog_sec: |
1382 | security_bpf_prog_free(prog->aux); | |
aaac3ba9 | 1383 | free_prog_nouncharge: |
09756af4 AS |
1384 | bpf_prog_free(prog); |
1385 | return err; | |
1386 | } | |
1387 | ||
6e71b04a | 1388 | #define BPF_OBJ_LAST_FIELD file_flags |
b2197755 DB |
1389 | |
1390 | static int bpf_obj_pin(const union bpf_attr *attr) | |
1391 | { | |
6e71b04a | 1392 | if (CHECK_ATTR(BPF_OBJ) || attr->file_flags != 0) |
b2197755 DB |
1393 | return -EINVAL; |
1394 | ||
535e7b4b | 1395 | return bpf_obj_pin_user(attr->bpf_fd, u64_to_user_ptr(attr->pathname)); |
b2197755 DB |
1396 | } |
1397 | ||
1398 | static int bpf_obj_get(const union bpf_attr *attr) | |
1399 | { | |
6e71b04a CF |
1400 | if (CHECK_ATTR(BPF_OBJ) || attr->bpf_fd != 0 || |
1401 | attr->file_flags & ~BPF_OBJ_FLAG_MASK) | |
b2197755 DB |
1402 | return -EINVAL; |
1403 | ||
6e71b04a CF |
1404 | return bpf_obj_get_user(u64_to_user_ptr(attr->pathname), |
1405 | attr->file_flags); | |
b2197755 DB |
1406 | } |
1407 | ||
c4f6699d AS |
1408 | struct bpf_raw_tracepoint { |
1409 | struct bpf_raw_event_map *btp; | |
1410 | struct bpf_prog *prog; | |
1411 | }; | |
1412 | ||
1413 | static int bpf_raw_tracepoint_release(struct inode *inode, struct file *filp) | |
1414 | { | |
1415 | struct bpf_raw_tracepoint *raw_tp = filp->private_data; | |
1416 | ||
1417 | if (raw_tp->prog) { | |
1418 | bpf_probe_unregister(raw_tp->btp, raw_tp->prog); | |
1419 | bpf_prog_put(raw_tp->prog); | |
1420 | } | |
1421 | kfree(raw_tp); | |
1422 | return 0; | |
1423 | } | |
1424 | ||
1425 | static const struct file_operations bpf_raw_tp_fops = { | |
1426 | .release = bpf_raw_tracepoint_release, | |
1427 | .read = bpf_dummy_read, | |
1428 | .write = bpf_dummy_write, | |
1429 | }; | |
1430 | ||
1431 | #define BPF_RAW_TRACEPOINT_OPEN_LAST_FIELD raw_tracepoint.prog_fd | |
1432 | ||
1433 | static int bpf_raw_tracepoint_open(const union bpf_attr *attr) | |
1434 | { | |
1435 | struct bpf_raw_tracepoint *raw_tp; | |
1436 | struct bpf_raw_event_map *btp; | |
1437 | struct bpf_prog *prog; | |
1438 | char tp_name[128]; | |
1439 | int tp_fd, err; | |
1440 | ||
1441 | if (strncpy_from_user(tp_name, u64_to_user_ptr(attr->raw_tracepoint.name), | |
1442 | sizeof(tp_name) - 1) < 0) | |
1443 | return -EFAULT; | |
1444 | tp_name[sizeof(tp_name) - 1] = 0; | |
1445 | ||
1446 | btp = bpf_find_raw_tracepoint(tp_name); | |
1447 | if (!btp) | |
1448 | return -ENOENT; | |
1449 | ||
1450 | raw_tp = kzalloc(sizeof(*raw_tp), GFP_USER); | |
1451 | if (!raw_tp) | |
1452 | return -ENOMEM; | |
1453 | raw_tp->btp = btp; | |
1454 | ||
1455 | prog = bpf_prog_get_type(attr->raw_tracepoint.prog_fd, | |
1456 | BPF_PROG_TYPE_RAW_TRACEPOINT); | |
1457 | if (IS_ERR(prog)) { | |
1458 | err = PTR_ERR(prog); | |
1459 | goto out_free_tp; | |
1460 | } | |
1461 | ||
1462 | err = bpf_probe_register(raw_tp->btp, prog); | |
1463 | if (err) | |
1464 | goto out_put_prog; | |
1465 | ||
1466 | raw_tp->prog = prog; | |
1467 | tp_fd = anon_inode_getfd("bpf-raw-tracepoint", &bpf_raw_tp_fops, raw_tp, | |
1468 | O_CLOEXEC); | |
1469 | if (tp_fd < 0) { | |
1470 | bpf_probe_unregister(raw_tp->btp, prog); | |
1471 | err = tp_fd; | |
1472 | goto out_put_prog; | |
1473 | } | |
1474 | return tp_fd; | |
1475 | ||
1476 | out_put_prog: | |
1477 | bpf_prog_put(prog); | |
1478 | out_free_tp: | |
1479 | kfree(raw_tp); | |
1480 | return err; | |
1481 | } | |
1482 | ||
f4324551 DM |
1483 | #ifdef CONFIG_CGROUP_BPF |
1484 | ||
33491588 AR |
1485 | static int bpf_prog_attach_check_attach_type(const struct bpf_prog *prog, |
1486 | enum bpf_attach_type attach_type) | |
1487 | { | |
1488 | switch (prog->type) { | |
1489 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
1490 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: | |
