2 * linux/kernel/seccomp.c
4 * Copyright 2004-2005 Andrea Arcangeli <andrea@cpushare.com>
6 * Copyright (C) 2012 Google, Inc.
7 * Will Drewry <wad@chromium.org>
9 * This defines a simple but solid secure-computing facility.
11 * Mode 1 uses a fixed list of allowed system calls.
12 * Mode 2 allows user-defined system call filters in the form
13 * of Berkeley Packet Filters/Linux Socket Filters.
16 #include <linux/refcount.h>
17 #include <linux/audit.h>
18 #include <linux/compat.h>
19 #include <linux/coredump.h>
20 #include <linux/kmemleak.h>
21 #include <linux/sched.h>
22 #include <linux/sched/task_stack.h>
23 #include <linux/seccomp.h>
24 #include <linux/slab.h>
25 #include <linux/syscalls.h>
26 #include <linux/sysctl.h>
28 #ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
29 #include <asm/syscall.h>
32 #ifdef CONFIG_SECCOMP_FILTER
33 #include <linux/filter.h>
34 #include <linux/pid.h>
35 #include <linux/ptrace.h>
36 #include <linux/security.h>
37 #include <linux/tracehook.h>
38 #include <linux/uaccess.h>
41 * struct seccomp_filter - container for seccomp BPF programs
43 * @usage: reference count to manage the object lifetime.
44 * get/put helpers should be used when accessing an instance
45 * outside of a lifetime-guarded section. In general, this
46 * is only needed for handling filters shared across tasks.
47 * @prev: points to a previously installed, or inherited, filter
48 * @prog: the BPF program to evaluate
50 * seccomp_filter objects are organized in a tree linked via the @prev
51 * pointer. For any task, it appears to be a singly-linked list starting
52 * with current->seccomp.filter, the most recently attached or inherited filter.
53 * However, multiple filters may share a @prev node, by way of fork(), which
54 * results in a unidirectional tree existing in memory. This is similar to
55 * how namespaces work.
57 * seccomp_filter objects should never be modified after being attached
58 * to a task_struct (other than @usage).
60 struct seccomp_filter
{
62 struct seccomp_filter
*prev
;
63 struct bpf_prog
*prog
;
66 /* Limit any path through the tree to 256KB worth of instructions. */
67 #define MAX_INSNS_PER_PATH ((1 << 18) / sizeof(struct sock_filter))
70 * Endianness is explicitly ignored and left for BPF program authors to manage
71 * as per the specific architecture.
73 static void populate_seccomp_data(struct seccomp_data
*sd
)
75 struct task_struct
*task
= current
;
76 struct pt_regs
*regs
= task_pt_regs(task
);
77 unsigned long args
[6];
79 sd
->nr
= syscall_get_nr(task
, regs
);
80 sd
->arch
= syscall_get_arch();
81 syscall_get_arguments(task
, regs
, 0, 6, args
);
82 sd
->args
[0] = args
[0];
83 sd
->args
[1] = args
[1];
84 sd
->args
[2] = args
[2];
85 sd
->args
[3] = args
[3];
86 sd
->args
[4] = args
[4];
87 sd
->args
[5] = args
[5];
88 sd
->instruction_pointer
= KSTK_EIP(task
);
92 * seccomp_check_filter - verify seccomp filter code
93 * @filter: filter to verify
94 * @flen: length of filter
96 * Takes a previously checked filter (by bpf_check_classic) and
97 * redirects all filter code that loads struct sk_buff data
98 * and related data through seccomp_bpf_load. It also
99 * enforces length and alignment checking of those loads.
101 * Returns 0 if the rule set is legal or -EINVAL if not.
