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> | |
9ef09e35 | 33 | #include <linux/nospec.h> |
99c55f7d | 34 | |
14dc6f04 MKL |
35 | #define IS_FD_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PROG_ARRAY || \ |
36 | (map)->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || \ | |
37 | (map)->map_type == BPF_MAP_TYPE_CGROUP_ARRAY || \ | |
38 | (map)->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS) | |
39 | #define IS_FD_HASH(map) ((map)->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) | |
40 | #define IS_FD_MAP(map) (IS_FD_ARRAY(map) || IS_FD_HASH(map)) | |
41 | ||
6e71b04a CF |
42 | #define BPF_OBJ_FLAG_MASK (BPF_F_RDONLY | BPF_F_WRONLY) |
43 | ||
b121d1e7 | 44 | DEFINE_PER_CPU(int, bpf_prog_active); |
dc4bb0e2 MKL |
45 | static DEFINE_IDR(prog_idr); |
46 | static DEFINE_SPINLOCK(prog_idr_lock); | |
f3f1c054 MKL |
47 | static DEFINE_IDR(map_idr); |
48 | static DEFINE_SPINLOCK(map_idr_lock); | |
b121d1e7 | 49 | |
1be7f75d AS |
50 | int sysctl_unprivileged_bpf_disabled __read_mostly; |
51 | ||
40077e0c JB |
52 | static const struct bpf_map_ops * const bpf_map_types[] = { |
53 | #define BPF_PROG_TYPE(_id, _ops) | |
54 | #define BPF_MAP_TYPE(_id, _ops) \ | |
55 | [_id] = &_ops, | |
56 | #include <linux/bpf_types.h> | |
57 | #undef BPF_PROG_TYPE | |
58 | #undef BPF_MAP_TYPE | |
59 | }; | |
99c55f7d | 60 | |
752ba56f MS |
61 | /* |
62 | * If we're handed a bigger struct than we know of, ensure all the unknown bits | |
63 | * are 0 - i.e. new user-space does not rely on any kernel feature extensions | |
64 | * we don't know about yet. | |
65 | * | |
66 | * There is a ToCToU between this function call and the following | |
67 | * copy_from_user() call. However, this is not a concern since this function is | |
68 | * meant to be a future-proofing of bits. | |
69 | */ | |
dcab51f1 MKL |
70 | int bpf_check_uarg_tail_zero(void __user *uaddr, |
71 | size_t expected_size, | |
72 | size_t actual_size) | |
58291a74 MS |
73 | { |
74 | unsigned char __user *addr; | |
75 | unsigned char __user *end; | |
76 | unsigned char val; | |
77 | int err; | |
78 | ||
752ba56f MS |
79 | if (unlikely(actual_size > PAGE_SIZE)) /* silly large */ |
80 | return -E2BIG; | |
81 | ||
82 | if (unlikely(!access_ok(VERIFY_READ, uaddr, actual_size))) | |
83 | return -EFAULT; | |
84 | ||
58291a74 MS |
85 | if (actual_size <= expected_size) |
86 | return 0; | |
87 | ||
88 | addr = uaddr + expected_size; | |
89 | end = uaddr + actual_size; | |
90 | ||
91 | for (; addr < end; addr++) { | |
92 | err = get_user(val, addr); | |
93 | if (err) | |
94 | return err; | |
95 | if (val) | |
96 | return -E2BIG; | |
97 | } | |
98 | ||
99 | return 0; | |
100 | } | |
101 | ||
a3884572 JK |
102 | const struct bpf_map_ops bpf_map_offload_ops = { |
103 | .map_alloc = bpf_map_offload_map_alloc, | |
104 | .map_free = bpf_map_offload_map_free, | |
e8d2bec0 | 105 | .map_check_btf = map_check_no_btf, |
a3884572 JK |
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 | ||
0a4c58f5 | 184 | static int bpf_charge_memlock(struct user_struct *user, u32 pages) |
aaac3ba9 | 185 | { |
0a4c58f5 | 186 | unsigned long memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; |
aaac3ba9 | 187 | |
0a4c58f5 RG |
188 | if (atomic_long_add_return(pages, &user->locked_vm) > memlock_limit) { |
189 | atomic_long_sub(pages, &user->locked_vm); | |
190 | return -EPERM; | |
191 | } | |
192 | return 0; | |
193 | } | |
aaac3ba9 | 194 | |
0a4c58f5 RG |
195 | static void bpf_uncharge_memlock(struct user_struct *user, u32 pages) |
196 | { | |
197 | atomic_long_sub(pages, &user->locked_vm); | |
198 | } | |
aaac3ba9 | 199 | |
0a4c58f5 RG |
200 | static int bpf_map_init_memlock(struct bpf_map *map) |
201 | { | |
202 | struct user_struct *user = get_current_user(); | |
203 | int ret; | |
aaac3ba9 | 204 | |
0a4c58f5 RG |
205 | ret = bpf_charge_memlock(user, map->pages); |
206 | if (ret) { | |
aaac3ba9 | 207 | free_uid(user); |
0a4c58f5 | 208 | return ret; |
aaac3ba9 AS |
209 | } |
210 | map->user = user; | |
0a4c58f5 | 211 | return ret; |
aaac3ba9 AS |
212 | } |
213 | ||
0a4c58f5 | 214 | static void bpf_map_release_memlock(struct bpf_map *map) |
aaac3ba9 AS |
215 | { |
216 | struct user_struct *user = map->user; | |
0a4c58f5 | 217 | bpf_uncharge_memlock(user, map->pages); |
aaac3ba9 AS |
218 | free_uid(user); |
219 | } | |
220 | ||
0a4c58f5 RG |
221 | int bpf_map_charge_memlock(struct bpf_map *map, u32 pages) |
222 | { | |
223 | int ret; | |
224 | ||
225 | ret = bpf_charge_memlock(map->user, pages); | |
226 | if (ret) | |
227 | return ret; | |
228 | map->pages += pages; | |
229 | return ret; | |
230 | } | |
231 | ||
232 | void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages) | |
233 | { | |
234 | bpf_uncharge_memlock(map->user, pages); | |
235 | map->pages -= pages; | |
236 | } | |
237 | ||
f3f1c054 MKL |
238 | static int bpf_map_alloc_id(struct bpf_map *map) |
239 | { | |
240 | int id; | |
241 | ||
b76354cd | 242 | idr_preload(GFP_KERNEL); |
f3f1c054 MKL |
243 | spin_lock_bh(&map_idr_lock); |
244 | id = idr_alloc_cyclic(&map_idr, map, 1, INT_MAX, GFP_ATOMIC); | |
245 | if (id > 0) | |
246 | map->id = id; | |
247 | spin_unlock_bh(&map_idr_lock); | |
b76354cd | 248 | idr_preload_end(); |
f3f1c054 MKL |
249 | |
250 | if (WARN_ON_ONCE(!id)) | |
251 | return -ENOSPC; | |
252 | ||
253 | return id > 0 ? 0 : id; | |
254 | } | |
255 | ||
a3884572 | 256 | void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock) |
f3f1c054 | 257 | { |
930651a7 ED |
258 | unsigned long flags; |
259 | ||
a3884572 JK |
260 | /* Offloaded maps are removed from the IDR store when their device |
261 | * disappears - even if someone holds an fd to them they are unusable, | |
262 | * the memory is gone, all ops will fail; they are simply waiting for | |
263 | * refcnt to drop to be freed. | |
264 | */ | |
265 | if (!map->id) | |
266 | return; | |
267 | ||
bd5f5f4e | 268 | if (do_idr_lock) |
930651a7 | 269 | spin_lock_irqsave(&map_idr_lock, flags); |
bd5f5f4e MKL |
270 | else |
271 | __acquire(&map_idr_lock); | |
272 | ||
f3f1c054 | 273 | idr_remove(&map_idr, map->id); |
a3884572 | 274 | map->id = 0; |
bd5f5f4e MKL |
275 | |
276 | if (do_idr_lock) | |
930651a7 | 277 | spin_unlock_irqrestore(&map_idr_lock, flags); |
bd5f5f4e MKL |
278 | else |
279 | __release(&map_idr_lock); | |
f3f1c054 MKL |
280 | } |
281 | ||
99c55f7d AS |
282 | /* called from workqueue */ |
283 | static void bpf_map_free_deferred(struct work_struct *work) | |
284 | { | |
285 | struct bpf_map *map = container_of(work, struct bpf_map, work); | |
286 | ||
0a4c58f5 | 287 | bpf_map_release_memlock(map); |
afdb09c7 | 288 | security_bpf_map_free(map); |
99c55f7d AS |
289 | /* implementation dependent freeing */ |
290 | map->ops->map_free(map); | |
291 | } | |
292 | ||
c9da161c DB |
293 | static void bpf_map_put_uref(struct bpf_map *map) |
294 | { | |
295 | if (atomic_dec_and_test(&map->usercnt)) { | |
ba6b8de4 JF |
296 | if (map->ops->map_release_uref) |
297 | map->ops->map_release_uref(map); | |
c9da161c DB |
298 | } |
299 | } | |
300 | ||
99c55f7d AS |
301 | /* decrement map refcnt and schedule it for freeing via workqueue |
302 | * (unrelying map implementation ops->map_free() might sleep) | |
303 | */ | |
bd5f5f4e | 304 | static void __bpf_map_put(struct bpf_map *map, bool do_idr_lock) |
99c55f7d AS |
305 | { |
306 | if (atomic_dec_and_test(&map->refcnt)) { | |
34ad5580 | 307 | /* bpf_map_free_id() must be called first */ |
bd5f5f4e | 308 | bpf_map_free_id(map, do_idr_lock); |
78958fca | 309 | btf_put(map->btf); |
99c55f7d AS |
310 | INIT_WORK(&map->work, bpf_map_free_deferred); |
311 | schedule_work(&map->work); | |
312 | } | |
313 | } | |
314 | ||
bd5f5f4e MKL |
315 | void bpf_map_put(struct bpf_map *map) |
316 | { | |
317 | __bpf_map_put(map, true); | |
318 | } | |
630a4d38 | 319 | EXPORT_SYMBOL_GPL(bpf_map_put); |
bd5f5f4e | 320 | |
c9da161c | 321 | void bpf_map_put_with_uref(struct bpf_map *map) |
99c55f7d | 322 | { |
c9da161c | 323 | bpf_map_put_uref(map); |
99c55f7d | 324 | bpf_map_put(map); |
c9da161c DB |
325 | } |
326 | ||
327 | static int bpf_map_release(struct inode *inode, struct file *filp) | |
328 | { | |
61d1b6a4 DB |
329 | struct bpf_map *map = filp->private_data; |
330 | ||
331 | if (map->ops->map_release) | |
332 | map->ops->map_release(map, filp); | |
333 | ||
334 | bpf_map_put_with_uref(map); | |
99c55f7d AS |
335 | return 0; |
336 | } | |
337 | ||
f99bf205 DB |
338 | #ifdef CONFIG_PROC_FS |
339 | static void bpf_map_show_fdinfo(struct seq_file *m, struct file *filp) | |
340 | { | |
341 | const struct bpf_map *map = filp->private_data; | |
21116b70 DB |
342 | const struct bpf_array *array; |
343 | u32 owner_prog_type = 0; | |
9780c0ab | 344 | u32 owner_jited = 0; |
21116b70 DB |
345 | |
346 | if (map->map_type == BPF_MAP_TYPE_PROG_ARRAY) { | |
347 | array = container_of(map, struct bpf_array, map); | |
348 | owner_prog_type = array->owner_prog_type; | |
9780c0ab | 349 | owner_jited = array->owner_jited; |
21116b70 | 350 | } |
f99bf205 DB |
351 | |
352 | seq_printf(m, | |
353 | "map_type:\t%u\n" | |
354 | "key_size:\t%u\n" | |
355 | "value_size:\t%u\n" | |
322cea2f | 356 | "max_entries:\t%u\n" |
21116b70 | 357 | "map_flags:\t%#x\n" |
4316b409 DB |
358 | "memlock:\t%llu\n" |
359 | "map_id:\t%u\n", | |
f99bf205 DB |
360 | map->map_type, |
361 | map->key_size, | |
362 | map->value_size, | |
322cea2f | 363 | map->max_entries, |
21116b70 | 364 | map->map_flags, |
4316b409 DB |
365 | map->pages * 1ULL << PAGE_SHIFT, |
366 | map->id); | |
21116b70 | 367 | |
9780c0ab | 368 | if (owner_prog_type) { |
21116b70 DB |
369 | seq_printf(m, "owner_prog_type:\t%u\n", |
370 | owner_prog_type); | |
9780c0ab DB |
371 | seq_printf(m, "owner_jited:\t%u\n", |
372 | owner_jited); | |
373 | } | |
f99bf205 DB |
374 | } |
375 | #endif | |
376 | ||
6e71b04a CF |
377 | static ssize_t bpf_dummy_read(struct file *filp, char __user *buf, size_t siz, |
378 | loff_t *ppos) | |
379 | { | |
380 | /* We need this handler such that alloc_file() enables | |
381 | * f_mode with FMODE_CAN_READ. | |
382 | */ | |
383 | return -EINVAL; | |
384 | } | |
385 | ||
386 | static ssize_t bpf_dummy_write(struct file *filp, const char __user *buf, | |
387 | size_t siz, loff_t *ppos) | |
388 | { | |
