1 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 * This program is free software; you can redistribute it and/or
4 * modify it under the terms of version 2 of the GNU General Public
5 * License as published by the Free Software Foundation.
7 * This program is distributed in the hope that it will be useful, but
8 * WITHOUT ANY WARRANTY; without even the implied warranty of
9 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
10 * General Public License for more details.
12 #include <linux/bpf.h>
13 #include <linux/bpf_trace.h>
14 #include <linux/syscalls.h>
15 #include <linux/slab.h>
16 #include <linux/vmalloc.h>
17 #include <linux/mmzone.h>
18 #include <linux/anon_inodes.h>
19 #include <linux/file.h>
20 #include <linux/license.h>
21 #include <linux/filter.h>
22 #include <linux/version.h>
23 #include <linux/kernel.h>
25 DEFINE_PER_CPU(int, bpf_prog_active
);
27 int sysctl_unprivileged_bpf_disabled __read_mostly
;
29 static LIST_HEAD(bpf_map_types
);
31 static struct bpf_map
*find_and_alloc_map(union bpf_attr
*attr
)
33 struct bpf_map_type_list
*tl
;
36 list_for_each_entry(tl
, &bpf_map_types
, list_node
) {
37 if (tl
->type
== attr
->map_type
) {
38 map
= tl
->ops
->map_alloc(attr
);
42 map
->map_type
= attr
->map_type
;
46 return ERR_PTR(-EINVAL
);
49 /* boot time registration of different map implementations */
50 void bpf_register_map_type(struct bpf_map_type_list
*tl
)
52 list_add(&tl
->list_node
, &bpf_map_types
);
55 void *bpf_map_area_alloc(size_t size
)
57 /* We definitely need __GFP_NORETRY, so OOM killer doesn't
58 * trigger under memory pressure as we really just want to
61 const gfp_t flags
= __GFP_NOWARN
| __GFP_NORETRY
| __GFP_ZERO
;
64 if (size
<= (PAGE_SIZE
<< PAGE_ALLOC_COSTLY_ORDER
)) {
65 area
= kmalloc(size
, GFP_USER
| flags
);
70 return __vmalloc(size
, GFP_KERNEL
| __GFP_HIGHMEM
| flags
,
74 void bpf_map_area_free(void *area
)
79 int bpf_map_precharge_memlock(u32 pages
)
81 struct user_struct
*user
= get_current_user();
82 unsigned long memlock_limit
, cur
;
84 memlock_limit
= rlimit(RLIMIT_MEMLOCK
) >> PAGE_SHIFT
;
85 cur
= atomic_long_read(&user
->locked_vm
);
87 if (cur
+ pages
> memlock_limit
)
92 static int bpf_map_charge_memlock(struct bpf_map
*map
)
94 struct user_struct
*user
= get_current_user();
95 unsigned long memlock_limit
;
97 memlock_limit
= rlimit(RLIMIT_MEMLOCK
) >> PAGE_SHIFT
;
99 atomic_long_add(map
->pages
, &user
->locked_vm
);
101 if (atomic_long_read(&user
->locked_vm
) > memlock_limit
) {
102 atomic_long_sub(map
->pages
, &user
->locked_vm
);
110 static void bpf_map_uncharge_memlock(struct bpf_map
*map
)
112 struct user_struct
*user
= map
->user
;
114 atomic_long_sub(map
->pages
, &user
->locked_vm
);
118 /* called from workqueue */
119 static void bpf_map_free_deferred(struct work_struct
*work
)
121 struct bpf_map
*map
= container_of(work
, struct bpf_map
, work
);
123 bpf_map_uncharge_memlock(map
);
124 /* implementation dependent freeing */
125 map
->ops
->map_free(map
);
128 static void bpf_map_put_uref(struct bpf_map
*map
)
130 if (atomic_dec_and_test(&map
->usercnt
)) {
131 if (map
->map_type
