Commit | Line | Data |
---|---|---|
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> | |
aef2feda | 5 | #include <linux/bpf-cgroup.h> |
a67edbf4 | 6 | #include <linux/bpf_trace.h> |
f4364dcf | 7 | #include <linux/bpf_lirc.h> |
4a1e7c0c | 8 | #include <linux/bpf_verifier.h> |
61df10c7 | 9 | #include <linux/bsearch.h> |
f56a653c | 10 | #include <linux/btf.h> |
99c55f7d AS |
11 | #include <linux/syscalls.h> |
12 | #include <linux/slab.h> | |
3f07c014 | 13 | #include <linux/sched/signal.h> |
d407bd25 DB |
14 | #include <linux/vmalloc.h> |
15 | #include <linux/mmzone.h> | |
99c55f7d | 16 | #include <linux/anon_inodes.h> |
41bdc4b4 | 17 | #include <linux/fdtable.h> |
db20fd2b | 18 | #include <linux/file.h> |
41bdc4b4 | 19 | #include <linux/fs.h> |
09756af4 AS |
20 | #include <linux/license.h> |
21 | #include <linux/filter.h> | |
535e7b4b | 22 | #include <linux/kernel.h> |
dc4bb0e2 | 23 | #include <linux/idr.h> |
cb4d2b3f MKL |
24 | #include <linux/cred.h> |
25 | #include <linux/timekeeping.h> | |
26 | #include <linux/ctype.h> | |
9ef09e35 | 27 | #include <linux/nospec.h> |
bae141f5 | 28 | #include <linux/audit.h> |
ccfe29eb | 29 | #include <uapi/linux/btf.h> |
ca5999fd | 30 | #include <linux/pgtable.h> |
9e4e01df | 31 | #include <linux/bpf_lsm.h> |
457f4436 | 32 | #include <linux/poll.h> |
4d7d7f69 | 33 | #include <linux/sort.h> |
a3fd7cee | 34 | #include <linux/bpf-netns.h> |
1e6c62a8 | 35 | #include <linux/rcupdate_trace.h> |
48edc1f7 | 36 | #include <linux/memcontrol.h> |
0dcac272 | 37 | #include <linux/trace_events.h> |
84601d6e | 38 | #include <net/netfilter/nf_bpf_link.h> |
99c55f7d | 39 | |
e420bed0 DB |
40 | #include <net/tcx.h> |
41 | ||
da765a2f DB |
42 | #define IS_FD_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || \ |
43 | (map)->map_type == BPF_MAP_TYPE_CGROUP_ARRAY || \ | |
44 | (map)->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS) | |
45 | #define IS_FD_PROG_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PROG_ARRAY) | |
14dc6f04 | 46 | #define IS_FD_HASH(map) ((map)->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) |
da765a2f DB |
47 | #define IS_FD_MAP(map) (IS_FD_ARRAY(map) || IS_FD_PROG_ARRAY(map) || \ |
48 | IS_FD_HASH(map)) | |
14dc6f04 | 49 | |
6e71b04a CF |
50 | #define BPF_OBJ_FLAG_MASK (BPF_F_RDONLY | BPF_F_WRONLY) |
51 | ||
b121d1e7 | 52 | DEFINE_PER_CPU(int, bpf_prog_active); |
dc4bb0e2 MKL |
53 | static DEFINE_IDR(prog_idr); |
54 | static DEFINE_SPINLOCK(prog_idr_lock); | |
f3f1c054 MKL |
55 | static DEFINE_IDR(map_idr); |
56 | static DEFINE_SPINLOCK(map_idr_lock); | |
a3b80e10 AN |
57 | static DEFINE_IDR(link_idr); |
58 | static DEFINE_SPINLOCK(link_idr_lock); | |
b121d1e7 | 59 | |
08389d88 DB |
60 | int sysctl_unprivileged_bpf_disabled __read_mostly = |
61 | IS_BUILTIN(CONFIG_BPF_UNPRIV_DEFAULT_OFF) ? 2 : 0; | |
1be7f75d | 62 | |
40077e0c | 63 | static const struct bpf_map_ops * const bpf_map_types[] = { |
91cc1a99 | 64 | #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) |
40077e0c JB |
65 | #define BPF_MAP_TYPE(_id, _ops) \ |
66 | [_id] = &_ops, | |
f2e10bff | 67 | #define BPF_LINK_TYPE(_id, _name) |
40077e0c JB |
68 | #include <linux/bpf_types.h> |
69 | #undef BPF_PROG_TYPE | |
70 | #undef BPF_MAP_TYPE | |
f2e10bff | 71 | #undef BPF_LINK_TYPE |
40077e0c | 72 | }; |
99c55f7d | 73 | |
752ba56f MS |
74 | /* |
75 | * If we're handed a bigger struct than we know of, ensure all the unknown bits | |
76 | * are 0 - i.e. new user-space does not rely on any kernel feature extensions | |
77 | * we don't know about yet. | |
78 | * | |
79 | * There is a ToCToU between this function call and the following | |
80 | * copy_from_user() call. However, this is not a concern since this function is | |
81 | * meant to be a future-proofing of bits. | |
82 | */ | |
af2ac3e1 | 83 | int bpf_check_uarg_tail_zero(bpfptr_t uaddr, |
dcab51f1 MKL |
84 | size_t expected_size, |
85 | size_t actual_size) | |
58291a74 | 86 | { |
b7e4b65f | 87 | int res; |
58291a74 | 88 | |
752ba56f MS |
89 | if (unlikely(actual_size > PAGE_SIZE)) /* silly large */ |
90 | return -E2BIG; | |
91 | ||
58291a74 MS |
92 | if (actual_size <= expected_size) |
93 | return 0; | |
94 | ||
af2ac3e1 AS |
95 | if (uaddr.is_kernel) |
96 | res = memchr_inv(uaddr.kernel + expected_size, 0, | |
97 | actual_size - expected_size) == NULL; | |
98 | else | |
99 | res = check_zeroed_user(uaddr.user + expected_size, | |
100 | actual_size - expected_size); | |
b7e4b65f AV |
101 | if (res < 0) |
102 | return res; | |
103 | return res ? 0 : -E2BIG; | |
58291a74 MS |
104 | } |
105 | ||
a3884572 | 106 | const struct bpf_map_ops bpf_map_offload_ops = { |
f4d05259 | 107 | .map_meta_equal = bpf_map_meta_equal, |
a3884572 JK |
108 | .map_alloc = bpf_map_offload_map_alloc, |
109 | .map_free = bpf_map_offload_map_free, | |
e8d2bec0 | 110 | .map_check_btf = map_check_no_btf, |
9629363c | 111 | .map_mem_usage = bpf_map_offload_map_mem_usage, |
a3884572 JK |
112 | }; |
113 | ||
353050be DB |
114 | static void bpf_map_write_active_inc(struct bpf_map *map) |
115 | { | |
116 | atomic64_inc(&map->writecnt); | |
117 | } | |
118 | ||
119 | static void bpf_map_write_active_dec(struct bpf_map *map) | |
120 | { | |
121 | atomic64_dec(&map->writecnt); | |
122 | } | |
123 | ||
124 | bool bpf_map_write_active(const struct bpf_map *map) | |
125 | { | |
126 | return atomic64_read(&map->writecnt) != 0; | |
127 | } | |
128 | ||
80ee81e0 | 129 | static u32 bpf_map_value_size(const struct bpf_map *map) |
15c14a3d BV |
130 | { |
131 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || | |
132 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH || | |
133 | map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY || | |
134 | map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE) | |
135 | return round_up(map->value_size, 8) * num_possible_cpus(); | |
136 | else if (IS_FD_MAP(map)) | |
137 | return sizeof(u32); | |
138 | else | |
139 | return map->value_size; | |
140 | } | |
141 | ||
142 | static void maybe_wait_bpf_programs(struct bpf_map *map) | |
143 | { | |
144 | /* Wait for any running BPF programs to complete so that | |
145 | * userspace, when we return to it, knows that all programs | |
146 | * that could be running use the new map value. | |
147 | */ | |
148 | if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS || | |
149 | map->map_type == BPF_MAP_TYPE_ARRAY_OF_MAPS) | |
150 | synchronize_rcu(); | |
151 | } | |
152 | ||
3af43ba4 HT |
153 | static int bpf_map_update_value(struct bpf_map *map, struct file *map_file, |
154 | void *key, void *value, __u64 flags) | |
15c14a3d BV |
155 | { |
156 | int err; | |
157 | ||
158 | /* Need to create a kthread, thus must support schedule */ | |
9d03ebc7 | 159 | if (bpf_map_is_offloaded(map)) { |
15c14a3d BV |
160 | return bpf_map_offload_update_elem(map, key, value, flags); |
161 | } else if (map->map_type == BPF_MAP_TYPE_CPUMAP || | |
15c14a3d BV |
162 | map->map_type == BPF_MAP_TYPE_STRUCT_OPS) { |
163 | return map->ops->map_update_elem(map, key, value, flags); | |
13b79d3f LB |
164 | } else if (map->map_type == BPF_MAP_TYPE_SOCKHASH || |
165 | map->map_type == BPF_MAP_TYPE_SOCKMAP) { | |
166 | return sock_map_update_elem_sys(map, key, value, flags); | |
15c14a3d | 167 | } else if (IS_FD_PROG_ARRAY(map)) { |
3af43ba4 | 168 | return bpf_fd_array_map_update_elem(map, map_file, key, value, |
15c14a3d BV |
169 | flags); |
170 | } | |
171 | ||
b6e5dae1 | 172 | bpf_disable_instrumentation(); |
15c14a3d BV |
173 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
174 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
175 | err = bpf_percpu_hash_update(map, key, value, flags); | |
176 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
177 | err = bpf_percpu_array_update(map, key, value, flags); | |
178 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE) { | |
179 | err = bpf_percpu_cgroup_storage_update(map, key, value, | |
180 | flags); | |
181 | } else if (IS_FD_ARRAY(map)) { | |
182 | rcu_read_lock(); | |
3af43ba4 | 183 | err = bpf_fd_array_map_update_elem(map, map_file, key, value, |
15c14a3d BV |
184 | flags); |
185 | rcu_read_unlock(); | |
186 | } else if (map->map_type == BPF_MAP_TYPE_HASH_OF_MAPS) { | |
187 | rcu_read_lock(); | |
3af43ba4 | 188 | err = bpf_fd_htab_map_update_elem(map, map_file, key, value, |
15c14a3d BV |
189 | flags); |
190 | rcu_read_unlock(); | |
191 | } else if (map->map_type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY) { | |
192 | /* rcu_read_lock() is not needed */ | |
193 | err = bpf_fd_reuseport_array_update_elem(map, key, value, | |
194 | flags); | |
195 | } else if (map->map_type == BPF_MAP_TYPE_QUEUE || | |
9330986c JK |
196 | map->map_type == BPF_MAP_TYPE_STACK || |
197 | map->map_type == BPF_MAP_TYPE_BLOOM_FILTER) { | |
15c14a3d BV |
198 | err = map->ops->map_push_elem(map, value, flags); |
199 | } else { | |
200 | rcu_read_lock(); | |
201 | err = map->ops->map_update_elem(map, key, value, flags); | |
202 | rcu_read_unlock(); | |
203 | } | |
b6e5dae1 | 204 | bpf_enable_instrumentation(); |
15c14a3d BV |
205 | maybe_wait_bpf_programs(map); |
206 | ||
207 | return err; | |
208 | } | |
209 | ||
210 | static int bpf_map_copy_value(struct bpf_map *map, void *key, void *value, | |
211 | __u64 flags) | |
212 | { | |
213 | void *ptr; | |
214 | int err; | |
215 | ||
9d03ebc7 | 216 | if (bpf_map_is_offloaded(map)) |
cb4d03ab | 217 | return bpf_map_offload_lookup_elem(map, key, value); |
15c14a3d | 218 | |
b6e5dae1 | 219 | bpf_disable_instrumentation(); |
15c14a3d BV |
220 | if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || |
221 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
222 | err = bpf_percpu_hash_copy(map, key, value); | |
223 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { | |
224 | err = bpf_percpu_array_copy(map, key, value); | |
225 | } else if (map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE) { | |
226 | err = bpf_percpu_cgroup_storage_copy(map, key, value); | |
227 | } else if (map->map_type == BPF_MAP_TYPE_STACK_TRACE) { | |
228 | err = bpf_stackmap_copy(map, key, value); | |
229 | } else if (IS_FD_ARRAY(map) || IS_FD_PROG_ARRAY(map)) { | |
230 | err = bpf_fd_array_map_lookup_elem(map, key, value); | |
231 | } else if (IS_FD_HASH(map)) { | |
232 | err = bpf_fd_htab_map_lookup_elem(map, key, value); | |
233 | } else if (map->map_type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY) { | |
234 | err = bpf_fd_reuseport_array_lookup_elem(map, key, value); | |
235 | } else if (map->map_type == BPF_MAP_TYPE_QUEUE || | |
9330986c JK |
236 | map->map_type == BPF_MAP_TYPE_STACK || |
237 | map->map_type == BPF_MAP_TYPE_BLOOM_FILTER) { | |
15c14a3d BV |
238 | err = map->ops->map_peek_elem(map, value); |
239 | } else if (map->map_type == BPF_MAP_TYPE_STRUCT_OPS) { | |
240 | /* struct_ops map requires directly updating "value" */ | |
241 | err = bpf_struct_ops_map_sys_lookup_elem(map, key, value); | |
242 | } else { | |
243 | rcu_read_lock(); | |
244 | if (map->ops->map_lookup_elem_sys_only) | |
245 | ptr = map->ops->map_lookup_elem_sys_only(map, key); | |
246 | else | |
247 | ptr = map->ops->map_lookup_elem(map, key); | |
248 | if (IS_ERR(ptr)) { | |
249 | err = PTR_ERR(ptr); | |
250 | } else if (!ptr) { | |
251 | err = -ENOENT; | |
252 | } else { | |
253 | err = 0; | |
254 | if (flags & BPF_F_LOCK) | |
255 | /* lock 'ptr' and copy everything but lock */ | |
256 | copy_map_value_locked(map, value, ptr, true); | |
257 | else | |
258 | copy_map_value(map, value, ptr); | |
68134668 AS |
259 | /* mask lock and timer, since value wasn't zero inited */ |
260 | check_and_init_map_value(map, value); | |
15c14a3d BV |
261 | } |
262 | rcu_read_unlock(); | |
263 | } | |
264 | ||
b6e5dae1 | 265 | bpf_enable_instrumentation(); |
15c14a3d BV |
266 | maybe_wait_bpf_programs(map); |
267 | ||
268 | return err; | |
269 | } | |
270 | ||
d5299b67 RG |
271 | /* Please, do not use this function outside from the map creation path |
272 | * (e.g. in map update path) without taking care of setting the active | |
273 | * memory cgroup (see at bpf_map_kmalloc_node() for example). | |
274 | */ | |
196e8ca7 | 275 | static void *__bpf_map_area_alloc(u64 size, int numa_node, bool mmapable) |
d407bd25 | 276 | { |
f01a7dbe MP |
277 | /* We really just want to fail instead of triggering OOM killer |
278 | * under memory pressure, therefore we set __GFP_NORETRY to kmalloc, | |
279 | * which is used for lower order allocation requests. | |
280 | * | |
281 | * It has been observed that higher order allocation requests done by | |
282 | * vmalloc with __GFP_NORETRY being set might fail due to not trying | |
283 | * to reclaim memory from the page cache, thus we set | |
284 | * __GFP_RETRY_MAYFAIL to avoid such situations. | |
d407bd25 | 285 | */ |
f01a7dbe | 286 | |
ee53cbfb | 287 | gfp_t gfp = bpf_memcg_flags(__GFP_NOWARN | __GFP_ZERO); |
041de93f CH |
288 | unsigned int flags = 0; |
289 | unsigned long align = 1; | |
d407bd25 DB |
290 | void *area; |
291 | ||
196e8ca7 DB |
292 | if (size >= SIZE_MAX) |
293 | return NULL; | |
294 | ||
fc970227 | 295 | /* kmalloc()'ed memory can't be mmap()'ed */ |
041de93f CH |
296 | if (mmapable) { |
297 | BUG_ON(!PAGE_ALIGNED(size)); | |
298 | align = SHMLBA; | |
299 | flags = VM_USERMAP; | |
300 | } else if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) { | |
301 | area = kmalloc_node(size, gfp | GFP_USER | __GFP_NORETRY, | |
f01a7dbe | 302 | numa_node); |
d407bd25 DB |
303 | if (area != NULL) |
304 | return area; | |
305 | } | |
041de93f CH |
306 | |
307 | return __vmalloc_node_range(size, align, VMALLOC_START, VMALLOC_END, | |
308 | gfp | GFP_KERNEL | __GFP_RETRY_MAYFAIL, PAGE_KERNEL, | |
309 | flags, numa_node, __builtin_return_address(0)); | |
d407bd25 DB |
310 | } |
311 | ||
196e8ca7 | 312 | void *bpf_map_area_alloc(u64 size, int numa_node) |
fc970227 AN |
313 | { |
314 | return __bpf_map_area_alloc(size, numa_node, false); | |
315 | } | |
316 | ||
196e8ca7 | 317 | void *bpf_map_area_mmapable_alloc(u64 size, int numa_node) |
fc970227 AN |
318 | { |
319 | return __bpf_map_area_alloc(size, numa_node, true); | |
320 | } | |
321 | ||
d407bd25 DB |
322 | void bpf_map_area_free(void *area) |
323 | { | |
324 | kvfree(area); | |
325 | } | |
326 | ||
be70bcd5 DB |
327 | static u32 bpf_map_flags_retain_permanent(u32 flags) |
328 | { | |
329 | /* Some map creation flags are not tied to the map object but | |
330 | * rather to the map fd instead, so they have no meaning upon | |
331 | * map object inspection since multiple file descriptors with | |
332 | * different (access) properties can exist here. Thus, given | |
333 | * this has zero meaning for the map itself, lets clear these | |
334 | * from here. | |
335 | */ | |
336 | return flags & ~(BPF_F_RDONLY | BPF_F_WRONLY); | |
337 | } | |
338 | ||
bd475643 JK |
339 | void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr) |
340 | { | |
341 | map->map_type = attr->map_type; | |
342 | map->key_size = attr->key_size; | |
343 | map->value_size = attr->value_size; | |
344 | map->max_entries = attr->max_entries; | |
be70bcd5 | 345 | map->map_flags = bpf_map_flags_retain_permanent(attr->map_flags); |
bd475643 | 346 | map->numa_node = bpf_map_attr_numa_node(attr); |
9330986c | 347 | map->map_extra = attr->map_extra; |
bd475643 JK |
348 | } |
349 | ||
f3f1c054 MKL |
350 | static int bpf_map_alloc_id(struct bpf_map *map) |
351 | { | |
352 | int id; | |
353 | ||
b76354cd | 354 | idr_preload(GFP_KERNEL); |
f3f1c054 MKL |
355 | spin_lock_bh(&map_idr_lock); |
356 | id = idr_alloc_cyclic(&map_idr, map, 1, INT_MAX, GFP_ATOMIC); | |
357 | if (id > 0) | |
358 | map->id = id; | |
359 | spin_unlock_bh(&map_idr_lock); | |
b76354cd | 360 | idr_preload_end(); |
f3f1c054 MKL |
361 | |
362 | if (WARN_ON_ONCE(!id)) | |
363 | return -ENOSPC; | |
364 | ||
365 | return id > 0 ? 0 : id; | |
366 | } | |
367 | ||
158e5e9e | 368 | void bpf_map_free_id(struct bpf_map *map) |
f3f1c054 | 369 | { |
930651a7 ED |
370 | unsigned long flags; |
371 | ||
a3884572 JK |
372 | /* Offloaded maps are removed from the IDR store when their device |
373 | * disappears - even if someone holds an fd to them they are unusable, | |
374 | * the memory is gone, all ops will fail; they are simply waiting for | |
375 | * refcnt to drop to be freed. | |
376 | */ | |
377 | if (!map->id) | |
378 | return; | |
379 | ||
158e5e9e | 380 | spin_lock_irqsave(&map_idr_lock, flags); |
bd5f5f4e | 381 | |
f3f1c054 | 382 | idr_remove(&map_idr, map->id); |
a3884572 | 383 | map->id = 0; |
bd5f5f4e | 384 | |
158e5e9e | 385 | spin_unlock_irqrestore(&map_idr_lock, flags); |
f3f1c054 MKL |
386 | } |
387 | ||
48edc1f7 RG |
388 | #ifdef CONFIG_MEMCG_KMEM |
389 | static void bpf_map_save_memcg(struct bpf_map *map) | |
390 | { | |
4201d9ab RG |
391 | /* Currently if a map is created by a process belonging to the root |
392 | * memory cgroup, get_obj_cgroup_from_current() will return NULL. | |
393 | * So we have to check map->objcg for being NULL each time it's | |
394 | * being used. | |
395 | */ | |
ee53cbfb YS |
396 | if (memcg_bpf_enabled()) |
397 | map->objcg = get_obj_cgroup_from_current(); | |
48edc1f7 RG |
398 | } |
399 | ||
400 | static void bpf_map_release_memcg(struct bpf_map *map) | |
401 | { | |
4201d9ab RG |
402 | if (map->objcg) |
403 | obj_cgroup_put(map->objcg); | |
404 | } | |
405 | ||
406 | static struct mem_cgroup *bpf_map_get_memcg(const struct bpf_map *map) | |
407 | { | |
408 | if (map->objcg) | |
409 | return get_mem_cgroup_from_objcg(map->objcg); | |
410 | ||
411 | return root_mem_cgroup; | |
48edc1f7 RG |
412 | } |
413 | ||
414 | void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, | |
415 | int node) | |
416 | { | |
4201d9ab | 417 | struct mem_cgroup *memcg, *old_memcg; |
48edc1f7 RG |
418 | void *ptr; |
419 | ||
4201d9ab RG |
420 | memcg = bpf_map_get_memcg(map); |
421 | old_memcg = set_active_memcg(memcg); | |
48edc1f7 RG |
422 | ptr = kmalloc_node(size, flags | __GFP_ACCOUNT, node); |
423 | set_active_memcg(old_memcg); | |
4201d9ab | 424 | mem_cgroup_put(memcg); |
48edc1f7 RG |
425 | |
426 | return ptr; | |
427 | } | |
428 | ||
429 | void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags) | |
430 | { | |
4201d9ab | 431 | struct mem_cgroup *memcg, *old_memcg; |
48edc1f7 RG |
432 | void *ptr; |
433 | ||
4201d9ab RG |
434 | memcg = bpf_map_get_memcg(map); |
435 | old_memcg = set_active_memcg(memcg); | |
48edc1f7 RG |
436 | ptr = kzalloc(size, flags | __GFP_ACCOUNT); |
437 | set_active_memcg(old_memcg); | |
4201d9ab | 438 | mem_cgroup_put(memcg); |
48edc1f7 RG |
439 | |
440 | return ptr; | |
441 | } | |
442 | ||
ddef81b5 YS |
443 | void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size, |
444 | gfp_t flags) | |
445 | { | |
446 | struct mem_cgroup *memcg, *old_memcg; | |
447 | void *ptr; | |
448 | ||
449 | memcg = bpf_map_get_memcg(map); | |
450 | old_memcg = set_active_memcg(memcg); | |
451 | ptr = kvcalloc(n, size, flags | __GFP_ACCOUNT); | |
452 | set_active_memcg(old_memcg); | |
453 | mem_cgroup_put(memcg); | |
454 | ||
455 | return ptr; | |
456 | } | |
457 | ||
48edc1f7 RG |
458 | void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, |
459 | size_t align, gfp_t flags) | |
460 | { | |
4201d9ab | 461 | struct mem_cgroup *memcg, *old_memcg; |
48edc1f7 RG |
462 | void __percpu *ptr; |
463 | ||
4201d9ab RG |
464 | memcg = bpf_map_get_memcg(map); |
465 | old_memcg = set_active_memcg(memcg); | |
48edc1f7 RG |
466 | ptr = __alloc_percpu_gfp(size, align, flags | __GFP_ACCOUNT); |
467 | set_active_memcg(old_memcg); | |
4201d9ab | 468 | mem_cgroup_put(memcg); |
48edc1f7 RG |
469 | |
470 | return ptr; | |
471 | } | |
472 | ||
473 | #else | |
474 | static void bpf_map_save_memcg(struct bpf_map *map) | |
475 | { | |
476 | } | |
477 | ||
478 | static void bpf_map_release_memcg(struct bpf_map *map) | |
479 | { | |
480 | } | |
481 | #endif | |
482 | ||
aa3496ac | 483 | static int btf_field_cmp(const void *a, const void *b) |
61df10c7 | 484 | { |
aa3496ac | 485 | const struct btf_field *f1 = a, *f2 = b; |
61df10c7 | 486 | |
aa3496ac | 487 | if (f1->offset < f2->offset) |
61df10c7 | 488 | return -1; |
aa3496ac | 489 | else if (f1->offset > f2->offset) |
61df10c7 KKD |
490 | return 1; |
491 | return 0; | |
492 | } | |
493 | ||
aa3496ac | 494 | struct btf_field *btf_record_find(const struct btf_record *rec, u32 offset, |
74843b57 | 495 | u32 field_mask) |
61df10c7 | 496 | { |
aa3496ac | 497 | struct btf_field *field; |
61df10c7 | 498 | |
74843b57 | 499 | if (IS_ERR_OR_NULL(rec) || !(rec->field_mask & field_mask)) |
aa3496ac KKD |
500 | return NULL; |
501 | field = bsearch(&offset, rec->fields, rec->cnt, sizeof(rec->fields[0]), btf_field_cmp); | |
74843b57 | 502 | if (!field || !(field->type & field_mask)) |
61df10c7 | 503 | return NULL; |
aa3496ac | 504 | return field; |
61df10c7 KKD |
505 | } |
506 | ||
aa3496ac | 507 | void btf_record_free(struct btf_record *rec) |
61df10c7 | 508 | { |
61df10c7 KKD |
509 | int i; |
510 | ||
aa3496ac | 511 | if (IS_ERR_OR_NULL(rec)) |
61df10c7 | 512 | return; |
aa3496ac KKD |
513 | for (i = 0; i < rec->cnt; i++) { |
514 | switch (rec->fields[i].type) { | |
515 | case BPF_KPTR_UNREF: | |
516 | case BPF_KPTR_REF: | |
517 | if (rec->fields[i].kptr.module) | |
518 | module_put(rec->fields[i].kptr.module); | |
519 | btf_put(rec->fields[i].kptr.btf); | |
520 | break; | |
f0c5941f | 521 | case BPF_LIST_HEAD: |
8ffa5cc1 | 522 | case BPF_LIST_NODE: |
9c395c1b DM |
523 | case BPF_RB_ROOT: |
524 | case BPF_RB_NODE: | |
525 | case BPF_SPIN_LOCK: | |
526 | case BPF_TIMER: | |
d54730b5 | 527 | case BPF_REFCOUNT: |
9c395c1b | 528 | /* Nothing to release */ |
f0c5941f | 529 | break; |
aa3496ac KKD |
530 | default: |
531 | WARN_ON_ONCE(1); | |
532 | continue; | |
533 | } | |
14a324f6 | 534 | } |
aa3496ac KKD |
535 | kfree(rec); |
536 | } | |
537 | ||
538 | void bpf_map_free_record(struct bpf_map *map) | |
539 | { | |
540 | btf_record_free(map->record); | |
541 | map->record = NULL; | |
61df10c7 KKD |
542 | } |
543 | ||
aa3496ac | 544 | struct btf_record *btf_record_dup(const struct btf_record *rec) |
61df10c7 | 545 | { |
aa3496ac KKD |
546 | const struct btf_field *fields; |
547 | struct btf_record *new_rec; | |
548 | int ret, size, i; | |
61df10c7 | 549 | |
aa3496ac KKD |
550 | if (IS_ERR_OR_NULL(rec)) |
551 | return NULL; | |
552 | size = offsetof(struct btf_record, fields[rec->cnt]); | |
553 | new_rec = kmemdup(rec, size, GFP_KERNEL | __GFP_NOWARN); | |
554 | if (!new_rec) | |
61df10c7 | 555 | return ERR_PTR(-ENOMEM); |
aa3496ac KKD |
556 | /* Do a deep copy of the btf_record */ |
557 | fields = rec->fields; | |
558 | new_rec->cnt = 0; | |
559 | for (i = 0; i < rec->cnt; i++) { | |
560 | switch (fields[i].type) { | |
561 | case BPF_KPTR_UNREF: | |
562 | case BPF_KPTR_REF: | |
563 | btf_get(fields[i].kptr.btf); | |
564 | if (fields[i].kptr.module && !try_module_get(fields[i].kptr.module)) { | |
565 | ret = -ENXIO; | |
566 | goto free; | |
14a324f6 | 567 | } |
aa3496ac | 568 | break; |
f0c5941f | 569 | case BPF_LIST_HEAD: |
8ffa5cc1 | 570 | case BPF_LIST_NODE: |
9c395c1b DM |
571 | case BPF_RB_ROOT: |
572 | case BPF_RB_NODE: | |
573 | case BPF_SPIN_LOCK: | |
574 | case BPF_TIMER: | |
d54730b5 | 575 | case BPF_REFCOUNT: |
9c395c1b | 576 | /* Nothing to acquire */ |
f0c5941f | 577 | break; |
aa3496ac KKD |
578 | default: |
579 | ret = -EFAULT; | |
580 | WARN_ON_ONCE(1); | |
581 | goto free; | |
14a324f6 | 582 | } |
aa3496ac | 583 | new_rec->cnt++; |
14a324f6 | 584 | } |
aa3496ac KKD |
585 | return new_rec; |
586 | free: | |
587 | btf_record_free(new_rec); | |
588 | return ERR_PTR(ret); | |
61df10c7 KKD |
589 | } |
590 | ||
aa3496ac | 591 | bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b) |
61df10c7 | 592 | { |
aa3496ac | 593 | bool a_has_fields = !IS_ERR_OR_NULL(rec_a), b_has_fields = !IS_ERR_OR_NULL(rec_b); |
61df10c7 KKD |
594 | int size; |
595 | ||
aa3496ac | 596 | if (!a_has_fields && !b_has_fields) |
61df10c7 | 597 | return true; |
aa3496ac | 598 | if (a_has_fields != b_has_fields) |
61df10c7 | 599 | return false; |
aa3496ac | 600 | if (rec_a->cnt != rec_b->cnt) |
61df10c7 | 601 | return false; |
aa3496ac | 602 | size = offsetof(struct btf_record, fields[rec_a->cnt]); |
c22dfdd2 KKD |
603 | /* btf_parse_fields uses kzalloc to allocate a btf_record, so unused |
604 | * members are zeroed out. So memcmp is safe to do without worrying | |
605 | * about padding/unused fields. | |
606 | * | |
607 | * While spin_lock, timer, and kptr have no relation to map BTF, | |
608 | * list_head metadata is specific to map BTF, the btf and value_rec | |
609 | * members in particular. btf is the map BTF, while value_rec points to | |
610 | * btf_record in that map BTF. | |
611 | * | |
612 | * So while by default, we don't rely on the map BTF (which the records | |
613 | * were parsed from) matching for both records, which is not backwards | |
614 | * compatible, in case list_head is part of it, we implicitly rely on | |
615 | * that by way of depending on memcmp succeeding for it. | |
616 | */ | |
aa3496ac | 617 | return !memcmp(rec_a, rec_b, size); |
61df10c7 KKD |
618 | } |
619 | ||
db559117 KKD |
620 | void bpf_obj_free_timer(const struct btf_record *rec, void *obj) |
621 | { | |
622 | if (WARN_ON_ONCE(!btf_record_has_field(rec, BPF_TIMER))) | |
623 | return; | |
624 | bpf_timer_cancel_and_free(obj + rec->timer_off); | |
625 | } | |
626 | ||
9e36a204 DM |
627 | extern void __bpf_obj_drop_impl(void *p, const struct btf_record *rec); |
628 | ||
aa3496ac | 629 | void bpf_obj_free_fields(const struct btf_record *rec, void *obj) |
14a324f6 | 630 | { |
aa3496ac | 631 | const struct btf_field *fields; |
14a324f6 KKD |
632 | int i; |
633 | ||
aa3496ac KKD |
634 | if (IS_ERR_OR_NULL(rec)) |
635 | return; | |
636 | fields = rec->fields; | |
637 | for (i = 0; i < rec->cnt; i++) { | |
c8e18754 | 638 | struct btf_struct_meta *pointee_struct_meta; |
aa3496ac KKD |
639 | const struct btf_field *field = &fields[i]; |
640 | void *field_ptr = obj + field->offset; | |
c8e18754 | 641 | void *xchgd_field; |
aa3496ac KKD |
642 | |
643 | switch (fields[i].type) { | |
db559117 KKD |
644 | case BPF_SPIN_LOCK: |
645 | break; | |
646 | case BPF_TIMER: | |
647 | bpf_timer_cancel_and_free(field_ptr); | |
648 | break; | |
aa3496ac KKD |
649 | case BPF_KPTR_UNREF: |
650 | WRITE_ONCE(*(u64 *)field_ptr, 0); | |
651 | break; | |
652 | case BPF_KPTR_REF: | |
c8e18754 | 653 | xchgd_field = (void *)xchg((unsigned long *)field_ptr, 0); |
1431d0b5 DV |
654 | if (!xchgd_field) |
655 | break; | |
656 | ||
c8e18754 DM |
657 | if (!btf_is_kernel(field->kptr.btf)) { |
658 | pointee_struct_meta = btf_find_struct_meta(field->kptr.btf, | |
659 | field->kptr.btf_id); | |
9e36a204 DM |
660 | migrate_disable(); |
