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> |
f4364dcf | 14 | #include <linux/bpf_lirc.h> |
f56a653c | 15 | #include <linux/btf.h> |
99c55f7d AS |
16 | #include <linux/syscalls.h> |
17 | #include <linux/slab.h> | |
3f07c014 | 18 | #include <linux/sched/signal.h> |
d407bd25 DB |
19 | #include <linux/vmalloc.h> |
20 | #include <linux/mmzone.h> | |
99c55f7d | 21 | #include <linux/anon_inodes.h> |
41bdc4b4 | 22 | #include <linux/fdtable.h> |
db20fd2b | 23 | #include <linux/file.h> |
41bdc4b4 | 24 | #include <linux/fs.h> |
09756af4 AS |
25 | #include <linux/license.h> |
26 | #include <linux/filter.h> | |
2541517c | 27 | #include <linux/version.h> |
535e7b4b | 28 | #include <linux/kernel.h> |
dc4bb0e2 | 29 | #include <linux/idr.h> |
cb4d2b3f MKL |
30 | #include <linux/cred.h> |
31 | #include <linux/timekeeping.h> | |
32 | #include <linux/ctype.h> | |
a26ca7c9 | 33 | #include <linux/btf.h> |
9ef09e35 | 34 | #include <linux/nospec.h> |
99c55f7d | 35 | |
14dc6f04 MKL |
36 | #define IS_FD_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PROG_ARRAY || \ |
37 | (map)->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || \ | |
38 | (map)->map_type == BPF_MAP_TYPE_CGROUP_ARRAY || \ | |
39 | (map)->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS) | |
40 | #define IS_FD_HASH(map) ((map)->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) | |
41 | #define IS_FD_MAP(map) (IS_FD_ARRAY(map) || IS_FD_HASH(map)) | |
42 | ||
6e71b04a CF |
43 | #define BPF_OBJ_FLAG_MASK (BPF_F_RDONLY | BPF_F_WRONLY) |
44 | ||
b121d1e7 | 45 | DEFINE_PER_CPU(int, bpf_prog_active); |
dc4bb0e2 MKL |
46 | static DEFINE_IDR(prog_idr); |
47 | static DEFINE_SPINLOCK(prog_idr_lock); | |
f3f1c054 MKL |
48 | static DEFINE_IDR(map_idr); |
49 | static DEFINE_SPINLOCK(map_idr_lock); | |
b121d1e7 | 50 | |
1be7f75d AS |
51 | int sysctl_unprivileged_bpf_disabled __read_mostly; |
52 | ||
40077e0c JB |
53 | static const struct bpf_map_ops * const bpf_map_types[] = { |
54 | #define BPF_PROG_TYPE(_id, _ops) | |
55 | #define BPF_MAP_TYPE(_id, _ops) \ | |
56 | [_id] = &_ops, | |
57 | #include <linux/bpf_types.h> | |
58 | #undef BPF_PROG_TYPE | |
59 | #undef BPF_MAP_TYPE | |
60 | }; | |
99c55f7d | 61 | |
752ba56f MS |
62 | /* |
63 | * If we're handed a bigger struct than we know of, ensure all the unknown bits | |
64 | * are 0 - i.e. new user-space does not rely on any kernel feature extensions | |
65 | * we don't know about yet. | |
66 | * | |
67 | * There is a ToCToU between this function call and the following | |
68 | * copy_from_user() call. However, this is not a concern since this function is | |
69 | * meant to be a future-proofing of bits. | |
70 | */ | |
dcab51f1 MKL |
71 | int bpf_check_uarg_tail_zero(void __user *uaddr, |
72 | size_t expected_size, | |
73 | size_t actual_size) | |
58291a74 MS |
74 | { |
75 | unsigned char __user *addr; | |
76 | unsigned char __user *end; | |
77 | unsigned char val; | |
78 | int err; | |
79 | ||
752ba56f MS |
80 | if (unlikely(actual_size > PAGE_SIZE)) /* silly large */ |
81 | return -E2BIG; | |
82 | ||
83 | if (unlikely(!access_ok(VERIFY_READ, uaddr, actual_size))) | |
84 | return -EFAULT; | |
85 | ||
58291a74 MS |
86 | if (actual_size <= expected_size) |
87 | return 0; | |
88 | ||
89 | addr = uaddr + expected_size; | |
90 | end = uaddr + actual_size; | |
91 | ||
92 | for (; addr < end; addr++) { | |
93 | err = get_user(val, addr); | |
94 | if (err) | |
95 | return err; | |
96 | if (val) | |
97 | return -E2BIG; | |
98 | } | |
99 | ||
100 | return 0; | |
101 | } | |
102 | ||
a3884572 JK |
103 | const struct bpf_map_ops bpf_map_offload_ops = { |
104 | .map_alloc = bpf_map_offload_map_alloc, | |
105 | .map_free = bpf_map_offload_map_free, | |
106 | }; | |
107 | ||
99c55f7d AS |
108 | static struct bpf_map *find_and_alloc_map(union bpf_attr *attr) |
109 | { | |
1110f3a9 | 110 | const struct bpf_map_ops *ops; |
9ef09e35 | 111 | u32 type = attr->map_type; |
99c55f7d | 112 | struct bpf_map *map; |
1110f3a9 | 113 | int err; |
99c55f7d | 114 | |
9ef09e35 | 115 | if (type >= ARRAY_SIZE(bpf_map_types)) |
1110f3a9 | 116 | return ERR_PTR(-EINVAL); |
9ef09e35 MR |
117 | type = array_index_nospec(type, ARRAY_SIZE(bpf_map_types)); |
118 | ops = bpf_map_types[type]; | |
1110f3a9 | 119 | if (!ops) |
40077e0c | 120 | return ERR_PTR(-EINVAL); |
99c55f7d | 121 | |
1110f3a9 JK |
122 | if (ops->map_alloc_check) { |
123 | err = ops->map_alloc_check(attr); | |
124 | if (err) | |
125 | return ERR_PTR(err); | |
126 | } | |
a3884572 JK |
127 | if (attr->map_ifindex) |
128 | ops = &bpf_map_offload_ops; | |
1110f3a9 | 129 | map = ops->map_alloc(attr); |
40077e0c JB |
130 | if (IS_ERR(map)) |
131 | return map; | |
1110f3a9 | 132 | map->ops = ops; |
9ef09e35 | 133 | map->map_type = type; |
40077e0c | 134 | return map; |
99c55f7d AS |
135 | } |
136 | ||
96eabe7a | 137 | void *bpf_map_area_alloc(size_t size, int numa_node) |
d407bd25 DB |
138 | { |
139 | /* We definitely need __GFP_NORETRY, so OOM killer doesn't | |
140 | * trigger under memory pressure as we really just want to | |
141 | * fail instead. | |
142 | */ | |
143 | const gfp_t flags = __GFP_NOWARN | __GFP_NORETRY | __GFP_ZERO; | |
144 | void *area; | |
145 | ||
146 | if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) { | |
96eabe7a | 147 | area = kmalloc_node(size, GFP_USER | flags, numa_node); |
d407bd25 DB |
148 | if (area != NULL) |
149 | return area; | |
150 | } | |
151 | ||
96eabe7a MKL |
152 | return __vmalloc_node_flags_caller(size, numa_node, GFP_KERNEL | flags, |
153 | __builtin_return_address(0)); | |
d407bd25 DB |
154 | } |
155 | ||
156 | void bpf_map_area_free(void *area) | |
157 | { | |
158 | kvfree(area); | |
159 | } | |
160 | ||
bd475643 JK |
161 | void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr) |
162 | { | |
163 | map->map_type = attr->map_type; | |
164 | map->key_size = attr->key_size; | |
165 | map->value_size = attr->value_size; | |
166 | map->max_entries = attr->max_entries; | |
167 | map->map_flags = attr->map_flags; | |
168 | map->numa_node = bpf_map_attr_numa_node(attr); | |
169 | } | |
170 | ||
6c905981 AS |
171 | int bpf_map_precharge_memlock(u32 pages) |
172 | { | |
173 | struct user_struct *user = get_current_user(); | |
174 | unsigned long memlock_limit, cur; | |
175 | ||
176 | memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
177 | cur = atomic_long_read(&user->locked_vm); | |
178 | free_uid(user); | |
179 | if (cur + pages > memlock_limit) | |
180 | return -EPERM; | |
181 | return 0; | |
182 | } | |
183 | ||
aaac3ba9 AS |
184 | static int bpf_map_charge_memlock(struct bpf_map *map) |
185 | { | |
186 | struct user_struct *user = get_current_user(); | |
187 | unsigned long memlock_limit; | |
188 | ||
189 | memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
190 | ||
191 | atomic_long_add(map->pages, &user->locked_vm); | |
192 | ||
193 | if (atomic_long_read(&user->locked_vm) > memlock_limit) { | |
194 | atomic_long_sub(map->pages, &user->locked_vm); | |
195 | free_uid(user); | |
196 | return -EPERM; | |
197 | } | |
198 | map->user = user; | |
199 | return 0; | |
200 | } | |
201 | ||
202 | static void bpf_map_uncharge_memlock(struct bpf_map *map) | |
203 | { | |
204 | struct user_struct *user = map->user; | |
205 | ||
206 | atomic_long_sub(map->pages, &user->locked_vm); | |
207 | free_uid(user); | |
208 | } | |
209 | ||
f3f1c054 MKL |
210 | static int bpf_map_alloc_id(struct bpf_map *map) |
211 | { | |
212 | int id; | |
213 | ||
b76354cd | 214 | idr_preload(GFP_KERNEL); |
f3f1c054 MKL |
215 | spin_lock_bh(&map_idr_lock); |
216 | id = idr_alloc_cyclic(&map_idr, map, 1, INT_MAX, GFP_ATOMIC); | |
217 | if (id > 0) | |
218 | map->id = id; | |
219 | spin_unlock_bh(&map_idr_lock); | |
b76354cd | 220 | idr_preload_end(); |
f3f1c054 MKL |
221 | |
222 | if (WARN_ON_ONCE(!id)) | |
223 | return -ENOSPC; | |
224 | ||
225 | return id > 0 ? 0 : id; | |
226 | } | |
227 | ||
a3884572 | 228 | void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock) |
f3f1c054 | 229 | { |
930651a7 ED |
230 | unsigned long flags; |
231 | ||
a3884572 JK |
232 | /* Offloaded maps are removed from the IDR store when their device |
233 | * disappears - even if someone holds an fd to them they are unusable, | |
234 | * the memory is gone, all ops will fail; they are simply waiting for | |
235 | * refcnt to drop to be freed. | |
236 | */ | |
237 | if (!map->id) | |
238 | return; | |
239 | ||
bd5f5f4e | 240 | if (do_idr_lock) |
930651a7 | 241 | spin_lock_irqsave(&map_idr_lock, flags); |
bd5f5f4e MKL |
242 | else |
243 | __acquire(&map_idr_lock); | |
244 | ||
f3f1c054 | 245 | idr_remove(&map_idr, map->id); |
a3884572 | 246 | map->id = 0; |
bd5f5f4e MKL |
247 | |
248 | if (do_idr_lock) | |
930651a7 | 249 | spin_unlock_irqrestore(&map_idr_lock, flags); |
bd5f5f4e MKL |
250 | else |
251 | __release(&map_idr_lock); | |
f3f1c054 MKL |
252 | } |
253 | ||
99c55f7d AS |
254 | /* called from workqueue */ |
255 | static void bpf_map_free_deferred(struct work_struct *work) | |
256 | { | |
257 | struct bpf_map *map = container_of(work, struct bpf_map, work); | |
258 | ||
aaac3ba9 | 259 | bpf_map_uncharge_memlock(map); |
afdb09c7 | 260 | security_bpf_map_free(map); |
99c55f7d AS |
261 | /* implementation dependent freeing */ |
262 | map->ops->map_free(map); | |
263 | } | |
264 | ||
c9da161c DB |
265 | static void bpf_map_put_uref(struct bpf_map *map) |
266 | { | |
267 | if (atomic_dec_and_test(&map->usercnt)) { | |
ba6b8de4 JF |
268 | if (map->ops->map_release_uref) |
269 | map->ops->map_release_uref(map); | |
c9da161c DB |
270 | } |
271 | } | |
272 | ||
99c55f7d AS |
273 | /* decrement map refcnt and schedule it for freeing via workqueue |
274 | * (unrelying map implementation ops->map_free() might sleep) | |
275 | */ | |
bd5f5f4e | 276 | static void __bpf_map_put(struct bpf_map *map, bool do_idr_lock) |
99c55f7d AS |
277 | { |
278 | if (atomic_dec_and_test(&map->refcnt)) { | |
34ad5580 | 279 | /* bpf_map_free_id() must be called first */ |
bd5f5f4e | 280 | bpf_map_free_id(map, do_idr_lock); |
78958fca | 281 | btf_put(map->btf); |
99c55f7d AS |
282 | INIT_WORK(&map->work, bpf_map_free_deferred); |
283 | schedule_work(&map->work); | |
284 | } | |
285 | } | |
286 | ||
bd5f5f4e MKL |
287 | void bpf_map_put(struct bpf_map *map) |
288 | { | |
289 | __bpf_map_put(map, true); | |
290 | } | |
630a4d38 | 291 | EXPORT_SYMBOL_GPL(bpf_map_put); |
bd5f5f4e | 292 | |
c9da161c | 293 | void bpf_map_put_with_uref(struct bpf_map *map) |