== BPF_MAP_TYPE_PROG_ARRAY
)
132 bpf_fd_array_map_clear(map
);
136 /* decrement map refcnt and schedule it for freeing via workqueue
137 * (unrelying map implementation ops->map_free() might sleep)
139 void bpf_map_put(struct bpf_map
*map
)
141 if (atomic_dec_and_test(&map
->refcnt
)) {
142 INIT_WORK(&map
->work
, bpf_map_free_deferred
);
143 schedule_work(&map
->work
);
147 void bpf_map_put_with_uref(struct bpf_map
*map
)
149 bpf_map_put_uref(map
);
153 static int bpf_map_release(struct inode
*inode
, struct file
*filp
)
155 struct bpf_map
*map
= filp
->private_data
;
157 if (map
->ops
->map_release
)
158 map
->ops
->map_release(map
, filp
);
160 bpf_map_put_with_uref(map
);
164 #ifdef CONFIG_PROC_FS
165 static void bpf_map_show_fdinfo(struct seq_file
*m
, struct file
*filp
)
167 const struct bpf_map
*map
= filp
->private_data
;
168 const struct bpf_array
*array
;
169 u32 owner_prog_type
= 0;
171 if (map
->map_type
== BPF_MAP_TYPE_PROG_ARRAY
) {
172 array
= container_of(map
, struct bpf_array
, map
);
173 owner_prog_type
= array
->owner_prog_type
;
188 map
->pages
* 1ULL << PAGE_SHIFT
);
191 seq_printf(m
, "owner_prog_type:\t%u\n",
196 static const struct file_operations bpf_map_fops
= {
197 #ifdef CONFIG_PROC_FS
198 .show_fdinfo
= bpf_map_show_fdinfo
,
200 .release
= bpf_map_release
,
203 int bpf_map_new_fd(struct bpf_map
*map
)
205 return anon_inode_getfd("bpf-map", &bpf_map_fops
, map
,
209 /* helper macro to check that unused fields 'union bpf_attr' are zero */
210 #define CHECK_ATTR(CMD) \
211 memchr_inv((void *) &attr->CMD##_LAST_FIELD + \
212 sizeof(attr->CMD##_LAST_FIELD), 0, \
214 offsetof(union bpf_attr, CMD##_LAST_FIELD) - \
215 sizeof(attr->CMD##_LAST_FIELD)) != NULL
217 #define BPF_MAP_CREATE_LAST_FIELD map_flags
218 /* called via syscall */
219 static int map_create(union bpf_attr
*attr
)
224 err
= CHECK_ATTR(BPF_MAP_CREATE
);
228 /* find map type and init map: hashtable vs rbtree vs bloom vs ... */
229 map
= find_and_alloc_map(attr
);
233 atomic_set(&map
->refcnt
, 1);
234 atomic_set(&map
->usercnt
, 1);
236 err
= bpf_map_charge_memlock(map
);
238 goto free_map_nouncharge
;
240 err
= bpf_map_new_fd(map
);
242 /* failed to allocate fd */
245 trace_bpf_map_create(map
, err
);
249 bpf_map_uncharge_memlock(map
);
251 map
->ops
->map_free(map
);
255 /* if error is returned, fd is released.
256 * On success caller should complete fd access with matching fdput()
258 struct bpf_map
*__bpf_map_get(struct fd f
)
261 return ERR_PTR(-EBADF
);
262 if (f
.file
->f_op
!= &bpf_map_fops
) {
264 return ERR_PTR(-EINVAL
);
267 return f
.file
->private_data
;
270 /* prog's and map's refcnt limit */
271 #define BPF_MAX_REFCNT 32768
273 struct bpf_map
*bpf_map_inc(struct bpf_map
*map
, bool uref
)
275 if (atomic_inc_return(&map
->refcnt
) > BPF_MAX_REFCNT
) {
276 atomic_dec(&map
->refcnt
);
277 return ERR_PTR(-EBUSY
);
280 atomic_inc(&map
->usercnt
);
284 struct bpf_map
*bpf_map_get_with_uref(u32 ufd
)
286 struct fd f
= fdget(ufd
);
289 map
= __bpf_map_get(f
);
293 map
= bpf_map_inc(map
, true);
299 int __weak
bpf_stackmap_copy(struct bpf_map
*map
, void *key
, void *value