103 static int seccomp_check_filter(struct sock_filter
*filter
, unsigned int flen
)
106 for (pc
= 0; pc
< flen
; pc
++) {
107 struct sock_filter
*ftest
= &filter
[pc
];
108 u16 code
= ftest
->code
;
112 case BPF_LD
| BPF_W
| BPF_ABS
:
113 ftest
->code
= BPF_LDX
| BPF_W
| BPF_ABS
;
114 /* 32-bit aligned and not out of bounds. */
115 if (k
>= sizeof(struct seccomp_data
) || k
& 3)
118 case BPF_LD
| BPF_W
| BPF_LEN
:
119 ftest
->code
= BPF_LD
| BPF_IMM
;
120 ftest
->k
= sizeof(struct seccomp_data
);
122 case BPF_LDX
| BPF_W
| BPF_LEN
:
123 ftest
->code
= BPF_LDX
| BPF_IMM
;
124 ftest
->k
= sizeof(struct seccomp_data
);
126 /* Explicitly include allowed calls. */
127 case BPF_RET
| BPF_K
:
128 case BPF_RET
| BPF_A
:
129 case BPF_ALU
| BPF_ADD
| BPF_K
:
130 case BPF_ALU
| BPF_ADD
| BPF_X
:
131 case BPF_ALU
| BPF_SUB
| BPF_K
:
132 case BPF_ALU
| BPF_SUB
| BPF_X
:
133 case BPF_ALU
| BPF_MUL
| BPF_K
:
134 case BPF_ALU
| BPF_MUL
| BPF_X
:
135 case BPF_ALU
| BPF_DIV
| BPF_K
:
136 case BPF_ALU
| BPF_DIV
| BPF_X
:
137 case BPF_ALU
| BPF_AND
| BPF_K
:
138 case BPF_ALU
| BPF_AND
| BPF_X
:
139 case BPF_ALU
| BPF_OR
| BPF_K
:
140 case BPF_ALU
| BPF_OR
| BPF_X
:
141 case BPF_ALU
| BPF_XOR
| BPF_K
:
142 case BPF_ALU
| BPF_XOR
| BPF_X
:
143 case BPF_ALU
| BPF_LSH
| BPF_K
:
144 case BPF_ALU
| BPF_LSH
| BPF_X
:
145 case BPF_ALU
| BPF_RSH
| BPF_K
:
146 case BPF_ALU
| BPF_RSH
| BPF_X
:
147 case BPF_ALU
| BPF_NEG
:
148 case BPF_LD
| BPF_IMM
:
149 case BPF_LDX
| BPF_IMM
:
150 case BPF_MISC
| BPF_TAX
:
151 case BPF_MISC
| BPF_TXA
:
152 case BPF_LD
| BPF_MEM
:
153 case BPF_LDX
| BPF_MEM
:
156 case BPF_JMP
| BPF_JA
:
157 case BPF_JMP
| BPF_JEQ
| BPF_K
:
158 case BPF_JMP
| BPF_JEQ
| BPF_X
:
159 case BPF_JMP
| BPF_JGE
| BPF_K
:
160 case BPF_JMP
| BPF_JGE
| BPF_X
:
161 case BPF_JMP
| BPF_JGT
| BPF_K
:
162 case BPF_JMP
| BPF_JGT
| BPF_X
:
163 case BPF_JMP
| BPF_JSET
| BPF_K
:
164 case BPF_JMP
| BPF_JSET
| BPF_X
:
174 * seccomp_run_filters - evaluates all seccomp filters against @sd
175 * @sd: optional seccomp data to be passed to filters
176 * @match: stores struct seccomp_filter that resulted in the return value,
177 * unless filter returned SECCOMP_RET_ALLOW, in which case it will
180 * Returns valid seccomp BPF response codes.
182 static u32
seccomp_run_filters(const struct seccomp_data
*sd
,
183 struct seccomp_filter
**match
)
185 struct seccomp_data sd_local
;
186 u32 ret
= SECCOMP_RET_ALLOW
;
187 /* Make sure cross-thread synced filter points somewhere sane. */
188 struct seccomp_filter
*f
=
189 lockless_dereference(current
->seccomp
.filter
);
191 /* Ensure unexpected behavior doesn't result in failing open. */
192 if (unlikely(WARN_ON(f
== NULL
)))
193 return SECCOMP_RET_KILL
;
196 populate_seccomp_data(&sd_local
);
201 * All filters in the list are evaluated and the lowest BPF return
202 * value always takes priority (ignoring the DATA).
