]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - kernel/bpf/syscall.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'pci-v5.15-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
[mirror_ubuntu-jammy-kernel.git] / kernel / bpf / syscall.c
CommitLineData
5b497af4 1// SPDX-License-Identifier: GPL-2.0-only
99c55f7d 2/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
99c55f7d
AS		 3 */
4#include <linux/bpf.h>
a67edbf4 5#include <linux/bpf_trace.h>
f4364dcf 6#include <linux/bpf_lirc.h>
4a1e7c0c 7#include <linux/bpf_verifier.h>
f56a653c 8#include <linux/btf.h>
99c55f7d
AS		 9#include <linux/syscalls.h>
10#include <linux/slab.h>
3f07c014 11#include <linux/sched/signal.h>
d407bd25
DB		 12#include <linux/vmalloc.h>
13#include <linux/mmzone.h>
99c55f7d 14#include <linux/anon_inodes.h>
41bdc4b4 15#include <linux/fdtable.h>
db20fd2b 16#include <linux/file.h>
41bdc4b4 17#include <linux/fs.h>
09756af4
AS		 18#include <linux/license.h>
19#include <linux/filter.h>
535e7b4b 20#include <linux/kernel.h>
dc4bb0e2 21#include <linux/idr.h>
cb4d2b3f
MKL		 22#include <linux/cred.h>
23#include <linux/timekeeping.h>
24#include <linux/ctype.h>
9ef09e35 25#include <linux/nospec.h>
bae141f5 26#include <linux/audit.h>
ccfe29eb 27#include <uapi/linux/btf.h>
ca5999fd 28#include <linux/pgtable.h>
9e4e01df 29#include <linux/bpf_lsm.h>
457f4436 30#include <linux/poll.h>
a3fd7cee 31#include <linux/bpf-netns.h>
1e6c62a8 32#include <linux/rcupdate_trace.h>
48edc1f7 33#include <linux/memcontrol.h>
99c55f7d 34
da765a2f
DB		 35#define IS_FD_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || \
36 (map)->map_type == BPF_MAP_TYPE_CGROUP_ARRAY || \
37 (map)->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS)
38#define IS_FD_PROG_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PROG_ARRAY)
14dc6f04 39#define IS_FD_HASH(map) ((map)->map_type == BPF_MAP_TYPE_HASH_OF_MAPS)
da765a2f
DB		 40#define IS_FD_MAP(map) (IS_FD_ARRAY(map) || IS_FD_PROG_ARRAY(map) || \
41 IS_FD_HASH(map))
14dc6f04 42
6e71b04a
CF		 43#define BPF_OBJ_FLAG_MASK (BPF_F_RDONLY | BPF_F_WRONLY)
44
b121d1e7 45DEFINE_PER_CPU(int, bpf_prog_active);
dc4bb0e2
MKL		 46static DEFINE_IDR(prog_idr);
47static DEFINE_SPINLOCK(prog_idr_lock);
f3f1c054
MKL		 48static DEFINE_IDR(map_idr);
49static DEFINE_SPINLOCK(map_idr_lock);
a3b80e10
AN		 50static DEFINE_IDR(link_idr);
51static DEFINE_SPINLOCK(link_idr_lock);
b121d1e7 52
08389d88
DB		 53int sysctl_unprivileged_bpf_disabled __read_mostly =
54 IS_BUILTIN(CONFIG_BPF_UNPRIV_DEFAULT_OFF) ? 2 : 0;
1be7f75d 55
40077e0c 56static const struct bpf_map_ops * const bpf_map_types[] = {
91cc1a99 57#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type)
40077e0c
JB		 58#define BPF_MAP_TYPE(_id, _ops) \
59 [_id] = &_ops,
f2e10bff 60#define BPF_LINK_TYPE(_id, _name)
40077e0c
JB		 61#include <linux/bpf_types.h>
62#undef BPF_PROG_TYPE
63#undef BPF_MAP_TYPE
f2e10bff 64#undef BPF_LINK_TYPE
40077e0c 65};
99c55f7d 66
752ba56f
MS		 67/*
68 * If we're handed a bigger struct than we know of, ensure all the unknown bits
69 * are 0 - i.e. new user-space does not rely on any kernel feature extensions
70 * we don't know about yet.
71 *
72 * There is a ToCToU between this function call and the following
73 * copy_from_user() call. However, this is not a concern since this function is
74 * meant to be a future-proofing of bits.
75 */
af2ac3e1 76int bpf_check_uarg_tail_zero(bpfptr_t uaddr,
dcab51f1
MKL		 77 size_t expected_size,
78 size_t actual_size)
58291a74 79{
b7e4b65f 80 int res;
58291a74 81
752ba56f
MS		 82 if (unlikely(actual_size > PAGE_SIZE)) /* silly large */
83 return -E2BIG;
84
58291a74
MS		 85 if (actual_size <= expected_size)
86 return 0;
87
af2ac3e1
AS		 88 if (uaddr.is_kernel)
89 res = memchr_inv(uaddr.kernel + expected_size, 0,
90 actual_size - expected_size) == NULL;
91 else
92 res = check_zeroed_user(uaddr.user + expected_size,
93 actual_size - expected_size);
b7e4b65f
AV		 94 if (res < 0)
95 return res;
96 return res ? 0 : -E2BIG;
58291a74
MS		 97}
98
a3884572 99const struct bpf_map_ops bpf_map_offload_ops = {
f4d05259 100 .map_meta_equal = bpf_map_meta_equal,
a3884572
JK		 101 .map_alloc = bpf_map_offload_map_alloc,
102 .map_free = bpf_map_offload_map_free,
e8d2bec0 103 .map_check_btf = map_check_no_btf,
a3884572
JK		 104};
105
99c55f7d
AS		 106static struct bpf_map *find_and_alloc_map(union bpf_attr *attr)
107{
1110f3a9 108 const struct bpf_map_ops *ops;
9ef09e35 109 u32 type = attr->map_type;
99c55f7d 110 struct bpf_map *map;
1110f3a9 111 int err;
99c55f7d 112
9ef09e35 113 if (type >= ARRAY_SIZE(bpf_map_types))
1110f3a9 114 return ERR_PTR(-EINVAL);
9ef09e35
MR		 115 type = array_index_nospec(type, ARRAY_SIZE(bpf_map_types));
116 ops = bpf_map_types[type];
1110f3a9 117 if (!ops)
40077e0c 118 return ERR_PTR(-EINVAL);
99c55f7d 119
1110f3a9
JK		 120 if (ops->map_alloc_check) {
121 err = ops->map_alloc_check(attr);
122 if (err)
123 return ERR_PTR(err);
124 }
a3884572
JK		 125 if (attr->map_ifindex)
126 ops = &bpf_map_offload_ops;
1110f3a9 127 map = ops->map_alloc(attr);
40077e0c
JB		 128 if (IS_ERR(map))
129 return map;
1110f3a9 130 map->ops = ops;
9ef09e35 131 map->map_type = type;
40077e0c 132 return map;
99c55f7d
AS		 133}
134
80ee81e0 135static u32 bpf_map_value_size(const struct bpf_map *map)
15c14a3d
BV		 136{
137 if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
138 map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH ||
139 map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY ||
140 map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE)
141 return round_up(map->value_size, 8) * num_possible_cpus();
142 else if (IS_FD_MAP(map))
143 return sizeof(u32);
144 else
145 return map->value_size;
146}
147
148static void maybe_wait_bpf_programs(struct bpf_map *map)
149{
150 /* Wait for any running BPF programs to complete so that
151 * userspace, when we return to it, knows that all programs
152 * that could be running use the new map value.
153 */
154 if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS ||
155 map->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS)
156 synchronize_rcu();
157}
158
159static int bpf_map_update_value(struct bpf_map *map, struct fd f, void *key,
160 void *value, __u64 flags)
161{
162 int err;
163
164 /* Need to create a kthread, thus must support schedule */
165 if (bpf_map_is_dev_bound(map)) {
166 return bpf_map_offload_update_elem(map, key, value, flags);
167 } else if (map->map_type == BPF_MAP_TYPE_CPUMAP ||
15c14a3d
BV		 168 map->map_type == BPF_MAP_TYPE_STRUCT_OPS) {
169 return map->ops->map_update_elem(map, key, value, flags);
13b79d3f
LB		 170 } else if (map->map_type == BPF_MAP_TYPE_SOCKHASH ||
171 map->map_type == BPF_MAP_TYPE_SOCKMAP) {
172 return sock_map_update_elem_sys(map, key, value, flags);
15c14a3d
BV		 173 } else if (IS_FD_PROG_ARRAY(map)) {
174 return bpf_fd_array_map_update_elem(map, f.file, key, value,
175 flags);
176 }
177
b6e5dae1 178 bpf_disable_instrumentation();
15c14a3d
BV		 179 if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
180 map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) {
181 err = bpf_percpu_hash_update(map, key, value, flags);
182 } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) {
183 err = bpf_percpu_array_update(map, key, value, flags);
184 } else if (map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE) {
185 err = bpf_percpu_cgroup_storage_update(map, key, value,
186 flags);
187 } else if (IS_FD_ARRAY(map)) {
188 rcu_read_lock();
189 err = bpf_fd_array_map_update_elem(map, f.file, key, value,
190 flags);
191 rcu_read_unlock();
192 } else if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) {
193 rcu_read_lock();
194 err = bpf_fd_htab_map_update_elem(map, f.file, key, value,
195 flags);
196 rcu_read_unlock();
197 } else if (map->map_type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY) {
198 /* rcu_read_lock() is not needed */
199 err = bpf_fd_reuseport_array_update_elem(map, key, value,
200 flags);
201 } else if (map->map_type == BPF_MAP_TYPE_QUEUE ||
202 map->map_type == BPF_MAP_TYPE_STACK) {
203 err = map->ops->map_push_elem(map, value, flags);
204 } else {
205 rcu_read_lock();
206 err = map->ops->map_update_elem(map, key, value, flags);
207 rcu_read_unlock();
208 }
b6e5dae1 209 bpf_enable_instrumentation();
15c14a3d
BV		 210 maybe_wait_bpf_programs(map);
211
212 return err;
213}
214
215static int bpf_map_copy_value(struct bpf_map *map, void *key, void *value,
216 __u64 flags)
217{
218 void *ptr;
219 int err;
220
cb4d03ab
BV		 221 if (bpf_map_is_dev_bound(map))
222 return bpf_map_offload_lookup_elem(map, key, value);
15c14a3d 223
b6e5dae1 224 bpf_disable_instrumentation();
15c14a3d
BV		 225 if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
226 map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) {
227 err = bpf_percpu_hash_copy(map, key, value);
228 } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) {
229 err = bpf_percpu_array_copy(map, key, value);
230 } else if (map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE) {
231 err = bpf_percpu_cgroup_storage_copy(map, key, value);
232 } else if (map->map_type == BPF_MAP_TYPE_STACK_TRACE) {
233 err = bpf_stackmap_copy(map, key, value);
234 } else if (IS_FD_ARRAY(map) || IS_FD_PROG_ARRAY(map)) {
235 err = bpf_fd_array_map_lookup_elem(map, key, value);
236 } else if (IS_FD_HASH(map)) {
237 err = bpf_fd_htab_map_lookup_elem(map, key, value);
238 } else if (map->map_type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY) {
239 err = bpf_fd_reuseport_array_lookup_elem(map, key, value);
240 } else if (map->map_type == BPF_MAP_TYPE_QUEUE ||
241 map->map_type == BPF_MAP_TYPE_STACK) {
242 err = map->ops->map_peek_elem(map, value);
243 } else if (map->map_type == BPF_MAP_TYPE_STRUCT_OPS) {
244 /* struct_ops map requires directly updating "value" */
245 err = bpf_struct_ops_map_sys_lookup_elem(map, key, value);
246 } else {
247 rcu_read_lock();
248 if (map->ops->map_lookup_elem_sys_only)
249 ptr = map->ops->map_lookup_elem_sys_only(map, key);
250 else
251 ptr = map->ops->map_lookup_elem(map, key);
252 if (IS_ERR(ptr)) {
253 err = PTR_ERR(ptr);
254 } else if (!ptr) {
255 err = -ENOENT;
256 } else {
257 err = 0;
258 if (flags & BPF_F_LOCK)
259 /* lock 'ptr' and copy everything but lock */
260 copy_map_value_locked(map, value, ptr, true);
261 else
262 copy_map_value(map, value, ptr);
68134668
AS		 263 /* mask lock and timer, since value wasn't zero inited */
264 check_and_init_map_value(map, value);
15c14a3d
BV		 265 }
266 rcu_read_unlock();
267 }
268
b6e5dae1 269 bpf_enable_instrumentation();
15c14a3d
BV		 270 maybe_wait_bpf_programs(map);
271
272 return err;
273}
274
d5299b67
RG		 275/* Please, do not use this function outside from the map creation path
276 * (e.g. in map update path) without taking care of setting the active
277 * memory cgroup (see at bpf_map_kmalloc_node() for example).
278 */
196e8ca7 279static void *__bpf_map_area_alloc(u64 size, int numa_node, bool mmapable)
d407bd25 280{
f01a7dbe
MP		 281 /* We really just want to fail instead of triggering OOM killer
282 * under memory pressure, therefore we set __GFP_NORETRY to kmalloc,
283 * which is used for lower order allocation requests.
284 *
285 * It has been observed that higher order allocation requests done by
286 * vmalloc with __GFP_NORETRY being set might fail due to not trying
287 * to reclaim memory from the page cache, thus we set
288 * __GFP_RETRY_MAYFAIL to avoid such situations.
d407bd25 289 */
f01a7dbe 290
d5299b67 291 const gfp_t gfp = __GFP_NOWARN | __GFP_ZERO | __GFP_ACCOUNT;
041de93f
CH		 292 unsigned int flags = 0;
293 unsigned long align = 1;
d407bd25
DB		 294 void *area;
295
196e8ca7
DB		 296 if (size >= SIZE_MAX)
297 return NULL;
298
fc970227 299 /* kmalloc()'ed memory can't be mmap()'ed */
041de93f
CH		 300 if (mmapable) {
301 BUG_ON(!PAGE_ALIGNED(size));
302 align = SHMLBA;
303 flags = VM_USERMAP;
304 } else if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
305 area = kmalloc_node(size, gfp | GFP_USER | __GFP_NORETRY,
f01a7dbe 306 numa_node);
d407bd25
DB		 307 if (area != NULL)
308 return area;
309 }
041de93f
CH		 310
311 return __vmalloc_node_range(size, align, VMALLOC_START, VMALLOC_END,
312 gfp | GFP_KERNEL | __GFP_RETRY_MAYFAIL, PAGE_KERNEL,
313 flags, numa_node, __builtin_return_address(0));
d407bd25
DB		 314}
315
196e8ca7 316void *bpf_map_area_alloc(u64 size, int numa_node)
fc970227
AN		 317{
318 return __bpf_map_area_alloc(size, numa_node, false);
319}
320
196e8ca7 321void *bpf_map_area_mmapable_alloc(u64 size, int numa_node)
fc970227
AN		 322{
323 return __bpf_map_area_alloc(size, numa_node, true);
324}
325
d407bd25
DB		 326void bpf_map_area_free(void *area)
327{
328 kvfree(area);
329}
330
be70bcd5
DB		 331static u32 bpf_map_flags_retain_permanent(u32 flags)
332{
333 /* Some map creation flags are not tied to the map object but
334 * rather to the map fd instead, so they have no meaning upon
335 * map object inspection since multiple file descriptors with
336 * different (access) properties can exist here. Thus, given
337 * this has zero meaning for the map itself, lets clear these
338 * from here.
339 */
340 return flags & ~(BPF_F_RDONLY | BPF_F_WRONLY);
341}
342
bd475643
JK		 343void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr)
344{
345 map->map_type = attr->map_type;
346 map->key_size = attr->key_size;
347 map->value_size = attr->value_size;
348 map->max_entries = attr->max_entries;
be70bcd5 349 map->map_flags = bpf_map_flags_retain_permanent(attr->map_flags);
bd475643
JK		 350 map->numa_node = bpf_map_attr_numa_node(attr);
351}
352
f3f1c054
MKL		 353static int bpf_map_alloc_id(struct bpf_map *map)
354{
355 int id;
356
b76354cd 357 idr_preload(GFP_KERNEL);
f3f1c054
MKL		 358 spin_lock_bh(&map_idr_lock);
359 id = idr_alloc_cyclic(&map_idr, map, 1, INT_MAX, GFP_ATOMIC);
360 if (id > 0)
361 map->id = id;
362 spin_unlock_bh(&map_idr_lock);
b76354cd 363 idr_preload_end();
f3f1c054
MKL		 364
365 if (WARN_ON_ONCE(!id))
366 return -ENOSPC;
367
368 return id > 0 ? 0 : id;
369}
370
a3884572 371void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock)
f3f1c054 372{
930651a7
ED		 373 unsigned long flags;
374
a3884572
JK		 375 /* Offloaded maps are removed from the IDR store when their device
376 * disappears - even if someone holds an fd to them they are unusable,
377 * the memory is gone, all ops will fail; they are simply waiting for
378 * refcnt to drop to be freed.
379 */
380 if (!map->id)
381 return;
382
bd5f5f4e 383 if (do_idr_lock)
930651a7 384 spin_lock_irqsave(&map_idr_lock, flags);
bd5f5f4e
MKL		 385 else
386 __acquire(&map_idr_lock);
387
f3f1c054 388 idr_remove(&map_idr, map->id);
a3884572 389 map->id = 0;
bd5f5f4e
MKL		 390
391 if (do_idr_lock)
930651a7 392 spin_unlock_irqrestore(&map_idr_lock, flags);
bd5f5f4e
MKL		 393 else
394 __release(&map_idr_lock);
f3f1c054
MKL		 395}
396
48edc1f7
RG		 397#ifdef CONFIG_MEMCG_KMEM
398static void bpf_map_save_memcg(struct bpf_map *map)
399{
400 map->memcg = get_mem_cgroup_from_mm(current->mm);
401}
402
403static void bpf_map_release_memcg(struct bpf_map *map)
404{
405 mem_cgroup_put(map->memcg);
406}
407
408void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
409 int node)
410{
411 struct mem_cgroup *old_memcg;
412 void *ptr;
413
414 old_memcg = set_active_memcg(map->memcg);
415 ptr = kmalloc_node(size, flags | __GFP_ACCOUNT, node);
416 set_active_memcg(old_memcg);
417
418 return ptr;
419}
420
421void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags)
422{
423 struct mem_cgroup *old_memcg;
424 void *ptr;
425
426 old_memcg = set_active_memcg(map->memcg);
427 ptr = kzalloc(size, flags | __GFP_ACCOUNT);
428 set_active_memcg(old_memcg);
429
430 return ptr;
431}
432
433void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
434 size_t align, gfp_t flags)
435{
436 struct mem_cgroup *old_memcg;
437 void __percpu *ptr;
438
439 old_memcg = set_active_memcg(map->memcg);
440 ptr = __alloc_percpu_gfp(size, align, flags | __GFP_ACCOUNT);
441 set_active_memcg(old_memcg);
442
443 return ptr;
444}
445
446#else
447static void bpf_map_save_memcg(struct bpf_map *map)
448{
449}
450
451static void bpf_map_release_memcg(struct bpf_map *map)
452{
453}
454#endif
455
99c55f7d
AS		 456/* called from workqueue */
457static void bpf_map_free_deferred(struct work_struct *work)
458{
459 struct bpf_map *map = container_of(work, struct bpf_map, work);
460
afdb09c7 461 security_bpf_map_free(map);
48edc1f7 462 bpf_map_release_memcg(map);
99c55f7d
AS		 463 /* implementation dependent freeing */
464 map->ops->map_free(map);
465}
466
c9da161c
DB		 467static void bpf_map_put_uref(struct bpf_map *map)
468{
1e0bd5a0 469 if (atomic64_dec_and_test(&map->usercnt)) {
ba6b8de4
JF		 470 if (map->ops->map_release_uref)
471 map->ops->map_release_uref(map);
c9da161c
DB		 472 }
473}
474
99c55f7d
AS		 475/* decrement map refcnt and schedule it for freeing via workqueue
476 * (unrelying map implementation ops->map_free() might sleep)
477 */
bd5f5f4e 478static void __bpf_map_put(struct bpf_map *map, bool do_idr_lock)
99c55f7d 479{
1e0bd5a0 480 if (atomic64_dec_and_test(&map->refcnt)) {
34ad5580 481 /* bpf_map_free_id() must be called first */
bd5f5f4e 482 bpf_map_free_id(map, do_idr_lock);
78958fca 483 btf_put(map->btf);
99c55f7d
AS		 484 INIT_WORK(&map->work, bpf_map_free_deferred);
485 schedule_work(&map->work);
486 }
487}
488
bd5f5f4e
MKL		 489void bpf_map_put(struct bpf_map *map)
490{
491 __bpf_map_put(map, true);
492}
630a4d38 493EXPORT_SYMBOL_GPL(bpf_map_put);
bd5f5f4e 494
c9da161c 495void bpf_map_put_with_uref(struct bpf_map *map)
99c55f7d 496{
c9da161c 497 bpf_map_put_uref(map);
99c55f7d 498 bpf_map_put(map);
c9da161c
DB		 499}
500
501static int bpf_map_release(struct inode *inode, struct file *filp)
502{
61d1b6a4
DB		 503 struct bpf_map *map = filp->private_data;
504
505 if (map->ops->map_release)
506 map->ops->map_release(map, filp);
507
508 bpf_map_put_with_uref(map);
99c55f7d
AS		 509 return 0;
510}
511
87df15de
DB		 512static fmode_t map_get_sys_perms(struct bpf_map *map, struct fd f)
513{
514 fmode_t mode = f.file->f_mode;
515
516 /* Our file permissions may have been overridden by global
517 * map permissions facing syscall side.
518 */
519 if (READ_ONCE(map->frozen))
520 mode &= ~FMODE_CAN_WRITE;
521 return mode;
522}
523
f99bf205 524#ifdef CONFIG_PROC_FS
80ee81e0
RG		 525/* Provides an approximation of the map's memory footprint.
526 * Used only to provide a backward compatibility and display
527 * a reasonable "memlock" info.
528 */
529static unsigned long bpf_map_memory_footprint(const struct bpf_map *map)
530{
531 unsigned long size;
532
533 size = round_up(map->key_size + bpf_map_value_size(map), 8);
534
535 return round_up(map->max_entries * size, PAGE_SIZE);
536}
537
f99bf205
DB		 538static void bpf_map_show_fdinfo(struct seq_file *m, struct file *filp)
539{
540 const struct bpf_map *map = filp->private_data;
21116b70 541 const struct bpf_array *array;
2beee5f5 542 u32 type = 0, jited = 0;
21116b70
DB		 543
544 if (map->map_type == BPF_MAP_TYPE_PROG_ARRAY) {
545 array = container_of(map, struct bpf_array, map);
2beee5f5
DB		 546 type = array->aux->type;
547 jited = array->aux->jited;
21116b70 548 }
f99bf205
DB		 549
550 seq_printf(m,
551 "map_type:\t%u\n"
552 "key_size:\t%u\n"
553 "value_size:\t%u\n"
322cea2f 554 "max_entries:\t%u\n"
21116b70 555 "map_flags:\t%#x\n"
80ee81e0 556 "memlock:\t%lu\n"
87df15de
DB		 557 "map_id:\t%u\n"
558 "frozen:\t%u\n",
f99bf205
DB		 559 map->map_type,
560 map->key_size,
561 map->value_size,
322cea2f 562 map->max_entries,
21116b70 563 map->map_flags,
80ee81e0 564 bpf_map_memory_footprint(map),
87df15de
DB		 565 map->id,
566 READ_ONCE(map->frozen));
2beee5f5
DB		 567 if (type) {
568 seq_printf(m, "owner_prog_type:\t%u\n", type);
569 seq_printf(m, "owner_jited:\t%u\n", jited);
9780c0ab 570 }
f99bf205
DB		 571}
572#endif
573
6e71b04a
CF		 574static ssize_t bpf_dummy_read(struct file *filp, char __user *buf, size_t siz,
575 loff_t *ppos)
576{
577 /* We need this handler such that alloc_file() enables
578 * f_mode with FMODE_CAN_READ.
579 */
580 return -EINVAL;
581}
582
583static ssize_t bpf_dummy_write(struct file *filp, const char __user *buf,
584 size_t siz, loff_t *ppos)
585{
586 /* We need this handler such that alloc_file() enables
587 * f_mode with FMODE_CAN_WRITE.