)
304 /* last field in 'union bpf_attr' used by this command */
305 #define BPF_MAP_LOOKUP_ELEM_LAST_FIELD value
307 static int map_lookup_elem(union bpf_attr
*attr
)
309 void __user
*ukey
= u64_to_user_ptr(attr
->key
);
310 void __user
*uvalue
= u64_to_user_ptr(attr
->value
);
311 int ufd
= attr
->map_fd
;
313 void *key
, *value
, *ptr
;
318 if (CHECK_ATTR(BPF_MAP_LOOKUP_ELEM
))
322 map
= __bpf_map_get(f
);
327 key
= kmalloc(map
->key_size
, GFP_USER
);
332 if (copy_from_user(key
, ukey
, map
->key_size
) != 0)
335 if (map
->map_type
== BPF_MAP_TYPE_PERCPU_HASH
||
336 map
->map_type
== BPF_MAP_TYPE_LRU_PERCPU_HASH
||
337 map
->map_type
== BPF_MAP_TYPE_PERCPU_ARRAY
)
338 value_size
= round_up(map
->value_size
, 8) * num_possible_cpus();
340 value_size
= map
->value_size
;
343 value
= kmalloc(value_size
, GFP_USER
| __GFP_NOWARN
);
347 if (map
->map_type
== BPF_MAP_TYPE_PERCPU_HASH
||
348 map
->map_type
== BPF_MAP_TYPE_LRU_PERCPU_HASH
) {
349 err
= bpf_percpu_hash_copy(map
, key
, value
);
350 } else if (map
->map_type
== BPF_MAP_TYPE_PERCPU_ARRAY
) {
351 err
= bpf_percpu_array_copy(map
, key
, value
);
352 } else if (map
->map_type
== BPF_MAP_TYPE_STACK_TRACE
) {
353 err
= bpf_stackmap_copy(map
, key
, value
);
356 ptr
= map
->ops
->map_lookup_elem(map
, key
);
358 memcpy(value
, ptr
, value_size
);
360 err
= ptr
? 0 : -ENOENT
;
367 if (copy_to_user(uvalue
, value
, value_size
) != 0)
370 trace_bpf_map_lookup_elem(map
, ufd
, key
, value
);
382 #define BPF_MAP_UPDATE_ELEM_LAST_FIELD flags
384 static int map_update_elem(union bpf_attr
*attr
)
386 void __user
*ukey
= u64_to_user_ptr(attr
->key
);
387 void __user
*uvalue
= u64_to_user_ptr(attr
->value
);
388 int ufd
= attr
->map_fd
;
395 if (CHECK_ATTR(BPF_MAP_UPDATE_ELEM
))
399 map
= __bpf_map_get(f
);
404 key
= kmalloc(map
->key_size
, GFP_USER
);
409 if (copy_from_user(key
, ukey
, map
->key_size
) != 0)
412 if (map
->map_type
== BPF_MAP_TYPE_PERCPU_HASH
||
413 map
->map_type
== BPF_MAP_TYPE_LRU_PERCPU_HASH
||
414 map
->map_type
== BPF_MAP_TYPE_PERCPU_ARRAY
)
415 value_size
= round_up(map
->value_size
, 8) * num_possible_cpus();
417 value_size
= map
->value_size
;
420 value
= kmalloc(value_size
, GFP_USER
| __GFP_NOWARN
);
425 if (copy_from_user(value
, uvalue
, value_size
) != 0)
428 /* must increment bpf_prog_active to avoid kprobe+bpf triggering from
429 * inside bpf map update or delete otherwise deadlocks are possible
432 __this_cpu_inc(bpf_prog_active
);
433 if (map
->map_type
== BPF_MAP_TYPE_PERCPU_HASH
||
434 map
->map_type
== BPF_MAP_TYPE_LRU_PERCPU_HASH
) {
435 err
= bpf_percpu_hash_update(map
, key
, value
, attr
->flags
);
436 } else if (map
->map_type
== BPF_MAP_TYPE_PERCPU_ARRAY
) {
437 err
= bpf_percpu_array_update(map
, key
, value
, attr
->flags
);
438 } else if (map
->map_type
== BPF_MAP_TYPE_PERF_EVENT_ARRAY
||
439 map
->map_type
== BPF_MAP_TYPE_PROG_ARRAY
||
440 map
->map_type
== BPF_MAP_TYPE_CGROUP_ARRAY
) {
442 err
= bpf_fd_array_map_update_elem(map
, f
.file
, key
, value
,
447 err
= map
->ops
->map_update_elem(map
, key
, value
, attr
->flags
);
450 __this_cpu_dec(bpf_prog_active