99c55f7d | 294 | { |
c9da161c | 295 | bpf_map_put_uref(map); |
99c55f7d | 296 | bpf_map_put(map); |
c9da161c DB |
297 | } |
298 | ||
299 | static int bpf_map_release(struct inode *inode, struct file *filp) | |
300 | { | |
61d1b6a4 DB |
301 | struct bpf_map *map = filp->private_data; |
302 | ||
303 | if (map->ops->map_release) | |
304 | map->ops->map_release(map, filp); | |
305 | ||
306 | bpf_map_put_with_uref(map); | |
99c55f7d AS |
307 | return 0; |
308 | } | |
309 | ||
f99bf205 DB |
310 | #ifdef CONFIG_PROC_FS |
311 | static void bpf_map_show_fdinfo(struct seq_file *m, struct file *filp) | |
312 | { | |
313 | const struct bpf_map *map = filp->private_data; | |
21116b70 DB |
314 | const struct bpf_array *array; |
315 | u32 owner_prog_type = 0; | |
9780c0ab | 316 | u32 owner_jited = 0; |
21116b70 DB |
317 | |
318 | if (map->map_type == BPF_MAP_TYPE_PROG_ARRAY) { | |
319 | array = container_of(map, struct bpf_array, map); | |
320 | owner_prog_type = array->owner_prog_type; | |
9780c0ab | 321 | owner_jited = array->owner_jited; |
21116b70 | 322 | } |
f99bf205 DB |
323 | |
324 | seq_printf(m, | |
325 | "map_type:\t%u\n" | |
326 | "key_size:\t%u\n" | |
327 | "value_size:\t%u\n" | |
322cea2f | 328 | "max_entries:\t%u\n" |
21116b70 | 329 | "map_flags:\t%#x\n" |
4316b409 DB |
330 | "memlock:\t%llu\n" |
331 | "map_id:\t%u\n", | |
f99bf205 DB |
332 | map->map_type, |
333 | map->key_size, | |
334 | map->value_size, | |
322cea2f | 335 | map->max_entries, |
21116b70 | 336 | map->map_flags, |
4316b409 DB |
337 | map->pages * 1ULL << PAGE_SHIFT, |
338 | map->id); | |
21116b70 | 339 | |
9780c0ab | 340 | if (owner_prog_type) { |
21116b70 DB |
341 | seq_printf(m, "owner_prog_type:\t%u\n", |
342 | owner_prog_type); | |
9780c0ab DB |
343 | seq_printf(m, "owner_jited:\t%u\n", |
344 | owner_jited); | |
345 | } | |
f99bf205 DB |
346 | } |
347 | #endif | |
348 | ||
6e71b04a CF |
349 | static ssize_t bpf_dummy_read(struct file *filp, char __user *buf, size_t siz, |
350 | loff_t *ppos) | |
351 | { | |
352 | /* We need this handler such that alloc_file() enables | |
353 | * f_mode with FMODE_CAN_READ. | |
354 | */ | |
355 | return -EINVAL; | |
356 | } | |
357 | ||
358 | static ssize_t bpf_dummy_write(struct file *filp, const char __user *buf, | |
359 | size_t siz, loff_t *ppos) | |
360 | { | |
361 | /* We need this handler such that alloc_file() enables | |
362 | * f_mode with FMODE_CAN_WRITE. | |
363 | */ | |
364 | return -EINVAL; | |
365 | } | |
366 | ||
f66e448c | 367 | const struct file_operations bpf_map_fops = { |
f99bf205 DB |
368 | #ifdef CONFIG_PROC_FS |
369 | .show_fdinfo = bpf_map_show_fdinfo, | |
370 | #endif | |
371 | .release = bpf_map_release, | |
6e71b04a CF |
372 | .read = bpf_dummy_read, |
373 | .write = bpf_dummy_write, | |
99c55f7d AS |
374 | }; |
375 | ||
6e71b04a | 376 | int bpf_map_new_fd(struct bpf_map *map, int flags) |
aa79781b | 377 | { |
afdb09c7 CF |
378 | int ret; |
379 | ||
380 | ret = security_bpf_map(map, OPEN_FMODE(flags)); | |
381 | if (ret < 0) | |
382 | return ret; | |
383 | ||
aa79781b | 384 | return anon_inode_getfd("bpf-map", &bpf_map_fops, map, |
6e71b04a CF |
385 | flags | O_CLOEXEC); |
386 | } | |
387 | ||
388 | int bpf_get_file_flag(int flags) | |
389 | { | |
390 | if ((flags & BPF_F_RDONLY) && (flags & BPF_F_WRONLY)) | |
391 | return -EINVAL; | |
392 | if (flags & BPF_F_RDONLY) | |
393 | return O_RDONLY; | |
394 | if (flags & BPF_F_WRONLY) | |
395 | return O_WRONLY; | |
396 | return O_RDWR; | |
aa79781b DB |
397 | } |
398 | ||
99c55f7d AS |
399 | /* helper macro to check that unused fields 'union bpf_attr' are zero */ |
400 | #define CHECK_ATTR(CMD) \ | |
401 | memchr_inv((void *) &attr->CMD##_LAST_FIELD + \ | |
402 | sizeof(attr->CMD##_LAST_FIELD), 0, \ | |
403 | sizeof(*attr) - \ | |
404 | offsetof(union bpf_attr, CMD##_LAST_FIELD) - \ | |
405 | sizeof(attr->CMD##_LAST_FIELD)) != NULL | |
406 | ||
cb4d2b3f MKL |
407 | /* dst and src must have at least BPF_OBJ_NAME_LEN number of bytes. |
408 | * Return 0 on success and < 0 on error. | |
409 | */ | |
410 | static int bpf_obj_name_cpy(char *dst, const char *src) | |
411 | { | |
412 | const char *end = src + BPF_OBJ_NAME_LEN; | |
413 | ||
473d9734 MKL |
414 | memset(dst, 0, BPF_OBJ_NAME_LEN); |
415 | ||
cb4d2b3f MKL |
416 | /* Copy all isalnum() and '_' char */ |
417 | while (src < end && *src) { | |
418 | if (!isalnum(*src) && *src != '_') | |
419 | return -EINVAL; | |
420 | *dst++ = *src++; | |
421 | } | |
422 | ||
423 | /* No '\0' found in BPF_OBJ_NAME_LEN number of bytes */ | |
424 | if (src == end) | |
425 | return -EINVAL; | |
426 | ||
cb4d2b3f MKL |
427 | return 0; |
428 | } | |
429 | ||
9b2cf328 | 430 | #define BPF_MAP_CREATE_LAST_FIELD btf_value_type_id |
99c55f7d AS |
431 | /* called via syscall */ |
432 | static int map_create(union bpf_attr *attr) | |
433 | { | |
96eabe7a | 434 | int numa_node = bpf_map_attr_numa_node(attr); |
99c55f7d | 435 | struct bpf_map *map; |
6e71b04a | 436 | int f_flags; |
99c55f7d AS |
437 | int err; |
438 | ||
439 | err = CHECK_ATTR(BPF_MAP_CREATE); | |
440 | if (err) | |
441 | return -EINVAL; | |
442 | ||
6e71b04a CF |
443 | f_flags = bpf_get_file_flag(attr->map_flags); |
444 | if (f_flags < 0) | |
445 | return f_flags; | |
446 | ||
96eabe7a | 447 | if (numa_node != NUMA_NO_NODE && |
96e5ae4e ED |
448 | ((unsigned int)numa_node >= nr_node_ids || |
449 | !node_online(numa_node))) | |
96eabe7a MKL |
450 | return -EINVAL; |
451 | ||
99c55f7d AS |
452 | /* find map type and init map: hashtable vs rbtree vs bloom vs ... */ |
453 | map = find_and_alloc_map(attr); | |
454 | if (IS_ERR(map)) | |
455 | return PTR_ERR(map); | |
456 | ||
ad5b177b MKL |
457 | err = bpf_obj_name_cpy(map->name, attr->map_name); |
458 | if (err) | |
459 | goto free_map_nouncharge; | |
460 | ||
99c55f7d | 461 | atomic_set(&map->refcnt, 1); |
c9da161c | 462 | atomic_set(&map->usercnt, 1); |
99c55f7d | 463 | |
a26ca7c9 | 464 | if (bpf_map_support_seq_show(map) && |
9b2cf328 | 465 | (attr->btf_key_type_id || attr->btf_value_type_id)) { |
a26ca7c9 MKL |
466 | struct btf *btf; |
467 | ||
9b2cf328 | 468 | if (!attr->btf_key_type_id || !attr->btf_value_type_id) { |
a26ca7c9 MKL |
469 | err = -EINVAL; |
470 | goto free_map_nouncharge; | |
471 | } | |
472 | ||
473 | btf = btf_get_by_fd(attr->btf_fd); | |
474 | if (IS_ERR(btf)) { | |
475 | err = PTR_ERR(btf); | |
476 | goto free_map_nouncharge; | |
477 | } | |
478 | ||
9b2cf328 MKL |
479 | err = map->ops->map_check_btf(map, btf, attr->btf_key_type_id, |
480 | attr->btf_value_type_id); | |
a26ca7c9 MKL |
481 | if (err) { |
482 | btf_put(btf); | |
483 | goto free_map_nouncharge; | |
484 | } | |
485 | ||
486 | map->btf = btf; | |
9b2cf328 MKL |
487 | map->btf_key_type_id = attr->btf_key_type_id; |
488 | map->btf_value_type_id = attr->btf_value_type_id; | |
a26ca7c9 MKL |
489 | } |
490 | ||
afdb09c7 | 491 | err = security_bpf_map_alloc(map); |
aaac3ba9 | 492 | if (err) |
20b2b24f | 493 | goto free_map_nouncharge; |
aaac3ba9 | 494 | |
afdb09c7 CF |
495 | err = bpf_map_charge_memlock(map); |
496 | if (err) | |
497 | goto free_map_sec; | |
498 | ||
f3f1c054 MKL |
499 | err = bpf_map_alloc_id(map); |
500 | if (err) | |
501 | goto free_map; | |
502 | ||
6e71b04a | 503 | err = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
504 | if (err < 0) { |
505 | /* failed to allocate fd. | |
506 | * bpf_map_put() is needed because the above | |
507 | * bpf_map_alloc_id() has published the map | |
508 | * to the userspace and the userspace may | |
509 | * have refcnt-ed it through BPF_MAP_GET_FD_BY_ID. | |
510 | */ | |
511 | bpf_map_put(map); | |
512 | return err; | |
513 | } | |
99c55f7d AS |
514 | |
515 | return err; | |
516 | ||
517 | free_map: | |
20b2b24f | 518 | bpf_map_uncharge_memlock(map); |
afdb09c7 CF |
519 | free_map_sec: |
520 | security_bpf_map_free(map); | |
20b2b24f | 521 | free_map_nouncharge: |
a26ca7c9 | 522 | btf_put(map->btf); |
99c55f7d AS |
523 | map->ops->map_free(map); |
524 | return err; | |
525 | } | |
526 | ||
db20fd2b AS |
527 | /* if error is returned, fd is released. |
528 | * On success caller should complete fd access with matching fdput() | |
529 | */ | |
c2101297 | 530 | struct bpf_map *__bpf_map_get(struct fd f) |
db20fd2b | 531 | { |
db20fd2b AS |
532 | if (!f.file) |
533 | return ERR_PTR(-EBADF); | |
db20fd2b AS |
534 | if (f.file->f_op != &bpf_map_fops) { |
535 | fdput(f); | |
536 | return ERR_PTR(-EINVAL); | |
537 | } | |
538 | ||
c2101297 DB |
539 | return f.file->private_data; |
540 | } | |
541 | ||
92117d84 AS |
542 | /* prog's and map's refcnt limit */ |
543 | #define BPF_MAX_REFCNT 32768 | |
544 | ||
545 | struct bpf_map *bpf_map_inc(struct bpf_map *map, bool uref) | |
c9da161c | 546 | { |
92117d84 AS |
547 | if (atomic_inc_return(&map->refcnt) > BPF_MAX_REFCNT) { |
548 | atomic_dec(&map->refcnt); | |
549 | return ERR_PTR(-EBUSY); | |
550 | } | |
c9da161c DB |
551 | if (uref) |
552 | atomic_inc(&map->usercnt); | |
92117d84 | 553 | return map; |
c9da161c | 554 | } |
630a4d38 | 555 | EXPORT_SYMBOL_GPL(bpf_map_inc); |
c9da161c DB |
556 | |
557 | struct bpf_map *bpf_map_get_with_uref(u32 ufd) | |
c2101297 DB |
558 | { |
559 | struct fd f = fdget(ufd); | |
560 | struct bpf_map *map; | |
561 | ||
562 | map = __bpf_map_get(f); | |
563 | if (IS_ERR(map)) | |
564 | return map; | |
565 | ||
92117d84 | 566 | map = bpf_map_inc(map, true); |
c2101297 | 567 | fdput(f); |
db20fd2b AS |
568 | |
569 | return map; | |
570 | } | |
571 | ||
bd5f5f4e MKL |
572 | /* map_idr_lock should have been held */ |
573 | static struct bpf_map *bpf_map_inc_not_zero(struct bpf_map *map, | |
574 | bool uref) | |
575 | { | |
576 | int refold; | |
577 | ||
578 | refold = __atomic_add_unless(&map->refcnt, 1, 0); | |
579 | ||
580 | if (refold >= BPF_MAX_REFCNT) { | |
581 | __bpf_map_put(map, false); | |
582 | return ERR_PTR(-EBUSY); | |
583 | } | |
584 | ||
585 | if (!refold) | |
586 | return ERR_PTR(-ENOENT); | |
587 | ||
588 | if (uref) | |
589 | atomic_inc(&map->usercnt); | |
590 | ||
591 | return map; | |
592 | } | |
593 | ||
b8cdc051 AS |
594 | int __weak bpf_stackmap_copy(struct bpf_map *map, void *key, void *value) |
595 | { | |
596 | return -ENOTSUPP; | |
597 | } | |
598 | ||
db20fd2b AS |
599 | /* last field in 'union bpf_attr' used by this command */ |
600 | #define BPF_MAP_LOOKUP_ELEM_LAST_FIELD value | |
601 | ||
602 | static int map_lookup_elem(union bpf_attr *attr) | |
603 | { | |
535e7b4b MS |