1491 | return attach_type == prog->expected_attach_type ? 0 : -EINVAL; | |
1492 | default: | |
1493 | return 0; | |
1494 | } | |
1495 | } | |
1496 | ||
464bc0fd | 1497 | #define BPF_PROG_ATTACH_LAST_FIELD attach_flags |
174a79ff | 1498 | |
4f738adb JF |
1499 | static int sockmap_get_from_fd(const union bpf_attr *attr, |
1500 | int type, bool attach) | |
174a79ff | 1501 | { |
5a67da2a | 1502 | struct bpf_prog *prog = NULL; |
174a79ff JF |
1503 | int ufd = attr->target_fd; |
1504 | struct bpf_map *map; | |
1505 | struct fd f; | |
1506 | int err; | |
1507 | ||
1508 | f = fdget(ufd); | |
1509 | map = __bpf_map_get(f); | |
1510 | if (IS_ERR(map)) | |
1511 | return PTR_ERR(map); | |
1512 | ||
5a67da2a | 1513 | if (attach) { |
4f738adb | 1514 | prog = bpf_prog_get_type(attr->attach_bpf_fd, type); |
5a67da2a JF |
1515 | if (IS_ERR(prog)) { |
1516 | fdput(f); | |
1517 | return PTR_ERR(prog); | |
1518 | } | |
174a79ff JF |
1519 | } |
1520 | ||
5a67da2a | 1521 | err = sock_map_prog(map, prog, attr->attach_type); |
174a79ff JF |
1522 | if (err) { |
1523 | fdput(f); | |
5a67da2a JF |
1524 | if (prog) |
1525 | bpf_prog_put(prog); | |
ae2b27b8 | 1526 | return err; |
174a79ff JF |
1527 | } |
1528 | ||
1529 | fdput(f); | |
ae2b27b8 | 1530 | return 0; |
174a79ff | 1531 | } |
f4324551 | 1532 | |
324bda9e AS |
1533 | #define BPF_F_ATTACH_MASK \ |
1534 | (BPF_F_ALLOW_OVERRIDE | BPF_F_ALLOW_MULTI) | |
1535 | ||
f4324551 DM |
1536 | static int bpf_prog_attach(const union bpf_attr *attr) |
1537 | { | |
7f677633 | 1538 | enum bpf_prog_type ptype; |
f4324551 DM |
1539 | struct bpf_prog *prog; |
1540 | struct cgroup *cgrp; | |
7f677633 | 1541 | int ret; |
f4324551 DM |
1542 | |
1543 | if (!capable(CAP_NET_ADMIN)) | |
1544 | return -EPERM; | |
1545 | ||
1546 | if (CHECK_ATTR(BPF_PROG_ATTACH)) | |
1547 | return -EINVAL; | |
1548 | ||
324bda9e | 1549 | if (attr->attach_flags & ~BPF_F_ATTACH_MASK) |
7f677633 AS |
1550 | return -EINVAL; |
1551 | ||
f4324551 DM |
1552 | switch (attr->attach_type) { |
1553 | case BPF_CGROUP_INET_INGRESS: | |
1554 | case BPF_CGROUP_INET_EGRESS: | |
b2cd1257 | 1555 | ptype = BPF_PROG_TYPE_CGROUP_SKB; |
f4324551 | 1556 | break; |
61023658 | 1557 | case BPF_CGROUP_INET_SOCK_CREATE: |
aac3fc32 AI |
1558 | case BPF_CGROUP_INET4_POST_BIND: |
1559 | case BPF_CGROUP_INET6_POST_BIND: | |
61023658 DA |
1560 | ptype = BPF_PROG_TYPE_CGROUP_SOCK; |
1561 | break; | |
4fbac77d AI |
1562 | case BPF_CGROUP_INET4_BIND: |
1563 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1564 | case BPF_CGROUP_INET4_CONNECT: |
1565 | case BPF_CGROUP_INET6_CONNECT: | |
4fbac77d AI |
1566 | ptype = BPF_PROG_TYPE_CGROUP_SOCK_ADDR; |
1567 | break; | |
40304b2a LB |
1568 | case BPF_CGROUP_SOCK_OPS: |
1569 | ptype = BPF_PROG_TYPE_SOCK_OPS; | |
1570 | break; | |
ebc614f6 RG |
1571 | case BPF_CGROUP_DEVICE: |
1572 | ptype = BPF_PROG_TYPE_CGROUP_DEVICE; | |
1573 | break; | |
4f738adb JF |
1574 | case BPF_SK_MSG_VERDICT: |
1575 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_MSG, true); | |
464bc0fd JF |
1576 | case BPF_SK_SKB_STREAM_PARSER: |
1577 | case BPF_SK_SKB_STREAM_VERDICT: | |
4f738adb | 1578 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_SKB, true); |
f4324551 DM |
1579 | default: |
1580 | return -EINVAL; | |
1581 | } | |
1582 | ||
b2cd1257 DA |
1583 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); |
1584 | if (IS_ERR(prog)) | |
1585 | return PTR_ERR(prog); | |
1586 | ||
5e43f899 AI |
1587 | if (bpf_prog_attach_check_attach_type(prog, attr->attach_type)) { |
1588 | bpf_prog_put(prog); | |
1589 | return -EINVAL; | |
1590 | } | |
1591 | ||
b2cd1257 DA |
1592 | cgrp = cgroup_get_from_fd(attr->target_fd); |
1593 | if (IS_ERR(cgrp)) { | |
1594 | bpf_prog_put(prog); | |
1595 | return PTR_ERR(cgrp); | |