204 for (; f
; f
= f
->prev
) {
205 u32 cur_ret
= BPF_PROG_RUN(f
->prog
, sd
);
207 if ((cur_ret
& SECCOMP_RET_ACTION
) < (ret
& SECCOMP_RET_ACTION
)) {
214 #endif /* CONFIG_SECCOMP_FILTER */
216 static inline bool seccomp_may_assign_mode(unsigned long seccomp_mode
)
218 assert_spin_locked(¤t
->sighand
->siglock
);
220 if (current
->seccomp
.mode
&& current
->seccomp
.mode
!= seccomp_mode
)
226 static inline void seccomp_assign_mode(struct task_struct
*task
,
227 unsigned long seccomp_mode
)
229 assert_spin_locked(&task
->sighand
->siglock
);
231 task
->seccomp
.mode
= seccomp_mode
;
233 * Make sure TIF_SECCOMP cannot be set before the mode (and
236 smp_mb__before_atomic();
237 set_tsk_thread_flag(task
, TIF_SECCOMP
);
240 #ifdef CONFIG_SECCOMP_FILTER
241 /* Returns 1 if the parent is an ancestor of the child. */
242 static int is_ancestor(struct seccomp_filter
*parent
,
243 struct seccomp_filter
*child
)
245 /* NULL is the root ancestor. */
248 for (; child
; child
= child
->prev
)
255 * seccomp_can_sync_threads: checks if all threads can be synchronized
257 * Expects sighand and cred_guard_mutex locks to be held.
259 * Returns 0 on success, -ve on error, or the pid of a thread which was
260 * either not in the correct seccomp mode or it did not have an ancestral
263 static inline pid_t
seccomp_can_sync_threads(void)
265 struct task_struct
*thread
, *caller
;
267 BUG_ON(!mutex_is_locked(¤t
->signal
->cred_guard_mutex
));
268 assert_spin_locked(¤t
->sighand
->siglock
);
270 /* Validate all threads being eligible for synchronization. */
272 for_each_thread(caller
, thread
) {
275 /* Skip current, since it is initiating the sync. */
276 if (thread
== caller
)
279 if (thread
->seccomp
.mode
== SECCOMP_MODE_DISABLED
||
280 (thread
->seccomp
.mode
== SECCOMP_MODE_FILTER
&&
281 is_ancestor(thread
->seccomp
.filter
,
282 caller
->seccomp
.filter
)))
285 /* Return the first thread that cannot be synchronized. */
286 failed
= task_pid_vnr(thread
);
287 /* If the pid cannot be resolved, then return -ESRCH */
288 if (unlikely(WARN_ON(failed
== 0)))
297 * seccomp_sync_threads: sets all threads to use current's filter
299 * Expects sighand and cred_guard_mutex locks to be held, and for
300 * seccomp_can_sync_threads() to have returned success already
301 * without dropping the locks.
304 static inline void seccomp_sync_threads(void)
306 struct task_struct
*thread
, *caller
;
308 BUG_ON(!mutex_is_locked(¤t
->signal
->cred_guard_mutex
));
309 assert_spin_locked(¤t
->sighand
->siglock
);
311 /* Synchronize all threads. */
313 for_each_thread(caller
, thread
) {
314 /* Skip current, since it needs no changes. */
315 if (thread
== caller
)
318 /* Get a task reference for the new leaf node. */
319 get_seccomp_filter(caller
);
321 * Drop the task reference to the shared ancestor since
322 * current's path will hold a reference. (This also
323 * allows a put before the assignment.)