389 | /* We need this handler such that alloc_file() enables | |
390 | * f_mode with FMODE_CAN_WRITE. | |
391 | */ | |
392 | return -EINVAL; | |
393 | } | |
394 | ||
f66e448c | 395 | const struct file_operations bpf_map_fops = { |
f99bf205 DB |
396 | #ifdef CONFIG_PROC_FS |
397 | .show_fdinfo = bpf_map_show_fdinfo, | |
398 | #endif | |
399 | .release = bpf_map_release, | |
6e71b04a CF |
400 | .read = bpf_dummy_read, |
401 | .write = bpf_dummy_write, | |
99c55f7d AS |
402 | }; |
403 | ||
6e71b04a | 404 | int bpf_map_new_fd(struct bpf_map *map, int flags) |
aa79781b | 405 | { |
afdb09c7 CF |
406 | int ret; |
407 | ||
408 | ret = security_bpf_map(map, OPEN_FMODE(flags)); | |
409 | if (ret < 0) | |
410 | return ret; | |
411 | ||
aa79781b | 412 | return anon_inode_getfd("bpf-map", &bpf_map_fops, map, |
6e71b04a CF |
413 | flags | O_CLOEXEC); |
414 | } | |
415 | ||
416 | int bpf_get_file_flag(int flags) | |
417 | { | |
418 | if ((flags & BPF_F_RDONLY) && (flags & BPF_F_WRONLY)) | |
419 | return -EINVAL; | |
420 | if (flags & BPF_F_RDONLY) | |
421 | return O_RDONLY; | |
422 | if (flags & BPF_F_WRONLY) | |
423 | return O_WRONLY; | |
424 | return O_RDWR; | |
aa79781b DB |
425 | } |
426 | ||
99c55f7d AS |
427 | /* helper macro to check that unused fields 'union bpf_attr' are zero */ |
428 | #define CHECK_ATTR(CMD) \ | |
429 | memchr_inv((void *) &attr->CMD##_LAST_FIELD + \ | |
430 | sizeof(attr->CMD##_LAST_FIELD), 0, \ | |
431 | sizeof(*attr) - \ | |
432 | offsetof(union bpf_attr, CMD##_LAST_FIELD) - \ | |
433 | sizeof(attr->CMD##_LAST_FIELD)) != NULL | |
434 | ||
cb4d2b3f MKL |
435 | /* dst and src must have at least BPF_OBJ_NAME_LEN number of bytes. |
436 | * Return 0 on success and < 0 on error. | |
437 | */ | |
438 | static int bpf_obj_name_cpy(char *dst, const char *src) | |
439 | { | |
440 | const char *end = src + BPF_OBJ_NAME_LEN; | |
441 | ||
473d9734 MKL |
442 | memset(dst, 0, BPF_OBJ_NAME_LEN); |
443 | ||
cb4d2b3f MKL |
444 | /* Copy all isalnum() and '_' char */ |
445 | while (src < end && *src) { | |
446 | if (!isalnum(*src) && *src != '_') | |
447 | return -EINVAL; | |
448 | *dst++ = *src++; | |
449 | } | |
450 | ||
451 | /* No '\0' found in BPF_OBJ_NAME_LEN number of bytes */ | |
452 | if (src == end) | |
453 | return -EINVAL; | |
454 | ||
cb4d2b3f MKL |
455 | return 0; |
456 | } | |
457 | ||
e8d2bec0 DB |
458 | int map_check_no_btf(const struct bpf_map *map, |
459 | const struct btf_type *key_type, | |
460 | const struct btf_type *value_type) | |
461 | { | |
462 | return -ENOTSUPP; | |
463 | } | |
464 | ||
465 | static int map_check_btf(const struct bpf_map *map, const struct btf *btf, | |
466 | u32 btf_key_id, u32 btf_value_id) | |
467 | { | |
468 | const struct btf_type *key_type, *value_type; | |
469 | u32 key_size, value_size; | |
470 | int ret = 0; | |
471 | ||
472 | key_type = btf_type_id_size(btf, &btf_key_id, &key_size); | |
473 | if (!key_type || key_size != map->key_size) | |
474 | return -EINVAL; | |
475 | ||
476 | value_type = btf_type_id_size(btf, &btf_value_id, &value_size); | |
477 | if (!value_type || value_size != map->value_size) | |
478 | return -EINVAL; | |
479 | ||
480 | if (map->ops->map_check_btf) | |
481 | ret = map->ops->map_check_btf(map, key_type, value_type); | |
482 | ||
483 | return ret; | |
484 | } | |
485 | ||
9b2cf328 | 486 | #define BPF_MAP_CREATE_LAST_FIELD btf_value_type_id |
99c55f7d AS |
487 | /* called via syscall */ |
488 | static int map_create(union bpf_attr *attr) | |
489 | { | |
96eabe7a | 490 | int numa_node = bpf_map_attr_numa_node(attr); |
99c55f7d | 491 | struct bpf_map *map; |
6e71b04a | 492 | int f_flags; |
99c55f7d AS |
493 | int err; |
494 | ||
495 | err = CHECK_ATTR(BPF_MAP_CREATE); | |
496 | if (err) | |
497 | return -EINVAL; | |
498 | ||
6e71b04a CF |
499 | f_flags = bpf_get_file_flag(attr->map_flags); |
500 | if (f_flags < 0) | |
501 | return f_flags; | |
502 | ||
96eabe7a | 503 | if (numa_node != NUMA_NO_NODE && |
96e5ae4e ED |
504 | ((unsigned int)numa_node >= nr_node_ids || |
505 | !node_online(numa_node))) | |
96eabe7a MKL |
506 | return -EINVAL; |
507 | ||
99c55f7d AS |
508 | /* find map type and init map: hashtable vs rbtree vs bloom vs ... */ |
509 | map = find_and_alloc_map(attr); | |
510 | if (IS_ERR(map)) | |
511 | return PTR_ERR(map); | |
512 | ||
ad5b177b MKL |
513 | err = bpf_obj_name_cpy(map->name, attr->map_name); |
514 | if (err) | |
515 | goto free_map_nouncharge; | |
516 | ||
99c55f7d | 517 | atomic_set(&map->refcnt, 1); |
c9da161c | 518 | atomic_set(&map->usercnt, 1); |
99c55f7d | 519 | |
e8d2bec0 | 520 | if (attr->btf_key_type_id || attr->btf_value_type_id) { |
a26ca7c9 MKL |
521 | struct btf *btf; |
522 | ||
9b2cf328 | 523 | if (!attr->btf_key_type_id || !attr->btf_value_type_id) { |
a26ca7c9 MKL |
524 | err = -EINVAL; |
525 | goto free_map_nouncharge; | |
526 | } | |
527 | ||
528 | btf = btf_get_by_fd(attr->btf_fd); | |
529 | if (IS_ERR(btf)) { | |
530 | err = PTR_ERR(btf); | |
531 | goto free_map_nouncharge; | |
532 | } | |
533 | ||
e8d2bec0 DB |
534 | err = map_check_btf(map, btf, attr->btf_key_type_id, |
535 | attr->btf_value_type_id); | |
a26ca7c9 MKL |
536 | if (err) { |
537 | btf_put(btf); | |
538 | goto free_map_nouncharge; | |
539 | } | |
540 | ||
541 | map->btf = btf; | |
9b2cf328 MKL |
542 | map->btf_key_type_id = attr->btf_key_type_id; |
543 | map->btf_value_type_id = attr->btf_value_type_id; | |
a26ca7c9 MKL |
544 | } |
545 | ||
afdb09c7 | 546 | err = security_bpf_map_alloc(map); |
aaac3ba9 | 547 | if (err) |
20b2b24f | 548 | goto free_map_nouncharge; |
aaac3ba9 | 549 | |
0a4c58f5 | 550 | err = bpf_map_init_memlock(map); |
afdb09c7 CF |
551 | if (err) |
552 | goto free_map_sec; | |
553 | ||
f3f1c054 MKL |
554 | err = bpf_map_alloc_id(map); |
555 | if (err) | |
556 | goto free_map; | |
557 | ||
6e71b04a | 558 | err = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
559 | if (err < 0) { |
560 | /* failed to allocate fd. | |
561 | * bpf_map_put() is needed because the above | |
562 | * bpf_map_alloc_id() has published the map | |
563 | * to the userspace and the userspace may | |
564 | * have refcnt-ed it through BPF_MAP_GET_FD_BY_ID. | |
565 | */ | |
566 | bpf_map_put(map); | |
567 | return err; | |
568 | } | |
99c55f7d AS |
569 | |
570 | return err; | |
571 | ||
572 | free_map: | |
0a4c58f5 | 573 | bpf_map_release_memlock(map); |
afdb09c7 CF |
574 | free_map_sec: |
575 | security_bpf_map_free(map); | |
20b2b24f | 576 | free_map_nouncharge: |
a26ca7c9 | 577 | btf_put(map->btf); |
99c55f7d AS |
578 | map->ops->map_free(map); |
579 | return err; | |
580 | } | |
581 | ||
db20fd2b AS |
582 | /* if error is returned, fd is released. |
583 | * On success caller should complete fd access with matching fdput() | |
584 | */ | |
c2101297 | 585 | struct bpf_map *__bpf_map_get(struct fd f) |
db20fd2b | 586 | { |
db20fd2b AS |
587 | if (!f.file) |
588 | return ERR_PTR(-EBADF); | |
db20fd2b AS |
589 | if (f.file->f_op != &bpf_map_fops) { |
590 | fdput(f); | |
591 | return ERR_PTR(-EINVAL); | |
592 | } | |
593 | ||
c2101297 DB |
594 | return f.file->private_data; |
595 | } | |
596 | ||
92117d84 AS |
597 | /* prog's and map's refcnt limit */ |
598 | #define BPF_MAX_REFCNT 32768 | |
599 | ||
600 | struct bpf_map *bpf_map_inc(struct bpf_map *map, bool uref) | |
c9da161c | 601 | { |
92117d84 AS |
602 | if (atomic_inc_return(&map->refcnt) > BPF_MAX_REFCNT) { |
603 | atomic_dec(&map->refcnt); | |
604 | return ERR_PTR(-EBUSY); | |
605 | } | |
c9da161c DB |
606 | if (uref) |
607 | atomic_inc(&map->usercnt); | |
92117d84 | 608 | return map; |
c9da161c | 609 | } |
630a4d38 | 610 | EXPORT_SYMBOL_GPL(bpf_map_inc); |
c9da161c DB |
611 | |
612 | struct bpf_map *bpf_map_get_with_uref(u32 ufd) | |
c2101297 DB |
613 | { |
614 | struct fd f = fdget(ufd); | |
615 | struct bpf_map *map; | |
616 | ||
617 | map = __bpf_map_get(f); | |
618 | if (IS_ERR(map)) | |
619 | return map; | |
620 | ||
92117d84 | 621 | map = bpf_map_inc(map, true); |
c2101297 | 622 | fdput(f); |
db20fd2b AS |
623 | |
624 | return map; | |
625 | } | |
626 | ||
bd5f5f4e MKL |
627 | /* map_idr_lock should have been held */ |
628 | static struct bpf_map *bpf_map_inc_not_zero(struct bpf_map *map, | |
629 | bool uref) | |
630 | { | |
631 | int refold; | |
632 | ||
bfc18e38 | 633 | refold = atomic_fetch_add_unless(&map->refcnt, 1, 0); |
bd5f5f4e MKL |
634 | |
635 | if (refold >= BPF_MAX_REFCNT) { | |
636 | __bpf_map_put(map, false); | |
637 | return ERR_PTR(-EBUSY); | |
638 | } | |
639 | ||
640 | if (!refold) | |
641 | return ERR_PTR(-ENOENT); | |
642 | ||
643 | if (uref) | |
644 | atomic_inc(&map->usercnt); | |
645 | ||
646 | return map; | |
647 | } | |
648 | ||
b8cdc051 AS |
649 | int __weak bpf_stackmap_copy(struct bpf_map *map, void *key, void *value) |
650 | { | |
651 | return -ENOTSUPP; | |
652 | } | |
653 | ||
db20fd2b AS |
654 | /* last field in 'union bpf_attr' used by this command */ |
655 | #define BPF_MAP_LOOKUP_ELEM_LAST_FIELD value | |
656 | ||
657 | static int map_lookup_elem(union bpf_attr *attr) | |
658 | { | |
535e7b4b MS |
659 | void __user *ukey = u64_to_user_ptr(attr->key); |
660 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 661 | int ufd = attr->map_fd; |
db20fd2b | 662 | struct bpf_map *map; |
8ebe667c | 663 | void *key, *value, *ptr; |
15a07b33 | 664 | u32 value_size; |
592867bf | 665 | struct fd f; |
db20fd2b AS |
666 | int err; |
667 | ||
668 | if (CHECK_ATTR(BPF_MAP_LOOKUP_ELEM)) | |
669 | return -EINVAL; | |
670 | ||
592867bf | 671 | f = fdget(ufd); |
c2101297 | 672 | map = __bpf_map_get(f); |
db20fd2b AS |
673 | if (IS_ERR(map)) |
674 | return PTR_ERR(map); | |
675 | ||
6e71b04a CF |
676 | if (!(f.file->f_mode & FMODE_CAN_READ)) { |
677 | err = -EPERM; | |
678 | goto err_put; | |
679 | } | |
680 | ||
e4448ed8 AV |
681 | key = memdup_user(ukey, map->key_size); |
682 | if (IS_ERR(key)) { | |
683 | err = PTR_ERR(key); | |
db20fd2b | 684 | goto err_put; |
e4448ed8 | 685 | } |
db20fd2b | 686 | |
15a07b33 | 687 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
8f844938 | 688 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || |
15a07b33 AS |
689 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) |
690 | value_size = round_up(map->value_size, 8) * num_possible_cpus(); | |