1596 | } | |
1597 | ||
324bda9e AS |
1598 | ret = cgroup_bpf_attach(cgrp, prog, attr->attach_type, |
1599 | attr->attach_flags); | |
7f677633 AS |
1600 | if (ret) |
1601 | bpf_prog_put(prog); | |
b2cd1257 DA |
1602 | cgroup_put(cgrp); |
1603 | ||
7f677633 | 1604 | return ret; |
f4324551 DM |
1605 | } |
1606 | ||
1607 | #define BPF_PROG_DETACH_LAST_FIELD attach_type | |
1608 | ||
1609 | static int bpf_prog_detach(const union bpf_attr *attr) | |
1610 | { | |
324bda9e AS |
1611 | enum bpf_prog_type ptype; |
1612 | struct bpf_prog *prog; | |
f4324551 | 1613 | struct cgroup *cgrp; |
7f677633 | 1614 | int ret; |
f4324551 DM |
1615 | |
1616 | if (!capable(CAP_NET_ADMIN)) | |
1617 | return -EPERM; | |
1618 | ||
1619 | if (CHECK_ATTR(BPF_PROG_DETACH)) | |
1620 | return -EINVAL; | |
1621 | ||
1622 | switch (attr->attach_type) { | |
1623 | case BPF_CGROUP_INET_INGRESS: | |
1624 | case BPF_CGROUP_INET_EGRESS: | |
324bda9e AS |
1625 | ptype = BPF_PROG_TYPE_CGROUP_SKB; |
1626 | break; | |
61023658 | 1627 | case BPF_CGROUP_INET_SOCK_CREATE: |
aac3fc32 AI |
1628 | case BPF_CGROUP_INET4_POST_BIND: |
1629 | case BPF_CGROUP_INET6_POST_BIND: | |
324bda9e AS |
1630 | ptype = BPF_PROG_TYPE_CGROUP_SOCK; |
1631 | break; | |
4fbac77d AI |
1632 | case BPF_CGROUP_INET4_BIND: |
1633 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1634 | case BPF_CGROUP_INET4_CONNECT: |
1635 | case BPF_CGROUP_INET6_CONNECT: | |
4fbac77d AI |
1636 | ptype = BPF_PROG_TYPE_CGROUP_SOCK_ADDR; |
1637 | break; | |
40304b2a | 1638 | case BPF_CGROUP_SOCK_OPS: |
324bda9e | 1639 | ptype = BPF_PROG_TYPE_SOCK_OPS; |
f4324551 | 1640 | break; |
ebc614f6 RG |
1641 | case BPF_CGROUP_DEVICE: |
1642 | ptype = BPF_PROG_TYPE_CGROUP_DEVICE; | |
1643 | break; | |
4f738adb JF |
1644 | case BPF_SK_MSG_VERDICT: |
1645 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_MSG, false); | |
5a67da2a JF |
1646 | case BPF_SK_SKB_STREAM_PARSER: |
1647 | case BPF_SK_SKB_STREAM_VERDICT: | |
4f738adb | 1648 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_SKB, false); |
f4324551 DM |
1649 | default: |
1650 | return -EINVAL; | |
1651 | } | |
1652 | ||
324bda9e AS |
1653 | cgrp = cgroup_get_from_fd(attr->target_fd); |
1654 | if (IS_ERR(cgrp)) | |
1655 | return PTR_ERR(cgrp); | |
1656 | ||
1657 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); | |
1658 | if (IS_ERR(prog)) | |
1659 | prog = NULL; | |
1660 | ||
1661 | ret = cgroup_bpf_detach(cgrp, prog, attr->attach_type, 0); | |
1662 | if (prog) | |
1663 | bpf_prog_put(prog); | |
1664 | cgroup_put(cgrp); | |
7f677633 | 1665 | return ret; |
f4324551 | 1666 | } |
40304b2a | 1667 | |
468e2f64 AS |
1668 | #define BPF_PROG_QUERY_LAST_FIELD query.prog_cnt |
1669 | ||
1670 | static int bpf_prog_query(const union bpf_attr *attr, | |
1671 | union bpf_attr __user *uattr) | |
1672 | { | |
1673 | struct cgroup *cgrp; | |
1674 | int ret; | |
1675 | ||
1676 | if (!capable(CAP_NET_ADMIN)) | |
1677 | return -EPERM; | |
1678 | if (CHECK_ATTR(BPF_PROG_QUERY)) | |
1679 | return -EINVAL; | |
1680 | if (attr->query.query_flags & ~BPF_F_QUERY_EFFECTIVE) | |
1681 | return -EINVAL; | |
1682 | ||
1683 | switch (attr->query.attach_type) { | |
1684 | case BPF_CGROUP_INET_INGRESS: | |
1685 | case BPF_CGROUP_INET_EGRESS: | |
1686 | case BPF_CGROUP_INET_SOCK_CREATE: | |
4fbac77d AI |
1687 | case BPF_CGROUP_INET4_BIND: |
1688 | case BPF_CGROUP_INET6_BIND: | |
aac3fc32 AI |
1689 | case BPF_CGROUP_INET4_POST_BIND: |
1690 | case BPF_CGROUP_INET6_POST_BIND: | |
d74bad4e AI |
1691 | case BPF_CGROUP_INET4_CONNECT: |
1692 | case BPF_CGROUP_INET6_CONNECT: | |
468e2f64 | 1693 | case BPF_CGROUP_SOCK_OPS: |
ebc614f6 | 1694 | case BPF_CGROUP_DEVICE: |
468e2f64 AS |
1695 | break; |
1696 | default: | |
1697 | return -EINVAL; | |