325 put_seccomp_filter(thread
);
326 smp_store_release(&thread
->seccomp
.filter
,
327 caller
->seccomp
.filter
);
330 * Don't let an unprivileged task work around
331 * the no_new_privs restriction by creating
332 * a thread that sets it up, enters seccomp,
335 if (task_no_new_privs(caller
))
336 task_set_no_new_privs(thread
);
339 * Opt the other thread into seccomp if needed.
340 * As threads are considered to be trust-realm
341 * equivalent (see ptrace_may_access), it is safe to
342 * allow one thread to transition the other.
344 if (thread
->seccomp
.mode
== SECCOMP_MODE_DISABLED
)
345 seccomp_assign_mode(thread
, SECCOMP_MODE_FILTER
);
350 * seccomp_prepare_filter: Prepares a seccomp filter for use.
351 * @fprog: BPF program to install
353 * Returns filter on success or an ERR_PTR on failure.
355 static struct seccomp_filter
*seccomp_prepare_filter(struct sock_fprog
*fprog
)
357 struct seccomp_filter
*sfilter
;
359 const bool save_orig
= IS_ENABLED(CONFIG_CHECKPOINT_RESTORE
);
361 if (fprog
->len
== 0 || fprog
->len
> BPF_MAXINSNS
)
362 return ERR_PTR(-EINVAL
);
364 BUG_ON(INT_MAX
/ fprog
->len
< sizeof(struct sock_filter
));
367 * Installing a seccomp filter requires that the task has
368 * CAP_SYS_ADMIN in its namespace or be running with no_new_privs.
369 * This avoids scenarios where unprivileged tasks can affect the
370 * behavior of privileged children.
372 if (!task_no_new_privs(current
) &&
373 security_capable_noaudit(current_cred(), current_user_ns(),
375 return ERR_PTR(-EACCES
);
377 /* Allocate a new seccomp_filter */
378 sfilter
= kzalloc(sizeof(*sfilter
), GFP_KERNEL
| __GFP_NOWARN
);
380 return ERR_PTR(-ENOMEM
);
382 ret
= bpf_prog_create_from_user(&sfilter
->prog
, fprog
,
383 seccomp_check_filter
, save_orig
);
389 refcount_set(&sfilter
->usage
, 1);
395 * seccomp_prepare_user_filter - prepares a user-supplied sock_fprog
396 * @user_filter: pointer to the user data containing a sock_fprog.
398 * Returns 0 on success and non-zero otherwise.
400 static struct seccomp_filter
*
401 seccomp_prepare_user_filter(const char __user
*user_filter
)
403 struct sock_fprog fprog
;
404 struct seccomp_filter
*filter
= ERR_PTR(-EFAULT
);
407 if (in_compat_syscall()) {
408 struct compat_sock_fprog fprog32
;
409 if (copy_from_user(&fprog32
, user_filter
, sizeof(fprog32
)))
411 fprog
.len
= fprog32
.len
;
412 fprog
.filter
= compat_ptr(fprog32
.filter
);
413 } else /* falls through to the if below. */
415 if (copy_from_user(&fprog
, user_filter
, sizeof(fprog
)))
417 filter
= seccomp_prepare_filter(&fprog
);
423 * seccomp_attach_filter: validate and attach filter
424 * @flags: flags to change filter behavior
425 * @filter: seccomp filter to add to the current process
427 * Caller must be holding current->sighand->siglock lock.
429 * Returns 0 on success, -ve on error.