);
454 trace_bpf_map_update_elem(map
, ufd
, key
, value
);
464 #define BPF_MAP_DELETE_ELEM_LAST_FIELD key
466 static int map_delete_elem(union bpf_attr
*attr
)
468 void __user
*ukey
= u64_to_user_ptr(attr
->key
);
469 int ufd
= attr
->map_fd
;
475 if (CHECK_ATTR(BPF_MAP_DELETE_ELEM
))
479 map
= __bpf_map_get(f
);
484 key
= kmalloc(map
->key_size
, GFP_USER
);
489 if (copy_from_user(key
, ukey
, map
->key_size
) != 0)
493 __this_cpu_inc(bpf_prog_active
);
495 err
= map
->ops
->map_delete_elem(map
, key
);
497 __this_cpu_dec(bpf_prog_active
);
501 trace_bpf_map_delete_elem(map
, ufd
, key
);
509 /* last field in 'union bpf_attr' used by this command */
510 #define BPF_MAP_GET_NEXT_KEY_LAST_FIELD next_key
512 static int map_get_next_key(union bpf_attr
*attr
)
514 void __user
*ukey
= u64_to_user_ptr(attr
->key
);
515 void __user
*unext_key
= u64_to_user_ptr(attr
->next_key
);
516 int ufd
= attr
->map_fd
;
518 void *key
, *next_key
;
522 if (CHECK_ATTR(BPF_MAP_GET_NEXT_KEY
))
526 map
= __bpf_map_get(f
);
531 key
= kmalloc(map
->key_size
, GFP_USER
);
536 if (copy_from_user(key
, ukey
, map
->key_size
) != 0)
540 next_key
= kmalloc(map
->key_size
, GFP_USER
);
545 err
= map
->ops
->map_get_next_key(map
, key
, next_key
);
551 if (copy_to_user(unext_key
, next_key
, map
->key_size
) != 0)
554 trace_bpf_map_next_key(map
, ufd
, key
, next_key
);
566 static LIST_HEAD(bpf_prog_types
);
568 static int find_prog_type(enum bpf_prog_type type
, struct bpf_prog
*prog
)
570 struct bpf_prog_type_list
*tl
;
572 list_for_each_entry(tl
, &bpf_prog_types
, list_node
) {
573 if (tl
->type
== type
) {
574 prog
->aux
->ops
= tl
->ops
;
583 void bpf_register_prog_type(struct bpf_prog_type_list
*tl
)
585 list_add(&tl
->list_node
, &bpf_prog_types
);
588 /* fixup insn->imm field of bpf_call instructions:
589 * if (insn->imm == BPF_FUNC_map_lookup_elem)
590 * insn->imm = bpf_map_lookup_elem - __bpf_call_base;
591 * else if (insn->imm == BPF_FUNC_map_update_elem)
592 * insn->imm = bpf_map_update_elem - __bpf_call_base;
595 * this function is called after eBPF program passed verification
597 static void fixup_bpf_calls(struct bpf_prog
*prog
)
599 const struct bpf_func_proto
*fn
;
602 for (i
= 0; i
< prog
->len
; i
++) {
603 struct bpf_insn
*insn
= &prog
->insnsi
[i
];
605 if (insn
->code
== (BPF_JMP
| BPF_CALL
)) {
606 /* we reach here when program has bpf_call instructions
607 * and it passed bpf_check(), means that
608 * ops->get_func_proto must have been supplied, check it
610 BUG_ON(!prog
->aux
->ops
->get_func_proto
);
612 if (insn
->imm
== BPF_FUNC_get_route_realm
)
613 prog
->dst_needed
= 1;
614 if (insn
->imm
== BPF_FUNC_get_prandom_u32
)
615 bpf_user_rnd_init_once();
616 if (insn
->imm
== BPF_FUNC_xdp_adjust_head
)
617 prog
->xdp_adjust_head
= 1;
618 if (insn
->imm
== BPF_FUNC_tail_call
) {
619 /* mark bpf_tail_call as different opcode
620 * to avoid conditional branch in
621 * interpeter for every normal call
622 * and to prevent accidental JITing by
623 * JIT compiler that doesn't support
631 fn
= prog
->aux
->ops
->get_func_proto(insn
->imm
);