588 */
589 return -EINVAL;
590}
591
fc970227
AN		 592/* called for any extra memory-mapped regions (except initial) */
593static void bpf_map_mmap_open(struct vm_area_struct *vma)
594{
595 struct bpf_map *map = vma->vm_file->private_data;
596
1f6cb19b 597 if (vma->vm_flags & VM_MAYWRITE) {
fc970227
AN		 598 mutex_lock(&map->freeze_mutex);
599 map->writecnt++;
600 mutex_unlock(&map->freeze_mutex);
601 }
602}
603
604/* called for all unmapped memory region (including initial) */
605static void bpf_map_mmap_close(struct vm_area_struct *vma)
606{
607 struct bpf_map *map = vma->vm_file->private_data;
608
1f6cb19b 609 if (vma->vm_flags & VM_MAYWRITE) {
fc970227
AN		 610 mutex_lock(&map->freeze_mutex);
611 map->writecnt--;
612 mutex_unlock(&map->freeze_mutex);
613 }
fc970227
AN		 614}
615
616static const struct vm_operations_struct bpf_map_default_vmops = {
617 .open = bpf_map_mmap_open,
618 .close = bpf_map_mmap_close,
619};
620
621static int bpf_map_mmap(struct file *filp, struct vm_area_struct *vma)
622{
623 struct bpf_map *map = filp->private_data;
624 int err;
625
68134668
AS		 626 if (!map->ops->map_mmap || map_value_has_spin_lock(map) ||
627 map_value_has_timer(map))
fc970227
AN		 628 return -ENOTSUPP;
629
630 if (!(vma->vm_flags & VM_SHARED))
631 return -EINVAL;
632
633 mutex_lock(&map->freeze_mutex);
634
dfeb376d
AN		 635 if (vma->vm_flags & VM_WRITE) {
636 if (map->frozen) {
637 err = -EPERM;
638 goto out;
639 }
640 /* map is meant to be read-only, so do not allow mapping as
641 * writable, because it's possible to leak a writable page
642 * reference and allows user-space to still modify it after
643 * freezing, while verifier will assume contents do not change
644 */
645 if (map->map_flags & BPF_F_RDONLY_PROG) {
646 err = -EACCES;
647 goto out;
648 }
fc970227
AN		 649 }
650
651 /* set default open/close callbacks */
652 vma->vm_ops = &bpf_map_default_vmops;
653 vma->vm_private_data = map;
1f6cb19b
AN		 654 vma->vm_flags &= ~VM_MAYEXEC;
655 if (!(vma->vm_flags & VM_WRITE))
656 /* disallow re-mapping with PROT_WRITE */
657 vma->vm_flags &= ~VM_MAYWRITE;
fc970227
AN		 658
659 err = map->ops->map_mmap(map, vma);
660 if (err)
661 goto out;
662
1f6cb19b 663 if (vma->vm_flags & VM_MAYWRITE)
fc970227
AN		 664 map->writecnt++;
665out:
666 mutex_unlock(&map->freeze_mutex);
667 return err;
668}
669
457f4436
AN		 670static __poll_t bpf_map_poll(struct file *filp, struct poll_table_struct *pts)
671{
672 struct bpf_map *map = filp->private_data;
673
674 if (map->ops->map_poll)
675 return map->ops->map_poll(map, filp, pts);
676
677 return EPOLLERR;
678}
679
f66e448c 680const struct file_operations bpf_map_fops = {
f99bf205
DB		 681#ifdef CONFIG_PROC_FS
682 .show_fdinfo = bpf_map_show_fdinfo,
683#endif
684 .release = bpf_map_release,
6e71b04a
CF		 685 .read = bpf_dummy_read,
686 .write = bpf_dummy_write,
fc970227 687 .mmap = bpf_map_mmap,
457f4436 688 .poll = bpf_map_poll,
99c55f7d
AS		 689};
690
6e71b04a 691int bpf_map_new_fd(struct bpf_map *map, int flags)
aa79781b 692{
afdb09c7
CF		 693 int ret;
694
695 ret = security_bpf_map(map, OPEN_FMODE(flags));
696 if (ret < 0)
697 return ret;
698
aa79781b 699 return anon_inode_getfd("bpf-map", &bpf_map_fops, map,
6e71b04a
CF		 700 flags | O_CLOEXEC);
701}
702
703int bpf_get_file_flag(int flags)
704{
705 if ((flags & BPF_F_RDONLY) && (flags & BPF_F_WRONLY))
706 return -EINVAL;
707 if (flags & BPF_F_RDONLY)
708 return O_RDONLY;
709 if (flags & BPF_F_WRONLY)
710 return O_WRONLY;
711 return O_RDWR;
aa79781b
DB		 712}
713
99c55f7d
AS		 714/* helper macro to check that unused fields 'union bpf_attr' are zero */
715#define CHECK_ATTR(CMD) \
716 memchr_inv((void *) &attr->CMD##_LAST_FIELD + \
717 sizeof(attr->CMD##_LAST_FIELD), 0, \
718 sizeof(*attr) - \
719 offsetof(union bpf_attr, CMD##_LAST_FIELD) - \
720 sizeof(attr->CMD##_LAST_FIELD)) != NULL
721
8e7ae251
MKL		 722/* dst and src must have at least "size" number of bytes.
723 * Return strlen on success and < 0 on error.
cb4d2b3f 724 */
8e7ae251 725int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size)
cb4d2b3f 726{
8e7ae251
MKL		 727 const char *end = src + size;
728 const char *orig_src = src;
cb4d2b3f 729
8e7ae251 730 memset(dst, 0, size);
3e0ddc4f 731 /* Copy all isalnum(), '_' and '.' chars. */
cb4d2b3f 732 while (src < end && *src) {
3e0ddc4f
DB		 733 if (!isalnum(*src) &&
734 *src != '_' && *src != '.')
cb4d2b3f
MKL		 735 return -EINVAL;
736 *dst++ = *src++;
737 }
738
8e7ae251 739 /* No '\0' found in "size" number of bytes */
cb4d2b3f
MKL		 740 if (src == end)
741 return -EINVAL;
742
8e7ae251 743 return src - orig_src;
cb4d2b3f
MKL		 744}
745
e8d2bec0 746int map_check_no_btf(const struct bpf_map *map,
1b2b234b 747 const struct btf *btf,
e8d2bec0
DB		 748 const struct btf_type *key_type,
749 const struct btf_type *value_type)
750{
751 return -ENOTSUPP;
752}
753
d83525ca 754static int map_check_btf(struct bpf_map *map, const struct btf *btf,
e8d2bec0
DB		 755 u32 btf_key_id, u32 btf_value_id)
756{
757 const struct btf_type *key_type, *value_type;
758 u32 key_size, value_size;
759 int ret = 0;
760
2824ecb7
DB		 761 /* Some maps allow key to be unspecified. */
762 if (btf_key_id) {
763 key_type = btf_type_id_size(btf, &btf_key_id, &key_size);
764 if (!key_type || key_size != map->key_size)
765 return -EINVAL;
766 } else {
767 key_type = btf_type_by_id(btf, 0);
768 if (!map->ops->map_check_btf)
769 return -EINVAL;
770 }
e8d2bec0
DB		 771
772 value_type = btf_type_id_size(btf, &btf_value_id, &value_size);
773 if (!value_type || value_size != map->value_size)
774 return -EINVAL;
775
d83525ca
AS		 776 map->spin_lock_off = btf_find_spin_lock(btf, value_type);
777
778 if (map_value_has_spin_lock(map)) {
591fe988
DB		 779 if (map->map_flags & BPF_F_RDONLY_PROG)
780 return -EACCES;
d83525ca 781 if (map->map_type != BPF_MAP_TYPE_HASH &&
e16d2f1a 782 map->map_type != BPF_MAP_TYPE_ARRAY &&
6ac99e8f 783 map->map_type != BPF_MAP_TYPE_CGROUP_STORAGE &&
8ea63684 784 map->map_type != BPF_MAP_TYPE_SK_STORAGE &&
4cf1bc1f
KS		 785 map->map_type != BPF_MAP_TYPE_INODE_STORAGE &&
786 map->map_type != BPF_MAP_TYPE_TASK_STORAGE)
d83525ca
AS		 787 return -ENOTSUPP;
788 if (map->spin_lock_off + sizeof(struct bpf_spin_lock) >
789 map->value_size) {
790 WARN_ONCE(1,
791 "verifier bug spin_lock_off %d value_size %d\n",
792 map->spin_lock_off, map->value_size);
793 return -EFAULT;
794 }
795 }
796
68134668
AS		 797 map->timer_off = btf_find_timer(btf, value_type);
798 if (map_value_has_timer(map)) {
799 if (map->map_flags & BPF_F_RDONLY_PROG)
800 return -EACCES;
801 if (map->map_type != BPF_MAP_TYPE_HASH &&
802 map->map_type != BPF_MAP_TYPE_LRU_HASH &&
803 map->map_type != BPF_MAP_TYPE_ARRAY)
804 return -EOPNOTSUPP;
805 }
806
e8d2bec0 807 if (map->ops->map_check_btf)
1b2b234b 808 ret = map->ops->map_check_btf(map, btf, key_type, value_type);
e8d2bec0
DB		 809
810 return ret;
811}
812
85d33df3 813#define BPF_MAP_CREATE_LAST_FIELD btf_vmlinux_value_type_id
99c55f7d
AS		 814/* called via syscall */
815static int map_create(union bpf_attr *attr)
816{
96eabe7a 817 int numa_node = bpf_map_attr_numa_node(attr);
99c55f7d 818 struct bpf_map *map;
6e71b04a 819 int f_flags;
99c55f7d
AS		 820 int err;
821
822 err = CHECK_ATTR(BPF_MAP_CREATE);
823 if (err)
824 return -EINVAL;
825
85d33df3
MKL		 826 if (attr->btf_vmlinux_value_type_id) {
827 if (attr->map_type != BPF_MAP_TYPE_STRUCT_OPS ||
828 attr->btf_key_type_id || attr->btf_value_type_id)
829 return -EINVAL;
830 } else if (attr->btf_key_type_id && !attr->btf_value_type_id) {
831 return -EINVAL;
832 }
833
6e71b04a
CF		 834 f_flags = bpf_get_file_flag(attr->map_flags);
835 if (f_flags < 0)
836 return f_flags;
837
96eabe7a 838 if (numa_node != NUMA_NO_NODE &&
96e5ae4e
ED		 839 ((unsigned int)numa_node >= nr_node_ids ||
840 !node_online(numa_node)))
96eabe7a
MKL		 841 return -EINVAL;
842
99c55f7d
AS		 843 /* find map type and init map: hashtable vs rbtree vs bloom vs ... */
844 map = find_and_alloc_map(attr);
845 if (IS_ERR(map))
846 return PTR_ERR(map);
847
8e7ae251
MKL		 848 err = bpf_obj_name_cpy(map->name, attr->map_name,
849 sizeof(attr->map_name));
850 if (err < 0)
b936ca64 851 goto free_map;
ad5b177b 852
1e0bd5a0
AN		 853 atomic64_set(&map->refcnt, 1);
854 atomic64_set(&map->usercnt, 1);
fc970227 855 mutex_init(&map->freeze_mutex);
99c55f7d 856
85d33df3 857 map->spin_lock_off = -EINVAL;
68134668 858 map->timer_off = -EINVAL;
85d33df3
MKL		 859 if (attr->btf_key_type_id || attr->btf_value_type_id ||
860 /* Even the map's value is a kernel's struct,
861 * the bpf_prog.o must have BTF to begin with
862 * to figure out the corresponding kernel's
863 * counter part. Thus, attr->btf_fd has
864 * to be valid also.
865 */
866 attr->btf_vmlinux_value_type_id) {
a26ca7c9
MKL		 867 struct btf *btf;
868
a26ca7c9
MKL		 869 btf = btf_get_by_fd(attr->btf_fd);
870 if (IS_ERR(btf)) {
871 err = PTR_ERR(btf);
b936ca64 872 goto free_map;
a26ca7c9 873 }
350a5c4d
AS		 874 if (btf_is_kernel(btf)) {
875 btf_put(btf);
876 err = -EACCES;
877 goto free_map;
878 }
85d33df3 879 map->btf = btf;
a26ca7c9 880
85d33df3
MKL		 881 if (attr->btf_value_type_id) {
882 err = map_check_btf(map, btf, attr->btf_key_type_id,
883 attr->btf_value_type_id);
884 if (err)
885 goto free_map;
a26ca7c9
MKL		 886 }
887
9b2cf328
MKL		 888 map->btf_key_type_id = attr->btf_key_type_id;
889 map->btf_value_type_id = attr->btf_value_type_id;
85d33df3
MKL		 890 map->btf_vmlinux_value_type_id =
891 attr->btf_vmlinux_value_type_id;
a26ca7c9
MKL		 892 }
893
afdb09c7 894 err = security_bpf_map_alloc(map);
aaac3ba9 895 if (err)
b936ca64 896 goto free_map;
afdb09c7 897
f3f1c054
MKL		 898 err = bpf_map_alloc_id(map);
899 if (err)
b936ca64 900 goto free_map_sec;
f3f1c054 901
48edc1f7
RG		 902 bpf_map_save_memcg(map);
903
6e71b04a 904 err = bpf_map_new_fd(map, f_flags);
bd5f5f4e
MKL		 905 if (err < 0) {
906 /* failed to allocate fd.
352d20d6 907 * bpf_map_put_with_uref() is needed because the above
bd5f5f4e
MKL		 908 * bpf_map_alloc_id() has published the map
909 * to the userspace and the userspace may
910 * have refcnt-ed it through BPF_MAP_GET_FD_BY_ID.
911 */
352d20d6 912 bpf_map_put_with_uref(map);
bd5f5f4e
MKL		 913 return err;
914 }
99c55f7d
AS		 915
916 return err;
917
afdb09c7
CF		 918free_map_sec:
919 security_bpf_map_free(map);
b936ca64 920free_map:
a26ca7c9 921 btf_put(map->btf);
99c55f7d
AS		 922 map->ops->map_free(map);
923 return err;
924}
925
db20fd2b
AS		 926/* if error is returned, fd is released.
927 * On success caller should complete fd access with matching fdput()
928 */
c2101297 929struct bpf_map *__bpf_map_get(struct fd f)
db20fd2b 930{
db20fd2b
AS		 931 if (!f.file)
932 return ERR_PTR(-EBADF);
db20fd2b
AS		 933 if (f.file->f_op != &bpf_map_fops) {
934 fdput(f);
935 return ERR_PTR(-EINVAL);
936 }
937
c2101297
DB		 938 return f.file->private_data;
939}
940
1e0bd5a0 941void bpf_map_inc(struct bpf_map *map)
c9da161c 942{
1e0bd5a0 943 atomic64_inc(&map->refcnt);
c9da161c 944}
630a4d38 945EXPORT_SYMBOL_GPL(bpf_map_inc);
c9da161c 946
1e0bd5a0
AN		 947void bpf_map_inc_with_uref(struct bpf_map *map)
948{
949 atomic64_inc(&map->refcnt);
950 atomic64_inc(&map->usercnt);
951}
952EXPORT_SYMBOL_GPL(bpf_map_inc_with_uref);
953
1ed4d924
MKL		 954struct bpf_map *bpf_map_get(u32 ufd)
955{
956 struct fd f = fdget(ufd);
957 struct bpf_map *map;
958
959 map = __bpf_map_get(f);
960 if (IS_ERR(map))
961 return map;
962
963 bpf_map_inc(map);
964 fdput(f);
965
966 return map;
967}
968
c9da161c 969struct bpf_map *bpf_map_get_with_uref(u32 ufd)
c2101297
DB		 970{
971 struct fd f = fdget(ufd);
972 struct bpf_map *map;
973
974 map = __bpf_map_get(f);
975 if (IS_ERR(map))
976 return map;
977
1e0bd5a0 978 bpf_map_inc_with_uref(map);
c2101297 979 fdput(f);
db20fd2b
AS		 980
981 return map;
982}
983
bd5f5f4e 984/* map_idr_lock should have been held */
1e0bd5a0 985static struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref)
bd5f5f4e
MKL		 986{
987 int refold;
988
1e0bd5a0 989 refold = atomic64_fetch_add_unless(&map->refcnt, 1, 0);
bd5f5f4e
MKL		 990 if (!refold)
991 return ERR_PTR(-ENOENT);
bd5f5f4e 992 if (uref)
1e0bd5a0 993 atomic64_inc(&map->usercnt);
bd5f5f4e
MKL		 994
995 return map;
996}
997
1e0bd5a0 998struct bpf_map *bpf_map_inc_not_zero(struct bpf_map *map)
b0e4701c
SF		 999{
1000 spin_lock_bh(&map_idr_lock);
1e0bd5a0 1001 map = __bpf_map_inc_not_zero(map, false);
b0e4701c
SF		 1002 spin_unlock_bh(&map_idr_lock);
1003
1004 return map;
1005}
1006EXPORT_SYMBOL_GPL(bpf_map_inc_not_zero);
1007
b8cdc051
AS		 1008int __weak bpf_stackmap_copy(struct bpf_map *map, void *key, void *value)
1009{
1010 return -ENOTSUPP;
1011}
1012
c9d29f46
MV		 1013static void *__bpf_copy_key(void __user *ukey, u64 key_size)
1014{
1015 if (key_size)
44779a4b 1016 return vmemdup_user(ukey, key_size);
c9d29f46
MV		 1017
1018 if (ukey)
1019 return ERR_PTR(-EINVAL);
1020
1021 return NULL;
1022}
1023
af2ac3e1
AS		 1024static void *___bpf_copy_key(bpfptr_t ukey, u64 key_size)
1025{
1026 if (key_size)
44779a4b 1027 return kvmemdup_bpfptr(ukey, key_size);
af2ac3e1
AS		 1028
1029 if (!bpfptr_is_null(ukey))
1030 return ERR_PTR(-EINVAL);
1031
1032 return NULL;
1033}
1034
db20fd2b 1035/* last field in 'union bpf_attr' used by this command */
96049f3a 1036#define BPF_MAP_LOOKUP_ELEM_LAST_FIELD flags
db20fd2b
AS		 1037
1038static int map_lookup_elem(union bpf_attr *attr)
1039{
535e7b4b
MS		 1040 void __user *ukey = u64_to_user_ptr(attr->key);
1041 void __user *uvalue = u64_to_user_ptr(attr->value);
db20fd2b 1042 int ufd = attr->map_fd;
db20fd2b 1043 struct bpf_map *map;
15c14a3d 1044 void *key, *value;
15a07b33 1045 u32 value_size;
592867bf 1046 struct fd f;
db20fd2b
AS		 1047 int err;
1048
1049 if (CHECK_ATTR(BPF_MAP_LOOKUP_ELEM))
1050 return -EINVAL;
1051
96049f3a
AS		 1052 if (attr->flags & ~BPF_F_LOCK)
1053 return -EINVAL;
1054
592867bf 1055 f = fdget(ufd);
c2101297 1056 map = __bpf_map_get(f);
db20fd2b
AS		 1057 if (IS_ERR(map))
1058 return PTR_ERR(map);
87df15de 1059 if (!(map_get_sys_perms(map, f) & FMODE_CAN_READ)) {
6e71b04a
CF		 1060 err = -EPERM;
1061 goto err_put;
1062 }
1063
96049f3a
AS		 1064 if ((attr->flags & BPF_F_LOCK) &&
1065 !map_value_has_spin_lock(map)) {
1066 err = -EINVAL;
1067 goto err_put;
1068 }
1069
c9d29f46 1070 key = __bpf_copy_key(ukey, map->key_size);
e4448ed8
AV		 1071 if (IS_ERR(key)) {
1072 err = PTR_ERR(key);
db20fd2b 1073 goto err_put;
e4448ed8 1074 }
db20fd2b 1075
15c14a3d 1076 value_size = bpf_map_value_size(map);
15a07b33 1077
8ebe667c 1078 err = -ENOMEM;
f0dce1d9 1079 value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN);
db20fd2b 1080 if (!value)
8ebe667c
AS		 1081 goto free_key;
1082
15c14a3d 1083 err = bpf_map_copy_value(map, key, value, attr->flags);
15a07b33 1084 if (err)
8ebe667c 1085 goto free_value;
db20fd2b
AS		 1086
1087 err = -EFAULT;
15a07b33 1088 if (copy_to_user(uvalue, value, value_size) != 0)
8ebe667c 1089 goto free_value;
db20fd2b
AS		 1090
1091 err = 0;
1092
8ebe667c 1093free_value:
f0dce1d9 1094 kvfree(value);
db20fd2b 1095free_key:
44779a4b 1096 kvfree(key);
db20fd2b
AS		 1097err_put:
1098 fdput(f);
1099 return err;
1100}
1101
1ae80cf3 1102
3274f520 1103#define BPF_MAP_UPDATE_ELEM_LAST_FIELD flags
db20fd2b 1104
af2ac3e1 1105static int map_update_elem(union bpf_attr *attr, bpfptr_t uattr)
db20fd2b 1106{
af2ac3e1
AS		 1107 bpfptr_t ukey = make_bpfptr(attr->key, uattr.is_kernel);
1108 bpfptr_t uvalue = make_bpfptr(attr->value, uattr.is_kernel);
db20fd2b 1109 int ufd = attr->map_fd;
db20fd2b
AS		 1110 struct bpf_map *map;
1111 void *key, *value;
15a07b33 1112 u32 value_size;
592867bf 1113 struct fd f;
db20fd2b
AS		 1114 int err;
1115
1116 if (CHECK_ATTR(BPF_MAP_UPDATE_ELEM))
1117 return -EINVAL;
1118
592867bf 1119 f = fdget(ufd);
c2101297 1120 map = __bpf_map_get(f);
db20fd2b
AS		 1121 if (IS_ERR(map))
1122 return PTR_ERR(map);
87df15de 1123 if (!(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) {
6e71b04a
CF		 1124 err = -EPERM;
1125 goto err_put;
1126 }
1127
96049f3a
AS		 1128 if ((attr->flags & BPF_F_LOCK) &&
1129 !map_value_has_spin_lock(map)) {
1130 err = -EINVAL;
1131 goto err_put;
1132 }
1133
af2ac3e1 1134 key = ___bpf_copy_key(ukey, map->key_size);
e4448ed8
AV		 1135 if (IS_ERR(key)) {
1136 err = PTR_ERR(key);
db20fd2b 1137 goto err_put;
e4448ed8 1138 }
db20fd2b 1139
f0dce1d9 1140 value_size = bpf_map_value_size(map);
15a07b33 1141
db20fd2b 1142 err = -ENOMEM;
f0dce1d9 1143 value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN);
db20fd2b
AS		 1144 if (!value)
1145 goto free_key;
1146
1147 err = -EFAULT;
af2ac3e1 1148 if (copy_from_bpfptr(value, uvalue, value_size) != 0)
db20fd2b
AS		 1149 goto free_value;
1150
15c14a3d 1151 err = bpf_map_update_value(map, f, key, value, attr->flags);
6710e112 1152
db20fd2b 1153free_value:
f0dce1d9 1154 kvfree(value);
db20fd2b 1155free_key:
44779a4b 1156 kvfree(key);
db20fd2b
AS		 1157err_put:
1158 fdput(f);
1159 return err;
1160}
1161
1162#define BPF_MAP_DELETE_ELEM_LAST_FIELD key
1163
1164static int map_delete_elem(union bpf_attr *attr)
1165{
535e7b4b 1166 void __user *ukey = u64_to_user_ptr(attr->key);
db20fd2b 1167 int ufd = attr->map_fd;
db20fd2b 1168 struct bpf_map *map;
592867bf 1169 struct fd f;
db20fd2b
AS		 1170 void *key;
1171 int err;
1172
1173 if (CHECK_ATTR(BPF_MAP_DELETE_ELEM))
1174 return -EINVAL;
1175
592867bf 1176 f = fdget(ufd);
c2101297 1177 map = __bpf_map_get(f);
db20fd2b
AS		 1178 if (IS_ERR(map))
1179 return PTR_ERR(map);
87df15de 1180 if (!(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) {
6e71b04a
CF		 1181 err = -EPERM;
1182 goto err_put;
1183 }
1184
c9d29f46 1185 key = __bpf_copy_key(ukey, map->key_size);
e4448ed8
AV		 1186 if (IS_ERR(key)) {
1187 err = PTR_ERR(key);
db20fd2b 1188 goto err_put;
e4448ed8 1189 }
db20fd2b 1190
a3884572
JK		 1191 if (bpf_map_is_dev_bound(map)) {
1192 err = bpf_map_offload_delete_elem(map, key);
1193 goto out;
85d33df3
MKL		 1194 } else if (IS_FD_PROG_ARRAY(map) ||
1195 map->map_type == BPF_MAP_TYPE_STRUCT_OPS) {
1196 /* These maps require sleepable context */
da765a2f
DB		 1197 err = map->ops->map_delete_elem(map, key);
1198 goto out;
a3884572
JK		 1199 }
1200
b6e5dae1 1201 bpf_disable_instrumentation();
db20fd2b
AS		 1202 rcu_read_lock();
1203 err = map->ops->map_delete_elem(map, key);
1204 rcu_read_unlock();
b6e5dae1 1205 bpf_enable_instrumentation();
1ae80cf3 1206 maybe_wait_bpf_programs(map);
a3884572 1207out:
44779a4b 1208 kvfree(key);
db20fd2b
AS		 1209err_put:
1210 fdput(f);
1211 return err;
1212}
1213
1214/* last field in 'union bpf_attr' used by this command */
1215#define BPF_MAP_GET_NEXT_KEY_LAST_FIELD next_key
1216
1217static int map_get_next_key(union bpf_attr *attr)
1218{
535e7b4b
MS		 1219 void __user *ukey = u64_to_user_ptr(attr->key);
1220 void __user *unext_key = u64_to_user_ptr(attr->next_key);
db20fd2b 1221 int ufd = attr->map_fd;
db20fd2b
AS		 1222 struct bpf_map *map;
1223 void *key, *next_key;
592867bf 1224 struct fd f;
db20fd2b
AS		 1225 int err;
1226
1227 if (CHECK_ATTR(BPF_MAP_GET_NEXT_KEY))
1228 return -EINVAL;
1229
592867bf 1230 f = fdget(ufd);
c2101297 1231 map = __bpf_map_get(f);
db20fd2b
AS		 1232 if (IS_ERR(map))
1233 return PTR_ERR(map);
87df15de 1234 if (!(map_get_sys_perms(map, f) & FMODE_CAN_READ)) {
6e71b04a
CF		 1235 err = -EPERM;
1236 goto err_put;
1237 }