431 static long seccomp_attach_filter(unsigned int flags
,
432 struct seccomp_filter
*filter
)
434 unsigned long total_insns
;
435 struct seccomp_filter
*walker
;
437 assert_spin_locked(¤t
->sighand
->siglock
);
439 /* Validate resulting filter length. */
440 total_insns
= filter
->prog
->len
;
441 for (walker
= current
->seccomp
.filter
; walker
; walker
= walker
->prev
)
442 total_insns
+= walker
->prog
->len
+ 4; /* 4 instr penalty */
443 if (total_insns
> MAX_INSNS_PER_PATH
)
446 /* If thread sync has been requested, check that it is possible. */
447 if (flags
& SECCOMP_FILTER_FLAG_TSYNC
) {
450 ret
= seccomp_can_sync_threads();
456 * If there is an existing filter, make it the prev and don't drop its
459 filter
->prev
= current
->seccomp
.filter
;
460 current
->seccomp
.filter
= filter
;
462 /* Now that the new filter is in place, synchronize to all threads. */
463 if (flags
& SECCOMP_FILTER_FLAG_TSYNC
)
464 seccomp_sync_threads();
469 /* get_seccomp_filter - increments the reference count of the filter on @tsk */
470 void get_seccomp_filter(struct task_struct
*tsk
)
472 struct seccomp_filter
*orig
= tsk
->seccomp
.filter
;
475 /* Reference count is bounded by the number of total processes. */
476 refcount_inc(&orig
->usage
);
479 static inline void seccomp_filter_free(struct seccomp_filter
*filter
)
482 bpf_prog_destroy(filter
->prog
);
487 /* put_seccomp_filter - decrements the ref count of tsk->seccomp.filter */
488 void put_seccomp_filter(struct task_struct
*tsk
)
490 struct seccomp_filter
*orig
= tsk
->seccomp
.filter
;
491 /* Clean up single-reference branches iteratively. */
492 while (orig
&& refcount_dec_and_test(&orig
->usage
)) {
493 struct seccomp_filter
*freeme
= orig
;
495 seccomp_filter_free(freeme
);
499 static void seccomp_init_siginfo(siginfo_t
*info
, int syscall
, int reason
)
501 memset(info
, 0, sizeof(*info
));
502 info
->si_signo
= SIGSYS
;
503 info
->si_code
= SYS_SECCOMP
;
504 info
->si_call_addr
= (void __user
*)KSTK_EIP(current
);
505 info
->si_errno
= reason
;
506 info
->si_arch
= syscall_get_arch();
507 info
->si_syscall
= syscall
;
511 * seccomp_send_sigsys - signals the task to allow in-process syscall emulation
512 * @syscall: syscall number to send to userland
513 * @reason: filter-supplied reason code to send to userland (via si_errno)
515 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
517 static void seccomp_send_sigsys(int syscall
, int reason
)
520 seccomp_init_siginfo(&info
, syscall
, reason
);
521 force_sig_info(SIGSYS
, &info
, current
);
523 #endif /* CONFIG_SECCOMP_FILTER */
526 * Secure computing mode 1 allows only read/write/exit/sigreturn.
527 * To be fully secure this must be combined with rlimit
528 * to limit the stack allocations too.
530 static const int mode1_syscalls
[] = {
531 __NR_seccomp_read
, __NR_seccomp_write
, __NR_seccomp_exit
, __NR_seccomp_sigreturn
,
532 0, /* null terminated */
535 static void __secure_computing_strict(int this_syscall
)
537 const int *syscall_whitelist
= mode1_syscalls
;
539 if (in_compat_syscall())
540 syscall_whitelist
= get_compat_mode1_syscalls();
543 if (*syscall_whitelist
== this_syscall
)
545 } while (*++syscall_whitelist
);
550 audit_seccomp(this_syscall
, SIGKILL
, SECCOMP_RET_KILL
);
554 #ifndef CONFIG_HAVE_ARCH_SECCOMP_FILTER
555 void secure_computing_strict(int this_syscall
)
557 int mode
= current
->seccomp
.mode
;
559 if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE
) &&
560 unlikely(current
->ptrace
& PT_SUSPEND_SECCOMP
))
563 if (mode
== SECCOMP_MODE_DISABLED
)
565 else if (mode
== SECCOMP_MODE_STRICT
)
566 __secure_computing_strict(this_syscall
);
572 #ifdef CONFIG_SECCOMP_FILTER
573 static int __seccomp_filter(int this_syscall
, const struct seccomp_data
*sd
,
574 const bool recheck_after_trace
)
576 u32 filter_ret
, action
;
577 struct seccomp_filter
*match
= NULL
;
581 * Make sure that any changes to mode from another thread have
582 * been seen after TIF_SECCOMP was seen.