14dc6f04 MKL |
691 | else if (IS_FD_MAP(map)) |
692 | value_size = sizeof(u32); | |
15a07b33 AS |
693 | else |
694 | value_size = map->value_size; | |
695 | ||
8ebe667c | 696 | err = -ENOMEM; |
15a07b33 | 697 | value = kmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b | 698 | if (!value) |
8ebe667c AS |
699 | goto free_key; |
700 | ||
a3884572 JK |
701 | if (bpf_map_is_dev_bound(map)) { |
702 | err = bpf_map_offload_lookup_elem(map, key, value); | |
703 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || | |
704 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
15a07b33 AS |
705 | err = bpf_percpu_hash_copy(map, key, value); |
706 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
707 | err = bpf_percpu_array_copy(map, key, value); | |
557c0c6e AS |
708 | } else if (map->map_type == BPF_MAP_TYPE_STACK_TRACE) { |
709 | err = bpf_stackmap_copy(map, key, value); | |
14dc6f04 MKL |
710 | } else if (IS_FD_ARRAY(map)) { |
711 | err = bpf_fd_array_map_lookup_elem(map, key, value); | |
712 | } else if (IS_FD_HASH(map)) { | |
713 | err = bpf_fd_htab_map_lookup_elem(map, key, value); | |
5dc4c4b7 MKL |
714 | } else if (map->map_type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY) { |
715 | err = bpf_fd_reuseport_array_lookup_elem(map, key, value); | |
15a07b33 AS |
716 | } else { |
717 | rcu_read_lock(); | |
718 | ptr = map->ops->map_lookup_elem(map, key); | |
719 | if (ptr) | |
720 | memcpy(value, ptr, value_size); | |
721 | rcu_read_unlock(); | |
722 | err = ptr ? 0 : -ENOENT; | |
723 | } | |
8ebe667c | 724 | |
15a07b33 | 725 | if (err) |
8ebe667c | 726 | goto free_value; |
db20fd2b AS |
727 | |
728 | err = -EFAULT; | |
15a07b33 | 729 | if (copy_to_user(uvalue, value, value_size) != 0) |
8ebe667c | 730 | goto free_value; |
db20fd2b AS |
731 | |
732 | err = 0; | |
733 | ||
8ebe667c AS |
734 | free_value: |
735 | kfree(value); | |
db20fd2b AS |
736 | free_key: |
737 | kfree(key); | |
738 | err_put: | |
739 | fdput(f); | |
740 | return err; | |
741 | } | |
742 | ||
3274f520 | 743 | #define BPF_MAP_UPDATE_ELEM_LAST_FIELD flags |
db20fd2b AS |
744 | |
745 | static int map_update_elem(union bpf_attr *attr) | |
746 | { | |
535e7b4b MS |
747 | void __user *ukey = u64_to_user_ptr(attr->key); |
748 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 749 | int ufd = attr->map_fd; |
db20fd2b AS |
750 | struct bpf_map *map; |
751 | void *key, *value; | |
15a07b33 | 752 | u32 value_size; |
592867bf | 753 | struct fd f; |
db20fd2b AS |
754 | int err; |
755 | ||
756 | if (CHECK_ATTR(BPF_MAP_UPDATE_ELEM)) | |
757 | return -EINVAL; | |
758 | ||
592867bf | 759 | f = fdget(ufd); |
c2101297 | 760 | map = __bpf_map_get(f); |
db20fd2b AS |
761 | if (IS_ERR(map)) |
762 | return PTR_ERR(map); | |
763 | ||
6e71b04a CF |
764 | if (!(f.file->f_mode & FMODE_CAN_WRITE)) { |
765 | err = -EPERM; | |
766 | goto err_put; | |
767 | } | |
768 | ||
e4448ed8 AV |
769 | key = memdup_user(ukey, map->key_size); |
770 | if (IS_ERR(key)) { | |
771 | err = PTR_ERR(key); | |
db20fd2b | 772 | goto err_put; |
e4448ed8 | 773 | } |
db20fd2b | 774 | |
15a07b33 | 775 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
8f844938 | 776 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || |
15a07b33 AS |
777 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) |
778 | value_size = round_up(map->value_size, 8) * num_possible_cpus(); | |
779 | else | |
780 | value_size = map->value_size; | |
781 | ||
db20fd2b | 782 | err = -ENOMEM; |
15a07b33 | 783 | value = kmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b AS |
784 | if (!value) |
785 | goto free_key; | |
786 | ||
787 | err = -EFAULT; | |
15a07b33 | 788 | if (copy_from_user(value, uvalue, value_size) != 0) |
db20fd2b AS |
789 | goto free_value; |
790 | ||
6710e112 | 791 | /* Need to create a kthread, thus must support schedule */ |
a3884572 JK |
792 | if (bpf_map_is_dev_bound(map)) { |
793 | err = bpf_map_offload_update_elem(map, key, value, attr->flags); | |
794 | goto out; | |
99ba2b5a JF |
795 | } else if (map->map_type == BPF_MAP_TYPE_CPUMAP || |
796 | map->map_type == BPF_MAP_TYPE_SOCKHASH || | |
797 | map->map_type == BPF_MAP_TYPE_SOCKMAP) { | |
6710e112 JDB |
798 | err = map->ops->map_update_elem(map, key, value, attr->flags); |
799 | goto out; | |
800 | } | |
801 | ||
b121d1e7 AS |
802 | /* must increment bpf_prog_active to avoid kprobe+bpf triggering from |
803 | * inside bpf map update or delete otherwise deadlocks are possible | |
804 | */ | |
805 | preempt_disable(); | |
806 | __this_cpu_inc(bpf_prog_active); | |
8f844938 MKL |
807 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
808 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
15a07b33 AS |
809 | err = bpf_percpu_hash_update(map, key, value, attr->flags); |
810 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
811 | err = bpf_percpu_array_update(map, key, value, attr->flags); | |
9c147b56 | 812 | } else if (IS_FD_ARRAY(map)) { |
d056a788 DB |
813 | rcu_read_lock(); |
814 | err = bpf_fd_array_map_update_elem(map, f.file, key, value, | |
815 | attr->flags); | |
816 | rcu_read_unlock(); | |
bcc6b1b7 MKL |
817 | } else if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) { |
818 | rcu_read_lock(); | |
819 | err = bpf_fd_htab_map_update_elem(map, f.file, key, value, | |
820 | attr->flags); | |
821 | rcu_read_unlock(); | |
5dc4c4b7 MKL |
822 | } else if (map->map_type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY) { |
823 | /* rcu_read_lock() is not needed */ | |
824 | err = bpf_fd_reuseport_array_update_elem(map, key, value, | |
825 | attr->flags); | |
15a07b33 AS |
826 | } else { |
827 | rcu_read_lock(); | |
828 | err = map->ops->map_update_elem(map, key, value, attr->flags); | |
829 | rcu_read_unlock(); | |
830 | } | |
b121d1e7 AS |
831 | __this_cpu_dec(bpf_prog_active); |
832 | preempt_enable(); | |
6710e112 | 833 | out: |
db20fd2b AS |
834 | free_value: |
835 | kfree(value); | |
836 | free_key: | |
837 | kfree(key); | |
838 | err_put: | |
839 | fdput(f); | |
840 | return err; | |
841 | } | |
842 | ||
843 | #define BPF_MAP_DELETE_ELEM_LAST_FIELD key | |
844 | ||
845 | static int map_delete_elem(union bpf_attr *attr) | |
846 | { | |
535e7b4b | 847 | void __user *ukey = u64_to_user_ptr(attr->key); |
db20fd2b | 848 | int ufd = attr->map_fd; |
db20fd2b | 849 | struct bpf_map *map; |
592867bf | 850 | struct fd f; |
db20fd2b AS |
851 | void *key; |
852 | int err; | |
853 | ||
854 | if (CHECK_ATTR(BPF_MAP_DELETE_ELEM)) | |
855 | return -EINVAL; | |
856 | ||
592867bf | 857 | f = fdget(ufd); |
c2101297 | 858 | map = __bpf_map_get(f); |
db20fd2b AS |
859 | if (IS_ERR(map)) |
860 | return PTR_ERR(map); | |
861 | ||
6e71b04a CF |
862 | if (!(f.file->f_mode & FMODE_CAN_WRITE)) { |
863 | err = -EPERM; | |
864 | goto err_put; | |
865 | } | |
866 | ||
e4448ed8 AV |
867 | key = memdup_user(ukey, map->key_size); |
868 | if (IS_ERR(key)) { | |
869 | err = PTR_ERR(key); | |
db20fd2b | 870 | goto err_put; |
e4448ed8 | 871 | } |
db20fd2b | 872 | |
a3884572 JK |
873 | if (bpf_map_is_dev_bound(map)) { |
874 | err = bpf_map_offload_delete_elem(map, key); | |
875 | goto out; | |
876 | } | |
877 | ||
b121d1e7 AS |
878 | preempt_disable(); |
879 | __this_cpu_inc(bpf_prog_active); | |
db20fd2b AS |
880 | rcu_read_lock(); |
881 | err = map->ops->map_delete_elem(map, key); | |
882 | rcu_read_unlock(); | |
b121d1e7 AS |
883 | __this_cpu_dec(bpf_prog_active); |
884 | preempt_enable(); | |
a3884572 | 885 | out: |
db20fd2b AS |
886 | kfree(key); |
887 | err_put: | |
888 | fdput(f); | |
889 | return err; | |
890 | } | |
891 | ||
892 | /* last field in 'union bpf_attr' used by this command */ | |
893 | #define BPF_MAP_GET_NEXT_KEY_LAST_FIELD next_key | |
894 | ||
895 | static int map_get_next_key(union bpf_attr *attr) | |
896 | { | |
535e7b4b MS |
897 | void __user *ukey = u64_to_user_ptr(attr->key); |
898 | void __user *unext_key = u64_to_user_ptr(attr->next_key); | |
db20fd2b | 899 | int ufd = attr->map_fd; |
db20fd2b AS |
900 | struct bpf_map *map; |
901 | void *key, *next_key; | |
592867bf | 902 | struct fd f; |
db20fd2b AS |
903 | int err; |
904 | ||
905 | if (CHECK_ATTR(BPF_MAP_GET_NEXT_KEY)) | |
906 | return -EINVAL; | |
907 | ||
592867bf | 908 | f = fdget(ufd); |
c2101297 | 909 | map = __bpf_map_get(f); |
db20fd2b AS |
910 | if (IS_ERR(map)) |
911 | return PTR_ERR(map); | |
912 | ||
6e71b04a CF |
913 | if (!(f.file->f_mode & FMODE_CAN_READ)) { |
914 | err = -EPERM; | |
915 | goto err_put; | |
916 | } | |
917 | ||
8fe45924 | 918 | if (ukey) { |
e4448ed8 AV |
919 | key = memdup_user(ukey, map->key_size); |
920 | if (IS_ERR(key)) { | |
921 | err = PTR_ERR(key); | |
8fe45924 | 922 | goto err_put; |
e4448ed8 | 923 | } |
8fe45924 TQ |
924 | } else { |
925 | key = NULL; | |
926 | } | |
db20fd2b AS |
927 | |
928 | err = -ENOMEM; | |
929 | next_key = kmalloc(map->key_size, GFP_USER); | |
930 | if (!next_key) | |
931 | goto free_key; | |
932 | ||
a3884572 JK |
933 | if (bpf_map_is_dev_bound(map)) { |
934 | err = bpf_map_offload_get_next_key(map, key, next_key); | |
935 | goto out; | |
936 | } | |
937 | ||
db20fd2b AS |
938 | rcu_read_lock(); |
939 | err = map->ops->map_get_next_key(map, key, next_key); | |
940 | rcu_read_unlock(); | |
a3884572 | 941 | out: |
db20fd2b AS |
942 | if (err) |
943 | goto free_next_key; | |
944 | ||
945 | err = -EFAULT; | |
946 | if (copy_to_user(unext_key, next_key, map->key_size) != 0) | |
947 | goto free_next_key; | |
948 | ||
949 | err = 0; | |
950 | ||
951 | free_next_key: | |
952 | kfree(next_key); | |
953 | free_key: | |
954 | kfree(key); | |
955 | err_put: | |
956 | fdput(f); | |
957 | return err; | |
958 | } | |
959 | ||
7de16e3a JK |
960 | static const struct bpf_prog_ops * const bpf_prog_types[] = { |
961 | #define BPF_PROG_TYPE(_id, _name) \ | |
962 | [_id] = & _name ## _prog_ops, | |
963 | #define BPF_MAP_TYPE(_id, _ops) | |
964 | #include <linux/bpf_types.h> | |
965 | #undef BPF_PROG_TYPE | |
966 | #undef BPF_MAP_TYPE | |
967 | }; | |
968 | ||
09756af4 AS |
969 | static int find_prog_type(enum bpf_prog_type type, struct bpf_prog *prog) |
970 | { | |
d0f1a451 DB |
971 | const struct bpf_prog_ops *ops; |
972 | ||
973 | if (type >= ARRAY_SIZE(bpf_prog_types)) | |
974 | return -EINVAL; | |