604 | void __user *ukey = u64_to_user_ptr(attr->key); |
605 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 606 | int ufd = attr->map_fd; |
db20fd2b | 607 | struct bpf_map *map; |
8ebe667c | 608 | void *key, *value, *ptr; |
15a07b33 | 609 | u32 value_size; |
592867bf | 610 | struct fd f; |
db20fd2b AS |
611 | int err; |
612 | ||
613 | if (CHECK_ATTR(BPF_MAP_LOOKUP_ELEM)) | |
614 | return -EINVAL; | |
615 | ||
592867bf | 616 | f = fdget(ufd); |
c2101297 | 617 | map = __bpf_map_get(f); |
db20fd2b AS |
618 | if (IS_ERR(map)) |
619 | return PTR_ERR(map); | |
620 | ||
6e71b04a CF |
621 | if (!(f.file->f_mode & FMODE_CAN_READ)) { |
622 | err = -EPERM; | |
623 | goto err_put; | |
624 | } | |
625 | ||
e4448ed8 AV |
626 | key = memdup_user(ukey, map->key_size); |
627 | if (IS_ERR(key)) { | |
628 | err = PTR_ERR(key); | |
db20fd2b | 629 | goto err_put; |
e4448ed8 | 630 | } |
db20fd2b | 631 | |
15a07b33 | 632 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
8f844938 | 633 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || |
15a07b33 AS |
634 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) |
635 | value_size = round_up(map->value_size, 8) * num_possible_cpus(); | |
14dc6f04 MKL |
636 | else if (IS_FD_MAP(map)) |
637 | value_size = sizeof(u32); | |
15a07b33 AS |
638 | else |
639 | value_size = map->value_size; | |
640 | ||
8ebe667c | 641 | err = -ENOMEM; |
15a07b33 | 642 | value = kmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b | 643 | if (!value) |
8ebe667c AS |
644 | goto free_key; |
645 | ||
a3884572 JK |
646 | if (bpf_map_is_dev_bound(map)) { |
647 | err = bpf_map_offload_lookup_elem(map, key, value); | |
648 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || | |
649 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
15a07b33 AS |
650 | err = bpf_percpu_hash_copy(map, key, value); |
651 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
652 | err = bpf_percpu_array_copy(map, key, value); | |
557c0c6e AS |
653 | } else if (map->map_type == BPF_MAP_TYPE_STACK_TRACE) { |
654 | err = bpf_stackmap_copy(map, key, value); | |
14dc6f04 MKL |
655 | } else if (IS_FD_ARRAY(map)) { |
656 | err = bpf_fd_array_map_lookup_elem(map, key, value); | |
657 | } else if (IS_FD_HASH(map)) { | |
658 | err = bpf_fd_htab_map_lookup_elem(map, key, value); | |
15a07b33 AS |
659 | } else { |
660 | rcu_read_lock(); | |
661 | ptr = map->ops->map_lookup_elem(map, key); | |
662 | if (ptr) | |
663 | memcpy(value, ptr, value_size); | |
664 | rcu_read_unlock(); | |
665 | err = ptr ? 0 : -ENOENT; | |
666 | } | |
8ebe667c | 667 | |
15a07b33 | 668 | if (err) |
8ebe667c | 669 | goto free_value; |
db20fd2b AS |
670 | |
671 | err = -EFAULT; | |
15a07b33 | 672 | if (copy_to_user(uvalue, value, value_size) != 0) |
8ebe667c | 673 | goto free_value; |
db20fd2b AS |
674 | |
675 | err = 0; | |
676 | ||
8ebe667c AS |
677 | free_value: |
678 | kfree(value); | |
db20fd2b AS |
679 | free_key: |
680 | kfree(key); | |
681 | err_put: | |
682 | fdput(f); | |
683 | return err; | |
684 | } | |
685 | ||
3274f520 | 686 | #define BPF_MAP_UPDATE_ELEM_LAST_FIELD flags |
db20fd2b AS |
687 | |
688 | static int map_update_elem(union bpf_attr *attr) | |
689 | { | |
535e7b4b MS |
690 | void __user *ukey = u64_to_user_ptr(attr->key); |
691 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 692 | int ufd = attr->map_fd; |
db20fd2b AS |
693 | struct bpf_map *map; |
694 | void *key, *value; | |
15a07b33 | 695 | u32 value_size; |
592867bf | 696 | struct fd f; |
db20fd2b AS |
697 | int err; |
698 | ||
699 | if (CHECK_ATTR(BPF_MAP_UPDATE_ELEM)) | |
700 | return -EINVAL; | |
701 | ||
592867bf | 702 | f = fdget(ufd); |
c2101297 | 703 | map = __bpf_map_get(f); |
db20fd2b AS |
704 | if (IS_ERR(map)) |
705 | return PTR_ERR(map); | |
706 | ||
6e71b04a CF |
707 | if (!(f.file->f_mode & FMODE_CAN_WRITE)) { |
708 | err = -EPERM; | |
709 | goto err_put; | |
710 | } | |
711 | ||
e4448ed8 AV |
712 | key = memdup_user(ukey, map->key_size); |
713 | if (IS_ERR(key)) { | |
714 | err = PTR_ERR(key); | |
db20fd2b | 715 | goto err_put; |
e4448ed8 | 716 | } |
db20fd2b | 717 | |
15a07b33 | 718 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
8f844938 | 719 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || |
15a07b33 AS |
720 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) |
721 | value_size = round_up(map->value_size, 8) * num_possible_cpus(); | |
722 | else | |
723 | value_size = map->value_size; | |
724 | ||
db20fd2b | 725 | err = -ENOMEM; |
15a07b33 | 726 | value = kmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b AS |
727 | if (!value) |
728 | goto free_key; | |
729 | ||
730 | err = -EFAULT; | |
15a07b33 | 731 | if (copy_from_user(value, uvalue, value_size) != 0) |
db20fd2b AS |
732 | goto free_value; |
733 | ||
6710e112 | 734 | /* Need to create a kthread, thus must support schedule */ |
a3884572 JK |
735 | if (bpf_map_is_dev_bound(map)) { |
736 | err = bpf_map_offload_update_elem(map, key, value, attr->flags); | |
737 | goto out; | |
738 | } else if (map->map_type == BPF_MAP_TYPE_CPUMAP) { | |
6710e112 JDB |
739 | err = map->ops->map_update_elem(map, key, value, attr->flags); |
740 | goto out; | |
741 | } | |
742 | ||
b121d1e7 AS |
743 | /* must increment bpf_prog_active to avoid kprobe+bpf triggering from |
744 | * inside bpf map update or delete otherwise deadlocks are possible | |
745 | */ | |
746 | preempt_disable(); | |
747 | __this_cpu_inc(bpf_prog_active); | |
8f844938 MKL |
748 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
749 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
15a07b33 AS |
750 | err = bpf_percpu_hash_update(map, key, value, attr->flags); |
751 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
752 | err = bpf_percpu_array_update(map, key, value, attr->flags); | |
9c147b56 | 753 | } else if (IS_FD_ARRAY(map)) { |
d056a788 DB |
754 | rcu_read_lock(); |
755 | err = bpf_fd_array_map_update_elem(map, f.file, key, value, | |
756 | attr->flags); | |
757 | rcu_read_unlock(); | |
bcc6b1b7 MKL |
758 | } else if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) { |
759 | rcu_read_lock(); | |
760 | err = bpf_fd_htab_map_update_elem(map, f.file, key, value, | |
761 | attr->flags); | |
762 | rcu_read_unlock(); | |
15a07b33 AS |
763 | } else { |
764 | rcu_read_lock(); | |
765 | err = map->ops->map_update_elem(map, key, value, attr->flags); | |
766 | rcu_read_unlock(); | |
767 | } | |
b121d1e7 AS |
768 | __this_cpu_dec(bpf_prog_active); |
769 | preempt_enable(); | |
6710e112 | 770 | out: |
db20fd2b AS |
771 | free_value: |
772 | kfree(value); | |
773 | free_key: | |
774 | kfree(key); | |
775 | err_put: | |
776 | fdput(f); | |
777 | return err; | |
778 | } | |
779 | ||
780 | #define BPF_MAP_DELETE_ELEM_LAST_FIELD key | |
781 | ||
782 | static int map_delete_elem(union bpf_attr *attr) | |
783 | { | |
535e7b4b | 784 | void __user *ukey = u64_to_user_ptr(attr->key); |
db20fd2b | 785 | int ufd = attr->map_fd; |
db20fd2b | 786 | struct bpf_map *map; |
592867bf | 787 | struct fd f; |
db20fd2b AS |
788 | void *key; |
789 | int err; | |
790 | ||
791 | if (CHECK_ATTR(BPF_MAP_DELETE_ELEM)) | |
792 | return -EINVAL; | |
793 | ||
592867bf | 794 | f = fdget(ufd); |
c2101297 | 795 | map = __bpf_map_get(f); |
db20fd2b AS |
796 | if (IS_ERR(map)) |
797 | return PTR_ERR(map); | |
798 | ||
6e71b04a CF |
799 | if (!(f.file->f_mode & FMODE_CAN_WRITE)) { |
800 | err = -EPERM; | |
801 | goto err_put; | |
802 | } | |
803 | ||
e4448ed8 AV |
804 | key = memdup_user(ukey, map->key_size); |
805 | if (IS_ERR(key)) { | |
806 | err = PTR_ERR(key); | |
db20fd2b | 807 | goto err_put; |
e4448ed8 | 808 | } |
db20fd2b | 809 | |
a3884572 JK |
810 | if (bpf_map_is_dev_bound(map)) { |
811 | err = bpf_map_offload_delete_elem(map, key); | |
812 | goto out; | |
813 | } | |
814 | ||
b121d1e7 AS |
815 | preempt_disable(); |
816 | __this_cpu_inc(bpf_prog_active); | |
db20fd2b AS |
817 | rcu_read_lock(); |
818 | err = map->ops->map_delete_elem(map, key); | |
819 | rcu_read_unlock(); | |
b121d1e7 AS |
820 | __this_cpu_dec(bpf_prog_active); |
821 | preempt_enable(); | |
a3884572 | 822 | out: |
db20fd2b AS |
823 | kfree(key); |
824 | err_put: | |
825 | fdput(f); | |
826 | return err; | |
827 | } | |
828 | ||
829 | /* last field in 'union bpf_attr' used by this command */ | |
830 | #define BPF_MAP_GET_NEXT_KEY_LAST_FIELD next_key | |
831 | ||
832 | static int map_get_next_key(union bpf_attr *attr) | |
833 | { | |
535e7b4b MS |
834 | void __user *ukey = u64_to_user_ptr(attr->key); |
835 | void __user *unext_key = u64_to_user_ptr(attr->next_key); | |
db20fd2b | 836 | int ufd = attr->map_fd; |
db20fd2b AS |
837 | struct bpf_map *map; |
838 | void *key, *next_key; | |
592867bf | 839 | struct fd f; |
db20fd2b AS |
840 | int err; |
841 | ||
842 | if (CHECK_ATTR(BPF_MAP_GET_NEXT_KEY)) | |
843 | return -EINVAL; | |
844 | ||
592867bf | 845 | f = fdget(ufd); |
c2101297 | 846 | map = __bpf_map_get(f); |
db20fd2b AS |
847 | if (IS_ERR(map)) |
848 | return PTR_ERR(map); | |
849 | ||
6e71b04a CF |
850 | if (!(f.file->f_mode & FMODE_CAN_READ)) { |
851 | err = -EPERM; | |
852 | goto err_put; | |
853 | } | |
854 | ||
8fe45924 | 855 | if (ukey) { |
e4448ed8 AV |
856 | key = memdup_user(ukey, map->key_size); |
857 | if (IS_ERR(key)) { | |
858 | err = PTR_ERR(key); | |
8fe45924 | 859 | goto err_put; |
e4448ed8 | 860 | } |
8fe45924 TQ |
861 | } else { |
862 | key = NULL; | |
863 | } | |
db20fd2b AS |
864 | |
865 | err = -ENOMEM; | |
866 | next_key = kmalloc(map->key_size, GFP_USER); | |
867 | if (!next_key) | |
868 | goto free_key; | |
869 | ||
a3884572 JK |
870 | if (bpf_map_is_dev_bound(map)) { |
871 | err = bpf_map_offload_get_next_key(map, key, next_key); | |
872 | goto out; | |
873 | } | |
874 | ||
db20fd2b AS |
875 | rcu_read_lock(); |
876 | err = map->ops->map_get_next_key(map, key, next_key); | |
877 | rcu_read_unlock(); | |
a3884572 | 878 | out: |
db20fd2b AS |
879 | if (err) |
880 | goto free_next_key; | |
881 | ||
882 | err = -EFAULT; | |
883 | if (copy_to_user(unext_key, next_key, map->key_size) != 0) | |
884 | goto free_next_key; | |
885 | ||
886 | err = 0; | |
887 | ||
888 | free_next_key: | |
889 | kfree(next_key); | |
890 | free_key: | |
891 | kfree(key); | |
892 | err_put: | |
893 | fdput(f); | |
894 | return err; | |
895 | } | |
896 | ||
7de16e3a JK |
897 | static const struct bpf_prog_ops * const bpf_prog_types[] = { |
898 | #define BPF_PROG_TYPE(_id, _name) \ | |
899 | [_id] = & _name ## _prog_ops, | |