1238
8fe45924 1239 if (ukey) {
c9d29f46 1240 key = __bpf_copy_key(ukey, map->key_size);
e4448ed8
AV		 1241 if (IS_ERR(key)) {
1242 err = PTR_ERR(key);
8fe45924 1243 goto err_put;
e4448ed8 1244 }
8fe45924
TQ		 1245 } else {
1246 key = NULL;
1247 }
db20fd2b
AS		 1248
1249 err = -ENOMEM;
44779a4b 1250 next_key = kvmalloc(map->key_size, GFP_USER);
db20fd2b
AS		 1251 if (!next_key)
1252 goto free_key;
1253
a3884572
JK		 1254 if (bpf_map_is_dev_bound(map)) {
1255 err = bpf_map_offload_get_next_key(map, key, next_key);
1256 goto out;
1257 }
1258
db20fd2b
AS		 1259 rcu_read_lock();
1260 err = map->ops->map_get_next_key(map, key, next_key);
1261 rcu_read_unlock();
a3884572 1262out:
db20fd2b
AS		 1263 if (err)
1264 goto free_next_key;
1265
1266 err = -EFAULT;
1267 if (copy_to_user(unext_key, next_key, map->key_size) != 0)
1268 goto free_next_key;
1269
1270 err = 0;
1271
1272free_next_key:
44779a4b 1273 kvfree(next_key);
db20fd2b 1274free_key:
44779a4b 1275 kvfree(key);
db20fd2b
AS		 1276err_put:
1277 fdput(f);
1278 return err;
1279}
1280
aa2e93b8
BV		 1281int generic_map_delete_batch(struct bpf_map *map,
1282 const union bpf_attr *attr,
1283 union bpf_attr __user *uattr)
1284{
1285 void __user *keys = u64_to_user_ptr(attr->batch.keys);
1286 u32 cp, max_count;
1287 int err = 0;
1288 void *key;
1289
1290 if (attr->batch.elem_flags & ~BPF_F_LOCK)
1291 return -EINVAL;
1292
1293 if ((attr->batch.elem_flags & BPF_F_LOCK) &&
1294 !map_value_has_spin_lock(map)) {
1295 return -EINVAL;
1296 }
1297
1298 max_count = attr->batch.count;
1299 if (!max_count)
1300 return 0;
1301
44779a4b 1302 key = kvmalloc(map->key_size, GFP_USER | __GFP_NOWARN);
2e3a94aa
BV		 1303 if (!key)
1304 return -ENOMEM;
1305
aa2e93b8 1306 for (cp = 0; cp < max_count; cp++) {
2e3a94aa
BV		 1307 err = -EFAULT;
1308 if (copy_from_user(key, keys + cp * map->key_size,
1309 map->key_size))
aa2e93b8 1310 break;
aa2e93b8
BV		 1311
1312 if (bpf_map_is_dev_bound(map)) {
1313 err = bpf_map_offload_delete_elem(map, key);
1314 break;
1315 }
1316
b6e5dae1 1317 bpf_disable_instrumentation();
aa2e93b8
BV		 1318 rcu_read_lock();
1319 err = map->ops->map_delete_elem(map, key);
1320 rcu_read_unlock();
b6e5dae1 1321 bpf_enable_instrumentation();
aa2e93b8
BV		 1322 maybe_wait_bpf_programs(map);
1323 if (err)
1324 break;
1325 }
1326 if (copy_to_user(&uattr->batch.count, &cp, sizeof(cp)))
1327 err = -EFAULT;
2e3a94aa 1328
44779a4b 1329 kvfree(key);
aa2e93b8
BV		 1330 return err;
1331}
1332
1333int generic_map_update_batch(struct bpf_map *map,
1334 const union bpf_attr *attr,
1335 union bpf_attr __user *uattr)
1336{
1337 void __user *values = u64_to_user_ptr(attr->batch.values);
1338 void __user *keys = u64_to_user_ptr(attr->batch.keys);
1339 u32 value_size, cp, max_count;
1340 int ufd = attr->map_fd;
1341 void *key, *value;
1342 struct fd f;
1343 int err = 0;
1344
1345 f = fdget(ufd);
1346 if (attr->batch.elem_flags & ~BPF_F_LOCK)
1347 return -EINVAL;
1348
1349 if ((attr->batch.elem_flags & BPF_F_LOCK) &&
1350 !map_value_has_spin_lock(map)) {
1351 return -EINVAL;
1352 }
1353
1354 value_size = bpf_map_value_size(map);
1355
1356 max_count = attr->batch.count;
1357 if (!max_count)
1358 return 0;
1359
44779a4b 1360 key = kvmalloc(map->key_size, GFP_USER | __GFP_NOWARN);
2e3a94aa
BV		 1361 if (!key)
1362 return -ENOMEM;
1363
f0dce1d9 1364 value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN);
2e3a94aa 1365 if (!value) {
44779a4b 1366 kvfree(key);
aa2e93b8 1367 return -ENOMEM;
2e3a94aa 1368 }
aa2e93b8
BV		 1369
1370 for (cp = 0; cp < max_count; cp++) {
aa2e93b8 1371 err = -EFAULT;
2e3a94aa
BV		 1372 if (copy_from_user(key, keys + cp * map->key_size,
1373 map->key_size) ||
1374 copy_from_user(value, values + cp * value_size, value_size))
aa2e93b8
BV		 1375 break;
1376
1377 err = bpf_map_update_value(map, f, key, value,
1378 attr->batch.elem_flags);
1379
1380 if (err)
1381 break;
1382 }
1383
1384 if (copy_to_user(&uattr->batch.count, &cp, sizeof(cp)))
1385 err = -EFAULT;
1386
f0dce1d9 1387 kvfree(value);
44779a4b 1388 kvfree(key);
aa2e93b8
BV		 1389 return err;
1390}
1391
cb4d03ab
BV		 1392#define MAP_LOOKUP_RETRIES 3
1393
1394int generic_map_lookup_batch(struct bpf_map *map,
1395 const union bpf_attr *attr,
1396 union bpf_attr __user *uattr)
1397{
1398 void __user *uobatch = u64_to_user_ptr(attr->batch.out_batch);
1399 void __user *ubatch = u64_to_user_ptr(attr->batch.in_batch);
1400 void __user *values = u64_to_user_ptr(attr->batch.values);
1401 void __user *keys = u64_to_user_ptr(attr->batch.keys);
1402 void *buf, *buf_prevkey, *prev_key, *key, *value;
1403 int err, retry = MAP_LOOKUP_RETRIES;
1404 u32 value_size, cp, max_count;
cb4d03ab
BV		 1405
1406 if (attr->batch.elem_flags & ~BPF_F_LOCK)
1407 return -EINVAL;
1408
1409 if ((attr->batch.elem_flags & BPF_F_LOCK) &&
1410 !map_value_has_spin_lock(map))
1411 return -EINVAL;
1412
1413 value_size = bpf_map_value_size(map);
1414
1415 max_count = attr->batch.count;
1416 if (!max_count)
1417 return 0;
1418
1419 if (put_user(0, &uattr->batch.count))
1420 return -EFAULT;
1421
44779a4b 1422 buf_prevkey = kvmalloc(map->key_size, GFP_USER | __GFP_NOWARN);
cb4d03ab
BV		 1423 if (!buf_prevkey)
1424 return -ENOMEM;
1425
f0dce1d9 1426 buf = kvmalloc(map->key_size + value_size, GFP_USER | __GFP_NOWARN);
cb4d03ab 1427 if (!buf) {
44779a4b 1428 kvfree(buf_prevkey);
cb4d03ab
BV		 1429 return -ENOMEM;
1430 }
1431
1432 err = -EFAULT;
cb4d03ab
BV		 1433 prev_key = NULL;
1434 if (ubatch && copy_from_user(buf_prevkey, ubatch, map->key_size))
1435 goto free_buf;
1436 key = buf;
1437 value = key + map->key_size;
1438 if (ubatch)
1439 prev_key = buf_prevkey;
1440
1441 for (cp = 0; cp < max_count;) {
1442 rcu_read_lock();
1443 err = map->ops->map_get_next_key(map, prev_key, key);
1444 rcu_read_unlock();
1445 if (err)
1446 break;
1447 err = bpf_map_copy_value(map, key, value,
1448 attr->batch.elem_flags);
1449
1450 if (err == -ENOENT) {
1451 if (retry) {
1452 retry--;
1453 continue;
1454 }
1455 err = -EINTR;
1456 break;
1457 }
1458
1459 if (err)
1460 goto free_buf;
1461
1462 if (copy_to_user(keys + cp * map->key_size, key,
1463 map->key_size)) {
1464 err = -EFAULT;
1465 goto free_buf;
1466 }
1467 if (copy_to_user(values + cp * value_size, value, value_size)) {
1468 err = -EFAULT;
1469 goto free_buf;
1470 }
1471
1472 if (!prev_key)
1473 prev_key = buf_prevkey;
1474
1475 swap(prev_key, key);
1476 retry = MAP_LOOKUP_RETRIES;
1477 cp++;
1478 }
1479
1480 if (err == -EFAULT)
1481 goto free_buf;
1482
1483 if ((copy_to_user(&uattr->batch.count, &cp, sizeof(cp)) ||
1484 (cp && copy_to_user(uobatch, prev_key, map->key_size))))
1485 err = -EFAULT;
1486
1487free_buf:
44779a4b 1488 kvfree(buf_prevkey);
f0dce1d9 1489 kvfree(buf);
cb4d03ab
BV		 1490 return err;
1491}
1492
3e87f192 1493#define BPF_MAP_LOOKUP_AND_DELETE_ELEM_LAST_FIELD flags
bd513cd0
MV		 1494
1495static int map_lookup_and_delete_elem(union bpf_attr *attr)
1496{
1497 void __user *ukey = u64_to_user_ptr(attr->key);
1498 void __user *uvalue = u64_to_user_ptr(attr->value);
1499 int ufd = attr->map_fd;
1500 struct bpf_map *map;
540fefc0 1501 void *key, *value;
bd513cd0
MV		 1502 u32 value_size;
1503 struct fd f;
1504 int err;
1505
1506 if (CHECK_ATTR(BPF_MAP_LOOKUP_AND_DELETE_ELEM))
1507 return -EINVAL;
1508
3e87f192
DS		 1509 if (attr->flags & ~BPF_F_LOCK)
1510 return -EINVAL;
1511
bd513cd0
MV		 1512 f = fdget(ufd);
1513 map = __bpf_map_get(f);
1514 if (IS_ERR(map))
1515 return PTR_ERR(map);
1ea0f912
AP		 1516 if (!(map_get_sys_perms(map, f) & FMODE_CAN_READ) ||
1517 !(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) {
bd513cd0
MV		 1518 err = -EPERM;
1519 goto err_put;
1520 }
1521
3e87f192
DS		 1522 if (attr->flags &&
1523 (map->map_type == BPF_MAP_TYPE_QUEUE ||
1524 map->map_type == BPF_MAP_TYPE_STACK)) {
1525 err = -EINVAL;
1526 goto err_put;
1527 }
1528
1529 if ((attr->flags & BPF_F_LOCK) &&
1530 !map_value_has_spin_lock(map)) {
1531 err = -EINVAL;
1532 goto err_put;
1533 }
1534
bd513cd0
MV		 1535 key = __bpf_copy_key(ukey, map->key_size);
1536 if (IS_ERR(key)) {
1537 err = PTR_ERR(key);
1538 goto err_put;
1539 }
1540
3e87f192 1541 value_size = bpf_map_value_size(map);
bd513cd0
MV		 1542
1543 err = -ENOMEM;
f0dce1d9 1544 value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN);
bd513cd0
MV		 1545 if (!value)
1546 goto free_key;
1547
3e87f192 1548 err = -ENOTSUPP;
bd513cd0
MV		 1549 if (map->map_type == BPF_MAP_TYPE_QUEUE ||
1550 map->map_type == BPF_MAP_TYPE_STACK) {
1551 err = map->ops->map_pop_elem(map, value);
3e87f192
DS		 1552 } else if (map->map_type == BPF_MAP_TYPE_HASH ||
1553 map->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
1554 map->map_type == BPF_MAP_TYPE_LRU_HASH ||
1555 map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) {
1556 if (!bpf_map_is_dev_bound(map)) {
1557 bpf_disable_instrumentation();
1558 rcu_read_lock();
1559 err = map->ops->map_lookup_and_delete_elem(map, key, value, attr->flags);
1560 rcu_read_unlock();
1561 bpf_enable_instrumentation();
1562 }
bd513cd0
MV		 1563 }
1564
1565 if (err)
1566 goto free_value;
1567
7f645462
WY		 1568 if (copy_to_user(uvalue, value, value_size) != 0) {
1569 err = -EFAULT;
bd513cd0 1570 goto free_value;
7f645462 1571 }
bd513cd0
MV		 1572
1573 err = 0;
1574
1575free_value:
f0dce1d9 1576 kvfree(value);
bd513cd0 1577free_key:
44779a4b 1578 kvfree(key);
bd513cd0
MV		 1579err_put:
1580 fdput(f);
1581 return err;
1582}
1583
87df15de
DB		 1584#define BPF_MAP_FREEZE_LAST_FIELD map_fd
1585
1586static int map_freeze(const union bpf_attr *attr)
1587{
1588 int err = 0, ufd = attr->map_fd;
1589 struct bpf_map *map;
1590 struct fd f;
1591
1592 if (CHECK_ATTR(BPF_MAP_FREEZE))
1593 return -EINVAL;
1594
1595 f = fdget(ufd);
1596 map = __bpf_map_get(f);
1597 if (IS_ERR(map))
1598 return PTR_ERR(map);
fc970227 1599
68134668
AS		 1600 if (map->map_type == BPF_MAP_TYPE_STRUCT_OPS ||
1601 map_value_has_timer(map)) {
849b4d94
MKL		 1602 fdput(f);
1603 return -ENOTSUPP;
1604 }
1605
fc970227
AN		 1606 mutex_lock(&map->freeze_mutex);
1607
1608 if (map->writecnt) {
1609 err = -EBUSY;
1610 goto err_put;
1611 }
87df15de
DB		 1612 if (READ_ONCE(map->frozen)) {
1613 err = -EBUSY;
1614 goto err_put;
1615 }
2c78ee89 1616 if (!bpf_capable()) {
87df15de
DB		 1617 err = -EPERM;
1618 goto err_put;
1619 }
1620
1621 WRITE_ONCE(map->frozen, true);
1622err_put:
fc970227 1623 mutex_unlock(&map->freeze_mutex);
87df15de
DB		 1624 fdput(f);
1625 return err;
1626}
1627
7de16e3a 1628static const struct bpf_prog_ops * const bpf_prog_types[] = {
91cc1a99 1629#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
7de16e3a
JK		 1630 [_id] = & _name ## _prog_ops,
1631#define BPF_MAP_TYPE(_id, _ops)
f2e10bff 1632#define BPF_LINK_TYPE(_id, _name)
7de16e3a
JK		 1633#include <linux/bpf_types.h>
1634#undef BPF_PROG_TYPE
1635#undef BPF_MAP_TYPE
f2e10bff 1636#undef BPF_LINK_TYPE
7de16e3a
JK		 1637};
1638
09756af4
AS		 1639static int find_prog_type(enum bpf_prog_type type, struct bpf_prog *prog)
1640{
d0f1a451
DB		 1641 const struct bpf_prog_ops *ops;
1642
1643 if (type >= ARRAY_SIZE(bpf_prog_types))
1644 return -EINVAL;
1645 type = array_index_nospec(type, ARRAY_SIZE(bpf_prog_types));
1646 ops = bpf_prog_types[type];
1647 if (!ops)
be9370a7 1648 return -EINVAL;
09756af4 1649
ab3f0063 1650 if (!bpf_prog_is_dev_bound(prog->aux))
d0f1a451 1651 prog->aux->ops = ops;
ab3f0063
JK		 1652 else
1653 prog->aux->ops = &bpf_offload_prog_ops;
be9370a7
JB		 1654 prog->type = type;
1655 return 0;
09756af4
AS		 1656}
1657
bae141f5
DB		 1658enum bpf_audit {
1659 BPF_AUDIT_LOAD,
1660 BPF_AUDIT_UNLOAD,
1661 BPF_AUDIT_MAX,
1662};
1663
1664static const char * const bpf_audit_str[BPF_AUDIT_MAX] = {
1665 [BPF_AUDIT_LOAD] = "LOAD",
1666 [BPF_AUDIT_UNLOAD] = "UNLOAD",
1667};
1668
1669static void bpf_audit_prog(const struct bpf_prog *prog, unsigned int op)
1670{
1671 struct audit_context *ctx = NULL;
1672 struct audit_buffer *ab;
1673
1674 if (WARN_ON_ONCE(op >= BPF_AUDIT_MAX))
1675 return;
1676 if (audit_enabled == AUDIT_OFF)
1677 return;
1678 if (op == BPF_AUDIT_LOAD)
1679 ctx = audit_context();
1680 ab = audit_log_start(ctx, GFP_ATOMIC, AUDIT_BPF);
1681 if (unlikely(!ab))
1682 return;
1683 audit_log_format(ab, "prog-id=%u op=%s",
1684 prog->aux->id, bpf_audit_str[op]);
1685 audit_log_end(ab);
1686}
1687
dc4bb0e2
MKL		 1688static int bpf_prog_alloc_id(struct bpf_prog *prog)
1689{
1690 int id;
1691
b76354cd 1692 idr_preload(GFP_KERNEL);
dc4bb0e2
MKL		 1693 spin_lock_bh(&prog_idr_lock);
1694 id = idr_alloc_cyclic(&prog_idr, prog, 1, INT_MAX, GFP_ATOMIC);
1695 if (id > 0)
1696 prog->aux->id = id;
1697 spin_unlock_bh(&prog_idr_lock);
b76354cd 1698 idr_preload_end();
dc4bb0e2
MKL		 1699
1700 /* id is in [1, INT_MAX) */
1701 if (WARN_ON_ONCE(!id))
1702 return -ENOSPC;
1703
1704 return id > 0 ? 0 : id;
1705}
1706
ad8ad79f 1707void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock)
dc4bb0e2 1708{
d809e134
AS		 1709 unsigned long flags;
1710
ad8ad79f
JK		 1711 /* cBPF to eBPF migrations are currently not in the idr store.
1712 * Offloaded programs are removed from the store when their device
1713 * disappears - even if someone grabs an fd to them they are unusable,
1714 * simply waiting for refcnt to drop to be freed.
1715 */
dc4bb0e2
MKL		 1716 if (!prog->aux->id)
1717 return;
1718
b16d9aa4 1719 if (do_idr_lock)
d809e134 1720 spin_lock_irqsave(&prog_idr_lock, flags);
b16d9aa4
MKL		 1721 else
1722 __acquire(&prog_idr_lock);
1723
dc4bb0e2 1724 idr_remove(&prog_idr, prog->aux->id);
ad8ad79f 1725 prog->aux->id = 0;
b16d9aa4
MKL		 1726
1727 if (do_idr_lock)
d809e134 1728 spin_unlock_irqrestore(&prog_idr_lock, flags);
b16d9aa4
MKL		 1729 else
1730 __release(&prog_idr_lock);
dc4bb0e2
MKL		 1731}
1732
1aacde3d 1733static void __bpf_prog_put_rcu(struct rcu_head *rcu)
abf2e7d6
AS		 1734{
1735 struct bpf_prog_aux *aux = container_of(rcu, struct bpf_prog_aux, rcu);
1736
3b4d9eb2 1737 kvfree(aux->func_info);
8c1b6e69 1738 kfree(aux->func_info_aux);
3ac1f01b 1739 free_uid(aux->user);
afdb09c7 1740 security_bpf_prog_free(aux);
abf2e7d6
AS		 1741 bpf_prog_free(aux->prog);
1742}
1743
cd7455f1
DB		 1744static void __bpf_prog_put_noref(struct bpf_prog *prog, bool deferred)
1745{
1746 bpf_prog_kallsyms_del_all(prog);
1747 btf_put(prog->aux->btf);
e16301fb
MKL		 1748 kvfree(prog->aux->jited_linfo);
1749 kvfree(prog->aux->linfo);
e6ac2450 1750 kfree(prog->aux->kfunc_tab);
22dc4a0f
AN		 1751 if (prog->aux->attach_btf)
1752 btf_put(prog->aux->attach_btf);
cd7455f1 1753
1e6c62a8
AS		 1754 if (deferred) {
1755 if (prog->aux->sleepable)
1756 call_rcu_tasks_trace(&prog->aux->rcu, __bpf_prog_put_rcu);
1757 else
1758 call_rcu(&prog->aux->rcu, __bpf_prog_put_rcu);
1759 } else {
cd7455f1 1760 __bpf_prog_put_rcu(&prog->aux->rcu);
1e6c62a8 1761 }
cd7455f1
DB		 1762}
1763
d809e134
AS		 1764static void bpf_prog_put_deferred(struct work_struct *work)
1765{
1766 struct bpf_prog_aux *aux;
1767 struct bpf_prog *prog;
1768
1769 aux = container_of(work, struct bpf_prog_aux, work);
1770 prog = aux->prog;
1771 perf_event_bpf_event(prog, PERF_BPF_EVENT_PROG_UNLOAD, 0);
1772 bpf_audit_prog(prog, BPF_AUDIT_UNLOAD);
1773 __bpf_prog_put_noref(prog, true);
1774}
1775
b16d9aa4 1776static void __bpf_prog_put(struct bpf_prog *prog, bool do_idr_lock)
09756af4 1777{
d809e134
AS		 1778 struct bpf_prog_aux *aux = prog->aux;
1779
1780 if (atomic64_dec_and_test(&aux->refcnt)) {
34ad5580 1781 /* bpf_prog_free_id() must be called first */
b16d9aa4 1782 bpf_prog_free_id(prog, do_idr_lock);
d809e134
AS		 1783
1784 if (in_irq() || irqs_disabled()) {
1785 INIT_WORK(&aux->work, bpf_prog_put_deferred);
1786 schedule_work(&aux->work);
1787 } else {
1788 bpf_prog_put_deferred(&aux->work);
1789 }
a67edbf4 1790 }
09756af4 1791}
b16d9aa4
MKL		 1792
1793void bpf_prog_put(struct bpf_prog *prog)
1794{
1795 __bpf_prog_put(prog, true);
1796}
e2e9b654 1797EXPORT_SYMBOL_GPL(bpf_prog_put);
09756af4
AS		 1798
1799static int bpf_prog_release(struct inode *inode, struct file *filp)
1800{
1801 struct bpf_prog *prog = filp->private_data;
1802
1aacde3d 1803 bpf_prog_put(prog);
09756af4
AS		 1804 return 0;
1805}
1806
492ecee8
AS		 1807static void bpf_prog_get_stats(const struct bpf_prog *prog,
1808 struct bpf_prog_stats *stats)
1809{
9ed9e9ba 1810 u64 nsecs = 0, cnt = 0, misses = 0;
492ecee8
AS		 1811 int cpu;
1812
1813 for_each_possible_cpu(cpu) {
1814 const struct bpf_prog_stats *st;
1815 unsigned int start;
9ed9e9ba 1816 u64 tnsecs, tcnt, tmisses;
492ecee8 1817
700d4796 1818 st = per_cpu_ptr(prog->stats, cpu);
492ecee8
AS		 1819 do {
1820 start = u64_stats_fetch_begin_irq(&st->syncp);
1821 tnsecs = st->nsecs;
1822 tcnt = st->cnt;
9ed9e9ba 1823 tmisses = st->misses;
492ecee8
AS		 1824 } while (u64_stats_fetch_retry_irq(&st->syncp, start));
1825 nsecs += tnsecs;
1826 cnt += tcnt;
9ed9e9ba 1827 misses += tmisses;
492ecee8
AS		 1828 }
1829 stats->nsecs = nsecs;
1830 stats->cnt = cnt;
9ed9e9ba 1831 stats->misses = misses;
492ecee8
AS		 1832}
1833
7bd509e3
DB		 1834#ifdef CONFIG_PROC_FS
1835static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
1836{
1837 const struct bpf_prog *prog = filp->private_data;
f1f7714e 1838 char prog_tag[sizeof(prog->tag) * 2 + 1] = { };
492ecee8 1839 struct bpf_prog_stats stats;
7bd509e3 1840
492ecee8 1841 bpf_prog_get_stats(prog, &stats);
f1f7714e 1842 bin2hex(prog_tag, prog->tag, sizeof(prog->tag));
7bd509e3
DB		 1843 seq_printf(m,
1844 "prog_type:\t%u\n"
1845 "prog_jited:\t%u\n"
f1f7714e 1846 "prog_tag:\t%s\n"
4316b409 1847 "memlock:\t%llu\n"
492ecee8
AS		 1848 "prog_id:\t%u\n"
1849 "run_time_ns:\t%llu\n"
9ed9e9ba
AS		 1850 "run_cnt:\t%llu\n"
1851 "recursion_misses:\t%llu\n",
7bd509e3
DB		 1852 prog->type,
1853 prog->jited,
f1f7714e 1854 prog_tag,
4316b409 1855 prog->pages * 1ULL << PAGE_SHIFT,
492ecee8
AS		 1856 prog->aux->id,
1857 stats.nsecs,
9ed9e9ba
AS		 1858 stats.cnt,
1859 stats.misses);
7bd509e3
DB		 1860}
1861#endif
1862
f66e448c 1863const struct file_operations bpf_prog_fops = {
7bd509e3
DB		 1864#ifdef CONFIG_PROC_FS
1865 .show_fdinfo = bpf_prog_show_fdinfo,
1866#endif
1867 .release = bpf_prog_release,
6e71b04a
CF		 1868 .read = bpf_dummy_read,
1869 .write = bpf_dummy_write,
09756af4
AS		 1870};
1871
b2197755 1872int bpf_prog_new_fd(struct bpf_prog *prog)
aa79781b 1873{
afdb09c7
CF		 1874 int ret;
1875
1876 ret = security_bpf_prog(prog);
1877 if (ret < 0)
1878 return ret;
1879
aa79781b
DB		 1880 return anon_inode_getfd("bpf-prog", &bpf_prog_fops, prog,
1881 O_RDWR | O_CLOEXEC);
1882}
1883
113214be 1884static struct bpf_prog *____bpf_prog_get(struct fd f)
09756af4 1885{
09756af4
AS		 1886 if (!f.file)
1887 return ERR_PTR(-EBADF);
09756af4
AS		 1888 if (f.file->f_op != &bpf_prog_fops) {
1889 fdput(f);
1890 return ERR_PTR(-EINVAL);
1891 }
1892
c2101297 1893 return f.file->private_data;
09756af4
AS		 1894}
1895
85192dbf 1896void bpf_prog_add(struct bpf_prog *prog, int i)
92117d84 1897{
85192dbf 1898 atomic64_add(i, &prog->aux->refcnt);
92117d84 1899}
59d3656d
BB		 1900EXPORT_SYMBOL_GPL(bpf_prog_add);
1901
c540594f
DB		 1902void bpf_prog_sub(struct bpf_prog *prog, int i)
1903{
1904 /* Only to be used for undoing previous bpf_prog_add() in some
1905 * error path. We still know that another entity in our call
1906 * path holds a reference to the program, thus atomic_sub() can
1907 * be safely used in such cases!