586 filter_ret
= seccomp_run_filters(sd
, &match
);
587 data
= filter_ret
& SECCOMP_RET_DATA
;
588 action
= filter_ret
& SECCOMP_RET_ACTION
;
591 case SECCOMP_RET_ERRNO
:
592 /* Set low-order bits as an errno, capped at MAX_ERRNO. */
593 if (data
> MAX_ERRNO
)
595 syscall_set_return_value(current
, task_pt_regs(current
),
599 case SECCOMP_RET_TRAP
:
600 /* Show the handler the original registers. */
601 syscall_rollback(current
, task_pt_regs(current
));
602 /* Let the filter pass back 16 bits of data. */
603 seccomp_send_sigsys(this_syscall
, data
);
606 case SECCOMP_RET_TRACE
:
607 /* We've been put in this state by the ptracer already. */
608 if (recheck_after_trace
)
611 /* ENOSYS these calls if there is no tracer attached. */
612 if (!ptrace_event_enabled(current
, PTRACE_EVENT_SECCOMP
)) {
613 syscall_set_return_value(current
,
614 task_pt_regs(current
),
619 /* Allow the BPF to provide the event message */
620 ptrace_event(PTRACE_EVENT_SECCOMP
, data
);
622 * The delivery of a fatal signal during event
623 * notification may silently skip tracer notification,
624 * which could leave us with a potentially unmodified
625 * syscall that the tracer would have liked to have
626 * changed. Since the process is about to die, we just
627 * force the syscall to be skipped and let the signal
628 * kill the process and correctly handle any tracer exit
631 if (fatal_signal_pending(current
))
633 /* Check if the tracer forced the syscall to be skipped. */
634 this_syscall
= syscall_get_nr(current
, task_pt_regs(current
));
635 if (this_syscall
< 0)
639 * Recheck the syscall, since it may have changed. This
640 * intentionally uses a NULL struct seccomp_data to force
641 * a reload of all registers. This does not goto skip since
642 * a skip would have already been reported.
644 if (__seccomp_filter(this_syscall
, NULL
, true))
649 case SECCOMP_RET_ALLOW
:
651 * Note that the "match" filter will always be NULL for
652 * this action since SECCOMP_RET_ALLOW is the starting
653 * state in seccomp_run_filters().
657 case SECCOMP_RET_KILL
:
659 audit_seccomp(this_syscall
, SIGSYS
, action
);
660 /* Dump core only if this is the last remaining thread. */
661 if (get_nr_threads(current
) == 1) {
664 /* Show the original registers in the dump. */
665 syscall_rollback(current
, task_pt_regs(current
));
666 /* Trigger a manual coredump since do_exit skips it. */
667 seccomp_init_siginfo(&info
, this_syscall
, data
);
676 audit_seccomp(this_syscall
, 0, action
);
680 static int __seccomp_filter(int this_syscall
, const struct seccomp_data
*sd
,
681 const bool recheck_after_trace
)
687 int __secure_computing(const struct seccomp_data
*sd
)
689 int mode
= current
->seccomp
.mode
;
692 if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE
) &&
693 unlikely(current
->ptrace
& PT_SUSPEND_SECCOMP
))
696 this_syscall
= sd
? sd
->nr
:
697 syscall_get_nr(current
, task_pt_regs(current
));
700 case SECCOMP_MODE_STRICT
:
701 __secure_computing_strict(this_syscall
); /* may call do_exit */
703 case SECCOMP_MODE_FILTER
:
704 return __seccomp_filter(this_syscall
, sd
, false);
709 #endif /* CONFIG_HAVE_ARCH_SECCOMP_FILTER */
711 long prctl_get_seccomp(void)
713 return current
->seccomp
.mode
;
717 * seccomp_set_mode_strict: internal function for setting strict seccomp
719 * Once current->seccomp.mode is non-zero, it may not be changed.