661 | __bpf_obj_drop_impl(xchgd_field, pointee_struct_meta ? | |
662 | pointee_struct_meta->record : | |
663 | NULL); | |
664 | migrate_enable(); | |
c8e18754 DM |
665 | } else { |
666 | field->kptr.dtor(xchgd_field); | |
667 | } | |
aa3496ac | 668 | break; |
f0c5941f KKD |
669 | case BPF_LIST_HEAD: |
670 | if (WARN_ON_ONCE(rec->spin_lock_off < 0)) | |
671 | continue; | |
672 | bpf_list_head_free(field, field_ptr, obj + rec->spin_lock_off); | |
673 | break; | |
9c395c1b DM |
674 | case BPF_RB_ROOT: |
675 | if (WARN_ON_ONCE(rec->spin_lock_off < 0)) | |
676 | continue; | |
677 | bpf_rb_root_free(field, field_ptr, obj + rec->spin_lock_off); | |
678 | break; | |
8ffa5cc1 | 679 | case BPF_LIST_NODE: |
9c395c1b | 680 | case BPF_RB_NODE: |
d54730b5 | 681 | case BPF_REFCOUNT: |
8ffa5cc1 | 682 | break; |
aa3496ac KKD |
683 | default: |
684 | WARN_ON_ONCE(1); | |
14a324f6 KKD |
685 | continue; |
686 | } | |
14a324f6 KKD |
687 | } |
688 | } | |
689 | ||
99c55f7d AS |
690 | /* called from workqueue */ |
691 | static void bpf_map_free_deferred(struct work_struct *work) | |
692 | { | |
693 | struct bpf_map *map = container_of(work, struct bpf_map, work); | |
d7f5ef65 | 694 | struct btf_record *rec = map->record; |
99c55f7d | 695 | |
afdb09c7 | 696 | security_bpf_map_free(map); |
48edc1f7 | 697 | bpf_map_release_memcg(map); |
d7f5ef65 | 698 | /* implementation dependent freeing */ |
99c55f7d | 699 | map->ops->map_free(map); |
cd2a8079 | 700 | /* Delay freeing of btf_record for maps, as map_free |
d7f5ef65 KKD |
701 | * callback usually needs access to them. It is better to do it here |
702 | * than require each callback to do the free itself manually. | |
703 | * | |
704 | * Note that the btf_record stashed in map->inner_map_meta->record was | |
705 | * already freed using the map_free callback for map in map case which | |
706 | * eventually calls bpf_map_free_meta, since inner_map_meta is only a | |
707 | * template bpf_map struct used during verification. | |
708 | */ | |
d7f5ef65 | 709 | btf_record_free(rec); |
99c55f7d AS |
710 | } |
711 | ||
c9da161c DB |
712 | static void bpf_map_put_uref(struct bpf_map *map) |
713 | { | |
1e0bd5a0 | 714 | if (atomic64_dec_and_test(&map->usercnt)) { |
ba6b8de4 JF |
715 | if (map->ops->map_release_uref) |
716 | map->ops->map_release_uref(map); | |
c9da161c DB |
717 | } |
718 | } | |
719 | ||
99c55f7d | 720 | /* decrement map refcnt and schedule it for freeing via workqueue |
158e5e9e | 721 | * (underlying map implementation ops->map_free() might sleep) |
99c55f7d | 722 | */ |
158e5e9e | 723 | void bpf_map_put(struct bpf_map *map) |
99c55f7d | 724 | { |
1e0bd5a0 | 725 | if (atomic64_dec_and_test(&map->refcnt)) { |
34ad5580 | 726 | /* bpf_map_free_id() must be called first */ |
158e5e9e | 727 | bpf_map_free_id(map); |
78958fca | 728 | btf_put(map->btf); |
99c55f7d | 729 | INIT_WORK(&map->work, bpf_map_free_deferred); |
8d5a8011 AS |
730 | /* Avoid spawning kworkers, since they all might contend |
731 | * for the same mutex like slab_mutex. | |
732 | */ | |
733 | queue_work(system_unbound_wq, &map->work); | |
99c55f7d AS |
734 | } |
735 | } | |
630a4d38 | 736 | EXPORT_SYMBOL_GPL(bpf_map_put); |
bd5f5f4e | 737 | |
c9da161c | 738 | void bpf_map_put_with_uref(struct bpf_map *map) |
99c55f7d | 739 | { |
c9da161c | 740 | bpf_map_put_uref(map); |
99c55f7d | 741 | bpf_map_put(map); |
c9da161c DB |
742 | } |
743 | ||
744 | static int bpf_map_release(struct inode *inode, struct file *filp) | |
745 | { | |
61d1b6a4 DB |
746 | struct bpf_map *map = filp->private_data; |
747 | ||
748 | if (map->ops->map_release) | |
749 | map->ops->map_release(map, filp); | |
750 | ||
751 | bpf_map_put_with_uref(map); | |
99c55f7d AS |
752 | return 0; |
753 | } | |
754 | ||
87df15de DB |
755 | static fmode_t map_get_sys_perms(struct bpf_map *map, struct fd f) |
756 | { | |
757 | fmode_t mode = f.file->f_mode; | |
758 | ||
759 | /* Our file permissions may have been overridden by global | |
760 | * map permissions facing syscall side. | |
761 | */ | |
762 | if (READ_ONCE(map->frozen)) | |
763 | mode &= ~FMODE_CAN_WRITE; | |
764 | return mode; | |
765 | } | |
766 | ||
f99bf205 | 767 | #ifdef CONFIG_PROC_FS |
90a5527d YS |
768 | /* Show the memory usage of a bpf map */ |
769 | static u64 bpf_map_memory_usage(const struct bpf_map *map) | |
80ee81e0 | 770 | { |
6b4a6ea2 | 771 | return map->ops->map_mem_usage(map); |
80ee81e0 RG |
772 | } |
773 | ||
f99bf205 DB |
774 | static void bpf_map_show_fdinfo(struct seq_file *m, struct file *filp) |
775 | { | |
f45d5b6c | 776 | struct bpf_map *map = filp->private_data; |
2beee5f5 | 777 | u32 type = 0, jited = 0; |
21116b70 | 778 | |
f45d5b6c THJ |
779 | if (map_type_contains_progs(map)) { |
780 | spin_lock(&map->owner.lock); | |
781 | type = map->owner.type; | |
782 | jited = map->owner.jited; | |
783 | spin_unlock(&map->owner.lock); | |
21116b70 | 784 | } |
f99bf205 DB |
785 | |
786 | seq_printf(m, | |
787 | "map_type:\t%u\n" | |
788 | "key_size:\t%u\n" | |
789 | "value_size:\t%u\n" | |
322cea2f | 790 | "max_entries:\t%u\n" |
21116b70 | 791 | "map_flags:\t%#x\n" |
9330986c | 792 | "map_extra:\t%#llx\n" |
90a5527d | 793 | "memlock:\t%llu\n" |
87df15de DB |
794 | "map_id:\t%u\n" |
795 | "frozen:\t%u\n", | |
f99bf205 DB |
796 | map->map_type, |
797 | map->key_size, | |
798 | map->value_size, | |
322cea2f | 799 | map->max_entries, |
21116b70 | 800 | map->map_flags, |
9330986c | 801 | (unsigned long long)map->map_extra, |
90a5527d | 802 | bpf_map_memory_usage(map), |
87df15de DB |
803 | map->id, |
804 | READ_ONCE(map->frozen)); | |
2beee5f5 DB |
805 | if (type) { |
806 | seq_printf(m, "owner_prog_type:\t%u\n", type); | |
807 | seq_printf(m, "owner_jited:\t%u\n", jited); | |
9780c0ab | 808 | } |
f99bf205 DB |
809 | } |
810 | #endif | |
811 | ||
6e71b04a CF |
812 | static ssize_t bpf_dummy_read(struct file *filp, char __user *buf, size_t siz, |
813 | loff_t *ppos) | |
814 | { | |
815 | /* We need this handler such that alloc_file() enables | |
816 | * f_mode with FMODE_CAN_READ. | |
817 | */ | |
818 | return -EINVAL; | |
819 | } | |
820 | ||
821 | static ssize_t bpf_dummy_write(struct file *filp, const char __user *buf, | |
822 | size_t siz, loff_t *ppos) | |
823 | { | |
824 | /* We need this handler such that alloc_file() enables | |
825 | * f_mode with FMODE_CAN_WRITE. | |
826 | */ | |
827 | return -EINVAL; | |
828 | } | |
829 | ||
fc970227 AN |
830 | /* called for any extra memory-mapped regions (except initial) */ |
831 | static void bpf_map_mmap_open(struct vm_area_struct *vma) | |
832 | { | |
833 | struct bpf_map *map = vma->vm_file->private_data; | |
834 | ||
353050be DB |
835 | if (vma->vm_flags & VM_MAYWRITE) |
836 | bpf_map_write_active_inc(map); | |
fc970227 AN |
837 | } |
838 | ||
839 | /* called for all unmapped memory region (including initial) */ | |
840 | static void bpf_map_mmap_close(struct vm_area_struct *vma) | |
841 | { | |
842 | struct bpf_map *map = vma->vm_file->private_data; | |
843 | ||
353050be DB |
844 | if (vma->vm_flags & VM_MAYWRITE) |
845 | bpf_map_write_active_dec(map); | |
fc970227 AN |
846 | } |
847 | ||
848 | static const struct vm_operations_struct bpf_map_default_vmops = { | |
849 | .open = bpf_map_mmap_open, | |
850 | .close = bpf_map_mmap_close, | |
851 | }; | |
852 | ||
853 | static int bpf_map_mmap(struct file *filp, struct vm_area_struct *vma) | |
854 | { | |
855 | struct bpf_map *map = filp->private_data; | |
856 | int err; | |
857 | ||
db559117 | 858 | if (!map->ops->map_mmap || !IS_ERR_OR_NULL(map->record)) |
fc970227 AN |
859 | return -ENOTSUPP; |
860 | ||
861 | if (!(vma->vm_flags & VM_SHARED)) | |
862 | return -EINVAL; | |
863 | ||
864 | mutex_lock(&map->freeze_mutex); | |
865 | ||
dfeb376d AN |
866 | if (vma->vm_flags & VM_WRITE) { |
867 | if (map->frozen) { | |
868 | err = -EPERM; | |
869 | goto out; | |
870 | } | |
871 | /* map is meant to be read-only, so do not allow mapping as | |
872 | * writable, because it's possible to leak a writable page | |
873 | * reference and allows user-space to still modify it after | |
874 | * freezing, while verifier will assume contents do not change | |
875 | */ | |
876 | if (map->map_flags & BPF_F_RDONLY_PROG) { | |
877 | err = -EACCES; | |
878 | goto out; | |
879 | } | |
fc970227 AN |
880 | } |
881 | ||
882 | /* set default open/close callbacks */ | |
883 | vma->vm_ops = &bpf_map_default_vmops; | |
884 | vma->vm_private_data = map; | |
1c71222e | 885 | vm_flags_clear(vma, VM_MAYEXEC); |
1f6cb19b AN |
886 | if (!(vma->vm_flags & VM_WRITE)) |
887 | /* disallow re-mapping with PROT_WRITE */ | |
1c71222e | 888 | vm_flags_clear(vma, VM_MAYWRITE); |
fc970227 AN |
889 | |
890 | err = map->ops->map_mmap(map, vma); | |
891 | if (err) | |
892 | goto out; | |
893 | ||
1f6cb19b | 894 | if (vma->vm_flags & VM_MAYWRITE) |
353050be | 895 | bpf_map_write_active_inc(map); |
fc970227 AN |
896 | out: |
897 | mutex_unlock(&map->freeze_mutex); | |
898 | return err; | |
899 | } | |
900 | ||
457f4436 AN |
901 | static __poll_t bpf_map_poll(struct file *filp, struct poll_table_struct *pts) |
902 | { | |
903 | struct bpf_map *map = filp->private_data; | |
904 | ||
905 | if (map->ops->map_poll) | |
906 | return map->ops->map_poll(map, filp, pts); | |
907 | ||
908 | return EPOLLERR; | |
909 | } | |
910 | ||
f66e448c | 911 | const struct file_operations bpf_map_fops = { |
f99bf205 DB |
912 | #ifdef CONFIG_PROC_FS |
913 | .show_fdinfo = bpf_map_show_fdinfo, | |
914 | #endif | |
915 | .release = bpf_map_release, | |
6e71b04a CF |
916 | .read = bpf_dummy_read, |
917 | .write = bpf_dummy_write, | |
fc970227 | 918 | .mmap = bpf_map_mmap, |
457f4436 | 919 | .poll = bpf_map_poll, |
99c55f7d AS |
920 | }; |
921 | ||
6e71b04a | 922 | int bpf_map_new_fd(struct bpf_map *map, int flags) |
aa79781b | 923 | { |
afdb09c7 CF |
924 | int ret; |
925 | ||
926 | ret = security_bpf_map(map, OPEN_FMODE(flags)); | |
927 | if (ret < 0) | |
928 | return ret; | |
929 | ||
aa79781b | 930 | return anon_inode_getfd("bpf-map", &bpf_map_fops, map, |
6e71b04a CF |
931 | flags | O_CLOEXEC); |
932 | } | |
933 | ||
934 | int bpf_get_file_flag(int flags) | |
935 | { | |
936 | if ((flags & BPF_F_RDONLY) && (flags & BPF_F_WRONLY)) | |
937 | return -EINVAL; | |
938 | if (flags & BPF_F_RDONLY) | |
939 | return O_RDONLY; | |
940 | if (flags & BPF_F_WRONLY) | |
941 | return O_WRONLY; | |
942 | return O_RDWR; | |
aa79781b DB |
943 | } |
944 | ||
99c55f7d AS |
945 | /* helper macro to check that unused fields 'union bpf_attr' are zero */ |
946 | #define CHECK_ATTR(CMD) \ | |
947 | memchr_inv((void *) &attr->CMD##_LAST_FIELD + \ | |
948 | sizeof(attr->CMD##_LAST_FIELD), 0, \ | |
949 | sizeof(*attr) - \ | |
950 | offsetof(union bpf_attr, CMD##_LAST_FIELD) - \ | |
951 | sizeof(attr->CMD##_LAST_FIELD)) != NULL | |
952 | ||
8e7ae251 MKL |
953 | /* dst and src must have at least "size" number of bytes. |
954 | * Return strlen on success and < 0 on error. | |
cb4d2b3f | 955 | */ |
8e7ae251 | 956 | int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size) |
cb4d2b3f | 957 | { |
8e7ae251 MKL |
958 | const char *end = src + size; |
959 | const char *orig_src = src; | |
cb4d2b3f | 960 | |
8e7ae251 | 961 | memset(dst, 0, size); |
3e0ddc4f | 962 | /* Copy all isalnum(), '_' and '.' chars. */ |
cb4d2b3f | 963 | while (src < end && *src) { |
3e0ddc4f DB |
964 | if (!isalnum(*src) && |
965 | *src != '_' && *src != '.') | |
cb4d2b3f MKL |
966 | return -EINVAL; |
967 | *dst++ = *src++; | |
968 | } | |
969 | ||
8e7ae251 | 970 | /* No '\0' found in "size" number of bytes */ |
cb4d2b3f MKL |
971 | if (src == end) |
972 | return -EINVAL; | |
973 | ||
8e7ae251 | 974 | return src - orig_src; |
cb4d2b3f MKL |
975 | } |
976 | ||
e8d2bec0 | 977 | int map_check_no_btf(const struct bpf_map *map, |
1b2b234b | 978 | const struct btf *btf, |
e8d2bec0 DB |
979 | const struct btf_type *key_type, |
980 | const struct btf_type *value_type) | |
981 | { | |
982 | return -ENOTSUPP; | |
983 | } | |
984 | ||
d83525ca | 985 | static int map_check_btf(struct bpf_map *map, const struct btf *btf, |
e8d2bec0 DB |
986 | u32 btf_key_id, u32 btf_value_id) |
987 | { | |
988 | const struct btf_type *key_type, *value_type; | |
989 | u32 key_size, value_size; | |
990 | int ret = 0; | |
991 | ||
2824ecb7 DB |
992 | /* Some maps allow key to be unspecified. */ |
993 | if (btf_key_id) { | |
994 | key_type = btf_type_id_size(btf, &btf_key_id, &key_size); | |
995 | if (!key_type || key_size != map->key_size) | |
996 | return -EINVAL; | |
997 | } else { | |
998 | key_type = btf_type_by_id(btf, 0); | |
999 | if (!map->ops->map_check_btf) | |
1000 | return -EINVAL; | |
1001 | } | |
e8d2bec0 DB |
1002 | |
1003 | value_type = btf_type_id_size(btf, &btf_value_id, &value_size); | |
1004 | if (!value_type || value_size != map->value_size) | |
1005 | return -EINVAL; | |
1006 | ||
f0c5941f | 1007 | map->record = btf_parse_fields(btf, value_type, |
9c395c1b | 1008 | BPF_SPIN_LOCK | BPF_TIMER | BPF_KPTR | BPF_LIST_HEAD | |
d54730b5 | 1009 | BPF_RB_ROOT | BPF_REFCOUNT, |
db559117 | 1010 | map->value_size); |
aa3496ac KKD |
1011 | if (!IS_ERR_OR_NULL(map->record)) { |
1012 | int i; | |
1013 | ||
61df10c7 KKD |
1014 | if (!bpf_capable()) { |
1015 | ret = -EPERM; | |
1016 | goto free_map_tab; | |
1017 | } | |
1018 | if (map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) { | |
1019 | ret = -EACCES; | |
1020 | goto free_map_tab; | |
1021 | } | |
aa3496ac KKD |
1022 | for (i = 0; i < sizeof(map->record->field_mask) * 8; i++) { |
1023 | switch (map->record->field_mask & (1 << i)) { | |
1024 | case 0: | |
1025 | continue; | |
db559117 KKD |
1026 | case BPF_SPIN_LOCK: |
1027 | if (map->map_type != BPF_MAP_TYPE_HASH && | |
1028 | map->map_type != BPF_MAP_TYPE_ARRAY && | |
1029 | map->map_type != BPF_MAP_TYPE_CGROUP_STORAGE && | |
1030 | map->map_type != BPF_MAP_TYPE_SK_STORAGE && | |
1031 | map->map_type != BPF_MAP_TYPE_INODE_STORAGE && | |
1032 | map->map_type != BPF_MAP_TYPE_TASK_STORAGE && | |
1033 | map->map_type != BPF_MAP_TYPE_CGRP_STORAGE) { | |
1034 | ret = -EOPNOTSUPP; | |
1035 | goto free_map_tab; | |
1036 | } | |
1037 | break; | |
1038 | case BPF_TIMER: | |
1039 | if (map->map_type != BPF_MAP_TYPE_HASH && | |
1040 | map->map_type != BPF_MAP_TYPE_LRU_HASH && | |
1041 | map->map_type != BPF_MAP_TYPE_ARRAY) { | |
c237bfa5 | 1042 | ret = -EOPNOTSUPP; |
db559117 KKD |
1043 | goto free_map_tab; |
1044 | } | |
1045 | break; | |
aa3496ac KKD |
1046 | case BPF_KPTR_UNREF: |
1047 | case BPF_KPTR_REF: | |
d54730b5 | 1048 | case BPF_REFCOUNT: |
aa3496ac | 1049 | if (map->map_type != BPF_MAP_TYPE_HASH && |
65334e64 | 1050 | map->map_type != BPF_MAP_TYPE_PERCPU_HASH && |
aa3496ac | 1051 | map->map_type != BPF_MAP_TYPE_LRU_HASH && |
65334e64 | 1052 | map->map_type != BPF_MAP_TYPE_LRU_PERCPU_HASH && |
aa3496ac | 1053 | map->map_type != BPF_MAP_TYPE_ARRAY && |
9db44fdd KKD |
1054 | map->map_type != BPF_MAP_TYPE_PERCPU_ARRAY && |
1055 | map->map_type != BPF_MAP_TYPE_SK_STORAGE && | |
1056 | map->map_type != BPF_MAP_TYPE_INODE_STORAGE && | |
1057 | map->map_type != BPF_MAP_TYPE_TASK_STORAGE && | |
1058 | map->map_type != BPF_MAP_TYPE_CGRP_STORAGE) { | |
f0c5941f KKD |
1059 | ret = -EOPNOTSUPP; |
1060 | goto free_map_tab; | |
1061 | } | |
1062 | break; | |
1063 | case BPF_LIST_HEAD: | |
9c395c1b | 1064 | case BPF_RB_ROOT: |
f0c5941f KKD |
1065 | if (map->map_type != BPF_MAP_TYPE_HASH && |
1066 | map->map_type != BPF_MAP_TYPE_LRU_HASH && | |
1067 | map->map_type != BPF_MAP_TYPE_ARRAY) { | |
aa3496ac KKD |
1068 | ret = -EOPNOTSUPP; |
1069 | goto free_map_tab; | |
1070 | } | |
1071 | break; | |
1072 | default: | |
1073 | /* Fail if map_type checks are missing for a field type */ | |
1074 | ret = -EOPNOTSUPP; | |
1075 | goto free_map_tab; | |
1076 | } | |
61df10c7 KKD |
1077 | } |
1078 | } | |
1079 | ||
865ce09a KKD |
1080 | ret = btf_check_and_fixup_fields(btf, map->record); |
1081 | if (ret < 0) | |
1082 | goto free_map_tab; | |
1083 | ||
61df10c7 | 1084 | if (map->ops->map_check_btf) { |
1b2b234b | 1085 | ret = map->ops->map_check_btf(map, btf, key_type, value_type); |
61df10c7 KKD |
1086 | if (ret < 0) |
1087 | goto free_map_tab; | |
1088 | } | |
e8d2bec0 | 1089 | |
61df10c7 KKD |
1090 | return ret; |
1091 | free_map_tab: | |
aa3496ac | 1092 | bpf_map_free_record(map); |
e8d2bec0 DB |
1093 | return ret; |
1094 | } | |
1095 | ||
9330986c | 1096 | #define BPF_MAP_CREATE_LAST_FIELD map_extra |
99c55f7d AS |
1097 | /* called via syscall */ |
1098 | static int map_create(union bpf_attr *attr) | |
1099 | { | |
22db4122 | 1100 | const struct bpf_map_ops *ops; |
96eabe7a | 1101 | int numa_node = bpf_map_attr_numa_node(attr); |
22db4122 | 1102 | u32 map_type = attr->map_type; |
99c55f7d | 1103 | struct bpf_map *map; |
6e71b04a | 1104 | int f_flags; |
99c55f7d AS |
1105 | int err; |
1106 | ||
1107 | err = CHECK_ATTR(BPF_MAP_CREATE); | |
1108 | if (err) | |
1109 | return -EINVAL; | |
1110 | ||
85d33df3 MKL |
1111 | if (attr->btf_vmlinux_value_type_id) { |
1112 | if (attr->map_type != BPF_MAP_TYPE_STRUCT_OPS || | |
1113 | attr->btf_key_type_id || attr->btf_value_type_id) | |
1114 | return -EINVAL; | |
1115 | } else if (attr->btf_key_type_id && !attr->btf_value_type_id) { | |
1116 | return -EINVAL; | |
1117 | } | |
1118 | ||
9330986c JK |
1119 | if (attr->map_type != BPF_MAP_TYPE_BLOOM_FILTER && |
1120 | attr->map_extra != 0) | |
1121 | return -EINVAL; | |
1122 | ||
6e71b04a CF |
1123 | f_flags = bpf_get_file_flag(attr->map_flags); |
1124 | if (f_flags < 0) | |
1125 | return f_flags; | |
1126 | ||
96eabe7a | 1127 | if (numa_node != NUMA_NO_NODE && |
96e5ae4e ED |
1128 | ((unsigned int)numa_node >= nr_node_ids || |
1129 | !node_online(numa_node))) | |
96eabe7a MKL |
1130 | return -EINVAL; |
1131 | ||
99c55f7d | 1132 | /* find map type and init map: hashtable vs rbtree vs bloom vs ... */ |
22db4122 AN |
1133 | map_type = attr->map_type; |
1134 | if (map_type >= ARRAY_SIZE(bpf_map_types)) | |
1135 | return -EINVAL; | |
1136 | map_type = array_index_nospec(map_type, ARRAY_SIZE(bpf_map_types)); | |
1137 | ops = bpf_map_types[map_type]; | |
1138 | if (!ops) | |
1139 | return -EINVAL; | |
1140 | ||
1141 | if (ops->map_alloc_check) { | |
1142 | err = ops->map_alloc_check(attr); | |
1143 | if (err) | |
1144 | return err; | |
1145 | } | |
1146 | if (attr->map_ifindex) | |
1147 | ops = &bpf_map_offload_ops; | |
1148 | if (!ops->map_mem_usage) | |
1149 | return -EINVAL; | |
1150 | ||
1d28635a AN |
1151 | /* Intent here is for unprivileged_bpf_disabled to block BPF map |
1152 | * creation for unprivileged users; other actions depend | |
1153 | * on fd availability and access to bpffs, so are dependent on | |
1154 | * object creation success. Even with unprivileged BPF disabled, | |
1155 | * capability checks are still carried out. | |
1156 | */ | |
1157 | if (sysctl_unprivileged_bpf_disabled && !bpf_capable()) | |
1158 | return -EPERM; | |
1159 | ||
6c3eba1c AN |
1160 | /* check privileged map type permissions */ |
1161 | switch (map_type) { | |
1162 | case BPF_MAP_TYPE_ARRAY: | |
1163 | case BPF_MAP_TYPE_PERCPU_ARRAY: | |
1164 | case BPF_MAP_TYPE_PROG_ARRAY: | |
1165 | case BPF_MAP_TYPE_PERF_EVENT_ARRAY: | |
1166 | case BPF_MAP_TYPE_CGROUP_ARRAY: | |
1167 | case BPF_MAP_TYPE_ARRAY_OF_MAPS: | |
1168 | case BPF_MAP_TYPE_HASH: | |
1169 | case BPF_MAP_TYPE_PERCPU_HASH: | |
1170 | case BPF_MAP_TYPE_HASH_OF_MAPS: | |
1171 | case BPF_MAP_TYPE_RINGBUF: | |
1172 | case BPF_MAP_TYPE_USER_RINGBUF: | |
1173 | case BPF_MAP_TYPE_CGROUP_STORAGE: | |
1174 | case BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE: | |
1175 | /* unprivileged */ | |
1176 | break; | |
1177 | case BPF_MAP_TYPE_SK_STORAGE: | |
1178 | case BPF_MAP_TYPE_INODE_STORAGE: | |
1179 | case BPF_MAP_TYPE_TASK_STORAGE: | |
1180 | case BPF_MAP_TYPE_CGRP_STORAGE: | |
1181 | case BPF_MAP_TYPE_BLOOM_FILTER: | |
1182 | case BPF_MAP_TYPE_LPM_TRIE: | |
1183 | case BPF_MAP_TYPE_REUSEPORT_SOCKARRAY: | |
1184 | case BPF_MAP_TYPE_STACK_TRACE: | |
1185 | case BPF_MAP_TYPE_QUEUE: | |
1186 | case BPF_MAP_TYPE_STACK: | |
1187 | case BPF_MAP_TYPE_LRU_HASH: | |
1188 | case BPF_MAP_TYPE_LRU_PERCPU_HASH: | |
1189 | case BPF_MAP_TYPE_STRUCT_OPS: | |
1190 | case BPF_MAP_TYPE_CPUMAP: | |
1191 | if (!bpf_capable()) | |
1192 | return -EPERM; | |
1193 | break; | |
1194 | case BPF_MAP_TYPE_SOCKMAP: | |
1195 | case BPF_MAP_TYPE_SOCKHASH: | |
1196 | case BPF_MAP_TYPE_DEVMAP: | |
1197 | case BPF_MAP_TYPE_DEVMAP_HASH: | |
1198 | case BPF_MAP_TYPE_XSKMAP: | |
1199 | if (!capable(CAP_NET_ADMIN)) | |
1200 | return -EPERM; | |
1201 | break; | |
1202 | default: | |
1203 | WARN(1, "unsupported map type %d", map_type); | |
1204 | return -EPERM; | |
1205 | } | |
1206 | ||
22db4122 | 1207 | map = ops->map_alloc(attr); |
99c55f7d AS |
1208 | if (IS_ERR(map)) |
1209 | return PTR_ERR(map); | |
22db4122 AN |
1210 | map->ops = ops; |
1211 | map->map_type = map_type; | |
99c55f7d | 1212 | |
8e7ae251 MKL |
1213 | err = bpf_obj_name_cpy(map->name, attr->map_name, |
1214 | sizeof(attr->map_name)); | |
1215 | if (err < 0) | |
b936ca64 | 1216 | goto free_map; |
ad5b177b | 1217 | |
1e0bd5a0 AN |
1218 | atomic64_set(&map->refcnt, 1); |
1219 | atomic64_set(&map->usercnt, 1); | |
fc970227 | 1220 | mutex_init(&map->freeze_mutex); |
f45d5b6c | 1221 | spin_lock_init(&map->owner.lock); |
99c55f7d | 1222 | |
85d33df3 MKL |
1223 | if (attr->btf_key_type_id || attr->btf_value_type_id || |
1224 | /* Even the map's value is a kernel's struct, | |
1225 | * the bpf_prog.o must have BTF to begin with | |
1226 | * to figure out the corresponding kernel's | |
1227 | * counter part. Thus, attr->btf_fd has | |
1228 | * to be valid also. | |
1229 | */ | |
1230 | attr->btf_vmlinux_value_type_id) { | |
a26ca7c9 MKL |
1231 | struct btf *btf; |
1232 | ||
a26ca7c9 MKL |
1233 | btf = btf_get_by_fd(attr->btf_fd); |
1234 | if (IS_ERR(btf)) { | |
1235 | err = PTR_ERR(btf); | |
b936ca64 | 1236 | goto free_map; |
a26ca7c9 | 1237 | } |
350a5c4d AS |
1238 | if (btf_is_kernel(btf)) { |
1239 | btf_put(btf); | |
1240 | err = -EACCES; | |
1241 | goto free_map; | |
1242 | } | |
85d33df3 | 1243 | map->btf = btf; |
a26ca7c9 | 1244 | |
85d33df3 MKL |
1245 | if (attr->btf_value_type_id) { |
1246 | err = map_check_btf(map, btf, attr->btf_key_type_id, | |
1247 | attr->btf_value_type_id); | |
1248 | if (err) | |
1249 | goto free_map; | |
a26ca7c9 MKL |
1250 | } |
1251 | ||
9b2cf328 MKL |
1252 | map->btf_key_type_id = attr->btf_key_type_id; |
1253 | map->btf_value_type_id = attr->btf_value_type_id; | |
85d33df3 MKL |
1254 | map->btf_vmlinux_value_type_id = |
1255 | attr->btf_vmlinux_value_type_id; | |
a26ca7c9 MKL |
1256 | } |
1257 | ||
4d7d7f69 KKD |
1258 | err = security_bpf_map_alloc(map); |
1259 | if (err) | |
cd2a8079 | 1260 | goto free_map; |
4d7d7f69 | 1261 | |
f3f1c054 MKL |
1262 | err = bpf_map_alloc_id(map); |
1263 | if (err) | |
b936ca64 | 1264 | goto free_map_sec; |
f3f1c054 | 1265 | |
48edc1f7 RG |
1266 | bpf_map_save_memcg(map); |
1267 | ||
6e71b04a | 1268 | err = bpf_map_new_fd(map, f_flags); |
bd5f5f4e MKL |
1269 | if (err < 0) { |
1270 | /* failed to allocate fd. | |
352d20d6 | 1271 | * bpf_map_put_with_uref() is needed because the above |
bd5f5f4e MKL |
1272 | * bpf_map_alloc_id() has published the map |
1273 | * to the userspace and the userspace may | |
1274 | * have refcnt-ed it through BPF_MAP_GET_FD_BY_ID. | |
1275 | */ | |
352d20d6 | 1276 | bpf_map_put_with_uref(map); |
bd5f5f4e MKL |
1277 | return err; |
1278 | } | |
99c55f7d AS |
1279 | |
1280 | return err; | |
1281 | ||
afdb09c7 CF |
1282 | free_map_sec: |
1283 | security_bpf_map_free(map); | |
b936ca64 | 1284 | free_map: |
a26ca7c9 | 1285 | btf_put(map->btf); |
99c55f7d AS |
1286 | map->ops->map_free(map); |
1287 | return err; | |
1288 | } | |
1289 | ||
db20fd2b AS |
1290 | /* if error is returned, fd is released. |
1291 | * On success caller should complete fd access with matching fdput() | |
1292 | */ | |
c2101297 | 1293 | struct bpf_map *__bpf_map_get(struct fd f) |
db20fd2b | 1294 | { |
db20fd2b AS |
1295 | if (!f.file) |
1296 | return ERR_PTR(-EBADF); | |
db20fd2b AS |
1297 | if (f.file->f_op != &bpf_map_fops) { |
1298 | fdput(f); | |
1299 | return ERR_PTR(-EINVAL); | |
1300 | } | |
1301 | ||
c2101297 DB |
1302 | return f.file->private_data; |
1303 | } | |
1304 | ||
1e0bd5a0 | 1305 | void bpf_map_inc(struct bpf_map *map) |
c9da161c | 1306 | { |
1e0bd5a0 | 1307 | atomic64_inc(&map->refcnt); |
c9da161c | 1308 | } |
630a4d38 | 1309 | EXPORT_SYMBOL_GPL(bpf_map_inc); |
c9da161c | 1310 | |
1e0bd5a0 AN |
1311 | void bpf_map_inc_with_uref(struct bpf_map *map) |
1312 | { | |
1313 | atomic64_inc(&map->refcnt); | |
1314 | atomic64_inc(&map->usercnt); | |
1315 | } | |
1316 | EXPORT_SYMBOL_GPL(bpf_map_inc_with_uref); | |
1317 | ||
1ed4d924 MKL |
1318 | struct bpf_map *bpf_map_get(u32 ufd) |
1319 | { | |
1320 | struct fd f = fdget(ufd); | |
1321 | struct bpf_map *map; | |
1322 | ||
1323 | map = __bpf_map_get(f); | |
1324 | if (IS_ERR(map)) | |
1325 | return map; | |
1326 | ||
1327 | bpf_map_inc(map); | |
1328 | fdput(f); | |
1329 | ||
1330 | return map; | |
1331 | } | |
b1d18a75 | 1332 | EXPORT_SYMBOL(bpf_map_get); |
1ed4d924 | 1333 | |
c9da161c | 1334 | struct bpf_map *bpf_map_get_with_uref(u32 ufd) |
c2101297 DB |
1335 | { |
1336 | struct fd f = fdget(ufd); | |
1337 | struct bpf_map *map; | |
1338 | ||
1339 | map = __bpf_map_get(f); | |
1340 | if (IS_ERR(map)) | |
1341 | return map; | |
1342 | ||
1e0bd5a0 | 1343 | bpf_map_inc_with_uref(map); |
c2101297 | 1344 | fdput(f); |
db20fd2b AS |
1345 | |
1346 | return map; | |
1347 | } | |
1348 | ||
b671c206 KFL |
1349 | /* map_idr_lock should have been held or the map should have been |
1350 | * protected by rcu read lock. | |
1351 | */ | |
1352 | struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref) | |
bd5f5f4e MKL |
1353 | { |
1354 | int refold; | |
1355 | ||
1e0bd5a0 | 1356 | refold = atomic64_fetch_add_unless(&map->refcnt, 1, 0); |
bd5f5f4e MKL |
1357 | if (!refold) |
1358 | return ERR_PTR(-ENOENT); | |
bd5f5f4e | 1359 | if (uref) |
1e0bd5a0 | 1360 | atomic64_inc(&map->usercnt); |
bd5f5f4e MKL |
1361 | |
1362 | return map; | |
1363 | } | |
1364 | ||
1e0bd5a0 | 1365 | struct bpf_map *bpf_map_inc_not_zero(struct bpf_map *map) |
b0e4701c SF |
1366 | { |
1367 | spin_lock_bh(&map_idr_lock); | |
1e0bd5a0 | 1368 | map = __bpf_map_inc_not_zero(map, false); |
b0e4701c SF |
1369 | spin_unlock_bh(&map_idr_lock); |
1370 | ||
1371 | return map; | |
1372 | } | |
1373 | EXPORT_SYMBOL_GPL(bpf_map_inc_not_zero); | |
1374 | ||
b8cdc051 AS |
1375 | int __weak bpf_stackmap_copy(struct bpf_map *map, void *key, void *value) |
1376 | { | |
1377 | return -ENOTSUPP; | |
1378 | } | |
1379 | ||
c9d29f46 MV |
1380 | static void *__bpf_copy_key(void __user *ukey, u64 key_size) |
1381 | { | |
1382 | if (key_size) | |