900 | #define BPF_MAP_TYPE(_id, _ops) | |
901 | #include <linux/bpf_types.h> | |
902 | #undef BPF_PROG_TYPE | |
903 | #undef BPF_MAP_TYPE | |
904 | }; | |
905 | ||
09756af4 AS |
906 | static int find_prog_type(enum bpf_prog_type type, struct bpf_prog *prog) |
907 | { | |
d0f1a451 DB |
908 | const struct bpf_prog_ops *ops; |
909 | ||
910 | if (type >= ARRAY_SIZE(bpf_prog_types)) | |
911 | return -EINVAL; | |
912 | type = array_index_nospec(type, ARRAY_SIZE(bpf_prog_types)); | |
913 | ops = bpf_prog_types[type]; | |
914 | if (!ops) | |
be9370a7 | 915 | return -EINVAL; |
09756af4 | 916 | |
ab3f0063 | 917 | if (!bpf_prog_is_dev_bound(prog->aux)) |
d0f1a451 | 918 | prog->aux->ops = ops; |
ab3f0063 JK |
919 | else |
920 | prog->aux->ops = &bpf_offload_prog_ops; | |
be9370a7 JB |
921 | prog->type = type; |
922 | return 0; | |
09756af4 AS |
923 | } |
924 | ||
925 | /* drop refcnt on maps used by eBPF program and free auxilary data */ | |
926 | static void free_used_maps(struct bpf_prog_aux *aux) | |
927 | { | |
928 | int i; | |
929 | ||
930 | for (i = 0; i < aux->used_map_cnt; i++) | |
931 | bpf_map_put(aux->used_maps[i]); | |
932 | ||
933 | kfree(aux->used_maps); | |
934 | } | |
935 | ||
5ccb071e DB |
936 | int __bpf_prog_charge(struct user_struct *user, u32 pages) |
937 | { | |
938 | unsigned long memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
939 | unsigned long user_bufs; | |
940 | ||
941 | if (user) { | |
942 | user_bufs = atomic_long_add_return(pages, &user->locked_vm); | |
943 | if (user_bufs > memlock_limit) { | |
944 | atomic_long_sub(pages, &user->locked_vm); | |
945 | return -EPERM; | |
946 | } | |
947 | } | |
948 | ||
949 | return 0; | |
950 | } | |
951 | ||
952 | void __bpf_prog_uncharge(struct user_struct *user, u32 pages) | |
953 | { | |
954 | if (user) | |
955 | atomic_long_sub(pages, &user->locked_vm); | |
956 | } | |
957 | ||
aaac3ba9 AS |
958 | static int bpf_prog_charge_memlock(struct bpf_prog *prog) |
959 | { | |
960 | struct user_struct *user = get_current_user(); | |
5ccb071e | 961 | int ret; |
aaac3ba9 | 962 | |
5ccb071e DB |
963 | ret = __bpf_prog_charge(user, prog->pages); |
964 | if (ret) { | |
aaac3ba9 | 965 | free_uid(user); |
5ccb071e | 966 | return ret; |
aaac3ba9 | 967 | } |
5ccb071e | 968 | |
aaac3ba9 AS |
969 | prog->aux->user = user; |
970 | return 0; | |
971 | } | |
972 | ||
973 | static void bpf_prog_uncharge_memlock(struct bpf_prog *prog) | |
974 | { | |
975 | struct user_struct *user = prog->aux->user; | |
976 | ||
5ccb071e | 977 | __bpf_prog_uncharge(user, prog->pages); |
aaac3ba9 AS |
978 | free_uid(user); |
979 | } | |
980 | ||
dc4bb0e2 MKL |
981 | static int bpf_prog_alloc_id(struct bpf_prog *prog) |
982 | { | |
983 | int id; | |
984 | ||
b76354cd | 985 | idr_preload(GFP_KERNEL); |
dc4bb0e2 MKL |
986 | spin_lock_bh(&prog_idr_lock); |
987 | id = idr_alloc_cyclic(&prog_idr, prog, 1, INT_MAX, GFP_ATOMIC); | |
988 | if (id > 0) | |
989 | prog->aux->id = id; | |
990 | spin_unlock_bh(&prog_idr_lock); | |
b76354cd | 991 | idr_preload_end(); |
dc4bb0e2 MKL |
992 | |
993 | /* id is in [1, INT_MAX) */ | |
994 | if (WARN_ON_ONCE(!id)) | |
995 | return -ENOSPC; | |
996 | ||
997 | return id > 0 ? 0 : id; | |
998 | } | |
999 | ||
ad8ad79f | 1000 | void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock) |
dc4bb0e2 | 1001 | { |
ad8ad79f JK |
1002 | /* cBPF to eBPF migrations are currently not in the idr store. |
1003 | * Offloaded programs are removed from the store when their device | |
1004 | * disappears - even if someone grabs an fd to them they are unusable, | |
1005 | * simply waiting for refcnt to drop to be freed. | |
1006 | */ | |
dc4bb0e2 MKL |
1007 | if (!prog->aux->id) |
1008 | return; | |
1009 | ||
b16d9aa4 MKL |
1010 | if (do_idr_lock) |
1011 | spin_lock_bh(&prog_idr_lock); | |
1012 | else | |
1013 | __acquire(&prog_idr_lock); | |
1014 | ||
dc4bb0e2 | 1015 | idr_remove(&prog_idr, prog->aux->id); |
ad8ad79f | 1016 | prog->aux->id = 0; |
b16d9aa4 MKL |
1017 | |
1018 | if (do_idr_lock) | |
1019 | spin_unlock_bh(&prog_idr_lock); | |
1020 | else | |
1021 | __release(&prog_idr_lock); | |
dc4bb0e2 MKL |
1022 | } |
1023 | ||
1aacde3d | 1024 | static void __bpf_prog_put_rcu(struct rcu_head *rcu) |
abf2e7d6 AS |
1025 | { |
1026 | struct bpf_prog_aux *aux = container_of(rcu, struct bpf_prog_aux, rcu); | |
1027 | ||
1028 | free_used_maps(aux); | |
aaac3ba9 | 1029 | bpf_prog_uncharge_memlock(aux->prog); |
afdb09c7 | 1030 | security_bpf_prog_free(aux); |
abf2e7d6 AS |
1031 | bpf_prog_free(aux->prog); |
1032 | } | |
1033 | ||
b16d9aa4 | 1034 | static void __bpf_prog_put(struct bpf_prog *prog, bool do_idr_lock) |
09756af4 | 1035 | { |
a67edbf4 | 1036 | if (atomic_dec_and_test(&prog->aux->refcnt)) { |
34ad5580 | 1037 | /* bpf_prog_free_id() must be called first */ |
b16d9aa4 | 1038 | bpf_prog_free_id(prog, do_idr_lock); |
7d1982b4 | 1039 | bpf_prog_kallsyms_del_all(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" |
4316b409 DB |
1070 | "memlock:\t%llu\n" |
1071 | "prog_id:\t%u\n", | |
7bd509e3 DB |
1072 | prog->type, |
1073 | prog->jited, | |
f1f7714e | 1074 | prog_tag, |
4316b409 DB |
1075 | prog->pages * 1ULL << PAGE_SHIFT, |
1076 | prog->aux->id); | |
7bd509e3 DB |
1077 | } |
1078 | #endif | |
1079 | ||
f66e448c | 1080 | const struct file_operations bpf_prog_fops = { |
7bd509e3 DB |
1081 | #ifdef CONFIG_PROC_FS |
1082 | .show_fdinfo = bpf_prog_show_fdinfo, | |
1083 | #endif | |
1084 | .release = bpf_prog_release, | |
6e71b04a CF |
1085 | .read = bpf_dummy_read, |
1086 | .write = bpf_dummy_write, | |
09756af4 AS |
1087 | }; |
1088 | ||
b2197755 | 1089 | int bpf_prog_new_fd(struct bpf_prog *prog) |
aa79781b | 1090 | { |
afdb09c7 CF |
1091 | int ret; |
1092 | ||
1093 | ret = security_bpf_prog(prog); | |
1094 | if (ret < 0) | |
1095 | return ret; | |
1096 | ||
aa79781b DB |
1097 | return anon_inode_getfd("bpf-prog", &bpf_prog_fops, prog, |
1098 | O_RDWR | O_CLOEXEC); | |
1099 | } | |
1100 | ||
113214be | 1101 | static struct bpf_prog *____bpf_prog_get(struct fd f) |
09756af4 | 1102 | { |
09756af4 AS |
1103 | if (!f.file) |
1104 | return ERR_PTR(-EBADF); | |
09756af4 AS |
1105 | if (f.file->f_op != &bpf_prog_fops) { |
1106 | fdput(f); | |
1107 | return ERR_PTR(-EINVAL); | |
1108 | } | |
1109 | ||
c2101297 | 1110 | return f.file->private_data; |
09756af4 AS |
1111 | } |
1112 | ||
59d3656d | 1113 | struct bpf_prog *bpf_prog_add(struct bpf_prog *prog, int i) |
92117d84 | 1114 | { |
59d3656d BB |
1115 | if (atomic_add_return(i, &prog->aux->refcnt) > BPF_MAX_REFCNT) { |
1116 | atomic_sub(i, &prog->aux->refcnt); | |
92117d84 AS |
1117 | return ERR_PTR(-EBUSY); |
1118 | } | |
1119 | return prog; | |
1120 | } | |
59d3656d BB |
1121 | EXPORT_SYMBOL_GPL(bpf_prog_add); |
1122 | ||
c540594f DB |
1123 | void bpf_prog_sub(struct bpf_prog *prog, int i) |
1124 | { | |
1125 | /* Only to be used for undoing previous bpf_prog_add() in some | |
1126 | * error path. We still know that another entity in our call | |
1127 | * path holds a reference to the program, thus atomic_sub() can | |
1128 | * be safely used in such cases! | |
1129 | */ | |
1130 | WARN_ON(atomic_sub_return(i, &prog->aux->refcnt) == 0); | |
1131 | } | |
1132 | EXPORT_SYMBOL_GPL(bpf_prog_sub); | |
1133 | ||
59d3656d BB |
1134 | struct bpf_prog *bpf_prog_inc(struct bpf_prog *prog) |
1135 | { | |
1136 | return bpf_prog_add(prog, 1); | |
1137 | } | |
97bc402d | 1138 | EXPORT_SYMBOL_GPL(bpf_prog_inc); |
92117d84 | 1139 | |
b16d9aa4 | 1140 | /* prog_idr_lock should have been held */ |
a6f6df69 | 1141 | struct bpf_prog *bpf_prog_inc_not_zero(struct bpf_prog *prog) |
b16d9aa4 MKL |
1142 | { |
1143 | int refold; | |
1144 | ||
1145 | refold = __atomic_add_unless(&prog->aux->refcnt, 1, 0); | |
1146 | ||
1147 | if (refold >= BPF_MAX_REFCNT) { | |
1148 | __bpf_prog_put(prog, false); | |
1149 | return ERR_PTR(-EBUSY); | |
1150 | } | |
1151 | ||
1152 | if (!refold) | |
1153 | return ERR_PTR(-ENOENT); | |
1154 | ||
1155 | return prog; | |
1156 | } | |
a6f6df69 | 1157 | EXPORT_SYMBOL_GPL(bpf_prog_inc_not_zero); |
b16d9aa4 | 1158 | |
040ee692 | 1159 | bool bpf_prog_get_ok(struct bpf_prog *prog, |
288b3de5 | 1160 | enum bpf_prog_type *attach_type, bool attach_drv) |
248f346f | 1161 | { |
288b3de5 JK |
1162 | /* not an attachment, just a refcount inc, always allow */ |
1163 | if (!attach_type) | |
1164 | return true; | |
248f346f JK |
1165 | |
1166 | if (prog->type != *attach_type) | |
1167 | return false; | |
288b3de5 | 1168 | if (bpf_prog_is_dev_bound(prog->aux) && !attach_drv) |
248f346f JK |
1169 | return false; |
1170 | ||
1171 | return true; | |
1172 | } | |
1173 | ||
1174 | static struct bpf_prog *__bpf_prog_get(u32 ufd, enum bpf_prog_type *attach_type, | |
288b3de5 | 1175 | bool attach_drv) |
09756af4 AS |
1176 | { |
1177 | struct fd f = fdget(ufd); | |
1178 | struct bpf_prog *prog; | |
1179 | ||
113214be | 1180 | prog = ____bpf_prog_get(f); |
09756af4 AS |
1181 | if (IS_ERR(prog)) |
1182 | return prog; | |
288b3de5 | 1183 | if (!bpf_prog_get_ok(prog, attach_type, attach_drv)) { |
113214be DB |
1184 | prog = ERR_PTR(-EINVAL); |
1185 | goto out; | |
1186 | } | |
09756af4 | 1187 | |
92117d84 | 1188 | prog = bpf_prog_inc(prog); |
113214be | 1189 | out: |
09756af4 AS |
1190 | fdput(f); |
1191 | return prog; | |
1192 | } | |
113214be DB |
1193 | |
1194 | struct bpf_prog *bpf_prog_get(u32 ufd) | |
1195 | { | |
288b3de5 | 1196 | return __bpf_prog_get(ufd, NULL, false); |
113214be DB |
1197 | } |
1198 | ||
248f346f | 1199 | struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, |
288b3de5 | 1200 | bool attach_drv) |
248f346f | 1201 | { |
4d220ed0 | 1202 | return __bpf_prog_get(ufd, &type, attach_drv); |
248f346f | 1203 | } |
6c8dfe21 | 1204 | EXPORT_SYMBOL_GPL(bpf_prog_get_type_dev); |
248f346f | 1205 | |
aac3fc32 AI |
1206 | /* Initially all BPF programs could be loaded w/o specifying |
1207 | * expected_attach_type. Later for some of them specifying expected_attach_type | |
1208 | * at load time became required so that program could be validated properly. | |
1209 | * Programs of types that are allowed to be loaded both w/ and w/o (for | |
1210 | * backward compatibility) expected_attach_type, should have the default attach | |
1211 | * type assigned to expected_attach_type for the latter case, so that it can be | |
1212 | * validated later at attach time. | |
1213 | * | |
1214 | * bpf_prog_load_fixup_attach_type() sets expected_attach_type in @attr if | |