632 /* all functions that have prototype and verifier allowed
633 * programs to call them, must be real in-kernel functions
636 insn
->imm
= fn
->func
- __bpf_call_base
;
641 /* drop refcnt on maps used by eBPF program and free auxilary data */
642 static void free_used_maps(struct bpf_prog_aux
*aux
)
646 for (i
= 0; i
< aux
->used_map_cnt
; i
++)
647 bpf_map_put(aux
->used_maps
[i
]);
649 kfree(aux
->used_maps
);
652 int __bpf_prog_charge(struct user_struct
*user
, u32 pages
)
654 unsigned long memlock_limit
= rlimit(RLIMIT_MEMLOCK
) >> PAGE_SHIFT
;
655 unsigned long user_bufs
;
658 user_bufs
= atomic_long_add_return(pages
, &user
->locked_vm
);
659 if (user_bufs
> memlock_limit
) {
660 atomic_long_sub(pages
, &user
->locked_vm
);
668 void __bpf_prog_uncharge(struct user_struct
*user
, u32 pages
)
671 atomic_long_sub(pages
, &user
->locked_vm
);
674 static int bpf_prog_charge_memlock(struct bpf_prog
*prog
)
676 struct user_struct
*user
= get_current_user();
679 ret
= __bpf_prog_charge(user
, prog
->pages
);
685 prog
->aux
->user
= user
;
689 static void bpf_prog_uncharge_memlock(struct bpf_prog
*prog
)
691 struct user_struct
*user
= prog
->aux
->user
;
693 __bpf_prog_uncharge(user
, prog
->pages
);
697 static void __bpf_prog_put_rcu(struct rcu_head
*rcu
)
699 struct bpf_prog_aux
*aux
= container_of(rcu
, struct bpf_prog_aux
, rcu
);
702 bpf_prog_uncharge_memlock(aux
->prog
);
703 bpf_prog_free(aux
->prog
);
706 void bpf_prog_put(struct bpf_prog
*prog
)
708 if (atomic_dec_and_test(&prog
->aux
->refcnt
)) {
709 trace_bpf_prog_put_rcu(prog
);
710 bpf_prog_kallsyms_del(prog
);
711 call_rcu(&prog
->aux
->rcu
, __bpf_prog_put_rcu
);
714 EXPORT_SYMBOL_GPL(bpf_prog_put
);
716 static int bpf_prog_release(struct inode
*inode
, struct file
*filp
)
718 struct bpf_prog
*prog
= filp
->private_data
;
724 #ifdef CONFIG_PROC_FS
725 static void bpf_prog_show_fdinfo(struct seq_file
*m
, struct file
*filp
)
727 const struct bpf_prog
*prog
= filp
->private_data
;
728 char prog_tag
[sizeof(prog
->tag
) * 2 + 1] = { };
730 bin2hex(prog_tag
, prog
->tag
, sizeof(prog
->tag
));
739 prog
->pages
* 1ULL << PAGE_SHIFT
);
743 static const struct file_operations bpf_prog_fops
= {
744 #ifdef CONFIG_PROC_FS
745 .show_fdinfo
= bpf_prog_show_fdinfo
,
747 .release
= bpf_prog_release
,
750 int bpf_prog_new_fd(struct bpf_prog
*prog
)
752 return anon_inode_getfd("bpf-prog", &bpf_prog_fops
, prog
,
756 static struct bpf_prog
*____bpf_prog_get(struct fd f
)
759 return ERR_PTR(-EBADF
);
760 if (f
.file
->f_op
!= &bpf_prog_fops
) {
762 return ERR_PTR(-EINVAL
);
765 return f
.file
->private_data
;
768 struct bpf_prog
*bpf_prog_add(struct bpf_prog
*prog
, int i
)
770 if (atomic_add_return(i
, &prog
->aux
->refcnt
) > BPF_MAX_REFCNT
) {
771 atomic_sub(i
, &prog
->aux
->refcnt
);
772 return ERR_PTR(-EBUSY
);
776 EXPORT_SYMBOL_GPL(bpf_prog_add
);
778 void bpf_prog_sub(struct bpf_prog
*prog
, int i
)
780 /* Only to be used for undoing previous bpf_prog_add() in some
781 * error path. We still know that another entity in our call
782 * path holds a reference to the program, thus atomic_sub() can
783 * be safely used in such cases!