44779a4b | 1383 | return vmemdup_user(ukey, key_size); |
c9d29f46 MV |
1384 | |
1385 | if (ukey) | |
1386 | return ERR_PTR(-EINVAL); | |
1387 | ||
1388 | return NULL; | |
1389 | } | |
1390 | ||
af2ac3e1 AS |
1391 | static void *___bpf_copy_key(bpfptr_t ukey, u64 key_size) |
1392 | { | |
1393 | if (key_size) | |
44779a4b | 1394 | return kvmemdup_bpfptr(ukey, key_size); |
af2ac3e1 AS |
1395 | |
1396 | if (!bpfptr_is_null(ukey)) | |
1397 | return ERR_PTR(-EINVAL); | |
1398 | ||
1399 | return NULL; | |
1400 | } | |
1401 | ||
db20fd2b | 1402 | /* last field in 'union bpf_attr' used by this command */ |
96049f3a | 1403 | #define BPF_MAP_LOOKUP_ELEM_LAST_FIELD flags |
db20fd2b AS |
1404 | |
1405 | static int map_lookup_elem(union bpf_attr *attr) | |
1406 | { | |
535e7b4b MS |
1407 | void __user *ukey = u64_to_user_ptr(attr->key); |
1408 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
db20fd2b | 1409 | int ufd = attr->map_fd; |
db20fd2b | 1410 | struct bpf_map *map; |
15c14a3d | 1411 | void *key, *value; |
15a07b33 | 1412 | u32 value_size; |
592867bf | 1413 | struct fd f; |
db20fd2b AS |
1414 | int err; |
1415 | ||
1416 | if (CHECK_ATTR(BPF_MAP_LOOKUP_ELEM)) | |
1417 | return -EINVAL; | |
1418 | ||
96049f3a AS |
1419 | if (attr->flags & ~BPF_F_LOCK) |
1420 | return -EINVAL; | |
1421 | ||
592867bf | 1422 | f = fdget(ufd); |
c2101297 | 1423 | map = __bpf_map_get(f); |
db20fd2b AS |
1424 | if (IS_ERR(map)) |
1425 | return PTR_ERR(map); | |
87df15de | 1426 | if (!(map_get_sys_perms(map, f) & FMODE_CAN_READ)) { |
6e71b04a CF |
1427 | err = -EPERM; |
1428 | goto err_put; | |
1429 | } | |
1430 | ||
96049f3a | 1431 | if ((attr->flags & BPF_F_LOCK) && |
db559117 | 1432 | !btf_record_has_field(map->record, BPF_SPIN_LOCK)) { |
96049f3a AS |
1433 | err = -EINVAL; |
1434 | goto err_put; | |
1435 | } | |
1436 | ||
c9d29f46 | 1437 | key = __bpf_copy_key(ukey, map->key_size); |
e4448ed8 AV |
1438 | if (IS_ERR(key)) { |
1439 | err = PTR_ERR(key); | |
db20fd2b | 1440 | goto err_put; |
e4448ed8 | 1441 | } |
db20fd2b | 1442 | |
15c14a3d | 1443 | value_size = bpf_map_value_size(map); |
15a07b33 | 1444 | |
8ebe667c | 1445 | err = -ENOMEM; |
f0dce1d9 | 1446 | value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN); |
db20fd2b | 1447 | if (!value) |
8ebe667c AS |
1448 | goto free_key; |
1449 | ||
9330986c JK |
1450 | if (map->map_type == BPF_MAP_TYPE_BLOOM_FILTER) { |
1451 | if (copy_from_user(value, uvalue, value_size)) | |
1452 | err = -EFAULT; | |
1453 | else | |
1454 | err = bpf_map_copy_value(map, key, value, attr->flags); | |
1455 | goto free_value; | |
1456 | } | |
1457 | ||
15c14a3d | 1458 | err = bpf_map_copy_value(map, key, value, attr->flags); |
15a07b33 | 1459 | if (err) |
8ebe667c | 1460 | goto free_value; |
db20fd2b AS |
1461 | |
1462 | err = -EFAULT; | |
15a07b33 | 1463 | if (copy_to_user(uvalue, value, value_size) != 0) |
8ebe667c | 1464 | goto free_value; |
db20fd2b AS |
1465 | |
1466 | err = 0; | |
1467 | ||
8ebe667c | 1468 | free_value: |
f0dce1d9 | 1469 | kvfree(value); |
db20fd2b | 1470 | free_key: |
44779a4b | 1471 | kvfree(key); |
db20fd2b AS |
1472 | err_put: |
1473 | fdput(f); | |
1474 | return err; | |
1475 | } | |
1476 | ||
1ae80cf3 | 1477 | |
3274f520 | 1478 | #define BPF_MAP_UPDATE_ELEM_LAST_FIELD flags |
db20fd2b | 1479 | |
af2ac3e1 | 1480 | static int map_update_elem(union bpf_attr *attr, bpfptr_t uattr) |
db20fd2b | 1481 | { |
af2ac3e1 AS |
1482 | bpfptr_t ukey = make_bpfptr(attr->key, uattr.is_kernel); |
1483 | bpfptr_t uvalue = make_bpfptr(attr->value, uattr.is_kernel); | |
db20fd2b | 1484 | int ufd = attr->map_fd; |
db20fd2b AS |
1485 | struct bpf_map *map; |
1486 | void *key, *value; | |
15a07b33 | 1487 | u32 value_size; |
592867bf | 1488 | struct fd f; |
db20fd2b AS |
1489 | int err; |
1490 | ||
1491 | if (CHECK_ATTR(BPF_MAP_UPDATE_ELEM)) | |
1492 | return -EINVAL; | |
1493 | ||
592867bf | 1494 | f = fdget(ufd); |
c2101297 | 1495 | map = __bpf_map_get(f); |
db20fd2b AS |
1496 | if (IS_ERR(map)) |
1497 | return PTR_ERR(map); | |
353050be | 1498 | bpf_map_write_active_inc(map); |
87df15de | 1499 | if (!(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) { |
6e71b04a CF |
1500 | err = -EPERM; |
1501 | goto err_put; | |
1502 | } | |
1503 | ||
96049f3a | 1504 | if ((attr->flags & BPF_F_LOCK) && |
db559117 | 1505 | !btf_record_has_field(map->record, BPF_SPIN_LOCK)) { |
96049f3a AS |
1506 | err = -EINVAL; |
1507 | goto err_put; | |
1508 | } | |
1509 | ||
af2ac3e1 | 1510 | key = ___bpf_copy_key(ukey, map->key_size); |
e4448ed8 AV |
1511 | if (IS_ERR(key)) { |
1512 | err = PTR_ERR(key); | |
db20fd2b | 1513 | goto err_put; |
e4448ed8 | 1514 | } |
db20fd2b | 1515 | |
f0dce1d9 | 1516 | value_size = bpf_map_value_size(map); |
a02c118e WY |
1517 | value = kvmemdup_bpfptr(uvalue, value_size); |
1518 | if (IS_ERR(value)) { | |
1519 | err = PTR_ERR(value); | |
db20fd2b | 1520 | goto free_key; |
a02c118e | 1521 | } |
db20fd2b | 1522 | |
3af43ba4 | 1523 | err = bpf_map_update_value(map, f.file, key, value, attr->flags); |
6710e112 | 1524 | |
f0dce1d9 | 1525 | kvfree(value); |
db20fd2b | 1526 | free_key: |
44779a4b | 1527 | kvfree(key); |
db20fd2b | 1528 | err_put: |
353050be | 1529 | bpf_map_write_active_dec(map); |
db20fd2b AS |
1530 | fdput(f); |
1531 | return err; | |
1532 | } | |
1533 | ||
1534 | #define BPF_MAP_DELETE_ELEM_LAST_FIELD key | |
1535 | ||
b88df697 | 1536 | static int map_delete_elem(union bpf_attr *attr, bpfptr_t uattr) |
db20fd2b | 1537 | { |
b88df697 | 1538 | bpfptr_t ukey = make_bpfptr(attr->key, uattr.is_kernel); |
db20fd2b | 1539 | int ufd = attr->map_fd; |
db20fd2b | 1540 | struct bpf_map *map; |
592867bf | 1541 | struct fd f; |
db20fd2b AS |
1542 | void *key; |
1543 | int err; | |
1544 | ||
1545 | if (CHECK_ATTR(BPF_MAP_DELETE_ELEM)) | |
1546 | return -EINVAL; | |
1547 | ||
592867bf | 1548 | f = fdget(ufd); |
c2101297 | 1549 | map = __bpf_map_get(f); |
db20fd2b AS |
1550 | if (IS_ERR(map)) |
1551 | return PTR_ERR(map); | |
353050be | 1552 | bpf_map_write_active_inc(map); |
87df15de | 1553 | if (!(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) { |
6e71b04a CF |
1554 | err = -EPERM; |
1555 | goto err_put; | |
1556 | } | |
1557 | ||
b88df697 | 1558 | key = ___bpf_copy_key(ukey, map->key_size); |
e4448ed8 AV |
1559 | if (IS_ERR(key)) { |
1560 | err = PTR_ERR(key); | |
db20fd2b | 1561 | goto err_put; |
e4448ed8 | 1562 | } |
db20fd2b | 1563 | |
9d03ebc7 | 1564 | if (bpf_map_is_offloaded(map)) { |
a3884572 JK |
1565 | err = bpf_map_offload_delete_elem(map, key); |
1566 | goto out; | |
85d33df3 MKL |
1567 | } else if (IS_FD_PROG_ARRAY(map) || |
1568 | map->map_type == BPF_MAP_TYPE_STRUCT_OPS) { | |
1569 | /* These maps require sleepable context */ | |
da765a2f DB |
1570 | err = map->ops->map_delete_elem(map, key); |
1571 | goto out; | |
a3884572 JK |
1572 | } |
1573 | ||
b6e5dae1 | 1574 | bpf_disable_instrumentation(); |
db20fd2b AS |
1575 | rcu_read_lock(); |
1576 | err = map->ops->map_delete_elem(map, key); | |
1577 | rcu_read_unlock(); | |
b6e5dae1 | 1578 | bpf_enable_instrumentation(); |
1ae80cf3 | 1579 | maybe_wait_bpf_programs(map); |
a3884572 | 1580 | out: |
44779a4b | 1581 | kvfree(key); |
db20fd2b | 1582 | err_put: |
353050be | 1583 | bpf_map_write_active_dec(map); |
db20fd2b AS |
1584 | fdput(f); |
1585 | return err; | |
1586 | } | |
1587 | ||
1588 | /* last field in 'union bpf_attr' used by this command */ | |
1589 | #define BPF_MAP_GET_NEXT_KEY_LAST_FIELD next_key | |
1590 | ||
1591 | static int map_get_next_key(union bpf_attr *attr) | |
1592 | { | |
535e7b4b MS |
1593 | void __user *ukey = u64_to_user_ptr(attr->key); |
1594 | void __user *unext_key = u64_to_user_ptr(attr->next_key); | |
db20fd2b | 1595 | int ufd = attr->map_fd; |
db20fd2b AS |
1596 | struct bpf_map *map; |
1597 | void *key, *next_key; | |
592867bf | 1598 | struct fd f; |
db20fd2b AS |
1599 | int err; |
1600 | ||
1601 | if (CHECK_ATTR(BPF_MAP_GET_NEXT_KEY)) | |
1602 | return -EINVAL; | |
1603 | ||
592867bf | 1604 | f = fdget(ufd); |
c2101297 | 1605 | map = __bpf_map_get(f); |
db20fd2b AS |
1606 | if (IS_ERR(map)) |
1607 | return PTR_ERR(map); | |
87df15de | 1608 | if (!(map_get_sys_perms(map, f) & FMODE_CAN_READ)) { |
6e71b04a CF |
1609 | err = -EPERM; |
1610 | goto err_put; | |
1611 | } | |
1612 | ||
8fe45924 | 1613 | if (ukey) { |
c9d29f46 | 1614 | key = __bpf_copy_key(ukey, map->key_size); |
e4448ed8 AV |
1615 | if (IS_ERR(key)) { |
1616 | err = PTR_ERR(key); | |
8fe45924 | 1617 | goto err_put; |
e4448ed8 | 1618 | } |
8fe45924 TQ |
1619 | } else { |
1620 | key = NULL; | |
1621 | } | |
db20fd2b AS |
1622 | |
1623 | err = -ENOMEM; | |
44779a4b | 1624 | next_key = kvmalloc(map->key_size, GFP_USER); |
db20fd2b AS |
1625 | if (!next_key) |
1626 | goto free_key; | |
1627 | ||
9d03ebc7 | 1628 | if (bpf_map_is_offloaded(map)) { |
a3884572 JK |
1629 | err = bpf_map_offload_get_next_key(map, key, next_key); |
1630 | goto out; | |
1631 | } | |
1632 | ||
db20fd2b AS |
1633 | rcu_read_lock(); |
1634 | err = map->ops->map_get_next_key(map, key, next_key); | |
1635 | rcu_read_unlock(); | |
a3884572 | 1636 | out: |
db20fd2b AS |
1637 | if (err) |
1638 | goto free_next_key; | |
1639 | ||
1640 | err = -EFAULT; | |
1641 | if (copy_to_user(unext_key, next_key, map->key_size) != 0) | |
1642 | goto free_next_key; | |
1643 | ||
1644 | err = 0; | |
1645 | ||
1646 | free_next_key: | |
44779a4b | 1647 | kvfree(next_key); |
db20fd2b | 1648 | free_key: |
44779a4b | 1649 | kvfree(key); |
db20fd2b AS |
1650 | err_put: |
1651 | fdput(f); | |
1652 | return err; | |
1653 | } | |
1654 | ||
aa2e93b8 BV |
1655 | int generic_map_delete_batch(struct bpf_map *map, |
1656 | const union bpf_attr *attr, | |
1657 | union bpf_attr __user *uattr) | |
1658 | { | |
1659 | void __user *keys = u64_to_user_ptr(attr->batch.keys); | |
1660 | u32 cp, max_count; | |
1661 | int err = 0; | |
1662 | void *key; | |
1663 | ||
1664 | if (attr->batch.elem_flags & ~BPF_F_LOCK) | |
1665 | return -EINVAL; | |
1666 | ||
1667 | if ((attr->batch.elem_flags & BPF_F_LOCK) && | |
db559117 | 1668 | !btf_record_has_field(map->record, BPF_SPIN_LOCK)) { |
aa2e93b8 BV |
1669 | return -EINVAL; |
1670 | } | |
1671 | ||
1672 | max_count = attr->batch.count; | |
1673 | if (!max_count) | |
1674 | return 0; | |
1675 | ||
44779a4b | 1676 | key = kvmalloc(map->key_size, GFP_USER | __GFP_NOWARN); |
2e3a94aa BV |
1677 | if (!key) |
1678 | return -ENOMEM; | |
1679 | ||
aa2e93b8 | 1680 | for (cp = 0; cp < max_count; cp++) { |
2e3a94aa BV |
1681 | err = -EFAULT; |
1682 | if (copy_from_user(key, keys + cp * map->key_size, | |
1683 | map->key_size)) | |
aa2e93b8 | 1684 | break; |
aa2e93b8 | 1685 | |
9d03ebc7 | 1686 | if (bpf_map_is_offloaded(map)) { |
aa2e93b8 BV |
1687 | err = bpf_map_offload_delete_elem(map, key); |
1688 | break; | |
1689 | } | |
1690 | ||
b6e5dae1 | 1691 | bpf_disable_instrumentation(); |
aa2e93b8 BV |
1692 | rcu_read_lock(); |
1693 | err = map->ops->map_delete_elem(map, key); | |
1694 | rcu_read_unlock(); | |
b6e5dae1 | 1695 | bpf_enable_instrumentation(); |
aa2e93b8 BV |
1696 | if (err) |
1697 | break; | |
75134f16 | 1698 | cond_resched(); |
aa2e93b8 BV |
1699 | } |
1700 | if (copy_to_user(&uattr->batch.count, &cp, sizeof(cp))) | |
1701 | err = -EFAULT; | |
2e3a94aa | 1702 | |
44779a4b | 1703 | kvfree(key); |
9087c6ff ED |
1704 | |
1705 | maybe_wait_bpf_programs(map); | |
aa2e93b8 BV |
1706 | return err; |
1707 | } | |
1708 | ||
3af43ba4 | 1709 | int generic_map_update_batch(struct bpf_map *map, struct file *map_file, |
aa2e93b8 BV |
1710 | const union bpf_attr *attr, |
1711 | union bpf_attr __user *uattr) | |
1712 | { | |
1713 | void __user *values = u64_to_user_ptr(attr->batch.values); | |
1714 | void __user *keys = u64_to_user_ptr(attr->batch.keys); | |
1715 | u32 value_size, cp, max_count; | |
aa2e93b8 | 1716 | void *key, *value; |
aa2e93b8 BV |
1717 | int err = 0; |
1718 | ||
aa2e93b8 BV |
1719 | if (attr->batch.elem_flags & ~BPF_F_LOCK) |
1720 | return -EINVAL; | |
1721 | ||
1722 | if ((attr->batch.elem_flags & BPF_F_LOCK) && | |
db559117 | 1723 | !btf_record_has_field(map->record, BPF_SPIN_LOCK)) { |
aa2e93b8 BV |
1724 | return -EINVAL; |
1725 | } | |
1726 | ||
1727 | value_size = bpf_map_value_size(map); | |
1728 | ||
1729 | max_count = attr->batch.count; | |
1730 | if (!max_count) | |
1731 | return 0; | |
1732 | ||
44779a4b | 1733 | key = kvmalloc(map->key_size, GFP_USER | __GFP_NOWARN); |
2e3a94aa BV |
1734 | if (!key) |
1735 | return -ENOMEM; | |
1736 | ||
f0dce1d9 | 1737 | value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN); |
2e3a94aa | 1738 | if (!value) { |
44779a4b | 1739 | kvfree(key); |
aa2e93b8 | 1740 | return -ENOMEM; |
2e3a94aa | 1741 | } |
aa2e93b8 BV |
1742 | |
1743 | for (cp = 0; cp < max_count; cp++) { | |
aa2e93b8 | 1744 | err = -EFAULT; |
2e3a94aa BV |
1745 | if (copy_from_user(key, keys + cp * map->key_size, |
1746 | map->key_size) || | |
1747 | copy_from_user(value, values + cp * value_size, value_size)) | |
aa2e93b8 BV |
1748 | break; |
1749 | ||
3af43ba4 | 1750 | err = bpf_map_update_value(map, map_file, key, value, |
aa2e93b8 BV |
1751 | attr->batch.elem_flags); |
1752 | ||
1753 | if (err) | |
1754 | break; | |
75134f16 | 1755 | cond_resched(); |
aa2e93b8 BV |
1756 | } |
1757 | ||
1758 | if (copy_to_user(&uattr->batch.count, &cp, sizeof(cp))) | |
1759 | err = -EFAULT; | |
1760 | ||
f0dce1d9 | 1761 | kvfree(value); |
44779a4b | 1762 | kvfree(key); |
aa2e93b8 BV |
1763 | return err; |
1764 | } | |
1765 | ||
cb4d03ab BV |
1766 | #define MAP_LOOKUP_RETRIES 3 |
1767 | ||
1768 | int generic_map_lookup_batch(struct bpf_map *map, | |
1769 | const union bpf_attr *attr, | |
1770 | union bpf_attr __user *uattr) | |
1771 | { | |
1772 | void __user *uobatch = u64_to_user_ptr(attr->batch.out_batch); | |
1773 | void __user *ubatch = u64_to_user_ptr(attr->batch.in_batch); | |
1774 | void __user *values = u64_to_user_ptr(attr->batch.values); | |
1775 | void __user *keys = u64_to_user_ptr(attr->batch.keys); | |
1776 | void *buf, *buf_prevkey, *prev_key, *key, *value; | |
1777 | int err, retry = MAP_LOOKUP_RETRIES; | |
1778 | u32 value_size, cp, max_count; | |
cb4d03ab BV |
1779 | |
1780 | if (attr->batch.elem_flags & ~BPF_F_LOCK) | |
1781 | return -EINVAL; | |
1782 | ||
1783 | if ((attr->batch.elem_flags & BPF_F_LOCK) && | |
db559117 | 1784 | !btf_record_has_field(map->record, BPF_SPIN_LOCK)) |
cb4d03ab BV |
1785 | return -EINVAL; |
1786 | ||
1787 | value_size = bpf_map_value_size(map); | |
1788 | ||
1789 | max_count = attr->batch.count; | |
1790 | if (!max_count) | |
1791 | return 0; | |
1792 | ||
1793 | if (put_user(0, &uattr->batch.count)) | |
1794 | return -EFAULT; | |
1795 | ||
44779a4b | 1796 | buf_prevkey = kvmalloc(map->key_size, GFP_USER | __GFP_NOWARN); |
cb4d03ab BV |
1797 | if (!buf_prevkey) |
1798 | return -ENOMEM; | |
1799 | ||
f0dce1d9 | 1800 | buf = kvmalloc(map->key_size + value_size, GFP_USER | __GFP_NOWARN); |
cb4d03ab | 1801 | if (!buf) { |
44779a4b | 1802 | kvfree(buf_prevkey); |
cb4d03ab BV |
1803 | return -ENOMEM; |
1804 | } | |
1805 | ||
1806 | err = -EFAULT; | |
cb4d03ab BV |
1807 | prev_key = NULL; |
1808 | if (ubatch && copy_from_user(buf_prevkey, ubatch, map->key_size)) | |
1809 | goto free_buf; | |
1810 | key = buf; | |
1811 | value = key + map->key_size; | |
1812 | if (ubatch) | |
1813 | prev_key = buf_prevkey; | |
1814 | ||
1815 | for (cp = 0; cp < max_count;) { | |
1816 | rcu_read_lock(); | |
1817 | err = map->ops->map_get_next_key(map, prev_key, key); | |
1818 | rcu_read_unlock(); | |
1819 | if (err) | |
1820 | break; | |
1821 | err = bpf_map_copy_value(map, key, value, | |
1822 | attr->batch.elem_flags); | |
1823 | ||
1824 | if (err == -ENOENT) { | |
1825 | if (retry) { | |
1826 | retry--; | |
1827 | continue; | |
1828 | } | |
1829 | err = -EINTR; | |
1830 | break; | |
1831 | } | |
1832 | ||
1833 | if (err) | |
1834 | goto free_buf; | |
1835 | ||
1836 | if (copy_to_user(keys + cp * map->key_size, key, | |
1837 | map->key_size)) { | |
1838 | err = -EFAULT; | |
1839 | goto free_buf; | |
1840 | } | |
1841 | if (copy_to_user(values + cp * value_size, value, value_size)) { | |
1842 | err = -EFAULT; | |
1843 | goto free_buf; | |
1844 | } | |
1845 | ||
1846 | if (!prev_key) | |
1847 | prev_key = buf_prevkey; | |
1848 | ||
1849 | swap(prev_key, key); | |
1850 | retry = MAP_LOOKUP_RETRIES; | |
1851 | cp++; | |
75134f16 | 1852 | cond_resched(); |
cb4d03ab BV |
1853 | } |
1854 | ||
1855 | if (err == -EFAULT) | |
1856 | goto free_buf; | |
1857 | ||
1858 | if ((copy_to_user(&uattr->batch.count, &cp, sizeof(cp)) || | |
1859 | (cp && copy_to_user(uobatch, prev_key, map->key_size)))) | |
1860 | err = -EFAULT; | |
1861 | ||
1862 | free_buf: | |
44779a4b | 1863 | kvfree(buf_prevkey); |
f0dce1d9 | 1864 | kvfree(buf); |
cb4d03ab BV |
1865 | return err; |
1866 | } | |
1867 | ||
3e87f192 | 1868 | #define BPF_MAP_LOOKUP_AND_DELETE_ELEM_LAST_FIELD flags |
bd513cd0 MV |
1869 | |
1870 | static int map_lookup_and_delete_elem(union bpf_attr *attr) | |
1871 | { | |
1872 | void __user *ukey = u64_to_user_ptr(attr->key); | |
1873 | void __user *uvalue = u64_to_user_ptr(attr->value); | |
1874 | int ufd = attr->map_fd; | |
1875 | struct bpf_map *map; | |
540fefc0 | 1876 | void *key, *value; |
bd513cd0 MV |
1877 | u32 value_size; |
1878 | struct fd f; | |
1879 | int err; | |
1880 | ||
1881 | if (CHECK_ATTR(BPF_MAP_LOOKUP_AND_DELETE_ELEM)) | |
1882 | return -EINVAL; | |
1883 | ||
3e87f192 DS |
1884 | if (attr->flags & ~BPF_F_LOCK) |
1885 | return -EINVAL; | |
1886 | ||
bd513cd0 MV |
1887 | f = fdget(ufd); |
1888 | map = __bpf_map_get(f); | |
1889 | if (IS_ERR(map)) | |
1890 | return PTR_ERR(map); | |
353050be | 1891 | bpf_map_write_active_inc(map); |
1ea0f912 AP |
1892 | if (!(map_get_sys_perms(map, f) & FMODE_CAN_READ) || |
1893 | !(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) { | |
bd513cd0 MV |
1894 | err = -EPERM; |
1895 | goto err_put; | |
1896 | } | |
1897 | ||
3e87f192 DS |
1898 | if (attr->flags && |
1899 | (map->map_type == BPF_MAP_TYPE_QUEUE || | |
1900 | map->map_type == BPF_MAP_TYPE_STACK)) { | |
1901 | err = -EINVAL; | |
1902 | goto err_put; | |
1903 | } | |
1904 | ||
1905 | if ((attr->flags & BPF_F_LOCK) && | |
db559117 | 1906 | !btf_record_has_field(map->record, BPF_SPIN_LOCK)) { |
3e87f192 DS |
1907 | err = -EINVAL; |
1908 | goto err_put; | |
1909 | } | |
1910 | ||
bd513cd0 MV |
1911 | key = __bpf_copy_key(ukey, map->key_size); |
1912 | if (IS_ERR(key)) { | |
1913 | err = PTR_ERR(key); | |
1914 | goto err_put; | |
1915 | } | |
1916 | ||
3e87f192 | 1917 | value_size = bpf_map_value_size(map); |
bd513cd0 MV |
1918 | |
1919 | err = -ENOMEM; | |
f0dce1d9 | 1920 | value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN); |
bd513cd0 MV |
1921 | if (!value) |
1922 | goto free_key; | |
1923 | ||
3e87f192 | 1924 | err = -ENOTSUPP; |
bd513cd0 MV |
1925 | if (map->map_type == BPF_MAP_TYPE_QUEUE || |
1926 | map->map_type == BPF_MAP_TYPE_STACK) { | |
1927 | err = map->ops->map_pop_elem(map, value); | |
3e87f192 DS |
1928 | } else if (map->map_type == BPF_MAP_TYPE_HASH || |
1929 | map->map_type == BPF_MAP_TYPE_PERCPU_HASH || | |
1930 | map->map_type == BPF_MAP_TYPE_LRU_HASH || | |
1931 | map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { | |
9d03ebc7 | 1932 | if (!bpf_map_is_offloaded(map)) { |
3e87f192 DS |
1933 | bpf_disable_instrumentation(); |
1934 | rcu_read_lock(); | |
1935 | err = map->ops->map_lookup_and_delete_elem(map, key, value, attr->flags); | |
1936 | rcu_read_unlock(); | |
1937 | bpf_enable_instrumentation(); | |
1938 | } | |
bd513cd0 MV |
1939 | } |
1940 | ||
1941 | if (err) | |
1942 | goto free_value; | |
1943 | ||
7f645462 WY |
1944 | if (copy_to_user(uvalue, value, value_size) != 0) { |
1945 | err = -EFAULT; | |
bd513cd0 | 1946 | goto free_value; |
7f645462 | 1947 | } |
bd513cd0 MV |
1948 | |
1949 | err = 0; | |
1950 | ||
1951 | free_value: | |
f0dce1d9 | 1952 | kvfree(value); |
bd513cd0 | 1953 | free_key: |
44779a4b | 1954 | kvfree(key); |
bd513cd0 | 1955 | err_put: |
353050be | 1956 | bpf_map_write_active_dec(map); |
bd513cd0 MV |
1957 | fdput(f); |
1958 | return err; | |
1959 | } | |
1960 | ||
87df15de DB |
1961 | #define BPF_MAP_FREEZE_LAST_FIELD map_fd |
1962 | ||
1963 | static int map_freeze(const union bpf_attr *attr) | |
1964 | { | |
1965 | int err = 0, ufd = attr->map_fd; | |
1966 | struct bpf_map *map; | |
1967 | struct fd f; | |
1968 | ||
1969 | if (CHECK_ATTR(BPF_MAP_FREEZE)) | |
1970 | return -EINVAL; | |
1971 | ||
1972 | f = fdget(ufd); | |
1973 | map = __bpf_map_get(f); | |
1974 | if (IS_ERR(map)) | |
1975 | return PTR_ERR(map); | |
fc970227 | 1976 | |
db559117 | 1977 | if (map->map_type == BPF_MAP_TYPE_STRUCT_OPS || !IS_ERR_OR_NULL(map->record)) { |
849b4d94 MKL |
1978 | fdput(f); |
1979 | return -ENOTSUPP; | |
1980 | } | |
1981 | ||
c4c84f6f | 1982 | if (!(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) { |
4266f41f DB |
1983 | fdput(f); |
1984 | return -EPERM; | |
c4c84f6f AN |
1985 | } |
1986 | ||
fc970227 | 1987 | mutex_lock(&map->freeze_mutex); |
353050be | 1988 | if (bpf_map_write_active(map)) { |
fc970227 AN |
1989 | err = -EBUSY; |
1990 | goto err_put; | |
1991 | } | |
87df15de DB |
1992 | if (READ_ONCE(map->frozen)) { |
1993 | err = -EBUSY; | |
1994 | goto err_put; | |
1995 | } | |
87df15de DB |
1996 | |
1997 | WRITE_ONCE(map->frozen, true); | |
1998 | err_put: | |
fc970227 | 1999 | mutex_unlock(&map->freeze_mutex); |
87df15de DB |
2000 | fdput(f); |
2001 | return err; | |
2002 | } | |
2003 | ||
7de16e3a | 2004 | static const struct bpf_prog_ops * const bpf_prog_types[] = { |
91cc1a99 | 2005 | #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ |
7de16e3a JK |
2006 | [_id] = & _name ## _prog_ops, |
2007 | #define BPF_MAP_TYPE(_id, _ops) | |
f2e10bff | 2008 | #define BPF_LINK_TYPE(_id, _name) |
7de16e3a JK |
2009 | #include <linux/bpf_types.h> |
2010 | #undef BPF_PROG_TYPE | |
2011 | #undef BPF_MAP_TYPE | |
f2e10bff | 2012 | #undef BPF_LINK_TYPE |
7de16e3a JK |
2013 | }; |
2014 | ||
09756af4 AS |
2015 | static int find_prog_type(enum bpf_prog_type type, struct bpf_prog *prog) |
2016 | { | |
d0f1a451 DB |
2017 | const struct bpf_prog_ops *ops; |
2018 | ||
2019 | if (type >= ARRAY_SIZE(bpf_prog_types)) | |
2020 | return -EINVAL; | |
2021 | type = array_index_nospec(type, ARRAY_SIZE(bpf_prog_types)); | |
2022 | ops = bpf_prog_types[type]; | |
2023 | if (!ops) | |
be9370a7 | 2024 | return -EINVAL; |
09756af4 | 2025 | |
9d03ebc7 | 2026 | if (!bpf_prog_is_offloaded(prog->aux)) |
d0f1a451 | 2027 | prog->aux->ops = ops; |
ab3f0063 JK |
2028 | else |
2029 | prog->aux->ops = &bpf_offload_prog_ops; | |
be9370a7 JB |
2030 | prog->type = type; |
2031 | return 0; | |
09756af4 AS |
2032 | } |
2033 | ||
bae141f5 DB |
2034 | enum bpf_audit { |
2035 | BPF_AUDIT_LOAD, | |
2036 | BPF_AUDIT_UNLOAD, | |
2037 | BPF_AUDIT_MAX, | |
2038 | }; | |
2039 | ||
2040 | static const char * const bpf_audit_str[BPF_AUDIT_MAX] = { | |
2041 | [BPF_AUDIT_LOAD] = "LOAD", | |
2042 | [BPF_AUDIT_UNLOAD] = "UNLOAD", | |
2043 | }; | |
2044 | ||
2045 | static void bpf_audit_prog(const struct bpf_prog *prog, unsigned int op) | |
2046 | { | |
2047 | struct audit_context *ctx = NULL; | |
2048 | struct audit_buffer *ab; | |
2049 | ||
2050 | if (WARN_ON_ONCE(op >= BPF_AUDIT_MAX)) | |
2051 | return; | |
2052 | if (audit_enabled == AUDIT_OFF) | |
2053 | return; | |
ef01f4e2 | 2054 | if (!in_irq() && !irqs_disabled()) |
bae141f5 DB |
2055 | ctx = audit_context(); |
2056 | ab = audit_log_start(ctx, GFP_ATOMIC, AUDIT_BPF); | |
2057 | if (unlikely(!ab)) | |
2058 | return; | |
2059 | audit_log_format(ab, "prog-id=%u op=%s", | |
2060 | prog->aux->id, bpf_audit_str[op]); | |
2061 | audit_log_end(ab); | |
2062 | } | |
2063 | ||
dc4bb0e2 MKL |
2064 | static int bpf_prog_alloc_id(struct bpf_prog *prog) |
2065 | { | |
2066 | int id; | |
2067 | ||
b76354cd | 2068 | idr_preload(GFP_KERNEL); |
dc4bb0e2 MKL |
2069 | spin_lock_bh(&prog_idr_lock); |
2070 | id = idr_alloc_cyclic(&prog_idr, prog, 1, INT_MAX, GFP_ATOMIC); | |
2071 | if (id > 0) | |
2072 | prog->aux->id = id; | |
2073 | spin_unlock_bh(&prog_idr_lock); | |
b76354cd | 2074 | idr_preload_end(); |
dc4bb0e2 MKL |
2075 | |
2076 | /* id is in [1, INT_MAX) */ | |
2077 | if (WARN_ON_ONCE(!id)) | |
2078 | return -ENOSPC; | |
2079 | ||
2080 | return id > 0 ? 0 : id; | |
2081 | } | |
2082 | ||
e7895f01 | 2083 | void bpf_prog_free_id(struct bpf_prog *prog) |
dc4bb0e2 | 2084 | { |
d809e134 AS |
2085 | unsigned long flags; |
2086 | ||
ad8ad79f JK |
2087 | /* cBPF to eBPF migrations are currently not in the idr store. |
2088 | * Offloaded programs are removed from the store when their device | |
2089 | * disappears - even if someone grabs an fd to them they are unusable, | |
2090 | * simply waiting for refcnt to drop to be freed. | |
2091 | */ | |
dc4bb0e2 MKL |
2092 | if (!prog->aux->id) |
2093 | return; | |
2094 | ||
e7895f01 | 2095 | spin_lock_irqsave(&prog_idr_lock, flags); |
dc4bb0e2 | 2096 | idr_remove(&prog_idr, prog->aux->id); |
ad8ad79f | 2097 | prog->aux->id = 0; |
e7895f01 | 2098 | spin_unlock_irqrestore(&prog_idr_lock, flags); |
dc4bb0e2 MKL |
2099 | } |
2100 | ||
1aacde3d | 2101 | static void __bpf_prog_put_rcu(struct rcu_head *rcu) |
abf2e7d6 AS |
2102 | { |
2103 | struct bpf_prog_aux *aux = container_of(rcu, struct bpf_prog_aux, rcu); | |
2104 | ||
3b4d9eb2 | 2105 | kvfree(aux->func_info); |
8c1b6e69 | 2106 | kfree(aux->func_info_aux); |
3ac1f01b | 2107 | free_uid(aux->user); |
afdb09c7 | 2108 | security_bpf_prog_free(aux); |
abf2e7d6 AS |
2109 | bpf_prog_free(aux->prog); |
2110 | } | |
2111 | ||
cd7455f1 DB |
2112 | static void __bpf_prog_put_noref(struct bpf_prog *prog, bool deferred) |
2113 | { | |
2114 | bpf_prog_kallsyms_del_all(prog); | |
2115 | btf_put(prog->aux->btf); | |
31bf1dbc | 2116 | module_put(prog->aux->mod); |
e16301fb MKL |
2117 | kvfree(prog->aux->jited_linfo); |
2118 | kvfree(prog->aux->linfo); | |
e6ac2450 | 2119 | kfree(prog->aux->kfunc_tab); |
22dc4a0f AN |
2120 | if (prog->aux->attach_btf) |
2121 | btf_put(prog->aux->attach_btf); | |
cd7455f1 | 2122 | |
1e6c62a8 AS |
2123 | if (deferred) { |
2124 | if (prog->aux->sleepable) | |
2125 | call_rcu_tasks_trace(&prog->aux->rcu, __bpf_prog_put_rcu); | |
2126 | else | |
2127 | call_rcu(&prog->aux->rcu, __bpf_prog_put_rcu); | |
2128 | } else { | |
cd7455f1 | 2129 | __bpf_prog_put_rcu(&prog->aux->rcu); |
1e6c62a8 | 2130 | } |
cd7455f1 DB |
2131 | } |
2132 | ||
d809e134 AS |
2133 | static void bpf_prog_put_deferred(struct work_struct *work) |
2134 | { | |
2135 | struct bpf_prog_aux *aux; | |
2136 | struct bpf_prog *prog; | |
2137 | ||
2138 | aux = container_of(work, struct bpf_prog_aux, work); | |
2139 | prog = aux->prog; | |
2140 | perf_event_bpf_event(prog, PERF_BPF_EVENT_PROG_UNLOAD, 0); | |
2141 | bpf_audit_prog(prog, BPF_AUDIT_UNLOAD); | |