1908 */
85192dbf 1909 WARN_ON(atomic64_sub_return(i, &prog->aux->refcnt) == 0);
c540594f
DB		 1910}
1911EXPORT_SYMBOL_GPL(bpf_prog_sub);
1912
85192dbf 1913void bpf_prog_inc(struct bpf_prog *prog)
59d3656d 1914{
85192dbf 1915 atomic64_inc(&prog->aux->refcnt);
59d3656d 1916}
97bc402d 1917EXPORT_SYMBOL_GPL(bpf_prog_inc);
92117d84 1918
b16d9aa4 1919/* prog_idr_lock should have been held */
a6f6df69 1920struct bpf_prog *bpf_prog_inc_not_zero(struct bpf_prog *prog)
b16d9aa4
MKL		 1921{
1922 int refold;
1923
85192dbf 1924 refold = atomic64_fetch_add_unless(&prog->aux->refcnt, 1, 0);
b16d9aa4
MKL		 1925
1926 if (!refold)
1927 return ERR_PTR(-ENOENT);
1928
1929 return prog;
1930}
a6f6df69 1931EXPORT_SYMBOL_GPL(bpf_prog_inc_not_zero);
b16d9aa4 1932
040ee692 1933bool bpf_prog_get_ok(struct bpf_prog *prog,
288b3de5 1934 enum bpf_prog_type *attach_type, bool attach_drv)
248f346f 1935{
288b3de5
JK		 1936 /* not an attachment, just a refcount inc, always allow */
1937 if (!attach_type)
1938 return true;
248f346f
JK		 1939
1940 if (prog->type != *attach_type)
1941 return false;
288b3de5 1942 if (bpf_prog_is_dev_bound(prog->aux) && !attach_drv)
248f346f
JK		 1943 return false;
1944
1945 return true;
1946}
1947
1948static struct bpf_prog *__bpf_prog_get(u32 ufd, enum bpf_prog_type *attach_type,
288b3de5 1949 bool attach_drv)
09756af4
AS		 1950{
1951 struct fd f = fdget(ufd);
1952 struct bpf_prog *prog;
1953
113214be 1954 prog = ____bpf_prog_get(f);
09756af4
AS		 1955 if (IS_ERR(prog))
1956 return prog;
288b3de5 1957 if (!bpf_prog_get_ok(prog, attach_type, attach_drv)) {
113214be
DB		 1958 prog = ERR_PTR(-EINVAL);
1959 goto out;
1960 }
09756af4 1961
85192dbf 1962 bpf_prog_inc(prog);
113214be 1963out:
09756af4
AS		 1964 fdput(f);
1965 return prog;
1966}
113214be
DB		 1967
1968struct bpf_prog *bpf_prog_get(u32 ufd)
1969{
288b3de5 1970 return __bpf_prog_get(ufd, NULL, false);
113214be
DB		 1971}
1972
248f346f 1973struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
288b3de5 1974 bool attach_drv)
248f346f 1975{
4d220ed0 1976 return __bpf_prog_get(ufd, &type, attach_drv);
248f346f 1977}
6c8dfe21 1978EXPORT_SYMBOL_GPL(bpf_prog_get_type_dev);
248f346f 1979
aac3fc32
AI		 1980/* Initially all BPF programs could be loaded w/o specifying
1981 * expected_attach_type. Later for some of them specifying expected_attach_type
1982 * at load time became required so that program could be validated properly.
1983 * Programs of types that are allowed to be loaded both w/ and w/o (for
1984 * backward compatibility) expected_attach_type, should have the default attach
1985 * type assigned to expected_attach_type for the latter case, so that it can be
1986 * validated later at attach time.
1987 *
1988 * bpf_prog_load_fixup_attach_type() sets expected_attach_type in @attr if
1989 * prog type requires it but has some attach types that have to be backward
1990 * compatible.
1991 */
1992static void bpf_prog_load_fixup_attach_type(union bpf_attr *attr)
1993{
1994 switch (attr->prog_type) {
1995 case BPF_PROG_TYPE_CGROUP_SOCK:
1996 /* Unfortunately BPF_ATTACH_TYPE_UNSPEC enumeration doesn't
1997 * exist so checking for non-zero is the way to go here.
1998 */
1999 if (!attr->expected_attach_type)
2000 attr->expected_attach_type =
2001 BPF_CGROUP_INET_SOCK_CREATE;
2002 break;
d5e4ddae
KI		 2003 case BPF_PROG_TYPE_SK_REUSEPORT:
2004 if (!attr->expected_attach_type)
2005 attr->expected_attach_type =
2006 BPF_SK_REUSEPORT_SELECT;
2007 break;
aac3fc32
AI		 2008 }
2009}
2010
5e43f899 2011static int
ccfe29eb
AS		 2012bpf_prog_load_check_attach(enum bpf_prog_type prog_type,
2013 enum bpf_attach_type expected_attach_type,
290248a5
AN		 2014 struct btf *attach_btf, u32 btf_id,
2015 struct bpf_prog *dst_prog)
5e43f899 2016{
27ae7997 2017 if (btf_id) {
c108e3c1
AS		 2018 if (btf_id > BTF_MAX_TYPE)
2019 return -EINVAL;
27ae7997 2020
290248a5
AN		 2021 if (!attach_btf && !dst_prog)
2022 return -EINVAL;
2023
27ae7997
MKL		 2024 switch (prog_type) {
2025 case BPF_PROG_TYPE_TRACING:
9e4e01df 2026 case BPF_PROG_TYPE_LSM:
27ae7997 2027 case BPF_PROG_TYPE_STRUCT_OPS:
be8704ff 2028 case BPF_PROG_TYPE_EXT:
27ae7997
MKL		 2029 break;
2030 default:
c108e3c1 2031 return -EINVAL;
27ae7997 2032 }
c108e3c1
AS		 2033 }
2034
290248a5
AN		 2035 if (attach_btf && (!btf_id || dst_prog))
2036 return -EINVAL;
2037
2038 if (dst_prog && prog_type != BPF_PROG_TYPE_TRACING &&
be8704ff 2039 prog_type != BPF_PROG_TYPE_EXT)
27ae7997
MKL		 2040 return -EINVAL;
2041
4fbac77d 2042 switch (prog_type) {
aac3fc32
AI		 2043 case BPF_PROG_TYPE_CGROUP_SOCK:
2044 switch (expected_attach_type) {
2045 case BPF_CGROUP_INET_SOCK_CREATE:
f5836749 2046 case BPF_CGROUP_INET_SOCK_RELEASE:
aac3fc32
AI		 2047 case BPF_CGROUP_INET4_POST_BIND:
2048 case BPF_CGROUP_INET6_POST_BIND:
2049 return 0;
2050 default:
2051 return -EINVAL;
2052 }
4fbac77d
AI		 2053 case BPF_PROG_TYPE_CGROUP_SOCK_ADDR:
2054 switch (expected_attach_type) {
2055 case BPF_CGROUP_INET4_BIND:
2056 case BPF_CGROUP_INET6_BIND:
d74bad4e
AI		 2057 case BPF_CGROUP_INET4_CONNECT:
2058 case BPF_CGROUP_INET6_CONNECT:
1b66d253
DB		 2059 case BPF_CGROUP_INET4_GETPEERNAME:
2060 case BPF_CGROUP_INET6_GETPEERNAME:
2061 case BPF_CGROUP_INET4_GETSOCKNAME:
2062 case BPF_CGROUP_INET6_GETSOCKNAME:
1cedee13
AI		 2063 case BPF_CGROUP_UDP4_SENDMSG:
2064 case BPF_CGROUP_UDP6_SENDMSG:
983695fa
DB		 2065 case BPF_CGROUP_UDP4_RECVMSG:
2066 case BPF_CGROUP_UDP6_RECVMSG:
4fbac77d
AI		 2067 return 0;
2068 default:
2069 return -EINVAL;
2070 }
5cf1e914 2071 case BPF_PROG_TYPE_CGROUP_SKB:
2072 switch (expected_attach_type) {
2073 case BPF_CGROUP_INET_INGRESS:
2074 case BPF_CGROUP_INET_EGRESS:
2075 return 0;
2076 default:
2077 return -EINVAL;
2078 }
0d01da6a
SF		 2079 case BPF_PROG_TYPE_CGROUP_SOCKOPT:
2080 switch (expected_attach_type) {
2081 case BPF_CGROUP_SETSOCKOPT:
2082 case BPF_CGROUP_GETSOCKOPT:
2083 return 0;
2084 default:
2085 return -EINVAL;
2086 }
e9ddbb77
JS		 2087 case BPF_PROG_TYPE_SK_LOOKUP:
2088 if (expected_attach_type == BPF_SK_LOOKUP)
2089 return 0;
2090 return -EINVAL;
d5e4ddae
KI		 2091 case BPF_PROG_TYPE_SK_REUSEPORT:
2092 switch (expected_attach_type) {
2093 case BPF_SK_REUSEPORT_SELECT:
2094 case BPF_SK_REUSEPORT_SELECT_OR_MIGRATE:
2095 return 0;
2096 default:
2097 return -EINVAL;
2098 }
79a7f8bd 2099 case BPF_PROG_TYPE_SYSCALL:
be8704ff
AS		 2100 case BPF_PROG_TYPE_EXT:
2101 if (expected_attach_type)
2102 return -EINVAL;
df561f66 2103 fallthrough;
4fbac77d
AI		 2104 default:
2105 return 0;
2106 }
5e43f899
AI		 2107}
2108
2c78ee89
AS		 2109static bool is_net_admin_prog_type(enum bpf_prog_type prog_type)
2110{
2111 switch (prog_type) {
2112 case BPF_PROG_TYPE_SCHED_CLS:
2113 case BPF_PROG_TYPE_SCHED_ACT:
2114 case BPF_PROG_TYPE_XDP:
2115 case BPF_PROG_TYPE_LWT_IN:
2116 case BPF_PROG_TYPE_LWT_OUT:
2117 case BPF_PROG_TYPE_LWT_XMIT:
2118 case BPF_PROG_TYPE_LWT_SEG6LOCAL:
2119 case BPF_PROG_TYPE_SK_SKB:
2120 case BPF_PROG_TYPE_SK_MSG:
2121 case BPF_PROG_TYPE_LIRC_MODE2:
2122 case BPF_PROG_TYPE_FLOW_DISSECTOR:
2123 case BPF_PROG_TYPE_CGROUP_DEVICE:
2124 case BPF_PROG_TYPE_CGROUP_SOCK:
2125 case BPF_PROG_TYPE_CGROUP_SOCK_ADDR:
2126 case BPF_PROG_TYPE_CGROUP_SOCKOPT:
2127 case BPF_PROG_TYPE_CGROUP_SYSCTL:
2128 case BPF_PROG_TYPE_SOCK_OPS:
2129 case BPF_PROG_TYPE_EXT: /* extends any prog */
2130 return true;
2131 case BPF_PROG_TYPE_CGROUP_SKB:
2132 /* always unpriv */
2133 case BPF_PROG_TYPE_SK_REUSEPORT:
2134 /* equivalent to SOCKET_FILTER. need CAP_BPF only */
2135 default:
2136 return false;
2137 }
2138}
2139
2140static bool is_perfmon_prog_type(enum bpf_prog_type prog_type)
2141{
2142 switch (prog_type) {
2143 case BPF_PROG_TYPE_KPROBE:
2144 case BPF_PROG_TYPE_TRACEPOINT:
2145 case BPF_PROG_TYPE_PERF_EVENT:
2146 case BPF_PROG_TYPE_RAW_TRACEPOINT:
2147 case BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE:
2148 case BPF_PROG_TYPE_TRACING:
2149 case BPF_PROG_TYPE_LSM:
2150 case BPF_PROG_TYPE_STRUCT_OPS: /* has access to struct sock */
2151 case BPF_PROG_TYPE_EXT: /* extends any prog */
2152 return true;
2153 default:
2154 return false;
2155 }
2156}
2157
09756af4 2158/* last field in 'union bpf_attr' used by this command */
387544bf 2159#define BPF_PROG_LOAD_LAST_FIELD fd_array
09756af4 2160
af2ac3e1 2161static int bpf_prog_load(union bpf_attr *attr, bpfptr_t uattr)
09756af4
AS		 2162{
2163 enum bpf_prog_type type = attr->prog_type;
290248a5
AN		 2164 struct bpf_prog *prog, *dst_prog = NULL;
2165 struct btf *attach_btf = NULL;
09756af4
AS		 2166 int err;
2167 char license[128];
2168 bool is_gpl;
2169
2170 if (CHECK_ATTR(BPF_PROG_LOAD))
2171 return -EINVAL;
2172
c240eff6
JW		 2173 if (attr->prog_flags & ~(BPF_F_STRICT_ALIGNMENT |
2174 BPF_F_ANY_ALIGNMENT |
10d274e8 2175 BPF_F_TEST_STATE_FREQ |
1e6c62a8 2176 BPF_F_SLEEPABLE |
c240eff6 2177 BPF_F_TEST_RND_HI32))
e07b98d9
DM		 2178 return -EINVAL;
2179
e9ee9efc
DM		 2180 if (!IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) &&
2181 (attr->prog_flags & BPF_F_ANY_ALIGNMENT) &&
2c78ee89 2182 !bpf_capable())
e9ee9efc
DM		 2183 return -EPERM;
2184
09756af4 2185 /* copy eBPF program license from user space */
af2ac3e1
AS		 2186 if (strncpy_from_bpfptr(license,
2187 make_bpfptr(attr->license, uattr.is_kernel),
2188 sizeof(license) - 1) < 0)
09756af4
AS		 2189 return -EFAULT;
2190 license[sizeof(license) - 1] = 0;
2191
2192 /* eBPF programs must be GPL compatible to use GPL-ed functions */
2193 is_gpl = license_is_gpl_compatible(license);
2194
c04c0d2b 2195 if (attr->insn_cnt == 0 ||
2c78ee89 2196 attr->insn_cnt > (bpf_capable() ? BPF_COMPLEXITY_LIMIT_INSNS : BPF_MAXINSNS))
ef0915ca 2197 return -E2BIG;
80b7d819
CF		 2198 if (type != BPF_PROG_TYPE_SOCKET_FILTER &&
2199 type != BPF_PROG_TYPE_CGROUP_SKB &&
2c78ee89
AS		 2200 !bpf_capable())
2201 return -EPERM;
2202
b338cb92 2203 if (is_net_admin_prog_type(type) && !capable(CAP_NET_ADMIN) && !capable(CAP_SYS_ADMIN))
2c78ee89
AS		 2204 return -EPERM;
2205 if (is_perfmon_prog_type(type) && !perfmon_capable())
1be7f75d
AS		 2206 return -EPERM;
2207
290248a5
AN		 2208 /* attach_prog_fd/attach_btf_obj_fd can specify fd of either bpf_prog
2209 * or btf, we need to check which one it is
2210 */
2211 if (attr->attach_prog_fd) {
2212 dst_prog = bpf_prog_get(attr->attach_prog_fd);
2213 if (IS_ERR(dst_prog)) {
2214 dst_prog = NULL;
2215 attach_btf = btf_get_by_fd(attr->attach_btf_obj_fd);
2216 if (IS_ERR(attach_btf))
2217 return -EINVAL;
2218 if (!btf_is_kernel(attach_btf)) {
8bdd8e27
AN		 2219 /* attaching through specifying bpf_prog's BTF
2220 * objects directly might be supported eventually
2221 */
290248a5 2222 btf_put(attach_btf);
8bdd8e27 2223 return -ENOTSUPP;
290248a5
AN		 2224 }
2225 }
2226 } else if (attr->attach_btf_id) {
2227 /* fall back to vmlinux BTF, if BTF type ID is specified */
2228 attach_btf = bpf_get_btf_vmlinux();
2229 if (IS_ERR(attach_btf))
2230 return PTR_ERR(attach_btf);
2231 if (!attach_btf)
2232 return -EINVAL;
2233 btf_get(attach_btf);
2234 }
2235
aac3fc32 2236 bpf_prog_load_fixup_attach_type(attr);
ccfe29eb 2237 if (bpf_prog_load_check_attach(type, attr->expected_attach_type,
290248a5
AN		 2238 attach_btf, attr->attach_btf_id,
2239 dst_prog)) {
2240 if (dst_prog)
2241 bpf_prog_put(dst_prog);
2242 if (attach_btf)
2243 btf_put(attach_btf);
5e43f899 2244 return -EINVAL;
290248a5 2245 }
5e43f899 2246
09756af4
AS		 2247 /* plain bpf_prog allocation */
2248 prog = bpf_prog_alloc(bpf_prog_size(attr->insn_cnt), GFP_USER);
290248a5
AN		 2249 if (!prog) {
2250 if (dst_prog)
2251 bpf_prog_put(dst_prog);
2252 if (attach_btf)
2253 btf_put(attach_btf);
09756af4 2254 return -ENOMEM;
290248a5 2255 }
09756af4 2256
5e43f899 2257 prog->expected_attach_type = attr->expected_attach_type;
290248a5 2258 prog->aux->attach_btf = attach_btf;
ccfe29eb 2259 prog->aux->attach_btf_id = attr->attach_btf_id;
290248a5 2260 prog->aux->dst_prog = dst_prog;
9a18eedb 2261 prog->aux->offload_requested = !!attr->prog_ifindex;
1e6c62a8 2262 prog->aux->sleepable = attr->prog_flags & BPF_F_SLEEPABLE;
9a18eedb 2263
afdb09c7 2264 err = security_bpf_prog_alloc(prog->aux);
aaac3ba9 2265 if (err)
3ac1f01b 2266 goto free_prog;
afdb09c7 2267
3ac1f01b 2268 prog->aux->user = get_current_user();
09756af4
AS		 2269 prog->len = attr->insn_cnt;
2270
2271 err = -EFAULT;
af2ac3e1
AS		 2272 if (copy_from_bpfptr(prog->insns,
2273 make_bpfptr(attr->insns, uattr.is_kernel),
2274 bpf_prog_insn_size(prog)) != 0)
3ac1f01b 2275 goto free_prog_sec;
09756af4
AS		 2276
2277 prog->orig_prog = NULL;
a91263d5 2278 prog->jited = 0;
09756af4 2279
85192dbf 2280 atomic64_set(&prog->aux->refcnt, 1);
a91263d5 2281 prog->gpl_compatible = is_gpl ? 1 : 0;
09756af4 2282
9a18eedb 2283 if (bpf_prog_is_dev_bound(prog->aux)) {
ab3f0063
JK		 2284 err = bpf_prog_offload_init(prog, attr);
2285 if (err)
3ac1f01b 2286 goto free_prog_sec;
ab3f0063
JK		 2287 }
2288
09756af4
AS		 2289 /* find program type: socket_filter vs tracing_filter */
2290 err = find_prog_type(type, prog);
2291 if (err < 0)
3ac1f01b 2292 goto free_prog_sec;
09756af4 2293
9285ec4c 2294 prog->aux->load_time = ktime_get_boottime_ns();
8e7ae251
MKL		 2295 err = bpf_obj_name_cpy(prog->aux->name, attr->prog_name,
2296 sizeof(attr->prog_name));
2297 if (err < 0)
3ac1f01b 2298 goto free_prog_sec;
cb4d2b3f 2299
09756af4 2300 /* run eBPF verifier */
838e9690 2301 err = bpf_check(&prog, attr, uattr);
09756af4
AS		 2302 if (err < 0)
2303 goto free_used_maps;
2304
9facc336 2305 prog = bpf_prog_select_runtime(prog, &err);
04fd61ab
AS		 2306 if (err < 0)
2307 goto free_used_maps;
09756af4 2308
dc4bb0e2
MKL		 2309 err = bpf_prog_alloc_id(prog);
2310 if (err)
2311 goto free_used_maps;
2312
c751798a
DB		 2313 /* Upon success of bpf_prog_alloc_id(), the BPF prog is
2314 * effectively publicly exposed. However, retrieving via
2315 * bpf_prog_get_fd_by_id() will take another reference,
2316 * therefore it cannot be gone underneath us.
2317 *
2318 * Only for the time /after/ successful bpf_prog_new_fd()
2319 * and before returning to userspace, we might just hold
2320 * one reference and any parallel close on that fd could
2321 * rip everything out. Hence, below notifications must
2322 * happen before bpf_prog_new_fd().
2323 *
2324 * Also, any failure handling from this point onwards must
2325 * be using bpf_prog_put() given the program is exposed.
2326 */
74451e66 2327 bpf_prog_kallsyms_add(prog);
6ee52e2a 2328 perf_event_bpf_event(prog, PERF_BPF_EVENT_PROG_LOAD, 0);
bae141f5 2329 bpf_audit_prog(prog, BPF_AUDIT_LOAD);
c751798a
DB		 2330
2331 err = bpf_prog_new_fd(prog);
2332 if (err < 0)
2333 bpf_prog_put(prog);
09756af4
AS		 2334 return err;
2335
2336free_used_maps:
cd7455f1
DB		 2337 /* In case we have subprogs, we need to wait for a grace
2338 * period before we can tear down JIT memory since symbols
2339 * are already exposed under kallsyms.
2340 */
2341 __bpf_prog_put_noref(prog, prog->aux->func_cnt);
2342 return err;
afdb09c7 2343free_prog_sec:
3ac1f01b 2344 free_uid(prog->aux->user);
afdb09c7 2345 security_bpf_prog_free(prog->aux);
3ac1f01b 2346free_prog:
22dc4a0f
AN		 2347 if (prog->aux->attach_btf)
2348 btf_put(prog->aux->attach_btf);
09756af4
AS		 2349 bpf_prog_free(prog);
2350 return err;
2351}
2352
6e71b04a 2353#define BPF_OBJ_LAST_FIELD file_flags
b2197755
DB		 2354
2355static int bpf_obj_pin(const union bpf_attr *attr)
2356{
6e71b04a 2357 if (CHECK_ATTR(BPF_OBJ) || attr->file_flags != 0)
b2197755
DB		 2358 return -EINVAL;
2359
535e7b4b 2360 return bpf_obj_pin_user(attr->bpf_fd, u64_to_user_ptr(attr->pathname));
b2197755
DB		 2361}
2362
2363static int bpf_obj_get(const union bpf_attr *attr)
2364{
6e71b04a
CF		 2365 if (CHECK_ATTR(BPF_OBJ) || attr->bpf_fd != 0 ||
2366 attr->file_flags & ~BPF_OBJ_FLAG_MASK)
b2197755
DB		 2367 return -EINVAL;
2368
6e71b04a
CF		 2369 return bpf_obj_get_user(u64_to_user_ptr(attr->pathname),
2370 attr->file_flags);
b2197755
DB		 2371}
2372
f2e10bff 2373void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
a3b80e10 2374 const struct bpf_link_ops *ops, struct bpf_prog *prog)
fec56f58 2375{
70ed506c 2376 atomic64_set(&link->refcnt, 1);
f2e10bff 2377 link->type = type;
a3b80e10 2378 link->id = 0;
70ed506c
AN		 2379 link->ops = ops;
2380 link->prog = prog;
2381}
2382
a3b80e10
AN		 2383static void bpf_link_free_id(int id)
2384{
2385 if (!id)
2386 return;
2387
2388 spin_lock_bh(&link_idr_lock);
2389 idr_remove(&link_idr, id);
2390 spin_unlock_bh(&link_idr_lock);
2391}
2392
98868668
AN		 2393/* Clean up bpf_link and corresponding anon_inode file and FD. After
2394 * anon_inode is created, bpf_link can't be just kfree()'d due to deferred
a3b80e10
AN		 2395 * anon_inode's release() call. This helper marksbpf_link as
2396 * defunct, releases anon_inode file and puts reserved FD. bpf_prog's refcnt
2397 * is not decremented, it's the responsibility of a calling code that failed
2398 * to complete bpf_link initialization.