785 WARN_ON(atomic_sub_return(i
, &prog
->aux
->refcnt
) == 0);
787 EXPORT_SYMBOL_GPL(bpf_prog_sub
);
789 struct bpf_prog
*bpf_prog_inc(struct bpf_prog
*prog
)
791 return bpf_prog_add(prog
, 1);
793 EXPORT_SYMBOL_GPL(bpf_prog_inc
);
795 static struct bpf_prog
*__bpf_prog_get(u32 ufd
, enum bpf_prog_type
*type
)
797 struct fd f
= fdget(ufd
);
798 struct bpf_prog
*prog
;
800 prog
= ____bpf_prog_get(f
);
803 if (type
&& prog
->type
!= *type
) {
804 prog
= ERR_PTR(-EINVAL
);
808 prog
= bpf_prog_inc(prog
);
814 struct bpf_prog
*bpf_prog_get(u32 ufd
)
816 return __bpf_prog_get(ufd
, NULL
);
819 struct bpf_prog
*bpf_prog_get_type(u32 ufd
, enum bpf_prog_type type
)
821 struct bpf_prog
*prog
= __bpf_prog_get(ufd
, &type
);
824 trace_bpf_prog_get_type(prog
);
827 EXPORT_SYMBOL_GPL(bpf_prog_get_type
);
829 /* last field in 'union bpf_attr' used by this command */
830 #define BPF_PROG_LOAD_LAST_FIELD kern_version
832 static int bpf_prog_load(union bpf_attr
*attr
)
834 enum bpf_prog_type type
= attr
->prog_type
;
835 struct bpf_prog
*prog
;
840 if (CHECK_ATTR(BPF_PROG_LOAD
))
843 /* copy eBPF program license from user space */
844 if (strncpy_from_user(license
, u64_to_user_ptr(attr
->license
),
845 sizeof(license
) - 1) < 0)
847 license
[sizeof(license
) - 1] = 0;
849 /* eBPF programs must be GPL compatible to use GPL-ed functions */
850 is_gpl
= license_is_gpl_compatible(license
);
852 if (attr
->insn_cnt
== 0 || attr
->insn_cnt
> BPF_MAXINSNS
)
855 if (type
== BPF_PROG_TYPE_KPROBE
&&
856 attr
->kern_version
!= LINUX_VERSION_CODE
)
859 if (type
!= BPF_PROG_TYPE_SOCKET_FILTER
&& !capable(CAP_SYS_ADMIN
))
862 /* plain bpf_prog allocation */
863 prog
= bpf_prog_alloc(bpf_prog_size(attr
->insn_cnt
), GFP_USER
);
867 err
= bpf_prog_charge_memlock(prog
);
869 goto free_prog_nouncharge
;
871 prog
->len
= attr
->insn_cnt
;
874 if (copy_from_user(prog
->insns
, u64_to_user_ptr(attr
->insns
),
875 bpf_prog_insn_size(prog
)) != 0)
878 prog
->orig_prog
= NULL
;
881 atomic_set(&prog
->aux
->refcnt
, 1);
882 prog
->gpl_compatible
= is_gpl
? 1 : 0;
884 /* find program type: socket_filter vs tracing_filter */
885 err
= find_prog_type(type
, prog
);
889 /* run eBPF verifier */
890 err
= bpf_check(&prog
, attr
);
894 /* fixup BPF_CALL->imm field */
895 fixup_bpf_calls(prog
);
897 /* eBPF program is ready to be JITed */
898 prog
= bpf_prog_select_runtime(prog
, &err
);
902 err
= bpf_prog_new_fd(prog
);
904 /* failed to allocate fd */
907 bpf_prog_kallsyms_add(prog
);
908 trace_bpf_prog_load(prog
, err
);
912 free_used_maps(prog
->aux
);
914 bpf_prog_uncharge_memlock(prog
);
915 free_prog_nouncharge
:
920 #define BPF_OBJ_LAST_FIELD bpf_fd
922 static int bpf_obj_pin(const union bpf_attr
*attr
)
924 if (CHECK_ATTR(BPF_OBJ
))
927 return bpf_obj_pin_user(attr
->bpf_fd
, u64_to_user_ptr(attr
->pathname
));
930 static int bpf_obj_get(const union bpf_attr
*attr
)
932 if (CHECK_ATTR(BPF_OBJ
) || attr