1698 | } | |
1699 | cgrp = cgroup_get_from_fd(attr->query.target_fd); | |
1700 | if (IS_ERR(cgrp)) | |
1701 | return PTR_ERR(cgrp); | |
1702 | ret = cgroup_bpf_query(cgrp, attr, uattr); | |
1703 | cgroup_put(cgrp); | |
1704 | return ret; | |
1705 | } | |
f4324551 DM |
1706 | #endif /* CONFIG_CGROUP_BPF */ |
1707 | ||
1cf1cae9 AS |
1708 | #define BPF_PROG_TEST_RUN_LAST_FIELD test.duration |
1709 | ||
1710 | static int bpf_prog_test_run(const union bpf_attr *attr, | |
1711 | union bpf_attr __user *uattr) | |
1712 | { | |
1713 | struct bpf_prog *prog; | |
1714 | int ret = -ENOTSUPP; | |
1715 | ||
61f3c964 AS |
1716 | if (!capable(CAP_SYS_ADMIN)) |
1717 | return -EPERM; | |
1cf1cae9 AS |
1718 | if (CHECK_ATTR(BPF_PROG_TEST_RUN)) |
1719 | return -EINVAL; | |
1720 | ||
1721 | prog = bpf_prog_get(attr->test.prog_fd); | |
1722 | if (IS_ERR(prog)) | |
1723 | return PTR_ERR(prog); | |
1724 | ||
1725 | if (prog->aux->ops->test_run) | |
1726 | ret = prog->aux->ops->test_run(prog, attr, uattr); | |
1727 | ||
1728 | bpf_prog_put(prog); | |
1729 | return ret; | |
1730 | } | |
1731 | ||
34ad5580 MKL |
1732 | #define BPF_OBJ_GET_NEXT_ID_LAST_FIELD next_id |
1733 | ||
1734 | static int bpf_obj_get_next_id(const union bpf_attr *attr, | |
1735 | union bpf_attr __user *uattr, | |
1736 | struct idr *idr, | |
1737 | spinlock_t *lock) | |
1738 | { | |
1739 | u32 next_id = attr->start_id; | |
1740 | int err = 0; | |
1741 | ||
1742 | if (CHECK_ATTR(BPF_OBJ_GET_NEXT_ID) || next_id >= INT_MAX) | |
1743 | return -EINVAL; | |
1744 | ||
1745 | if (!capable(CAP_SYS_ADMIN)) | |
1746 | return -EPERM; | |
1747 | ||
1748 | next_id++; | |
1749 | spin_lock_bh(lock); | |
1750 | if (!idr_get_next(idr, &next_id)) | |
1751 | err = -ENOENT; | |
1752 | spin_unlock_bh(lock); | |
1753 | ||
1754 | if (!err) | |
1755 | err = put_user(next_id, &uattr->next_id); | |
1756 | ||
1757 | return err; | |
1758 | } | |
1759 | ||
b16d9aa4 MKL |
1760 | #define BPF_PROG_GET_FD_BY_ID_LAST_FIELD prog_id |
1761 | ||
1762 | static int bpf_prog_get_fd_by_id(const union bpf_attr *attr) | |
1763 | { | |
1764 | struct bpf_prog *prog; | |
1765 | u32 id = attr->prog_id; | |
1766 | int fd; | |
1767 | ||
1768 | if (CHECK_ATTR(BPF_PROG_GET_FD_BY_ID)) | |
1769 | return -EINVAL; | |
1770 | ||
1771 | if (!capable(CAP_SYS_ADMIN)) | |
1772 | return -EPERM; | |
1773 | ||
1774 | spin_lock_bh(&prog_idr_lock); | |
1775 | prog = idr_find(&prog_idr, id); | |
1776 | if (prog) | |
1777 | prog = bpf_prog_inc_not_zero(prog); | |
1778 | else | |
1779 | prog = ERR_PTR(-ENOENT); | |
1780 | spin_unlock_bh(&prog_idr_lock); | |
1781 | ||
1782 | if (IS_ERR(prog)) | |
1783 | return PTR_ERR(prog); | |
1784 | ||
1785 | fd = bpf_prog_new_fd(prog); | |
1786 | if (fd < 0) | |
1787 | bpf_prog_put(prog); | |
1788 | ||
1789 | return fd; | |
1790 | } | |
1791 | ||
6e71b04a | 1792 | #define BPF_MAP_GET_FD_BY_ID_LAST_FIELD open_flags |
bd5f5f4e MKL |
1793 | |
1794 | static int bpf_map_get_fd_by_id(const union bpf_attr *attr) | |
1795 | { | |
1796 | struct bpf_map *map; | |
1797 | u32 id = attr->map_id; | |
6e71b04a | 1798 | int f_flags; |
bd5f5f4e MKL |
1799 | int fd; |
1800 | ||
6e71b04a CF |
1801 | if (CHECK_ATTR(BPF_MAP_GET_FD_BY_ID) || |
1802 | attr->open_flags & ~BPF_OBJ_FLAG_MASK) | |
bd5f5f4e MKL |
1803 | return -EINVAL; |
1804 | ||
1805 | if (!capable(CAP_SYS_ADMIN)) | |
1806 | return -EPERM; | |
1807 | ||
6e71b04a CF |
1808 | f_flags = bpf_get_file_flag(attr->open_flags); |
1809 | if (f_flags < 0) | |
1810 | return f_flags; | |
1811 | ||
bd5f5f4e MKL |
1812 | spin_lock_bh(&map_idr_lock); |
1813 | map = idr_find(&map_idr, id); | |
1814 | if (map) | |
1815 | map = bpf_map_inc_not_zero(map, true); | |
1816 | else | |
1817 | map = ERR_PTR(-ENOENT); | |
1818 | spin_unlock_bh(&map_idr_lock); | |