98868668 2399 */
a3b80e10 2400void bpf_link_cleanup(struct bpf_link_primer *primer)
babf3164 2401{
a3b80e10
AN		 2402 primer->link->prog = NULL;
2403 bpf_link_free_id(primer->id);
2404 fput(primer->file);
2405 put_unused_fd(primer->fd);
babf3164
AN		 2406}
2407
70ed506c
AN		 2408void bpf_link_inc(struct bpf_link *link)
2409{
2410 atomic64_inc(&link->refcnt);
2411}
2412
2413/* bpf_link_free is guaranteed to be called from process context */
2414static void bpf_link_free(struct bpf_link *link)
2415{
a3b80e10 2416 bpf_link_free_id(link->id);
babf3164
AN		 2417 if (link->prog) {
2418 /* detach BPF program, clean up used resources */
2419 link->ops->release(link);
2420 bpf_prog_put(link->prog);
2421 }
2422 /* free bpf_link and its containing memory */
2423 link->ops->dealloc(link);
70ed506c
AN		 2424}
2425
2426static void bpf_link_put_deferred(struct work_struct *work)
2427{
2428 struct bpf_link *link = container_of(work, struct bpf_link, work);
2429
2430 bpf_link_free(link);
2431}
2432
2433/* bpf_link_put can be called from atomic context, but ensures that resources
2434 * are freed from process context
2435 */
2436void bpf_link_put(struct bpf_link *link)
2437{
2438 if (!atomic64_dec_and_test(&link->refcnt))
2439 return;
2440
f00f2f7f
AS		 2441 if (in_atomic()) {
2442 INIT_WORK(&link->work, bpf_link_put_deferred);
2443 schedule_work(&link->work);
2444 } else {
2445 bpf_link_free(link);
2446 }
70ed506c
AN		 2447}
2448
2449static int bpf_link_release(struct inode *inode, struct file *filp)
2450{
2451 struct bpf_link *link = filp->private_data;
2452
2453 bpf_link_put(link);
fec56f58
AS		 2454 return 0;
2455}
2456
70ed506c 2457#ifdef CONFIG_PROC_FS
f2e10bff
AN		 2458#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type)
2459#define BPF_MAP_TYPE(_id, _ops)
2460#define BPF_LINK_TYPE(_id, _name) [_id] = #_name,
2461static const char *bpf_link_type_strs[] = {
2462 [BPF_LINK_TYPE_UNSPEC] = "<invalid>",
2463#include <linux/bpf_types.h>
2464};
2465#undef BPF_PROG_TYPE
2466#undef BPF_MAP_TYPE
2467#undef BPF_LINK_TYPE
70ed506c
AN		 2468
2469static void bpf_link_show_fdinfo(struct seq_file *m, struct file *filp)
2470{
2471 const struct bpf_link *link = filp->private_data;
2472 const struct bpf_prog *prog = link->prog;
2473 char prog_tag[sizeof(prog->tag) * 2 + 1] = { };
70ed506c
AN		 2474
2475 bin2hex(prog_tag, prog->tag, sizeof(prog->tag));
2476 seq_printf(m,
2477 "link_type:\t%s\n"
a3b80e10 2478 "link_id:\t%u\n"
70ed506c
AN		 2479 "prog_tag:\t%s\n"
2480 "prog_id:\t%u\n",
f2e10bff 2481 bpf_link_type_strs[link->type],
a3b80e10 2482 link->id,
70ed506c
AN		 2483 prog_tag,
2484 prog->aux->id);
f2e10bff
AN		 2485 if (link->ops->show_fdinfo)
2486 link->ops->show_fdinfo(link, m);
70ed506c
AN		 2487}
2488#endif
2489
6f302bfb 2490static const struct file_operations bpf_link_fops = {
70ed506c
AN		 2491#ifdef CONFIG_PROC_FS
2492 .show_fdinfo = bpf_link_show_fdinfo,
2493#endif
2494 .release = bpf_link_release,
fec56f58
AS		 2495 .read = bpf_dummy_read,
2496 .write = bpf_dummy_write,
2497};
2498
a3b80e10 2499static int bpf_link_alloc_id(struct bpf_link *link)
70ed506c 2500{
a3b80e10 2501 int id;
70ed506c 2502
a3b80e10
AN		 2503 idr_preload(GFP_KERNEL);
2504 spin_lock_bh(&link_idr_lock);
2505 id = idr_alloc_cyclic(&link_idr, link, 1, INT_MAX, GFP_ATOMIC);
2506 spin_unlock_bh(&link_idr_lock);
2507 idr_preload_end();
70ed506c 2508
a3b80e10
AN		 2509 return id;
2510}
2511
2512/* Prepare bpf_link to be exposed to user-space by allocating anon_inode file,
2513 * reserving unused FD and allocating ID from link_idr. This is to be paired
2514 * with bpf_link_settle() to install FD and ID and expose bpf_link to
2515 * user-space, if bpf_link is successfully attached. If not, bpf_link and
2516 * pre-allocated resources are to be freed with bpf_cleanup() call. All the
2517 * transient state is passed around in struct bpf_link_primer.
2518 * This is preferred way to create and initialize bpf_link, especially when
2519 * there are complicated and expensive operations inbetween creating bpf_link
2520 * itself and attaching it to BPF hook. By using bpf_link_prime() and
2521 * bpf_link_settle() kernel code using bpf_link doesn't have to perform
2522 * expensive (and potentially failing) roll back operations in a rare case
2523 * that file, FD, or ID can't be allocated.
babf3164 2524 */
a3b80e10 2525int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer)
babf3164
AN		 2526{
2527 struct file *file;
a3b80e10 2528 int fd, id;
babf3164
AN		 2529
2530 fd = get_unused_fd_flags(O_CLOEXEC);
2531 if (fd < 0)
a3b80e10 2532 return fd;
babf3164 2533
babf3164 2534
a3b80e10
AN		 2535 id = bpf_link_alloc_id(link);
2536 if (id < 0) {
2537 put_unused_fd(fd);
a3b80e10
AN		 2538 return id;
2539 }
babf3164
AN		 2540
2541 file = anon_inode_getfile("bpf_link", &bpf_link_fops, link, O_CLOEXEC);
2542 if (IS_ERR(file)) {
138c6767 2543 bpf_link_free_id(id);
babf3164 2544 put_unused_fd(fd);
138c6767 2545 return PTR_ERR(file);
babf3164
AN		 2546 }
2547
a3b80e10
AN		 2548 primer->link = link;
2549 primer->file = file;
2550 primer->fd = fd;
2551 primer->id = id;
2552 return 0;
2553}
2554
2555int bpf_link_settle(struct bpf_link_primer *primer)
2556{
2557 /* make bpf_link fetchable by ID */
2558 spin_lock_bh(&link_idr_lock);
2559 primer->link->id = primer->id;
2560 spin_unlock_bh(&link_idr_lock);
2561 /* make bpf_link fetchable by FD */
2562 fd_install(primer->fd, primer->file);
2563 /* pass through installed FD */
2564 return primer->fd;
2565}
2566
2567int bpf_link_new_fd(struct bpf_link *link)
2568{
2569 return anon_inode_getfd("bpf-link", &bpf_link_fops, link, O_CLOEXEC);
babf3164
AN		 2570}
2571
70ed506c
AN		 2572struct bpf_link *bpf_link_get_from_fd(u32 ufd)
2573{
2574 struct fd f = fdget(ufd);
2575 struct bpf_link *link;
2576
2577 if (!f.file)
2578 return ERR_PTR(-EBADF);
2579 if (f.file->f_op != &bpf_link_fops) {
2580 fdput(f);
2581 return ERR_PTR(-EINVAL);
2582 }
2583
2584 link = f.file->private_data;
2585 bpf_link_inc(link);
2586 fdput(f);
2587
2588 return link;
2589}
2590
2591struct bpf_tracing_link {
2592 struct bpf_link link;
f2e10bff 2593 enum bpf_attach_type attach_type;
3aac1ead
THJ		 2594 struct bpf_trampoline *trampoline;
2595 struct bpf_prog *tgt_prog;
70ed506c
AN		 2596};
2597
2598static void bpf_tracing_link_release(struct bpf_link *link)
babf3164 2599{
3aac1ead
THJ		 2600 struct bpf_tracing_link *tr_link =
2601 container_of(link, struct bpf_tracing_link, link);
2602
2603 WARN_ON_ONCE(bpf_trampoline_unlink_prog(link->prog,
2604 tr_link->trampoline));
2605
2606 bpf_trampoline_put(tr_link->trampoline);
2607
2608 /* tgt_prog is NULL if target is a kernel function */
2609 if (tr_link->tgt_prog)
2610 bpf_prog_put(tr_link->tgt_prog);
babf3164
AN		 2611}
2612
2613static void bpf_tracing_link_dealloc(struct bpf_link *link)
70ed506c
AN		 2614{
2615 struct bpf_tracing_link *tr_link =
2616 container_of(link, struct bpf_tracing_link, link);
2617
70ed506c
AN		 2618 kfree(tr_link);
2619}
2620
f2e10bff
AN		 2621static void bpf_tracing_link_show_fdinfo(const struct bpf_link *link,
2622 struct seq_file *seq)
2623{
2624 struct bpf_tracing_link *tr_link =
2625 container_of(link, struct bpf_tracing_link, link);
2626
2627 seq_printf(seq,
2628 "attach_type:\t%d\n",
2629 tr_link->attach_type);
2630}
2631
2632static int bpf_tracing_link_fill_link_info(const struct bpf_link *link,
2633 struct bpf_link_info *info)
2634{
2635 struct bpf_tracing_link *tr_link =
2636 container_of(link, struct bpf_tracing_link, link);
2637
2638 info->tracing.attach_type = tr_link->attach_type;
441e8c66
THJ		 2639 bpf_trampoline_unpack_key(tr_link->trampoline->key,
2640 &info->tracing.target_obj_id,
2641 &info->tracing.target_btf_id);
f2e10bff
AN		 2642
2643 return 0;
2644}
2645
70ed506c
AN		 2646static const struct bpf_link_ops bpf_tracing_link_lops = {
2647 .release = bpf_tracing_link_release,
babf3164 2648 .dealloc = bpf_tracing_link_dealloc,
f2e10bff
AN		 2649 .show_fdinfo = bpf_tracing_link_show_fdinfo,
2650 .fill_link_info = bpf_tracing_link_fill_link_info,
70ed506c
AN		 2651};
2652
4a1e7c0c
THJ		 2653static int bpf_tracing_prog_attach(struct bpf_prog *prog,
2654 int tgt_prog_fd,
2655 u32 btf_id)
fec56f58 2656{
a3b80e10 2657 struct bpf_link_primer link_primer;
3aac1ead 2658 struct bpf_prog *tgt_prog = NULL;
4a1e7c0c 2659 struct bpf_trampoline *tr = NULL;
70ed506c 2660 struct bpf_tracing_link *link;
4a1e7c0c 2661 u64 key = 0;
a3b80e10 2662 int err;
fec56f58 2663
9e4e01df
KS		 2664 switch (prog->type) {
2665 case BPF_PROG_TYPE_TRACING:
2666 if (prog->expected_attach_type != BPF_TRACE_FENTRY &&
2667 prog->expected_attach_type != BPF_TRACE_FEXIT &&
2668 prog->expected_attach_type != BPF_MODIFY_RETURN) {
2669 err = -EINVAL;
2670 goto out_put_prog;
2671 }
2672 break;
2673 case BPF_PROG_TYPE_EXT:
2674 if (prog->expected_attach_type != 0) {
2675 err = -EINVAL;
2676 goto out_put_prog;
2677 }
2678 break;
2679 case BPF_PROG_TYPE_LSM:
2680 if (prog->expected_attach_type != BPF_LSM_MAC) {
2681 err = -EINVAL;
2682 goto out_put_prog;
2683 }
2684 break;
2685 default:
fec56f58
AS		 2686 err = -EINVAL;
2687 goto out_put_prog;
2688 }
2689
4a1e7c0c
THJ		 2690 if (!!tgt_prog_fd != !!btf_id) {
2691 err = -EINVAL;
2692 goto out_put_prog;
2693 }
2694
2695 if (tgt_prog_fd) {
2696 /* For now we only allow new targets for BPF_PROG_TYPE_EXT */
2697 if (prog->type != BPF_PROG_TYPE_EXT) {
2698 err = -EINVAL;
2699 goto out_put_prog;
2700 }
2701
2702 tgt_prog = bpf_prog_get(tgt_prog_fd);
2703 if (IS_ERR(tgt_prog)) {
2704 err = PTR_ERR(tgt_prog);
2705 tgt_prog = NULL;
2706 goto out_put_prog;
2707 }
2708
22dc4a0f 2709 key = bpf_trampoline_compute_key(tgt_prog, NULL, btf_id);
4a1e7c0c
THJ		 2710 }
2711
70ed506c
AN		 2712 link = kzalloc(sizeof(*link), GFP_USER);
2713 if (!link) {
2714 err = -ENOMEM;
2715 goto out_put_prog;
2716 }
f2e10bff
AN		 2717 bpf_link_init(&link->link, BPF_LINK_TYPE_TRACING,
2718 &bpf_tracing_link_lops, prog);
2719 link->attach_type = prog->expected_attach_type;
70ed506c 2720
3aac1ead
THJ		 2721 mutex_lock(&prog->aux->dst_mutex);
2722
4a1e7c0c
THJ		 2723 /* There are a few possible cases here:
2724 *
2725 * - if prog->aux->dst_trampoline is set, the program was just loaded
2726 * and not yet attached to anything, so we can use the values stored
2727 * in prog->aux
2728 *
2729 * - if prog->aux->dst_trampoline is NULL, the program has already been
2730 * attached to a target and its initial target was cleared (below)
2731 *
2732 * - if tgt_prog != NULL, the caller specified tgt_prog_fd +
2733 * target_btf_id using the link_create API.
2734 *
2735 * - if tgt_prog == NULL when this function was called using the old
f3a95075
JO		 2736 * raw_tracepoint_open API, and we need a target from prog->aux
2737 *
2738 * - if prog->aux->dst_trampoline and tgt_prog is NULL, the program
2739 * was detached and is going for re-attachment.
4a1e7c0c
THJ		 2740 */
2741 if (!prog->aux->dst_trampoline && !tgt_prog) {
f3a95075
JO		 2742 /*
2743 * Allow re-attach for TRACING and LSM programs. If it's
2744 * currently linked, bpf_trampoline_link_prog will fail.
2745 * EXT programs need to specify tgt_prog_fd, so they
2746 * re-attach in separate code path.
2747 */
2748 if (prog->type != BPF_PROG_TYPE_TRACING &&
2749 prog->type != BPF_PROG_TYPE_LSM) {
2750 err = -EINVAL;
2751 goto out_unlock;
2752 }
2753 btf_id = prog->aux->attach_btf_id;
2754 key = bpf_trampoline_compute_key(NULL, prog->aux->attach_btf, btf_id);
babf3164 2755 }
4a1e7c0c
THJ		 2756
2757 if (!prog->aux->dst_trampoline ||
2758 (key && key != prog->aux->dst_trampoline->key)) {
2759 /* If there is no saved target, or the specified target is
2760 * different from the destination specified at load time, we
2761 * need a new trampoline and a check for compatibility
2762 */
2763 struct bpf_attach_target_info tgt_info = {};
2764
2765 err = bpf_check_attach_target(NULL, prog, tgt_prog, btf_id,
2766 &tgt_info);
2767 if (err)
2768 goto out_unlock;
2769
2770 tr = bpf_trampoline_get(key, &tgt_info);
2771 if (!tr) {
2772 err = -ENOMEM;
2773 goto out_unlock;
2774 }
2775 } else {
2776 /* The caller didn't specify a target, or the target was the
2777 * same as the destination supplied during program load. This
2778 * means we can reuse the trampoline and reference from program
2779 * load time, and there is no need to allocate a new one. This
2780 * can only happen once for any program, as the saved values in
2781 * prog->aux are cleared below.
2782 */
2783 tr = prog->aux->dst_trampoline;
2784 tgt_prog = prog->aux->dst_prog;
2785 }
3aac1ead
THJ		 2786
2787 err = bpf_link_prime(&link->link, &link_primer);
2788 if (err)
2789 goto out_unlock;
fec56f58 2790
3aac1ead 2791 err = bpf_trampoline_link_prog(prog, tr);
babf3164 2792 if (err) {
a3b80e10 2793 bpf_link_cleanup(&link_primer);
3aac1ead
THJ		 2794 link = NULL;
2795 goto out_unlock;
fec56f58 2796 }
babf3164 2797
3aac1ead
THJ		 2798 link->tgt_prog = tgt_prog;
2799 link->trampoline = tr;
2800
4a1e7c0c
THJ		 2801 /* Always clear the trampoline and target prog from prog->aux to make
2802 * sure the original attach destination is not kept alive after a
2803 * program is (re-)attached to another target.
2804 */
2805 if (prog->aux->dst_prog &&
2806 (tgt_prog_fd || tr != prog->aux->dst_trampoline))
2807 /* got extra prog ref from syscall, or attaching to different prog */
2808 bpf_prog_put(prog->aux->dst_prog);
2809 if (prog->aux->dst_trampoline && tr != prog->aux->dst_trampoline)
2810 /* we allocated a new trampoline, so free the old one */
2811 bpf_trampoline_put(prog->aux->dst_trampoline);
2812
3aac1ead
THJ		 2813 prog->aux->dst_prog = NULL;
2814 prog->aux->dst_trampoline = NULL;
2815 mutex_unlock(&prog->aux->dst_mutex);
2816
a3b80e10 2817 return bpf_link_settle(&link_primer);
3aac1ead 2818out_unlock:
4a1e7c0c
THJ		 2819 if (tr && tr != prog->aux->dst_trampoline)
2820 bpf_trampoline_put(tr);
3aac1ead
THJ		 2821 mutex_unlock(&prog->aux->dst_mutex);
2822 kfree(link);
fec56f58 2823out_put_prog:
4a1e7c0c
THJ		 2824 if (tgt_prog_fd && tgt_prog)
2825 bpf_prog_put(tgt_prog);
fec56f58
AS		 2826 return err;
2827}
2828
70ed506c
AN		 2829struct bpf_raw_tp_link {
2830 struct bpf_link link;
c4f6699d 2831 struct bpf_raw_event_map *btp;
c4f6699d
AS		 2832};
2833
70ed506c 2834static void bpf_raw_tp_link_release(struct bpf_link *link)
c4f6699d 2835{
70ed506c
AN		 2836 struct bpf_raw_tp_link *raw_tp =
2837 container_of(link, struct bpf_raw_tp_link, link);
c4f6699d 2838
70ed506c 2839 bpf_probe_unregister(raw_tp->btp, raw_tp->link.prog);
a38d1107 2840 bpf_put_raw_tracepoint(raw_tp->btp);
babf3164
AN		 2841}
2842
2843static void bpf_raw_tp_link_dealloc(struct bpf_link *link)
2844{
2845 struct bpf_raw_tp_link *raw_tp =
2846 container_of(link, struct bpf_raw_tp_link, link);
2847
c4f6699d 2848 kfree(raw_tp);
c4f6699d
AS		 2849}
2850
f2e10bff
AN		 2851static void bpf_raw_tp_link_show_fdinfo(const struct bpf_link *link,
2852 struct seq_file *seq)
2853{
2854 struct bpf_raw_tp_link *raw_tp_link =
2855 container_of(link, struct bpf_raw_tp_link, link);
2856
2857 seq_printf(seq,
2858 "tp_name:\t%s\n",
2859 raw_tp_link->btp->tp->name);
2860}
2861
2862static int bpf_raw_tp_link_fill_link_info(const struct bpf_link *link,
2863 struct bpf_link_info *info)
2864{
2865 struct bpf_raw_tp_link *raw_tp_link =
2866 container_of(link, struct bpf_raw_tp_link, link);
2867 char __user *ubuf = u64_to_user_ptr(info->raw_tracepoint.tp_name);
2868 const char *tp_name = raw_tp_link->btp->tp->name;
2869 u32 ulen = info->raw_tracepoint.tp_name_len;
2870 size_t tp_len = strlen(tp_name);
2871
b474959d 2872 if (!ulen ^ !ubuf)
f2e10bff
AN		 2873 return -EINVAL;
2874
2875 info->raw_tracepoint.tp_name_len = tp_len + 1;
2876
2877 if (!ubuf)
2878 return 0;
2879
2880 if (ulen >= tp_len + 1) {
2881 if (copy_to_user(ubuf, tp_name, tp_len + 1))
2882 return -EFAULT;
2883 } else {
2884 char zero = '\0';
2885
2886 if (copy_to_user(ubuf, tp_name, ulen - 1))
2887 return -EFAULT;
2888 if (put_user(zero, ubuf + ulen - 1))
2889 return -EFAULT;
2890 return -ENOSPC;
2891 }
2892
2893 return 0;
2894}
2895
a3b80e10 2896static const struct bpf_link_ops bpf_raw_tp_link_lops = {
70ed506c 2897 .release = bpf_raw_tp_link_release,
babf3164 2898 .dealloc = bpf_raw_tp_link_dealloc,
f2e10bff
AN		 2899 .show_fdinfo = bpf_raw_tp_link_show_fdinfo,
2900 .fill_link_info = bpf_raw_tp_link_fill_link_info,
c4f6699d
AS		 2901};
2902
b89fbfbb
AN		 2903#ifdef CONFIG_PERF_EVENTS
2904struct bpf_perf_link {
2905 struct bpf_link link;
2906 struct file *perf_file;
2907};
2908
2909static void bpf_perf_link_release(struct bpf_link *link)
2910{
2911 struct bpf_perf_link *perf_link = container_of(link, struct bpf_perf_link, link);
2912 struct perf_event *event = perf_link->perf_file->private_data;
2913
2914 perf_event_free_bpf_prog(event);
2915 fput(perf_link->perf_file);
2916}
2917
2918static void bpf_perf_link_dealloc(struct bpf_link *link)
2919{
2920 struct bpf_perf_link *perf_link = container_of(link, struct bpf_perf_link, link);
2921
2922 kfree(perf_link);
2923}
2924
2925static const struct bpf_link_ops bpf_perf_link_lops = {
2926 .release = bpf_perf_link_release,
2927 .dealloc = bpf_perf_link_dealloc,
2928};
2929
2930static int bpf_perf_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
2931{
2932 struct bpf_link_primer link_primer;
2933 struct bpf_perf_link *link;
2934 struct perf_event *event;
2935 struct file *perf_file;
2936 int err;
2937
2938 if (attr->link_create.flags)
2939 return -EINVAL;
2940
2941 perf_file = perf_event_get(attr->link_create.target_fd);
2942 if (IS_ERR(perf_file))
2943 return PTR_ERR(perf_file);
2944
2945 link = kzalloc(sizeof(*link), GFP_USER);
2946 if (!link) {
2947 err = -ENOMEM;
2948 goto out_put_file;
2949 }
2950 bpf_link_init(&link->link, BPF_LINK_TYPE_PERF_EVENT, &bpf_perf_link_lops, prog);
2951 link->perf_file = perf_file;
2952
2953 err = bpf_link_prime(&link->link, &link_primer);
2954 if (err) {
2955 kfree(link);
2956 goto out_put_file;
2957 }
2958
2959 event = perf_file->private_data;
82e6b1ee 2960 err = perf_event_set_bpf_prog(event, prog, attr->link_create.perf_event.bpf_cookie);
b89fbfbb
AN		 2961 if (err) {
2962 bpf_link_cleanup(&link_primer);
2963 goto out_put_file;
2964 }
2965 /* perf_event_set_bpf_prog() doesn't take its own refcnt on prog */
2966 bpf_prog_inc(prog);
2967
2968 return bpf_link_settle(&link_primer);
2969
2970out_put_file:
2971 fput(perf_file);
2972 return err;
2973}
2974#endif /* CONFIG_PERF_EVENTS */
2975
c4f6699d
AS		 2976#define BPF_RAW_TRACEPOINT_OPEN_LAST_FIELD raw_tracepoint.prog_fd
2977
2978static int bpf_raw_tracepoint_open(const union bpf_attr *attr)
2979{
a3b80e10 2980 struct bpf_link_primer link_primer;
babf3164 2981 struct bpf_raw_tp_link *link;
c4f6699d
AS		 2982 struct bpf_raw_event_map *btp;
2983 struct bpf_prog *prog;
ac4414b5
AS		 2984 const char *tp_name;
2985 char buf[128];
a3b80e10 2986 int err;
c4f6699d 2987
ac4414b5
AS		 2988 if (CHECK_ATTR(BPF_RAW_TRACEPOINT_OPEN))
2989 return -EINVAL;
2990
2991 prog = bpf_prog_get(attr->raw_tracepoint.prog_fd);
2992 if (IS_ERR(prog))
2993 return PTR_ERR(prog);
2994
9e4e01df
KS		 2995 switch (prog->type) {
2996 case BPF_PROG_TYPE_TRACING:
2997 case BPF_PROG_TYPE_EXT:
2998 case BPF_PROG_TYPE_LSM:
ac4414b5 2999 if (attr->raw_tracepoint.name) {
fec56f58
AS		 3000 /* The attach point for this category of programs
3001 * should be specified via btf_id during program load.
ac4414b5
AS		 3002 */
3003 err = -EINVAL;
3004 goto out_put_prog;
3005 }
9e4e01df
KS		 3006 if (prog->type == BPF_PROG_TYPE_TRACING &&
3007 prog->expected_attach_type == BPF_TRACE_RAW_TP) {
fec56f58 3008 tp_name = prog->aux->attach_func_name;
9e4e01df
KS		 3009 break;
3010 }
5541075a
JO		 3011 err = bpf_tracing_prog_attach(prog, 0, 0);
3012 if (err >= 0)
3013 return err;
3014 goto out_put_prog;
9e4e01df
KS		 3015 case BPF_PROG_TYPE_RAW_TRACEPOINT:
3016 case BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE:
ac4414b5
AS		 3017 if (strncpy_from_user(buf,
3018 u64_to_user_ptr(attr->raw_tracepoint.name),
3019 sizeof(buf) - 1) < 0) {
3020 err = -EFAULT;
3021 goto out_put_prog;
3022 }
3023 buf[sizeof(buf) - 1] = 0;
3024 tp_name = buf;
9e4e01df
KS		 3025 break;
3026 default:
3027 err = -EINVAL;
3028 goto out_put_prog;
ac4414b5 3029 }
c4f6699d 3030
a38d1107 3031 btp = bpf_get_raw_tracepoint(tp_name);
ac4414b5
AS		 3032 if (!btp) {
3033 err = -ENOENT;
3034 goto out_put_prog;
3035 }
c4f6699d 3036
babf3164
AN		 3037 link = kzalloc(sizeof(*link), GFP_USER);
3038 if (!link) {
a38d1107
MM		 3039 err = -ENOMEM;
3040 goto out_put_btp;
3041 }
f2e10bff
AN		 3042 bpf_link_init(&link->link, BPF_LINK_TYPE_RAW_TRACEPOINT,
3043 &bpf_raw_tp_link_lops, prog);
babf3164 3044 link->btp = btp;
c4f6699d 3045
a3b80e10
AN		 3046 err = bpf_link_prime(&link->link, &link_primer);
3047 if (err) {
babf3164 3048 kfree(link);
babf3164
AN		 3049 goto out_put_btp;
3050 }
c4f6699d 3051
babf3164
AN		 3052 err = bpf_probe_register(link->btp, prog);
3053 if (err) {
a3b80e10 3054 bpf_link_cleanup(&link_primer);
babf3164 3055 goto out_put_btp;
c4f6699d 3056 }
babf3164 3057
a3b80e10 3058 return bpf_link_settle(&link_primer);
c4f6699d 3059
a38d1107
MM		 3060out_put_btp:
3061 bpf_put_raw_tracepoint(btp);
ac4414b5
AS		 3062out_put_prog:
3063 bpf_prog_put(prog);
c4f6699d
AS		 3064 return err;
3065}
3066
33491588
AR		 3067static int bpf_prog_attach_check_attach_type(const struct bpf_prog *prog,
3068 enum bpf_attach_type attach_type)
3069{
3070 switch (prog->type) {
3071 case BPF_PROG_TYPE_CGROUP_SOCK:
3072 case BPF_PROG_TYPE_CGROUP_SOCK_ADDR:
0d01da6a 3073 case BPF_PROG_TYPE_CGROUP_SOCKOPT:
e9ddbb77 3074 case BPF_PROG_TYPE_SK_LOOKUP:
33491588 3075 return attach_type == prog->expected_attach_type ? 0 : -EINVAL;
5cf1e914 3076 case BPF_PROG_TYPE_CGROUP_SKB:
2c78ee89
AS		 3077 if (!capable(CAP_NET_ADMIN))
3078 /* cg-skb progs can be loaded by unpriv user.