1215 | * prog type requires it but has some attach types that have to be backward | |
1216 | * compatible. | |
1217 | */ | |
1218 | static void bpf_prog_load_fixup_attach_type(union bpf_attr *attr) | |
1219 | { | |
1220 | switch (attr->prog_type) { | |
1221 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
1222 | /* Unfortunately BPF_ATTACH_TYPE_UNSPEC enumeration doesn't | |
1223 | * exist so checking for non-zero is the way to go here. | |
1224 | */ | |
1225 | if (!attr->expected_attach_type) | |
1226 | attr->expected_attach_type = | |
1227 | BPF_CGROUP_INET_SOCK_CREATE; | |
1228 | break; | |
1229 | } | |
1230 | } | |
1231 | ||
5e43f899 AI |
1232 | static int |
1233 | bpf_prog_load_check_attach_type(enum bpf_prog_type prog_type, | |
1234 | enum bpf_attach_type expected_attach_type) | |
1235 | { | |
4fbac77d | 1236 | switch (prog_type) { |
aac3fc32 AI |
1237 | case BPF_PROG_TYPE_CGROUP_SOCK: |
1238 | switch (expected_attach_type) { | |
1239 | case BPF_CGROUP_INET_SOCK_CREATE: | |
1240 | case BPF_CGROUP_INET4_POST_BIND: | |
1241 | case BPF_CGROUP_INET6_POST_BIND: | |
1242 | return 0; | |
1243 | default: | |
1244 | return -EINVAL; | |
1245 | } | |
4fbac77d AI |
1246 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: |
1247 | switch (expected_attach_type) { | |
1248 | case BPF_CGROUP_INET4_BIND: | |
1249 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1250 | case BPF_CGROUP_INET4_CONNECT: |
1251 | case BPF_CGROUP_INET6_CONNECT: | |
1cedee13 AI |
1252 | case BPF_CGROUP_UDP4_SENDMSG: |
1253 | case BPF_CGROUP_UDP6_SENDMSG: | |
4fbac77d AI |
1254 | return 0; |
1255 | default: | |
1256 | return -EINVAL; | |
1257 | } | |
1258 | default: | |
1259 | return 0; | |
1260 | } | |
5e43f899 AI |
1261 | } |
1262 | ||
09756af4 | 1263 | /* last field in 'union bpf_attr' used by this command */ |
5e43f899 | 1264 | #define BPF_PROG_LOAD_LAST_FIELD expected_attach_type |
09756af4 AS |
1265 | |
1266 | static int bpf_prog_load(union bpf_attr *attr) | |
1267 | { | |
1268 | enum bpf_prog_type type = attr->prog_type; | |
1269 | struct bpf_prog *prog; | |
1270 | int err; | |
1271 | char license[128]; | |
1272 | bool is_gpl; | |
1273 | ||
1274 | if (CHECK_ATTR(BPF_PROG_LOAD)) | |
1275 | return -EINVAL; | |
1276 | ||
e07b98d9 DM |
1277 | if (attr->prog_flags & ~BPF_F_STRICT_ALIGNMENT) |
1278 | return -EINVAL; | |
1279 | ||
09756af4 | 1280 | /* copy eBPF program license from user space */ |
535e7b4b | 1281 | if (strncpy_from_user(license, u64_to_user_ptr(attr->license), |
09756af4 AS |
1282 | sizeof(license) - 1) < 0) |
1283 | return -EFAULT; | |
1284 | license[sizeof(license) - 1] = 0; | |
1285 | ||
1286 | /* eBPF programs must be GPL compatible to use GPL-ed functions */ | |
1287 | is_gpl = license_is_gpl_compatible(license); | |
1288 | ||
ef0915ca DB |
1289 | if (attr->insn_cnt == 0 || attr->insn_cnt > BPF_MAXINSNS) |
1290 | return -E2BIG; | |
09756af4 | 1291 | |
2541517c AS |
1292 | if (type == BPF_PROG_TYPE_KPROBE && |
1293 | attr->kern_version != LINUX_VERSION_CODE) | |
1294 | return -EINVAL; | |
1295 | ||
80b7d819 CF |
1296 | if (type != BPF_PROG_TYPE_SOCKET_FILTER && |
1297 | type != BPF_PROG_TYPE_CGROUP_SKB && | |
1298 | !capable(CAP_SYS_ADMIN)) | |
1be7f75d AS |
1299 | return -EPERM; |
1300 | ||
aac3fc32 | 1301 | bpf_prog_load_fixup_attach_type(attr); |
5e43f899 AI |
1302 | if (bpf_prog_load_check_attach_type(type, attr->expected_attach_type)) |
1303 | return -EINVAL; | |
1304 | ||
09756af4 AS |
1305 | /* plain bpf_prog allocation */ |
1306 | prog = bpf_prog_alloc(bpf_prog_size(attr->insn_cnt), GFP_USER); | |
1307 | if (!prog) | |
1308 | return -ENOMEM; | |
1309 | ||
5e43f899 AI |
1310 | prog->expected_attach_type = attr->expected_attach_type; |
1311 | ||
9a18eedb JK |
1312 | prog->aux->offload_requested = !!attr->prog_ifindex; |
1313 | ||
afdb09c7 | 1314 | err = security_bpf_prog_alloc(prog->aux); |
aaac3ba9 AS |
1315 | if (err) |
1316 | goto free_prog_nouncharge; | |
1317 | ||
afdb09c7 CF |
1318 | err = bpf_prog_charge_memlock(prog); |
1319 | if (err) | |
1320 | goto free_prog_sec; | |
1321 | ||
09756af4 AS |
1322 | prog->len = attr->insn_cnt; |
1323 | ||
1324 | err = -EFAULT; | |
535e7b4b | 1325 | if (copy_from_user(prog->insns, u64_to_user_ptr(attr->insns), |
aafe6ae9 | 1326 | bpf_prog_insn_size(prog)) != 0) |
09756af4 AS |
1327 | goto free_prog; |
1328 | ||
1329 | prog->orig_prog = NULL; | |
a91263d5 | 1330 | prog->jited = 0; |
09756af4 AS |
1331 | |
1332 | atomic_set(&prog->aux->refcnt, 1); | |
a91263d5 | 1333 | prog->gpl_compatible = is_gpl ? 1 : 0; |
09756af4 | 1334 | |
9a18eedb | 1335 | if (bpf_prog_is_dev_bound(prog->aux)) { |
ab3f0063 JK |
1336 | err = bpf_prog_offload_init(prog, attr); |
1337 | if (err) | |
1338 | goto free_prog; | |
1339 | } | |
1340 | ||
09756af4 AS |
1341 | /* find program type: socket_filter vs tracing_filter */ |
1342 | err = find_prog_type(type, prog); | |
1343 | if (err < 0) | |
1344 | goto free_prog; | |
1345 | ||
cb4d2b3f MKL |
1346 | prog->aux->load_time = ktime_get_boot_ns(); |
1347 | err = bpf_obj_name_cpy(prog->aux->name, attr->prog_name); | |
1348 | if (err) | |
1349 | goto free_prog; | |
1350 | ||
09756af4 | 1351 | /* run eBPF verifier */ |
9bac3d6d | 1352 | err = bpf_check(&prog, attr); |
09756af4 AS |
1353 | if (err < 0) |
1354 | goto free_used_maps; | |
1355 | ||
9facc336 | 1356 | prog = bpf_prog_select_runtime(prog, &err); |
04fd61ab AS |
1357 | if (err < 0) |
1358 | goto free_used_maps; | |
09756af4 | 1359 | |
dc4bb0e2 MKL |
1360 | err = bpf_prog_alloc_id(prog); |
1361 | if (err) | |
1362 | goto free_used_maps; | |
1363 | ||
aa79781b | 1364 | err = bpf_prog_new_fd(prog); |
b16d9aa4 MKL |
1365 | if (err < 0) { |
1366 | /* failed to allocate fd. | |
1367 | * bpf_prog_put() is needed because the above | |
1368 | * bpf_prog_alloc_id() has published the prog | |
1369 | * to the userspace and the userspace may | |
1370 | * have refcnt-ed it through BPF_PROG_GET_FD_BY_ID. | |
1371 | */ | |
1372 | bpf_prog_put(prog); | |
1373 | return err; | |
1374 | } | |
09756af4 | 1375 | |
74451e66 | 1376 | bpf_prog_kallsyms_add(prog); |
09756af4 AS |
1377 | return err; |
1378 | ||
1379 | free_used_maps: | |
7d1982b4 | 1380 | bpf_prog_kallsyms_del_subprogs(prog); |
09756af4 AS |
1381 | free_used_maps(prog->aux); |
1382 | free_prog: | |
aaac3ba9 | 1383 | bpf_prog_uncharge_memlock(prog); |
afdb09c7 CF |
1384 | free_prog_sec: |
1385 | security_bpf_prog_free(prog->aux); | |
aaac3ba9 | 1386 | free_prog_nouncharge: |
09756af4 AS |
1387 | bpf_prog_free(prog); |
1388 | return err; | |
1389 | } | |
1390 | ||
6e71b04a | 1391 | #define BPF_OBJ_LAST_FIELD file_flags |
b2197755 DB |
1392 | |
1393 | static int bpf_obj_pin(const union bpf_attr *attr) | |
1394 | { | |
6e71b04a | 1395 | if (CHECK_ATTR(BPF_OBJ) || attr->file_flags != 0) |
b2197755 DB |
1396 | return -EINVAL; |
1397 | ||
535e7b4b | 1398 | return bpf_obj_pin_user(attr->bpf_fd, u64_to_user_ptr(attr->pathname)); |
b2197755 DB |
1399 | } |
1400 | ||
1401 | static int bpf_obj_get(const union bpf_attr *attr) | |
1402 | { | |
6e71b04a CF |
1403 | if (CHECK_ATTR(BPF_OBJ) || attr->bpf_fd != 0 || |
1404 | attr->file_flags & ~BPF_OBJ_FLAG_MASK) | |
b2197755 DB |
1405 | return -EINVAL; |
1406 | ||
6e71b04a CF |
1407 | return bpf_obj_get_user(u64_to_user_ptr(attr->pathname), |
1408 | attr->file_flags); | |
b2197755 DB |
1409 | } |
1410 | ||
c4f6699d AS |
1411 | struct bpf_raw_tracepoint { |
1412 | struct bpf_raw_event_map *btp; | |
1413 | struct bpf_prog *prog; | |
1414 | }; | |
1415 | ||
1416 | static int bpf_raw_tracepoint_release(struct inode *inode, struct file *filp) | |
1417 | { | |
1418 | struct bpf_raw_tracepoint *raw_tp = filp->private_data; | |
1419 | ||
1420 | if (raw_tp->prog) { | |
1421 | bpf_probe_unregister(raw_tp->btp, raw_tp->prog); | |
1422 | bpf_prog_put(raw_tp->prog); | |
1423 | } | |
1424 | kfree(raw_tp); | |
1425 | return 0; | |
1426 | } | |
1427 | ||
1428 | static const struct file_operations bpf_raw_tp_fops = { | |
1429 | .release = bpf_raw_tracepoint_release, | |
1430 | .read = bpf_dummy_read, | |
1431 | .write = bpf_dummy_write, | |
1432 | }; | |
1433 | ||
1434 | #define BPF_RAW_TRACEPOINT_OPEN_LAST_FIELD raw_tracepoint.prog_fd | |
1435 | ||
1436 | static int bpf_raw_tracepoint_open(const union bpf_attr *attr) | |
1437 | { | |
1438 | struct bpf_raw_tracepoint *raw_tp; | |
1439 | struct bpf_raw_event_map *btp; | |
1440 | struct bpf_prog *prog; | |
1441 | char tp_name[128]; | |
1442 | int tp_fd, err; | |
1443 | ||
1444 | if (strncpy_from_user(tp_name, u64_to_user_ptr(attr->raw_tracepoint.name), | |
1445 | sizeof(tp_name) - 1) < 0) | |
1446 | return -EFAULT; | |
1447 | tp_name[sizeof(tp_name) - 1] = 0; | |
1448 | ||
1449 | btp = bpf_find_raw_tracepoint(tp_name); | |
1450 | if (!btp) | |
1451 | return -ENOENT; | |
1452 | ||
1453 | raw_tp = kzalloc(sizeof(*raw_tp), GFP_USER); | |
1454 | if (!raw_tp) | |
1455 | return -ENOMEM; | |
1456 | raw_tp->btp = btp; | |
1457 | ||
1458 | prog = bpf_prog_get_type(attr->raw_tracepoint.prog_fd, | |
1459 | BPF_PROG_TYPE_RAW_TRACEPOINT); | |
1460 | if (IS_ERR(prog)) { | |
1461 | err = PTR_ERR(prog); | |
1462 | goto out_free_tp; | |
1463 | } | |
1464 | ||
1465 | err = bpf_probe_register(raw_tp->btp, prog); | |
1466 | if (err) | |
1467 | goto out_put_prog; | |
1468 | ||
1469 | raw_tp->prog = prog; | |
1470 | tp_fd = anon_inode_getfd("bpf-raw-tracepoint", &bpf_raw_tp_fops, raw_tp, | |
1471 | O_CLOEXEC); | |
1472 | if (tp_fd < 0) { | |
1473 | bpf_probe_unregister(raw_tp->btp, prog); | |
1474 | err = tp_fd; | |
1475 | goto out_put_prog; | |
1476 | } | |
1477 | return tp_fd; | |
1478 | ||
1479 | out_put_prog: | |
1480 | bpf_prog_put(prog); | |
1481 | out_free_tp: | |
1482 | kfree(raw_tp); | |
1483 | return err; | |
1484 | } | |
1485 | ||
f4324551 DM |
1486 | #ifdef CONFIG_CGROUP_BPF |
1487 | ||
33491588 AR |
1488 | static int bpf_prog_attach_check_attach_type(const struct bpf_prog *prog, |
1489 | enum bpf_attach_type attach_type) | |
1490 | { | |
1491 | switch (prog->type) { | |
1492 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
1493 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: | |
1494 | return attach_type == prog->expected_attach_type ? 0 : -EINVAL; | |
1495 | default: | |
1496 | return 0; | |
1497 | } | |
1498 | } | |
1499 | ||
464bc0fd | 1500 | #define BPF_PROG_ATTACH_LAST_FIELD attach_flags |
174a79ff | 1501 | |
4f738adb JF |
1502 | static int sockmap_get_from_fd(const union bpf_attr *attr, |
1503 | int type, bool attach) | |
174a79ff | 1504 | { |