->bpf_fd
!= 0)
935 return bpf_obj_get_user(u64_to_user_ptr(attr
->pathname
));
938 #ifdef CONFIG_CGROUP_BPF
940 #define BPF_PROG_ATTACH_LAST_FIELD attach_flags
942 static int bpf_prog_attach(const union bpf_attr
*attr
)
944 enum bpf_prog_type ptype
;
945 struct bpf_prog
*prog
;
949 if (!capable(CAP_NET_ADMIN
))
952 if (CHECK_ATTR(BPF_PROG_ATTACH
))
955 if (attr
->attach_flags
& ~BPF_F_ALLOW_OVERRIDE
)
958 switch (attr
->attach_type
) {
959 case BPF_CGROUP_INET_INGRESS
:
960 case BPF_CGROUP_INET_EGRESS
:
961 ptype
= BPF_PROG_TYPE_CGROUP_SKB
;
963 case BPF_CGROUP_INET_SOCK_CREATE
:
964 ptype
= BPF_PROG_TYPE_CGROUP_SOCK
;
970 prog
= bpf_prog_get_type(attr
->attach_bpf_fd
, ptype
);
972 return PTR_ERR(prog
);
974 cgrp
= cgroup_get_from_fd(attr
->target_fd
);
977 return PTR_ERR(cgrp
);
980 ret
= cgroup_bpf_update(cgrp
, prog
, attr
->attach_type
,
981 attr
->attach_flags
& BPF_F_ALLOW_OVERRIDE
);
989 #define BPF_PROG_DETACH_LAST_FIELD attach_type
991 static int bpf_prog_detach(const union bpf_attr
*attr
)
996 if (!capable(CAP_NET_ADMIN
))
999 if (CHECK_ATTR(BPF_PROG_DETACH
))
1002 switch (attr
->attach_type
) {
1003 case BPF_CGROUP_INET_INGRESS
:
1004 case BPF_CGROUP_INET_EGRESS
:
1005 case BPF_CGROUP_INET_SOCK_CREATE
:
1006 cgrp
= cgroup_get_from_fd(attr
->target_fd
);
1008 return PTR_ERR(cgrp
);
1010 ret
= cgroup_bpf_update(cgrp
, NULL
, attr
->attach_type
, false);
1020 #endif /* CONFIG_CGROUP_BPF */
1022 SYSCALL_DEFINE3(bpf
, int, cmd
, union bpf_attr __user
*, uattr
, unsigned int, size
)
1024 union bpf_attr attr
= {};
1027 if (!capable(CAP_SYS_ADMIN
) && sysctl_unprivileged_bpf_disabled
)
1030 if (!access_ok(VERIFY_READ
, uattr
, 1))
1033 if (size
> PAGE_SIZE
) /* silly large */
1036 /* If we're handed a bigger struct than we know of,
1037 * ensure all the unknown bits are 0 - i.e. new
1038 * user-space does not rely on any kernel feature
1039 * extensions we dont know about yet.
1041 if (size
> sizeof(attr
)) {
1042 unsigned char __user
*addr
;
1043 unsigned char __user
*end
;
1046 addr
= (void __user
*)uattr
+ sizeof(attr
);
1047 end
= (void __user
*)uattr
+ size
;
1049 for (; addr
< end
; addr
++) {
1050 err
= get_user(val
, addr
);
1056 size
= sizeof(attr
);
1059 /* copy attributes from user space, may be less than sizeof(bpf_attr) */
1060 if (copy_from_user(&attr
, uattr
, size
) != 0)
1064 case BPF_MAP_CREATE
:
1065 err
= map_create(&attr
);
1067 case BPF_MAP_LOOKUP_ELEM
:
1068 err
= map_lookup_elem(&attr
);
1070 case BPF_MAP_UPDATE_ELEM
:
1071 err
= map_update_elem(&attr
);
1073 case BPF_MAP_DELETE_ELEM
:
1074 err
= map_delete_elem(&attr
);
1076 case BPF_MAP_GET_NEXT_KEY
:
1077 err
= map_get_next_key(&attr
);
1080 err
= bpf_prog_load(&attr
);
1083 err
= bpf_obj_pin(&attr
);
1086 err
= bpf_obj_get(&attr
);
1089 #ifdef CONFIG_CGROUP_BPF
1090 case BPF_PROG_ATTACH
:
1091 err
= bpf_prog_attach(&attr
);
1093 case BPF_PROG_DETACH
:
1094 err
= bpf_prog_detach(&attr
);