e7895f01 | 2142 | bpf_prog_free_id(prog); |
d809e134 AS |
2143 | __bpf_prog_put_noref(prog, true); |
2144 | } | |
2145 | ||
e7895f01 | 2146 | static void __bpf_prog_put(struct bpf_prog *prog) |
09756af4 | 2147 | { |
d809e134 AS |
2148 | struct bpf_prog_aux *aux = prog->aux; |
2149 | ||
2150 | if (atomic64_dec_and_test(&aux->refcnt)) { | |
d809e134 AS |
2151 | if (in_irq() || irqs_disabled()) { |
2152 | INIT_WORK(&aux->work, bpf_prog_put_deferred); | |
2153 | schedule_work(&aux->work); | |
2154 | } else { | |
2155 | bpf_prog_put_deferred(&aux->work); | |
2156 | } | |
a67edbf4 | 2157 | } |
09756af4 | 2158 | } |
b16d9aa4 MKL |
2159 | |
2160 | void bpf_prog_put(struct bpf_prog *prog) | |
2161 | { | |
e7895f01 | 2162 | __bpf_prog_put(prog); |
b16d9aa4 | 2163 | } |
e2e9b654 | 2164 | EXPORT_SYMBOL_GPL(bpf_prog_put); |
09756af4 AS |
2165 | |
2166 | static int bpf_prog_release(struct inode *inode, struct file *filp) | |
2167 | { | |
2168 | struct bpf_prog *prog = filp->private_data; | |
2169 | ||
1aacde3d | 2170 | bpf_prog_put(prog); |
09756af4 AS |
2171 | return 0; |
2172 | } | |
2173 | ||
61a0abae ED |
2174 | struct bpf_prog_kstats { |
2175 | u64 nsecs; | |
2176 | u64 cnt; | |
2177 | u64 misses; | |
2178 | }; | |
2179 | ||
05b24ff9 JO |
2180 | void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog) |
2181 | { | |
2182 | struct bpf_prog_stats *stats; | |
2183 | unsigned int flags; | |
2184 | ||
2185 | stats = this_cpu_ptr(prog->stats); | |
2186 | flags = u64_stats_update_begin_irqsave(&stats->syncp); | |
2187 | u64_stats_inc(&stats->misses); | |
2188 | u64_stats_update_end_irqrestore(&stats->syncp, flags); | |
2189 | } | |
2190 | ||
492ecee8 | 2191 | static void bpf_prog_get_stats(const struct bpf_prog *prog, |
61a0abae | 2192 | struct bpf_prog_kstats *stats) |
492ecee8 | 2193 | { |
9ed9e9ba | 2194 | u64 nsecs = 0, cnt = 0, misses = 0; |
492ecee8 AS |
2195 | int cpu; |
2196 | ||
2197 | for_each_possible_cpu(cpu) { | |
2198 | const struct bpf_prog_stats *st; | |
2199 | unsigned int start; | |
9ed9e9ba | 2200 | u64 tnsecs, tcnt, tmisses; |
492ecee8 | 2201 | |
700d4796 | 2202 | st = per_cpu_ptr(prog->stats, cpu); |
492ecee8 | 2203 | do { |
97c4090b | 2204 | start = u64_stats_fetch_begin(&st->syncp); |
61a0abae ED |
2205 | tnsecs = u64_stats_read(&st->nsecs); |
2206 | tcnt = u64_stats_read(&st->cnt); | |
2207 | tmisses = u64_stats_read(&st->misses); | |
97c4090b | 2208 | } while (u64_stats_fetch_retry(&st->syncp, start)); |
492ecee8 AS |
2209 | nsecs += tnsecs; |
2210 | cnt += tcnt; | |
9ed9e9ba | 2211 | misses += tmisses; |
492ecee8 AS |
2212 | } |
2213 | stats->nsecs = nsecs; | |
2214 | stats->cnt = cnt; | |
9ed9e9ba | 2215 | stats->misses = misses; |
492ecee8 AS |
2216 | } |
2217 | ||
7bd509e3 DB |
2218 | #ifdef CONFIG_PROC_FS |
2219 | static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp) | |
2220 | { | |
2221 | const struct bpf_prog *prog = filp->private_data; | |
f1f7714e | 2222 | char prog_tag[sizeof(prog->tag) * 2 + 1] = { }; |
61a0abae | 2223 | struct bpf_prog_kstats stats; |
7bd509e3 | 2224 | |
492ecee8 | 2225 | bpf_prog_get_stats(prog, &stats); |
f1f7714e | 2226 | bin2hex(prog_tag, prog->tag, sizeof(prog->tag)); |
7bd509e3 DB |
2227 | seq_printf(m, |
2228 | "prog_type:\t%u\n" | |
2229 | "prog_jited:\t%u\n" | |
f1f7714e | 2230 | "prog_tag:\t%s\n" |
4316b409 | 2231 | "memlock:\t%llu\n" |
492ecee8 AS |
2232 | "prog_id:\t%u\n" |
2233 | "run_time_ns:\t%llu\n" | |
9ed9e9ba | 2234 | "run_cnt:\t%llu\n" |
aba64c7d DM |
2235 | "recursion_misses:\t%llu\n" |
2236 | "verified_insns:\t%u\n", | |
7bd509e3 DB |
2237 | prog->type, |
2238 | prog->jited, | |
f1f7714e | 2239 | prog_tag, |
4316b409 | 2240 | prog->pages * 1ULL << PAGE_SHIFT, |
492ecee8 AS |
2241 | prog->aux->id, |
2242 | stats.nsecs, | |
9ed9e9ba | 2243 | stats.cnt, |
aba64c7d DM |
2244 | stats.misses, |
2245 | prog->aux->verified_insns); | |
7bd509e3 DB |
2246 | } |
2247 | #endif | |
2248 | ||
f66e448c | 2249 | const struct file_operations bpf_prog_fops = { |
7bd509e3 DB |
2250 | #ifdef CONFIG_PROC_FS |
2251 | .show_fdinfo = bpf_prog_show_fdinfo, | |
2252 | #endif | |
2253 | .release = bpf_prog_release, | |
6e71b04a CF |
2254 | .read = bpf_dummy_read, |
2255 | .write = bpf_dummy_write, | |
09756af4 AS |
2256 | }; |
2257 | ||
b2197755 | 2258 | int bpf_prog_new_fd(struct bpf_prog *prog) |
aa79781b | 2259 | { |
afdb09c7 CF |
2260 | int ret; |
2261 | ||
2262 | ret = security_bpf_prog(prog); | |
2263 | if (ret < 0) | |
2264 | return ret; | |
2265 | ||
aa79781b DB |
2266 | return anon_inode_getfd("bpf-prog", &bpf_prog_fops, prog, |
2267 | O_RDWR | O_CLOEXEC); | |
2268 | } | |
2269 | ||
113214be | 2270 | static struct bpf_prog *____bpf_prog_get(struct fd f) |
09756af4 | 2271 | { |
09756af4 AS |
2272 | if (!f.file) |
2273 | return ERR_PTR(-EBADF); | |
09756af4 AS |
2274 | if (f.file->f_op != &bpf_prog_fops) { |
2275 | fdput(f); | |
2276 | return ERR_PTR(-EINVAL); | |
2277 | } | |
2278 | ||
c2101297 | 2279 | return f.file->private_data; |
09756af4 AS |
2280 | } |
2281 | ||
85192dbf | 2282 | void bpf_prog_add(struct bpf_prog *prog, int i) |
92117d84 | 2283 | { |
85192dbf | 2284 | atomic64_add(i, &prog->aux->refcnt); |
92117d84 | 2285 | } |
59d3656d BB |
2286 | EXPORT_SYMBOL_GPL(bpf_prog_add); |
2287 | ||
c540594f DB |
2288 | void bpf_prog_sub(struct bpf_prog *prog, int i) |
2289 | { | |
2290 | /* Only to be used for undoing previous bpf_prog_add() in some | |
2291 | * error path. We still know that another entity in our call | |
2292 | * path holds a reference to the program, thus atomic_sub() can | |
2293 | * be safely used in such cases! | |
2294 | */ | |
85192dbf | 2295 | WARN_ON(atomic64_sub_return(i, &prog->aux->refcnt) == 0); |
c540594f DB |
2296 | } |
2297 | EXPORT_SYMBOL_GPL(bpf_prog_sub); | |
2298 | ||
85192dbf | 2299 | void bpf_prog_inc(struct bpf_prog *prog) |
59d3656d | 2300 | { |
85192dbf | 2301 | atomic64_inc(&prog->aux->refcnt); |
59d3656d | 2302 | } |
97bc402d | 2303 | EXPORT_SYMBOL_GPL(bpf_prog_inc); |
92117d84 | 2304 | |
b16d9aa4 | 2305 | /* prog_idr_lock should have been held */ |
a6f6df69 | 2306 | struct bpf_prog *bpf_prog_inc_not_zero(struct bpf_prog *prog) |
b16d9aa4 MKL |
2307 | { |
2308 | int refold; | |
2309 | ||
85192dbf | 2310 | refold = atomic64_fetch_add_unless(&prog->aux->refcnt, 1, 0); |
b16d9aa4 MKL |
2311 | |
2312 | if (!refold) | |
2313 | return ERR_PTR(-ENOENT); | |
2314 | ||
2315 | return prog; | |
2316 | } | |
a6f6df69 | 2317 | EXPORT_SYMBOL_GPL(bpf_prog_inc_not_zero); |
b16d9aa4 | 2318 | |
040ee692 | 2319 | bool bpf_prog_get_ok(struct bpf_prog *prog, |
288b3de5 | 2320 | enum bpf_prog_type *attach_type, bool attach_drv) |
248f346f | 2321 | { |
288b3de5 JK |
2322 | /* not an attachment, just a refcount inc, always allow */ |
2323 | if (!attach_type) | |
2324 | return true; | |
248f346f JK |
2325 | |
2326 | if (prog->type != *attach_type) | |
2327 | return false; | |
9d03ebc7 | 2328 | if (bpf_prog_is_offloaded(prog->aux) && !attach_drv) |
248f346f JK |
2329 | return false; |
2330 | ||
2331 | return true; | |
2332 | } | |
2333 | ||
2334 | static struct bpf_prog *__bpf_prog_get(u32 ufd, enum bpf_prog_type *attach_type, | |
288b3de5 | 2335 | bool attach_drv) |
09756af4 AS |
2336 | { |
2337 | struct fd f = fdget(ufd); | |
2338 | struct bpf_prog *prog; | |
2339 | ||
113214be | 2340 | prog = ____bpf_prog_get(f); |
09756af4 AS |
2341 | if (IS_ERR(prog)) |
2342 | return prog; | |
288b3de5 | 2343 | if (!bpf_prog_get_ok(prog, attach_type, attach_drv)) { |
113214be DB |
2344 | prog = ERR_PTR(-EINVAL); |
2345 | goto out; | |
2346 | } | |
09756af4 | 2347 | |
85192dbf | 2348 | bpf_prog_inc(prog); |
113214be | 2349 | out: |
09756af4 AS |
2350 | fdput(f); |
2351 | return prog; | |
2352 | } | |
113214be DB |
2353 | |
2354 | struct bpf_prog *bpf_prog_get(u32 ufd) | |
2355 | { | |
288b3de5 | 2356 | return __bpf_prog_get(ufd, NULL, false); |
113214be DB |
2357 | } |
2358 | ||
248f346f | 2359 | struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, |
288b3de5 | 2360 | bool attach_drv) |
248f346f | 2361 | { |
4d220ed0 | 2362 | return __bpf_prog_get(ufd, &type, attach_drv); |
248f346f | 2363 | } |
6c8dfe21 | 2364 | EXPORT_SYMBOL_GPL(bpf_prog_get_type_dev); |
248f346f | 2365 | |
aac3fc32 AI |
2366 | /* Initially all BPF programs could be loaded w/o specifying |
2367 | * expected_attach_type. Later for some of them specifying expected_attach_type | |
2368 | * at load time became required so that program could be validated properly. | |
2369 | * Programs of types that are allowed to be loaded both w/ and w/o (for | |
2370 | * backward compatibility) expected_attach_type, should have the default attach | |
2371 | * type assigned to expected_attach_type for the latter case, so that it can be | |
2372 | * validated later at attach time. | |
2373 | * | |
2374 | * bpf_prog_load_fixup_attach_type() sets expected_attach_type in @attr if | |
2375 | * prog type requires it but has some attach types that have to be backward | |
2376 | * compatible. | |
2377 | */ | |
2378 | static void bpf_prog_load_fixup_attach_type(union bpf_attr *attr) | |
2379 | { | |
2380 | switch (attr->prog_type) { | |
2381 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
2382 | /* Unfortunately BPF_ATTACH_TYPE_UNSPEC enumeration doesn't | |
2383 | * exist so checking for non-zero is the way to go here. | |
2384 | */ | |
2385 | if (!attr->expected_attach_type) | |
2386 | attr->expected_attach_type = | |
2387 | BPF_CGROUP_INET_SOCK_CREATE; | |
2388 | break; | |
d5e4ddae KI |
2389 | case BPF_PROG_TYPE_SK_REUSEPORT: |
2390 | if (!attr->expected_attach_type) | |
2391 | attr->expected_attach_type = | |
2392 | BPF_SK_REUSEPORT_SELECT; | |
2393 | break; | |
aac3fc32 AI |
2394 | } |
2395 | } | |
2396 | ||
5e43f899 | 2397 | static int |
ccfe29eb AS |
2398 | bpf_prog_load_check_attach(enum bpf_prog_type prog_type, |
2399 | enum bpf_attach_type expected_attach_type, | |
290248a5 AN |
2400 | struct btf *attach_btf, u32 btf_id, |
2401 | struct bpf_prog *dst_prog) | |
5e43f899 | 2402 | { |
27ae7997 | 2403 | if (btf_id) { |
c108e3c1 AS |
2404 | if (btf_id > BTF_MAX_TYPE) |
2405 | return -EINVAL; | |
27ae7997 | 2406 | |
290248a5 AN |
2407 | if (!attach_btf && !dst_prog) |
2408 | return -EINVAL; | |
2409 | ||
27ae7997 MKL |
2410 | switch (prog_type) { |
2411 | case BPF_PROG_TYPE_TRACING: | |
9e4e01df | 2412 | case BPF_PROG_TYPE_LSM: |
27ae7997 | 2413 | case BPF_PROG_TYPE_STRUCT_OPS: |
be8704ff | 2414 | case BPF_PROG_TYPE_EXT: |
27ae7997 MKL |
2415 | break; |
2416 | default: | |
c108e3c1 | 2417 | return -EINVAL; |
27ae7997 | 2418 | } |
c108e3c1 AS |
2419 | } |
2420 | ||
290248a5 AN |
2421 | if (attach_btf && (!btf_id || dst_prog)) |
2422 | return -EINVAL; | |
2423 | ||
2424 | if (dst_prog && prog_type != BPF_PROG_TYPE_TRACING && | |
be8704ff | 2425 | prog_type != BPF_PROG_TYPE_EXT) |
27ae7997 MKL |
2426 | return -EINVAL; |
2427 | ||
4fbac77d | 2428 | switch (prog_type) { |
aac3fc32 AI |
2429 | case BPF_PROG_TYPE_CGROUP_SOCK: |
2430 | switch (expected_attach_type) { | |
2431 | case BPF_CGROUP_INET_SOCK_CREATE: | |
f5836749 | 2432 | case BPF_CGROUP_INET_SOCK_RELEASE: |
aac3fc32 AI |
2433 | case BPF_CGROUP_INET4_POST_BIND: |
2434 | case BPF_CGROUP_INET6_POST_BIND: | |
2435 | return 0; | |
2436 | default: | |
2437 | return -EINVAL; | |
2438 | } | |
4fbac77d AI |
2439 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: |
2440 | switch (expected_attach_type) { | |
2441 | case BPF_CGROUP_INET4_BIND: | |
2442 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
2443 | case BPF_CGROUP_INET4_CONNECT: |
2444 | case BPF_CGROUP_INET6_CONNECT: | |
1b66d253 DB |
2445 | case BPF_CGROUP_INET4_GETPEERNAME: |
2446 | case BPF_CGROUP_INET6_GETPEERNAME: | |
2447 | case BPF_CGROUP_INET4_GETSOCKNAME: | |
2448 | case BPF_CGROUP_INET6_GETSOCKNAME: | |
1cedee13 AI |
2449 | case BPF_CGROUP_UDP4_SENDMSG: |
2450 | case BPF_CGROUP_UDP6_SENDMSG: | |
983695fa DB |
2451 | case BPF_CGROUP_UDP4_RECVMSG: |
2452 | case BPF_CGROUP_UDP6_RECVMSG: | |
4fbac77d AI |
2453 | return 0; |
2454 | default: | |
2455 | return -EINVAL; | |
2456 | } | |
5cf1e914 | 2457 | case BPF_PROG_TYPE_CGROUP_SKB: |
2458 | switch (expected_attach_type) { | |
2459 | case BPF_CGROUP_INET_INGRESS: | |
2460 | case BPF_CGROUP_INET_EGRESS: | |
2461 | return 0; | |
2462 | default: | |
2463 | return -EINVAL; | |
2464 | } | |
0d01da6a SF |
2465 | case BPF_PROG_TYPE_CGROUP_SOCKOPT: |
2466 | switch (expected_attach_type) { | |
2467 | case BPF_CGROUP_SETSOCKOPT: | |
2468 | case BPF_CGROUP_GETSOCKOPT: | |
2469 | return 0; | |
2470 | default: | |
2471 | return -EINVAL; | |
2472 | } | |
e9ddbb77 JS |
2473 | case BPF_PROG_TYPE_SK_LOOKUP: |
2474 | if (expected_attach_type == BPF_SK_LOOKUP) | |
2475 | return 0; | |
2476 | return -EINVAL; | |
d5e4ddae KI |
2477 | case BPF_PROG_TYPE_SK_REUSEPORT: |
2478 | switch (expected_attach_type) { | |
2479 | case BPF_SK_REUSEPORT_SELECT: | |
2480 | case BPF_SK_REUSEPORT_SELECT_OR_MIGRATE: | |
2481 | return 0; | |
2482 | default: | |
2483 | return -EINVAL; | |
2484 | } | |
132328e8 FW |
2485 | case BPF_PROG_TYPE_NETFILTER: |
2486 | if (expected_attach_type == BPF_NETFILTER) | |
2487 | return 0; | |
2488 | return -EINVAL; | |
79a7f8bd | 2489 | case BPF_PROG_TYPE_SYSCALL: |
be8704ff AS |
2490 | case BPF_PROG_TYPE_EXT: |
2491 | if (expected_attach_type) | |
2492 | return -EINVAL; | |
df561f66 | 2493 | fallthrough; |
4fbac77d AI |
2494 | default: |
2495 | return 0; | |
2496 | } | |
5e43f899 AI |
2497 | } |
2498 | ||
2c78ee89 AS |
2499 | static bool is_net_admin_prog_type(enum bpf_prog_type prog_type) |
2500 | { | |
2501 | switch (prog_type) { | |
2502 | case BPF_PROG_TYPE_SCHED_CLS: | |
2503 | case BPF_PROG_TYPE_SCHED_ACT: | |
2504 | case BPF_PROG_TYPE_XDP: | |
2505 | case BPF_PROG_TYPE_LWT_IN: | |
2506 | case BPF_PROG_TYPE_LWT_OUT: | |
2507 | case BPF_PROG_TYPE_LWT_XMIT: | |
2508 | case BPF_PROG_TYPE_LWT_SEG6LOCAL: | |
2509 | case BPF_PROG_TYPE_SK_SKB: | |
2510 | case BPF_PROG_TYPE_SK_MSG: | |
2c78ee89 AS |
2511 | case BPF_PROG_TYPE_FLOW_DISSECTOR: |
2512 | case BPF_PROG_TYPE_CGROUP_DEVICE: | |
2513 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
2514 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: | |
2515 | case BPF_PROG_TYPE_CGROUP_SOCKOPT: | |
2516 | case BPF_PROG_TYPE_CGROUP_SYSCTL: | |
2517 | case BPF_PROG_TYPE_SOCK_OPS: | |
2518 | case BPF_PROG_TYPE_EXT: /* extends any prog */ | |
84601d6e | 2519 | case BPF_PROG_TYPE_NETFILTER: |
2c78ee89 AS |
2520 | return true; |
2521 | case BPF_PROG_TYPE_CGROUP_SKB: | |
2522 | /* always unpriv */ | |
2523 | case BPF_PROG_TYPE_SK_REUSEPORT: | |
2524 | /* equivalent to SOCKET_FILTER. need CAP_BPF only */ | |
2525 | default: | |
2526 | return false; | |
2527 | } | |
2528 | } | |
2529 | ||
2530 | static bool is_perfmon_prog_type(enum bpf_prog_type prog_type) | |
2531 | { | |
2532 | switch (prog_type) { | |
2533 | case BPF_PROG_TYPE_KPROBE: | |
2534 | case BPF_PROG_TYPE_TRACEPOINT: | |
2535 | case BPF_PROG_TYPE_PERF_EVENT: | |
2536 | case BPF_PROG_TYPE_RAW_TRACEPOINT: | |
2537 | case BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE: | |
2538 | case BPF_PROG_TYPE_TRACING: | |
2539 | case BPF_PROG_TYPE_LSM: | |
2540 | case BPF_PROG_TYPE_STRUCT_OPS: /* has access to struct sock */ | |
2541 | case BPF_PROG_TYPE_EXT: /* extends any prog */ | |
2542 | return true; | |
2543 | default: | |
2544 | return false; | |
2545 | } | |
2546 | } | |
2547 | ||
09756af4 | 2548 | /* last field in 'union bpf_attr' used by this command */ |
47a71c1f | 2549 | #define BPF_PROG_LOAD_LAST_FIELD log_true_size |
09756af4 | 2550 | |
47a71c1f | 2551 | static int bpf_prog_load(union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size) |
09756af4 AS |
2552 | { |
2553 | enum bpf_prog_type type = attr->prog_type; | |
290248a5 AN |
2554 | struct bpf_prog *prog, *dst_prog = NULL; |
2555 | struct btf *attach_btf = NULL; | |
09756af4 AS |
2556 | int err; |
2557 | char license[128]; | |
09756af4 AS |
2558 | |
2559 | if (CHECK_ATTR(BPF_PROG_LOAD)) | |
2560 | return -EINVAL; | |
2561 | ||
c240eff6 JW |
2562 | if (attr->prog_flags & ~(BPF_F_STRICT_ALIGNMENT | |
2563 | BPF_F_ANY_ALIGNMENT | | |
10d274e8 | 2564 | BPF_F_TEST_STATE_FREQ | |
1e6c62a8 | 2565 | BPF_F_SLEEPABLE | |
c2f2cdbe | 2566 | BPF_F_TEST_RND_HI32 | |
2b3486bc SF |
2567 | BPF_F_XDP_HAS_FRAGS | |
2568 | BPF_F_XDP_DEV_BOUND_ONLY)) | |
e07b98d9 DM |
2569 | return -EINVAL; |
2570 | ||
e9ee9efc DM |
2571 | if (!IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && |
2572 | (attr->prog_flags & BPF_F_ANY_ALIGNMENT) && | |
2c78ee89 | 2573 | !bpf_capable()) |
e9ee9efc DM |
2574 | return -EPERM; |
2575 | ||
1d28635a AN |
2576 | /* Intent here is for unprivileged_bpf_disabled to block BPF program |
2577 | * creation for unprivileged users; other actions depend | |
2578 | * on fd availability and access to bpffs, so are dependent on | |
2579 | * object creation success. Even with unprivileged BPF disabled, | |
2580 | * capability checks are still carried out for these | |
2581 | * and other operations. | |
2582 | */ | |
2583 | if (sysctl_unprivileged_bpf_disabled && !bpf_capable()) | |
2584 | return -EPERM; | |
09756af4 | 2585 | |
c04c0d2b | 2586 | if (attr->insn_cnt == 0 || |
2c78ee89 | 2587 | attr->insn_cnt > (bpf_capable() ? BPF_COMPLEXITY_LIMIT_INSNS : BPF_MAXINSNS)) |
ef0915ca | 2588 | return -E2BIG; |
80b7d819 CF |
2589 | if (type != BPF_PROG_TYPE_SOCKET_FILTER && |
2590 | type != BPF_PROG_TYPE_CGROUP_SKB && | |
2c78ee89 AS |
2591 | !bpf_capable()) |
2592 | return -EPERM; | |
2593 | ||
b338cb92 | 2594 | if (is_net_admin_prog_type(type) && !capable(CAP_NET_ADMIN) && !capable(CAP_SYS_ADMIN)) |
2c78ee89 AS |
2595 | return -EPERM; |
2596 | if (is_perfmon_prog_type(type) && !perfmon_capable()) | |
1be7f75d AS |
2597 | return -EPERM; |
2598 | ||
290248a5 AN |
2599 | /* attach_prog_fd/attach_btf_obj_fd can specify fd of either bpf_prog |
2600 | * or btf, we need to check which one it is | |
2601 | */ | |
2602 | if (attr->attach_prog_fd) { | |
2603 | dst_prog = bpf_prog_get(attr->attach_prog_fd); | |
2604 | if (IS_ERR(dst_prog)) { | |
2605 | dst_prog = NULL; | |
2606 | attach_btf = btf_get_by_fd(attr->attach_btf_obj_fd); | |
2607 | if (IS_ERR(attach_btf)) | |
2608 | return -EINVAL; | |
2609 | if (!btf_is_kernel(attach_btf)) { | |
8bdd8e27 AN |
2610 | /* attaching through specifying bpf_prog's BTF |
2611 | * objects directly might be supported eventually | |
2612 | */ | |
290248a5 | 2613 | btf_put(attach_btf); |
8bdd8e27 | 2614 | return -ENOTSUPP; |
290248a5 AN |
2615 | } |
2616 | } | |
2617 | } else if (attr->attach_btf_id) { | |
2618 | /* fall back to vmlinux BTF, if BTF type ID is specified */ | |
2619 | attach_btf = bpf_get_btf_vmlinux(); | |
2620 | if (IS_ERR(attach_btf)) | |
2621 | return PTR_ERR(attach_btf); | |
2622 | if (!attach_btf) | |
2623 | return -EINVAL; | |
2624 | btf_get(attach_btf); | |
2625 | } | |
2626 | ||
aac3fc32 | 2627 | bpf_prog_load_fixup_attach_type(attr); |
ccfe29eb | 2628 | if (bpf_prog_load_check_attach(type, attr->expected_attach_type, |
290248a5 AN |
2629 | attach_btf, attr->attach_btf_id, |
2630 | dst_prog)) { | |
2631 | if (dst_prog) | |
2632 | bpf_prog_put(dst_prog); | |
2633 | if (attach_btf) | |
2634 | btf_put(attach_btf); | |
5e43f899 | 2635 | return -EINVAL; |
290248a5 | 2636 | } |
5e43f899 | 2637 | |
09756af4 AS |
2638 | /* plain bpf_prog allocation */ |
2639 | prog = bpf_prog_alloc(bpf_prog_size(attr->insn_cnt), GFP_USER); | |
290248a5 AN |
2640 | if (!prog) { |
2641 | if (dst_prog) | |
2642 | bpf_prog_put(dst_prog); | |
2643 | if (attach_btf) | |
2644 | btf_put(attach_btf); | |
09756af4 | 2645 | return -ENOMEM; |
290248a5 | 2646 | } |
09756af4 | 2647 | |
5e43f899 | 2648 | prog->expected_attach_type = attr->expected_attach_type; |
290248a5 | 2649 | prog->aux->attach_btf = attach_btf; |
ccfe29eb | 2650 | prog->aux->attach_btf_id = attr->attach_btf_id; |
290248a5 | 2651 | prog->aux->dst_prog = dst_prog; |
2b3486bc | 2652 | prog->aux->dev_bound = !!attr->prog_ifindex; |
1e6c62a8 | 2653 | prog->aux->sleepable = attr->prog_flags & BPF_F_SLEEPABLE; |
c2f2cdbe | 2654 | prog->aux->xdp_has_frags = attr->prog_flags & BPF_F_XDP_HAS_FRAGS; |
9a18eedb | 2655 | |
afdb09c7 | 2656 | err = security_bpf_prog_alloc(prog->aux); |
aaac3ba9 | 2657 | if (err) |
3ac1f01b | 2658 | goto free_prog; |
afdb09c7 | 2659 | |
3ac1f01b | 2660 | prog->aux->user = get_current_user(); |
09756af4 AS |
2661 | prog->len = attr->insn_cnt; |
2662 | ||
2663 | err = -EFAULT; | |
af2ac3e1 AS |
2664 | if (copy_from_bpfptr(prog->insns, |
2665 | make_bpfptr(attr->insns, uattr.is_kernel), | |
2666 | bpf_prog_insn_size(prog)) != 0) | |
3ac1f01b | 2667 | goto free_prog_sec; |
7f6719f7 AN |
2668 | /* copy eBPF program license from user space */ |
2669 | if (strncpy_from_bpfptr(license, | |
2670 | make_bpfptr(attr->license, uattr.is_kernel), | |
2671 | sizeof(license) - 1) < 0) | |
2672 | goto free_prog_sec; | |
2673 | license[sizeof(license) - 1] = 0; | |
2674 | ||
2675 | /* eBPF programs must be GPL compatible to use GPL-ed functions */ | |
2676 | prog->gpl_compatible = license_is_gpl_compatible(license) ? 1 : 0; | |
09756af4 AS |
2677 | |
2678 | prog->orig_prog = NULL; | |
a91263d5 | 2679 | prog->jited = 0; |
09756af4 | 2680 | |
85192dbf | 2681 | atomic64_set(&prog->aux->refcnt, 1); |
09756af4 | 2682 | |
9a18eedb | 2683 | if (bpf_prog_is_dev_bound(prog->aux)) { |
2b3486bc | 2684 | err = bpf_prog_dev_bound_init(prog, attr); |
ab3f0063 | 2685 | if (err) |
3ac1f01b | 2686 | goto free_prog_sec; |
ab3f0063 JK |
2687 | } |
2688 | ||
fd7c211d THJ |
2689 | if (type == BPF_PROG_TYPE_EXT && dst_prog && |
2690 | bpf_prog_is_dev_bound(dst_prog->aux)) { | |
2691 | err = bpf_prog_dev_bound_inherit(prog, dst_prog); | |
ab3f0063 | 2692 | if (err) |
3ac1f01b | 2693 | goto free_prog_sec; |
ab3f0063 JK |
2694 | } |
2695 | ||
09756af4 AS |
2696 | /* find program type: socket_filter vs tracing_filter */ |
2697 | err = find_prog_type(type, prog); | |
2698 | if (err < 0) | |
3ac1f01b | 2699 | goto free_prog_sec; |
09756af4 | 2700 | |
9285ec4c | 2701 | prog->aux->load_time = ktime_get_boottime_ns(); |
8e7ae251 MKL |
2702 | err = bpf_obj_name_cpy(prog->aux->name, attr->prog_name, |
2703 | sizeof(attr->prog_name)); | |
2704 | if (err < 0) | |
3ac1f01b | 2705 | goto free_prog_sec; |
cb4d2b3f | 2706 | |
09756af4 | 2707 | /* run eBPF verifier */ |
47a71c1f | 2708 | err = bpf_check(&prog, attr, uattr, uattr_size); |
09756af4 AS |
2709 | if (err < 0) |
2710 | goto free_used_maps; | |
2711 | ||
9facc336 | 2712 | prog = bpf_prog_select_runtime(prog, &err); |
04fd61ab AS |
2713 | if (err < 0) |
2714 | goto free_used_maps; | |
09756af4 | 2715 | |
dc4bb0e2 MKL |
2716 | err = bpf_prog_alloc_id(prog); |
2717 | if (err) | |
2718 | goto free_used_maps; | |
2719 | ||
c751798a DB |
2720 | /* Upon success of bpf_prog_alloc_id(), the BPF prog is |
2721 | * effectively publicly exposed. However, retrieving via | |
2722 | * bpf_prog_get_fd_by_id() will take another reference, | |
2723 | * therefore it cannot be gone underneath us. | |
2724 | * | |
2725 | * Only for the time /after/ successful bpf_prog_new_fd() | |
2726 | * and before returning to userspace, we might just hold | |
2727 | * one reference and any parallel close on that fd could | |
2728 | * rip everything out. Hence, below notifications must | |
2729 | * happen before bpf_prog_new_fd(). | |
2730 | * | |
2731 | * Also, any failure handling from this point onwards must | |
2732 | * be using bpf_prog_put() given the program is exposed. | |
2733 | */ | |
74451e66 | 2734 | bpf_prog_kallsyms_add(prog); |
6ee52e2a | 2735 | perf_event_bpf_event(prog, PERF_BPF_EVENT_PROG_LOAD, 0); |
bae141f5 | 2736 | bpf_audit_prog(prog, BPF_AUDIT_LOAD); |
c751798a DB |
2737 | |
2738 | err = bpf_prog_new_fd(prog); | |
2739 | if (err < 0) | |
2740 | bpf_prog_put(prog); | |
09756af4 AS |
2741 | return err; |
2742 | ||
2743 | free_used_maps: | |
cd7455f1 DB |
2744 | /* In case we have subprogs, we need to wait for a grace |
2745 | * period before we can tear down JIT memory since symbols | |
2746 | * are already exposed under kallsyms. | |
2747 | */ | |
2748 | __bpf_prog_put_noref(prog, prog->aux->func_cnt); | |
2749 | return err; | |
afdb09c7 | 2750 | free_prog_sec: |
3ac1f01b | 2751 | free_uid(prog->aux->user); |
afdb09c7 | 2752 | security_bpf_prog_free(prog->aux); |
3ac1f01b | 2753 | free_prog: |
22dc4a0f AN |
2754 | if (prog->aux->attach_btf) |
2755 | btf_put(prog->aux->attach_btf); | |
09756af4 AS |
2756 | bpf_prog_free(prog); |
2757 | return err; | |
2758 | } | |
2759 | ||
cb8edce2 | 2760 | #define BPF_OBJ_LAST_FIELD path_fd |
b2197755 DB |
2761 | |
2762 | static int bpf_obj_pin(const union bpf_attr *attr) | |
2763 | { | |
cb8edce2 AN |
2764 | int path_fd; |
2765 | ||
2766 | if (CHECK_ATTR(BPF_OBJ) || attr->file_flags & ~BPF_F_PATH_FD) | |
2767 | return -EINVAL; | |
2768 | ||
2769 | /* path_fd has to be accompanied by BPF_F_PATH_FD flag */ | |
2770 | if (!(attr->file_flags & BPF_F_PATH_FD) && attr->path_fd) | |
b2197755 DB |
2771 | return -EINVAL; |
2772 | ||
cb8edce2 AN |
2773 | path_fd = attr->file_flags & BPF_F_PATH_FD ? attr->path_fd : AT_FDCWD; |
2774 | return bpf_obj_pin_user(attr->bpf_fd, path_fd, | |
2775 | u64_to_user_ptr(attr->pathname)); | |
b2197755 DB |
2776 | } |
2777 | ||
2778 | static int bpf_obj_get(const union bpf_attr *attr) | |
2779 | { | |
cb8edce2 AN |
2780 | int path_fd; |
2781 | ||
6e71b04a | 2782 | if (CHECK_ATTR(BPF_OBJ) || attr->bpf_fd != 0 || |
cb8edce2 AN |
2783 | attr->file_flags & ~(BPF_OBJ_FLAG_MASK | BPF_F_PATH_FD)) |
2784 | return -EINVAL; | |
2785 | ||
2786 | /* path_fd has to be accompanied by BPF_F_PATH_FD flag */ | |
2787 | if (!(attr->file_flags & BPF_F_PATH_FD) && attr->path_fd) | |
b2197755 DB |
2788 | return -EINVAL; |
2789 | ||
cb8edce2 AN |
2790 | path_fd = attr->file_flags & BPF_F_PATH_FD ? attr->path_fd : AT_FDCWD; |
2791 | return bpf_obj_get_user(path_fd, u64_to_user_ptr(attr->pathname), | |
6e71b04a | 2792 | attr->file_flags); |
b2197755 DB |
2793 | } |
2794 | ||
f2e10bff | 2795 | void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, |
a3b80e10 | 2796 | const struct bpf_link_ops *ops, struct bpf_prog *prog) |
fec56f58 | 2797 | { |
70ed506c | 2798 | atomic64_set(&link->refcnt, 1); |
f2e10bff | 2799 | link->type = type; |
a3b80e10 | 2800 | link->id = 0; |
70ed506c AN |
2801 | link->ops = ops; |
2802 | link->prog = prog; | |
2803 | } | |
2804 | ||
a3b80e10 AN |
2805 | static void bpf_link_free_id(int id) |
2806 | { | |
2807 | if (!id) | |
2808 | return; | |
2809 | ||
2810 | spin_lock_bh(&link_idr_lock); | |
2811 | idr_remove(&link_idr, id); | |
2812 | spin_unlock_bh(&link_idr_lock); | |
2813 | } | |
2814 | ||
98868668 AN |
2815 | /* Clean up bpf_link and corresponding anon_inode file and FD. After |