975 | type = array_index_nospec(type, ARRAY_SIZE(bpf_prog_types)); | |
976 | ops = bpf_prog_types[type]; | |
977 | if (!ops) | |
be9370a7 | 978 | return -EINVAL; |
09756af4 | 979 | |
ab3f0063 | 980 | if (!bpf_prog_is_dev_bound(prog->aux)) |
d0f1a451 | 981 | prog->aux->ops = ops; |
ab3f0063 JK |
982 | else |
983 | prog->aux->ops = &bpf_offload_prog_ops; | |
be9370a7 JB |
984 | prog->type = type; |
985 | return 0; | |
09756af4 AS |
986 | } |
987 | ||
988 | /* drop refcnt on maps used by eBPF program and free auxilary data */ | |
989 | static void free_used_maps(struct bpf_prog_aux *aux) | |
990 | { | |
991 | int i; | |
992 | ||
de9cbbaa RG |
993 | if (aux->cgroup_storage) |
994 | bpf_cgroup_storage_release(aux->prog, aux->cgroup_storage); | |
995 | ||
09756af4 AS |
996 | for (i = 0; i < aux->used_map_cnt; i++) |
997 | bpf_map_put(aux->used_maps[i]); | |
998 | ||
999 | kfree(aux->used_maps); | |
1000 | } | |
1001 | ||
5ccb071e DB |
1002 | int __bpf_prog_charge(struct user_struct *user, u32 pages) |
1003 | { | |
1004 | unsigned long memlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
1005 | unsigned long user_bufs; | |
1006 | ||
1007 | if (user) { | |
1008 | user_bufs = atomic_long_add_return(pages, &user->locked_vm); | |
1009 | if (user_bufs > memlock_limit) { | |
1010 | atomic_long_sub(pages, &user->locked_vm); | |
1011 | return -EPERM; | |
1012 | } | |
1013 | } | |
1014 | ||
1015 | return 0; | |
1016 | } | |
1017 | ||
1018 | void __bpf_prog_uncharge(struct user_struct *user, u32 pages) | |
1019 | { | |
1020 | if (user) | |
1021 | atomic_long_sub(pages, &user->locked_vm); | |
1022 | } | |
1023 | ||
aaac3ba9 AS |
1024 | static int bpf_prog_charge_memlock(struct bpf_prog *prog) |
1025 | { | |
1026 | struct user_struct *user = get_current_user(); | |
5ccb071e | 1027 | int ret; |
aaac3ba9 | 1028 | |
5ccb071e DB |
1029 | ret = __bpf_prog_charge(user, prog->pages); |
1030 | if (ret) { | |
aaac3ba9 | 1031 | free_uid(user); |
5ccb071e | 1032 | return ret; |
aaac3ba9 | 1033 | } |
5ccb071e | 1034 | |
aaac3ba9 AS |
1035 | prog->aux->user = user; |
1036 | return 0; | |
1037 | } | |
1038 | ||
1039 | static void bpf_prog_uncharge_memlock(struct bpf_prog *prog) | |
1040 | { | |
1041 | struct user_struct *user = prog->aux->user; | |
1042 | ||
5ccb071e | 1043 | __bpf_prog_uncharge(user, prog->pages); |
aaac3ba9 AS |
1044 | free_uid(user); |
1045 | } | |
1046 | ||
dc4bb0e2 MKL |
1047 | static int bpf_prog_alloc_id(struct bpf_prog *prog) |
1048 | { | |
1049 | int id; | |
1050 | ||
b76354cd | 1051 | idr_preload(GFP_KERNEL); |
dc4bb0e2 MKL |
1052 | spin_lock_bh(&prog_idr_lock); |
1053 | id = idr_alloc_cyclic(&prog_idr, prog, 1, INT_MAX, GFP_ATOMIC); | |
1054 | if (id > 0) | |
1055 | prog->aux->id = id; | |
1056 | spin_unlock_bh(&prog_idr_lock); | |
b76354cd | 1057 | idr_preload_end(); |
dc4bb0e2 MKL |
1058 | |
1059 | /* id is in [1, INT_MAX) */ | |
1060 | if (WARN_ON_ONCE(!id)) | |
1061 | return -ENOSPC; | |
1062 | ||
1063 | return id > 0 ? 0 : id; | |
1064 | } | |
1065 | ||
ad8ad79f | 1066 | void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock) |
dc4bb0e2 | 1067 | { |
ad8ad79f JK |
1068 | /* cBPF to eBPF migrations are currently not in the idr store. |
1069 | * Offloaded programs are removed from the store when their device | |
1070 | * disappears - even if someone grabs an fd to them they are unusable, | |
1071 | * simply waiting for refcnt to drop to be freed. | |
1072 | */ | |
dc4bb0e2 MKL |
1073 | if (!prog->aux->id) |
1074 | return; | |
1075 | ||
b16d9aa4 MKL |
1076 | if (do_idr_lock) |
1077 | spin_lock_bh(&prog_idr_lock); | |
1078 | else | |
1079 | __acquire(&prog_idr_lock); | |
1080 | ||
dc4bb0e2 | 1081 | idr_remove(&prog_idr, prog->aux->id); |
ad8ad79f | 1082 | prog->aux->id = 0; |
b16d9aa4 MKL |
1083 | |
1084 | if (do_idr_lock) | |
1085 | spin_unlock_bh(&prog_idr_lock); | |
1086 | else | |
1087 | __release(&prog_idr_lock); | |
dc4bb0e2 MKL |
1088 | } |
1089 | ||
1aacde3d | 1090 | static void __bpf_prog_put_rcu(struct rcu_head *rcu) |
abf2e7d6 AS |
1091 | { |
1092 | struct bpf_prog_aux *aux = container_of(rcu, struct bpf_prog_aux, rcu); | |
1093 | ||
1094 | free_used_maps(aux); | |
aaac3ba9 | 1095 | bpf_prog_uncharge_memlock(aux->prog); |
afdb09c7 | 1096 | security_bpf_prog_free(aux); |
abf2e7d6 AS |
1097 | bpf_prog_free(aux->prog); |
1098 | } | |
1099 | ||
b16d9aa4 | 1100 | static void __bpf_prog_put(struct bpf_prog *prog, bool do_idr_lock) |
09756af4 | 1101 | { |
a67edbf4 | 1102 | if (atomic_dec_and_test(&prog->aux->refcnt)) { |
34ad5580 | 1103 | /* bpf_prog_free_id() must be called first */ |
b16d9aa4 | 1104 | bpf_prog_free_id(prog, do_idr_lock); |
7d1982b4 | 1105 | bpf_prog_kallsyms_del_all(prog); |
4f74d809 | 1106 | |
1aacde3d | 1107 | call_rcu(&prog->aux->rcu, __bpf_prog_put_rcu); |
a67edbf4 | 1108 | } |
09756af4 | 1109 | } |
b16d9aa4 MKL |
1110 | |
1111 | void bpf_prog_put(struct bpf_prog *prog) | |
1112 | { | |
1113 | __bpf_prog_put(prog, true); | |
1114 | } | |
e2e9b654 | 1115 | EXPORT_SYMBOL_GPL(bpf_prog_put); |
09756af4 AS |
1116 | |
1117 | static int bpf_prog_release(struct inode *inode, struct file *filp) | |
1118 | { | |
1119 | struct bpf_prog *prog = filp->private_data; | |
1120 | ||
1aacde3d | 1121 | bpf_prog_put(prog); |
09756af4 AS |
1122 | return 0; |
1123 | } | |
1124 | ||
7bd509e3 DB |
1125 | #ifdef CONFIG_PROC_FS |
1126 | static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp) | |
1127 | { | |
1128 | const struct bpf_prog *prog = filp->private_data; | |
f1f7714e | 1129 | char prog_tag[sizeof(prog->tag) * 2 + 1] = { }; |
7bd509e3 | 1130 | |
f1f7714e | 1131 | bin2hex(prog_tag, prog->tag, sizeof(prog->tag)); |
7bd509e3 DB |
1132 | seq_printf(m, |
1133 | "prog_type:\t%u\n" | |
1134 | "prog_jited:\t%u\n" | |
f1f7714e | 1135 | "prog_tag:\t%s\n" |
4316b409 DB |
1136 | "memlock:\t%llu\n" |
1137 | "prog_id:\t%u\n", | |
7bd509e3 DB |
1138 | prog->type, |
1139 | prog->jited, | |
f1f7714e | 1140 | prog_tag, |
4316b409 DB |
1141 | prog->pages * 1ULL << PAGE_SHIFT, |
1142 | prog->aux->id); | |
7bd509e3 DB |
1143 | } |
1144 | #endif | |
1145 | ||
f66e448c | 1146 | const struct file_operations bpf_prog_fops = { |
7bd509e3 DB |
1147 | #ifdef CONFIG_PROC_FS |
1148 | .show_fdinfo = bpf_prog_show_fdinfo, | |
1149 | #endif | |
1150 | .release = bpf_prog_release, | |
6e71b04a CF |
1151 | .read = bpf_dummy_read, |
1152 | .write = bpf_dummy_write, | |
09756af4 AS |
1153 | }; |
1154 | ||
b2197755 | 1155 | int bpf_prog_new_fd(struct bpf_prog *prog) |
aa79781b | 1156 | { |
afdb09c7 CF |
1157 | int ret; |
1158 | ||
1159 | ret = security_bpf_prog(prog); | |
1160 | if (ret < 0) | |
1161 | return ret; | |
1162 | ||
aa79781b DB |
1163 | return anon_inode_getfd("bpf-prog", &bpf_prog_fops, prog, |
1164 | O_RDWR | O_CLOEXEC); | |
1165 | } | |
1166 | ||
113214be | 1167 | static struct bpf_prog *____bpf_prog_get(struct fd f) |
09756af4 | 1168 | { |
09756af4 AS |
1169 | if (!f.file) |
1170 | return ERR_PTR(-EBADF); | |
09756af4 AS |
1171 | if (f.file->f_op != &bpf_prog_fops) { |
1172 | fdput(f); | |
1173 | return ERR_PTR(-EINVAL); | |
1174 | } | |
1175 | ||
c2101297 | 1176 | return f.file->private_data; |
09756af4 AS |
1177 | } |
1178 | ||
59d3656d | 1179 | struct bpf_prog *bpf_prog_add(struct bpf_prog *prog, int i) |
92117d84 | 1180 | { |
59d3656d BB |
1181 | if (atomic_add_return(i, &prog->aux->refcnt) > BPF_MAX_REFCNT) { |
1182 | atomic_sub(i, &prog->aux->refcnt); | |
92117d84 AS |
1183 | return ERR_PTR(-EBUSY); |
1184 | } | |
1185 | return prog; | |
1186 | } | |
59d3656d BB |
1187 | EXPORT_SYMBOL_GPL(bpf_prog_add); |
1188 | ||
c540594f DB |
1189 | void bpf_prog_sub(struct bpf_prog *prog, int i) |
1190 | { | |
1191 | /* Only to be used for undoing previous bpf_prog_add() in some | |
1192 | * error path. We still know that another entity in our call | |
1193 | * path holds a reference to the program, thus atomic_sub() can | |
1194 | * be safely used in such cases! | |
1195 | */ | |
1196 | WARN_ON(atomic_sub_return(i, &prog->aux->refcnt) == 0); | |
1197 | } | |
1198 | EXPORT_SYMBOL_GPL(bpf_prog_sub); | |
1199 | ||
59d3656d BB |
1200 | struct bpf_prog *bpf_prog_inc(struct bpf_prog *prog) |
1201 | { | |
1202 | return bpf_prog_add(prog, 1); | |
1203 | } | |
97bc402d | 1204 | EXPORT_SYMBOL_GPL(bpf_prog_inc); |
92117d84 | 1205 | |
b16d9aa4 | 1206 | /* prog_idr_lock should have been held */ |
a6f6df69 | 1207 | struct bpf_prog *bpf_prog_inc_not_zero(struct bpf_prog *prog) |
b16d9aa4 MKL |
1208 | { |
1209 | int refold; | |
1210 | ||
bfc18e38 | 1211 | refold = atomic_fetch_add_unless(&prog->aux->refcnt, 1, 0); |
b16d9aa4 MKL |
1212 | |
1213 | if (refold >= BPF_MAX_REFCNT) { | |
1214 | __bpf_prog_put(prog, false); | |
1215 | return ERR_PTR(-EBUSY); | |
1216 | } | |
1217 | ||
1218 | if (!refold) | |
1219 | return ERR_PTR(-ENOENT); | |
1220 | ||
1221 | return prog; | |
1222 | } | |
a6f6df69 | 1223 | EXPORT_SYMBOL_GPL(bpf_prog_inc_not_zero); |
b16d9aa4 | 1224 | |
040ee692 | 1225 | bool bpf_prog_get_ok(struct bpf_prog *prog, |
288b3de5 | 1226 | enum bpf_prog_type *attach_type, bool attach_drv) |
248f346f | 1227 | { |
288b3de5 JK |
1228 | /* not an attachment, just a refcount inc, always allow */ |
1229 | if (!attach_type) | |
1230 | return true; | |
248f346f JK |
1231 | |
1232 | if (prog->type != *attach_type) | |
1233 | return false; | |
288b3de5 | 1234 | if (bpf_prog_is_dev_bound(prog->aux) && !attach_drv) |
248f346f JK |
1235 | return false; |
1236 | ||
1237 | return true; | |
1238 | } | |
1239 | ||
1240 | static struct bpf_prog *__bpf_prog_get(u32 ufd, enum bpf_prog_type *attach_type, | |
288b3de5 | 1241 | bool attach_drv) |
09756af4 AS |
1242 | { |
1243 | struct fd f = fdget(ufd); | |
1244 | struct bpf_prog *prog; | |
1245 | ||
113214be | 1246 | prog = ____bpf_prog_get(f); |
09756af4 AS |
1247 | if (IS_ERR(prog)) |
1248 | return prog; | |
288b3de5 | 1249 | if (!bpf_prog_get_ok(prog, attach_type, attach_drv)) { |
113214be DB |
1250 | prog = ERR_PTR(-EINVAL); |
1251 | goto out; | |
1252 | } | |
09756af4 | 1253 | |
92117d84 | 1254 | prog = bpf_prog_inc(prog); |
113214be | 1255 | out: |
09756af4 AS |
1256 | fdput(f); |
1257 | return prog; | |
1258 | } | |
113214be DB |
1259 | |
1260 | struct bpf_prog *bpf_prog_get(u32 ufd) | |
1261 | { | |
288b3de5 | 1262 | return __bpf_prog_get(ufd, NULL, false); |