5a67da2a | 1505 | struct bpf_prog *prog = NULL; |
174a79ff JF |
1506 | int ufd = attr->target_fd; |
1507 | struct bpf_map *map; | |
1508 | struct fd f; | |
1509 | int err; | |
1510 | ||
1511 | f = fdget(ufd); | |
1512 | map = __bpf_map_get(f); | |
1513 | if (IS_ERR(map)) | |
1514 | return PTR_ERR(map); | |
1515 | ||
5a67da2a | 1516 | if (attach) { |
4f738adb | 1517 | prog = bpf_prog_get_type(attr->attach_bpf_fd, type); |
5a67da2a JF |
1518 | if (IS_ERR(prog)) { |
1519 | fdput(f); | |
1520 | return PTR_ERR(prog); | |
1521 | } | |
174a79ff JF |
1522 | } |
1523 | ||
5a67da2a | 1524 | err = sock_map_prog(map, prog, attr->attach_type); |
174a79ff JF |
1525 | if (err) { |
1526 | fdput(f); | |
5a67da2a JF |
1527 | if (prog) |
1528 | bpf_prog_put(prog); | |
ae2b27b8 | 1529 | return err; |
174a79ff JF |
1530 | } |
1531 | ||
1532 | fdput(f); | |
ae2b27b8 | 1533 | return 0; |
174a79ff | 1534 | } |
f4324551 | 1535 | |
324bda9e AS |
1536 | #define BPF_F_ATTACH_MASK \ |
1537 | (BPF_F_ALLOW_OVERRIDE | BPF_F_ALLOW_MULTI) | |
1538 | ||
f4324551 DM |
1539 | static int bpf_prog_attach(const union bpf_attr *attr) |
1540 | { | |
7f677633 | 1541 | enum bpf_prog_type ptype; |
f4324551 DM |
1542 | struct bpf_prog *prog; |
1543 | struct cgroup *cgrp; | |
7f677633 | 1544 | int ret; |
f4324551 DM |
1545 | |
1546 | if (!capable(CAP_NET_ADMIN)) | |
1547 | return -EPERM; | |
1548 | ||
1549 | if (CHECK_ATTR(BPF_PROG_ATTACH)) | |
1550 | return -EINVAL; | |
1551 | ||
324bda9e | 1552 | if (attr->attach_flags & ~BPF_F_ATTACH_MASK) |
7f677633 AS |
1553 | return -EINVAL; |
1554 | ||
f4324551 DM |
1555 | switch (attr->attach_type) { |
1556 | case BPF_CGROUP_INET_INGRESS: | |
1557 | case BPF_CGROUP_INET_EGRESS: | |
b2cd1257 | 1558 | ptype = BPF_PROG_TYPE_CGROUP_SKB; |
f4324551 | 1559 | break; |
61023658 | 1560 | case BPF_CGROUP_INET_SOCK_CREATE: |
aac3fc32 AI |
1561 | case BPF_CGROUP_INET4_POST_BIND: |
1562 | case BPF_CGROUP_INET6_POST_BIND: | |
61023658 DA |
1563 | ptype = BPF_PROG_TYPE_CGROUP_SOCK; |
1564 | break; | |
4fbac77d AI |
1565 | case BPF_CGROUP_INET4_BIND: |
1566 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1567 | case BPF_CGROUP_INET4_CONNECT: |
1568 | case BPF_CGROUP_INET6_CONNECT: | |
1cedee13 AI |
1569 | case BPF_CGROUP_UDP4_SENDMSG: |
1570 | case BPF_CGROUP_UDP6_SENDMSG: | |
4fbac77d AI |
1571 | ptype = BPF_PROG_TYPE_CGROUP_SOCK_ADDR; |
1572 | break; | |
40304b2a LB |
1573 | case BPF_CGROUP_SOCK_OPS: |
1574 | ptype = BPF_PROG_TYPE_SOCK_OPS; | |
1575 | break; | |
ebc614f6 RG |
1576 | case BPF_CGROUP_DEVICE: |
1577 | ptype = BPF_PROG_TYPE_CGROUP_DEVICE; | |
1578 | break; | |
4f738adb JF |
1579 | case BPF_SK_MSG_VERDICT: |
1580 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_MSG, true); | |
464bc0fd JF |
1581 | case BPF_SK_SKB_STREAM_PARSER: |
1582 | case BPF_SK_SKB_STREAM_VERDICT: | |
4f738adb | 1583 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_SKB, true); |
f4364dcf SY |
1584 | case BPF_LIRC_MODE2: |
1585 | return lirc_prog_attach(attr); | |
f4324551 DM |
1586 | default: |
1587 | return -EINVAL; | |
1588 | } | |
1589 | ||
b2cd1257 DA |
1590 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); |
1591 | if (IS_ERR(prog)) | |
1592 | return PTR_ERR(prog); | |
1593 | ||
5e43f899 AI |
1594 | if (bpf_prog_attach_check_attach_type(prog, attr->attach_type)) { |
1595 | bpf_prog_put(prog); | |
1596 | return -EINVAL; | |
1597 | } | |
1598 | ||
b2cd1257 DA |
1599 | cgrp = cgroup_get_from_fd(attr->target_fd); |
1600 | if (IS_ERR(cgrp)) { | |
1601 | bpf_prog_put(prog); | |
1602 | return PTR_ERR(cgrp); | |
1603 | } | |
1604 | ||
324bda9e AS |
1605 | ret = cgroup_bpf_attach(cgrp, prog, attr->attach_type, |
1606 | attr->attach_flags); | |
7f677633 AS |
1607 | if (ret) |
1608 | bpf_prog_put(prog); | |
b2cd1257 DA |
1609 | cgroup_put(cgrp); |
1610 | ||
7f677633 | 1611 | return ret; |
f4324551 DM |
1612 | } |
1613 | ||
1614 | #define BPF_PROG_DETACH_LAST_FIELD attach_type | |
1615 | ||
1616 | static int bpf_prog_detach(const union bpf_attr *attr) | |
1617 | { | |
324bda9e AS |
1618 | enum bpf_prog_type ptype; |
1619 | struct bpf_prog *prog; | |
f4324551 | 1620 | struct cgroup *cgrp; |
7f677633 | 1621 | int ret; |
f4324551 DM |
1622 | |
1623 | if (!capable(CAP_NET_ADMIN)) | |
1624 | return -EPERM; | |
1625 | ||
1626 | if (CHECK_ATTR(BPF_PROG_DETACH)) | |
1627 | return -EINVAL; | |
1628 | ||
1629 | switch (attr->attach_type) { | |
1630 | case BPF_CGROUP_INET_INGRESS: | |
1631 | case BPF_CGROUP_INET_EGRESS: | |
324bda9e AS |
1632 | ptype = BPF_PROG_TYPE_CGROUP_SKB; |
1633 | break; | |
61023658 | 1634 | case BPF_CGROUP_INET_SOCK_CREATE: |
aac3fc32 AI |
1635 | case BPF_CGROUP_INET4_POST_BIND: |
1636 | case BPF_CGROUP_INET6_POST_BIND: | |
324bda9e AS |
1637 | ptype = BPF_PROG_TYPE_CGROUP_SOCK; |
1638 | break; | |
4fbac77d AI |
1639 | case BPF_CGROUP_INET4_BIND: |
1640 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1641 | case BPF_CGROUP_INET4_CONNECT: |
1642 | case BPF_CGROUP_INET6_CONNECT: | |
1cedee13 AI |
1643 | case BPF_CGROUP_UDP4_SENDMSG: |
1644 | case BPF_CGROUP_UDP6_SENDMSG: | |
4fbac77d AI |
1645 | ptype = BPF_PROG_TYPE_CGROUP_SOCK_ADDR; |
1646 | break; | |
40304b2a | 1647 | case BPF_CGROUP_SOCK_OPS: |
324bda9e | 1648 | ptype = BPF_PROG_TYPE_SOCK_OPS; |
f4324551 | 1649 | break; |
ebc614f6 RG |
1650 | case BPF_CGROUP_DEVICE: |
1651 | ptype = BPF_PROG_TYPE_CGROUP_DEVICE; | |
1652 | break; | |
4f738adb JF |
1653 | case BPF_SK_MSG_VERDICT: |
1654 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_MSG, false); | |
5a67da2a JF |
1655 | case BPF_SK_SKB_STREAM_PARSER: |
1656 | case BPF_SK_SKB_STREAM_VERDICT: | |
4f738adb | 1657 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_SKB, false); |
f4364dcf SY |
1658 | case BPF_LIRC_MODE2: |
1659 | return lirc_prog_detach(attr); | |
f4324551 DM |
1660 | default: |
1661 | return -EINVAL; | |
1662 | } | |
1663 | ||
324bda9e AS |
1664 | cgrp = cgroup_get_from_fd(attr->target_fd); |
1665 | if (IS_ERR(cgrp)) | |
1666 | return PTR_ERR(cgrp); | |
1667 | ||
1668 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); | |
1669 | if (IS_ERR(prog)) | |
1670 | prog = NULL; | |
1671 | ||
1672 | ret = cgroup_bpf_detach(cgrp, prog, attr->attach_type, 0); | |
1673 | if (prog) | |
1674 | bpf_prog_put(prog); | |
1675 | cgroup_put(cgrp); | |
7f677633 | 1676 | return ret; |
f4324551 | 1677 | } |
40304b2a | 1678 | |
468e2f64 AS |
1679 | #define BPF_PROG_QUERY_LAST_FIELD query.prog_cnt |
1680 | ||
1681 | static int bpf_prog_query(const union bpf_attr *attr, | |
1682 | union bpf_attr __user *uattr) | |
1683 | { | |
1684 | struct cgroup *cgrp; | |
1685 | int ret; | |
1686 | ||
1687 | if (!capable(CAP_NET_ADMIN)) | |
1688 | return -EPERM; | |
1689 | if (CHECK_ATTR(BPF_PROG_QUERY)) | |
1690 | return -EINVAL; | |
1691 | if (attr->query.query_flags & ~BPF_F_QUERY_EFFECTIVE) | |
1692 | return -EINVAL; | |
1693 | ||
1694 | switch (attr->query.attach_type) { | |
1695 | case BPF_CGROUP_INET_INGRESS: | |
1696 | case BPF_CGROUP_INET_EGRESS: | |
1697 | case BPF_CGROUP_INET_SOCK_CREATE: | |
4fbac77d AI |
1698 | case BPF_CGROUP_INET4_BIND: |
1699 | case BPF_CGROUP_INET6_BIND: | |
aac3fc32 AI |
1700 | case BPF_CGROUP_INET4_POST_BIND: |
1701 | case BPF_CGROUP_INET6_POST_BIND: | |
d74bad4e AI |
1702 | case BPF_CGROUP_INET4_CONNECT: |
1703 | case BPF_CGROUP_INET6_CONNECT: | |
1cedee13 AI |
1704 | case BPF_CGROUP_UDP4_SENDMSG: |
1705 | case BPF_CGROUP_UDP6_SENDMSG: | |
468e2f64 | 1706 | case BPF_CGROUP_SOCK_OPS: |
ebc614f6 | 1707 | case BPF_CGROUP_DEVICE: |
468e2f64 | 1708 | break; |
f4364dcf SY |
1709 | case BPF_LIRC_MODE2: |
1710 | return lirc_prog_query(attr, uattr); | |
468e2f64 AS |
1711 | default: |
1712 | return -EINVAL; | |
1713 | } | |
1714 | cgrp = cgroup_get_from_fd(attr->query.target_fd); | |
1715 | if (IS_ERR(cgrp)) | |
1716 | return PTR_ERR(cgrp); | |
1717 | ret = cgroup_bpf_query(cgrp, attr, uattr); | |
1718 | cgroup_put(cgrp); | |
1719 | return ret; | |
1720 | } | |
f4324551 DM |
1721 | #endif /* CONFIG_CGROUP_BPF */ |
1722 | ||
1cf1cae9 AS |
1723 | #define BPF_PROG_TEST_RUN_LAST_FIELD test.duration |
1724 | ||
1725 | static int bpf_prog_test_run(const union bpf_attr *attr, | |
1726 | union bpf_attr __user *uattr) | |
1727 | { | |
1728 | struct bpf_prog *prog; | |
1729 | int ret = -ENOTSUPP; | |
1730 | ||
61f3c964 AS |
1731 | if (!capable(CAP_SYS_ADMIN)) |
1732 | return -EPERM; | |
1cf1cae9 AS |
1733 | if (CHECK_ATTR(BPF_PROG_TEST_RUN)) |
1734 | return -EINVAL; | |
1735 | ||
1736 | prog = bpf_prog_get(attr->test.prog_fd); | |
1737 | if (IS_ERR(prog)) | |
1738 | return PTR_ERR(prog); | |
1739 | ||
1740 | if (prog->aux->ops->test_run) | |
1741 | ret = prog->aux->ops->test_run(prog, attr, uattr); | |
1742 | ||
1743 | bpf_prog_put(prog); | |
1744 | return ret; | |
1745 | } | |
1746 | ||
34ad5580 MKL |
1747 | #define BPF_OBJ_GET_NEXT_ID_LAST_FIELD next_id |
1748 | ||
1749 | static int bpf_obj_get_next_id(const union bpf_attr *attr, | |
1750 | union bpf_attr __user *uattr, | |
1751 | struct idr *idr, | |
1752 | spinlock_t *lock) | |
1753 | { | |
1754 | u32 next_id = attr->start_id; | |
1755 | int err = 0; | |
1756 | ||
1757 | if (CHECK_ATTR(BPF_OBJ_GET_NEXT_ID) || next_id >= INT_MAX) | |
1758 | return -EINVAL; | |
1759 | ||
1760 | if (!capable(CAP_SYS_ADMIN)) | |
1761 | return -EPERM; | |
1762 | ||
1763 | next_id++; | |
1764 | spin_lock_bh(lock); | |
1765 | if (!idr_get_next(idr, &next_id)) | |
1766 | err = -ENOENT; | |
1767 | spin_unlock_bh(lock); | |
1768 | ||
1769 | if (!err) | |
1770 | err = put_user(next_id, &uattr->next_id); | |
1771 | ||
1772 | return err; | |
1773 | } | |
1774 | ||
b16d9aa4 MKL |
1775 | #define BPF_PROG_GET_FD_BY_ID_LAST_FIELD prog_id |
1776 | ||
1777 | static int bpf_prog_get_fd_by_id(const union bpf_attr *attr) | |
1778 | { | |
1779 | struct bpf_prog *prog; | |
1780 | u32 id = attr->prog_id; | |
1781 | int fd; | |
1782 | ||
1783 | if (CHECK_ATTR(BPF_PROG_GET_FD_BY_ID)) | |
1784 | return -EINVAL; | |
1785 | ||
1786 | if (!capable(CAP_SYS_ADMIN)) | |
1787 | return -EPERM; | |
1788 | ||
1789 | spin_lock_bh(&prog_idr_lock); | |
1790 | prog = idr_find(&prog_idr, id); | |
1791 | if (prog) | |
1792 | prog = bpf_prog_inc_not_zero(prog); | |
1793 | else | |
1794 | prog = ERR_PTR(-ENOENT); | |
1795 | spin_unlock_bh(&prog_idr_lock); | |
1796 | ||
1797 | if (IS_ERR(prog)) | |
1798 | return PTR_ERR(prog); | |
1799 | ||
1800 | fd = bpf_prog_new_fd(prog); | |
1801 | if (fd < 0) | |
1802 | bpf_prog_put(prog); | |
1803 | ||