3079 * check permissions at attach time.
3080 */
3081 return -EPERM;
5cf1e914 3082 return prog->enforce_expected_attach_type &&
3083 prog->expected_attach_type != attach_type ?
3084 -EINVAL : 0;
33491588
AR		 3085 default:
3086 return 0;
3087 }
3088}
3089
e28784e3
AN		 3090static enum bpf_prog_type
3091attach_type_to_prog_type(enum bpf_attach_type attach_type)
f4324551 3092{
e28784e3 3093 switch (attach_type) {
f4324551
DM		 3094 case BPF_CGROUP_INET_INGRESS:
3095 case BPF_CGROUP_INET_EGRESS:
e28784e3 3096 return BPF_PROG_TYPE_CGROUP_SKB;
61023658 3097 case BPF_CGROUP_INET_SOCK_CREATE:
f5836749 3098 case BPF_CGROUP_INET_SOCK_RELEASE:
aac3fc32
AI		 3099 case BPF_CGROUP_INET4_POST_BIND:
3100 case BPF_CGROUP_INET6_POST_BIND:
e28784e3 3101 return BPF_PROG_TYPE_CGROUP_SOCK;
4fbac77d
AI		 3102 case BPF_CGROUP_INET4_BIND:
3103 case BPF_CGROUP_INET6_BIND:
d74bad4e
AI		 3104 case BPF_CGROUP_INET4_CONNECT:
3105 case BPF_CGROUP_INET6_CONNECT:
1b66d253
DB		 3106 case BPF_CGROUP_INET4_GETPEERNAME:
3107 case BPF_CGROUP_INET6_GETPEERNAME:
3108 case BPF_CGROUP_INET4_GETSOCKNAME:
3109 case BPF_CGROUP_INET6_GETSOCKNAME:
1cedee13
AI		 3110 case BPF_CGROUP_UDP4_SENDMSG:
3111 case BPF_CGROUP_UDP6_SENDMSG:
983695fa
DB		 3112 case BPF_CGROUP_UDP4_RECVMSG:
3113 case BPF_CGROUP_UDP6_RECVMSG:
e28784e3 3114 return BPF_PROG_TYPE_CGROUP_SOCK_ADDR;
40304b2a 3115 case BPF_CGROUP_SOCK_OPS:
e28784e3 3116 return BPF_PROG_TYPE_SOCK_OPS;
ebc614f6 3117 case BPF_CGROUP_DEVICE:
e28784e3 3118 return BPF_PROG_TYPE_CGROUP_DEVICE;
4f738adb 3119 case BPF_SK_MSG_VERDICT:
e28784e3 3120 return BPF_PROG_TYPE_SK_MSG;
464bc0fd
JF		 3121 case BPF_SK_SKB_STREAM_PARSER:
3122 case BPF_SK_SKB_STREAM_VERDICT:
a7ba4558 3123 case BPF_SK_SKB_VERDICT:
e28784e3 3124 return BPF_PROG_TYPE_SK_SKB;
f4364dcf 3125 case BPF_LIRC_MODE2:
e28784e3 3126 return BPF_PROG_TYPE_LIRC_MODE2;
d58e468b 3127 case BPF_FLOW_DISSECTOR:
e28784e3 3128 return BPF_PROG_TYPE_FLOW_DISSECTOR;
7b146ceb 3129 case BPF_CGROUP_SYSCTL:
e28784e3 3130 return BPF_PROG_TYPE_CGROUP_SYSCTL;
0d01da6a
SF		 3131 case BPF_CGROUP_GETSOCKOPT:
3132 case BPF_CGROUP_SETSOCKOPT:
e28784e3 3133 return BPF_PROG_TYPE_CGROUP_SOCKOPT;
de4e05ca
YS		 3134 case BPF_TRACE_ITER:
3135 return BPF_PROG_TYPE_TRACING;
e9ddbb77
JS		 3136 case BPF_SK_LOOKUP:
3137 return BPF_PROG_TYPE_SK_LOOKUP;
aa8d3a71
AN		 3138 case BPF_XDP:
3139 return BPF_PROG_TYPE_XDP;
f4324551 3140 default:
e28784e3 3141 return BPF_PROG_TYPE_UNSPEC;
f4324551 3142 }
e28784e3
AN		 3143}
3144
3145#define BPF_PROG_ATTACH_LAST_FIELD replace_bpf_fd
3146
3147#define BPF_F_ATTACH_MASK \
3148 (BPF_F_ALLOW_OVERRIDE | BPF_F_ALLOW_MULTI | BPF_F_REPLACE)
3149
3150static int bpf_prog_attach(const union bpf_attr *attr)
3151{
3152 enum bpf_prog_type ptype;
3153 struct bpf_prog *prog;
3154 int ret;
3155
e28784e3
AN		 3156 if (CHECK_ATTR(BPF_PROG_ATTACH))
3157 return -EINVAL;
3158
3159 if (attr->attach_flags & ~BPF_F_ATTACH_MASK)
3160 return -EINVAL;
3161
3162 ptype = attach_type_to_prog_type(attr->attach_type);
3163 if (ptype == BPF_PROG_TYPE_UNSPEC)
3164 return -EINVAL;
f4324551 3165
b2cd1257
DA		 3166 prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype);
3167 if (IS_ERR(prog))
3168 return PTR_ERR(prog);
3169
5e43f899
AI		 3170 if (bpf_prog_attach_check_attach_type(prog, attr->attach_type)) {
3171 bpf_prog_put(prog);
3172 return -EINVAL;
3173 }
3174
fdb5c453
SY		 3175 switch (ptype) {
3176 case BPF_PROG_TYPE_SK_SKB:
3177 case BPF_PROG_TYPE_SK_MSG:
604326b4 3178 ret = sock_map_get_from_fd(attr, prog);
fdb5c453
SY		 3179 break;
3180 case BPF_PROG_TYPE_LIRC_MODE2:
3181 ret = lirc_prog_attach(attr, prog);
3182 break;
d58e468b 3183 case BPF_PROG_TYPE_FLOW_DISSECTOR:
a3fd7cee 3184 ret = netns_bpf_prog_attach(attr, prog);
d58e468b 3185 break;
e28784e3
AN		 3186 case BPF_PROG_TYPE_CGROUP_DEVICE:
3187 case BPF_PROG_TYPE_CGROUP_SKB:
3188 case BPF_PROG_TYPE_CGROUP_SOCK:
3189 case BPF_PROG_TYPE_CGROUP_SOCK_ADDR:
3190 case BPF_PROG_TYPE_CGROUP_SOCKOPT:
3191 case BPF_PROG_TYPE_CGROUP_SYSCTL:
3192 case BPF_PROG_TYPE_SOCK_OPS:
fdb5c453 3193 ret = cgroup_bpf_prog_attach(attr, ptype, prog);
e28784e3
AN		 3194 break;
3195 default:
3196 ret = -EINVAL;
b2cd1257
DA		 3197 }
3198
7f677633
AS		 3199 if (ret)
3200 bpf_prog_put(prog);
7f677633 3201 return ret;
f4324551
DM		 3202}
3203
3204#define BPF_PROG_DETACH_LAST_FIELD attach_type
3205
3206static int bpf_prog_detach(const union bpf_attr *attr)
3207{
324bda9e 3208 enum bpf_prog_type ptype;
f4324551 3209
f4324551
DM		 3210 if (CHECK_ATTR(BPF_PROG_DETACH))
3211 return -EINVAL;
3212
e28784e3
AN		 3213 ptype = attach_type_to_prog_type(attr->attach_type);
3214
3215 switch (ptype) {
3216 case BPF_PROG_TYPE_SK_MSG:
3217 case BPF_PROG_TYPE_SK_SKB:
bb0de313 3218 return sock_map_prog_detach(attr, ptype);
e28784e3 3219 case BPF_PROG_TYPE_LIRC_MODE2:
f4364dcf 3220 return lirc_prog_detach(attr);
e28784e3 3221 case BPF_PROG_TYPE_FLOW_DISSECTOR:
4ac2add6 3222 return netns_bpf_prog_detach(attr, ptype);
e28784e3
AN		 3223 case BPF_PROG_TYPE_CGROUP_DEVICE:
3224 case BPF_PROG_TYPE_CGROUP_SKB:
3225 case BPF_PROG_TYPE_CGROUP_SOCK:
3226 case BPF_PROG_TYPE_CGROUP_SOCK_ADDR:
3227 case BPF_PROG_TYPE_CGROUP_SOCKOPT:
3228 case BPF_PROG_TYPE_CGROUP_SYSCTL:
3229 case BPF_PROG_TYPE_SOCK_OPS:
3230 return cgroup_bpf_prog_detach(attr, ptype);
f4324551
DM		 3231 default:
3232 return -EINVAL;
3233 }
f4324551 3234}
40304b2a 3235
468e2f64
AS		 3236#define BPF_PROG_QUERY_LAST_FIELD query.prog_cnt
3237
3238static int bpf_prog_query(const union bpf_attr *attr,
3239 union bpf_attr __user *uattr)
3240{
468e2f64
AS		 3241 if (!capable(CAP_NET_ADMIN))
3242 return -EPERM;
3243 if (CHECK_ATTR(BPF_PROG_QUERY))
3244 return -EINVAL;
3245 if (attr->query.query_flags & ~BPF_F_QUERY_EFFECTIVE)
3246 return -EINVAL;
3247
3248 switch (attr->query.attach_type) {
3249 case BPF_CGROUP_INET_INGRESS:
3250 case BPF_CGROUP_INET_EGRESS:
3251 case BPF_CGROUP_INET_SOCK_CREATE:
f5836749 3252 case BPF_CGROUP_INET_SOCK_RELEASE:
4fbac77d
AI		 3253 case BPF_CGROUP_INET4_BIND:
3254 case BPF_CGROUP_INET6_BIND:
aac3fc32
AI		 3255 case BPF_CGROUP_INET4_POST_BIND:
3256 case BPF_CGROUP_INET6_POST_BIND:
d74bad4e
AI		 3257 case BPF_CGROUP_INET4_CONNECT:
3258 case BPF_CGROUP_INET6_CONNECT:
1b66d253
DB		 3259 case BPF_CGROUP_INET4_GETPEERNAME:
3260 case BPF_CGROUP_INET6_GETPEERNAME:
3261 case BPF_CGROUP_INET4_GETSOCKNAME:
3262 case BPF_CGROUP_INET6_GETSOCKNAME:
1cedee13
AI		 3263 case BPF_CGROUP_UDP4_SENDMSG:
3264 case BPF_CGROUP_UDP6_SENDMSG:
983695fa
DB		 3265 case BPF_CGROUP_UDP4_RECVMSG:
3266 case BPF_CGROUP_UDP6_RECVMSG:
468e2f64 3267 case BPF_CGROUP_SOCK_OPS:
ebc614f6 3268 case BPF_CGROUP_DEVICE:
7b146ceb 3269 case BPF_CGROUP_SYSCTL:
0d01da6a
SF		 3270 case BPF_CGROUP_GETSOCKOPT:
3271 case BPF_CGROUP_SETSOCKOPT:
e28784e3 3272 return cgroup_bpf_prog_query(attr, uattr);
f4364dcf
SY		 3273 case BPF_LIRC_MODE2:
3274 return lirc_prog_query(attr, uattr);
118c8e9a 3275 case BPF_FLOW_DISSECTOR:
e9ddbb77 3276 case BPF_SK_LOOKUP:
a3fd7cee 3277 return netns_bpf_prog_query(attr, uattr);
468e2f64
AS		 3278 default:
3279 return -EINVAL;
3280 }
468e2f64 3281}
f4324551 3282
1b4d60ec 3283#define BPF_PROG_TEST_RUN_LAST_FIELD test.cpu
1cf1cae9
AS		 3284
3285static int bpf_prog_test_run(const union bpf_attr *attr,
3286 union bpf_attr __user *uattr)
3287{
3288 struct bpf_prog *prog;
3289 int ret = -ENOTSUPP;
3290
3291 if (CHECK_ATTR(BPF_PROG_TEST_RUN))
3292 return -EINVAL;
3293
b0b9395d
SF		 3294 if ((attr->test.ctx_size_in && !attr->test.ctx_in) ||
3295 (!attr->test.ctx_size_in && attr->test.ctx_in))
3296 return -EINVAL;
3297
3298 if ((attr->test.ctx_size_out && !attr->test.ctx_out) ||
3299 (!attr->test.ctx_size_out && attr->test.ctx_out))
3300 return -EINVAL;
3301
1cf1cae9
AS		 3302 prog = bpf_prog_get(attr->test.prog_fd);
3303 if (IS_ERR(prog))
3304 return PTR_ERR(prog);
3305
3306 if (prog->aux->ops->test_run)
3307 ret = prog->aux->ops->test_run(prog, attr, uattr);
3308
3309 bpf_prog_put(prog);
3310 return ret;
3311}
3312
34ad5580
MKL		 3313#define BPF_OBJ_GET_NEXT_ID_LAST_FIELD next_id
3314
3315static int bpf_obj_get_next_id(const union bpf_attr *attr,
3316 union bpf_attr __user *uattr,
3317 struct idr *idr,
3318 spinlock_t *lock)
3319{
3320 u32 next_id = attr->start_id;
3321 int err = 0;
3322
3323 if (CHECK_ATTR(BPF_OBJ_GET_NEXT_ID) || next_id >= INT_MAX)
3324 return -EINVAL;
3325
3326 if (!capable(CAP_SYS_ADMIN))
3327 return -EPERM;
3328
3329 next_id++;
3330 spin_lock_bh(lock);
3331 if (!idr_get_next(idr, &next_id))
3332 err = -ENOENT;
3333 spin_unlock_bh(lock);
3334
3335 if (!err)
3336 err = put_user(next_id, &uattr->next_id);
3337
3338 return err;
3339}
3340
6086d29d
YS		 3341struct bpf_map *bpf_map_get_curr_or_next(u32 *id)
3342{
3343 struct bpf_map *map;
3344
3345 spin_lock_bh(&map_idr_lock);
3346again:
3347 map = idr_get_next(&map_idr, id);
3348 if (map) {
3349 map = __bpf_map_inc_not_zero(map, false);
3350 if (IS_ERR(map)) {
3351 (*id)++;
3352 goto again;
3353 }
3354 }
3355 spin_unlock_bh(&map_idr_lock);
3356
3357 return map;
3358}
3359
a228a64f
AS		 3360struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id)
3361{
3362 struct bpf_prog *prog;
3363
3364 spin_lock_bh(&prog_idr_lock);
3365again:
3366 prog = idr_get_next(&prog_idr, id);
3367 if (prog) {
3368 prog = bpf_prog_inc_not_zero(prog);
3369 if (IS_ERR(prog)) {
3370 (*id)++;
3371 goto again;
3372 }
3373 }
3374 spin_unlock_bh(&prog_idr_lock);
3375
3376 return prog;
3377}
3378
b16d9aa4
MKL		 3379#define BPF_PROG_GET_FD_BY_ID_LAST_FIELD prog_id
3380
7e6897f9 3381struct bpf_prog *bpf_prog_by_id(u32 id)
b16d9aa4
MKL		 3382{
3383 struct bpf_prog *prog;
b16d9aa4 3384
7e6897f9
BT		 3385 if (!id)
3386 return ERR_PTR(-ENOENT);
b16d9aa4
MKL		 3387
3388 spin_lock_bh(&prog_idr_lock);
3389 prog = idr_find(&prog_idr, id);
3390 if (prog)
3391 prog = bpf_prog_inc_not_zero(prog);
3392 else
3393 prog = ERR_PTR(-ENOENT);
3394 spin_unlock_bh(&prog_idr_lock);
7e6897f9
BT		 3395 return prog;
3396}
3397
3398static int bpf_prog_get_fd_by_id(const union bpf_attr *attr)
3399{
3400 struct bpf_prog *prog;
3401 u32 id = attr->prog_id;
3402 int fd;
3403
3404 if (CHECK_ATTR(BPF_PROG_GET_FD_BY_ID))
3405 return -EINVAL;
3406
3407 if (!capable(CAP_SYS_ADMIN))
3408 return -EPERM;
b16d9aa4 3409
7e6897f9 3410 prog = bpf_prog_by_id(id);
b16d9aa4
MKL		 3411 if (IS_ERR(prog))
3412 return PTR_ERR(prog);
3413
3414 fd = bpf_prog_new_fd(prog);
3415 if (fd < 0)
3416 bpf_prog_put(prog);
3417
3418 return fd;
3419}
3420
6e71b04a 3421#define BPF_MAP_GET_FD_BY_ID_LAST_FIELD open_flags
bd5f5f4e
MKL		 3422
3423static int bpf_map_get_fd_by_id(const union bpf_attr *attr)
3424{
3425 struct bpf_map *map;
3426 u32 id = attr->map_id;
6e71b04a 3427 int f_flags;
bd5f5f4e
MKL		 3428 int fd;
3429
6e71b04a
CF		 3430 if (CHECK_ATTR(BPF_MAP_GET_FD_BY_ID) ||
3431 attr->open_flags & ~BPF_OBJ_FLAG_MASK)
bd5f5f4e
MKL		 3432 return -EINVAL;
3433
3434 if (!capable(CAP_SYS_ADMIN))
3435 return -EPERM;
3436
6e71b04a
CF		 3437 f_flags = bpf_get_file_flag(attr->open_flags);
3438 if (f_flags < 0)
3439 return f_flags;
3440
bd5f5f4e
MKL		 3441 spin_lock_bh(&map_idr_lock);
3442 map = idr_find(&map_idr, id);
3443 if (map)
b0e4701c 3444 map = __bpf_map_inc_not_zero(map, true);
bd5f5f4e
MKL		 3445 else
3446 map = ERR_PTR(-ENOENT);
3447 spin_unlock_bh(&map_idr_lock);
3448
3449 if (IS_ERR(map))
3450 return PTR_ERR(map);
3451
6e71b04a 3452 fd = bpf_map_new_fd(map, f_flags);
bd5f5f4e 3453 if (fd < 0)
781e6282 3454 bpf_map_put_with_uref(map);
bd5f5f4e
MKL		 3455
3456 return fd;
3457}
3458
7105e828 3459static const struct bpf_map *bpf_map_from_imm(const struct bpf_prog *prog,
d8eca5bb
DB		 3460 unsigned long addr, u32 *off,
3461 u32 *type)
7105e828 3462{
d8eca5bb 3463 const struct bpf_map *map;
7105e828
DB		 3464 int i;
3465
984fe94f 3466 mutex_lock(&prog->aux->used_maps_mutex);
d8eca5bb
DB		 3467 for (i = 0, *off = 0; i < prog->aux->used_map_cnt; i++) {
3468 map = prog->aux->used_maps[i];
3469 if (map == (void *)addr) {
3470 *type = BPF_PSEUDO_MAP_FD;
984fe94f 3471 goto out;
d8eca5bb
DB		 3472 }
3473 if (!map->ops->map_direct_value_meta)
3474 continue;
3475 if (!map->ops->map_direct_value_meta(map, addr, off)) {
3476 *type = BPF_PSEUDO_MAP_VALUE;
984fe94f 3477 goto out;
d8eca5bb
DB		 3478 }
3479 }
984fe94f 3480 map = NULL;
d8eca5bb 3481
984fe94f
YZ		 3482out:
3483 mutex_unlock(&prog->aux->used_maps_mutex);
3484 return map;
7105e828
DB		 3485}
3486
63960260
KC		 3487static struct bpf_insn *bpf_insn_prepare_dump(const struct bpf_prog *prog,
3488 const struct cred *f_cred)
7105e828
DB		 3489{
3490 const struct bpf_map *map;
3491 struct bpf_insn *insns;
d8eca5bb 3492 u32 off, type;
7105e828 3493 u64 imm;
29fcb05b 3494 u8 code;
7105e828
DB		 3495 int i;
3496
3497 insns = kmemdup(prog->insnsi, bpf_prog_insn_size(prog),
3498 GFP_USER);
3499 if (!insns)
3500 return insns;
3501
3502 for (i = 0; i < prog->len; i++) {
29fcb05b
AN		 3503 code = insns[i].code;
3504
3505 if (code == (BPF_JMP | BPF_TAIL_CALL)) {
7105e828
DB		 3506 insns[i].code = BPF_JMP | BPF_CALL;
3507 insns[i].imm = BPF_FUNC_tail_call;
3508 /* fall-through */
3509 }
29fcb05b
AN		 3510 if (code == (BPF_JMP | BPF_CALL) ||
3511 code == (BPF_JMP | BPF_CALL_ARGS)) {
3512 if (code == (BPF_JMP | BPF_CALL_ARGS))
7105e828 3513 insns[i].code = BPF_JMP | BPF_CALL;
63960260 3514 if (!bpf_dump_raw_ok(f_cred))
7105e828
DB		 3515 insns[i].imm = 0;
3516 continue;
3517 }
29fcb05b
AN		 3518 if (BPF_CLASS(code) == BPF_LDX && BPF_MODE(code) == BPF_PROBE_MEM) {
3519 insns[i].code = BPF_LDX | BPF_SIZE(code) | BPF_MEM;
3520 continue;
3521 }
7105e828 3522
29fcb05b 3523 if (code != (BPF_LD | BPF_IMM | BPF_DW))
7105e828
DB		 3524 continue;
3525
3526 imm = ((u64)insns[i + 1].imm << 32) | (u32)insns[i].imm;
d8eca5bb 3527 map = bpf_map_from_imm(prog, imm, &off, &type);
7105e828 3528 if (map) {
d8eca5bb 3529 insns[i].src_reg = type;
7105e828 3530 insns[i].imm = map->id;
d8eca5bb 3531 insns[i + 1].imm = off;
7105e828
DB		 3532 continue;
3533 }
7105e828
DB		 3534 }
3535
3536 return insns;
3537}
3538
c454a46b
MKL		 3539static int set_info_rec_size(struct bpf_prog_info *info)
3540{
3541 /*
3542 * Ensure info.*_rec_size is the same as kernel expected size
3543 *
3544 * or
3545 *
3546 * Only allow zero *_rec_size if both _rec_size and _cnt are
3547 * zero. In this case, the kernel will set the expected
3548 * _rec_size back to the info.