2816 | * anon_inode is created, bpf_link can't be just kfree()'d due to deferred | |
89ae89f5 | 2817 | * anon_inode's release() call. This helper marks bpf_link as |
a3b80e10 AN |
2818 | * defunct, releases anon_inode file and puts reserved FD. bpf_prog's refcnt |
2819 | * is not decremented, it's the responsibility of a calling code that failed | |
2820 | * to complete bpf_link initialization. | |
89ae89f5 JO |
2821 | * This helper eventually calls link's dealloc callback, but does not call |
2822 | * link's release callback. | |
98868668 | 2823 | */ |
a3b80e10 | 2824 | void bpf_link_cleanup(struct bpf_link_primer *primer) |
babf3164 | 2825 | { |
a3b80e10 AN |
2826 | primer->link->prog = NULL; |
2827 | bpf_link_free_id(primer->id); | |
2828 | fput(primer->file); | |
2829 | put_unused_fd(primer->fd); | |
babf3164 AN |
2830 | } |
2831 | ||
70ed506c AN |
2832 | void bpf_link_inc(struct bpf_link *link) |
2833 | { | |
2834 | atomic64_inc(&link->refcnt); | |
2835 | } | |
2836 | ||
2837 | /* bpf_link_free is guaranteed to be called from process context */ | |
2838 | static void bpf_link_free(struct bpf_link *link) | |
2839 | { | |
a3b80e10 | 2840 | bpf_link_free_id(link->id); |
babf3164 AN |
2841 | if (link->prog) { |
2842 | /* detach BPF program, clean up used resources */ | |
2843 | link->ops->release(link); | |
2844 | bpf_prog_put(link->prog); | |
2845 | } | |
2846 | /* free bpf_link and its containing memory */ | |
2847 | link->ops->dealloc(link); | |
70ed506c AN |
2848 | } |
2849 | ||
2850 | static void bpf_link_put_deferred(struct work_struct *work) | |
2851 | { | |
2852 | struct bpf_link *link = container_of(work, struct bpf_link, work); | |
2853 | ||
2854 | bpf_link_free(link); | |
2855 | } | |
2856 | ||
ab5d47bd SAS |
2857 | /* bpf_link_put might be called from atomic context. It needs to be called |
2858 | * from sleepable context in order to acquire sleeping locks during the process. | |
70ed506c AN |
2859 | */ |
2860 | void bpf_link_put(struct bpf_link *link) | |
2861 | { | |
2862 | if (!atomic64_dec_and_test(&link->refcnt)) | |
2863 | return; | |
2864 | ||
ab5d47bd SAS |
2865 | INIT_WORK(&link->work, bpf_link_put_deferred); |
2866 | schedule_work(&link->work); | |
70ed506c | 2867 | } |
cb80ddc6 | 2868 | EXPORT_SYMBOL(bpf_link_put); |
70ed506c | 2869 | |
ab5d47bd SAS |
2870 | static void bpf_link_put_direct(struct bpf_link *link) |
2871 | { | |
2872 | if (!atomic64_dec_and_test(&link->refcnt)) | |
2873 | return; | |
2874 | bpf_link_free(link); | |
2875 | } | |
2876 | ||
70ed506c AN |
2877 | static int bpf_link_release(struct inode *inode, struct file *filp) |
2878 | { | |
2879 | struct bpf_link *link = filp->private_data; | |
2880 | ||
ab5d47bd | 2881 | bpf_link_put_direct(link); |
fec56f58 AS |
2882 | return 0; |
2883 | } | |
2884 | ||
70ed506c | 2885 | #ifdef CONFIG_PROC_FS |
f2e10bff AN |
2886 | #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) |
2887 | #define BPF_MAP_TYPE(_id, _ops) | |
2888 | #define BPF_LINK_TYPE(_id, _name) [_id] = #_name, | |
2889 | static const char *bpf_link_type_strs[] = { | |
2890 | [BPF_LINK_TYPE_UNSPEC] = "<invalid>", | |
2891 | #include <linux/bpf_types.h> | |
2892 | }; | |
2893 | #undef BPF_PROG_TYPE | |
2894 | #undef BPF_MAP_TYPE | |
2895 | #undef BPF_LINK_TYPE | |
70ed506c AN |
2896 | |
2897 | static void bpf_link_show_fdinfo(struct seq_file *m, struct file *filp) | |
2898 | { | |
2899 | const struct bpf_link *link = filp->private_data; | |
2900 | const struct bpf_prog *prog = link->prog; | |
2901 | char prog_tag[sizeof(prog->tag) * 2 + 1] = { }; | |
70ed506c | 2902 | |
70ed506c AN |
2903 | seq_printf(m, |
2904 | "link_type:\t%s\n" | |
68b04864 | 2905 | "link_id:\t%u\n", |
f2e10bff | 2906 | bpf_link_type_strs[link->type], |
68b04864 KFL |
2907 | link->id); |
2908 | if (prog) { | |
2909 | bin2hex(prog_tag, prog->tag, sizeof(prog->tag)); | |
2910 | seq_printf(m, | |
2911 | "prog_tag:\t%s\n" | |
2912 | "prog_id:\t%u\n", | |
2913 | prog_tag, | |
2914 | prog->aux->id); | |
2915 | } | |
f2e10bff AN |
2916 | if (link->ops->show_fdinfo) |
2917 | link->ops->show_fdinfo(link, m); | |
70ed506c AN |
2918 | } |
2919 | #endif | |
2920 | ||
6f302bfb | 2921 | static const struct file_operations bpf_link_fops = { |
70ed506c AN |
2922 | #ifdef CONFIG_PROC_FS |
2923 | .show_fdinfo = bpf_link_show_fdinfo, | |
2924 | #endif | |
2925 | .release = bpf_link_release, | |
fec56f58 AS |
2926 | .read = bpf_dummy_read, |
2927 | .write = bpf_dummy_write, | |
2928 | }; | |
2929 | ||
a3b80e10 | 2930 | static int bpf_link_alloc_id(struct bpf_link *link) |
70ed506c | 2931 | { |
a3b80e10 | 2932 | int id; |
70ed506c | 2933 | |
a3b80e10 AN |
2934 | idr_preload(GFP_KERNEL); |
2935 | spin_lock_bh(&link_idr_lock); | |
2936 | id = idr_alloc_cyclic(&link_idr, link, 1, INT_MAX, GFP_ATOMIC); | |
2937 | spin_unlock_bh(&link_idr_lock); | |
2938 | idr_preload_end(); | |
70ed506c | 2939 | |
a3b80e10 AN |
2940 | return id; |
2941 | } | |
2942 | ||
2943 | /* Prepare bpf_link to be exposed to user-space by allocating anon_inode file, | |
2944 | * reserving unused FD and allocating ID from link_idr. This is to be paired | |
2945 | * with bpf_link_settle() to install FD and ID and expose bpf_link to | |
2946 | * user-space, if bpf_link is successfully attached. If not, bpf_link and | |
2947 | * pre-allocated resources are to be freed with bpf_cleanup() call. All the | |
2948 | * transient state is passed around in struct bpf_link_primer. | |
2949 | * This is preferred way to create and initialize bpf_link, especially when | |
c561d110 | 2950 | * there are complicated and expensive operations in between creating bpf_link |
a3b80e10 AN |
2951 | * itself and attaching it to BPF hook. By using bpf_link_prime() and |
2952 | * bpf_link_settle() kernel code using bpf_link doesn't have to perform | |
2953 | * expensive (and potentially failing) roll back operations in a rare case | |
2954 | * that file, FD, or ID can't be allocated. | |
babf3164 | 2955 | */ |
a3b80e10 | 2956 | int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer) |
babf3164 AN |
2957 | { |
2958 | struct file *file; | |
a3b80e10 | 2959 | int fd, id; |
babf3164 AN |
2960 | |
2961 | fd = get_unused_fd_flags(O_CLOEXEC); | |
2962 | if (fd < 0) | |
a3b80e10 | 2963 | return fd; |
babf3164 | 2964 | |
babf3164 | 2965 | |
a3b80e10 AN |
2966 | id = bpf_link_alloc_id(link); |
2967 | if (id < 0) { | |
2968 | put_unused_fd(fd); | |
a3b80e10 AN |
2969 | return id; |
2970 | } | |
babf3164 AN |
2971 | |
2972 | file = anon_inode_getfile("bpf_link", &bpf_link_fops, link, O_CLOEXEC); | |
2973 | if (IS_ERR(file)) { | |
138c6767 | 2974 | bpf_link_free_id(id); |
babf3164 | 2975 | put_unused_fd(fd); |
138c6767 | 2976 | return PTR_ERR(file); |
babf3164 AN |
2977 | } |
2978 | ||
a3b80e10 AN |
2979 | primer->link = link; |
2980 | primer->file = file; | |
2981 | primer->fd = fd; | |
2982 | primer->id = id; | |
2983 | return 0; | |
2984 | } | |
2985 | ||
2986 | int bpf_link_settle(struct bpf_link_primer *primer) | |
2987 | { | |
2988 | /* make bpf_link fetchable by ID */ | |
2989 | spin_lock_bh(&link_idr_lock); | |
2990 | primer->link->id = primer->id; | |
2991 | spin_unlock_bh(&link_idr_lock); | |
2992 | /* make bpf_link fetchable by FD */ | |
2993 | fd_install(primer->fd, primer->file); | |
2994 | /* pass through installed FD */ | |
2995 | return primer->fd; | |
2996 | } | |
2997 | ||
2998 | int bpf_link_new_fd(struct bpf_link *link) | |
2999 | { | |
3000 | return anon_inode_getfd("bpf-link", &bpf_link_fops, link, O_CLOEXEC); | |
babf3164 AN |
3001 | } |
3002 | ||
70ed506c AN |
3003 | struct bpf_link *bpf_link_get_from_fd(u32 ufd) |
3004 | { | |
3005 | struct fd f = fdget(ufd); | |
3006 | struct bpf_link *link; | |
3007 | ||
3008 | if (!f.file) | |
3009 | return ERR_PTR(-EBADF); | |
3010 | if (f.file->f_op != &bpf_link_fops) { | |
3011 | fdput(f); | |
3012 | return ERR_PTR(-EINVAL); | |
3013 | } | |
3014 | ||
3015 | link = f.file->private_data; | |
3016 | bpf_link_inc(link); | |
3017 | fdput(f); | |
3018 | ||
3019 | return link; | |
3020 | } | |
cb80ddc6 | 3021 | EXPORT_SYMBOL(bpf_link_get_from_fd); |
70ed506c | 3022 | |
70ed506c | 3023 | static void bpf_tracing_link_release(struct bpf_link *link) |
babf3164 | 3024 | { |
3aac1ead | 3025 | struct bpf_tracing_link *tr_link = |
f7e0beaf | 3026 | container_of(link, struct bpf_tracing_link, link.link); |
3aac1ead | 3027 | |
f7e0beaf | 3028 | WARN_ON_ONCE(bpf_trampoline_unlink_prog(&tr_link->link, |
3aac1ead THJ |
3029 | tr_link->trampoline)); |
3030 | ||
3031 | bpf_trampoline_put(tr_link->trampoline); | |
3032 | ||
3033 | /* tgt_prog is NULL if target is a kernel function */ | |
3034 | if (tr_link->tgt_prog) | |
3035 | bpf_prog_put(tr_link->tgt_prog); | |
babf3164 AN |
3036 | } |
3037 | ||
3038 | static void bpf_tracing_link_dealloc(struct bpf_link *link) | |
70ed506c AN |
3039 | { |
3040 | struct bpf_tracing_link *tr_link = | |
f7e0beaf | 3041 | container_of(link, struct bpf_tracing_link, link.link); |
70ed506c | 3042 | |
70ed506c AN |
3043 | kfree(tr_link); |
3044 | } | |
3045 | ||
f2e10bff AN |
3046 | static void bpf_tracing_link_show_fdinfo(const struct bpf_link *link, |
3047 | struct seq_file *seq) | |
3048 | { | |
3049 | struct bpf_tracing_link *tr_link = | |
f7e0beaf | 3050 | container_of(link, struct bpf_tracing_link, link.link); |
e859e429 | 3051 | u32 target_btf_id, target_obj_id; |
f2e10bff | 3052 | |
e859e429 YS |
3053 | bpf_trampoline_unpack_key(tr_link->trampoline->key, |
3054 | &target_obj_id, &target_btf_id); | |
f2e10bff | 3055 | seq_printf(seq, |
e859e429 YS |
3056 | "attach_type:\t%d\n" |
3057 | "target_obj_id:\t%u\n" | |
3058 | "target_btf_id:\t%u\n", | |
3059 | tr_link->attach_type, | |
3060 | target_obj_id, | |
3061 | target_btf_id); | |
f2e10bff AN |
3062 | } |
3063 | ||
3064 | static int bpf_tracing_link_fill_link_info(const struct bpf_link *link, | |
3065 | struct bpf_link_info *info) | |
3066 | { | |
3067 | struct bpf_tracing_link *tr_link = | |
f7e0beaf | 3068 | container_of(link, struct bpf_tracing_link, link.link); |
f2e10bff AN |
3069 | |
3070 | info->tracing.attach_type = tr_link->attach_type; | |
441e8c66 THJ |
3071 | bpf_trampoline_unpack_key(tr_link->trampoline->key, |
3072 | &info->tracing.target_obj_id, | |
3073 | &info->tracing.target_btf_id); | |
f2e10bff AN |
3074 | |
3075 | return 0; | |
3076 | } | |
3077 | ||
70ed506c AN |
3078 | static const struct bpf_link_ops bpf_tracing_link_lops = { |
3079 | .release = bpf_tracing_link_release, | |
babf3164 | 3080 | .dealloc = bpf_tracing_link_dealloc, |
f2e10bff AN |
3081 | .show_fdinfo = bpf_tracing_link_show_fdinfo, |
3082 | .fill_link_info = bpf_tracing_link_fill_link_info, | |
70ed506c AN |
3083 | }; |
3084 | ||
4a1e7c0c THJ |
3085 | static int bpf_tracing_prog_attach(struct bpf_prog *prog, |
3086 | int tgt_prog_fd, | |
2fcc8241 KFL |
3087 | u32 btf_id, |
3088 | u64 bpf_cookie) | |
fec56f58 | 3089 | { |
a3b80e10 | 3090 | struct bpf_link_primer link_primer; |
3aac1ead | 3091 | struct bpf_prog *tgt_prog = NULL; |
4a1e7c0c | 3092 | struct bpf_trampoline *tr = NULL; |
70ed506c | 3093 | struct bpf_tracing_link *link; |
4a1e7c0c | 3094 | u64 key = 0; |
a3b80e10 | 3095 | int err; |
fec56f58 | 3096 | |
9e4e01df KS |
3097 | switch (prog->type) { |
3098 | case BPF_PROG_TYPE_TRACING: | |
3099 | if (prog->expected_attach_type != BPF_TRACE_FENTRY && | |
3100 | prog->expected_attach_type != BPF_TRACE_FEXIT && | |
3101 | prog->expected_attach_type != BPF_MODIFY_RETURN) { | |
3102 | err = -EINVAL; | |
3103 | goto out_put_prog; | |
3104 | } | |
3105 | break; | |
3106 | case BPF_PROG_TYPE_EXT: | |
3107 | if (prog->expected_attach_type != 0) { | |
3108 | err = -EINVAL; | |
3109 | goto out_put_prog; | |
3110 | } | |
3111 | break; | |
3112 | case BPF_PROG_TYPE_LSM: | |
3113 | if (prog->expected_attach_type != BPF_LSM_MAC) { | |
3114 | err = -EINVAL; | |
3115 | goto out_put_prog; | |
3116 | } | |
3117 | break; | |
3118 | default: | |
fec56f58 AS |
3119 | err = -EINVAL; |
3120 | goto out_put_prog; | |
3121 | } | |
3122 | ||
4a1e7c0c THJ |
3123 | if (!!tgt_prog_fd != !!btf_id) { |
3124 | err = -EINVAL; | |
3125 | goto out_put_prog; | |
3126 | } | |
3127 | ||
3128 | if (tgt_prog_fd) { | |
3129 | /* For now we only allow new targets for BPF_PROG_TYPE_EXT */ | |
3130 | if (prog->type != BPF_PROG_TYPE_EXT) { | |
3131 | err = -EINVAL; | |
3132 | goto out_put_prog; | |
3133 | } | |
3134 | ||
3135 | tgt_prog = bpf_prog_get(tgt_prog_fd); | |
3136 | if (IS_ERR(tgt_prog)) { | |
3137 | err = PTR_ERR(tgt_prog); | |
3138 | tgt_prog = NULL; | |
3139 | goto out_put_prog; | |
3140 | } | |
3141 | ||
22dc4a0f | 3142 | key = bpf_trampoline_compute_key(tgt_prog, NULL, btf_id); |
4a1e7c0c THJ |
3143 | } |
3144 | ||
70ed506c AN |
3145 | link = kzalloc(sizeof(*link), GFP_USER); |
3146 | if (!link) { | |
3147 | err = -ENOMEM; | |
3148 | goto out_put_prog; | |
3149 | } | |
f7e0beaf | 3150 | bpf_link_init(&link->link.link, BPF_LINK_TYPE_TRACING, |
f2e10bff AN |
3151 | &bpf_tracing_link_lops, prog); |
3152 | link->attach_type = prog->expected_attach_type; | |
2fcc8241 | 3153 | link->link.cookie = bpf_cookie; |
70ed506c | 3154 | |
3aac1ead THJ |
3155 | mutex_lock(&prog->aux->dst_mutex); |
3156 | ||
4a1e7c0c THJ |
3157 | /* There are a few possible cases here: |
3158 | * | |
3159 | * - if prog->aux->dst_trampoline is set, the program was just loaded | |
3160 | * and not yet attached to anything, so we can use the values stored | |
3161 | * in prog->aux | |
3162 | * | |
3163 | * - if prog->aux->dst_trampoline is NULL, the program has already been | |
3164 | * attached to a target and its initial target was cleared (below) | |
3165 | * | |
3166 | * - if tgt_prog != NULL, the caller specified tgt_prog_fd + | |
3167 | * target_btf_id using the link_create API. | |
3168 | * | |
3169 | * - if tgt_prog == NULL when this function was called using the old | |
f3a95075 JO |
3170 | * raw_tracepoint_open API, and we need a target from prog->aux |
3171 | * | |
3172 | * - if prog->aux->dst_trampoline and tgt_prog is NULL, the program | |
3173 | * was detached and is going for re-attachment. | |
4a1e7c0c THJ |
3174 | */ |
3175 | if (!prog->aux->dst_trampoline && !tgt_prog) { | |
f3a95075 JO |
3176 | /* |
3177 | * Allow re-attach for TRACING and LSM programs. If it's | |
3178 | * currently linked, bpf_trampoline_link_prog will fail. | |
3179 | * EXT programs need to specify tgt_prog_fd, so they | |
3180 | * re-attach in separate code path. | |
3181 | */ | |
3182 | if (prog->type != BPF_PROG_TYPE_TRACING && | |
3183 | prog->type != BPF_PROG_TYPE_LSM) { | |
3184 | err = -EINVAL; | |
3185 | goto out_unlock; | |
3186 | } | |
3187 | btf_id = prog->aux->attach_btf_id; | |
3188 | key = bpf_trampoline_compute_key(NULL, prog->aux->attach_btf, btf_id); | |
babf3164 | 3189 | } |
4a1e7c0c THJ |
3190 | |
3191 | if (!prog->aux->dst_trampoline || | |
3192 | (key && key != prog->aux->dst_trampoline->key)) { | |
3193 | /* If there is no saved target, or the specified target is | |
3194 | * different from the destination specified at load time, we | |
3195 | * need a new trampoline and a check for compatibility | |
3196 | */ | |
3197 | struct bpf_attach_target_info tgt_info = {}; | |
3198 | ||
3199 | err = bpf_check_attach_target(NULL, prog, tgt_prog, btf_id, | |
3200 | &tgt_info); | |
3201 | if (err) | |
3202 | goto out_unlock; | |
3203 | ||
31bf1dbc VM |
3204 | if (tgt_info.tgt_mod) { |
3205 | module_put(prog->aux->mod); | |
3206 | prog->aux->mod = tgt_info.tgt_mod; | |
3207 | } | |
3208 | ||
4a1e7c0c THJ |
3209 | tr = bpf_trampoline_get(key, &tgt_info); |
3210 | if (!tr) { | |
3211 | err = -ENOMEM; | |
3212 | goto out_unlock; | |
3213 | } | |
3214 | } else { | |
3215 | /* The caller didn't specify a target, or the target was the | |
3216 | * same as the destination supplied during program load. This | |
3217 | * means we can reuse the trampoline and reference from program | |
3218 | * load time, and there is no need to allocate a new one. This | |
3219 | * can only happen once for any program, as the saved values in | |
3220 | * prog->aux are cleared below. | |
3221 | */ | |
3222 | tr = prog->aux->dst_trampoline; | |
3223 | tgt_prog = prog->aux->dst_prog; | |
3224 | } | |
3aac1ead | 3225 | |
f7e0beaf | 3226 | err = bpf_link_prime(&link->link.link, &link_primer); |
3aac1ead THJ |
3227 | if (err) |
3228 | goto out_unlock; | |
fec56f58 | 3229 | |
f7e0beaf | 3230 | err = bpf_trampoline_link_prog(&link->link, tr); |
babf3164 | 3231 | if (err) { |
a3b80e10 | 3232 | bpf_link_cleanup(&link_primer); |
3aac1ead THJ |
3233 | link = NULL; |
3234 | goto out_unlock; | |
fec56f58 | 3235 | } |
babf3164 | 3236 | |
3aac1ead THJ |
3237 | link->tgt_prog = tgt_prog; |
3238 | link->trampoline = tr; | |
3239 | ||
4a1e7c0c THJ |
3240 | /* Always clear the trampoline and target prog from prog->aux to make |
3241 | * sure the original attach destination is not kept alive after a | |
3242 | * program is (re-)attached to another target. | |
3243 | */ | |
3244 | if (prog->aux->dst_prog && | |
3245 | (tgt_prog_fd || tr != prog->aux->dst_trampoline)) | |
3246 | /* got extra prog ref from syscall, or attaching to different prog */ | |
3247 | bpf_prog_put(prog->aux->dst_prog); | |
3248 | if (prog->aux->dst_trampoline && tr != prog->aux->dst_trampoline) | |
3249 | /* we allocated a new trampoline, so free the old one */ | |
3250 | bpf_trampoline_put(prog->aux->dst_trampoline); | |
3251 | ||
3aac1ead THJ |
3252 | prog->aux->dst_prog = NULL; |
3253 | prog->aux->dst_trampoline = NULL; | |
3254 | mutex_unlock(&prog->aux->dst_mutex); | |
3255 | ||
a3b80e10 | 3256 | return bpf_link_settle(&link_primer); |
3aac1ead | 3257 | out_unlock: |
4a1e7c0c THJ |
3258 | if (tr && tr != prog->aux->dst_trampoline) |
3259 | bpf_trampoline_put(tr); | |
3aac1ead THJ |
3260 | mutex_unlock(&prog->aux->dst_mutex); |
3261 | kfree(link); | |
fec56f58 | 3262 | out_put_prog: |
4a1e7c0c THJ |
3263 | if (tgt_prog_fd && tgt_prog) |
3264 | bpf_prog_put(tgt_prog); | |
fec56f58 AS |
3265 | return err; |
3266 | } | |
3267 | ||
70ed506c AN |
3268 | struct bpf_raw_tp_link { |
3269 | struct bpf_link link; | |
c4f6699d | 3270 | struct bpf_raw_event_map *btp; |
c4f6699d AS |
3271 | }; |
3272 | ||
70ed506c | 3273 | static void bpf_raw_tp_link_release(struct bpf_link *link) |
c4f6699d | 3274 | { |
70ed506c AN |
3275 | struct bpf_raw_tp_link *raw_tp = |
3276 | container_of(link, struct bpf_raw_tp_link, link); | |
c4f6699d | 3277 | |
70ed506c | 3278 | bpf_probe_unregister(raw_tp->btp, raw_tp->link.prog); |
a38d1107 | 3279 | bpf_put_raw_tracepoint(raw_tp->btp); |
babf3164 AN |
3280 | } |
3281 | ||
3282 | static void bpf_raw_tp_link_dealloc(struct bpf_link *link) | |
3283 | { | |
3284 | struct bpf_raw_tp_link *raw_tp = | |
3285 | container_of(link, struct bpf_raw_tp_link, link); | |
3286 | ||
c4f6699d | 3287 | kfree(raw_tp); |
c4f6699d AS |
3288 | } |
3289 | ||
f2e10bff AN |
3290 | static void bpf_raw_tp_link_show_fdinfo(const struct bpf_link *link, |
3291 | struct seq_file *seq) | |
3292 | { | |
3293 | struct bpf_raw_tp_link *raw_tp_link = | |
3294 | container_of(link, struct bpf_raw_tp_link, link); | |
3295 | ||
3296 | seq_printf(seq, | |
3297 | "tp_name:\t%s\n", | |
3298 | raw_tp_link->btp->tp->name); | |
3299 | } | |
3300 | ||
57d48537 YS |
3301 | static int bpf_copy_to_user(char __user *ubuf, const char *buf, u32 ulen, |
3302 | u32 len) | |
3303 | { | |
3304 | if (ulen >= len + 1) { | |
3305 | if (copy_to_user(ubuf, buf, len + 1)) | |
3306 | return -EFAULT; | |
3307 | } else { | |
3308 | char zero = '\0'; | |
3309 | ||
3310 | if (copy_to_user(ubuf, buf, ulen - 1)) | |
3311 | return -EFAULT; | |
3312 | if (put_user(zero, ubuf + ulen - 1)) | |
3313 | return -EFAULT; | |
3314 | return -ENOSPC; | |
3315 | } | |
3316 | ||
3317 | return 0; | |
3318 | } | |
3319 | ||
f2e10bff AN |
3320 | static int bpf_raw_tp_link_fill_link_info(const struct bpf_link *link, |
3321 | struct bpf_link_info *info) | |
3322 | { | |
3323 | struct bpf_raw_tp_link *raw_tp_link = | |
3324 | container_of(link, struct bpf_raw_tp_link, link); | |
3325 | char __user *ubuf = u64_to_user_ptr(info->raw_tracepoint.tp_name); | |
3326 | const char *tp_name = raw_tp_link->btp->tp->name; | |
3327 | u32 ulen = info->raw_tracepoint.tp_name_len; | |
3328 | size_t tp_len = strlen(tp_name); | |
3329 | ||
b474959d | 3330 | if (!ulen ^ !ubuf) |
f2e10bff AN |
3331 | return -EINVAL; |
3332 | ||
3333 | info->raw_tracepoint.tp_name_len = tp_len + 1; | |
3334 | ||
3335 | if (!ubuf) | |
3336 | return 0; | |
3337 | ||
57d48537 | 3338 | return bpf_copy_to_user(ubuf, tp_name, ulen, tp_len); |
f2e10bff AN |
3339 | } |
3340 | ||
a3b80e10 | 3341 | static const struct bpf_link_ops bpf_raw_tp_link_lops = { |
70ed506c | 3342 | .release = bpf_raw_tp_link_release, |
babf3164 | 3343 | .dealloc = bpf_raw_tp_link_dealloc, |
f2e10bff AN |
3344 | .show_fdinfo = bpf_raw_tp_link_show_fdinfo, |
3345 | .fill_link_info = bpf_raw_tp_link_fill_link_info, | |
c4f6699d AS |
3346 | }; |
3347 | ||
b89fbfbb AN |
3348 | #ifdef CONFIG_PERF_EVENTS |
3349 | struct bpf_perf_link { | |
3350 | struct bpf_link link; | |
3351 | struct file *perf_file; | |
3352 | }; | |
3353 | ||
3354 | static void bpf_perf_link_release(struct bpf_link *link) | |
3355 | { | |
3356 | struct bpf_perf_link *perf_link = container_of(link, struct bpf_perf_link, link); | |
3357 | struct perf_event *event = perf_link->perf_file->private_data; | |
3358 | ||
3359 | perf_event_free_bpf_prog(event); | |
3360 | fput(perf_link->perf_file); | |
3361 | } | |
3362 | ||
3363 | static void bpf_perf_link_dealloc(struct bpf_link *link) | |
3364 | { | |
3365 | struct bpf_perf_link *perf_link = container_of(link, struct bpf_perf_link, link); | |
3366 | ||
3367 | kfree(perf_link); | |
3368 | } | |
3369 | ||
1b715e1b YS |
3370 | static int bpf_perf_link_fill_common(const struct perf_event *event, |
3371 | char __user *uname, u32 ulen, | |
3372 | u64 *probe_offset, u64 *probe_addr, | |
3373 | u32 *fd_type) | |
3374 | { | |
3375 | const char *buf; | |
3376 | u32 prog_id; | |
3377 | size_t len; | |
3378 | int err; | |
3379 | ||
3380 | if (!ulen ^ !uname) | |
3381 | return -EINVAL; | |
1b715e1b YS |
3382 | |
3383 | err = bpf_get_perf_event_info(event, &prog_id, fd_type, &buf, | |
3384 | probe_offset, probe_addr); | |
3385 | if (err) | |
3386 | return err; | |
0aa35162 YS |
3387 | if (!uname) |
3388 | return 0; | |
1b715e1b YS |
3389 | if (buf) { |
3390 | len = strlen(buf); | |
3391 | err = bpf_copy_to_user(uname, buf, ulen, len); | |
3392 | if (err) | |
3393 | return err; | |
3394 | } else { | |
3395 | char zero = '\0'; | |
3396 | ||
3397 | if (put_user(zero, uname)) | |
3398 | return -EFAULT; | |
3399 | } | |
3400 | return 0; | |
3401 | } | |
3402 | ||
3403 | #ifdef CONFIG_KPROBE_EVENTS | |
3404 | static int bpf_perf_link_fill_kprobe(const struct perf_event *event, | |
3405 | struct bpf_link_info *info) | |
3406 | { | |
3407 | char __user *uname; | |
3408 | u64 addr, offset; | |
3409 | u32 ulen, type; | |
3410 | int err; | |
3411 | ||
3412 | uname = u64_to_user_ptr(info->perf_event.kprobe.func_name); | |
3413 | ulen = info->perf_event.kprobe.name_len; | |
3414 | err = bpf_perf_link_fill_common(event, uname, ulen, &offset, &addr, | |
3415 | &type); | |
3416 | if (err) | |
3417 | return err; | |
3418 | if (type == BPF_FD_TYPE_KRETPROBE) | |
3419 | info->perf_event.type = BPF_PERF_EVENT_KRETPROBE; | |
3420 | else | |
3421 | info->perf_event.type = BPF_PERF_EVENT_KPROBE; | |
3422 | ||
3423 | info->perf_event.kprobe.offset = offset; | |
3424 | if (!kallsyms_show_value(current_cred())) | |
3425 | addr = 0; | |
3426 | info->perf_event.kprobe.addr = addr; | |
3427 | return 0; | |
3428 | } | |
3429 | #endif | |
3430 | ||
3431 | #ifdef CONFIG_UPROBE_EVENTS | |
3432 | static int bpf_perf_link_fill_uprobe(const struct perf_event *event, | |
3433 | struct bpf_link_info *info) | |
3434 | { | |
3435 | char __user *uname; | |
3436 | u64 addr, offset; | |
3437 | u32 ulen, type; | |
3438 | int err; | |
3439 | ||
3440 | uname = u64_to_user_ptr(info->perf_event.uprobe.file_name); | |
3441 | ulen = info->perf_event.uprobe.name_len; | |
3442 | err = bpf_perf_link_fill_common(event, uname, ulen, &offset, &addr, | |
3443 | &type); | |
3444 | if (err) | |
3445 | return err; | |
3446 | ||
3447 | if (type == BPF_FD_TYPE_URETPROBE) | |
3448 | info->perf_event.type = BPF_PERF_EVENT_URETPROBE; | |
3449 | else | |
3450 | info->perf_event.type = BPF_PERF_EVENT_UPROBE; | |
3451 | info->perf_event.uprobe.offset = offset; | |
3452 | return 0; | |
3453 | } | |
3454 | #endif | |
3455 | ||
3456 | static int bpf_perf_link_fill_probe(const struct perf_event *event, | |
3457 | struct bpf_link_info *info) | |
3458 | { | |
3459 | #ifdef CONFIG_KPROBE_EVENTS | |
3460 | if (event->tp_event->flags & TRACE_EVENT_FL_KPROBE) | |
3461 | return bpf_perf_link_fill_kprobe(event, info); | |
3462 | #endif | |
3463 | #ifdef CONFIG_UPROBE_EVENTS | |
3464 | if (event->tp_event->flags & TRACE_EVENT_FL_UPROBE) | |
3465 | return bpf_perf_link_fill_uprobe(event, info); | |
3466 | #endif | |
3467 | return -EOPNOTSUPP; | |
3468 | } | |
3469 | ||
3470 | static int bpf_perf_link_fill_tracepoint(const struct perf_event *event, | |
3471 | struct bpf_link_info *info) | |
3472 | { | |
3473 | char __user *uname; | |
3474 | u32 ulen; | |
3475 | ||
3476 | uname = u64_to_user_ptr(info->perf_event.tracepoint.tp_name); | |
3477 | ulen = info->perf_event.tracepoint.name_len; | |
3478 | info->perf_event.type = BPF_PERF_EVENT_TRACEPOINT; | |
3479 | return bpf_perf_link_fill_common(event, uname, ulen, NULL, NULL, NULL); | |
3480 | } | |
3481 | ||
3482 | static int bpf_perf_link_fill_perf_event(const struct perf_event *event, | |
3483 | struct bpf_link_info *info) | |
3484 | { | |
3485 | info->perf_event.event.type = event->attr.type; | |
3486 | info->perf_event.event.config = event->attr.config; | |
3487 | info->perf_event.type = BPF_PERF_EVENT_EVENT; | |
3488 | return 0; | |
3489 | } | |
3490 | ||
3491 | static int bpf_perf_link_fill_link_info(const struct bpf_link *link, | |
3492 | struct bpf_link_info *info) | |
3493 | { | |
3494 | struct bpf_perf_link *perf_link; | |
3495 | const struct perf_event *event; | |
3496 | ||
3497 | perf_link = container_of(link, struct bpf_perf_link, link); | |
3498 | event = perf_get_event(perf_link->perf_file); | |
3499 | if (IS_ERR(event)) | |
3500 | return PTR_ERR(event); | |
3501 | ||
3502 | switch (event->prog->type) { | |
3503 | case BPF_PROG_TYPE_PERF_EVENT: | |
3504 | return bpf_perf_link_fill_perf_event(event, info); | |
3505 | case BPF_PROG_TYPE_TRACEPOINT: | |
3506 | return bpf_perf_link_fill_tracepoint(event, info); | |
3507 | case BPF_PROG_TYPE_KPROBE: | |
3508 | return bpf_perf_link_fill_probe(event, info); | |
3509 | default: | |
3510 | return -EOPNOTSUPP; | |
3511 | } | |
3512 | } | |
3513 | ||
b89fbfbb AN |