1804 | return fd; | |
1805 | } | |
1806 | ||
6e71b04a | 1807 | #define BPF_MAP_GET_FD_BY_ID_LAST_FIELD open_flags |
bd5f5f4e MKL |
1808 | |
1809 | static int bpf_map_get_fd_by_id(const union bpf_attr *attr) | |
1810 | { | |
1811 | struct bpf_map *map; | |
1812 | u32 id = attr->map_id; | |
6e71b04a | 1813 | int f_flags; |
bd5f5f4e MKL |
1814 | int fd; |
1815 | ||
6e71b04a CF |
1816 | if (CHECK_ATTR(BPF_MAP_GET_FD_BY_ID) || |
1817 | attr->open_flags & ~BPF_OBJ_FLAG_MASK) | |
bd5f5f4e MKL |
1818 | return -EINVAL; |
1819 | ||
1820 | if (!capable(CAP_SYS_ADMIN)) | |
1821 | return -EPERM; | |
1822 | ||
6e71b04a CF |
1823 | f_flags = bpf_get_file_flag(attr->open_flags); |
1824 | if (f_flags < 0) | |
1825 | return f_flags; | |
1826 | ||
bd5f5f4e MKL |
1827 | spin_lock_bh(&map_idr_lock); |
1828 | map = idr_find(&map_idr, id); | |
1829 | if (map) | |
1830 | map = bpf_map_inc_not_zero(map, true); | |
1831 | else | |
1832 | map = ERR_PTR(-ENOENT); | |
1833 | spin_unlock_bh(&map_idr_lock); | |
1834 | ||
1835 | if (IS_ERR(map)) | |
1836 | return PTR_ERR(map); | |
1837 | ||
6e71b04a | 1838 | fd = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
1839 | if (fd < 0) |
1840 | bpf_map_put(map); | |
1841 | ||
1842 | return fd; | |
1843 | } | |
1844 | ||
7105e828 DB |
1845 | static const struct bpf_map *bpf_map_from_imm(const struct bpf_prog *prog, |
1846 | unsigned long addr) | |
1847 | { | |
1848 | int i; | |
1849 | ||
1850 | for (i = 0; i < prog->aux->used_map_cnt; i++) | |
1851 | if (prog->aux->used_maps[i] == (void *)addr) | |
1852 | return prog->aux->used_maps[i]; | |
1853 | return NULL; | |
1854 | } | |
1855 | ||
1856 | static struct bpf_insn *bpf_insn_prepare_dump(const struct bpf_prog *prog) | |
1857 | { | |
1858 | const struct bpf_map *map; | |
1859 | struct bpf_insn *insns; | |
1860 | u64 imm; | |
1861 | int i; | |
1862 | ||
1863 | insns = kmemdup(prog->insnsi, bpf_prog_insn_size(prog), | |
1864 | GFP_USER); | |
1865 | if (!insns) | |
1866 | return insns; | |
1867 | ||
1868 | for (i = 0; i < prog->len; i++) { | |
1869 | if (insns[i].code == (BPF_JMP | BPF_TAIL_CALL)) { | |
1870 | insns[i].code = BPF_JMP | BPF_CALL; | |
1871 | insns[i].imm = BPF_FUNC_tail_call; | |
1872 | /* fall-through */ | |
1873 | } | |
1874 | if (insns[i].code == (BPF_JMP | BPF_CALL) || | |
1875 | insns[i].code == (BPF_JMP | BPF_CALL_ARGS)) { | |
1876 | if (insns[i].code == (BPF_JMP | BPF_CALL_ARGS)) | |
1877 | insns[i].code = BPF_JMP | BPF_CALL; | |
1878 | if (!bpf_dump_raw_ok()) | |
1879 | insns[i].imm = 0; | |
1880 | continue; | |
1881 | } | |
1882 | ||
1883 | if (insns[i].code != (BPF_LD | BPF_IMM | BPF_DW)) | |
1884 | continue; | |
1885 | ||
1886 | imm = ((u64)insns[i + 1].imm << 32) | (u32)insns[i].imm; | |
1887 | map = bpf_map_from_imm(prog, imm); | |
1888 | if (map) { | |
1889 | insns[i].src_reg = BPF_PSEUDO_MAP_FD; | |
1890 | insns[i].imm = map->id; | |
1891 | insns[i + 1].imm = 0; | |
1892 | continue; | |
1893 | } | |
1894 | ||
1895 | if (!bpf_dump_raw_ok() && | |
1896 | imm == (unsigned long)prog->aux) { | |
1897 | insns[i].imm = 0; | |
1898 | insns[i + 1].imm = 0; | |
1899 | continue; | |
1900 | } | |
1901 | } | |
1902 | ||
1903 | return insns; | |
1904 | } | |
1905 | ||
1e270976 MKL |
1906 | static int bpf_prog_get_info_by_fd(struct bpf_prog *prog, |
1907 | const union bpf_attr *attr, | |
1908 | union bpf_attr __user *uattr) | |
1909 | { | |
1910 | struct bpf_prog_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
1911 | struct bpf_prog_info info = {}; | |
1912 | u32 info_len = attr->info.info_len; | |
1913 | char __user *uinsns; | |
1914 | u32 ulen; | |
1915 | int err; | |
1916 | ||
dcab51f1 | 1917 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(info), info_len); |
1e270976 MKL |
1918 | if (err) |
1919 | return err; | |
1920 | info_len = min_t(u32, sizeof(info), info_len); | |
1921 | ||
1922 | if (copy_from_user(&info, uinfo, info_len)) | |
89b09689 | 1923 | return -EFAULT; |
1e270976 MKL |
1924 | |
1925 | info.type = prog->type; | |
1926 | info.id = prog->aux->id; | |
cb4d2b3f MKL |
1927 | info.load_time = prog->aux->load_time; |
1928 | info.created_by_uid = from_kuid_munged(current_user_ns(), | |
1929 | prog->aux->user->uid); | |
b85fab0e | 1930 | info.gpl_compatible = prog->gpl_compatible; |
1e270976 MKL |
1931 | |
1932 | memcpy(info.tag, prog->tag, sizeof(prog->tag)); | |
cb4d2b3f MKL |
1933 | memcpy(info.name, prog->aux->name, sizeof(prog->aux->name)); |
1934 | ||
1935 | ulen = info.nr_map_ids; | |
1936 | info.nr_map_ids = prog->aux->used_map_cnt; | |
1937 | ulen = min_t(u32, info.nr_map_ids, ulen); | |
1938 | if (ulen) { | |
721e08da | 1939 | u32 __user *user_map_ids = u64_to_user_ptr(info.map_ids); |
cb4d2b3f MKL |
1940 | u32 i; |
1941 | ||
1942 | for (i = 0; i < ulen; i++) | |
1943 | if (put_user(prog->aux->used_maps[i]->id, | |
1944 | &user_map_ids[i])) | |
1945 | return -EFAULT; | |
1946 | } | |
1e270976 MKL |
1947 | |
1948 | if (!capable(CAP_SYS_ADMIN)) { | |
1949 | info.jited_prog_len = 0; | |
1950 | info.xlated_prog_len = 0; | |
dbecd738 | 1951 | info.nr_jited_ksyms = 0; |
1e270976 MKL |
1952 | goto done; |
1953 | } | |
1954 | ||
1e270976 | 1955 | ulen = info.xlated_prog_len; |
9975a54b | 1956 | info.xlated_prog_len = bpf_prog_insn_size(prog); |
1e270976 | 1957 | if (info.xlated_prog_len && ulen) { |
7105e828 DB |
1958 | struct bpf_insn *insns_sanitized; |
1959 | bool fault; | |
1960 | ||
1961 | if (prog->blinded && !bpf_dump_raw_ok()) { | |
1962 | info.xlated_prog_insns = 0; | |
1963 | goto done; | |
1964 | } | |
1965 | insns_sanitized = bpf_insn_prepare_dump(prog); | |
1966 | if (!insns_sanitized) | |
1967 | return -ENOMEM; | |
1e270976 MKL |
1968 | uinsns = u64_to_user_ptr(info.xlated_prog_insns); |
1969 | ulen = min_t(u32, info.xlated_prog_len, ulen); | |
7105e828 DB |
1970 | fault = copy_to_user(uinsns, insns_sanitized, ulen); |
1971 | kfree(insns_sanitized); | |
1972 | if (fault) | |
1e270976 MKL |
1973 | return -EFAULT; |
1974 | } | |
1975 | ||
675fc275 JK |
1976 | if (bpf_prog_is_dev_bound(prog->aux)) { |
1977 | err = bpf_prog_offload_info_fill(&info, prog); | |
1978 | if (err) | |
1979 | return err; | |
fcfb126d JW |
1980 | goto done; |
1981 | } | |
1982 | ||
1983 | /* NOTE: the following code is supposed to be skipped for offload. | |
1984 | * bpf_prog_offload_info_fill() is the place to fill similar fields | |
1985 | * for offload. | |
1986 | */ | |
1987 | ulen = info.jited_prog_len; | |
4d56a76e SD |
1988 | if (prog->aux->func_cnt) { |
1989 | u32 i; | |
1990 | ||
1991 | info.jited_prog_len = 0; | |
1992 | for (i = 0; i < prog->aux->func_cnt; i++) | |
1993 | info.jited_prog_len += prog->aux->func[i]->jited_len; | |
1994 | } else { | |
1995 | info.jited_prog_len = prog->jited_len; | |
1996 | } | |
1997 | ||
fcfb126d JW |
1998 | if (info.jited_prog_len && ulen) { |
1999 | if (bpf_dump_raw_ok()) { | |
2000 | uinsns = u64_to_user_ptr(info.jited_prog_insns); | |
2001 | ulen = min_t(u32, info.jited_prog_len, ulen); | |
4d56a76e SD |
2002 | |
2003 | /* for multi-function programs, copy the JITed | |
2004 | * instructions for all the functions | |
2005 | */ | |
2006 | if (prog->aux->func_cnt) { | |
2007 | u32 len, free, i; | |
2008 | u8 *img; | |
2009 | ||
2010 | free = ulen; | |
2011 | for (i = 0; i < prog->aux->func_cnt; i++) { | |
2012 | len = prog->aux->func[i]->jited_len; | |
2013 | len = min_t(u32, len, free); | |
2014 | img = (u8 *) prog->aux->func[i]->bpf_func; | |
2015 | if (copy_to_user(uinsns, img, len)) | |
2016 | return -EFAULT; | |
2017 | uinsns += len; | |
2018 | free -= len; | |
2019 | if (!free) | |
2020 | break; | |
2021 | } | |
2022 | } else { | |
2023 | if (copy_to_user(uinsns, prog->bpf_func, ulen)) | |
2024 | return -EFAULT; | |
2025 | } | |
fcfb126d JW |
2026 | } else { |
2027 | info.jited_prog_insns = 0; | |
2028 | } | |
675fc275 JK |
2029 | } |
2030 | ||
dbecd738 SD |
2031 | ulen = info.nr_jited_ksyms; |
2032 | info.nr_jited_ksyms = prog->aux->func_cnt; | |
2033 | if (info.nr_jited_ksyms && ulen) { | |
2034 | if (bpf_dump_raw_ok()) { | |
2035 | u64 __user *user_ksyms; | |
2036 | ulong ksym_addr; | |
2037 | u32 i; | |
2038 | ||
2039 | /* copy the address of the kernel symbol | |
2040 | * corresponding to each function | |
2041 | */ | |
2042 | ulen = min_t(u32, info.nr_jited_ksyms, ulen); | |
2043 | user_ksyms = u64_to_user_ptr(info.jited_ksyms); | |
2044 | for (i = 0; i < ulen; i++) { | |
2045 | ksym_addr = (ulong) prog->aux->func[i]->bpf_func; | |
2046 | ksym_addr &= PAGE_MASK; | |
2047 | if (put_user((u64) ksym_addr, &user_ksyms[i])) | |
2048 | return -EFAULT; | |
2049 | } | |
2050 | } else { | |
2051 | info.jited_ksyms = 0; | |
2052 | } | |
2053 | } | |
2054 | ||
815581c1 SD |
2055 | ulen = info.nr_jited_func_lens; |
2056 | info.nr_jited_func_lens = prog->aux->func_cnt; | |
2057 | if (info.nr_jited_func_lens && ulen) { | |
2058 | if (bpf_dump_raw_ok()) { | |
2059 | u32 __user *user_lens; | |
2060 | u32 func_len, i; | |
2061 | ||
2062 | /* copy the JITed image lengths for each function */ | |
2063 | ulen = min_t(u32, info.nr_jited_func_lens, ulen); | |
2064 | user_lens = u64_to_user_ptr(info.jited_func_lens); | |
2065 | for (i = 0; i < ulen; i++) { | |
2066 | func_len = prog->aux->func[i]->jited_len; | |
2067 | if (put_user(func_len, &user_lens[i])) | |
2068 | return -EFAULT; | |
2069 | } | |
2070 | } else { | |
2071 | info.jited_func_lens = 0; | |
2072 | } | |
2073 | } | |
2074 | ||
1e270976 MKL |
2075 | done: |
2076 | if (copy_to_user(uinfo, &info, info_len) || | |
2077 | put_user(info_len, &uattr->info.info_len)) | |
2078 | return -EFAULT; | |
2079 | ||
2080 | return 0; | |
2081 | } | |
2082 | ||
2083 | static int bpf_map_get_info_by_fd(struct bpf_map *map, | |
2084 | const union bpf_attr *attr, | |
2085 | union bpf_attr __user *uattr) | |
2086 | { | |
2087 | struct bpf_map_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
2088 | struct bpf_map_info info = {}; | |
2089 | u32 info_len = attr->info.info_len; | |
2090 | int err; | |
2091 | ||
dcab51f1 | 2092 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(info), info_len); |
1e270976 MKL |
2093 | if (err) |
2094 | return err; | |
2095 | info_len = min_t(u32, sizeof(info), info_len); | |
2096 | ||
2097 | info.type = map->map_type; | |
2098 | info.id = map->id; | |
2099 | info.key_size = map->key_size; | |
2100 | info.value_size = map->value_size; | |
2101 | info.max_entries = map->max_entries; | |
2102 | info.map_flags = map->map_flags; | |