3549 */
3550
11d8b82d 3551 if ((info->nr_func_info || info->func_info_rec_size) &&
c454a46b
MKL		 3552 info->func_info_rec_size != sizeof(struct bpf_func_info))
3553 return -EINVAL;
3554
11d8b82d 3555 if ((info->nr_line_info || info->line_info_rec_size) &&
c454a46b
MKL		 3556 info->line_info_rec_size != sizeof(struct bpf_line_info))
3557 return -EINVAL;
3558
11d8b82d 3559 if ((info->nr_jited_line_info || info->jited_line_info_rec_size) &&
c454a46b
MKL		 3560 info->jited_line_info_rec_size != sizeof(__u64))
3561 return -EINVAL;
3562
3563 info->func_info_rec_size = sizeof(struct bpf_func_info);
3564 info->line_info_rec_size = sizeof(struct bpf_line_info);
3565 info->jited_line_info_rec_size = sizeof(__u64);
3566
3567 return 0;
3568}
3569
63960260
KC		 3570static int bpf_prog_get_info_by_fd(struct file *file,
3571 struct bpf_prog *prog,
1e270976
MKL		 3572 const union bpf_attr *attr,
3573 union bpf_attr __user *uattr)
3574{
3575 struct bpf_prog_info __user *uinfo = u64_to_user_ptr(attr->info.info);
5c6f2588 3576 struct bpf_prog_info info;
1e270976 3577 u32 info_len = attr->info.info_len;
5f8f8b93 3578 struct bpf_prog_stats stats;
1e270976
MKL		 3579 char __user *uinsns;
3580 u32 ulen;
3581 int err;
3582
af2ac3e1 3583 err = bpf_check_uarg_tail_zero(USER_BPFPTR(uinfo), sizeof(info), info_len);
1e270976
MKL		 3584 if (err)
3585 return err;
3586 info_len = min_t(u32, sizeof(info), info_len);
3587
5c6f2588 3588 memset(&info, 0, sizeof(info));
1e270976 3589 if (copy_from_user(&info, uinfo, info_len))
89b09689 3590 return -EFAULT;
1e270976
MKL		 3591
3592 info.type = prog->type;
3593 info.id = prog->aux->id;
cb4d2b3f
MKL		 3594 info.load_time = prog->aux->load_time;
3595 info.created_by_uid = from_kuid_munged(current_user_ns(),
3596 prog->aux->user->uid);
b85fab0e 3597 info.gpl_compatible = prog->gpl_compatible;
1e270976
MKL		 3598
3599 memcpy(info.tag, prog->tag, sizeof(prog->tag));
cb4d2b3f
MKL		 3600 memcpy(info.name, prog->aux->name, sizeof(prog->aux->name));
3601
984fe94f 3602 mutex_lock(&prog->aux->used_maps_mutex);
cb4d2b3f
MKL		 3603 ulen = info.nr_map_ids;
3604 info.nr_map_ids = prog->aux->used_map_cnt;
3605 ulen = min_t(u32, info.nr_map_ids, ulen);
3606 if (ulen) {
721e08da 3607 u32 __user *user_map_ids = u64_to_user_ptr(info.map_ids);
cb4d2b3f
MKL		 3608 u32 i;
3609
3610 for (i = 0; i < ulen; i++)
3611 if (put_user(prog->aux->used_maps[i]->id,
984fe94f
YZ		 3612 &user_map_ids[i])) {
3613 mutex_unlock(&prog->aux->used_maps_mutex);
cb4d2b3f 3614 return -EFAULT;
984fe94f 3615 }
cb4d2b3f 3616 }
984fe94f 3617 mutex_unlock(&prog->aux->used_maps_mutex);
1e270976 3618
c454a46b
MKL		 3619 err = set_info_rec_size(&info);
3620 if (err)
3621 return err;
7337224f 3622
5f8f8b93
AS		 3623 bpf_prog_get_stats(prog, &stats);
3624 info.run_time_ns = stats.nsecs;
3625 info.run_cnt = stats.cnt;
9ed9e9ba 3626 info.recursion_misses = stats.misses;
5f8f8b93 3627
2c78ee89 3628 if (!bpf_capable()) {
1e270976
MKL		 3629 info.jited_prog_len = 0;
3630 info.xlated_prog_len = 0;
dbecd738 3631 info.nr_jited_ksyms = 0;
28c2fae7 3632 info.nr_jited_func_lens = 0;
11d8b82d
YS		 3633 info.nr_func_info = 0;
3634 info.nr_line_info = 0;
3635 info.nr_jited_line_info = 0;
1e270976
MKL		 3636 goto done;
3637 }
3638
1e270976 3639 ulen = info.xlated_prog_len;
9975a54b 3640 info.xlated_prog_len = bpf_prog_insn_size(prog);
1e270976 3641 if (info.xlated_prog_len && ulen) {
7105e828
DB		 3642 struct bpf_insn *insns_sanitized;
3643 bool fault;
3644
63960260 3645 if (prog->blinded && !bpf_dump_raw_ok(file->f_cred)) {
7105e828
DB		 3646 info.xlated_prog_insns = 0;
3647 goto done;
3648 }
63960260 3649 insns_sanitized = bpf_insn_prepare_dump(prog, file->f_cred);
7105e828
DB		 3650 if (!insns_sanitized)
3651 return -ENOMEM;
1e270976
MKL		 3652 uinsns = u64_to_user_ptr(info.xlated_prog_insns);
3653 ulen = min_t(u32, info.xlated_prog_len, ulen);
7105e828
DB		 3654 fault = copy_to_user(uinsns, insns_sanitized, ulen);
3655 kfree(insns_sanitized);
3656 if (fault)
1e270976
MKL		 3657 return -EFAULT;
3658 }
3659
675fc275
JK		 3660 if (bpf_prog_is_dev_bound(prog->aux)) {
3661 err = bpf_prog_offload_info_fill(&info, prog);
3662 if (err)
3663 return err;
fcfb126d
JW		 3664 goto done;
3665 }
3666
3667 /* NOTE: the following code is supposed to be skipped for offload.
3668 * bpf_prog_offload_info_fill() is the place to fill similar fields
3669 * for offload.
3670 */
3671 ulen = info.jited_prog_len;
4d56a76e
SD		 3672 if (prog->aux->func_cnt) {
3673 u32 i;
3674
3675 info.jited_prog_len = 0;
3676 for (i = 0; i < prog->aux->func_cnt; i++)
3677 info.jited_prog_len += prog->aux->func[i]->jited_len;
3678 } else {
3679 info.jited_prog_len = prog->jited_len;
3680 }
3681
fcfb126d 3682 if (info.jited_prog_len && ulen) {
63960260 3683 if (bpf_dump_raw_ok(file->f_cred)) {
fcfb126d
JW		 3684 uinsns = u64_to_user_ptr(info.jited_prog_insns);
3685 ulen = min_t(u32, info.jited_prog_len, ulen);
4d56a76e
SD		 3686
3687 /* for multi-function programs, copy the JITed
3688 * instructions for all the functions
3689 */
3690 if (prog->aux->func_cnt) {
3691 u32 len, free, i;
3692 u8 *img;
3693
3694 free = ulen;
3695 for (i = 0; i < prog->aux->func_cnt; i++) {
3696 len = prog->aux->func[i]->jited_len;
3697 len = min_t(u32, len, free);
3698 img = (u8 *) prog->aux->func[i]->bpf_func;
3699 if (copy_to_user(uinsns, img, len))
3700 return -EFAULT;
3701 uinsns += len;
3702 free -= len;
3703 if (!free)
3704 break;
3705 }
3706 } else {
3707 if (copy_to_user(uinsns, prog->bpf_func, ulen))
3708 return -EFAULT;
3709 }
fcfb126d
JW		 3710 } else {
3711 info.jited_prog_insns = 0;
3712 }
675fc275
JK		 3713 }
3714
dbecd738 3715 ulen = info.nr_jited_ksyms;
ff1889fc 3716 info.nr_jited_ksyms = prog->aux->func_cnt ? : 1;
7a5725dd 3717 if (ulen) {
63960260 3718 if (bpf_dump_raw_ok(file->f_cred)) {
ff1889fc 3719 unsigned long ksym_addr;
dbecd738 3720 u64 __user *user_ksyms;
dbecd738
SD		 3721 u32 i;
3722
3723 /* copy the address of the kernel symbol
3724 * corresponding to each function
3725 */
3726 ulen = min_t(u32, info.nr_jited_ksyms, ulen);
3727 user_ksyms = u64_to_user_ptr(info.jited_ksyms);
ff1889fc
SL		 3728 if (prog->aux->func_cnt) {
3729 for (i = 0; i < ulen; i++) {
3730 ksym_addr = (unsigned long)
3731 prog->aux->func[i]->bpf_func;
3732 if (put_user((u64) ksym_addr,
3733 &user_ksyms[i]))
3734 return -EFAULT;
3735 }
3736 } else {
3737 ksym_addr = (unsigned long) prog->bpf_func;
3738 if (put_user((u64) ksym_addr, &user_ksyms[0]))
dbecd738
SD		 3739 return -EFAULT;
3740 }
3741 } else {
3742 info.jited_ksyms = 0;
3743 }
3744 }
3745
815581c1 3746 ulen = info.nr_jited_func_lens;
ff1889fc 3747 info.nr_jited_func_lens = prog->aux->func_cnt ? : 1;
7a5725dd 3748 if (ulen) {
63960260 3749 if (bpf_dump_raw_ok(file->f_cred)) {
815581c1
SD		 3750 u32 __user *user_lens;
3751 u32 func_len, i;
3752
3753 /* copy the JITed image lengths for each function */
3754 ulen = min_t(u32, info.nr_jited_func_lens, ulen);
3755 user_lens = u64_to_user_ptr(info.jited_func_lens);
ff1889fc
SL		 3756 if (prog->aux->func_cnt) {
3757 for (i = 0; i < ulen; i++) {
3758 func_len =
3759 prog->aux->func[i]->jited_len;
3760 if (put_user(func_len, &user_lens[i]))
3761 return -EFAULT;
3762 }
3763 } else {
3764 func_len = prog->jited_len;
3765 if (put_user(func_len, &user_lens[0]))
815581c1
SD		 3766 return -EFAULT;
3767 }
3768 } else {
3769 info.jited_func_lens = 0;
3770 }
3771 }
3772
7337224f 3773 if (prog->aux->btf)
22dc4a0f 3774 info.btf_id = btf_obj_id(prog->aux->btf);
838e9690 3775
11d8b82d
YS		 3776 ulen = info.nr_func_info;
3777 info.nr_func_info = prog->aux->func_info_cnt;
3778 if (info.nr_func_info && ulen) {
9e794163 3779 char __user *user_finfo;
7337224f 3780
9e794163
MKL		 3781 user_finfo = u64_to_user_ptr(info.func_info);
3782 ulen = min_t(u32, info.nr_func_info, ulen);
3783 if (copy_to_user(user_finfo, prog->aux->func_info,
3784 info.func_info_rec_size * ulen))
3785 return -EFAULT;
838e9690
YS		 3786 }
3787
11d8b82d
YS		 3788 ulen = info.nr_line_info;
3789 info.nr_line_info = prog->aux->nr_linfo;
3790 if (info.nr_line_info && ulen) {
9e794163 3791 __u8 __user *user_linfo;
c454a46b 3792
9e794163
MKL		 3793 user_linfo = u64_to_user_ptr(info.line_info);
3794 ulen = min_t(u32, info.nr_line_info, ulen);
3795 if (copy_to_user(user_linfo, prog->aux->linfo,
3796 info.line_info_rec_size * ulen))
3797 return -EFAULT;
c454a46b
MKL		 3798 }
3799
11d8b82d 3800 ulen = info.nr_jited_line_info;
c454a46b 3801 if (prog->aux->jited_linfo)
11d8b82d 3802 info.nr_jited_line_info = prog->aux->nr_linfo;
c454a46b 3803 else
11d8b82d
YS		 3804 info.nr_jited_line_info = 0;
3805 if (info.nr_jited_line_info && ulen) {
63960260 3806 if (bpf_dump_raw_ok(file->f_cred)) {
c454a46b
MKL		 3807 __u64 __user *user_linfo;
3808 u32 i;
3809
3810 user_linfo = u64_to_user_ptr(info.jited_line_info);
11d8b82d 3811 ulen = min_t(u32, info.nr_jited_line_info, ulen);
c454a46b
MKL		 3812 for (i = 0; i < ulen; i++) {
3813 if (put_user((__u64)(long)prog->aux->jited_linfo[i],
3814 &user_linfo[i]))
3815 return -EFAULT;
3816 }
3817 } else {
3818 info.jited_line_info = 0;
3819 }
3820 }
3821
c872bdb3
SL		 3822 ulen = info.nr_prog_tags;
3823 info.nr_prog_tags = prog->aux->func_cnt ? : 1;
3824 if (ulen) {
3825 __u8 __user (*user_prog_tags)[BPF_TAG_SIZE];
3826 u32 i;
3827
3828 user_prog_tags = u64_to_user_ptr(info.prog_tags);
3829 ulen = min_t(u32, info.nr_prog_tags, ulen);
3830 if (prog->aux->func_cnt) {
3831 for (i = 0; i < ulen; i++) {
3832 if (copy_to_user(user_prog_tags[i],
3833 prog->aux->func[i]->tag,
3834 BPF_TAG_SIZE))
3835 return -EFAULT;
3836 }
3837 } else {
3838 if (copy_to_user(user_prog_tags[0],
3839 prog->tag, BPF_TAG_SIZE))
3840 return -EFAULT;
3841 }
3842 }
3843
1e270976
MKL		 3844done:
3845 if (copy_to_user(uinfo, &info, info_len) ||
3846 put_user(info_len, &uattr->info.info_len))
3847 return -EFAULT;
3848
3849 return 0;
3850}
3851
63960260
KC		 3852static int bpf_map_get_info_by_fd(struct file *file,
3853 struct bpf_map *map,
1e270976
MKL		 3854 const union bpf_attr *attr,
3855 union bpf_attr __user *uattr)
3856{
3857 struct bpf_map_info __user *uinfo = u64_to_user_ptr(attr->info.info);
5c6f2588 3858 struct bpf_map_info info;
1e270976
MKL		 3859 u32 info_len = attr->info.info_len;
3860 int err;
3861
af2ac3e1 3862 err = bpf_check_uarg_tail_zero(USER_BPFPTR(uinfo), sizeof(info), info_len);
1e270976
MKL		 3863 if (err)
3864 return err;
3865 info_len = min_t(u32, sizeof(info), info_len);
3866
5c6f2588 3867 memset(&info, 0, sizeof(info));
1e270976
MKL		 3868 info.type = map->map_type;
3869 info.id = map->id;
3870 info.key_size = map->key_size;
3871 info.value_size = map->value_size;
3872 info.max_entries = map->max_entries;
3873 info.map_flags = map->map_flags;
ad5b177b 3874 memcpy(info.name, map->name, sizeof(map->name));
1e270976 3875
78958fca 3876 if (map->btf) {
22dc4a0f 3877 info.btf_id = btf_obj_id(map->btf);
9b2cf328
MKL		 3878 info.btf_key_type_id = map->btf_key_type_id;
3879 info.btf_value_type_id = map->btf_value_type_id;
78958fca 3880 }
85d33df3 3881 info.btf_vmlinux_value_type_id = map->btf_vmlinux_value_type_id;
78958fca 3882
52775b33
JK		 3883 if (bpf_map_is_dev_bound(map)) {
3884 err = bpf_map_offload_info_fill(&info, map);
3885 if (err)
3886 return err;
3887 }
3888
1e270976
MKL		 3889 if (copy_to_user(uinfo, &info, info_len) ||
3890 put_user(info_len, &uattr->info.info_len))
3891 return -EFAULT;
3892
3893 return 0;
3894}
3895
63960260
KC		 3896static int bpf_btf_get_info_by_fd(struct file *file,
3897 struct btf *btf,
62dab84c
MKL		 3898 const union bpf_attr *attr,
3899 union bpf_attr __user *uattr)
3900{
3901 struct bpf_btf_info __user *uinfo = u64_to_user_ptr(attr->info.info);
3902 u32 info_len = attr->info.info_len;
3903 int err;
3904
af2ac3e1 3905 err = bpf_check_uarg_tail_zero(USER_BPFPTR(uinfo), sizeof(*uinfo), info_len);
62dab84c
MKL		 3906 if (err)
3907 return err;
3908
3909 return btf_get_info_by_fd(btf, attr, uattr);
3910}
3911
63960260
KC		 3912static int bpf_link_get_info_by_fd(struct file *file,
3913 struct bpf_link *link,
f2e10bff
AN		 3914 const union bpf_attr *attr,
3915 union bpf_attr __user *uattr)
3916{
3917 struct bpf_link_info __user *uinfo = u64_to_user_ptr(attr->info.info);
3918 struct bpf_link_info info;
3919 u32 info_len = attr->info.info_len;
3920 int err;
3921
af2ac3e1 3922 err = bpf_check_uarg_tail_zero(USER_BPFPTR(uinfo), sizeof(info), info_len);
f2e10bff
AN		 3923 if (err)
3924 return err;
3925 info_len = min_t(u32, sizeof(info), info_len);
3926
3927 memset(&info, 0, sizeof(info));
3928 if (copy_from_user(&info, uinfo, info_len))
3929 return -EFAULT;
3930
3931 info.type = link->type;
3932 info.id = link->id;
3933 info.prog_id = link->prog->aux->id;
3934
3935 if (link->ops->fill_link_info) {
3936 err = link->ops->fill_link_info(link, &info);
3937 if (err)
3938 return err;
3939 }
3940
3941 if (copy_to_user(uinfo, &info, info_len) ||
3942 put_user(info_len, &uattr->info.info_len))
3943 return -EFAULT;
3944
3945 return 0;
3946}
3947
3948
1e270976
MKL		 3949#define BPF_OBJ_GET_INFO_BY_FD_LAST_FIELD info.info
3950
3951static int bpf_obj_get_info_by_fd(const union bpf_attr *attr,
3952 union bpf_attr __user *uattr)
3953{
3954 int ufd = attr->info.bpf_fd;
3955 struct fd f;
3956 int err;
3957
3958 if (CHECK_ATTR(BPF_OBJ_GET_INFO_BY_FD))
3959 return -EINVAL;
3960
3961 f = fdget(ufd);
3962 if (!f.file)
3963 return -EBADFD;
3964
3965 if (f.file->f_op == &bpf_prog_fops)
63960260 3966 err = bpf_prog_get_info_by_fd(f.file, f.file->private_data, attr,
1e270976
MKL		 3967 uattr);
3968 else if (f.file->f_op == &bpf_map_fops)
63960260 3969 err = bpf_map_get_info_by_fd(f.file, f.file->private_data, attr,
1e270976 3970 uattr);
60197cfb 3971 else if (f.file->f_op == &btf_fops)
63960260 3972 err = bpf_btf_get_info_by_fd(f.file, f.file->private_data, attr, uattr);
f2e10bff 3973 else if (f.file->f_op == &bpf_link_fops)
63960260 3974 err = bpf_link_get_info_by_fd(f.file, f.file->private_data,
f2e10bff 3975 attr, uattr);
1e270976
MKL		 3976 else
3977 err = -EINVAL;
3978
3979 fdput(f);
3980 return err;
3981}
3982
f56a653c
MKL		 3983#define BPF_BTF_LOAD_LAST_FIELD btf_log_level
3984
c571bd75 3985static int bpf_btf_load(const union bpf_attr *attr, bpfptr_t uattr)
f56a653c
MKL		 3986{
3987 if (CHECK_ATTR(BPF_BTF_LOAD))
3988 return -EINVAL;
3989
2c78ee89 3990 if (!bpf_capable())
f56a653c
MKL		 3991 return -EPERM;
3992
c571bd75 3993 return btf_new_fd(attr, uattr);
f56a653c
MKL		 3994}
3995
78958fca
MKL		 3996#define BPF_BTF_GET_FD_BY_ID_LAST_FIELD btf_id
3997
3998static int bpf_btf_get_fd_by_id(const union bpf_attr *attr)
3999{
4000 if (CHECK_ATTR(BPF_BTF_GET_FD_BY_ID))
4001 return -EINVAL;
4002
4003 if (!capable(CAP_SYS_ADMIN))
4004 return -EPERM;
4005
4006 return btf_get_fd_by_id(attr->btf_id);
4007}
4008
41bdc4b4
YS		 4009static int bpf_task_fd_query_copy(const union bpf_attr *attr,
4010 union bpf_attr __user *uattr,
4011 u32 prog_id, u32 fd_type,
4012 const char *buf, u64 probe_offset,
4013 u64 probe_addr)
4014{
4015 char __user *ubuf = u64_to_user_ptr(attr->task_fd_query.buf);
4016 u32 len = buf ? strlen(buf) : 0, input_len;
4017 int err = 0;
4018
4019 if (put_user(len, &uattr->task_fd_query.buf_len))
4020 return -EFAULT;
4021 input_len = attr->task_fd_query.buf_len;
4022 if (input_len && ubuf) {
4023 if (!len) {
4024 /* nothing to copy, just make ubuf NULL terminated */
4025 char zero = '\0';
4026
4027 if (put_user(zero, ubuf))
4028 return -EFAULT;
4029 } else if (input_len >= len + 1) {
4030 /* ubuf can hold the string with NULL terminator */
4031 if (copy_to_user(ubuf, buf, len + 1))
4032 return -EFAULT;
4033 } else {
4034 /* ubuf cannot hold the string with NULL terminator,
4035 * do a partial copy with NULL terminator.