3514 | static const struct bpf_link_ops bpf_perf_link_lops = { |
3515 | .release = bpf_perf_link_release, | |
3516 | .dealloc = bpf_perf_link_dealloc, | |
1b715e1b | 3517 | .fill_link_info = bpf_perf_link_fill_link_info, |
b89fbfbb AN |
3518 | }; |
3519 | ||
3520 | static int bpf_perf_link_attach(const union bpf_attr *attr, struct bpf_prog *prog) | |
3521 | { | |
3522 | struct bpf_link_primer link_primer; | |
3523 | struct bpf_perf_link *link; | |
3524 | struct perf_event *event; | |
3525 | struct file *perf_file; | |
3526 | int err; | |
3527 | ||
3528 | if (attr->link_create.flags) | |
3529 | return -EINVAL; | |
3530 | ||
3531 | perf_file = perf_event_get(attr->link_create.target_fd); | |
3532 | if (IS_ERR(perf_file)) | |
3533 | return PTR_ERR(perf_file); | |
3534 | ||
3535 | link = kzalloc(sizeof(*link), GFP_USER); | |
3536 | if (!link) { | |
3537 | err = -ENOMEM; | |
3538 | goto out_put_file; | |
3539 | } | |
3540 | bpf_link_init(&link->link, BPF_LINK_TYPE_PERF_EVENT, &bpf_perf_link_lops, prog); | |
3541 | link->perf_file = perf_file; | |
3542 | ||
3543 | err = bpf_link_prime(&link->link, &link_primer); | |
3544 | if (err) { | |
3545 | kfree(link); | |
3546 | goto out_put_file; | |
3547 | } | |
3548 | ||
3549 | event = perf_file->private_data; | |
82e6b1ee | 3550 | err = perf_event_set_bpf_prog(event, prog, attr->link_create.perf_event.bpf_cookie); |
b89fbfbb AN |
3551 | if (err) { |
3552 | bpf_link_cleanup(&link_primer); | |
3553 | goto out_put_file; | |
3554 | } | |
3555 | /* perf_event_set_bpf_prog() doesn't take its own refcnt on prog */ | |
3556 | bpf_prog_inc(prog); | |
3557 | ||
3558 | return bpf_link_settle(&link_primer); | |
3559 | ||
3560 | out_put_file: | |
3561 | fput(perf_file); | |
3562 | return err; | |
3563 | } | |
0dcac272 JO |
3564 | #else |
3565 | static int bpf_perf_link_attach(const union bpf_attr *attr, struct bpf_prog *prog) | |
3566 | { | |
3567 | return -EOPNOTSUPP; | |
3568 | } | |
b89fbfbb AN |
3569 | #endif /* CONFIG_PERF_EVENTS */ |
3570 | ||
df86ca0d AN |
3571 | static int bpf_raw_tp_link_attach(struct bpf_prog *prog, |
3572 | const char __user *user_tp_name) | |
c4f6699d | 3573 | { |
a3b80e10 | 3574 | struct bpf_link_primer link_primer; |
babf3164 | 3575 | struct bpf_raw_tp_link *link; |
c4f6699d | 3576 | struct bpf_raw_event_map *btp; |
ac4414b5 AS |
3577 | const char *tp_name; |
3578 | char buf[128]; | |
a3b80e10 | 3579 | int err; |
c4f6699d | 3580 | |
9e4e01df KS |
3581 | switch (prog->type) { |
3582 | case BPF_PROG_TYPE_TRACING: | |
3583 | case BPF_PROG_TYPE_EXT: | |
3584 | case BPF_PROG_TYPE_LSM: | |
df86ca0d | 3585 | if (user_tp_name) |
fec56f58 AS |
3586 | /* The attach point for this category of programs |
3587 | * should be specified via btf_id during program load. | |
ac4414b5 | 3588 | */ |
df86ca0d | 3589 | return -EINVAL; |
9e4e01df KS |
3590 | if (prog->type == BPF_PROG_TYPE_TRACING && |
3591 | prog->expected_attach_type == BPF_TRACE_RAW_TP) { | |
fec56f58 | 3592 | tp_name = prog->aux->attach_func_name; |
9e4e01df KS |
3593 | break; |
3594 | } | |
2fcc8241 | 3595 | return bpf_tracing_prog_attach(prog, 0, 0, 0); |
9e4e01df KS |
3596 | case BPF_PROG_TYPE_RAW_TRACEPOINT: |
3597 | case BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE: | |
df86ca0d AN |
3598 | if (strncpy_from_user(buf, user_tp_name, sizeof(buf) - 1) < 0) |
3599 | return -EFAULT; | |
ac4414b5 AS |
3600 | buf[sizeof(buf) - 1] = 0; |
3601 | tp_name = buf; | |
9e4e01df KS |
3602 | break; |
3603 | default: | |
df86ca0d | 3604 | return -EINVAL; |
ac4414b5 | 3605 | } |
c4f6699d | 3606 | |
a38d1107 | 3607 | btp = bpf_get_raw_tracepoint(tp_name); |
df86ca0d AN |
3608 | if (!btp) |
3609 | return -ENOENT; | |
c4f6699d | 3610 | |
babf3164 AN |
3611 | link = kzalloc(sizeof(*link), GFP_USER); |
3612 | if (!link) { | |
a38d1107 MM |
3613 | err = -ENOMEM; |
3614 | goto out_put_btp; | |
3615 | } | |
f2e10bff AN |
3616 | bpf_link_init(&link->link, BPF_LINK_TYPE_RAW_TRACEPOINT, |
3617 | &bpf_raw_tp_link_lops, prog); | |
babf3164 | 3618 | link->btp = btp; |
c4f6699d | 3619 | |
a3b80e10 AN |
3620 | err = bpf_link_prime(&link->link, &link_primer); |
3621 | if (err) { | |
babf3164 | 3622 | kfree(link); |
babf3164 AN |
3623 | goto out_put_btp; |
3624 | } | |
c4f6699d | 3625 | |
babf3164 AN |
3626 | err = bpf_probe_register(link->btp, prog); |
3627 | if (err) { | |
a3b80e10 | 3628 | bpf_link_cleanup(&link_primer); |
babf3164 | 3629 | goto out_put_btp; |
c4f6699d | 3630 | } |
babf3164 | 3631 | |
a3b80e10 | 3632 | return bpf_link_settle(&link_primer); |
c4f6699d | 3633 | |
a38d1107 MM |
3634 | out_put_btp: |
3635 | bpf_put_raw_tracepoint(btp); | |
c4f6699d AS |
3636 | return err; |
3637 | } | |
3638 | ||
df86ca0d AN |
3639 | #define BPF_RAW_TRACEPOINT_OPEN_LAST_FIELD raw_tracepoint.prog_fd |
3640 | ||
3641 | static int bpf_raw_tracepoint_open(const union bpf_attr *attr) | |
3642 | { | |
3643 | struct bpf_prog *prog; | |
3644 | int fd; | |
3645 | ||
3646 | if (CHECK_ATTR(BPF_RAW_TRACEPOINT_OPEN)) | |
3647 | return -EINVAL; | |
3648 | ||
3649 | prog = bpf_prog_get(attr->raw_tracepoint.prog_fd); | |
3650 | if (IS_ERR(prog)) | |
3651 | return PTR_ERR(prog); | |
3652 | ||
3653 | fd = bpf_raw_tp_link_attach(prog, u64_to_user_ptr(attr->raw_tracepoint.name)); | |
3654 | if (fd < 0) | |
3655 | bpf_prog_put(prog); | |
3656 | return fd; | |
3657 | } | |
3658 | ||
e28784e3 AN |
3659 | static enum bpf_prog_type |
3660 | attach_type_to_prog_type(enum bpf_attach_type attach_type) | |
f4324551 | 3661 | { |
e28784e3 | 3662 | switch (attach_type) { |
f4324551 DM |
3663 | case BPF_CGROUP_INET_INGRESS: |
3664 | case BPF_CGROUP_INET_EGRESS: | |
e28784e3 | 3665 | return BPF_PROG_TYPE_CGROUP_SKB; |
61023658 | 3666 | case BPF_CGROUP_INET_SOCK_CREATE: |
f5836749 | 3667 | case BPF_CGROUP_INET_SOCK_RELEASE: |
aac3fc32 AI |
3668 | case BPF_CGROUP_INET4_POST_BIND: |
3669 | case BPF_CGROUP_INET6_POST_BIND: | |
e28784e3 | 3670 | return BPF_PROG_TYPE_CGROUP_SOCK; |
4fbac77d AI |
3671 | case BPF_CGROUP_INET4_BIND: |
3672 | case BPF_CGROUP_INET6_BIND: | |
d74bad4e AI |
3673 | case BPF_CGROUP_INET4_CONNECT: |
3674 | case BPF_CGROUP_INET6_CONNECT: | |
1b66d253 DB |
3675 | case BPF_CGROUP_INET4_GETPEERNAME: |
3676 | case BPF_CGROUP_INET6_GETPEERNAME: | |
3677 | case BPF_CGROUP_INET4_GETSOCKNAME: | |
3678 | case BPF_CGROUP_INET6_GETSOCKNAME: | |
1cedee13 AI |
3679 | case BPF_CGROUP_UDP4_SENDMSG: |
3680 | case BPF_CGROUP_UDP6_SENDMSG: | |
983695fa DB |
3681 | case BPF_CGROUP_UDP4_RECVMSG: |
3682 | case BPF_CGROUP_UDP6_RECVMSG: | |
e28784e3 | 3683 | return BPF_PROG_TYPE_CGROUP_SOCK_ADDR; |
40304b2a | 3684 | case BPF_CGROUP_SOCK_OPS: |
e28784e3 | 3685 | return BPF_PROG_TYPE_SOCK_OPS; |
ebc614f6 | 3686 | case BPF_CGROUP_DEVICE: |
e28784e3 | 3687 | return BPF_PROG_TYPE_CGROUP_DEVICE; |
4f738adb | 3688 | case BPF_SK_MSG_VERDICT: |
e28784e3 | 3689 | return BPF_PROG_TYPE_SK_MSG; |
464bc0fd JF |
3690 | case BPF_SK_SKB_STREAM_PARSER: |
3691 | case BPF_SK_SKB_STREAM_VERDICT: | |
a7ba4558 | 3692 | case BPF_SK_SKB_VERDICT: |
e28784e3 | 3693 | return BPF_PROG_TYPE_SK_SKB; |
f4364dcf | 3694 | case BPF_LIRC_MODE2: |
e28784e3 | 3695 | return BPF_PROG_TYPE_LIRC_MODE2; |
d58e468b | 3696 | case BPF_FLOW_DISSECTOR: |
e28784e3 | 3697 | return BPF_PROG_TYPE_FLOW_DISSECTOR; |
7b146ceb | 3698 | case BPF_CGROUP_SYSCTL: |
e28784e3 | 3699 | return BPF_PROG_TYPE_CGROUP_SYSCTL; |
0d01da6a SF |
3700 | case BPF_CGROUP_GETSOCKOPT: |
3701 | case BPF_CGROUP_SETSOCKOPT: | |
e28784e3 | 3702 | return BPF_PROG_TYPE_CGROUP_SOCKOPT; |
de4e05ca | 3703 | case BPF_TRACE_ITER: |
df86ca0d AN |
3704 | case BPF_TRACE_RAW_TP: |
3705 | case BPF_TRACE_FENTRY: | |
3706 | case BPF_TRACE_FEXIT: | |
3707 | case BPF_MODIFY_RETURN: | |
de4e05ca | 3708 | return BPF_PROG_TYPE_TRACING; |
df86ca0d AN |
3709 | case BPF_LSM_MAC: |
3710 | return BPF_PROG_TYPE_LSM; | |
e9ddbb77 JS |
3711 | case BPF_SK_LOOKUP: |
3712 | return BPF_PROG_TYPE_SK_LOOKUP; | |
aa8d3a71 AN |
3713 | case BPF_XDP: |
3714 | return BPF_PROG_TYPE_XDP; | |
69fd337a SF |
3715 | case BPF_LSM_CGROUP: |
3716 | return BPF_PROG_TYPE_LSM; | |
e420bed0 DB |
3717 | case BPF_TCX_INGRESS: |
3718 | case BPF_TCX_EGRESS: | |
3719 | return BPF_PROG_TYPE_SCHED_CLS; | |
f4324551 | 3720 | default: |
e28784e3 | 3721 | return BPF_PROG_TYPE_UNSPEC; |
f4324551 | 3722 | } |
e28784e3 AN |
3723 | } |
3724 | ||
3505cb9f JO |
3725 | static int bpf_prog_attach_check_attach_type(const struct bpf_prog *prog, |
3726 | enum bpf_attach_type attach_type) | |
3727 | { | |
3728 | enum bpf_prog_type ptype; | |
3729 | ||
3730 | switch (prog->type) { | |
3731 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
3732 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: | |
3733 | case BPF_PROG_TYPE_CGROUP_SOCKOPT: | |
3734 | case BPF_PROG_TYPE_SK_LOOKUP: | |
3735 | return attach_type == prog->expected_attach_type ? 0 : -EINVAL; | |
3736 | case BPF_PROG_TYPE_CGROUP_SKB: | |
3737 | if (!capable(CAP_NET_ADMIN)) | |
3738 | /* cg-skb progs can be loaded by unpriv user. | |
3739 | * check permissions at attach time. | |
3740 | */ | |
3741 | return -EPERM; | |
3742 | return prog->enforce_expected_attach_type && | |
3743 | prog->expected_attach_type != attach_type ? | |
3744 | -EINVAL : 0; | |
3745 | case BPF_PROG_TYPE_EXT: | |
3746 | return 0; | |
3747 | case BPF_PROG_TYPE_NETFILTER: | |
3748 | if (attach_type != BPF_NETFILTER) | |
3749 | return -EINVAL; | |
3750 | return 0; | |
3751 | case BPF_PROG_TYPE_PERF_EVENT: | |
3752 | case BPF_PROG_TYPE_TRACEPOINT: | |
3753 | if (attach_type != BPF_PERF_EVENT) | |
3754 | return -EINVAL; | |
3755 | return 0; | |
3756 | case BPF_PROG_TYPE_KPROBE: | |
3757 | if (prog->expected_attach_type == BPF_TRACE_KPROBE_MULTI && | |
3758 | attach_type != BPF_TRACE_KPROBE_MULTI) | |
3759 | return -EINVAL; | |
89ae89f5 JO |
3760 | if (prog->expected_attach_type == BPF_TRACE_UPROBE_MULTI && |
3761 | attach_type != BPF_TRACE_UPROBE_MULTI) | |
3762 | return -EINVAL; | |
3505cb9f | 3763 | if (attach_type != BPF_PERF_EVENT && |
89ae89f5 JO |
3764 | attach_type != BPF_TRACE_KPROBE_MULTI && |
3765 | attach_type != BPF_TRACE_UPROBE_MULTI) | |
3505cb9f JO |
3766 | return -EINVAL; |
3767 | return 0; | |
3768 | case BPF_PROG_TYPE_SCHED_CLS: | |
3769 | if (attach_type != BPF_TCX_INGRESS && | |
3770 | attach_type != BPF_TCX_EGRESS) | |
3771 | return -EINVAL; | |
3772 | return 0; | |
3773 | default: | |
3774 | ptype = attach_type_to_prog_type(attach_type); | |
3775 | if (ptype == BPF_PROG_TYPE_UNSPEC || ptype != prog->type) | |
3776 | return -EINVAL; | |
3777 | return 0; | |
3778 | } | |
3779 | } | |
3780 | ||
e420bed0 DB |
3781 | #define BPF_PROG_ATTACH_LAST_FIELD expected_revision |
3782 | ||
3783 | #define BPF_F_ATTACH_MASK_BASE \ | |
3784 | (BPF_F_ALLOW_OVERRIDE | \ | |
3785 | BPF_F_ALLOW_MULTI | \ | |
3786 | BPF_F_REPLACE) | |
e28784e3 | 3787 | |
e420bed0 DB |
3788 | #define BPF_F_ATTACH_MASK_MPROG \ |
3789 | (BPF_F_REPLACE | \ | |
3790 | BPF_F_BEFORE | \ | |
3791 | BPF_F_AFTER | \ | |
3792 | BPF_F_ID | \ | |
3793 | BPF_F_LINK) | |
e28784e3 AN |
3794 | |
3795 | static int bpf_prog_attach(const union bpf_attr *attr) | |
3796 | { | |
3797 | enum bpf_prog_type ptype; | |
3798 | struct bpf_prog *prog; | |
3799 | int ret; | |
3800 | ||
e28784e3 AN |
3801 | if (CHECK_ATTR(BPF_PROG_ATTACH)) |
3802 | return -EINVAL; | |
3803 | ||
e28784e3 AN |
3804 | ptype = attach_type_to_prog_type(attr->attach_type); |
3805 | if (ptype == BPF_PROG_TYPE_UNSPEC) | |
3806 | return -EINVAL; | |
ba62d611 LB |
3807 | if (bpf_mprog_supported(ptype)) { |
3808 | if (attr->attach_flags & ~BPF_F_ATTACH_MASK_MPROG) | |
3809 | return -EINVAL; | |
3810 | } else { | |
3811 | if (attr->attach_flags & ~BPF_F_ATTACH_MASK_BASE) | |
3812 | return -EINVAL; | |
3813 | if (attr->relative_fd || | |
3814 | attr->expected_revision) | |
3815 | return -EINVAL; | |
3816 | } | |
f4324551 | 3817 | |
b2cd1257 DA |
3818 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); |
3819 | if (IS_ERR(prog)) | |
3820 | return PTR_ERR(prog); | |
3821 | ||
5e43f899 AI |
3822 | if (bpf_prog_attach_check_attach_type(prog, attr->attach_type)) { |
3823 | bpf_prog_put(prog); | |
3824 | return -EINVAL; | |
3825 | } | |
3826 | ||
fdb5c453 SY |
3827 | switch (ptype) { |
3828 | case BPF_PROG_TYPE_SK_SKB: | |
3829 | case BPF_PROG_TYPE_SK_MSG: | |
604326b4 | 3830 | ret = sock_map_get_from_fd(attr, prog); |
fdb5c453 SY |
3831 | break; |
3832 | case BPF_PROG_TYPE_LIRC_MODE2: | |
3833 | ret = lirc_prog_attach(attr, prog); | |
3834 | break; | |
d58e468b | 3835 | case BPF_PROG_TYPE_FLOW_DISSECTOR: |
a3fd7cee | 3836 | ret = netns_bpf_prog_attach(attr, prog); |
d58e468b | 3837 | break; |
e28784e3 AN |
3838 | case BPF_PROG_TYPE_CGROUP_DEVICE: |
3839 | case BPF_PROG_TYPE_CGROUP_SKB: | |
3840 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
3841 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: | |
3842 | case BPF_PROG_TYPE_CGROUP_SOCKOPT: | |
3843 | case BPF_PROG_TYPE_CGROUP_SYSCTL: | |
3844 | case BPF_PROG_TYPE_SOCK_OPS: | |
69fd337a SF |
3845 | case BPF_PROG_TYPE_LSM: |
3846 | if (ptype == BPF_PROG_TYPE_LSM && | |
3847 | prog->expected_attach_type != BPF_LSM_CGROUP) | |
e89f3edf ML |
3848 | ret = -EINVAL; |
3849 | else | |
3850 | ret = cgroup_bpf_prog_attach(attr, ptype, prog); | |
e28784e3 | 3851 | break; |
e420bed0 DB |
3852 | case BPF_PROG_TYPE_SCHED_CLS: |
3853 | ret = tcx_prog_attach(attr, prog); | |
3854 | break; | |
e28784e3 AN |
3855 | default: |
3856 | ret = -EINVAL; | |
b2cd1257 DA |
3857 | } |
3858 | ||
7f677633 AS |
3859 | if (ret) |
3860 | bpf_prog_put(prog); | |
7f677633 | 3861 | return ret; |
f4324551 DM |
3862 | } |
3863 | ||
e420bed0 | 3864 | #define BPF_PROG_DETACH_LAST_FIELD expected_revision |
f4324551 DM |
3865 | |
3866 | static int bpf_prog_detach(const union bpf_attr *attr) | |
3867 | { | |
e420bed0 | 3868 | struct bpf_prog *prog = NULL; |
324bda9e | 3869 | enum bpf_prog_type ptype; |
e420bed0 | 3870 | int ret; |
f4324551 | 3871 | |
f4324551 DM |
3872 | if (CHECK_ATTR(BPF_PROG_DETACH)) |
3873 | return -EINVAL; | |
3874 | ||
e28784e3 | 3875 | ptype = attach_type_to_prog_type(attr->attach_type); |
e420bed0 DB |
3876 | if (bpf_mprog_supported(ptype)) { |
3877 | if (ptype == BPF_PROG_TYPE_UNSPEC) | |
3878 | return -EINVAL; | |
3879 | if (attr->attach_flags & ~BPF_F_ATTACH_MASK_MPROG) | |
3880 | return -EINVAL; | |
3881 | if (attr->attach_bpf_fd) { | |
3882 | prog = bpf_prog_get_type(attr->attach_bpf_fd, ptype); | |
3883 | if (IS_ERR(prog)) | |
3884 | return PTR_ERR(prog); | |
3885 | } | |
ba62d611 LB |
3886 | } else if (attr->attach_flags || |
3887 | attr->relative_fd || | |
3888 | attr->expected_revision) { | |
3889 | return -EINVAL; | |
e420bed0 | 3890 | } |
e28784e3 AN |
3891 | |
3892 | switch (ptype) { | |
3893 | case BPF_PROG_TYPE_SK_MSG: | |
3894 | case BPF_PROG_TYPE_SK_SKB: | |
e420bed0 DB |
3895 | ret = sock_map_prog_detach(attr, ptype); |
3896 | break; | |
e28784e3 | 3897 | case BPF_PROG_TYPE_LIRC_MODE2: |
e420bed0 DB |
3898 | ret = lirc_prog_detach(attr); |
3899 | break; | |
e28784e3 | 3900 | case BPF_PROG_TYPE_FLOW_DISSECTOR: |
e420bed0 DB |
3901 | ret = netns_bpf_prog_detach(attr, ptype); |
3902 | break; | |
e28784e3 AN |
3903 | case BPF_PROG_TYPE_CGROUP_DEVICE: |
3904 | case BPF_PROG_TYPE_CGROUP_SKB: | |
3905 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
3906 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: | |
3907 | case BPF_PROG_TYPE_CGROUP_SOCKOPT: | |
3908 | case BPF_PROG_TYPE_CGROUP_SYSCTL: | |
3909 | case BPF_PROG_TYPE_SOCK_OPS: | |
69fd337a | 3910 | case BPF_PROG_TYPE_LSM: |
e420bed0 DB |
3911 | ret = cgroup_bpf_prog_detach(attr, ptype); |
3912 | break; | |
3913 | case BPF_PROG_TYPE_SCHED_CLS: | |
3914 | ret = tcx_prog_detach(attr, prog); | |
3915 | break; | |
f4324551 | 3916 | default: |
e420bed0 | 3917 | ret = -EINVAL; |
f4324551 | 3918 | } |
e420bed0 DB |
3919 | |
3920 | if (prog) | |
3921 | bpf_prog_put(prog); | |
3922 | return ret; | |
f4324551 | 3923 | } |
40304b2a | 3924 | |
a4fe7838 | 3925 | #define BPF_PROG_QUERY_LAST_FIELD query.revision |
468e2f64 AS |
3926 | |
3927 | static int bpf_prog_query(const union bpf_attr *attr, | |
3928 | union bpf_attr __user *uattr) | |
3929 | { | |
468e2f64 AS |
3930 | if (!capable(CAP_NET_ADMIN)) |
3931 | return -EPERM; | |
3932 | if (CHECK_ATTR(BPF_PROG_QUERY)) | |
3933 | return -EINVAL; | |
3934 | if (attr->query.query_flags & ~BPF_F_QUERY_EFFECTIVE) | |
3935 | return -EINVAL; | |
3936 | ||
3937 | switch (attr->query.attach_type) { | |
3938 | case BPF_CGROUP_INET_INGRESS: | |
3939 | case BPF_CGROUP_INET_EGRESS: | |
3940 | case BPF_CGROUP_INET_SOCK_CREATE: | |
f5836749 | 3941 | case BPF_CGROUP_INET_SOCK_RELEASE: |
4fbac77d AI |
3942 | case BPF_CGROUP_INET4_BIND: |
3943 | case BPF_CGROUP_INET6_BIND: | |
aac3fc32 AI |
3944 | case BPF_CGROUP_INET4_POST_BIND: |
3945 | case BPF_CGROUP_INET6_POST_BIND: | |
d74bad4e AI |
3946 | case BPF_CGROUP_INET4_CONNECT: |
3947 | case BPF_CGROUP_INET6_CONNECT: | |
1b66d253 DB |
3948 | case BPF_CGROUP_INET4_GETPEERNAME: |
3949 | case BPF_CGROUP_INET6_GETPEERNAME: | |
3950 | case BPF_CGROUP_INET4_GETSOCKNAME: | |
3951 | case BPF_CGROUP_INET6_GETSOCKNAME: | |
1cedee13 AI |
3952 | case BPF_CGROUP_UDP4_SENDMSG: |
3953 | case BPF_CGROUP_UDP6_SENDMSG: | |
983695fa DB |
3954 | case BPF_CGROUP_UDP4_RECVMSG: |
3955 | case BPF_CGROUP_UDP6_RECVMSG: | |
468e2f64 | 3956 | case BPF_CGROUP_SOCK_OPS: |
ebc614f6 | 3957 | case BPF_CGROUP_DEVICE: |
7b146ceb | 3958 | case BPF_CGROUP_SYSCTL: |
0d01da6a SF |
3959 | case BPF_CGROUP_GETSOCKOPT: |
3960 | case BPF_CGROUP_SETSOCKOPT: | |
b79c9fc9 | 3961 | case BPF_LSM_CGROUP: |
e28784e3 | 3962 | return cgroup_bpf_prog_query(attr, uattr); |
f4364dcf SY |
3963 | case BPF_LIRC_MODE2: |
3964 | return lirc_prog_query(attr, uattr); | |
118c8e9a | 3965 | case BPF_FLOW_DISSECTOR: |
e9ddbb77 | 3966 | case BPF_SK_LOOKUP: |
a3fd7cee | 3967 | return netns_bpf_prog_query(attr, uattr); |
748cd572 DZ |
3968 | case BPF_SK_SKB_STREAM_PARSER: |
3969 | case BPF_SK_SKB_STREAM_VERDICT: | |
3970 | case BPF_SK_MSG_VERDICT: | |
3971 | case BPF_SK_SKB_VERDICT: | |
3972 | return sock_map_bpf_prog_query(attr, uattr); | |
e420bed0 DB |
3973 | case BPF_TCX_INGRESS: |
3974 | case BPF_TCX_EGRESS: | |
3975 | return tcx_prog_query(attr, uattr); | |
468e2f64 AS |
3976 | default: |
3977 | return -EINVAL; | |
3978 | } | |
468e2f64 | 3979 | } |
f4324551 | 3980 | |
b530e9e1 | 3981 | #define BPF_PROG_TEST_RUN_LAST_FIELD test.batch_size |
1cf1cae9 AS |
3982 | |
3983 | static int bpf_prog_test_run(const union bpf_attr *attr, | |
3984 | union bpf_attr __user *uattr) | |
3985 | { | |
3986 | struct bpf_prog *prog; | |
3987 | int ret = -ENOTSUPP; | |
3988 | ||
3989 | if (CHECK_ATTR(BPF_PROG_TEST_RUN)) | |
3990 | return -EINVAL; | |
3991 | ||
b0b9395d SF |
3992 | if ((attr->test.ctx_size_in && !attr->test.ctx_in) || |
3993 | (!attr->test.ctx_size_in && attr->test.ctx_in)) | |
3994 | return -EINVAL; | |
3995 | ||
3996 | if ((attr->test.ctx_size_out && !attr->test.ctx_out) || | |
3997 | (!attr->test.ctx_size_out && attr->test.ctx_out)) | |
3998 | return -EINVAL; | |
3999 | ||
1cf1cae9 AS |
4000 | prog = bpf_prog_get(attr->test.prog_fd); |
4001 | if (IS_ERR(prog)) | |
4002 | return PTR_ERR(prog); | |
4003 | ||
4004 | if (prog->aux->ops->test_run) | |
4005 | ret = prog->aux->ops->test_run(prog, attr, uattr); | |
4006 | ||
4007 | bpf_prog_put(prog); | |
4008 | return ret; | |
4009 | } | |
4010 | ||
34ad5580 MKL |
4011 | #define BPF_OBJ_GET_NEXT_ID_LAST_FIELD next_id |
4012 | ||
4013 | static int bpf_obj_get_next_id(const union bpf_attr *attr, | |
4014 | union bpf_attr __user *uattr, | |
4015 | struct idr *idr, | |
4016 | spinlock_t *lock) | |
4017 | { | |
4018 | u32 next_id = attr->start_id; | |
4019 | int err = 0; | |
4020 | ||
4021 | if (CHECK_ATTR(BPF_OBJ_GET_NEXT_ID) || next_id >= INT_MAX) | |
4022 | return -EINVAL; | |
4023 | ||
4024 | if (!capable(CAP_SYS_ADMIN)) | |
4025 | return -EPERM; | |
4026 | ||
4027 | next_id++; | |
4028 | spin_lock_bh(lock); | |
4029 | if (!idr_get_next(idr, &next_id)) | |
4030 | err = -ENOENT; | |
4031 | spin_unlock_bh(lock); | |
4032 | ||
4033 | if (!err) | |
4034 | err = put_user(next_id, &uattr->next_id); | |
4035 | ||
4036 | return err; | |
4037 | } | |
4038 | ||
6086d29d YS |
4039 | struct bpf_map *bpf_map_get_curr_or_next(u32 *id) |
4040 | { | |
4041 | struct bpf_map *map; | |
4042 | ||
4043 | spin_lock_bh(&map_idr_lock); | |
4044 | again: | |
4045 | map = idr_get_next(&map_idr, id); | |
4046 | if (map) { | |
4047 | map = __bpf_map_inc_not_zero(map, false); | |
4048 | if (IS_ERR(map)) { | |
4049 | (*id)++; | |
4050 | goto again; | |
4051 | } | |
4052 | } | |
4053 | spin_unlock_bh(&map_idr_lock); | |
4054 | ||
4055 | return map; | |
4056 | } | |
4057 | ||
a228a64f AS |
4058 | struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id) |
4059 | { | |
4060 | struct bpf_prog *prog; | |
4061 | ||
4062 | spin_lock_bh(&prog_idr_lock); | |
4063 | again: | |
4064 | prog = idr_get_next(&prog_idr, id); | |
4065 | if (prog) { | |
4066 | prog = bpf_prog_inc_not_zero(prog); | |
4067 | if (IS_ERR(prog)) { | |
4068 | (*id)++; | |
4069 | goto again; | |
4070 | } | |
4071 | } | |
4072 | spin_unlock_bh(&prog_idr_lock); | |
4073 | ||
4074 | return prog; | |
4075 | } | |
4076 | ||
b16d9aa4 MKL |
4077 | #define BPF_PROG_GET_FD_BY_ID_LAST_FIELD prog_id |
4078 | ||
7e6897f9 | 4079 | struct bpf_prog *bpf_prog_by_id(u32 id) |
b16d9aa4 MKL |
4080 | { |
4081 | struct bpf_prog *prog; | |
b16d9aa4 | 4082 | |
7e6897f9 BT |
4083 | if (!id) |
4084 | return ERR_PTR(-ENOENT); | |
b16d9aa4 MKL |
4085 | |
4086 | spin_lock_bh(&prog_idr_lock); | |
4087 | prog = idr_find(&prog_idr, id); | |
4088 | if (prog) | |
4089 | prog = bpf_prog_inc_not_zero(prog); | |
4090 | else | |
4091 | prog = ERR_PTR(-ENOENT); | |
4092 | spin_unlock_bh(&prog_idr_lock); | |
7e6897f9 BT |
4093 | return prog; |
4094 | } | |
4095 | ||
4096 | static int bpf_prog_get_fd_by_id(const union bpf_attr *attr) | |
4097 | { | |
4098 | struct bpf_prog *prog; | |
4099 | u32 id = attr->prog_id; | |
4100 | int fd; | |
4101 | ||
4102 | if (CHECK_ATTR(BPF_PROG_GET_FD_BY_ID)) | |
4103 | return -EINVAL; | |
4104 | ||
4105 | if (!capable(CAP_SYS_ADMIN)) | |
4106 | return -EPERM; | |
b16d9aa4 | 4107 | |
7e6897f9 | 4108 | prog = bpf_prog_by_id(id); |
b16d9aa4 MKL |
4109 | if (IS_ERR(prog)) |
4110 | return PTR_ERR(prog); | |
4111 | ||
4112 | fd = bpf_prog_new_fd(prog); | |
4113 | if (fd < 0) | |
4114 | bpf_prog_put(prog); | |
4115 | ||
4116 | return fd; | |
4117 | } | |
4118 | ||
6e71b04a | 4119 | #define BPF_MAP_GET_FD_BY_ID_LAST_FIELD open_flags |
bd5f5f4e MKL |
4120 | |
4121 | static int bpf_map_get_fd_by_id(const union bpf_attr *attr) | |
4122 | { | |
4123 | struct bpf_map *map; | |
4124 | u32 id = attr->map_id; | |
6e71b04a | 4125 | int f_flags; |
bd5f5f4e MKL |
4126 | int fd; |
4127 | ||
6e71b04a CF |
4128 | if (CHECK_ATTR(BPF_MAP_GET_FD_BY_ID) || |
4129 | attr->open_flags & ~BPF_OBJ_FLAG_MASK) | |
bd5f5f4e MKL |
4130 | return -EINVAL; |
4131 | ||
4132 | if (!capable(CAP_SYS_ADMIN)) | |
4133 | return -EPERM; | |
4134 | ||
6e71b04a CF |
4135 | f_flags = bpf_get_file_flag(attr->open_flags); |
4136 | if (f_flags < 0) | |
4137 | return f_flags; | |
4138 | ||
bd5f5f4e MKL |
4139 | spin_lock_bh(&map_idr_lock); |
4140 | map = idr_find(&map_idr, id); | |
4141 | if (map) | |
b0e4701c | 4142 | map = __bpf_map_inc_not_zero(map, true); |
bd5f5f4e MKL |
4143 | else |
4144 | map = ERR_PTR(-ENOENT); | |
4145 | spin_unlock_bh(&map_idr_lock); | |
4146 | ||
4147 | if (IS_ERR(map)) | |
4148 | return PTR_ERR(map); | |
4149 | ||
6e71b04a | 4150 | fd = bpf_map_new_fd(map, f_flags); |
bd5f5f4e | 4151 | if (fd < 0) |
781e6282 | 4152 | bpf_map_put_with_uref(map); |
bd5f5f4e MKL |
4153 | |
4154 | return fd; | |
4155 | } | |
4156 | ||
7105e828 | 4157 | static const struct bpf_map *bpf_map_from_imm(const struct bpf_prog *prog, |
d8eca5bb DB |
4158 | unsigned long addr, u32 *off, |
4159 | u32 *type) | |
7105e828 | 4160 | { |
d8eca5bb | 4161 | const struct bpf_map *map; |
7105e828 DB |
4162 | int i; |
4163 | ||
984fe94f | 4164 | mutex_lock(&prog->aux->used_maps_mutex); |
d8eca5bb DB |
4165 | for (i = 0, *off = 0; i < prog->aux->used_map_cnt; i++) { |
4166 | map = prog->aux->used_maps[i]; | |
4167 | if (map == (void *)addr) { | |
4168 | *type = BPF_PSEUDO_MAP_FD; | |
984fe94f | 4169 | goto out; |
d8eca5bb DB |
4170 | } |
4171 | if (!map->ops->map_direct_value_meta) | |
4172 | continue; | |
4173 | if (!map->ops->map_direct_value_meta(map, addr, off)) { | |
4174 | *type = BPF_PSEUDO_MAP_VALUE; | |
984fe94f | 4175 | goto out; |
d8eca5bb DB |
4176 | } |
4177 | } | |
984fe94f | 4178 | map = NULL; |
d8eca5bb | 4179 | |
984fe94f YZ |
4180 | out: |
4181 | mutex_unlock(&prog->aux->used_maps_mutex); | |
4182 | return map; | |
7105e828 DB |
4183 | } |
4184 | ||
63960260 KC |
4185 | static struct bpf_insn *bpf_insn_prepare_dump(const struct bpf_prog *prog, |
4186 | const struct cred *f_cred) | |
7105e828 DB |
4187 | { |
4188 | const struct bpf_map *map; | |
4189 | struct bpf_insn *insns; | |
d8eca5bb | 4190 | u32 off, type; |
7105e828 | 4191 | u64 imm; |
29fcb05b | 4192 | u8 code; |
7105e828 DB |
4193 | int i; |
4194 | ||
4195 | insns = kmemdup(prog->insnsi, bpf_prog_insn_size(prog), | |
4196 | GFP_USER); | |
4197 | if (!insns) | |
4198 | return insns; | |
4199 | ||
4200 | for (i = 0; i < prog->len; i++) { | |
29fcb05b AN |
4201 | code = insns[i].code; |
4202 | ||
4203 | if (code == (BPF_JMP | BPF_TAIL_CALL)) { | |
7105e828 DB |
4204 | insns[i].code = BPF_JMP | BPF_CALL; |
4205 | insns[i].imm = BPF_FUNC_tail_call; | |
4206 | /* fall-through */ | |
4207 | } | |
29fcb05b AN |
4208 | if (code == (BPF_JMP | BPF_CALL) || |
4209 | code == (BPF_JMP | BPF_CALL_ARGS)) { | |
4210 | if (code == (BPF_JMP | BPF_CALL_ARGS)) | |
7105e828 | 4211 | insns[i].code = BPF_JMP | BPF_CALL; |
63960260 | 4212 | if (!bpf_dump_raw_ok(f_cred)) |
7105e828 DB |
4213 | insns[i].imm = 0; |
4214 | continue; | |
4215 | } | |
29fcb05b AN |
4216 | if (BPF_CLASS(code) == BPF_LDX && BPF_MODE(code) == BPF_PROBE_MEM) { |
4217 | insns[i].code = BPF_LDX | BPF_SIZE(code) | BPF_MEM; | |
4218 | continue; | |
4219 | } | |
7105e828 | 4220 | |
29fcb05b | 4221 | if (code != (BPF_LD | BPF_IMM | BPF_DW)) |
7105e828 DB |
4222 | continue; |
4223 | ||
4224 | imm = ((u64)insns[i + 1].imm << 32) | (u32)insns[i].imm; | |
d8eca5bb | 4225 | map = bpf_map_from_imm(prog, imm, &off, &type); |
7105e828 | 4226 | if (map) { |
d8eca5bb | 4227 | insns[i].src_reg = type; |
7105e828 | 4228 | insns[i].imm = map->id; |
d8eca5bb | 4229 | insns[i + 1].imm = off; |
7105e828 DB |
4230 | continue; |
4231 | } | |
7105e828 DB |
4232 | } |
4233 | ||
4234 | return insns; | |
4235 | } | |
4236 | ||
c454a46b MKL |
4237 | static int set_info_rec_size(struct bpf_prog_info *info) |
4238 | { | |
4239 | /* | |