ad5b177b | 2103 | memcpy(info.name, map->name, sizeof(map->name)); |
1e270976 | 2104 | |
78958fca MKL |
2105 | if (map->btf) { |
2106 | info.btf_id = btf_id(map->btf); | |
9b2cf328 MKL |
2107 | info.btf_key_type_id = map->btf_key_type_id; |
2108 | info.btf_value_type_id = map->btf_value_type_id; | |
78958fca MKL |
2109 | } |
2110 | ||
52775b33 JK |
2111 | if (bpf_map_is_dev_bound(map)) { |
2112 | err = bpf_map_offload_info_fill(&info, map); | |
2113 | if (err) | |
2114 | return err; | |
2115 | } | |
2116 | ||
1e270976 MKL |
2117 | if (copy_to_user(uinfo, &info, info_len) || |
2118 | put_user(info_len, &uattr->info.info_len)) | |
2119 | return -EFAULT; | |
2120 | ||
2121 | return 0; | |
2122 | } | |
2123 | ||
62dab84c MKL |
2124 | static int bpf_btf_get_info_by_fd(struct btf *btf, |
2125 | const union bpf_attr *attr, | |
2126 | union bpf_attr __user *uattr) | |
2127 | { | |
2128 | struct bpf_btf_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
2129 | u32 info_len = attr->info.info_len; | |
2130 | int err; | |
2131 | ||
dcab51f1 | 2132 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(*uinfo), info_len); |
62dab84c MKL |
2133 | if (err) |
2134 | return err; | |
2135 | ||
2136 | return btf_get_info_by_fd(btf, attr, uattr); | |
2137 | } | |
2138 | ||
1e270976 MKL |
2139 | #define BPF_OBJ_GET_INFO_BY_FD_LAST_FIELD info.info |
2140 | ||
2141 | static int bpf_obj_get_info_by_fd(const union bpf_attr *attr, | |
2142 | union bpf_attr __user *uattr) | |
2143 | { | |
2144 | int ufd = attr->info.bpf_fd; | |
2145 | struct fd f; | |
2146 | int err; | |
2147 | ||
2148 | if (CHECK_ATTR(BPF_OBJ_GET_INFO_BY_FD)) | |
2149 | return -EINVAL; | |
2150 | ||
2151 | f = fdget(ufd); | |
2152 | if (!f.file) | |
2153 | return -EBADFD; | |
2154 | ||
2155 | if (f.file->f_op == &bpf_prog_fops) | |
2156 | err = bpf_prog_get_info_by_fd(f.file->private_data, attr, | |
2157 | uattr); | |
2158 | else if (f.file->f_op == &bpf_map_fops) | |
2159 | err = bpf_map_get_info_by_fd(f.file->private_data, attr, | |
2160 | uattr); | |
60197cfb | 2161 | else if (f.file->f_op == &btf_fops) |
62dab84c | 2162 | err = bpf_btf_get_info_by_fd(f.file->private_data, attr, uattr); |
1e270976 MKL |
2163 | else |
2164 | err = -EINVAL; | |
2165 | ||
2166 | fdput(f); | |
2167 | return err; | |
2168 | } | |
2169 | ||
f56a653c MKL |
2170 | #define BPF_BTF_LOAD_LAST_FIELD btf_log_level |
2171 | ||
2172 | static int bpf_btf_load(const union bpf_attr *attr) | |
2173 | { | |
2174 | if (CHECK_ATTR(BPF_BTF_LOAD)) | |
2175 | return -EINVAL; | |
2176 | ||
2177 | if (!capable(CAP_SYS_ADMIN)) | |
2178 | return -EPERM; | |
2179 | ||
2180 | return btf_new_fd(attr); | |
2181 | } | |
2182 | ||
78958fca MKL |
2183 | #define BPF_BTF_GET_FD_BY_ID_LAST_FIELD btf_id |
2184 | ||
2185 | static int bpf_btf_get_fd_by_id(const union bpf_attr *attr) | |
2186 | { | |
2187 | if (CHECK_ATTR(BPF_BTF_GET_FD_BY_ID)) | |
2188 | return -EINVAL; | |
2189 | ||
2190 | if (!capable(CAP_SYS_ADMIN)) | |
2191 | return -EPERM; | |
2192 | ||
2193 | return btf_get_fd_by_id(attr->btf_id); | |
2194 | } | |
2195 | ||
41bdc4b4 YS |
2196 | static int bpf_task_fd_query_copy(const union bpf_attr *attr, |
2197 | union bpf_attr __user *uattr, | |
2198 | u32 prog_id, u32 fd_type, | |
2199 | const char *buf, u64 probe_offset, | |
2200 | u64 probe_addr) | |
2201 | { | |
2202 | char __user *ubuf = u64_to_user_ptr(attr->task_fd_query.buf); | |
2203 | u32 len = buf ? strlen(buf) : 0, input_len; | |
2204 | int err = 0; | |
2205 | ||
2206 | if (put_user(len, &uattr->task_fd_query.buf_len)) | |
2207 | return -EFAULT; | |
2208 | input_len = attr->task_fd_query.buf_len; | |
2209 | if (input_len && ubuf) { | |
2210 | if (!len) { | |
2211 | /* nothing to copy, just make ubuf NULL terminated */ | |
2212 | char zero = '\0'; | |
2213 | ||
2214 | if (put_user(zero, ubuf)) | |
2215 | return -EFAULT; | |
2216 | } else if (input_len >= len + 1) { | |
2217 | /* ubuf can hold the string with NULL terminator */ | |
2218 | if (copy_to_user(ubuf, buf, len + 1)) | |
2219 | return -EFAULT; | |
2220 | } else { | |
2221 | /* ubuf cannot hold the string with NULL terminator, | |
2222 | * do a partial copy with NULL terminator. | |
2223 | */ | |
2224 | char zero = '\0'; | |
2225 | ||
2226 | err = -ENOSPC; | |
2227 | if (copy_to_user(ubuf, buf, input_len - 1)) | |
2228 | return -EFAULT; | |
2229 | if (put_user(zero, ubuf + input_len - 1)) | |
2230 | return -EFAULT; | |
2231 | } | |
2232 | } | |
2233 | ||
2234 | if (put_user(prog_id, &uattr->task_fd_query.prog_id) || | |
2235 | put_user(fd_type, &uattr->task_fd_query.fd_type) || | |
2236 | put_user(probe_offset, &uattr->task_fd_query.probe_offset) || | |
2237 | put_user(probe_addr, &uattr->task_fd_query.probe_addr)) | |
2238 | return -EFAULT; | |
2239 | ||
2240 | return err; | |
2241 | } | |
2242 | ||
2243 | #define BPF_TASK_FD_QUERY_LAST_FIELD task_fd_query.probe_addr | |
2244 | ||
2245 | static int bpf_task_fd_query(const union bpf_attr *attr, | |
2246 | union bpf_attr __user *uattr) | |
2247 | { | |
2248 | pid_t pid = attr->task_fd_query.pid; | |
2249 | u32 fd = attr->task_fd_query.fd; | |
2250 | const struct perf_event *event; | |
2251 | struct files_struct *files; | |
2252 | struct task_struct *task; | |
2253 | struct file *file; | |
2254 | int err; | |
2255 | ||
2256 | if (CHECK_ATTR(BPF_TASK_FD_QUERY)) | |
2257 | return -EINVAL; | |
2258 | ||
2259 | if (!capable(CAP_SYS_ADMIN)) | |
2260 | return -EPERM; | |
2261 | ||
2262 | if (attr->task_fd_query.flags != 0) | |
2263 | return -EINVAL; | |
2264 | ||
2265 | task = get_pid_task(find_vpid(pid), PIDTYPE_PID); | |
2266 | if (!task) | |
2267 | return -ENOENT; | |
2268 | ||
2269 | files = get_files_struct(task); | |
2270 | put_task_struct(task); | |
2271 | if (!files) | |
2272 | return -ENOENT; | |
2273 | ||
2274 | err = 0; | |
2275 | spin_lock(&files->file_lock); | |
2276 | file = fcheck_files(files, fd); | |
2277 | if (!file) | |
2278 | err = -EBADF; | |
2279 | else | |
2280 | get_file(file); | |
2281 | spin_unlock(&files->file_lock); | |
2282 | put_files_struct(files); | |
2283 | ||
2284 | if (err) | |
2285 | goto out; | |
2286 | ||
2287 | if (file->f_op == &bpf_raw_tp_fops) { | |
2288 | struct bpf_raw_tracepoint *raw_tp = file->private_data; | |
2289 | struct bpf_raw_event_map *btp = raw_tp->btp; | |
2290 | ||
2291 | err = bpf_task_fd_query_copy(attr, uattr, | |
2292 | raw_tp->prog->aux->id, | |
2293 | BPF_FD_TYPE_RAW_TRACEPOINT, | |
2294 | btp->tp->name, 0, 0); | |
2295 | goto put_file; | |
2296 | } | |
2297 | ||
2298 | event = perf_get_event(file); | |
2299 | if (!IS_ERR(event)) { | |
2300 | u64 probe_offset, probe_addr; | |
2301 | u32 prog_id, fd_type; | |
2302 | const char *buf; | |
2303 | ||
2304 | err = bpf_get_perf_event_info(event, &prog_id, &fd_type, | |
2305 | &buf, &probe_offset, | |
2306 | &probe_addr); | |
2307 | if (!err) | |
2308 | err = bpf_task_fd_query_copy(attr, uattr, prog_id, | |
2309 | fd_type, buf, | |
2310 | probe_offset, | |
2311 | probe_addr); | |
2312 | goto put_file; | |
2313 | } | |
2314 | ||
2315 | err = -ENOTSUPP; | |
2316 | put_file: | |
2317 | fput(file); | |
2318 | out: | |
2319 | return err; | |
2320 | } | |
2321 | ||
99c55f7d AS |
2322 | SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size) |
2323 | { | |
2324 | union bpf_attr attr = {}; | |
2325 | int err; | |
2326 | ||
0fa4fe85 | 2327 | if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN)) |
99c55f7d AS |
2328 | return -EPERM; |
2329 | ||
dcab51f1 | 2330 | err = bpf_check_uarg_tail_zero(uattr, sizeof(attr), size); |
1e270976 MKL |
2331 | if (err) |
2332 | return err; | |
2333 | size = min_t(u32, size, sizeof(attr)); | |
99c55f7d AS |
2334 | |
2335 | /* copy attributes from user space, may be less than sizeof(bpf_attr) */ | |
2336 | if (copy_from_user(&attr, uattr, size) != 0) | |
2337 | return -EFAULT; | |
2338 | ||
afdb09c7 CF |
2339 | err = security_bpf(cmd, &attr, size); |
2340 | if (err < 0) | |
2341 | return err; | |
2342 | ||
99c55f7d AS |
2343 | switch (cmd) { |
2344 | case BPF_MAP_CREATE: | |
2345 | err = map_create(&attr); | |
2346 | break; | |
db20fd2b AS |
2347 | case BPF_MAP_LOOKUP_ELEM: |
2348 | err = map_lookup_elem(&attr); | |
2349 | break; | |
2350 | case BPF_MAP_UPDATE_ELEM: | |
2351 | err = map_update_elem(&attr); | |
2352 | break; | |
2353 | case BPF_MAP_DELETE_ELEM: | |
2354 | err = map_delete_elem(&attr); | |
2355 | break; | |
2356 | case BPF_MAP_GET_NEXT_KEY: | |
2357 | err = map_get_next_key(&attr); | |
2358 | break; | |
09756af4 AS |
2359 | case BPF_PROG_LOAD: |
2360 | err = bpf_prog_load(&attr); | |
2361 | break; | |
b2197755 DB |
2362 | case BPF_OBJ_PIN: |
2363 | err = bpf_obj_pin(&attr); | |
2364 | break; | |
2365 | case BPF_OBJ_GET: | |
2366 | err = bpf_obj_get(&attr); | |
2367 | break; | |
f4324551 DM |
2368 | #ifdef CONFIG_CGROUP_BPF |
2369 | case BPF_PROG_ATTACH: | |
2370 | err = bpf_prog_attach(&attr); | |
2371 | break; | |
2372 | case BPF_PROG_DETACH: | |
2373 | err = bpf_prog_detach(&attr); | |
2374 | break; | |
468e2f64 AS |
2375 | case BPF_PROG_QUERY: |
2376 | err = bpf_prog_query(&attr, uattr); | |
2377 | break; | |
f4324551 | 2378 | #endif |
1cf1cae9 AS |
2379 | case BPF_PROG_TEST_RUN: |
2380 | err = bpf_prog_test_run(&attr, uattr); | |
2381 | break; | |
34ad5580 MKL |
2382 | case BPF_PROG_GET_NEXT_ID: |
2383 | err = bpf_obj_get_next_id(&attr, uattr, | |
2384 | &prog_idr, &prog_idr_lock); | |
2385 | break; | |
2386 | case BPF_MAP_GET_NEXT_ID: | |
2387 | err = bpf_obj_get_next_id(&attr, uattr, | |
2388 | &map_idr, &map_idr_lock); | |
2389 | break; | |
b16d9aa4 MKL |
2390 | case BPF_PROG_GET_FD_BY_ID: |
2391 | err = bpf_prog_get_fd_by_id(&attr); | |
2392 | break; | |
bd5f5f4e MKL |
2393 | case BPF_MAP_GET_FD_BY_ID: |
2394 | err = bpf_map_get_fd_by_id(&attr); | |
2395 | break; | |
1e270976 MKL |
2396 | case BPF_OBJ_GET_INFO_BY_FD: |
2397 | err = bpf_obj_get_info_by_fd(&attr, uattr); | |
2398 | break; | |
c4f6699d AS |
2399 | case BPF_RAW_TRACEPOINT_OPEN: |
2400 | err = bpf_raw_tracepoint_open(&attr); | |
2401 | break; | |
f56a653c MKL |
2402 | case BPF_BTF_LOAD: |
2403 | err = bpf_btf_load(&attr); | |
2404 | break; | |
78958fca MKL |
2405 | case BPF_BTF_GET_FD_BY_ID: |
2406 | err = bpf_btf_get_fd_by_id(&attr); | |
2407 | break; | |
41bdc4b4 YS |
2408 | case BPF_TASK_FD_QUERY: |
2409 | err = bpf_task_fd_query(&attr, uattr); | |
2410 | break; | |
99c55f7d AS |
2411 | default: |
2412 | err = -EINVAL; | |
2413 | break; | |
2414 | } | |
2415 | ||
2416 | return err; | |
2417 | } |