113214be DB |
1263 | } |
1264 | ||
248f346f | 1265 | struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, |
288b3de5 | 1266 | bool attach_drv) |
248f346f | 1267 | { |
4d220ed0 | 1268 | return __bpf_prog_get(ufd, &type, attach_drv); |
248f346f | 1269 | } |
6c8dfe21 | 1270 | EXPORT_SYMBOL_GPL(bpf_prog_get_type_dev); |
248f346f | 1271 | |
aac3fc32 AI |
1272 | /* Initially all BPF programs could be loaded w/o specifying |
1273 | * expected_attach_type. Later for some of them specifying expected_attach_type | |
1274 | * at load time became required so that program could be validated properly. | |
1275 | * Programs of types that are allowed to be loaded both w/ and w/o (for | |
1276 | * backward compatibility) expected_attach_type, should have the default attach | |
1277 | * type assigned to expected_attach_type for the latter case, so that it can be | |
1278 | * validated later at attach time. | |
1279 | * | |
1280 | * bpf_prog_load_fixup_attach_type() sets expected_attach_type in @attr if | |
1281 | * prog type requires it but has some attach types that have to be backward | |
1282 | * compatible. | |
1283 | */ | |
1284 | static void bpf_prog_load_fixup_attach_type(union bpf_attr *attr) | |
1285 | { | |
1286 | switch (attr->prog_type) { | |
1287 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
1288 | /* Unfortunately BPF_ATTACH_TYPE_UNSPEC enumeration doesn't | |
1289 | * exist so checking for non-zero is the way to go here. | |
1290 | */ | |
1291 | if (!attr->expected_attach_type) | |
1292 | attr->expected_attach_type = | |
1293 | BPF_CGROUP_INET_SOCK_CREATE; | |
1294 | break; | |
1295 | } | |
1296 | } | |
1297 | ||
5e43f899 AI |
1298 | static int |
1299 | bpf_prog_load_check_attach_type(enum bpf_prog_type prog_type, | |
1300 | enum bpf_attach_type expected_attach_type) | |
1301 | { | |
4fbac77d | 1302 | switch (prog_type) { |
aac3fc32 AI |
1303 | case BPF_PROG_TYPE_CGROUP_SOCK: |
1304 | switch (expected_attach_type) { | |
1305 | case BPF_CGROUP_INET_SOCK_CREATE: | |
1306 | case BPF_CGROUP_INET4_POST_BIND: | |
1307 | case BPF_CGROUP_INET6_POST_BIND: | |
1308 | return 0; | |
1309 | default: | |
1310 | return -EINVAL; | |
1311 | } | |
4fbac77d AI |
1312 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: |
1313 | switch (expected_attach_type) { | |
1314 | case BPF_CGROUP_INET4_BIND: | |
1315 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1316 | case BPF_CGROUP_INET4_CONNECT: |
1317 | case BPF_CGROUP_INET6_CONNECT: | |
1cedee13 AI |
1318 | case BPF_CGROUP_UDP4_SENDMSG: |
1319 | case BPF_CGROUP_UDP6_SENDMSG: | |
4fbac77d AI |
1320 | return 0; |
1321 | default: | |
1322 | return -EINVAL; | |
1323 | } | |
1324 | default: | |
1325 | return 0; | |
1326 | } | |
5e43f899 AI |
1327 | } |
1328 | ||
09756af4 | 1329 | /* last field in 'union bpf_attr' used by this command */ |
5e43f899 | 1330 | #define BPF_PROG_LOAD_LAST_FIELD expected_attach_type |
09756af4 AS |
1331 | |
1332 | static int bpf_prog_load(union bpf_attr *attr) | |
1333 | { | |
1334 | enum bpf_prog_type type = attr->prog_type; | |
1335 | struct bpf_prog *prog; | |
1336 | int err; | |
1337 | char license[128]; | |
1338 | bool is_gpl; | |
1339 | ||
1340 | if (CHECK_ATTR(BPF_PROG_LOAD)) | |
1341 | return -EINVAL; | |
1342 | ||
e07b98d9 DM |
1343 | if (attr->prog_flags & ~BPF_F_STRICT_ALIGNMENT) |
1344 | return -EINVAL; | |
1345 | ||
09756af4 | 1346 | /* copy eBPF program license from user space */ |
535e7b4b | 1347 | if (strncpy_from_user(license, u64_to_user_ptr(attr->license), |
09756af4 AS |
1348 | sizeof(license) - 1) < 0) |
1349 | return -EFAULT; | |
1350 | license[sizeof(license) - 1] = 0; | |
1351 | ||
1352 | /* eBPF programs must be GPL compatible to use GPL-ed functions */ | |
1353 | is_gpl = license_is_gpl_compatible(license); | |
1354 | ||
ef0915ca DB |
1355 | if (attr->insn_cnt == 0 || attr->insn_cnt > BPF_MAXINSNS) |
1356 | return -E2BIG; | |
09756af4 | 1357 | |
2541517c AS |
1358 | if (type == BPF_PROG_TYPE_KPROBE && |
1359 | attr->kern_version != LINUX_VERSION_CODE) | |
1360 | return -EINVAL; | |
1361 | ||
80b7d819 CF |
1362 | if (type != BPF_PROG_TYPE_SOCKET_FILTER && |
1363 | type != BPF_PROG_TYPE_CGROUP_SKB && | |
1364 | !capable(CAP_SYS_ADMIN)) | |
1be7f75d AS |
1365 | return -EPERM; |
1366 | ||
aac3fc32 | 1367 | bpf_prog_load_fixup_attach_type(attr); |
5e43f899 AI |
1368 | if (bpf_prog_load_check_attach_type(type, attr->expected_attach_type)) |
1369 | return -EINVAL; | |
1370 | ||
09756af4 AS |
1371 | /* plain bpf_prog allocation */ |
1372 | prog = bpf_prog_alloc(bpf_prog_size(attr->insn_cnt), GFP_USER); | |
1373 | if (!prog) | |
1374 | return -ENOMEM; | |
1375 | ||
5e43f899 AI |
1376 | prog->expected_attach_type = attr->expected_attach_type; |
1377 | ||
9a18eedb JK |
1378 | prog->aux->offload_requested = !!attr->prog_ifindex; |
1379 | ||
afdb09c7 | 1380 | err = security_bpf_prog_alloc(prog->aux); |
aaac3ba9 AS |
1381 | if (err) |
1382 | goto free_prog_nouncharge; | |
1383 | ||
afdb09c7 CF |
1384 | err = bpf_prog_charge_memlock(prog); |
1385 | if (err) | |
1386 | goto free_prog_sec; | |
1387 | ||
09756af4 AS |
1388 | prog->len = attr->insn_cnt; |
1389 | ||
1390 | err = -EFAULT; | |
535e7b4b | 1391 | if (copy_from_user(prog->insns, u64_to_user_ptr(attr->insns), |
aafe6ae9 | 1392 | bpf_prog_insn_size(prog)) != 0) |
09756af4 AS |
1393 | goto free_prog; |
1394 | ||
1395 | prog->orig_prog = NULL; | |
a91263d5 | 1396 | prog->jited = 0; |
09756af4 AS |
1397 | |
1398 | atomic_set(&prog->aux->refcnt, 1); | |
a91263d5 | 1399 | prog->gpl_compatible = is_gpl ? 1 : 0; |
09756af4 | 1400 | |
9a18eedb | 1401 | if (bpf_prog_is_dev_bound(prog->aux)) { |
ab3f0063 JK |
1402 | err = bpf_prog_offload_init(prog, attr); |
1403 | if (err) | |
1404 | goto free_prog; | |
1405 | } | |
1406 | ||
09756af4 AS |
1407 | /* find program type: socket_filter vs tracing_filter */ |
1408 | err = find_prog_type(type, prog); | |
1409 | if (err < 0) | |
1410 | goto free_prog; | |
1411 | ||
cb4d2b3f MKL |
1412 | prog->aux->load_time = ktime_get_boot_ns(); |
1413 | err = bpf_obj_name_cpy(prog->aux->name, attr->prog_name); | |
1414 | if (err) | |
1415 | goto free_prog; | |
1416 | ||
09756af4 | 1417 | /* run eBPF verifier */ |
9bac3d6d | 1418 | err = bpf_check(&prog, attr); |
09756af4 AS |
1419 | if (err < 0) |
1420 | goto free_used_maps; | |
1421 | ||
9facc336 | 1422 | prog = bpf_prog_select_runtime(prog, &err); |
04fd61ab AS |
1423 | if (err < 0) |
1424 | goto free_used_maps; | |
09756af4 | 1425 | |
dc4bb0e2 MKL |
1426 | err = bpf_prog_alloc_id(prog); |
1427 | if (err) | |
1428 | goto free_used_maps; | |
1429 | ||
aa79781b | 1430 | err = bpf_prog_new_fd(prog); |
b16d9aa4 MKL |
1431 | if (err < 0) { |
1432 | /* failed to allocate fd. | |
1433 | * bpf_prog_put() is needed because the above | |
1434 | * bpf_prog_alloc_id() has published the prog | |
1435 | * to the userspace and the userspace may | |
1436 | * have refcnt-ed it through BPF_PROG_GET_FD_BY_ID. | |
1437 | */ | |
1438 | bpf_prog_put(prog); | |
1439 | return err; | |
1440 | } | |
09756af4 | 1441 | |
74451e66 | 1442 | bpf_prog_kallsyms_add(prog); |
09756af4 AS |
1443 | return err; |
1444 | ||
1445 | free_used_maps: | |
7d1982b4 | 1446 | bpf_prog_kallsyms_del_subprogs(prog); |
09756af4 AS |
1447 | free_used_maps(prog->aux); |
1448 | free_prog: | |
aaac3ba9 | 1449 | bpf_prog_uncharge_memlock(prog); |
afdb09c7 CF |
1450 | free_prog_sec: |
1451 | security_bpf_prog_free(prog->aux); | |
aaac3ba9 | 1452 | free_prog_nouncharge: |
09756af4 AS |
1453 | bpf_prog_free(prog); |
1454 | return err; | |
1455 | } | |
1456 | ||
6e71b04a | 1457 | #define BPF_OBJ_LAST_FIELD file_flags |
b2197755 DB |
1458 | |
1459 | static int bpf_obj_pin(const union bpf_attr *attr) | |
1460 | { | |
6e71b04a | 1461 | if (CHECK_ATTR(BPF_OBJ) || attr->file_flags != 0) |
b2197755 DB |
1462 | return -EINVAL; |
1463 | ||
535e7b4b | 1464 | return bpf_obj_pin_user(attr->bpf_fd, u64_to_user_ptr(attr->pathname)); |
b2197755 DB |
1465 | } |
1466 | ||
1467 | static int bpf_obj_get(const union bpf_attr *attr) | |
1468 | { | |
6e71b04a CF |
1469 | if (CHECK_ATTR(BPF_OBJ) || attr->bpf_fd != 0 || |
1470 | attr->file_flags & ~BPF_OBJ_FLAG_MASK) | |
b2197755 DB |
1471 | return -EINVAL; |
1472 | ||
6e71b04a CF |
1473 | return bpf_obj_get_user(u64_to_user_ptr(attr->pathname), |
1474 | attr->file_flags); | |
b2197755 DB |
1475 | } |
1476 | ||
c4f6699d AS |
1477 | struct bpf_raw_tracepoint { |
1478 | struct bpf_raw_event_map *btp; | |
1479 | struct bpf_prog *prog; | |
1480 | }; | |
1481 | ||
1482 | static int bpf_raw_tracepoint_release(struct inode *inode, struct file *filp) | |
1483 | { | |
1484 | struct bpf_raw_tracepoint *raw_tp = filp->private_data; | |
1485 | ||
1486 | if (raw_tp->prog) { | |
1487 | bpf_probe_unregister(raw_tp->btp, raw_tp->prog); | |
1488 | bpf_prog_put(raw_tp->prog); | |
1489 | } | |
1490 | kfree(raw_tp); | |
1491 | return 0; | |
1492 | } | |
1493 | ||
1494 | static const struct file_operations bpf_raw_tp_fops = { | |
1495 | .release = bpf_raw_tracepoint_release, | |
1496 | .read = bpf_dummy_read, | |
1497 | .write = bpf_dummy_write, | |
1498 | }; | |
1499 | ||
1500 | #define BPF_RAW_TRACEPOINT_OPEN_LAST_FIELD raw_tracepoint.prog_fd | |
1501 | ||
1502 | static int bpf_raw_tracepoint_open(const union bpf_attr *attr) | |
1503 | { | |
1504 | struct bpf_raw_tracepoint *raw_tp; | |
1505 | struct bpf_raw_event_map *btp; | |
1506 | struct bpf_prog *prog; | |
1507 | char tp_name[128]; | |
1508 | int tp_fd, err; | |
1509 | ||
1510 | if (strncpy_from_user(tp_name, u64_to_user_ptr(attr->raw_tracepoint.name), | |
1511 | sizeof(tp_name) - 1) < 0) | |
1512 | return -EFAULT; | |
1513 | tp_name[sizeof(tp_name) - 1] = 0; | |
1514 | ||
1515 | btp = bpf_find_raw_tracepoint(tp_name); | |
1516 | if (!btp) | |
1517 | return -ENOENT; | |
1518 | ||
1519 | raw_tp = kzalloc(sizeof(*raw_tp), GFP_USER); | |
1520 | if (!raw_tp) | |
1521 | return -ENOMEM; | |
1522 | raw_tp->btp = btp; | |
1523 | ||
1524 | prog = bpf_prog_get_type(attr->raw_tracepoint.prog_fd, | |
1525 | BPF_PROG_TYPE_RAW_TRACEPOINT); | |
1526 | if (IS_ERR(prog)) { | |
1527 | err = PTR_ERR(prog); | |
1528 | goto out_free_tp; | |
1529 | } | |
1530 | ||
1531 | err = bpf_probe_register(raw_tp->btp, prog); | |
1532 | if (err) | |