4240 | * Ensure info.*_rec_size is the same as kernel expected size | |
4241 | * | |
4242 | * or | |
4243 | * | |
4244 | * Only allow zero *_rec_size if both _rec_size and _cnt are | |
4245 | * zero. In this case, the kernel will set the expected | |
4246 | * _rec_size back to the info. | |
4247 | */ | |
4248 | ||
11d8b82d | 4249 | if ((info->nr_func_info || info->func_info_rec_size) && |
c454a46b MKL |
4250 | info->func_info_rec_size != sizeof(struct bpf_func_info)) |
4251 | return -EINVAL; | |
4252 | ||
11d8b82d | 4253 | if ((info->nr_line_info || info->line_info_rec_size) && |
c454a46b MKL |
4254 | info->line_info_rec_size != sizeof(struct bpf_line_info)) |
4255 | return -EINVAL; | |
4256 | ||
11d8b82d | 4257 | if ((info->nr_jited_line_info || info->jited_line_info_rec_size) && |
c454a46b MKL |
4258 | info->jited_line_info_rec_size != sizeof(__u64)) |
4259 | return -EINVAL; | |
4260 | ||
4261 | info->func_info_rec_size = sizeof(struct bpf_func_info); | |
4262 | info->line_info_rec_size = sizeof(struct bpf_line_info); | |
4263 | info->jited_line_info_rec_size = sizeof(__u64); | |
4264 | ||
4265 | return 0; | |
4266 | } | |
4267 | ||
63960260 KC |
4268 | static int bpf_prog_get_info_by_fd(struct file *file, |
4269 | struct bpf_prog *prog, | |
1e270976 MKL |
4270 | const union bpf_attr *attr, |
4271 | union bpf_attr __user *uattr) | |
4272 | { | |
4273 | struct bpf_prog_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
6644aabb | 4274 | struct btf *attach_btf = bpf_prog_get_target_btf(prog); |
5c6f2588 | 4275 | struct bpf_prog_info info; |
1e270976 | 4276 | u32 info_len = attr->info.info_len; |
61a0abae | 4277 | struct bpf_prog_kstats stats; |
1e270976 MKL |
4278 | char __user *uinsns; |
4279 | u32 ulen; | |
4280 | int err; | |
4281 | ||
af2ac3e1 | 4282 | err = bpf_check_uarg_tail_zero(USER_BPFPTR(uinfo), sizeof(info), info_len); |
1e270976 MKL |
4283 | if (err) |
4284 | return err; | |
4285 | info_len = min_t(u32, sizeof(info), info_len); | |
4286 | ||
5c6f2588 | 4287 | memset(&info, 0, sizeof(info)); |
1e270976 | 4288 | if (copy_from_user(&info, uinfo, info_len)) |
89b09689 | 4289 | return -EFAULT; |
1e270976 MKL |
4290 | |
4291 | info.type = prog->type; | |
4292 | info.id = prog->aux->id; | |
cb4d2b3f MKL |
4293 | info.load_time = prog->aux->load_time; |
4294 | info.created_by_uid = from_kuid_munged(current_user_ns(), | |
4295 | prog->aux->user->uid); | |
b85fab0e | 4296 | info.gpl_compatible = prog->gpl_compatible; |
1e270976 MKL |
4297 | |
4298 | memcpy(info.tag, prog->tag, sizeof(prog->tag)); | |
cb4d2b3f MKL |
4299 | memcpy(info.name, prog->aux->name, sizeof(prog->aux->name)); |
4300 | ||
984fe94f | 4301 | mutex_lock(&prog->aux->used_maps_mutex); |
cb4d2b3f MKL |
4302 | ulen = info.nr_map_ids; |
4303 | info.nr_map_ids = prog->aux->used_map_cnt; | |
4304 | ulen = min_t(u32, info.nr_map_ids, ulen); | |
4305 | if (ulen) { | |
721e08da | 4306 | u32 __user *user_map_ids = u64_to_user_ptr(info.map_ids); |
cb4d2b3f MKL |
4307 | u32 i; |
4308 | ||
4309 | for (i = 0; i < ulen; i++) | |
4310 | if (put_user(prog->aux->used_maps[i]->id, | |
984fe94f YZ |
4311 | &user_map_ids[i])) { |
4312 | mutex_unlock(&prog->aux->used_maps_mutex); | |
cb4d2b3f | 4313 | return -EFAULT; |
984fe94f | 4314 | } |
cb4d2b3f | 4315 | } |
984fe94f | 4316 | mutex_unlock(&prog->aux->used_maps_mutex); |
1e270976 | 4317 | |
c454a46b MKL |
4318 | err = set_info_rec_size(&info); |
4319 | if (err) | |
4320 | return err; | |
7337224f | 4321 | |
5f8f8b93 AS |
4322 | bpf_prog_get_stats(prog, &stats); |
4323 | info.run_time_ns = stats.nsecs; | |
4324 | info.run_cnt = stats.cnt; | |
9ed9e9ba | 4325 | info.recursion_misses = stats.misses; |
5f8f8b93 | 4326 | |
aba64c7d DM |
4327 | info.verified_insns = prog->aux->verified_insns; |
4328 | ||
2c78ee89 | 4329 | if (!bpf_capable()) { |
1e270976 MKL |
4330 | info.jited_prog_len = 0; |
4331 | info.xlated_prog_len = 0; | |
dbecd738 | 4332 | info.nr_jited_ksyms = 0; |
28c2fae7 | 4333 | info.nr_jited_func_lens = 0; |
11d8b82d YS |
4334 | info.nr_func_info = 0; |
4335 | info.nr_line_info = 0; | |
4336 | info.nr_jited_line_info = 0; | |
1e270976 MKL |
4337 | goto done; |
4338 | } | |
4339 | ||
1e270976 | 4340 | ulen = info.xlated_prog_len; |
9975a54b | 4341 | info.xlated_prog_len = bpf_prog_insn_size(prog); |
1e270976 | 4342 | if (info.xlated_prog_len && ulen) { |
7105e828 DB |
4343 | struct bpf_insn *insns_sanitized; |
4344 | bool fault; | |
4345 | ||
63960260 | 4346 | if (prog->blinded && !bpf_dump_raw_ok(file->f_cred)) { |
7105e828 DB |
4347 | info.xlated_prog_insns = 0; |
4348 | goto done; | |
4349 | } | |
63960260 | 4350 | insns_sanitized = bpf_insn_prepare_dump(prog, file->f_cred); |
7105e828 DB |
4351 | if (!insns_sanitized) |
4352 | return -ENOMEM; | |
1e270976 MKL |
4353 | uinsns = u64_to_user_ptr(info.xlated_prog_insns); |
4354 | ulen = min_t(u32, info.xlated_prog_len, ulen); | |
7105e828 DB |
4355 | fault = copy_to_user(uinsns, insns_sanitized, ulen); |
4356 | kfree(insns_sanitized); | |
4357 | if (fault) | |
1e270976 MKL |
4358 | return -EFAULT; |
4359 | } | |
4360 | ||
9d03ebc7 | 4361 | if (bpf_prog_is_offloaded(prog->aux)) { |
675fc275 JK |
4362 | err = bpf_prog_offload_info_fill(&info, prog); |
4363 | if (err) | |
4364 | return err; | |
fcfb126d JW |
4365 | goto done; |
4366 | } | |
4367 | ||
4368 | /* NOTE: the following code is supposed to be skipped for offload. | |
4369 | * bpf_prog_offload_info_fill() is the place to fill similar fields | |
4370 | * for offload. | |
4371 | */ | |
4372 | ulen = info.jited_prog_len; | |
4d56a76e SD |
4373 | if (prog->aux->func_cnt) { |
4374 | u32 i; | |
4375 | ||
4376 | info.jited_prog_len = 0; | |
4377 | for (i = 0; i < prog->aux->func_cnt; i++) | |
4378 | info.jited_prog_len += prog->aux->func[i]->jited_len; | |
4379 | } else { | |
4380 | info.jited_prog_len = prog->jited_len; | |
4381 | } | |
4382 | ||
fcfb126d | 4383 | if (info.jited_prog_len && ulen) { |
63960260 | 4384 | if (bpf_dump_raw_ok(file->f_cred)) { |
fcfb126d JW |
4385 | uinsns = u64_to_user_ptr(info.jited_prog_insns); |
4386 | ulen = min_t(u32, info.jited_prog_len, ulen); | |
4d56a76e SD |
4387 | |
4388 | /* for multi-function programs, copy the JITed | |
4389 | * instructions for all the functions | |
4390 | */ | |
4391 | if (prog->aux->func_cnt) { | |
4392 | u32 len, free, i; | |
4393 | u8 *img; | |
4394 | ||
4395 | free = ulen; | |
4396 | for (i = 0; i < prog->aux->func_cnt; i++) { | |
4397 | len = prog->aux->func[i]->jited_len; | |
4398 | len = min_t(u32, len, free); | |
4399 | img = (u8 *) prog->aux->func[i]->bpf_func; | |
4400 | if (copy_to_user(uinsns, img, len)) | |
4401 | return -EFAULT; | |
4402 | uinsns += len; | |
4403 | free -= len; | |
4404 | if (!free) | |
4405 | break; | |
4406 | } | |
4407 | } else { | |
4408 | if (copy_to_user(uinsns, prog->bpf_func, ulen)) | |
4409 | return -EFAULT; | |
4410 | } | |
fcfb126d JW |
4411 | } else { |
4412 | info.jited_prog_insns = 0; | |
4413 | } | |
675fc275 JK |
4414 | } |
4415 | ||
dbecd738 | 4416 | ulen = info.nr_jited_ksyms; |
ff1889fc | 4417 | info.nr_jited_ksyms = prog->aux->func_cnt ? : 1; |
7a5725dd | 4418 | if (ulen) { |
63960260 | 4419 | if (bpf_dump_raw_ok(file->f_cred)) { |
ff1889fc | 4420 | unsigned long ksym_addr; |
dbecd738 | 4421 | u64 __user *user_ksyms; |
dbecd738 SD |
4422 | u32 i; |
4423 | ||
4424 | /* copy the address of the kernel symbol | |
4425 | * corresponding to each function | |
4426 | */ | |
4427 | ulen = min_t(u32, info.nr_jited_ksyms, ulen); | |
4428 | user_ksyms = u64_to_user_ptr(info.jited_ksyms); | |
ff1889fc SL |
4429 | if (prog->aux->func_cnt) { |
4430 | for (i = 0; i < ulen; i++) { | |
4431 | ksym_addr = (unsigned long) | |
4432 | prog->aux->func[i]->bpf_func; | |
4433 | if (put_user((u64) ksym_addr, | |
4434 | &user_ksyms[i])) | |
4435 | return -EFAULT; | |
4436 | } | |
4437 | } else { | |
4438 | ksym_addr = (unsigned long) prog->bpf_func; | |
4439 | if (put_user((u64) ksym_addr, &user_ksyms[0])) | |
dbecd738 SD |
4440 | return -EFAULT; |
4441 | } | |
4442 | } else { | |
4443 | info.jited_ksyms = 0; | |
4444 | } | |
4445 | } | |
4446 | ||
815581c1 | 4447 | ulen = info.nr_jited_func_lens; |
ff1889fc | 4448 | info.nr_jited_func_lens = prog->aux->func_cnt ? : 1; |
7a5725dd | 4449 | if (ulen) { |
63960260 | 4450 | if (bpf_dump_raw_ok(file->f_cred)) { |
815581c1 SD |
4451 | u32 __user *user_lens; |
4452 | u32 func_len, i; | |
4453 | ||
4454 | /* copy the JITed image lengths for each function */ | |
4455 | ulen = min_t(u32, info.nr_jited_func_lens, ulen); | |
4456 | user_lens = u64_to_user_ptr(info.jited_func_lens); | |
ff1889fc SL |
4457 | if (prog->aux->func_cnt) { |
4458 | for (i = 0; i < ulen; i++) { | |
4459 | func_len = | |
4460 | prog->aux->func[i]->jited_len; | |
4461 | if (put_user(func_len, &user_lens[i])) | |
4462 | return -EFAULT; | |
4463 | } | |
4464 | } else { | |
4465 | func_len = prog->jited_len; | |
4466 | if (put_user(func_len, &user_lens[0])) | |
815581c1 SD |
4467 | return -EFAULT; |
4468 | } | |
4469 | } else { | |
4470 | info.jited_func_lens = 0; | |
4471 | } | |
4472 | } | |
4473 | ||
7337224f | 4474 | if (prog->aux->btf) |
22dc4a0f | 4475 | info.btf_id = btf_obj_id(prog->aux->btf); |
b79c9fc9 | 4476 | info.attach_btf_id = prog->aux->attach_btf_id; |
6644aabb SF |
4477 | if (attach_btf) |
4478 | info.attach_btf_obj_id = btf_obj_id(attach_btf); | |
838e9690 | 4479 | |
11d8b82d YS |
4480 | ulen = info.nr_func_info; |
4481 | info.nr_func_info = prog->aux->func_info_cnt; | |
4482 | if (info.nr_func_info && ulen) { | |
9e794163 | 4483 | char __user *user_finfo; |
7337224f | 4484 | |
9e794163 MKL |
4485 | user_finfo = u64_to_user_ptr(info.func_info); |
4486 | ulen = min_t(u32, info.nr_func_info, ulen); | |
4487 | if (copy_to_user(user_finfo, prog->aux->func_info, | |
4488 | info.func_info_rec_size * ulen)) | |
4489 | return -EFAULT; | |
838e9690 YS |
4490 | } |
4491 | ||
11d8b82d YS |
4492 | ulen = info.nr_line_info; |
4493 | info.nr_line_info = prog->aux->nr_linfo; | |
4494 | if (info.nr_line_info && ulen) { | |
9e794163 | 4495 | __u8 __user *user_linfo; |
c454a46b | 4496 | |
9e794163 MKL |
4497 | user_linfo = u64_to_user_ptr(info.line_info); |
4498 | ulen = min_t(u32, info.nr_line_info, ulen); | |
4499 | if (copy_to_user(user_linfo, prog->aux->linfo, | |
4500 | info.line_info_rec_size * ulen)) | |
4501 | return -EFAULT; | |
c454a46b MKL |
4502 | } |
4503 | ||
11d8b82d | 4504 | ulen = info.nr_jited_line_info; |
c454a46b | 4505 | if (prog->aux->jited_linfo) |
11d8b82d | 4506 | info.nr_jited_line_info = prog->aux->nr_linfo; |
c454a46b | 4507 | else |
11d8b82d YS |
4508 | info.nr_jited_line_info = 0; |
4509 | if (info.nr_jited_line_info && ulen) { | |
63960260 | 4510 | if (bpf_dump_raw_ok(file->f_cred)) { |
2cd00852 | 4511 | unsigned long line_addr; |
c454a46b MKL |
4512 | __u64 __user *user_linfo; |
4513 | u32 i; | |
4514 | ||
4515 | user_linfo = u64_to_user_ptr(info.jited_line_info); | |
11d8b82d | 4516 | ulen = min_t(u32, info.nr_jited_line_info, ulen); |
c454a46b | 4517 | for (i = 0; i < ulen; i++) { |
2cd00852 PL |
4518 | line_addr = (unsigned long)prog->aux->jited_linfo[i]; |
4519 | if (put_user((__u64)line_addr, &user_linfo[i])) | |
c454a46b MKL |
4520 | return -EFAULT; |
4521 | } | |
4522 | } else { | |
4523 | info.jited_line_info = 0; | |
4524 | } | |
4525 | } | |
4526 | ||
c872bdb3 SL |
4527 | ulen = info.nr_prog_tags; |
4528 | info.nr_prog_tags = prog->aux->func_cnt ? : 1; | |
4529 | if (ulen) { | |
4530 | __u8 __user (*user_prog_tags)[BPF_TAG_SIZE]; | |
4531 | u32 i; | |
4532 | ||
4533 | user_prog_tags = u64_to_user_ptr(info.prog_tags); | |
4534 | ulen = min_t(u32, info.nr_prog_tags, ulen); | |
4535 | if (prog->aux->func_cnt) { | |
4536 | for (i = 0; i < ulen; i++) { | |
4537 | if (copy_to_user(user_prog_tags[i], | |
4538 | prog->aux->func[i]->tag, | |
4539 | BPF_TAG_SIZE)) | |
4540 | return -EFAULT; | |
4541 | } | |
4542 | } else { | |
4543 | if (copy_to_user(user_prog_tags[0], | |
4544 | prog->tag, BPF_TAG_SIZE)) | |
4545 | return -EFAULT; | |
4546 | } | |
4547 | } | |
4548 | ||
1e270976 MKL |
4549 | done: |
4550 | if (copy_to_user(uinfo, &info, info_len) || | |
4551 | put_user(info_len, &uattr->info.info_len)) | |
4552 | return -EFAULT; | |
4553 | ||
4554 | return 0; | |
4555 | } | |
4556 | ||
63960260 KC |
4557 | static int bpf_map_get_info_by_fd(struct file *file, |
4558 | struct bpf_map *map, | |
1e270976 MKL |
4559 | const union bpf_attr *attr, |
4560 | union bpf_attr __user *uattr) | |
4561 | { | |
4562 | struct bpf_map_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
5c6f2588 | 4563 | struct bpf_map_info info; |
1e270976 MKL |
4564 | u32 info_len = attr->info.info_len; |
4565 | int err; | |
4566 | ||
af2ac3e1 | 4567 | err = bpf_check_uarg_tail_zero(USER_BPFPTR(uinfo), sizeof(info), info_len); |
1e270976 MKL |
4568 | if (err) |
4569 | return err; | |
4570 | info_len = min_t(u32, sizeof(info), info_len); | |
4571 | ||
5c6f2588 | 4572 | memset(&info, 0, sizeof(info)); |
1e270976 MKL |
4573 | info.type = map->map_type; |
4574 | info.id = map->id; | |
4575 | info.key_size = map->key_size; | |
4576 | info.value_size = map->value_size; | |
4577 | info.max_entries = map->max_entries; | |
4578 | info.map_flags = map->map_flags; | |
9330986c | 4579 | info.map_extra = map->map_extra; |
ad5b177b | 4580 | memcpy(info.name, map->name, sizeof(map->name)); |
1e270976 | 4581 | |
78958fca | 4582 | if (map->btf) { |
22dc4a0f | 4583 | info.btf_id = btf_obj_id(map->btf); |
9b2cf328 MKL |
4584 | info.btf_key_type_id = map->btf_key_type_id; |
4585 | info.btf_value_type_id = map->btf_value_type_id; | |
78958fca | 4586 | } |
85d33df3 | 4587 | info.btf_vmlinux_value_type_id = map->btf_vmlinux_value_type_id; |
78958fca | 4588 | |
9d03ebc7 | 4589 | if (bpf_map_is_offloaded(map)) { |
52775b33 JK |
4590 | err = bpf_map_offload_info_fill(&info, map); |
4591 | if (err) | |
4592 | return err; | |
4593 | } | |
4594 | ||
1e270976 MKL |
4595 | if (copy_to_user(uinfo, &info, info_len) || |
4596 | put_user(info_len, &uattr->info.info_len)) | |
4597 | return -EFAULT; | |
4598 | ||
4599 | return 0; | |
4600 | } | |
4601 | ||
63960260 KC |
4602 | static int bpf_btf_get_info_by_fd(struct file *file, |
4603 | struct btf *btf, | |
62dab84c MKL |
4604 | const union bpf_attr *attr, |
4605 | union bpf_attr __user *uattr) | |
4606 | { | |
4607 | struct bpf_btf_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
4608 | u32 info_len = attr->info.info_len; | |
4609 | int err; | |
4610 | ||
af2ac3e1 | 4611 | err = bpf_check_uarg_tail_zero(USER_BPFPTR(uinfo), sizeof(*uinfo), info_len); |
62dab84c MKL |
4612 | if (err) |
4613 | return err; | |
4614 | ||
4615 | return btf_get_info_by_fd(btf, attr, uattr); | |
4616 | } | |
4617 | ||
63960260 KC |
4618 | static int bpf_link_get_info_by_fd(struct file *file, |
4619 | struct bpf_link *link, | |
f2e10bff AN |
4620 | const union bpf_attr *attr, |
4621 | union bpf_attr __user *uattr) | |
4622 | { | |
4623 | struct bpf_link_info __user *uinfo = u64_to_user_ptr(attr->info.info); | |
4624 | struct bpf_link_info info; | |
4625 | u32 info_len = attr->info.info_len; | |
4626 | int err; | |
4627 | ||
af2ac3e1 | 4628 | err = bpf_check_uarg_tail_zero(USER_BPFPTR(uinfo), sizeof(info), info_len); |
f2e10bff AN |
4629 | if (err) |
4630 | return err; | |
4631 | info_len = min_t(u32, sizeof(info), info_len); | |
4632 | ||
4633 | memset(&info, 0, sizeof(info)); | |
4634 | if (copy_from_user(&info, uinfo, info_len)) | |
4635 | return -EFAULT; | |
4636 | ||
4637 | info.type = link->type; | |
4638 | info.id = link->id; | |
68b04864 KFL |
4639 | if (link->prog) |
4640 | info.prog_id = link->prog->aux->id; | |
f2e10bff AN |
4641 | |
4642 | if (link->ops->fill_link_info) { | |
4643 | err = link->ops->fill_link_info(link, &info); | |
4644 | if (err) | |
4645 | return err; | |
4646 | } | |
4647 | ||
4648 | if (copy_to_user(uinfo, &info, info_len) || | |
4649 | put_user(info_len, &uattr->info.info_len)) | |
4650 | return -EFAULT; | |
4651 | ||
4652 | return 0; | |
4653 | } | |
4654 | ||
4655 | ||
1e270976 MKL |
4656 | #define BPF_OBJ_GET_INFO_BY_FD_LAST_FIELD info.info |
4657 | ||
4658 | static int bpf_obj_get_info_by_fd(const union bpf_attr *attr, | |
4659 | union bpf_attr __user *uattr) | |
4660 | { | |
4661 | int ufd = attr->info.bpf_fd; | |
4662 | struct fd f; | |
4663 | int err; | |
4664 | ||
4665 | if (CHECK_ATTR(BPF_OBJ_GET_INFO_BY_FD)) | |
4666 | return -EINVAL; | |
4667 | ||
4668 | f = fdget(ufd); | |
4669 | if (!f.file) | |
4670 | return -EBADFD; | |
4671 | ||
4672 | if (f.file->f_op == &bpf_prog_fops) | |
63960260 | 4673 | err = bpf_prog_get_info_by_fd(f.file, f.file->private_data, attr, |
1e270976 MKL |
4674 | uattr); |
4675 | else if (f.file->f_op == &bpf_map_fops) | |
63960260 | 4676 | err = bpf_map_get_info_by_fd(f.file, f.file->private_data, attr, |
1e270976 | 4677 | uattr); |
60197cfb | 4678 | else if (f.file->f_op == &btf_fops) |
63960260 | 4679 | err = bpf_btf_get_info_by_fd(f.file, f.file->private_data, attr, uattr); |
f2e10bff | 4680 | else if (f.file->f_op == &bpf_link_fops) |
63960260 | 4681 | err = bpf_link_get_info_by_fd(f.file, f.file->private_data, |
f2e10bff | 4682 | attr, uattr); |
1e270976 MKL |
4683 | else |
4684 | err = -EINVAL; | |
4685 | ||
4686 | fdput(f); | |
4687 | return err; | |
4688 | } | |
4689 | ||
47a71c1f | 4690 | #define BPF_BTF_LOAD_LAST_FIELD btf_log_true_size |
f56a653c | 4691 | |
47a71c1f | 4692 | static int bpf_btf_load(const union bpf_attr *attr, bpfptr_t uattr, __u32 uattr_size) |
f56a653c MKL |
4693 | { |
4694 | if (CHECK_ATTR(BPF_BTF_LOAD)) | |
4695 | return -EINVAL; | |
4696 | ||
2c78ee89 | 4697 | if (!bpf_capable()) |
f56a653c MKL |
4698 | return -EPERM; |
4699 | ||
47a71c1f | 4700 | return btf_new_fd(attr, uattr, uattr_size); |
f56a653c MKL |
4701 | } |
4702 | ||
78958fca MKL |
4703 | #define BPF_BTF_GET_FD_BY_ID_LAST_FIELD btf_id |
4704 | ||
4705 | static int bpf_btf_get_fd_by_id(const union bpf_attr *attr) | |
4706 | { | |
4707 | if (CHECK_ATTR(BPF_BTF_GET_FD_BY_ID)) | |
4708 | return -EINVAL; | |
4709 | ||
4710 | if (!capable(CAP_SYS_ADMIN)) | |
4711 | return -EPERM; | |
4712 | ||
4713 | return btf_get_fd_by_id(attr->btf_id); | |
4714 | } | |
4715 | ||
41bdc4b4 YS |
4716 | static int bpf_task_fd_query_copy(const union bpf_attr *attr, |
4717 | union bpf_attr __user *uattr, | |
4718 | u32 prog_id, u32 fd_type, | |
4719 | const char *buf, u64 probe_offset, | |
4720 | u64 probe_addr) | |
4721 | { | |
4722 | char __user *ubuf = u64_to_user_ptr(attr->task_fd_query.buf); | |
4723 | u32 len = buf ? strlen(buf) : 0, input_len; | |
4724 | int err = 0; | |
4725 | ||
4726 | if (put_user(len, &uattr->task_fd_query.buf_len)) | |
4727 | return -EFAULT; | |
4728 | input_len = attr->task_fd_query.buf_len; | |
4729 | if (input_len && ubuf) { | |
4730 | if (!len) { | |
4731 | /* nothing to copy, just make ubuf NULL terminated */ | |
4732 | char zero = '\0'; | |
4733 | ||
4734 | if (put_user(zero, ubuf)) | |
4735 | return -EFAULT; | |
4736 | } else if (input_len >= len + 1) { | |
4737 | /* ubuf can hold the string with NULL terminator */ | |
4738 | if (copy_to_user(ubuf, buf, len + 1)) | |
4739 | return -EFAULT; | |
4740 | } else { | |
4741 | /* ubuf cannot hold the string with NULL terminator, | |
4742 | * do a partial copy with NULL terminator. | |
4743 | */ | |
4744 | char zero = '\0'; | |
4745 | ||
4746 | err = -ENOSPC; | |
4747 | if (copy_to_user(ubuf, buf, input_len - 1)) | |
4748 | return -EFAULT; | |
4749 | if (put_user(zero, ubuf + input_len - 1)) | |
4750 | return -EFAULT; | |
4751 | } | |
4752 | } | |
4753 | ||
4754 | if (put_user(prog_id, &uattr->task_fd_query.prog_id) || | |
4755 | put_user(fd_type, &uattr->task_fd_query.fd_type) || | |
4756 | put_user(probe_offset, &uattr->task_fd_query.probe_offset) || | |
4757 | put_user(probe_addr, &uattr->task_fd_query.probe_addr)) | |
4758 | return -EFAULT; | |
4759 | ||
4760 | return err; | |
4761 | } | |
4762 | ||
4763 | #define BPF_TASK_FD_QUERY_LAST_FIELD task_fd_query.probe_addr | |
4764 | ||
4765 | static int bpf_task_fd_query(const union bpf_attr *attr, | |
4766 | union bpf_attr __user *uattr) | |
4767 | { | |
4768 | pid_t pid = attr->task_fd_query.pid; | |
4769 | u32 fd = attr->task_fd_query.fd; | |
4770 | const struct perf_event *event; | |
41bdc4b4 YS |
4771 | struct task_struct *task; |
4772 | struct file *file; | |
4773 | int err; | |
4774 | ||
4775 | if (CHECK_ATTR(BPF_TASK_FD_QUERY)) | |
4776 | return -EINVAL; | |
4777 | ||
4778 | if (!capable(CAP_SYS_ADMIN)) | |
4779 | return -EPERM; | |
4780 | ||
4781 | if (attr->task_fd_query.flags != 0) | |
4782 | return -EINVAL; | |
4783 | ||
83c10cc3 | 4784 | rcu_read_lock(); |
41bdc4b4 | 4785 | task = get_pid_task(find_vpid(pid), PIDTYPE_PID); |
83c10cc3 | 4786 | rcu_read_unlock(); |
41bdc4b4 YS |
4787 | if (!task) |
4788 | return -ENOENT; | |
4789 | ||
41bdc4b4 | 4790 | err = 0; |
b48845af EB |
4791 | file = fget_task(task, fd); |
4792 | put_task_struct(task); | |
41bdc4b4 | 4793 | if (!file) |
b48845af | 4794 | return -EBADF; |
41bdc4b4 | 4795 | |
70ed506c AN |
4796 | if (file->f_op == &bpf_link_fops) { |
4797 | struct bpf_link *link = file->private_data; | |
41bdc4b4 | 4798 | |
a3b80e10 | 4799 | if (link->ops == &bpf_raw_tp_link_lops) { |
70ed506c AN |
4800 | struct bpf_raw_tp_link *raw_tp = |
4801 | container_of(link, struct bpf_raw_tp_link, link); | |
4802 | struct bpf_raw_event_map *btp = raw_tp->btp; | |
4803 | ||
4804 | err = bpf_task_fd_query_copy(attr, uattr, | |
4805 | raw_tp->link.prog->aux->id, | |
4806 | BPF_FD_TYPE_RAW_TRACEPOINT, | |
4807 | btp->tp->name, 0, 0); | |
4808 | goto put_file; | |
4809 | } | |
4810 | goto out_not_supp; | |
41bdc4b4 YS |
4811 | } |
4812 | ||
4813 | event = perf_get_event(file); | |
4814 | if (!IS_ERR(event)) { | |
4815 | u64 probe_offset, probe_addr; | |
4816 | u32 prog_id, fd_type; | |
4817 | const char *buf; | |
4818 | ||
4819 | err = bpf_get_perf_event_info(event, &prog_id, &fd_type, | |
4820 | &buf, &probe_offset, | |
4821 | &probe_addr); | |
4822 | if (!err) | |
4823 | err = bpf_task_fd_query_copy(attr, uattr, prog_id, | |
4824 | fd_type, buf, | |
4825 | probe_offset, | |
4826 | probe_addr); | |
4827 | goto put_file; | |
4828 | } | |
4829 | ||
70ed506c | 4830 | out_not_supp: |
41bdc4b4 YS |
4831 | err = -ENOTSUPP; |
4832 | put_file: | |
4833 | fput(file); | |
41bdc4b4 YS |
4834 | return err; |
4835 | } | |
4836 | ||
cb4d03ab BV |
4837 | #define BPF_MAP_BATCH_LAST_FIELD batch.flags |
4838 | ||
3af43ba4 | 4839 | #define BPF_DO_BATCH(fn, ...) \ |
cb4d03ab BV |
4840 | do { \ |
4841 | if (!fn) { \ | |
4842 | err = -ENOTSUPP; \ | |
4843 | goto err_put; \ | |
4844 | } \ | |
3af43ba4 | 4845 | err = fn(__VA_ARGS__); \ |
cb4d03ab BV |
4846 | } while (0) |
4847 | ||
4848 | static int bpf_map_do_batch(const union bpf_attr *attr, | |
4849 | union bpf_attr __user *uattr, | |
4850 | int cmd) | |
4851 | { | |
353050be DB |
4852 | bool has_read = cmd == BPF_MAP_LOOKUP_BATCH || |
4853 | cmd == BPF_MAP_LOOKUP_AND_DELETE_BATCH; | |
4854 | bool has_write = cmd != BPF_MAP_LOOKUP_BATCH; | |
cb4d03ab BV |
4855 | struct bpf_map *map; |
4856 | int err, ufd; | |
4857 | struct fd f; | |
4858 | ||
4859 | if (CHECK_ATTR(BPF_MAP_BATCH)) | |
4860 | return -EINVAL; | |
4861 | ||
4862 | ufd = attr->batch.map_fd; | |
4863 | f = fdget(ufd); | |
4864 | map = __bpf_map_get(f); | |
4865 | if (IS_ERR(map)) | |
4866 | return PTR_ERR(map); | |
353050be DB |
4867 | if (has_write) |
4868 | bpf_map_write_active_inc(map); | |
4869 | if (has_read && !(map_get_sys_perms(map, f) & FMODE_CAN_READ)) { | |
cb4d03ab BV |
4870 | err = -EPERM; |
4871 | goto err_put; | |
4872 | } | |
353050be | 4873 | if (has_write && !(map_get_sys_perms(map, f) & FMODE_CAN_WRITE)) { |
cb4d03ab BV |
4874 | err = -EPERM; |
4875 | goto err_put; | |
4876 | } | |
4877 | ||
4878 | if (cmd == BPF_MAP_LOOKUP_BATCH) | |
3af43ba4 | 4879 | BPF_DO_BATCH(map->ops->map_lookup_batch, map, attr, uattr); |
05799638 | 4880 | else if (cmd == BPF_MAP_LOOKUP_AND_DELETE_BATCH) |
3af43ba4 | 4881 | BPF_DO_BATCH(map->ops->map_lookup_and_delete_batch, map, attr, uattr); |
aa2e93b8 | 4882 | else if (cmd == BPF_MAP_UPDATE_BATCH) |
3af43ba4 | 4883 | BPF_DO_BATCH(map->ops->map_update_batch, map, f.file, attr, uattr); |
aa2e93b8 | 4884 | else |
3af43ba4 | 4885 | BPF_DO_BATCH(map->ops->map_delete_batch, map, attr, uattr); |
cb4d03ab | 4886 | err_put: |
353050be DB |
4887 | if (has_write) |
4888 | bpf_map_write_active_dec(map); | |
cb4d03ab BV |
4889 | fdput(f); |
4890 | return err; | |
4891 | } | |
4892 | ||
b733eead | 4893 | #define BPF_LINK_CREATE_LAST_FIELD link_create.uprobe_multi.pid |
af2ac3e1 | 4894 | static int link_create(union bpf_attr *attr, bpfptr_t uattr) |
af6eea57 | 4895 | { |
af6eea57 AN |
4896 | struct bpf_prog *prog; |
4897 | int ret; | |
4898 | ||
af6eea57 AN |
4899 | if (CHECK_ATTR(BPF_LINK_CREATE)) |
4900 | return -EINVAL; | |
4901 | ||
68b04864 KFL |
4902 | if (attr->link_create.attach_type == BPF_STRUCT_OPS) |
4903 | return bpf_struct_ops_link_create(attr); | |
4904 | ||
4a1e7c0c | 4905 | prog = bpf_prog_get(attr->link_create.prog_fd); |
af6eea57 AN |
4906 | if (IS_ERR(prog)) |
4907 | return PTR_ERR(prog); | |
4908 | ||
4909 | ret = bpf_prog_attach_check_attach_type(prog, | |
4910 | attr->link_create.attach_type); | |
4911 | if (ret) | |
4a1e7c0c THJ |
4912 | goto out; |
4913 | ||
df86ca0d | 4914 | switch (prog->type) { |
af6eea57 AN |
4915 | case BPF_PROG_TYPE_CGROUP_SKB: |
4916 | case BPF_PROG_TYPE_CGROUP_SOCK: | |
4917 | case BPF_PROG_TYPE_CGROUP_SOCK_ADDR: | |
4918 | case BPF_PROG_TYPE_SOCK_OPS: | |
4919 | case BPF_PROG_TYPE_CGROUP_DEVICE: | |
4920 | case BPF_PROG_TYPE_CGROUP_SYSCTL: | |
4921 | case BPF_PROG_TYPE_CGROUP_SOCKOPT: | |
4922 | ret = cgroup_bpf_link_attach(attr, prog); | |
4923 | break; | |
df86ca0d AN |
4924 | case BPF_PROG_TYPE_EXT: |
4925 | ret = bpf_tracing_prog_attach(prog, | |
4926 | attr->link_create.target_fd, | |
2fcc8241 KFL |
4927 | attr->link_create.target_btf_id, |
4928 | attr->link_create.tracing.cookie); | |
df86ca0d AN |
4929 | break; |
4930 | case BPF_PROG_TYPE_LSM: | |
de4e05ca | 4931 | case BPF_PROG_TYPE_TRACING: |
df86ca0d AN |
4932 | if (attr->link_create.attach_type != prog->expected_attach_type) { |
4933 | ret = -EINVAL; | |
4934 | goto out; | |
4935 | } | |
4936 | if (prog->expected_attach_type == BPF_TRACE_RAW_TP) | |
4937 | ret = bpf_raw_tp_link_attach(prog, NULL); | |
4938 | else if (prog->expected_attach_type == BPF_TRACE_ITER) | |
4939 | ret = bpf_iter_link_attach(attr, uattr, prog); | |
69fd337a SF |
4940 | else if (prog->expected_attach_type == BPF_LSM_CGROUP) |
4941 | ret = cgroup_bpf_link_attach(attr, prog); | |
df86ca0d AN |
4942 | else |
4943 | ret = bpf_tracing_prog_attach(prog, | |
4944 | attr->link_create.target_fd, | |
2fcc8241 KFL |