1819 | ||
1820 | if (IS_ERR(map)) | |
1821 | return PTR_ERR(map); | |
1822 | ||
6e71b04a | 1823 | fd = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
1824 | if (fd < 0) |
1825 | bpf_map_put(map); | |
1826 | ||
1827 | return fd; | |
1828 | } | |
1829 | ||
7105e828 DB |
1830 | static const struct bpf_map *bpf_map_from_imm(const struct bpf_prog *prog, |
1831 | unsigned long addr) | |
1832 | { | |
1833 | int i; | |
1834 | ||
1835 | for (i = 0; i < prog->aux->used_map_cnt; i++) | |
1836 | if (prog->aux->used_maps[i] == (void *)addr) | |
1837 | return prog->aux->used_maps[i]; | |
1838 | return NULL; | |
1839 | } | |
1840 | ||
1841 | static struct bpf_insn *bpf_insn_prepare_dump(const struct bpf_prog *prog) | |
1842 | { | |
1843 | const struct bpf_map *map; | |
1844 | struct bpf_insn *insns; | |
1845 | u64 imm; | |
1846 | int i; | |
1847 | ||
1848 | insns = kmemdup(prog->insnsi, bpf_prog_insn_size(prog), | |
1849 | GFP_USER); | |
1850 | if (!insns) | |
1851 | return insns; | |
1852 | ||
1853 | for (i = 0; i < prog->len; i++) { | |
1854 | if (insns[i].code == (BPF_JMP | BPF_TAIL_CALL)) { | |
1855 | insns[i].code = BPF_JMP | BPF_CALL; | |
1856 | insns[i].imm = BPF_FUNC_tail_call; | |
1857 | /* fall-through */ | |
1858 | } | |
1859 | if (insns[i].code == (BPF_JMP | BPF_CALL) || | |
1860 | insns[i].code == (BPF_JMP | BPF_CALL_ARGS)) { | |
1861 | if (insns[i].code == (BPF_JMP | BPF_CALL_ARGS)) | |
1862 | insns[i].code = BPF_JMP | BPF_CALL; | |
1863 | if (!bpf_dump_raw_ok()) | |
1864 | insns[i].imm = 0; | |
1865 | continue; | |
1866 | } | |
1867 | ||
1868 | if (insns[i].code != (BPF_LD | BPF_IMM | BPF_DW)) | |
1869 | continue; | |
1870 | ||
1871 | imm = ((u64)insns[i + 1].imm << 32) | (u32)insns[i].imm; | |
1872 | map = bpf_map_from_imm(prog, imm); | |
1873 | if (map) { | |
1874 | insns[i].src_reg = BPF_PSEUDO_MAP_FD; | |
1875 | insns[i].imm = map->id; | |
1876 | insns[i + 1].imm = 0; | |
1877 | continue; | |
1878 | } | |
1879 | ||
1880 | if (!bpf_dump_raw_ok() && | |
1881 | imm == (unsigned long)prog->aux) { | |
1882 | insns[i].imm = 0; | |
1883 | insns[i + 1].imm = 0; | |
1884 | continue; | |
1885 | } | |
1886 | } | |
1887 | ||
1888 | return insns; | |
1889 | } | |
1890 | ||
1e270976 MKL |
1891 | static int bpf_prog_get_info_by_fd(struct bpf_prog *prog, |
1892 | const union bpf_attr *attr, | |
1893 | union bpf_attr __user *uattr) | |
1894 | { | |
1895 | struct bpf_prog_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
1896 | struct bpf_prog_info info = {}; | |
1897 | u32 info_len = attr->info.info_len; | |
1898 | char __user *uinsns; | |
1899 | u32 ulen; | |
1900 | int err; | |
1901 | ||
dcab51f1 | 1902 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(info), info_len); |
1e270976 MKL |
1903 | if (err) |
1904 | return err; | |
1905 | info_len = min_t(u32, sizeof(info), info_len); | |
1906 | ||
1907 | if (copy_from_user(&info, uinfo, info_len)) | |
89b09689 | 1908 | return -EFAULT; |
1e270976 MKL |
1909 | |
1910 | info.type = prog->type; | |
1911 | info.id = prog->aux->id; | |
cb4d2b3f MKL |
1912 | info.load_time = prog->aux->load_time; |
1913 | info.created_by_uid = from_kuid_munged(current_user_ns(), | |
1914 | prog->aux->user->uid); | |
b85fab0e | 1915 | info.gpl_compatible = prog->gpl_compatible; |
1e270976 MKL |
1916 | |
1917 | memcpy(info.tag, prog->tag, sizeof(prog->tag)); | |
cb4d2b3f MKL |
1918 | memcpy(info.name, prog->aux->name, sizeof(prog->aux->name)); |
1919 | ||
1920 | ulen = info.nr_map_ids; | |
1921 | info.nr_map_ids = prog->aux->used_map_cnt; | |
1922 | ulen = min_t(u32, info.nr_map_ids, ulen); | |
1923 | if (ulen) { | |
721e08da | 1924 | u32 __user *user_map_ids = u64_to_user_ptr(info.map_ids); |
cb4d2b3f MKL |
1925 | u32 i; |
1926 | ||