1533 | goto out_put_prog; | |
1534 | ||
1535 | raw_tp->prog = prog; | |
1536 | tp_fd = anon_inode_getfd("bpf-raw-tracepoint", &bpf_raw_tp_fops, raw_tp, | |
1537 | O_CLOEXEC); | |
1538 | if (tp_fd < 0) { | |
1539 | bpf_probe_unregister(raw_tp->btp, prog); | |
1540 | err = tp_fd; | |
1541 | goto out_put_prog; | |
1542 | } | |
1543 | return tp_fd; | |
1544 | ||
1545 | out_put_prog: | |
1546 | bpf_prog_put(prog); | |
1547 | out_free_tp: | |
1548 | kfree(raw_tp); | |
1549 | return err; | |
1550 | } | |
1551 | ||
33491588 AR |
1552 | static int bpf_prog_attach_check_attach_type(const struct bpf_prog *prog, |
1553 | enum bpf_attach_type attach_type) | |
1554 | { | |
1555 | switch (prog->type) { | |
1556 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
1557 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: | |
1558 | return attach_type == prog->expected_attach_type ? 0 : -EINVAL; | |
1559 | default: | |
1560 | return 0; | |
1561 | } | |
1562 | } | |
1563 | ||
464bc0fd | 1564 | #define BPF_PROG_ATTACH_LAST_FIELD attach_flags |
174a79ff | 1565 | |
324bda9e AS |
1566 | #define BPF_F_ATTACH_MASK \ |
1567 | (BPF_F_ALLOW_OVERRIDE | BPF_F_ALLOW_MULTI) | |
1568 | ||
f4324551 DM |
1569 | static int bpf_prog_attach(const union bpf_attr *attr) |
1570 | { | |
7f677633 | 1571 | enum bpf_prog_type ptype; |
f4324551 | 1572 | struct bpf_prog *prog; |
7f677633 | 1573 | int ret; |
f4324551 DM |
1574 | |
1575 | if (!capable(CAP_NET_ADMIN)) | |
1576 | return -EPERM; | |
1577 | ||
1578 | if (CHECK_ATTR(BPF_PROG_ATTACH)) | |
1579 | return -EINVAL; | |
1580 | ||
324bda9e | 1581 | if (attr->attach_flags & ~BPF_F_ATTACH_MASK) |
7f677633 AS |
1582 | return -EINVAL; |
1583 | ||
f4324551 DM |
1584 | switch (attr->attach_type) { |
1585 | case BPF_CGROUP_INET_INGRESS: | |
1586 | case BPF_CGROUP_INET_EGRESS: | |
b2cd1257 | 1587 | ptype = BPF_PROG_TYPE_CGROUP_SKB; |
f4324551 | 1588 | break; |
61023658 | 1589 | case BPF_CGROUP_INET_SOCK_CREATE: |
aac3fc32 AI |
1590 | case BPF_CGROUP_INET4_POST_BIND: |
1591 | case BPF_CGROUP_INET6_POST_BIND: | |
61023658 DA |
1592 | ptype = BPF_PROG_TYPE_CGROUP_SOCK; |
1593 | break; | |
4fbac77d AI |
1594 | case BPF_CGROUP_INET4_BIND: |
1595 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1596 | case BPF_CGROUP_INET4_CONNECT: |
1597 | case BPF_CGROUP_INET6_CONNECT: | |
1cedee13 AI |
1598 | case BPF_CGROUP_UDP4_SENDMSG: |
1599 | case BPF_CGROUP_UDP6_SENDMSG: | |
4fbac77d AI |
1600 | ptype = BPF_PROG_TYPE_CGROUP_SOCK_ADDR; |
1601 | break; | |
40304b2a LB |
1602 | case BPF_CGROUP_SOCK_OPS: |
1603 | ptype = BPF_PROG_TYPE_SOCK_OPS; | |
1604 | break; | |
ebc614f6 RG |
1605 | case BPF_CGROUP_DEVICE: |
1606 | ptype = BPF_PROG_TYPE_CGROUP_DEVICE; | |
1607 | break; | |
4f738adb | 1608 | case BPF_SK_MSG_VERDICT: |
fdb5c453 SY |
1609 | ptype = BPF_PROG_TYPE_SK_MSG; |
1610 | break; | |
464bc0fd JF |
1611 | case BPF_SK_SKB_STREAM_PARSER: |
1612 | case BPF_SK_SKB_STREAM_VERDICT: | |
fdb5c453 SY |
1613 | ptype = BPF_PROG_TYPE_SK_SKB; |
1614 | break; | |
f4364dcf | 1615 | case BPF_LIRC_MODE2: |
fdb5c453 SY |
1616 | ptype = BPF_PROG_TYPE_LIRC_MODE2; |
1617 | break; | |
f4324551 DM |
1618 | default: |
1619 | return -EINVAL; | |
1620 | } | |
1621 | ||
b2cd1257 DA |
1622 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); |
1623 | if (IS_ERR(prog)) | |
1624 | return PTR_ERR(prog); | |
1625 | ||
5e43f899 AI |
1626 | if (bpf_prog_attach_check_attach_type(prog, attr->attach_type)) { |
1627 | bpf_prog_put(prog); | |
1628 | return -EINVAL; | |
1629 | } | |
1630 | ||
fdb5c453 SY |
1631 | switch (ptype) { |
1632 | case BPF_PROG_TYPE_SK_SKB: | |
1633 | case BPF_PROG_TYPE_SK_MSG: | |
1634 | ret = sockmap_get_from_fd(attr, ptype, prog); | |
1635 | break; | |
1636 | case BPF_PROG_TYPE_LIRC_MODE2: | |
1637 | ret = lirc_prog_attach(attr, prog); | |
1638 | break; | |
1639 | default: | |
1640 | ret = cgroup_bpf_prog_attach(attr, ptype, prog); | |
b2cd1257 DA |
1641 | } |
1642 | ||
7f677633 AS |
1643 | if (ret) |
1644 | bpf_prog_put(prog); | |
7f677633 | 1645 | return ret; |
f4324551 DM |
1646 | } |
1647 | ||
1648 | #define BPF_PROG_DETACH_LAST_FIELD attach_type | |
1649 | ||
1650 | static int bpf_prog_detach(const union bpf_attr *attr) | |
1651 | { | |
324bda9e | 1652 | enum bpf_prog_type ptype; |
f4324551 DM |
1653 | |
1654 | if (!capable(CAP_NET_ADMIN)) | |
1655 | return -EPERM; | |
1656 | ||
1657 | if (CHECK_ATTR(BPF_PROG_DETACH)) | |
1658 | return -EINVAL; | |
1659 | ||
1660 | switch (attr->attach_type) { | |
1661 | case BPF_CGROUP_INET_INGRESS: | |
1662 | case BPF_CGROUP_INET_EGRESS: | |
324bda9e AS |
1663 | ptype = BPF_PROG_TYPE_CGROUP_SKB; |
1664 | break; | |
61023658 | 1665 | case BPF_CGROUP_INET_SOCK_CREATE: |
aac3fc32 AI |
1666 | case BPF_CGROUP_INET4_POST_BIND: |
1667 | case BPF_CGROUP_INET6_POST_BIND: | |
324bda9e AS |
1668 | ptype = BPF_PROG_TYPE_CGROUP_SOCK; |
1669 | break; | |
4fbac77d AI |
1670 | case BPF_CGROUP_INET4_BIND: |
1671 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
1672 | case BPF_CGROUP_INET4_CONNECT: |
1673 | case BPF_CGROUP_INET6_CONNECT: | |
1cedee13 AI |
1674 | case BPF_CGROUP_UDP4_SENDMSG: |
1675 | case BPF_CGROUP_UDP6_SENDMSG: | |
4fbac77d AI |
1676 | ptype = BPF_PROG_TYPE_CGROUP_SOCK_ADDR; |
1677 | break; | |
40304b2a | 1678 | case BPF_CGROUP_SOCK_OPS: |
324bda9e | 1679 | ptype = BPF_PROG_TYPE_SOCK_OPS; |
f4324551 | 1680 | break; |
ebc614f6 RG |
1681 | case BPF_CGROUP_DEVICE: |
1682 | ptype = BPF_PROG_TYPE_CGROUP_DEVICE; | |
1683 | break; | |
4f738adb | 1684 | case BPF_SK_MSG_VERDICT: |
fdb5c453 | 1685 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_MSG, NULL); |
5a67da2a JF |
1686 | case BPF_SK_SKB_STREAM_PARSER: |
1687 | case BPF_SK_SKB_STREAM_VERDICT: | |
fdb5c453 | 1688 | return sockmap_get_from_fd(attr, BPF_PROG_TYPE_SK_SKB, NULL); |
f4364dcf SY |
1689 | case BPF_LIRC_MODE2: |
1690 | return lirc_prog_detach(attr); | |
f4324551 DM |
1691 | default: |
1692 | return -EINVAL; | |
1693 | } | |
1694 | ||
fdb5c453 | 1695 | return cgroup_bpf_prog_detach(attr, ptype); |
f4324551 | 1696 | } |
40304b2a | 1697 | |
468e2f64 AS |
1698 | #define BPF_PROG_QUERY_LAST_FIELD query.prog_cnt |
1699 | ||
1700 | static int bpf_prog_query(const union bpf_attr *attr, | |
1701 | union bpf_attr __user *uattr) | |
1702 | { | |
468e2f64 AS |
1703 | if (!capable(CAP_NET_ADMIN)) |
1704 | return -EPERM; | |
1705 | if (CHECK_ATTR(BPF_PROG_QUERY)) | |
1706 | return -EINVAL; | |
1707 | if (attr->query.query_flags & ~BPF_F_QUERY_EFFECTIVE) | |
1708 | return -EINVAL; | |
1709 | ||
1710 | switch (attr->query.attach_type) { | |
1711 | case BPF_CGROUP_INET_INGRESS: | |
1712 | case BPF_CGROUP_INET_EGRESS: | |
1713 | case BPF_CGROUP_INET_SOCK_CREATE: | |
4fbac77d AI |
1714 | case BPF_CGROUP_INET4_BIND: |
1715 | case BPF_CGROUP_INET6_BIND: | |
aac3fc32 AI |
1716 | case BPF_CGROUP_INET4_POST_BIND: |
1717 | case BPF_CGROUP_INET6_POST_BIND: | |
d74bad4e AI |
1718 | case BPF_CGROUP_INET4_CONNECT: |
1719 | case BPF_CGROUP_INET6_CONNECT: | |
1cedee13 AI |
1720 | case BPF_CGROUP_UDP4_SENDMSG: |
1721 | case BPF_CGROUP_UDP6_SENDMSG: | |
468e2f64 | 1722 | case BPF_CGROUP_SOCK_OPS: |
ebc614f6 | 1723 | case BPF_CGROUP_DEVICE: |
468e2f64 | 1724 | break; |
f4364dcf SY |
1725 | case BPF_LIRC_MODE2: |
1726 | return lirc_prog_query(attr, uattr); | |
468e2f64 AS |
1727 | default: |
1728 | return -EINVAL; | |
1729 | } | |
fdb5c453 SY |
1730 | |
1731 | return cgroup_bpf_prog_query(attr, uattr); | |
468e2f64 | 1732 | } |
f4324551 | 1733 | |
1cf1cae9 AS |
1734 | #define BPF_PROG_TEST_RUN_LAST_FIELD test.duration |
1735 | ||
1736 | static int bpf_prog_test_run(const union bpf_attr *attr, | |
1737 | union bpf_attr __user *uattr) | |
1738 | { | |
1739 | struct bpf_prog *prog; | |
1740 | int ret = -ENOTSUPP; | |
1741 | ||
61f3c964 AS |
1742 | if (!capable(CAP_SYS_ADMIN)) |
1743 | return -EPERM; | |
1cf1cae9 AS |
1744 | if (CHECK_ATTR(BPF_PROG_TEST_RUN)) |
1745 | return -EINVAL; | |
1746 | ||
1747 | prog = bpf_prog_get(attr->test.prog_fd); | |
1748 | if (IS_ERR(prog)) | |
1749 | return PTR_ERR(prog); | |
1750 | ||
1751 | if (prog->aux->ops->test_run) | |
1752 | ret = prog->aux->ops->test_run(prog, attr, uattr); | |
1753 | ||
1754 | bpf_prog_put(prog); | |
1755 | return ret; | |
1756 | } | |
1757 | ||
34ad5580 MKL |
1758 | #define BPF_OBJ_GET_NEXT_ID_LAST_FIELD next_id |
1759 | ||
1760 | static int bpf_obj_get_next_id(const union bpf_attr *attr, | |
1761 | union bpf_attr __user *uattr, | |
1762 | struct idr *idr, | |
1763 | spinlock_t *lock) | |
1764 | { | |
1765 | u32 next_id = attr->start_id; | |
1766 | int err = 0; | |
1767 | ||
1768 | if (CHECK_ATTR(BPF_OBJ_GET_NEXT_ID) || next_id >= INT_MAX) | |
1769 | return -EINVAL; | |
1770 | ||
1771 | if (!capable(CAP_SYS_ADMIN)) | |
1772 | return -EPERM; | |
1773 | ||
1774 | next_id++; | |
1775 | spin_lock_bh(lock); | |
1776 | if (!idr_get_next(idr, &next_id)) | |
1777 | err = -ENOENT; | |
1778 | spin_unlock_bh(lock); | |
1779 | ||
1780 | if (!err) | |
1781 | err = put_user(next_id, &uattr->next_id); | |
1782 | ||
1783 | return err; | |
1784 | } | |
1785 | ||
b16d9aa4 MKL |
1786 | #define BPF_PROG_GET_FD_BY_ID_LAST_FIELD prog_id |
1787 | ||
1788 | static int bpf_prog_get_fd_by_id(const union bpf_attr *attr) | |
1789 | { | |
1790 | struct bpf_prog *prog; | |
1791 | u32 id = attr->prog_id; | |
1792 | int fd; | |
1793 | ||
1794 | if (CHECK_ATTR(BPF_PROG_GET_FD_BY_ID)) | |
1795 | return -EINVAL; | |
1796 | ||
1797 | if (!capable(CAP_SYS_ADMIN)) | |
1798 | return -EPERM; | |
1799 | ||
1800 | spin_lock_bh(&prog_idr_lock); | |
1801 | prog = idr_find(&prog_idr, id); | |
1802 | if (prog) | |
1803 | prog = bpf_prog_inc_not_zero(prog); | |
1804 | else | |
1805 | prog = ERR_PTR(-ENOENT); | |
1806 | spin_unlock_bh(&prog_idr_lock); | |
1807 | ||
1808 | if (IS_ERR(prog)) | |
1809 | return PTR_ERR(prog); | |
1810 | ||
1811 | fd = bpf_prog_new_fd(prog); | |
1812 | if (fd < 0) | |
1813 | bpf_prog_put(prog); | |
1814 | ||
1815 | return fd; | |
1816 | } | |
1817 | ||
6e71b04a | 1818 | #define BPF_MAP_GET_FD_BY_ID_LAST_FIELD open_flags |
bd5f5f4e MKL |
1819 | |
1820 | static int bpf_map_get_fd_by_id(const union bpf_attr *attr) | |
1821 | { | |
1822 | struct bpf_map *map; | |
1823 | u32 id = attr->map_id; | |
6e71b04a | 1824 | int f_flags; |
bd5f5f4e MKL |
1825 | int fd; |