4945 | attr->link_create.target_btf_id, |
4946 | attr->link_create.tracing.cookie); | |
de4e05ca | 4947 | break; |
7f045a49 | 4948 | case BPF_PROG_TYPE_FLOW_DISSECTOR: |
e9ddbb77 | 4949 | case BPF_PROG_TYPE_SK_LOOKUP: |
7f045a49 JS |
4950 | ret = netns_bpf_link_create(attr, prog); |
4951 | break; | |
310ad797 | 4952 | #ifdef CONFIG_NET |
aa8d3a71 AN |
4953 | case BPF_PROG_TYPE_XDP: |
4954 | ret = bpf_xdp_link_attach(attr, prog); | |
84601d6e | 4955 | break; |
e420bed0 DB |
4956 | case BPF_PROG_TYPE_SCHED_CLS: |
4957 | ret = tcx_link_attach(attr, prog); | |
4958 | break; | |
84601d6e FW |
4959 | case BPF_PROG_TYPE_NETFILTER: |
4960 | ret = bpf_nf_link_attach(attr, prog); | |
aa8d3a71 | 4961 | break; |
b89fbfbb | 4962 | #endif |
b89fbfbb AN |
4963 | case BPF_PROG_TYPE_PERF_EVENT: |
4964 | case BPF_PROG_TYPE_TRACEPOINT: | |
b89fbfbb AN |
4965 | ret = bpf_perf_link_attach(attr, prog); |
4966 | break; | |
0dcac272 JO |
4967 | case BPF_PROG_TYPE_KPROBE: |
4968 | if (attr->link_create.attach_type == BPF_PERF_EVENT) | |
4969 | ret = bpf_perf_link_attach(attr, prog); | |
89ae89f5 | 4970 | else if (attr->link_create.attach_type == BPF_TRACE_KPROBE_MULTI) |
0dcac272 | 4971 | ret = bpf_kprobe_multi_link_attach(attr, prog); |
89ae89f5 JO |
4972 | else if (attr->link_create.attach_type == BPF_TRACE_UPROBE_MULTI) |
4973 | ret = bpf_uprobe_multi_link_attach(attr, prog); | |
0dcac272 | 4974 | break; |
af6eea57 AN |
4975 | default: |
4976 | ret = -EINVAL; | |
4977 | } | |
4978 | ||
4a1e7c0c | 4979 | out: |
af6eea57 AN |
4980 | if (ret < 0) |
4981 | bpf_prog_put(prog); | |
4982 | return ret; | |
4983 | } | |
4984 | ||
aef56f2e KFL |
4985 | static int link_update_map(struct bpf_link *link, union bpf_attr *attr) |
4986 | { | |
4987 | struct bpf_map *new_map, *old_map = NULL; | |
4988 | int ret; | |
4989 | ||
4990 | new_map = bpf_map_get(attr->link_update.new_map_fd); | |
4991 | if (IS_ERR(new_map)) | |
55fbae05 | 4992 | return PTR_ERR(new_map); |
aef56f2e KFL |
4993 | |
4994 | if (attr->link_update.flags & BPF_F_REPLACE) { | |
4995 | old_map = bpf_map_get(attr->link_update.old_map_fd); | |
4996 | if (IS_ERR(old_map)) { | |
55fbae05 | 4997 | ret = PTR_ERR(old_map); |
aef56f2e KFL |
4998 | goto out_put; |
4999 | } | |
5000 | } else if (attr->link_update.old_map_fd) { | |
5001 | ret = -EINVAL; | |
5002 | goto out_put; | |
5003 | } | |
5004 | ||
5005 | ret = link->ops->update_map(link, new_map, old_map); | |
5006 | ||
5007 | if (old_map) | |
5008 | bpf_map_put(old_map); | |
5009 | out_put: | |
5010 | bpf_map_put(new_map); | |
5011 | return ret; | |
5012 | } | |
5013 | ||
0c991ebc AN |
5014 | #define BPF_LINK_UPDATE_LAST_FIELD link_update.old_prog_fd |
5015 | ||
5016 | static int link_update(union bpf_attr *attr) | |
5017 | { | |
5018 | struct bpf_prog *old_prog = NULL, *new_prog; | |
5019 | struct bpf_link *link; | |
5020 | u32 flags; | |
5021 | int ret; | |
5022 | ||
0c991ebc AN |
5023 | if (CHECK_ATTR(BPF_LINK_UPDATE)) |
5024 | return -EINVAL; | |
5025 | ||
5026 | flags = attr->link_update.flags; | |
5027 | if (flags & ~BPF_F_REPLACE) | |
5028 | return -EINVAL; | |
5029 | ||
5030 | link = bpf_link_get_from_fd(attr->link_update.link_fd); | |
5031 | if (IS_ERR(link)) | |
5032 | return PTR_ERR(link); | |
5033 | ||
aef56f2e KFL |
5034 | if (link->ops->update_map) { |
5035 | ret = link_update_map(link, attr); | |
5036 | goto out_put_link; | |
5037 | } | |
5038 | ||
0c991ebc | 5039 | new_prog = bpf_prog_get(attr->link_update.new_prog_fd); |
4adb7a4a AN |
5040 | if (IS_ERR(new_prog)) { |
5041 | ret = PTR_ERR(new_prog); | |
5042 | goto out_put_link; | |
5043 | } | |
0c991ebc AN |
5044 | |
5045 | if (flags & BPF_F_REPLACE) { | |
5046 | old_prog = bpf_prog_get(attr->link_update.old_prog_fd); | |
5047 | if (IS_ERR(old_prog)) { | |
5048 | ret = PTR_ERR(old_prog); | |
5049 | old_prog = NULL; | |
5050 | goto out_put_progs; | |
5051 | } | |
4adb7a4a AN |
5052 | } else if (attr->link_update.old_prog_fd) { |
5053 | ret = -EINVAL; | |
5054 | goto out_put_progs; | |
0c991ebc AN |
5055 | } |
5056 | ||
f9d04127 AN |
5057 | if (link->ops->update_prog) |
5058 | ret = link->ops->update_prog(link, new_prog, old_prog); | |
5059 | else | |
fe537393 | 5060 | ret = -EINVAL; |
0c991ebc AN |
5061 | |
5062 | out_put_progs: | |
5063 | if (old_prog) | |
5064 | bpf_prog_put(old_prog); | |
5065 | if (ret) | |
5066 | bpf_prog_put(new_prog); | |
4adb7a4a | 5067 | out_put_link: |
ab5d47bd | 5068 | bpf_link_put_direct(link); |
0c991ebc AN |
5069 | return ret; |
5070 | } | |
5071 | ||
73b11c2a AN |
5072 | #define BPF_LINK_DETACH_LAST_FIELD link_detach.link_fd |
5073 | ||
5074 | static int link_detach(union bpf_attr *attr) | |
5075 | { | |
5076 | struct bpf_link *link; | |
5077 | int ret; | |
5078 | ||
5079 | if (CHECK_ATTR(BPF_LINK_DETACH)) | |
5080 | return -EINVAL; | |
5081 | ||
5082 | link = bpf_link_get_from_fd(attr->link_detach.link_fd); | |
5083 | if (IS_ERR(link)) | |
5084 | return PTR_ERR(link); | |
5085 | ||
5086 | if (link->ops->detach) | |
5087 | ret = link->ops->detach(link); | |
5088 | else | |
5089 | ret = -EOPNOTSUPP; | |
5090 | ||
ab5d47bd | 5091 | bpf_link_put_direct(link); |
73b11c2a AN |
5092 | return ret; |
5093 | } | |
5094 | ||
005142b8 | 5095 | static struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link) |
2d602c8c | 5096 | { |
005142b8 | 5097 | return atomic64_fetch_add_unless(&link->refcnt, 1, 0) ? link : ERR_PTR(-ENOENT); |
2d602c8c AN |
5098 | } |
5099 | ||
005142b8 | 5100 | struct bpf_link *bpf_link_by_id(u32 id) |
2d602c8c AN |
5101 | { |
5102 | struct bpf_link *link; | |
2d602c8c | 5103 | |
005142b8 AS |
5104 | if (!id) |
5105 | return ERR_PTR(-ENOENT); | |
2d602c8c AN |
5106 | |
5107 | spin_lock_bh(&link_idr_lock); | |
2d602c8c | 5108 | /* before link is "settled", ID is 0, pretend it doesn't exist yet */ |
005142b8 | 5109 | link = idr_find(&link_idr, id); |
2d602c8c AN |
5110 | if (link) { |
5111 | if (link->id) | |
005142b8 | 5112 | link = bpf_link_inc_not_zero(link); |
2d602c8c | 5113 | else |
005142b8 | 5114 | link = ERR_PTR(-EAGAIN); |
2d602c8c | 5115 | } else { |
005142b8 | 5116 | link = ERR_PTR(-ENOENT); |
2d602c8c AN |
5117 | } |
5118 | spin_unlock_bh(&link_idr_lock); | |
005142b8 AS |
5119 | return link; |
5120 | } | |
2d602c8c | 5121 | |
9f883612 DD |
5122 | struct bpf_link *bpf_link_get_curr_or_next(u32 *id) |
5123 | { | |
5124 | struct bpf_link *link; | |
5125 | ||
5126 | spin_lock_bh(&link_idr_lock); | |
5127 | again: | |
5128 | link = idr_get_next(&link_idr, id); | |
5129 | if (link) { | |
5130 | link = bpf_link_inc_not_zero(link); | |
5131 | if (IS_ERR(link)) { | |
5132 | (*id)++; | |
5133 | goto again; | |
5134 | } | |
5135 | } | |
5136 | spin_unlock_bh(&link_idr_lock); | |
5137 | ||
5138 | return link; | |
5139 | } | |
5140 | ||
005142b8 AS |
5141 | #define BPF_LINK_GET_FD_BY_ID_LAST_FIELD link_id |
5142 | ||
5143 | static int bpf_link_get_fd_by_id(const union bpf_attr *attr) | |
5144 | { | |
5145 | struct bpf_link *link; | |
5146 | u32 id = attr->link_id; | |
5147 | int fd; | |
5148 | ||
5149 | if (CHECK_ATTR(BPF_LINK_GET_FD_BY_ID)) | |
5150 | return -EINVAL; | |
5151 | ||
5152 | if (!capable(CAP_SYS_ADMIN)) | |
5153 | return -EPERM; | |
5154 | ||
5155 | link = bpf_link_by_id(id); | |
5156 | if (IS_ERR(link)) | |
5157 | return PTR_ERR(link); | |
2d602c8c AN |
5158 | |
5159 | fd = bpf_link_new_fd(link); | |
5160 | if (fd < 0) | |
ab5d47bd | 5161 | bpf_link_put_direct(link); |
2d602c8c AN |
5162 | |
5163 | return fd; | |
5164 | } | |
5165 | ||
d46edd67 SL |
5166 | DEFINE_MUTEX(bpf_stats_enabled_mutex); |
5167 | ||
5168 | static int bpf_stats_release(struct inode *inode, struct file *file) | |
5169 | { | |
5170 | mutex_lock(&bpf_stats_enabled_mutex); | |
5171 | static_key_slow_dec(&bpf_stats_enabled_key.key); | |
5172 | mutex_unlock(&bpf_stats_enabled_mutex); | |
5173 | return 0; | |
5174 | } | |
5175 | ||
5176 | static const struct file_operations bpf_stats_fops = { | |
5177 | .release = bpf_stats_release, | |
5178 | }; | |
5179 | ||
5180 | static int bpf_enable_runtime_stats(void) | |
5181 | { | |
5182 | int fd; | |
5183 | ||
5184 | mutex_lock(&bpf_stats_enabled_mutex); | |
5185 | ||
5186 | /* Set a very high limit to avoid overflow */ | |
5187 | if (static_key_count(&bpf_stats_enabled_key.key) > INT_MAX / 2) { | |
5188 | mutex_unlock(&bpf_stats_enabled_mutex); | |
5189 | return -EBUSY; | |
5190 | } | |
5191 | ||
5192 | fd = anon_inode_getfd("bpf-stats", &bpf_stats_fops, NULL, O_CLOEXEC); | |
5193 | if (fd >= 0) | |
5194 | static_key_slow_inc(&bpf_stats_enabled_key.key); | |
5195 | ||
5196 | mutex_unlock(&bpf_stats_enabled_mutex); | |
5197 | return fd; | |
5198 | } | |
5199 | ||
5200 | #define BPF_ENABLE_STATS_LAST_FIELD enable_stats.type | |
5201 | ||
5202 | static int bpf_enable_stats(union bpf_attr *attr) | |
5203 | { | |
5204 | ||
5205 | if (CHECK_ATTR(BPF_ENABLE_STATS)) | |
5206 | return -EINVAL; | |
5207 | ||
5208 | if (!capable(CAP_SYS_ADMIN)) | |
5209 | return -EPERM; | |
5210 | ||
5211 | switch (attr->enable_stats.type) { | |
5212 | case BPF_STATS_RUN_TIME: | |
5213 | return bpf_enable_runtime_stats(); | |
5214 | default: | |
5215 | break; | |
5216 | } | |
5217 | return -EINVAL; | |
5218 | } | |
5219 | ||
ac51d99b YS |
5220 | #define BPF_ITER_CREATE_LAST_FIELD iter_create.flags |
5221 | ||
5222 | static int bpf_iter_create(union bpf_attr *attr) | |
5223 | { | |
5224 | struct bpf_link *link; | |
5225 | int err; | |
5226 | ||
5227 | if (CHECK_ATTR(BPF_ITER_CREATE)) | |
5228 | return -EINVAL; | |
5229 | ||
5230 | if (attr->iter_create.flags) | |
5231 | return -EINVAL; | |
5232 | ||
5233 | link = bpf_link_get_from_fd(attr->iter_create.link_fd); | |
5234 | if (IS_ERR(link)) | |
5235 | return PTR_ERR(link); | |
5236 | ||
5237 | err = bpf_iter_new_fd(link); | |
ab5d47bd | 5238 | bpf_link_put_direct(link); |
ac51d99b YS |
5239 | |
5240 | return err; | |
5241 | } | |
5242 | ||
ef15314a YZ |
5243 | #define BPF_PROG_BIND_MAP_LAST_FIELD prog_bind_map.flags |
5244 | ||
5245 | static int bpf_prog_bind_map(union bpf_attr *attr) | |
5246 | { | |
5247 | struct bpf_prog *prog; | |
5248 | struct bpf_map *map; | |
5249 | struct bpf_map **used_maps_old, **used_maps_new; | |
5250 | int i, ret = 0; | |
5251 | ||
5252 | if (CHECK_ATTR(BPF_PROG_BIND_MAP)) | |
5253 | return -EINVAL; | |
5254 | ||
5255 | if (attr->prog_bind_map.flags) | |
5256 | return -EINVAL; | |
5257 | ||
5258 | prog = bpf_prog_get(attr->prog_bind_map.prog_fd); | |
5259 | if (IS_ERR(prog)) | |
5260 | return PTR_ERR(prog); | |
5261 | ||
5262 | map = bpf_map_get(attr->prog_bind_map.map_fd); | |
5263 | if (IS_ERR(map)) { | |
5264 | ret = PTR_ERR(map); | |
5265 | goto out_prog_put; | |
5266 | } | |
5267 | ||
5268 | mutex_lock(&prog->aux->used_maps_mutex); | |
5269 | ||
5270 | used_maps_old = prog->aux->used_maps; | |
5271 | ||
5272 | for (i = 0; i < prog->aux->used_map_cnt; i++) | |
1028ae40 SF |
5273 | if (used_maps_old[i] == map) { |
5274 | bpf_map_put(map); | |
ef15314a | 5275 | goto out_unlock; |
1028ae40 | 5276 | } |
ef15314a YZ |
5277 | |
5278 | used_maps_new = kmalloc_array(prog->aux->used_map_cnt + 1, | |
5279 | sizeof(used_maps_new[0]), | |
5280 | GFP_KERNEL); | |
5281 | if (!used_maps_new) { | |
5282 | ret = -ENOMEM; | |
5283 | goto out_unlock; | |
5284 | } | |
5285 | ||
5286 | memcpy(used_maps_new, used_maps_old, | |
5287 | sizeof(used_maps_old[0]) * prog->aux->used_map_cnt); | |
5288 | used_maps_new[prog->aux->used_map_cnt] = map; | |
5289 | ||
5290 | prog->aux->used_map_cnt++; | |
5291 | prog->aux->used_maps = used_maps_new; | |
5292 | ||
5293 | kfree(used_maps_old); | |
5294 | ||
5295 | out_unlock: | |
5296 | mutex_unlock(&prog->aux->used_maps_mutex); | |
5297 | ||
5298 | if (ret) | |
5299 | bpf_map_put(map); | |
5300 | out_prog_put: | |
5301 | bpf_prog_put(prog); | |
5302 | return ret; | |
5303 | } | |
5304 | ||
af2ac3e1 | 5305 | static int __sys_bpf(int cmd, bpfptr_t uattr, unsigned int size) |
99c55f7d | 5306 | { |
8096f229 | 5307 | union bpf_attr attr; |
99c55f7d AS |
5308 | int err; |
5309 | ||
dcab51f1 | 5310 | err = bpf_check_uarg_tail_zero(uattr, sizeof(attr), size); |
1e270976 MKL |
5311 | if (err) |
5312 | return err; | |
5313 | size = min_t(u32, size, sizeof(attr)); | |
99c55f7d AS |
5314 | |
5315 | /* copy attributes from user space, may be less than sizeof(bpf_attr) */ | |
8096f229 | 5316 | memset(&attr, 0, sizeof(attr)); |
af2ac3e1 | 5317 | if (copy_from_bpfptr(&attr, uattr, size) != 0) |
99c55f7d AS |
5318 | return -EFAULT; |
5319 | ||
afdb09c7 CF |
5320 | err = security_bpf(cmd, &attr, size); |
5321 | if (err < 0) | |
5322 | return err; | |
5323 | ||
99c55f7d AS |
5324 | switch (cmd) { |
5325 | case BPF_MAP_CREATE: | |
5326 | err = map_create(&attr); | |
5327 | break; | |
db20fd2b AS |
5328 | case BPF_MAP_LOOKUP_ELEM: |
5329 | err = map_lookup_elem(&attr); | |
5330 | break; | |
5331 | case BPF_MAP_UPDATE_ELEM: | |
af2ac3e1 | 5332 | err = map_update_elem(&attr, uattr); |
db20fd2b AS |
5333 | break; |
5334 | case BPF_MAP_DELETE_ELEM: | |
b88df697 | 5335 | err = map_delete_elem(&attr, uattr); |
db20fd2b AS |
5336 | break; |
5337 | case BPF_MAP_GET_NEXT_KEY: | |
5338 | err = map_get_next_key(&attr); | |
5339 | break; | |
87df15de DB |
5340 | case BPF_MAP_FREEZE: |
5341 | err = map_freeze(&attr); | |
5342 | break; | |
09756af4 | 5343 | case BPF_PROG_LOAD: |
47a71c1f | 5344 | err = bpf_prog_load(&attr, uattr, size); |
09756af4 | 5345 | break; |
b2197755 DB |
5346 | case BPF_OBJ_PIN: |
5347 | err = bpf_obj_pin(&attr); | |
5348 | break; | |
5349 | case BPF_OBJ_GET: | |
5350 | err = bpf_obj_get(&attr); | |
5351 | break; | |
f4324551 DM |
5352 | case BPF_PROG_ATTACH: |
5353 | err = bpf_prog_attach(&attr); | |
5354 | break; | |
5355 | case BPF_PROG_DETACH: | |
5356 | err = bpf_prog_detach(&attr); | |
5357 | break; | |
468e2f64 | 5358 | case BPF_PROG_QUERY: |
af2ac3e1 | 5359 | err = bpf_prog_query(&attr, uattr.user); |
468e2f64 | 5360 | break; |
1cf1cae9 | 5361 | case BPF_PROG_TEST_RUN: |
af2ac3e1 | 5362 | err = bpf_prog_test_run(&attr, uattr.user); |
1cf1cae9 | 5363 | break; |
34ad5580 | 5364 | case BPF_PROG_GET_NEXT_ID: |
af2ac3e1 | 5365 | err = bpf_obj_get_next_id(&attr, uattr.user, |
34ad5580 MKL |
5366 | &prog_idr, &prog_idr_lock); |
5367 | break; | |
5368 | case BPF_MAP_GET_NEXT_ID: | |
af2ac3e1 | 5369 | err = bpf_obj_get_next_id(&attr, uattr.user, |
34ad5580 MKL |
5370 | &map_idr, &map_idr_lock); |
5371 | break; | |
1b9ed84e | 5372 | case BPF_BTF_GET_NEXT_ID: |
af2ac3e1 | 5373 | err = bpf_obj_get_next_id(&attr, uattr.user, |
1b9ed84e QM |
5374 | &btf_idr, &btf_idr_lock); |
5375 | break; | |
b16d9aa4 MKL |
5376 | case BPF_PROG_GET_FD_BY_ID: |
5377 | err = bpf_prog_get_fd_by_id(&attr); | |
5378 | break; | |
bd5f5f4e MKL |
5379 | case BPF_MAP_GET_FD_BY_ID: |
5380 | err = bpf_map_get_fd_by_id(&attr); | |
5381 | break; | |
1e270976 | 5382 | case BPF_OBJ_GET_INFO_BY_FD: |
af2ac3e1 | 5383 | err = bpf_obj_get_info_by_fd(&attr, uattr.user); |
1e270976 | 5384 | break; |
c4f6699d AS |
5385 | case BPF_RAW_TRACEPOINT_OPEN: |
5386 | err = bpf_raw_tracepoint_open(&attr); | |
5387 | break; | |
f56a653c | 5388 | case BPF_BTF_LOAD: |
47a71c1f | 5389 | err = bpf_btf_load(&attr, uattr, size); |
f56a653c | 5390 | break; |
78958fca MKL |
5391 | case BPF_BTF_GET_FD_BY_ID: |
5392 | err = bpf_btf_get_fd_by_id(&attr); | |
5393 | break; | |
41bdc4b4 | 5394 | case BPF_TASK_FD_QUERY: |
af2ac3e1 | 5395 | err = bpf_task_fd_query(&attr, uattr.user); |
41bdc4b4 | 5396 | break; |
bd513cd0 MV |
5397 | case BPF_MAP_LOOKUP_AND_DELETE_ELEM: |
5398 | err = map_lookup_and_delete_elem(&attr); | |
5399 | break; | |
cb4d03ab | 5400 | case BPF_MAP_LOOKUP_BATCH: |
af2ac3e1 | 5401 | err = bpf_map_do_batch(&attr, uattr.user, BPF_MAP_LOOKUP_BATCH); |
cb4d03ab | 5402 | break; |
05799638 | 5403 | case BPF_MAP_LOOKUP_AND_DELETE_BATCH: |
af2ac3e1 | 5404 | err = bpf_map_do_batch(&attr, uattr.user, |
05799638 YS |
5405 | BPF_MAP_LOOKUP_AND_DELETE_BATCH); |
5406 | break; | |
aa2e93b8 | 5407 | case BPF_MAP_UPDATE_BATCH: |
af2ac3e1 | 5408 | err = bpf_map_do_batch(&attr, uattr.user, BPF_MAP_UPDATE_BATCH); |
aa2e93b8 BV |
5409 | break; |
5410 | case BPF_MAP_DELETE_BATCH: | |
af2ac3e1 | 5411 | err = bpf_map_do_batch(&attr, uattr.user, BPF_MAP_DELETE_BATCH); |
aa2e93b8 | 5412 | break; |
af6eea57 | 5413 | case BPF_LINK_CREATE: |
af2ac3e1 | 5414 | err = link_create(&attr, uattr); |
af6eea57 | 5415 | break; |
0c991ebc AN |
5416 | case BPF_LINK_UPDATE: |
5417 | err = link_update(&attr); | |
5418 | break; | |
2d602c8c AN |
5419 | case BPF_LINK_GET_FD_BY_ID: |
5420 | err = bpf_link_get_fd_by_id(&attr); | |
5421 | break; | |
5422 | case BPF_LINK_GET_NEXT_ID: | |
af2ac3e1 | 5423 | err = bpf_obj_get_next_id(&attr, uattr.user, |
2d602c8c AN |
5424 | &link_idr, &link_idr_lock); |
5425 | break; | |
d46edd67 SL |
5426 | case BPF_ENABLE_STATS: |
5427 | err = bpf_enable_stats(&attr); | |
5428 | break; | |
ac51d99b YS |
5429 | case BPF_ITER_CREATE: |
5430 | err = bpf_iter_create(&attr); | |
5431 | break; | |
73b11c2a AN |
5432 | case BPF_LINK_DETACH: |
5433 | err = link_detach(&attr); | |
5434 | break; | |
ef15314a YZ |
5435 | case BPF_PROG_BIND_MAP: |
5436 | err = bpf_prog_bind_map(&attr); | |
5437 | break; | |
99c55f7d AS |
5438 | default: |
5439 | err = -EINVAL; | |
5440 | break; | |
5441 | } | |
5442 | ||
5443 | return err; | |
5444 | } | |
79a7f8bd | 5445 | |
af2ac3e1 AS |
5446 | SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size) |
5447 | { | |
5448 | return __sys_bpf(cmd, USER_BPFPTR(uattr), size); | |
5449 | } | |
5450 | ||
79a7f8bd AS |
5451 | static bool syscall_prog_is_valid_access(int off, int size, |
5452 | enum bpf_access_type type, | |
5453 | const struct bpf_prog *prog, | |
5454 | struct bpf_insn_access_aux *info) | |
5455 | { | |
5456 | if (off < 0 || off >= U16_MAX) | |
5457 | return false; | |
5458 | if (off % size != 0) | |
5459 | return false; | |
5460 | return true; | |
5461 | } | |
5462 | ||
b1d18a75 | 5463 | BPF_CALL_3(bpf_sys_bpf, int, cmd, union bpf_attr *, attr, u32, attr_size) |
79a7f8bd | 5464 | { |
af2ac3e1 AS |
5465 | switch (cmd) { |
5466 | case BPF_MAP_CREATE: | |
b88df697 | 5467 | case BPF_MAP_DELETE_ELEM: |
af2ac3e1 AS |
5468 | case BPF_MAP_UPDATE_ELEM: |
5469 | case BPF_MAP_FREEZE: | |
b88df697 | 5470 | case BPF_MAP_GET_FD_BY_ID: |
af2ac3e1 | 5471 | case BPF_PROG_LOAD: |
c571bd75 | 5472 | case BPF_BTF_LOAD: |
b1d18a75 AS |
5473 | case BPF_LINK_CREATE: |
5474 | case BPF_RAW_TRACEPOINT_OPEN: | |
af2ac3e1 | 5475 | break; |
86f44fce AS |
5476 | default: |
5477 | return -EINVAL; | |
5478 | } | |
5479 | return __sys_bpf(cmd, KERNEL_BPFPTR(attr), attr_size); | |
5480 | } | |
5481 | ||
4e4588f1 AS |
5482 | |
5483 | /* To shut up -Wmissing-prototypes. | |
5484 | * This function is used by the kernel light skeleton | |
5485 | * to load bpf programs when modules are loaded or during kernel boot. | |
5486 | * See tools/lib/bpf/skel_internal.h | |
5487 | */ | |
5488 | int kern_sys_bpf(int cmd, union bpf_attr *attr, unsigned int size); | |
5489 | ||
86f44fce AS |
5490 | int kern_sys_bpf(int cmd, union bpf_attr *attr, unsigned int size) |
5491 | { | |
5492 | struct bpf_prog * __maybe_unused prog; | |
5493 | struct bpf_tramp_run_ctx __maybe_unused run_ctx; | |
5494 | ||
5495 | switch (cmd) { | |
b1d18a75 AS |
5496 | #ifdef CONFIG_BPF_JIT /* __bpf_prog_enter_sleepable used by trampoline and JIT */ |
5497 | case BPF_PROG_TEST_RUN: | |
5498 | if (attr->test.data_in || attr->test.data_out || | |
5499 | attr->test.ctx_out || attr->test.duration || | |
5500 | attr->test.repeat || attr->test.flags) | |
5501 | return -EINVAL; | |
5502 | ||
5503 | prog = bpf_prog_get_type(attr->test.prog_fd, BPF_PROG_TYPE_SYSCALL); | |
5504 | if (IS_ERR(prog)) | |
5505 | return PTR_ERR(prog); | |
5506 | ||
5507 | if (attr->test.ctx_size_in < prog->aux->max_ctx_offset || | |
5508 | attr->test.ctx_size_in > U16_MAX) { | |
5509 | bpf_prog_put(prog); | |
5510 | return -EINVAL; | |
5511 | } | |
5512 | ||
e384c7b7 | 5513 | run_ctx.bpf_cookie = 0; |
271de525 | 5514 | if (!__bpf_prog_enter_sleepable_recur(prog, &run_ctx)) { |
b1d18a75 | 5515 | /* recursion detected */ |
7645629f | 5516 | __bpf_prog_exit_sleepable_recur(prog, 0, &run_ctx); |
b1d18a75 AS |
5517 | bpf_prog_put(prog); |
5518 | return -EBUSY; | |
5519 | } | |
5520 | attr->test.retval = bpf_prog_run(prog, (void *) (long) attr->test.ctx_in); | |
271de525 MKL |
5521 | __bpf_prog_exit_sleepable_recur(prog, 0 /* bpf_prog_run does runtime stats */, |
5522 | &run_ctx); | |
b1d18a75 AS |
5523 | bpf_prog_put(prog); |
5524 | return 0; | |
5525 | #endif | |
af2ac3e1 | 5526 | default: |
86f44fce | 5527 | return ____bpf_sys_bpf(cmd, attr, size); |
af2ac3e1 | 5528 | } |
79a7f8bd | 5529 | } |
86f44fce | 5530 | EXPORT_SYMBOL(kern_sys_bpf); |
79a7f8bd | 5531 | |
3a2daa72 | 5532 | static const struct bpf_func_proto bpf_sys_bpf_proto = { |
79a7f8bd AS |
5533 | .func = bpf_sys_bpf, |
5534 | .gpl_only = false, | |
5535 | .ret_type = RET_INTEGER, | |
5536 | .arg1_type = ARG_ANYTHING, | |
216e3cd2 | 5537 | .arg2_type = ARG_PTR_TO_MEM | MEM_RDONLY, |
79a7f8bd AS |
5538 | .arg3_type = ARG_CONST_SIZE, |
5539 | }; | |
5540 | ||
5541 | const struct bpf_func_proto * __weak | |
5542 | tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) | |
5543 | { | |
5544 | return bpf_base_func_proto(func_id); | |
5545 | } | |
5546 | ||
3abea089 AS |
5547 | BPF_CALL_1(bpf_sys_close, u32, fd) |
5548 | { | |
5549 | /* When bpf program calls this helper there should not be | |
5550 | * an fdget() without matching completed fdput(). | |
5551 | * This helper is allowed in the following callchain only: | |
5552 | * sys_bpf->prog_test_run->bpf_prog->bpf_sys_close | |
5553 | */ | |
5554 | return close_fd(fd); | |
5555 | } | |
5556 | ||
3a2daa72 | 5557 | static const struct bpf_func_proto bpf_sys_close_proto = { |
3abea089 AS |
5558 | .func = bpf_sys_close, |
5559 | .gpl_only = false, | |
5560 | .ret_type = RET_INTEGER, | |
5561 | .arg1_type = ARG_ANYTHING, | |
5562 | }; | |
5563 | ||
d6aef08a KKD |
5564 | BPF_CALL_4(bpf_kallsyms_lookup_name, const char *, name, int, name_sz, int, flags, u64 *, res) |
5565 | { | |
5566 | if (flags) | |
5567 | return -EINVAL; | |
5568 | ||
5569 | if (name_sz <= 1 || name[name_sz - 1]) | |
5570 | return -EINVAL; | |
5571 | ||
5572 | if (!bpf_dump_raw_ok(current_cred())) | |
5573 | return -EPERM; | |
5574 | ||
5575 | *res = kallsyms_lookup_name(name); | |
5576 | return *res ? 0 : -ENOENT; | |
5577 | } | |
5578 | ||
dc368e1c | 5579 | static const struct bpf_func_proto bpf_kallsyms_lookup_name_proto = { |
d6aef08a KKD |
5580 | .func = bpf_kallsyms_lookup_name, |
5581 | .gpl_only = false, | |
5582 | .ret_type = RET_INTEGER, | |
5583 | .arg1_type = ARG_PTR_TO_MEM, | |
d4efb170 | 5584 | .arg2_type = ARG_CONST_SIZE_OR_ZERO, |
d6aef08a KKD |
5585 | .arg3_type = ARG_ANYTHING, |
5586 | .arg4_type = ARG_PTR_TO_LONG, | |
5587 | }; | |
5588 | ||
79a7f8bd AS |
5589 | static const struct bpf_func_proto * |
5590 | syscall_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) | |
5591 | { | |
5592 | switch (func_id) { | |
5593 | case BPF_FUNC_sys_bpf: | |
14b20b78 | 5594 | return !perfmon_capable() ? NULL : &bpf_sys_bpf_proto; |
3d78417b AS |
5595 | case BPF_FUNC_btf_find_by_name_kind: |
5596 | return &bpf_btf_find_by_name_kind_proto; | |
3abea089 AS |
5597 | case BPF_FUNC_sys_close: |
5598 | return &bpf_sys_close_proto; | |
d6aef08a KKD |
5599 | case BPF_FUNC_kallsyms_lookup_name: |
5600 | return &bpf_kallsyms_lookup_name_proto; | |
79a7f8bd AS |
5601 | default: |
5602 | return tracing_prog_func_proto(func_id, prog); | |
5603 | } | |
5604 | } | |
5605 | ||
5606 | const struct bpf_verifier_ops bpf_syscall_verifier_ops = { | |
5607 | .get_func_proto = syscall_prog_func_proto, | |
5608 | .is_valid_access = syscall_prog_is_valid_access, | |
5609 | }; | |
5610 | ||
5611 | const struct bpf_prog_ops bpf_syscall_prog_ops = { | |
5612 | .test_run = bpf_prog_test_run_syscall, | |
5613 | }; | |
2900005e YZ |
5614 | |
5615 | #ifdef CONFIG_SYSCTL | |
5616 | static int bpf_stats_handler(struct ctl_table *table, int write, | |
5617 | void *buffer, size_t *lenp, loff_t *ppos) | |
5618 | { | |
5619 | struct static_key *key = (struct static_key *)table->data; | |
5620 | static int saved_val; | |
5621 | int val, ret; | |
5622 | struct ctl_table tmp = { | |
5623 | .data = &val, | |
5624 | .maxlen = sizeof(val), | |
5625 | .mode = table->mode, | |
5626 | .extra1 = SYSCTL_ZERO, | |
5627 | .extra2 = SYSCTL_ONE, | |
5628 | }; | |
5629 | ||
5630 | if (write && !capable(CAP_SYS_ADMIN)) | |
5631 | return -EPERM; | |
5632 | ||
5633 | mutex_lock(&bpf_stats_enabled_mutex); | |
5634 | val = saved_val; | |
5635 | ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); | |
5636 | if (write && !ret && val != saved_val) { | |
5637 | if (val) | |
5638 | static_key_slow_inc(key); | |
5639 | else | |
5640 | static_key_slow_dec(key); | |
5641 | saved_val = val; | |
5642 | } | |
5643 | mutex_unlock(&bpf_stats_enabled_mutex); | |
5644 | return ret; | |
5645 | } | |
5646 | ||
5647 | void __weak unpriv_ebpf_notify(int new_state) | |
5648 | { | |
5649 | } | |
5650 | ||
5651 | static int bpf_unpriv_handler(struct ctl_table *table, int write, | |
5652 | void *buffer, size_t *lenp, loff_t *ppos) | |
5653 | { | |
5654 | int ret, unpriv_enable = *(int *)table->data; | |
5655 | bool locked_state = unpriv_enable == 1; | |
5656 | struct ctl_table tmp = *table; | |
5657 | ||
5658 | if (write && !capable(CAP_SYS_ADMIN)) | |
5659 | return -EPERM; | |
5660 | ||
5661 | tmp.data = &unpriv_enable; | |
5662 | ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); | |
5663 | if (write && !ret) { | |
5664 | if (locked_state && unpriv_enable != 1) | |
5665 | return -EPERM; | |
5666 | *(int *)table->data = unpriv_enable; | |
5667 | } | |
5668 | ||
fedf9920 KFL |
5669 | if (write) |
5670 | unpriv_ebpf_notify(unpriv_enable); | |
2900005e YZ |
5671 | |
5672 | return ret; | |
5673 | } | |
5674 | ||
5675 | static struct ctl_table bpf_syscall_table[] = { | |
5676 | { | |
5677 | .procname = "unprivileged_bpf_disabled", | |
5678 | .data = &sysctl_unprivileged_bpf_disabled, | |
5679 | .maxlen = sizeof(sysctl_unprivileged_bpf_disabled), | |
5680 | .mode = 0644, | |
5681 | .proc_handler = bpf_unpriv_handler, | |
5682 | .extra1 = SYSCTL_ZERO, | |
5683 | .extra2 = SYSCTL_TWO, | |
5684 | }, | |
5685 | { | |
5686 | .procname = "bpf_stats_enabled", | |
5687 | .data = &bpf_stats_enabled_key.key, | |
2900005e YZ |
5688 | .mode = 0644, |
5689 | .proc_handler = bpf_stats_handler, | |
5690 | }, | |
5691 | { } | |
5692 | }; | |
5693 | ||
5694 | static int __init bpf_syscall_sysctl_init(void) | |
5695 | { | |
5696 | register_sysctl_init("kernel", bpf_syscall_table); | |
5697 | return 0; | |
5698 | } | |
5699 | late_initcall(bpf_syscall_sysctl_init); | |
5700 | #endif /* CONFIG_SYSCTL */ |