1927 | for (i = 0; i < ulen; i++) | |
1928 | if (put_user(prog->aux->used_maps[i]->id, | |
1929 | &user_map_ids[i])) | |
1930 | return -EFAULT; | |
1931 | } | |
1e270976 MKL |
1932 | |
1933 | if (!capable(CAP_SYS_ADMIN)) { | |
1934 | info.jited_prog_len = 0; | |
1935 | info.xlated_prog_len = 0; | |
dbecd738 | 1936 | info.nr_jited_ksyms = 0; |
1e270976 MKL |
1937 | goto done; |
1938 | } | |
1939 | ||
1e270976 | 1940 | ulen = info.xlated_prog_len; |
9975a54b | 1941 | info.xlated_prog_len = bpf_prog_insn_size(prog); |
1e270976 | 1942 | if (info.xlated_prog_len && ulen) { |
7105e828 DB |
1943 | struct bpf_insn *insns_sanitized; |
1944 | bool fault; | |
1945 | ||
1946 | if (prog->blinded && !bpf_dump_raw_ok()) { | |
1947 | info.xlated_prog_insns = 0; | |
1948 | goto done; | |
1949 | } | |
1950 | insns_sanitized = bpf_insn_prepare_dump(prog); | |
1951 | if (!insns_sanitized) | |
1952 | return -ENOMEM; | |
1e270976 MKL |
1953 | uinsns = u64_to_user_ptr(info.xlated_prog_insns); |
1954 | ulen = min_t(u32, info.xlated_prog_len, ulen); | |
7105e828 DB |
1955 | fault = copy_to_user(uinsns, insns_sanitized, ulen); |
1956 | kfree(insns_sanitized); | |
1957 | if (fault) | |
1e270976 MKL |
1958 | return -EFAULT; |
1959 | } | |
1960 | ||
675fc275 JK |
1961 | if (bpf_prog_is_dev_bound(prog->aux)) { |
1962 | err = bpf_prog_offload_info_fill(&info, prog); | |
1963 | if (err) | |
1964 | return err; | |
fcfb126d JW |
1965 | goto done; |
1966 | } | |
1967 | ||
1968 | /* NOTE: the following code is supposed to be skipped for offload. | |
1969 | * bpf_prog_offload_info_fill() is the place to fill similar fields | |
1970 | * for offload. | |
1971 | */ | |
1972 | ulen = info.jited_prog_len; | |
4d56a76e SD |
1973 | if (prog->aux->func_cnt) { |
1974 | u32 i; | |
1975 | ||
1976 | info.jited_prog_len = 0; | |
1977 | for (i = 0; i < prog->aux->func_cnt; i++) | |
1978 | info.jited_prog_len += prog->aux->func[i]->jited_len; | |
1979 | } else { | |
1980 | info.jited_prog_len = prog->jited_len; | |
1981 | } | |
1982 | ||
fcfb126d JW |
1983 | if (info.jited_prog_len && ulen) { |
1984 | if (bpf_dump_raw_ok()) { | |
1985 | uinsns = u64_to_user_ptr(info.jited_prog_insns); | |
1986 | ulen = min_t(u32, info.jited_prog_len, ulen); | |
4d56a76e SD |
1987 | |
1988 | /* for multi-function programs, copy the JITed | |
1989 | * instructions for all the functions | |
1990 | */ | |
1991 | if (prog->aux->func_cnt) { | |
1992 | u32 len, free, i; | |
1993 | u8 *img; | |
1994 | ||
1995 | free = ulen; | |
1996 | for (i = 0; i < prog->aux->func_cnt; i++) { | |
1997 | len = prog->aux->func[i]->jited_len; | |
1998 | len = min_t(u32, len, free); | |
1999 | img = (u8 *) prog->aux->func[i]->bpf_func; | |
2000 | if (copy_to_user(uinsns, img, len)) | |
2001 | return -EFAULT; | |
2002 | uinsns += len; | |
2003 | free -= len; | |
2004 | if (!free) | |
2005 | break; | |
2006 | } | |
2007 | } else { | |
2008 | if (copy_to_user(uinsns, prog->bpf_func, ulen)) | |
2009 | return -EFAULT; | |
2010 | } | |
fcfb126d JW |
2011 | } else { |
2012 | info.jited_prog_insns = 0; | |
2013 | } | |
675fc275 JK |
2014 | } |
2015 | ||
dbecd738 SD |
2016 | ulen = info.nr_jited_ksyms; |
2017 | info.nr_jited_ksyms = prog->aux->func_cnt; | |
2018 | if (info.nr_jited_ksyms && ulen) { | |
2019 | if (bpf_dump_raw_ok()) { | |
2020 | u64 __user *user_ksyms; | |
2021 | ulong ksym_addr; | |
2022 | u32 i; | |
2023 | ||
2024 | /* copy the address of the kernel symbol | |
2025 | * corresponding to each function | |
2026 | */ | |
2027 | ulen = min_t(u32, info.nr_jited_ksyms, ulen); | |
2028 | user_ksyms = u64_to_user_ptr(info.jited_ksyms); | |
2029 | for (i = 0; i < ulen; i++) { | |
2030 | ksym_addr = (ulong) prog->aux->func[i]->bpf_func; | |
2031 | ksym_addr &= PAGE_MASK; | |