1826 | ||
6e71b04a CF |
1827 | if (CHECK_ATTR(BPF_MAP_GET_FD_BY_ID) || |
1828 | attr->open_flags & ~BPF_OBJ_FLAG_MASK) | |
bd5f5f4e MKL |
1829 | return -EINVAL; |
1830 | ||
1831 | if (!capable(CAP_SYS_ADMIN)) | |
1832 | return -EPERM; | |
1833 | ||
6e71b04a CF |
1834 | f_flags = bpf_get_file_flag(attr->open_flags); |
1835 | if (f_flags < 0) | |
1836 | return f_flags; | |
1837 | ||
bd5f5f4e MKL |
1838 | spin_lock_bh(&map_idr_lock); |
1839 | map = idr_find(&map_idr, id); | |
1840 | if (map) | |
1841 | map = bpf_map_inc_not_zero(map, true); | |
1842 | else | |
1843 | map = ERR_PTR(-ENOENT); | |
1844 | spin_unlock_bh(&map_idr_lock); | |
1845 | ||
1846 | if (IS_ERR(map)) | |
1847 | return PTR_ERR(map); | |
1848 | ||
6e71b04a | 1849 | fd = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
1850 | if (fd < 0) |
1851 | bpf_map_put(map); | |
1852 | ||
1853 | return fd; | |
1854 | } | |
1855 | ||
7105e828 DB |
1856 | static const struct bpf_map *bpf_map_from_imm(const struct bpf_prog *prog, |
1857 | unsigned long addr) | |
1858 | { | |
1859 | int i; | |
1860 | ||
1861 | for (i = 0; i < prog->aux->used_map_cnt; i++) | |
1862 | if (prog->aux->used_maps[i] == (void *)addr) | |
1863 | return prog->aux->used_maps[i]; | |
1864 | return NULL; | |
1865 | } | |
1866 | ||
1867 | static struct bpf_insn *bpf_insn_prepare_dump(const struct bpf_prog *prog) | |
1868 | { | |
1869 | const struct bpf_map *map; | |
1870 | struct bpf_insn *insns; | |
1871 | u64 imm; | |
1872 | int i; | |
1873 | ||
1874 | insns = kmemdup(prog->insnsi, bpf_prog_insn_size(prog), | |
1875 | GFP_USER); | |
1876 | if (!insns) | |
1877 | return insns; | |
1878 | ||
1879 | for (i = 0; i < prog->len; i++) { | |
1880 | if (insns[i].code == (BPF_JMP | BPF_TAIL_CALL)) { | |
1881 | insns[i].code = BPF_JMP | BPF_CALL; | |
1882 | insns[i].imm = BPF_FUNC_tail_call; | |
1883 | /* fall-through */ | |
1884 | } | |
1885 | if (insns[i].code == (BPF_JMP | BPF_CALL) || | |
1886 | insns[i].code == (BPF_JMP | BPF_CALL_ARGS)) { | |
1887 | if (insns[i].code == (BPF_JMP | BPF_CALL_ARGS)) | |
1888 | insns[i].code = BPF_JMP | BPF_CALL; | |
1889 | if (!bpf_dump_raw_ok()) | |
1890 | insns[i].imm = 0; | |
1891 | continue; | |
1892 | } | |
1893 | ||
1894 | if (insns[i].code != (BPF_LD | BPF_IMM | BPF_DW)) | |
1895 | continue; | |
1896 | ||
1897 | imm = ((u64)insns[i + 1].imm << 32) | (u32)insns[i].imm; | |
1898 | map = bpf_map_from_imm(prog, imm); | |
1899 | if (map) { | |
1900 | insns[i].src_reg = BPF_PSEUDO_MAP_FD; | |
1901 | insns[i].imm = map->id; | |
1902 | insns[i + 1].imm = 0; | |
1903 | continue; | |
1904 | } | |
1905 | ||
1906 | if (!bpf_dump_raw_ok() && | |
1907 | imm == (unsigned long)prog->aux) { | |
1908 | insns[i].imm = 0; | |
1909 | insns[i + 1].imm = 0; | |
1910 | continue; | |
1911 | } | |
1912 | } | |
1913 | ||
1914 | return insns; | |
1915 | } | |
1916 | ||
1e270976 MKL |
1917 | static int bpf_prog_get_info_by_fd(struct bpf_prog *prog, |
1918 | const union bpf_attr *attr, | |
1919 | union bpf_attr __user *uattr) | |
1920 | { | |
1921 | struct bpf_prog_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
1922 | struct bpf_prog_info info = {}; | |
1923 | u32 info_len = attr->info.info_len; | |
1924 | char __user *uinsns; | |
1925 | u32 ulen; | |
1926 | int err; | |
1927 | ||
dcab51f1 | 1928 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(info), info_len); |
1e270976 MKL |
1929 | if (err) |
1930 | return err; | |
1931 | info_len = min_t(u32, sizeof(info), info_len); | |
1932 | ||
1933 | if (copy_from_user(&info, uinfo, info_len)) | |
89b09689 | 1934 | return -EFAULT; |
1e270976 MKL |
1935 | |
1936 | info.type = prog->type; | |
1937 | info.id = prog->aux->id; | |
cb4d2b3f MKL |
1938 | info.load_time = prog->aux->load_time; |
1939 | info.created_by_uid = from_kuid_munged(current_user_ns(), | |
1940 | prog->aux->user->uid); | |
b85fab0e | 1941 | info.gpl_compatible = prog->gpl_compatible; |
1e270976 MKL |
1942 | |
1943 | memcpy(info.tag, prog->tag, sizeof(prog->tag)); | |
cb4d2b3f MKL |
1944 | memcpy(info.name, prog->aux->name, sizeof(prog->aux->name)); |
1945 | ||
1946 | ulen = info.nr_map_ids; | |
1947 | info.nr_map_ids = prog->aux->used_map_cnt; | |
1948 | ulen = min_t(u32, info.nr_map_ids, ulen); | |
1949 | if (ulen) { | |
721e08da | 1950 | u32 __user *user_map_ids = u64_to_user_ptr(info.map_ids); |
cb4d2b3f MKL |
1951 | u32 i; |
1952 | ||
1953 | for (i = 0; i < ulen; i++) | |
1954 | if (put_user(prog->aux->used_maps[i]->id, | |
1955 | &user_map_ids[i])) | |
1956 | return -EFAULT; | |
1957 | } | |
1e270976 MKL |
1958 | |
1959 | if (!capable(CAP_SYS_ADMIN)) { | |
1960 | info.jited_prog_len = 0; | |
1961 | info.xlated_prog_len = 0; | |
dbecd738 | 1962 | info.nr_jited_ksyms = 0; |
1e270976 MKL |
1963 | goto done; |
1964 | } | |
1965 | ||
1e270976 | 1966 | ulen = info.xlated_prog_len; |
9975a54b | 1967 | info.xlated_prog_len = bpf_prog_insn_size(prog); |
1e270976 | 1968 | if (info.xlated_prog_len && ulen) { |
7105e828 DB |
1969 | struct bpf_insn *insns_sanitized; |
1970 | bool fault; | |
1971 | ||
1972 | if (prog->blinded && !bpf_dump_raw_ok()) { | |
1973 | info.xlated_prog_insns = 0; | |
1974 | goto done; | |
1975 | } | |
1976 | insns_sanitized = bpf_insn_prepare_dump(prog); | |
1977 | if (!insns_sanitized) | |
1978 | return -ENOMEM; | |
1e270976 MKL |
1979 | uinsns = u64_to_user_ptr(info.xlated_prog_insns); |
1980 | ulen = min_t(u32, info.xlated_prog_len, ulen); | |
7105e828 DB |
1981 | fault = copy_to_user(uinsns, insns_sanitized, ulen); |
1982 | kfree(insns_sanitized); | |
1983 | if (fault) | |
1e270976 MKL |
1984 | return -EFAULT; |
1985 | } | |
1986 | ||
675fc275 JK |
1987 | if (bpf_prog_is_dev_bound(prog->aux)) { |
1988 | err = bpf_prog_offload_info_fill(&info, prog); | |
1989 | if (err) | |
1990 | return err; | |
fcfb126d JW |
1991 | goto done; |
1992 | } | |
1993 | ||
1994 | /* NOTE: the following code is supposed to be skipped for offload. | |
1995 | * bpf_prog_offload_info_fill() is the place to fill similar fields | |
1996 | * for offload. | |
1997 | */ | |
1998 | ulen = info.jited_prog_len; | |
4d56a76e SD |
1999 | if (prog->aux->func_cnt) { |
2000 | u32 i; | |
2001 | ||
2002 | info.jited_prog_len = 0; | |
2003 | for (i = 0; i < prog->aux->func_cnt; i++) | |
2004 | info.jited_prog_len += prog->aux->func[i]->jited_len; | |
2005 | } else { | |
2006 | info.jited_prog_len = prog->jited_len; | |
2007 | } | |
2008 | ||
fcfb126d JW |
2009 | if (info.jited_prog_len && ulen) { |
2010 | if (bpf_dump_raw_ok()) { | |
2011 | uinsns = u64_to_user_ptr(info.jited_prog_insns); | |
2012 | ulen = min_t(u32, info.jited_prog_len, ulen); | |
4d56a76e SD |
2013 | |
2014 | /* for multi-function programs, copy the JITed | |
2015 | * instructions for all the functions | |
2016 | */ | |
2017 | if (prog->aux->func_cnt) { | |
2018 | u32 len, free, i; | |
2019 | u8 *img; | |
2020 | ||
2021 | free = ulen; | |
2022 | for (i = 0; i < prog->aux->func_cnt; i++) { | |
2023 | len = prog->aux->func[i]->jited_len; | |
2024 | len = min_t(u32, len, free); | |
2025 | img = (u8 *) prog->aux->func[i]->bpf_func; | |
2026 | if (copy_to_user(uinsns, img, len)) | |
2027 | return -EFAULT; | |
2028 | uinsns += len; | |
2029 | free -= len; | |
2030 | if (!free) | |
2031 | break; | |
2032 | } | |
2033 | } else { | |
2034 | if (copy_to_user(uinsns, prog->bpf_func, ulen)) | |
2035 | return -EFAULT; | |
2036 | } | |
fcfb126d JW |
2037 | } else { |
2038 | info.jited_prog_insns = 0; | |
2039 | } | |
675fc275 JK |
2040 | } |
2041 | ||
dbecd738 SD |
2042 | ulen = info.nr_jited_ksyms; |
2043 | info.nr_jited_ksyms = prog->aux->func_cnt; | |
2044 | if (info.nr_jited_ksyms && ulen) { | |
2045 | if (bpf_dump_raw_ok()) { | |
2046 | u64 __user *user_ksyms; | |
2047 | ulong ksym_addr; | |
2048 | u32 i; | |
2049 | ||
2050 | /* copy the address of the kernel symbol | |
2051 | * corresponding to each function | |
2052 | */ | |
2053 | ulen = min_t(u32, info.nr_jited_ksyms, ulen); | |
2054 | user_ksyms = u64_to_user_ptr(info.jited_ksyms); | |
2055 | for (i = 0; i < ulen; i++) { | |
2056 | ksym_addr = (ulong) prog->aux->func[i]->bpf_func; | |
2057 | ksym_addr &= PAGE_MASK; | |
2058 | if (put_user((u64) ksym_addr, &user_ksyms[i])) | |
2059 | return -EFAULT; | |
2060 | } | |
2061 | } else { | |
2062 | info.jited_ksyms = 0; | |
2063 | } | |
2064 | } | |
2065 | ||
815581c1 SD |
2066 | ulen = info.nr_jited_func_lens; |
2067 | info.nr_jited_func_lens = prog->aux->func_cnt; | |
2068 | if (info.nr_jited_func_lens && ulen) { | |
2069 | if (bpf_dump_raw_ok()) { | |
2070 | u32 __user *user_lens; | |
2071 | u32 func_len, i; | |
2072 | ||
2073 | /* copy the JITed image lengths for each function */ | |
2074 | ulen = min_t(u32, info.nr_jited_func_lens, ulen); | |
2075 | user_lens = u64_to_user_ptr(info.jited_func_lens); | |
2076 | for (i = 0; i < ulen; i++) { | |
2077 | func_len = prog->aux->func[i]->jited_len; | |
2078 | if (put_user(func_len, &user_lens[i])) | |
2079 | return -EFAULT; | |
2080 | } | |
2081 | } else { | |
2082 | info.jited_func_lens = 0; | |
2083 | } | |
2084 | } | |
2085 | ||
1e270976 MKL |
2086 | done: |
2087 | if (copy_to_user(uinfo, &info, info_len) || | |
2088 | put_user(info_len, &uattr->info.info_len)) | |
2089 | return -EFAULT; | |
2090 | ||
2091 | return 0; | |
2092 | } | |
2093 | ||
2094 | static int bpf_map_get_info_by_fd(struct bpf_map *map, | |
2095 | const union bpf_attr *attr, | |
2096 | union bpf_attr __user *uattr) | |
2097 | { | |
2098 | struct bpf_map_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
2099 | struct bpf_map_info info = {}; | |
2100 | u32 info_len = attr->info.info_len; | |
2101 | int err; | |
2102 | ||
dcab51f1 | 2103 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(info), info_len); |
1e270976 MKL |
2104 | if (err) |
2105 | return err; | |
2106 | info_len = min_t(u32, sizeof(info), info_len); | |
2107 | ||
2108 | info.type = map->map_type; | |
2109 | info.id = map->id; | |
2110 | info.key_size = map->key_size; | |
2111 | info.value_size = map->value_size; | |
2112 | info.max_entries = map->max_entries; | |
2113 | info.map_flags = map->map_flags; | |
ad5b177b | 2114 | memcpy(info.name, map->name, sizeof(map->name)); |
1e270976 | 2115 | |
78958fca MKL |
2116 | if (map->btf) { |
2117 | info.btf_id = btf_id(map->btf); | |