4036 */
4037 char zero = '\0';
4038
4039 err = -ENOSPC;
4040 if (copy_to_user(ubuf, buf, input_len - 1))
4041 return -EFAULT;
4042 if (put_user(zero, ubuf + input_len - 1))
4043 return -EFAULT;
4044 }
4045 }
4046
4047 if (put_user(prog_id, &uattr->task_fd_query.prog_id) ||
4048 put_user(fd_type, &uattr->task_fd_query.fd_type) ||
4049 put_user(probe_offset, &uattr->task_fd_query.probe_offset) ||
4050 put_user(probe_addr, &uattr->task_fd_query.probe_addr))
4051 return -EFAULT;
4052
4053 return err;
4054}
4055
4056#define BPF_TASK_FD_QUERY_LAST_FIELD task_fd_query.probe_addr
4057
4058static int bpf_task_fd_query(const union bpf_attr *attr,
4059 union bpf_attr __user *uattr)
4060{
4061 pid_t pid = attr->task_fd_query.pid;
4062 u32 fd = attr->task_fd_query.fd;
4063 const struct perf_event *event;
41bdc4b4
YS		 4064 struct task_struct *task;
4065 struct file *file;
4066 int err;
4067
4068 if (CHECK_ATTR(BPF_TASK_FD_QUERY))
4069 return -EINVAL;
4070
4071 if (!capable(CAP_SYS_ADMIN))
4072 return -EPERM;
4073
4074 if (attr->task_fd_query.flags != 0)
4075 return -EINVAL;
4076
4077 task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
4078 if (!task)
4079 return -ENOENT;
4080
41bdc4b4 4081 err = 0;
b48845af
EB		 4082 file = fget_task(task, fd);
4083 put_task_struct(task);
41bdc4b4 4084 if (!file)
b48845af 4085 return -EBADF;
41bdc4b4 4086
70ed506c
AN		 4087 if (file->f_op == &bpf_link_fops) {
4088 struct bpf_link *link = file->private_data;
41bdc4b4 4089
a3b80e10 4090 if (link->ops == &bpf_raw_tp_link_lops) {
70ed506c
AN		 4091 struct bpf_raw_tp_link *raw_tp =
4092 container_of(link, struct bpf_raw_tp_link, link);
4093 struct bpf_raw_event_map *btp = raw_tp->btp;
4094
4095 err = bpf_task_fd_query_copy(attr, uattr,
4096 raw_tp->link.prog->aux->id,
4097 BPF_FD_TYPE_RAW_TRACEPOINT,
4098 btp->tp->name, 0, 0);
4099 goto put_file;
4100 }
4101 goto out_not_supp;
41bdc4b4
YS		 4102 }
4103
4104 event = perf_get_event(file);
4105 if (!IS_ERR(event)) {
4106 u64 probe_offset, probe_addr;
4107 u32 prog_id, fd_type;
4108 const char *buf;
4109
4110 err = bpf_get_perf_event_info(event, &prog_id, &fd_type,
4111 &buf, &probe_offset,
4112 &probe_addr);
4113 if (!err)
4114 err = bpf_task_fd_query_copy(attr, uattr, prog_id,
4115 fd_type, buf,
4116 probe_offset,
4117 probe_addr);
4118 goto put_file;
4119 }
4120
70ed506c 4121out_not_supp:
41bdc4b4
YS		 4122 err = -ENOTSUPP;
4123put_file:
4124 fput(file);
41bdc4b4
YS		 4125 return err;
4126}
4127
cb4d03ab
BV		 4128#define BPF_MAP_BATCH_LAST_FIELD batch.flags
4129
4130#define BPF_DO_BATCH(fn) \
4131 do { \
4132 if (!fn) { \
4133 err = -ENOTSUPP; \
4134 goto err_put; \
4135 } \
4136 err = fn(map, attr, uattr); \
4137 } while (0)
4138
4139static int bpf_map_do_batch(const union bpf_attr *attr,
4140 union bpf_attr __user *uattr,
4141 int cmd)
4142{
4143 struct bpf_map *map;
4144 int err, ufd;
4145 struct fd f;
4146
4147 if (CHECK_ATTR(BPF_MAP_BATCH))
4148 return -EINVAL;
4149
4150 ufd = attr->batch.map_fd;
4151 f = fdget(ufd);
4152 map = __bpf_map_get(f);
4153 if (IS_ERR(map))
4154 return PTR_ERR(map);
4155
05799638
YS		 4156 if ((cmd == BPF_MAP_LOOKUP_BATCH ||
4157 cmd == BPF_MAP_LOOKUP_AND_DELETE_BATCH) &&
cb4d03ab
BV		 4158 !(map_get_sys_perms(map, f) & FMODE_CAN_READ)) {
4159 err = -EPERM;
4160 goto err_put;
4161 }
4162
4163 if (cmd != BPF_MAP_LOOKUP_BATCH &&
4164 !(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) {
4165 err = -EPERM;
4166 goto err_put;
4167 }
4168
4169 if (cmd == BPF_MAP_LOOKUP_BATCH)
4170 BPF_DO_BATCH(map->ops->map_lookup_batch);
05799638
YS		 4171 else if (cmd == BPF_MAP_LOOKUP_AND_DELETE_BATCH)
4172 BPF_DO_BATCH(map->ops->map_lookup_and_delete_batch);
aa2e93b8
BV		 4173 else if (cmd == BPF_MAP_UPDATE_BATCH)
4174 BPF_DO_BATCH(map->ops->map_update_batch);
4175 else
4176 BPF_DO_BATCH(map->ops->map_delete_batch);
cb4d03ab
BV		 4177
4178err_put:
4179 fdput(f);
4180 return err;
4181}
4182
af2ac3e1
AS		 4183static int tracing_bpf_link_attach(const union bpf_attr *attr, bpfptr_t uattr,
4184 struct bpf_prog *prog)
de4e05ca 4185{
4a1e7c0c
THJ		 4186 if (attr->link_create.attach_type != prog->expected_attach_type)
4187 return -EINVAL;
de4e05ca 4188
4a1e7c0c 4189 if (prog->expected_attach_type == BPF_TRACE_ITER)
af2ac3e1 4190 return bpf_iter_link_attach(attr, uattr, prog);
4a1e7c0c
THJ		 4191 else if (prog->type == BPF_PROG_TYPE_EXT)
4192 return bpf_tracing_prog_attach(prog,
4193 attr->link_create.target_fd,
4194 attr->link_create.target_btf_id);
de4e05ca
YS		 4195 return -EINVAL;
4196}
4197
5e7b3020 4198#define BPF_LINK_CREATE_LAST_FIELD link_create.iter_info_len
af2ac3e1 4199static int link_create(union bpf_attr *attr, bpfptr_t uattr)
af6eea57
AN		 4200{
4201 enum bpf_prog_type ptype;
4202 struct bpf_prog *prog;
4203 int ret;
4204
af6eea57
AN		 4205 if (CHECK_ATTR(BPF_LINK_CREATE))
4206 return -EINVAL;
4207
4a1e7c0c 4208 prog = bpf_prog_get(attr->link_create.prog_fd);
af6eea57
AN		 4209 if (IS_ERR(prog))
4210 return PTR_ERR(prog);
4211
4212 ret = bpf_prog_attach_check_attach_type(prog,
4213 attr->link_create.attach_type);
4214 if (ret)
4a1e7c0c
THJ		 4215 goto out;
4216
b89fbfbb
AN		 4217 switch (prog->type) {
4218 case BPF_PROG_TYPE_EXT:
af2ac3e1 4219 ret = tracing_bpf_link_attach(attr, uattr, prog);
4a1e7c0c 4220 goto out;
b89fbfbb
AN		 4221 case BPF_PROG_TYPE_PERF_EVENT:
4222 case BPF_PROG_TYPE_KPROBE:
4223 case BPF_PROG_TYPE_TRACEPOINT:
4224 if (attr->link_create.attach_type != BPF_PERF_EVENT) {
4225 ret = -EINVAL;
4226 goto out;
4227 }
4228 ptype = prog->type;
4229 break;
4230 default:
4231 ptype = attach_type_to_prog_type(attr->link_create.attach_type);
4232 if (ptype == BPF_PROG_TYPE_UNSPEC || ptype != prog->type) {
4233 ret = -EINVAL;
4234 goto out;
4235 }
4236 break;
4a1e7c0c 4237 }
af6eea57
AN		 4238
4239 switch (ptype) {
4240 case BPF_PROG_TYPE_CGROUP_SKB:
4241 case BPF_PROG_TYPE_CGROUP_SOCK:
4242 case BPF_PROG_TYPE_CGROUP_SOCK_ADDR:
4243 case BPF_PROG_TYPE_SOCK_OPS:
4244 case BPF_PROG_TYPE_CGROUP_DEVICE:
4245 case BPF_PROG_TYPE_CGROUP_SYSCTL:
4246 case BPF_PROG_TYPE_CGROUP_SOCKOPT:
4247 ret = cgroup_bpf_link_attach(attr, prog);
4248 break;
de4e05ca 4249 case BPF_PROG_TYPE_TRACING:
af2ac3e1 4250 ret = tracing_bpf_link_attach(attr, uattr, prog);
de4e05ca 4251 break;
7f045a49 4252 case BPF_PROG_TYPE_FLOW_DISSECTOR:
e9ddbb77 4253 case BPF_PROG_TYPE_SK_LOOKUP:
7f045a49
JS		 4254 ret = netns_bpf_link_create(attr, prog);
4255 break;
310ad797 4256#ifdef CONFIG_NET
aa8d3a71
AN		 4257 case BPF_PROG_TYPE_XDP:
4258 ret = bpf_xdp_link_attach(attr, prog);
4259 break;
b89fbfbb
AN		 4260#endif
4261#ifdef CONFIG_PERF_EVENTS
4262 case BPF_PROG_TYPE_PERF_EVENT:
4263 case BPF_PROG_TYPE_TRACEPOINT:
4264 case BPF_PROG_TYPE_KPROBE:
4265 ret = bpf_perf_link_attach(attr, prog);
4266 break;
310ad797 4267#endif
af6eea57
AN		 4268 default:
4269 ret = -EINVAL;
4270 }
4271
4a1e7c0c 4272out:
af6eea57
AN		 4273 if (ret < 0)
4274 bpf_prog_put(prog);
4275 return ret;
4276}
4277
0c991ebc
AN		 4278#define BPF_LINK_UPDATE_LAST_FIELD link_update.old_prog_fd
4279
4280static int link_update(union bpf_attr *attr)
4281{
4282 struct bpf_prog *old_prog = NULL, *new_prog;
4283 struct bpf_link *link;
4284 u32 flags;
4285 int ret;
4286
0c991ebc
AN		 4287 if (CHECK_ATTR(BPF_LINK_UPDATE))
4288 return -EINVAL;
4289
4290 flags = attr->link_update.flags;
4291 if (flags & ~BPF_F_REPLACE)
4292 return -EINVAL;
4293
4294 link = bpf_link_get_from_fd(attr->link_update.link_fd);
4295 if (IS_ERR(link))
4296 return PTR_ERR(link);
4297
4298 new_prog = bpf_prog_get(attr->link_update.new_prog_fd);
4adb7a4a
AN		 4299 if (IS_ERR(new_prog)) {
4300 ret = PTR_ERR(new_prog);
4301 goto out_put_link;
4302 }
0c991ebc
AN		 4303
4304 if (flags & BPF_F_REPLACE) {
4305 old_prog = bpf_prog_get(attr->link_update.old_prog_fd);
4306 if (IS_ERR(old_prog)) {
4307 ret = PTR_ERR(old_prog);
4308 old_prog = NULL;
4309 goto out_put_progs;
4310 }
4adb7a4a
AN		 4311 } else if (attr->link_update.old_prog_fd) {
4312 ret = -EINVAL;
4313 goto out_put_progs;
0c991ebc
AN		 4314 }
4315
f9d04127
AN		 4316 if (link->ops->update_prog)
4317 ret = link->ops->update_prog(link, new_prog, old_prog);
4318 else
fe537393 4319 ret = -EINVAL;
0c991ebc
AN		 4320
4321out_put_progs:
4322 if (old_prog)
4323 bpf_prog_put(old_prog);
4324 if (ret)
4325 bpf_prog_put(new_prog);
4adb7a4a
AN		 4326out_put_link:
4327 bpf_link_put(link);
0c991ebc
AN		 4328 return ret;
4329}
4330
73b11c2a
AN		 4331#define BPF_LINK_DETACH_LAST_FIELD link_detach.link_fd
4332
4333static int link_detach(union bpf_attr *attr)
4334{
4335 struct bpf_link *link;
4336 int ret;
4337
4338 if (CHECK_ATTR(BPF_LINK_DETACH))
4339 return -EINVAL;
4340
4341 link = bpf_link_get_from_fd(attr->link_detach.link_fd);
4342 if (IS_ERR(link))
4343 return PTR_ERR(link);
4344
4345 if (link->ops->detach)
4346 ret = link->ops->detach(link);
4347 else
4348 ret = -EOPNOTSUPP;
4349
4350 bpf_link_put(link);
4351 return ret;
4352}
4353
005142b8 4354static struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link)
2d602c8c 4355{
005142b8 4356 return atomic64_fetch_add_unless(&link->refcnt, 1, 0) ? link : ERR_PTR(-ENOENT);
2d602c8c
AN		 4357}
4358
005142b8 4359struct bpf_link *bpf_link_by_id(u32 id)
2d602c8c
AN		 4360{
4361 struct bpf_link *link;
2d602c8c 4362
005142b8
AS		 4363 if (!id)
4364 return ERR_PTR(-ENOENT);
2d602c8c
AN		 4365
4366 spin_lock_bh(&link_idr_lock);
2d602c8c 4367 /* before link is "settled", ID is 0, pretend it doesn't exist yet */
005142b8 4368 link = idr_find(&link_idr, id);
2d602c8c
AN		 4369 if (link) {
4370 if (link->id)
005142b8 4371 link = bpf_link_inc_not_zero(link);
2d602c8c 4372 else
005142b8 4373 link = ERR_PTR(-EAGAIN);
2d602c8c 4374 } else {
005142b8 4375 link = ERR_PTR(-ENOENT);
2d602c8c
AN		 4376 }
4377 spin_unlock_bh(&link_idr_lock);
005142b8
AS		 4378 return link;
4379}
2d602c8c 4380
005142b8
AS		 4381#define BPF_LINK_GET_FD_BY_ID_LAST_FIELD link_id
4382
4383static int bpf_link_get_fd_by_id(const union bpf_attr *attr)
4384{
4385 struct bpf_link *link;
4386 u32 id = attr->link_id;
4387 int fd;
4388
4389 if (CHECK_ATTR(BPF_LINK_GET_FD_BY_ID))
4390 return -EINVAL;
4391
4392 if (!capable(CAP_SYS_ADMIN))
4393 return -EPERM;
4394
4395 link = bpf_link_by_id(id);
4396 if (IS_ERR(link))
4397 return PTR_ERR(link);
2d602c8c
AN		 4398
4399 fd = bpf_link_new_fd(link);
4400 if (fd < 0)
4401 bpf_link_put(link);
4402
4403 return fd;
4404}
4405
d46edd67
SL		 4406DEFINE_MUTEX(bpf_stats_enabled_mutex);
4407
4408static int bpf_stats_release(struct inode *inode, struct file *file)
4409{
4410 mutex_lock(&bpf_stats_enabled_mutex);
4411 static_key_slow_dec(&bpf_stats_enabled_key.key);
4412 mutex_unlock(&bpf_stats_enabled_mutex);
4413 return 0;
4414}
4415
4416static const struct file_operations bpf_stats_fops = {
4417 .release = bpf_stats_release,
4418};
4419
4420static int bpf_enable_runtime_stats(void)
4421{
4422 int fd;
4423
4424 mutex_lock(&bpf_stats_enabled_mutex);
4425
4426 /* Set a very high limit to avoid overflow */
4427 if (static_key_count(&bpf_stats_enabled_key.key) > INT_MAX / 2) {
4428 mutex_unlock(&bpf_stats_enabled_mutex);
4429 return -EBUSY;
4430 }
4431
4432 fd = anon_inode_getfd("bpf-stats", &bpf_stats_fops, NULL, O_CLOEXEC);
4433 if (fd >= 0)
4434 static_key_slow_inc(&bpf_stats_enabled_key.key);
4435
4436 mutex_unlock(&bpf_stats_enabled_mutex);
4437 return fd;
4438}
4439
4440#define BPF_ENABLE_STATS_LAST_FIELD enable_stats.type
4441
4442static int bpf_enable_stats(union bpf_attr *attr)
4443{
4444
4445 if (CHECK_ATTR(BPF_ENABLE_STATS))
4446 return -EINVAL;
4447
4448 if (!capable(CAP_SYS_ADMIN))
4449 return -EPERM;
4450
4451 switch (attr->enable_stats.type) {
4452 case BPF_STATS_RUN_TIME:
4453 return bpf_enable_runtime_stats();
4454 default:
4455 break;
4456 }
4457 return -EINVAL;
4458}
4459
ac51d99b
YS		 4460#define BPF_ITER_CREATE_LAST_FIELD iter_create.flags
4461
4462static int bpf_iter_create(union bpf_attr *attr)
4463{
4464 struct bpf_link *link;
4465 int err;
4466
4467 if (CHECK_ATTR(BPF_ITER_CREATE))
4468 return -EINVAL;
4469
4470 if (attr->iter_create.flags)
4471 return -EINVAL;
4472
4473 link = bpf_link_get_from_fd(attr->iter_create.link_fd);
4474 if (IS_ERR(link))
4475 return PTR_ERR(link);
4476
4477 err = bpf_iter_new_fd(link);
4478 bpf_link_put(link);
4479
4480 return err;
4481}
4482
ef15314a
YZ		 4483#define BPF_PROG_BIND_MAP_LAST_FIELD prog_bind_map.flags
4484
4485static int bpf_prog_bind_map(union bpf_attr *attr)
4486{
4487 struct bpf_prog *prog;
4488 struct bpf_map *map;
4489 struct bpf_map **used_maps_old, **used_maps_new;
4490 int i, ret = 0;
4491
4492 if (CHECK_ATTR(BPF_PROG_BIND_MAP))
4493 return -EINVAL;
4494
4495 if (attr->prog_bind_map.flags)
4496 return -EINVAL;
4497
4498 prog = bpf_prog_get(attr->prog_bind_map.prog_fd);
4499 if (IS_ERR(prog))
4500 return PTR_ERR(prog);
4501
4502 map = bpf_map_get(attr->prog_bind_map.map_fd);
4503 if (IS_ERR(map)) {
4504 ret = PTR_ERR(map);
4505 goto out_prog_put;
4506 }
4507
4508 mutex_lock(&prog->aux->used_maps_mutex);
4509
4510 used_maps_old = prog->aux->used_maps;
4511
4512 for (i = 0; i < prog->aux->used_map_cnt; i++)
1028ae40
SF		 4513 if (used_maps_old[i] == map) {
4514 bpf_map_put(map);
ef15314a 4515 goto out_unlock;
1028ae40 4516 }
ef15314a
YZ		 4517
4518 used_maps_new = kmalloc_array(prog->aux->used_map_cnt + 1,
4519 sizeof(used_maps_new[0]),
4520 GFP_KERNEL);
4521 if (!used_maps_new) {
4522 ret = -ENOMEM;
4523 goto out_unlock;
4524 }
4525
4526 memcpy(used_maps_new, used_maps_old,
4527 sizeof(used_maps_old[0]) * prog->aux->used_map_cnt);
4528 used_maps_new[prog->aux->used_map_cnt] = map;
4529
4530 prog->aux->used_map_cnt++;
4531 prog->aux->used_maps = used_maps_new;
4532
4533 kfree(used_maps_old);
4534
4535out_unlock:
4536 mutex_unlock(&prog->aux->used_maps_mutex);
4537
4538 if (ret)
4539 bpf_map_put(map);
4540out_prog_put:
4541 bpf_prog_put(prog);
4542 return ret;
4543}
4544
af2ac3e1 4545static int __sys_bpf(int cmd, bpfptr_t uattr, unsigned int size)
99c55f7d 4546{
8096f229 4547 union bpf_attr attr;
99c55f7d
AS		 4548 int err;
4549
2c78ee89 4550 if (sysctl_unprivileged_bpf_disabled && !bpf_capable())
99c55f7d
AS		 4551 return -EPERM;
4552
dcab51f1 4553 err = bpf_check_uarg_tail_zero(uattr, sizeof(attr), size);
1e270976
MKL		 4554 if (err)
4555 return err;
4556 size = min_t(u32, size, sizeof(attr));
99c55f7d
AS		 4557
4558 /* copy attributes from user space, may be less than sizeof(bpf_attr) */
8096f229 4559 memset(&attr, 0, sizeof(attr));
af2ac3e1 4560 if (copy_from_bpfptr(&attr, uattr, size) != 0)
99c55f7d
AS		 4561 return -EFAULT;
4562
afdb09c7
CF		 4563 err = security_bpf(cmd, &attr, size);
4564 if (err < 0)
4565 return err;
4566
99c55f7d
AS		 4567 switch (cmd) {
4568 case BPF_MAP_CREATE:
4569 err = map_create(&attr);
4570 break;
db20fd2b
AS		 4571 case BPF_MAP_LOOKUP_ELEM:
4572 err = map_lookup_elem(&attr);
4573 break;
4574 case BPF_MAP_UPDATE_ELEM:
af2ac3e1 4575 err = map_update_elem(&attr, uattr);
db20fd2b
AS		 4576 break;
4577 case BPF_MAP_DELETE_ELEM:
4578 err = map_delete_elem(&attr);
4579 break;
4580 case BPF_MAP_GET_NEXT_KEY:
4581 err = map_get_next_key(&attr);
4582 break;
87df15de
DB		 4583 case BPF_MAP_FREEZE:
4584 err = map_freeze(&attr);
4585 break;
09756af4 4586 case BPF_PROG_LOAD:
838e9690 4587 err = bpf_prog_load(&attr, uattr);
09756af4 4588 break;
b2197755
DB		 4589 case BPF_OBJ_PIN:
4590 err = bpf_obj_pin(&attr);
4591 break;
4592 case BPF_OBJ_GET:
4593 err = bpf_obj_get(&attr);
4594 break;
f4324551
DM		 4595 case BPF_PROG_ATTACH:
4596 err = bpf_prog_attach(&attr);
4597 break;
4598 case BPF_PROG_DETACH:
4599 err = bpf_prog_detach(&attr);
4600 break;
468e2f64 4601 case BPF_PROG_QUERY:
af2ac3e1 4602 err = bpf_prog_query(&attr, uattr.user);
468e2f64 4603 break;
1cf1cae9 4604 case BPF_PROG_TEST_RUN:
af2ac3e1 4605 err = bpf_prog_test_run(&attr, uattr.user);
1cf1cae9 4606 break;
34ad5580 4607 case BPF_PROG_GET_NEXT_ID:
af2ac3e1 4608 err = bpf_obj_get_next_id(&attr, uattr.user,
34ad5580
MKL		 4609 &prog_idr, &prog_idr_lock);
4610 break;
4611 case BPF_MAP_GET_NEXT_ID:
af2ac3e1 4612 err = bpf_obj_get_next_id(&attr, uattr.user,
34ad5580
MKL		 4613 &map_idr, &map_idr_lock);
4614 break;
1b9ed84e 4615 case BPF_BTF_GET_NEXT_ID:
af2ac3e1 4616 err = bpf_obj_get_next_id(&attr, uattr.user,
1b9ed84e
QM		 4617 &btf_idr, &btf_idr_lock);
4618 break;
b16d9aa4
MKL		 4619 case BPF_PROG_GET_FD_BY_ID:
4620 err = bpf_prog_get_fd_by_id(&attr);
4621 break;
bd5f5f4e
MKL		 4622 case BPF_MAP_GET_FD_BY_ID:
4623 err = bpf_map_get_fd_by_id(&attr);
4624 break;
1e270976 4625 case BPF_OBJ_GET_INFO_BY_FD:
af2ac3e1 4626 err = bpf_obj_get_info_by_fd(&attr, uattr.user);
1e270976 4627 break;
c4f6699d
AS		 4628 case BPF_RAW_TRACEPOINT_OPEN:
4629 err = bpf_raw_tracepoint_open(&attr);
4630 break;
f56a653c 4631 case BPF_BTF_LOAD:
c571bd75 4632 err = bpf_btf_load(&attr, uattr);
f56a653c 4633 break;
78958fca
MKL		 4634 case BPF_BTF_GET_FD_BY_ID:
4635 err = bpf_btf_get_fd_by_id(&attr);
4636 break;
41bdc4b4 4637 case BPF_TASK_FD_QUERY:
af2ac3e1 4638 err = bpf_task_fd_query(&attr, uattr.user);
41bdc4b4 4639 break;
bd513cd0
MV		 4640 case BPF_MAP_LOOKUP_AND_DELETE_ELEM:
4641 err = map_lookup_and_delete_elem(&attr);
4642 break;
cb4d03ab 4643 case BPF_MAP_LOOKUP_BATCH:
af2ac3e1 4644 err = bpf_map_do_batch(&attr, uattr.user, BPF_MAP_LOOKUP_BATCH);
cb4d03ab 4645 break;
05799638 4646 case BPF_MAP_LOOKUP_AND_DELETE_BATCH:
af2ac3e1 4647 err = bpf_map_do_batch(&attr, uattr.user,
05799638
YS		 4648 BPF_MAP_LOOKUP_AND_DELETE_BATCH);
4649 break;
aa2e93b8 4650 case BPF_MAP_UPDATE_BATCH:
af2ac3e1 4651 err = bpf_map_do_batch(&attr, uattr.user, BPF_MAP_UPDATE_BATCH);
aa2e93b8
BV		 4652 break;
4653 case BPF_MAP_DELETE_BATCH:
af2ac3e1 4654 err = bpf_map_do_batch(&attr, uattr.user, BPF_MAP_DELETE_BATCH);
aa2e93b8 4655 break;
af6eea57 4656 case BPF_LINK_CREATE:
af2ac3e1 4657 err = link_create(&attr, uattr);
af6eea57 4658 break;
0c991ebc
AN		 4659 case BPF_LINK_UPDATE:
4660 err = link_update(&attr);
4661 break;
2d602c8c
AN		 4662 case BPF_LINK_GET_FD_BY_ID:
4663 err = bpf_link_get_fd_by_id(&attr);
4664 break;
4665 case BPF_LINK_GET_NEXT_ID:
af2ac3e1 4666 err = bpf_obj_get_next_id(&attr, uattr.user,
2d602c8c
AN		 4667 &link_idr, &link_idr_lock);
4668 break;
d46edd67
SL		 4669 case BPF_ENABLE_STATS:
4670 err = bpf_enable_stats(&attr);
4671 break;
ac51d99b
YS		 4672 case BPF_ITER_CREATE:
4673 err = bpf_iter_create(&attr);
4674 break;
73b11c2a
AN		 4675 case BPF_LINK_DETACH:
4676 err = link_detach(&attr);
4677 break;
ef15314a
YZ		 4678 case BPF_PROG_BIND_MAP:
4679 err = bpf_prog_bind_map(&attr);
4680 break;
99c55f7d
AS		 4681 default:
4682 err = -EINVAL;
4683 break;
4684 }
4685
4686 return err;
4687}
79a7f8bd 4688
af2ac3e1
AS		 4689SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size)
4690{
4691 return __sys_bpf(cmd, USER_BPFPTR(uattr), size);
4692}
4693
79a7f8bd
AS		 4694static bool syscall_prog_is_valid_access(int off, int size,
4695 enum bpf_access_type type,
4696 const struct bpf_prog *prog,
4697 struct bpf_insn_access_aux *info)
4698{
4699 if (off < 0 || off >= U16_MAX)
4700 return false;
4701 if (off % size != 0)
4702 return false;
4703 return true;
4704}
4705
4706BPF_CALL_3(bpf_sys_bpf, int, cmd, void *, attr, u32, attr_size)
4707{
af2ac3e1
AS		 4708 switch (cmd) {
4709 case BPF_MAP_CREATE:
4710 case BPF_MAP_UPDATE_ELEM:
4711 case BPF_MAP_FREEZE:
4712 case BPF_PROG_LOAD:
c571bd75 4713 case BPF_BTF_LOAD:
af2ac3e1
AS		 4714 break;
4715 /* case BPF_PROG_TEST_RUN:
4716 * is not part of this list to prevent recursive test_run
4717 */
4718 default:
4719 return -EINVAL;
4720 }
4721 return __sys_bpf(cmd, KERNEL_BPFPTR(attr), attr_size);
79a7f8bd
AS		 4722}
4723
3a2daa72 4724static const struct bpf_func_proto bpf_sys_bpf_proto = {
79a7f8bd
AS		 4725 .func = bpf_sys_bpf,
4726 .gpl_only = false,
4727 .ret_type = RET_INTEGER,
4728 .arg1_type = ARG_ANYTHING,
4729 .arg2_type = ARG_PTR_TO_MEM,
4730 .arg3_type = ARG_CONST_SIZE,
4731};
4732
4733const struct bpf_func_proto * __weak
4734tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
4735{
4736 return bpf_base_func_proto(func_id);
4737}
4738
3abea089
AS		 4739BPF_CALL_1(bpf_sys_close, u32, fd)
4740{
4741 /* When bpf program calls this helper there should not be
4742 * an fdget() without matching completed fdput().
4743 * This helper is allowed in the following callchain only:
4744 * sys_bpf->prog_test_run->bpf_prog->bpf_sys_close
4745 */
4746 return close_fd(fd);
4747}
4748
3a2daa72 4749static const struct bpf_func_proto bpf_sys_close_proto = {
3abea089
AS		 4750 .func = bpf_sys_close,
4751 .gpl_only = false,
4752 .ret_type = RET_INTEGER,
4753 .arg1_type = ARG_ANYTHING,
4754};
4755
79a7f8bd
AS		 4756static const struct bpf_func_proto *
4757syscall_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
4758{
4759 switch (func_id) {
4760 case BPF_FUNC_sys_bpf:
4761 return &bpf_sys_bpf_proto;
3d78417b
AS		 4762 case BPF_FUNC_btf_find_by_name_kind:
4763 return &bpf_btf_find_by_name_kind_proto;
3abea089
AS		 4764 case BPF_FUNC_sys_close:
4765 return &bpf_sys_close_proto;
79a7f8bd
AS		 4766 default:
4767 return tracing_prog_func_proto(func_id, prog);
4768 }
4769}
4770
4771const struct bpf_verifier_ops bpf_syscall_verifier_ops = {
4772 .get_func_proto = syscall_prog_func_proto,
4773 .is_valid_access = syscall_prog_is_valid_access,
4774};
4775
4776const struct bpf_prog_ops bpf_syscall_prog_ops = {
4777 .test_run = bpf_prog_test_run_syscall,
4778};
Ubuntu Jammy Linux kernel mirror