2032 | if (put_user((u64) ksym_addr, &user_ksyms[i])) | |
2033 | return -EFAULT; | |
2034 | } | |
2035 | } else { | |
2036 | info.jited_ksyms = 0; | |
2037 | } | |
2038 | } | |
2039 | ||
1e270976 MKL |
2040 | done: |
2041 | if (copy_to_user(uinfo, &info, info_len) || | |
2042 | put_user(info_len, &uattr->info.info_len)) | |
2043 | return -EFAULT; | |
2044 | ||
2045 | return 0; | |
2046 | } | |
2047 | ||
2048 | static int bpf_map_get_info_by_fd(struct bpf_map *map, | |
2049 | const union bpf_attr *attr, | |
2050 | union bpf_attr __user *uattr) | |
2051 | { | |
2052 | struct bpf_map_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
2053 | struct bpf_map_info info = {}; | |
2054 | u32 info_len = attr->info.info_len; | |
2055 | int err; | |
2056 | ||
dcab51f1 | 2057 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(info), info_len); |
1e270976 MKL |
2058 | if (err) |
2059 | return err; | |
2060 | info_len = min_t(u32, sizeof(info), info_len); | |
2061 | ||
2062 | info.type = map->map_type; | |
2063 | info.id = map->id; | |
2064 | info.key_size = map->key_size; | |
2065 | info.value_size = map->value_size; | |
2066 | info.max_entries = map->max_entries; | |
2067 | info.map_flags = map->map_flags; | |
ad5b177b | 2068 | memcpy(info.name, map->name, sizeof(map->name)); |
1e270976 | 2069 | |
78958fca MKL |
2070 | if (map->btf) { |
2071 | info.btf_id = btf_id(map->btf); | |
9b2cf328 MKL |
2072 | info.btf_key_type_id = map->btf_key_type_id; |
2073 | info.btf_value_type_id = map->btf_value_type_id; | |
78958fca MKL |
2074 | } |
2075 | ||
52775b33 JK |
2076 | if (bpf_map_is_dev_bound(map)) { |
2077 | err = bpf_map_offload_info_fill(&info, map); | |
2078 | if (err) | |
2079 | return err; | |
2080 | } | |
2081 | ||
1e270976 MKL |
2082 | if (copy_to_user(uinfo, &info, info_len) || |
2083 | put_user(info_len, &uattr->info.info_len)) | |
2084 | return -EFAULT; | |
2085 | ||
2086 | return 0; | |
2087 | } | |
2088 | ||
62dab84c MKL |
2089 | static int bpf_btf_get_info_by_fd(struct btf *btf, |
2090 | const union bpf_attr *attr, | |
2091 | union bpf_attr __user *uattr) | |
2092 | { | |
2093 | struct bpf_btf_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
2094 | u32 info_len = attr->info.info_len; | |
2095 | int err; | |
2096 | ||
dcab51f1 | 2097 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(*uinfo), info_len); |
62dab84c MKL |
2098 | if (err) |
2099 | return err; | |
2100 | ||
2101 | return btf_get_info_by_fd(btf, attr, uattr); | |
2102 | } | |
2103 | ||
1e270976 MKL |
2104 | #define BPF_OBJ_GET_INFO_BY_FD_LAST_FIELD info.info |
2105 | ||
2106 | static int bpf_obj_get_info_by_fd(const union bpf_attr *attr, | |
2107 | union bpf_attr __user *uattr) | |
2108 | { | |
2109 | int ufd = attr->info.bpf_fd; | |
2110 | struct fd f; | |
2111 | int err; | |
2112 | ||
2113 | if (CHECK_ATTR(BPF_OBJ_GET_INFO_BY_FD)) | |
2114 | return -EINVAL; | |
2115 | ||
2116 | f = fdget(ufd); | |
2117 | if (!f.file) | |
2118 | return -EBADFD; | |
2119 | ||
2120 | if (f.file->f_op == &bpf_prog_fops) | |
2121 | err = bpf_prog_get_info_by_fd(f.file->private_data, attr, | |
2122 | uattr); | |
2123 | else if (f.file->f_op == &bpf_map_fops) | |
2124 | err = bpf_map_get_info_by_fd(f.file->private_data, attr, | |
2125 | uattr); | |
60197cfb | 2126 | else if (f.file->f_op == &btf_fops) |
62dab84c | 2127 | err = bpf_btf_get_info_by_fd(f.file->private_data, attr, uattr); |
1e270976 MKL |
2128 | else |
2129 | err = -EINVAL; | |
2130 | ||
2131 | fdput(f); | |
2132 | return err; | |
2133 | } | |
2134 | ||
f56a653c MKL |
2135 | #define BPF_BTF_LOAD_LAST_FIELD btf_log_level |
2136 | ||
2137 | static int bpf_btf_load(const union bpf_attr *attr) | |
2138 | { | |
2139 | if (CHECK_ATTR(BPF_BTF_LOAD)) | |
2140 | return -EINVAL; | |