9b2cf328 MKL |
2118 | info.btf_key_type_id = map->btf_key_type_id; |
2119 | info.btf_value_type_id = map->btf_value_type_id; | |
78958fca MKL |
2120 | } |
2121 | ||
52775b33 JK |
2122 | if (bpf_map_is_dev_bound(map)) { |
2123 | err = bpf_map_offload_info_fill(&info, map); | |
2124 | if (err) | |
2125 | return err; | |
2126 | } | |
2127 | ||
1e270976 MKL |
2128 | if (copy_to_user(uinfo, &info, info_len) || |
2129 | put_user(info_len, &uattr->info.info_len)) | |
2130 | return -EFAULT; | |
2131 | ||
2132 | return 0; | |
2133 | } | |
2134 | ||
62dab84c MKL |
2135 | static int bpf_btf_get_info_by_fd(struct btf *btf, |
2136 | const union bpf_attr *attr, | |
2137 | union bpf_attr __user *uattr) | |
2138 | { | |
2139 | struct bpf_btf_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
2140 | u32 info_len = attr->info.info_len; | |
2141 | int err; | |
2142 | ||
dcab51f1 | 2143 | err = bpf_check_uarg_tail_zero(uinfo, sizeof(*uinfo), info_len); |
62dab84c MKL |
2144 | if (err) |
2145 | return err; | |
2146 | ||
2147 | return btf_get_info_by_fd(btf, attr, uattr); | |
2148 | } | |
2149 | ||
1e270976 MKL |
2150 | #define BPF_OBJ_GET_INFO_BY_FD_LAST_FIELD info.info |
2151 | ||
2152 | static int bpf_obj_get_info_by_fd(const union bpf_attr *attr, | |
2153 | union bpf_attr __user *uattr) | |
2154 | { | |
2155 | int ufd = attr->info.bpf_fd; | |
2156 | struct fd f; | |
2157 | int err; | |
2158 | ||
2159 | if (CHECK_ATTR(BPF_OBJ_GET_INFO_BY_FD)) | |
2160 | return -EINVAL; | |
2161 | ||
2162 | f = fdget(ufd); | |
2163 | if (!f.file) | |
2164 | return -EBADFD; | |
2165 | ||
2166 | if (f.file->f_op == &bpf_prog_fops) | |
2167 | err = bpf_prog_get_info_by_fd(f.file->private_data, attr, | |
2168 | uattr); | |
2169 | else if (f.file->f_op == &bpf_map_fops) | |
2170 | err = bpf_map_get_info_by_fd(f.file->private_data, attr, | |
2171 | uattr); | |
60197cfb | 2172 | else if (f.file->f_op == &btf_fops) |
62dab84c | 2173 | err = bpf_btf_get_info_by_fd(f.file->private_data, attr, uattr); |
1e270976 MKL |
2174 | else |
2175 | err = -EINVAL; | |
2176 | ||
2177 | fdput(f); | |
2178 | return err; | |
2179 | } | |
2180 | ||
f56a653c MKL |
2181 | #define BPF_BTF_LOAD_LAST_FIELD btf_log_level |
2182 | ||
2183 | static int bpf_btf_load(const union bpf_attr *attr) | |
2184 | { | |
2185 | if (CHECK_ATTR(BPF_BTF_LOAD)) | |
2186 | return -EINVAL; | |
2187 | ||
2188 | if (!capable(CAP_SYS_ADMIN)) | |
2189 | return -EPERM; | |
2190 | ||
2191 | return btf_new_fd(attr); | |
2192 | } | |
2193 | ||
78958fca MKL |
2194 | #define BPF_BTF_GET_FD_BY_ID_LAST_FIELD btf_id |
2195 | ||
2196 | static int bpf_btf_get_fd_by_id(const union bpf_attr *attr) | |
2197 | { | |
2198 | if (CHECK_ATTR(BPF_BTF_GET_FD_BY_ID)) | |
2199 | return -EINVAL; | |
2200 | ||
2201 | if (!capable(CAP_SYS_ADMIN)) | |
2202 | return -EPERM; | |
2203 | ||
2204 | return btf_get_fd_by_id(attr->btf_id); | |
2205 | } | |
2206 | ||
41bdc4b4 YS |
2207 | static int bpf_task_fd_query_copy(const union bpf_attr *attr, |
2208 | union bpf_attr __user *uattr, | |
2209 | u32 prog_id, u32 fd_type, | |
2210 | const char *buf, u64 probe_offset, | |
2211 | u64 probe_addr) | |
2212 | { | |
2213 | char __user *ubuf = u64_to_user_ptr(attr->task_fd_query.buf); | |
2214 | u32 len = buf ? strlen(buf) : 0, input_len; | |
2215 | int err = 0; | |
2216 | ||
2217 | if (put_user(len, &uattr->task_fd_query.buf_len)) | |
2218 | return -EFAULT; | |
2219 | input_len = attr->task_fd_query.buf_len; | |
2220 | if (input_len && ubuf) { | |
2221 | if (!len) { | |
2222 | /* nothing to copy, just make ubuf NULL terminated */ | |
2223 | char zero = '\0'; | |
2224 | ||
2225 | if (put_user(zero, ubuf)) | |
2226 | return -EFAULT; | |
2227 | } else if (input_len >= len + 1) { | |
2228 | /* ubuf can hold the string with NULL terminator */ | |
2229 | if (copy_to_user(ubuf, buf, len + 1)) | |
2230 | return -EFAULT; | |
2231 | } else { | |
2232 | /* ubuf cannot hold the string with NULL terminator, | |
2233 | * do a partial copy with NULL terminator. | |
2234 | */ | |
2235 | char zero = '\0'; | |
2236 | ||
2237 | err = -ENOSPC; | |
2238 | if (copy_to_user(ubuf, buf, input_len - 1)) | |
2239 | return -EFAULT; | |
2240 | if (put_user(zero, ubuf + input_len - 1)) | |
2241 | return -EFAULT; | |
2242 | } | |
2243 | } | |
2244 | ||
2245 | if (put_user(prog_id, &uattr->task_fd_query.prog_id) || | |
2246 | put_user(fd_type, &uattr->task_fd_query.fd_type) || | |
2247 | put_user(probe_offset, &uattr->task_fd_query.probe_offset) || | |
2248 | put_user(probe_addr, &uattr->task_fd_query.probe_addr)) | |
2249 | return -EFAULT; | |
2250 | ||
2251 | return err; | |
2252 | } | |
2253 | ||
2254 | #define BPF_TASK_FD_QUERY_LAST_FIELD task_fd_query.probe_addr | |
2255 | ||
2256 | static int bpf_task_fd_query(const union bpf_attr *attr, | |
2257 | union bpf_attr __user *uattr) | |
2258 | { | |
2259 | pid_t pid = attr->task_fd_query.pid; | |
2260 | u32 fd = attr->task_fd_query.fd; | |
2261 | const struct perf_event *event; | |
2262 | struct files_struct *files; | |
2263 | struct task_struct *task; | |
2264 | struct file *file; | |
2265 | int err; | |
2266 | ||
2267 | if (CHECK_ATTR(BPF_TASK_FD_QUERY)) | |
2268 | return -EINVAL; | |
2269 | ||
2270 | if (!capable(CAP_SYS_ADMIN)) | |
2271 | return -EPERM; | |
2272 | ||
2273 | if (attr->task_fd_query.flags != 0) | |
2274 | return -EINVAL; | |
2275 | ||
2276 | task = get_pid_task(find_vpid(pid), PIDTYPE_PID); | |
2277 | if (!task) | |
2278 | return -ENOENT; | |
2279 | ||
2280 | files = get_files_struct(task); | |
2281 | put_task_struct(task); | |
2282 | if (!files) | |
2283 | return -ENOENT; | |
2284 | ||
2285 | err = 0; | |
2286 | spin_lock(&files->file_lock); | |
2287 | file = fcheck_files(files, fd); | |
2288 | if (!file) | |
2289 | err = -EBADF; | |
2290 | else | |
2291 | get_file(file); | |
2292 | spin_unlock(&files->file_lock); | |
2293 | put_files_struct(files); | |
2294 | ||
2295 | if (err) | |
2296 | goto out; | |
2297 | ||
2298 | if (file->f_op == &bpf_raw_tp_fops) { | |
2299 | struct bpf_raw_tracepoint *raw_tp = file->private_data; | |
2300 | struct bpf_raw_event_map *btp = raw_tp->btp; | |
2301 | ||
2302 | err = bpf_task_fd_query_copy(attr, uattr, | |
2303 | raw_tp->prog->aux->id, | |
2304 | BPF_FD_TYPE_RAW_TRACEPOINT, | |
2305 | btp->tp->name, 0, 0); | |
2306 | goto put_file; | |
2307 | } | |
2308 | ||
2309 | event = perf_get_event(file); | |
2310 | if (!IS_ERR(event)) { | |
2311 | u64 probe_offset, probe_addr; | |
2312 | u32 prog_id, fd_type; | |
2313 | const char *buf; | |
2314 | ||
2315 | err = bpf_get_perf_event_info(event, &prog_id, &fd_type, | |
2316 | &buf, &probe_offset, | |
2317 | &probe_addr); | |
2318 | if (!err) | |
2319 | err = bpf_task_fd_query_copy(attr, uattr, prog_id, | |
2320 | fd_type, buf, | |
2321 | probe_offset, | |
2322 | probe_addr); | |
2323 | goto put_file; | |
2324 | } | |
2325 | ||
2326 | err = -ENOTSUPP; | |
2327 | put_file: | |
2328 | fput(file); | |
2329 | out: | |
2330 | return err; | |
2331 | } | |
2332 | ||
99c55f7d AS |
2333 | SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size) |
2334 | { | |
2335 | union bpf_attr attr = {}; | |
2336 | int err; | |
2337 | ||
0fa4fe85 | 2338 | if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN)) |
99c55f7d AS |
2339 | return -EPERM; |
2340 | ||
dcab51f1 | 2341 | err = bpf_check_uarg_tail_zero(uattr, sizeof(attr), size); |
1e270976 MKL |
2342 | if (err) |
2343 | return err; | |
2344 | size = min_t(u32, size, sizeof(attr)); | |
99c55f7d AS |
2345 | |
2346 | /* copy attributes from user space, may be less than sizeof(bpf_attr) */ | |
2347 | if (copy_from_user(&attr, uattr, size) != 0) | |
2348 | return -EFAULT; | |
2349 | ||
afdb09c7 CF |
2350 | err = security_bpf(cmd, &attr, size); |
2351 | if (err < 0) | |
2352 | return err; | |
2353 | ||
99c55f7d AS |
2354 | switch (cmd) { |
2355 | case BPF_MAP_CREATE: | |
2356 | err = map_create(&attr); | |
2357 | break; | |
db20fd2b AS |
2358 | case BPF_MAP_LOOKUP_ELEM: |
2359 | err = map_lookup_elem(&attr); | |
2360 | break; | |
2361 | case BPF_MAP_UPDATE_ELEM: | |
2362 | err = map_update_elem(&attr); | |
2363 | break; | |
2364 | case BPF_MAP_DELETE_ELEM: | |
2365 | err = map_delete_elem(&attr); | |
2366 | break; | |
2367 | case BPF_MAP_GET_NEXT_KEY: | |
2368 | err = map_get_next_key(&attr); | |
2369 | break; | |
09756af4 AS |
2370 | case BPF_PROG_LOAD: |
2371 | err = bpf_prog_load(&attr); | |
2372 | break; | |
b2197755 DB |
2373 | case BPF_OBJ_PIN: |
2374 | err = bpf_obj_pin(&attr); | |
2375 | break; | |
2376 | case BPF_OBJ_GET: | |
2377 | err = bpf_obj_get(&attr); | |
2378 | break; | |
f4324551 DM |
2379 | case BPF_PROG_ATTACH: |
2380 | err = bpf_prog_attach(&attr); | |
2381 | break; | |
2382 | case BPF_PROG_DETACH: | |
2383 | err = bpf_prog_detach(&attr); | |
2384 | break; | |
468e2f64 AS |
2385 | case BPF_PROG_QUERY: |
2386 | err = bpf_prog_query(&attr, uattr); | |
2387 | break; | |
1cf1cae9 AS |
2388 | case BPF_PROG_TEST_RUN: |
2389 | err = bpf_prog_test_run(&attr, uattr); | |
2390 | break; | |
34ad5580 MKL |
2391 | case BPF_PROG_GET_NEXT_ID: |
2392 | err = bpf_obj_get_next_id(&attr, uattr, | |
2393 | &prog_idr, &prog_idr_lock); | |
2394 | break; | |
2395 | case BPF_MAP_GET_NEXT_ID: | |
2396 | err = bpf_obj_get_next_id(&attr, uattr, | |
2397 | &map_idr, &map_idr_lock); | |
2398 | break; | |
b16d9aa4 MKL |
2399 | case BPF_PROG_GET_FD_BY_ID: |
2400 | err = bpf_prog_get_fd_by_id(&attr); | |
2401 | break; | |
bd5f5f4e MKL |
2402 | case BPF_MAP_GET_FD_BY_ID: |
2403 | err = bpf_map_get_fd_by_id(&attr); | |
2404 | break; | |
1e270976 MKL |
2405 | case BPF_OBJ_GET_INFO_BY_FD: |
2406 | err = bpf_obj_get_info_by_fd(&attr, uattr); | |
2407 | break; | |
c4f6699d AS |
2408 | case BPF_RAW_TRACEPOINT_OPEN: |
2409 | err = bpf_raw_tracepoint_open(&attr); | |
2410 | break; | |
f56a653c MKL |
2411 | case BPF_BTF_LOAD: |
2412 | err = bpf_btf_load(&attr); | |
2413 | break; | |
78958fca MKL |
2414 | case BPF_BTF_GET_FD_BY_ID: |
2415 | err = bpf_btf_get_fd_by_id(&attr); | |
2416 | break; | |
41bdc4b4 YS |
2417 | case BPF_TASK_FD_QUERY: |
2418 | err = bpf_task_fd_query(&attr, uattr); | |
2419 | break; | |
99c55f7d AS |
2420 | default: |
2421 | err = -EINVAL; | |
2422 | break; | |
2423 | } | |
2424 | ||
2425 | return err; | |
2426 | } |