2141 | ||
2142 | if (!capable(CAP_SYS_ADMIN)) | |
2143 | return -EPERM; | |
2144 | ||
2145 | return btf_new_fd(attr); | |
2146 | } | |
2147 | ||
78958fca MKL |
2148 | #define BPF_BTF_GET_FD_BY_ID_LAST_FIELD btf_id |
2149 | ||
2150 | static int bpf_btf_get_fd_by_id(const union bpf_attr *attr) | |
2151 | { | |
2152 | if (CHECK_ATTR(BPF_BTF_GET_FD_BY_ID)) | |
2153 | return -EINVAL; | |
2154 | ||
2155 | if (!capable(CAP_SYS_ADMIN)) | |
2156 | return -EPERM; | |
2157 | ||
2158 | return btf_get_fd_by_id(attr->btf_id); | |
2159 | } | |
2160 | ||
99c55f7d AS |
2161 | SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size) |
2162 | { | |
2163 | union bpf_attr attr = {}; | |
2164 | int err; | |
2165 | ||
0fa4fe85 | 2166 | if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN)) |
99c55f7d AS |
2167 | return -EPERM; |
2168 | ||
dcab51f1 | 2169 | err = bpf_check_uarg_tail_zero(uattr, sizeof(attr), size); |
1e270976 MKL |
2170 | if (err) |
2171 | return err; | |
2172 | size = min_t(u32, size, sizeof(attr)); | |
99c55f7d AS |
2173 | |
2174 | /* copy attributes from user space, may be less than sizeof(bpf_attr) */ | |
2175 | if (copy_from_user(&attr, uattr, size) != 0) | |
2176 | return -EFAULT; | |
2177 | ||
afdb09c7 CF |
2178 | err = security_bpf(cmd, &attr, size); |
2179 | if (err < 0) | |
2180 | return err; | |
2181 | ||
99c55f7d AS |
2182 | switch (cmd) { |
2183 | case BPF_MAP_CREATE: | |
2184 | err = map_create(&attr); | |
2185 | break; | |
db20fd2b AS |
2186 | case BPF_MAP_LOOKUP_ELEM: |
2187 | err = map_lookup_elem(&attr); | |
2188 | break; | |
2189 | case BPF_MAP_UPDATE_ELEM: | |
2190 | err = map_update_elem(&attr); | |
2191 | break; | |
2192 | case BPF_MAP_DELETE_ELEM: | |
2193 | err = map_delete_elem(&attr); | |
2194 | break; | |
2195 | case BPF_MAP_GET_NEXT_KEY: | |
2196 | err = map_get_next_key(&attr); | |
2197 | break; | |
09756af4 AS |
2198 | case BPF_PROG_LOAD: |
2199 | err = bpf_prog_load(&attr); | |
2200 | break; | |
b2197755 DB |
2201 | case BPF_OBJ_PIN: |
2202 | err = bpf_obj_pin(&attr); | |
2203 | break; | |
2204 | case BPF_OBJ_GET: | |
2205 | err = bpf_obj_get(&attr); | |
2206 | break; | |
f4324551 DM |
2207 | #ifdef CONFIG_CGROUP_BPF |
2208 | case BPF_PROG_ATTACH: | |
2209 | err = bpf_prog_attach(&attr); | |
2210 | break; | |
2211 | case BPF_PROG_DETACH: | |
2212 | err = bpf_prog_detach(&attr); | |
2213 | break; | |
468e2f64 AS |
2214 | case BPF_PROG_QUERY: |
2215 | err = bpf_prog_query(&attr, uattr); | |
2216 | break; | |
f4324551 | 2217 | #endif |
1cf1cae9 AS |
2218 | case BPF_PROG_TEST_RUN: |
2219 | err = bpf_prog_test_run(&attr, uattr); | |
2220 | break; | |
34ad5580 MKL |
2221 | case BPF_PROG_GET_NEXT_ID: |
2222 | err = bpf_obj_get_next_id(&attr, uattr, | |
2223 | &prog_idr, &prog_idr_lock); | |
2224 | break; | |
2225 | case BPF_MAP_GET_NEXT_ID: | |
2226 | err = bpf_obj_get_next_id(&attr, uattr, | |
2227 | &map_idr, &map_idr_lock); | |
2228 | break; | |
b16d9aa4 MKL |
2229 | case BPF_PROG_GET_FD_BY_ID: |
2230 | err = bpf_prog_get_fd_by_id(&attr); | |
2231 | break; | |
bd5f5f4e MKL |
2232 | case BPF_MAP_GET_FD_BY_ID: |
2233 | err = bpf_map_get_fd_by_id(&attr); | |
2234 | break; | |
1e270976 MKL |
2235 | case BPF_OBJ_GET_INFO_BY_FD: |
2236 | err = bpf_obj_get_info_by_fd(&attr, uattr); | |
2237 | break; | |
c4f6699d AS |
2238 | case BPF_RAW_TRACEPOINT_OPEN: |
2239 | err = bpf_raw_tracepoint_open(&attr); | |
2240 | break; | |
f56a653c MKL |
2241 | case BPF_BTF_LOAD: |
2242 | err = bpf_btf_load(&attr); | |
2243 | break; | |
78958fca MKL |
2244 | case BPF_BTF_GET_FD_BY_ID: |
2245 | err = bpf_btf_get_fd_by_id(&attr); | |
2246 | break; | |
99c55f7d AS |
2247 | default: |
2248 | err = -EINVAL; | |
2249 | break; | |
2250 | } | |
2251 | ||
2252 | return err; | |
2253 | } |