721 * Returns 0 on success or -EINVAL on failure.
723 static long seccomp_set_mode_strict(void)
725 const unsigned long seccomp_mode
= SECCOMP_MODE_STRICT
;
728 spin_lock_irq(¤t
->sighand
->siglock
);
730 if (!seccomp_may_assign_mode(seccomp_mode
))
736 seccomp_assign_mode(current
, seccomp_mode
);
740 spin_unlock_irq(¤t
->sighand
->siglock
);
745 #ifdef CONFIG_SECCOMP_FILTER
747 * seccomp_set_mode_filter: internal function for setting seccomp filter
748 * @flags: flags to change filter behavior
749 * @filter: struct sock_fprog containing filter
751 * This function may be called repeatedly to install additional filters.
752 * Every filter successfully installed will be evaluated (in reverse order)
753 * for each system call the task makes.
755 * Once current->seccomp.mode is non-zero, it may not be changed.
757 * Returns 0 on success or -EINVAL on failure.
759 static long seccomp_set_mode_filter(unsigned int flags
,
760 const char __user
*filter
)
762 const unsigned long seccomp_mode
= SECCOMP_MODE_FILTER
;
763 struct seccomp_filter
*prepared
= NULL
;
766 /* Validate flags. */
767 if (flags
& ~SECCOMP_FILTER_FLAG_MASK
)
770 /* Prepare the new filter before holding any locks. */
771 prepared
= seccomp_prepare_user_filter(filter
);
772 if (IS_ERR(prepared
))
773 return PTR_ERR(prepared
);
776 * Make sure we cannot change seccomp or nnp state via TSYNC
777 * while another thread is in the middle of calling exec.
779 if (flags
& SECCOMP_FILTER_FLAG_TSYNC
&&
780 mutex_lock_killable(¤t
->signal
->cred_guard_mutex
))
783 spin_lock_irq(¤t
->sighand
->siglock
);
785 if (!seccomp_may_assign_mode(seccomp_mode
))
788 ret
= seccomp_attach_filter(flags
, prepared
);
791 /* Do not free the successfully attached filter. */
794 seccomp_assign_mode(current
, seccomp_mode
);
796 spin_unlock_irq(¤t
->sighand
->siglock
);
797 if (flags
& SECCOMP_FILTER_FLAG_TSYNC
)
798 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
800 seccomp_filter_free(prepared
);
804 static inline long seccomp_set_mode_filter(unsigned int flags
,
805 const char __user
*filter
)
811 static long seccomp_get_action_avail(const char __user
*uaction
)
815 if (copy_from_user(&action
, uaction
, sizeof(action
)))
819 case SECCOMP_RET_KILL
:
820 case SECCOMP_RET_TRAP
:
821 case SECCOMP_RET_ERRNO
:
822 case SECCOMP_RET_TRACE
:
823 case SECCOMP_RET_ALLOW
:
832 /* Common entry point for both prctl and syscall. */
833 static long do_seccomp(unsigned int op
, unsigned int flags
,
834 const char __user
*uargs
)
837 case SECCOMP_SET_MODE_STRICT
:
838 if (flags
!= 0 || uargs
!= NULL
)
840 return seccomp_set_mode_strict();
841 case SECCOMP_SET_MODE_FILTER
:
842 return seccomp_set_mode_filter(flags
, uargs
);
843 case SECCOMP_GET_ACTION_AVAIL
:
847 return seccomp_get_action_avail(uargs
);
853 SYSCALL_DEFINE3(seccomp
, unsigned int, op
, unsigned int, flags
,
854 const char __user
*, uargs
)
856 return do_seccomp(op
, flags
, uargs
);
860 * prctl_set_seccomp: configures current->seccomp.mode
861 * @seccomp_mode: requested mode to use
862 * @filter: optional struct sock_fprog for use with SECCOMP_MODE_FILTER
864 * Returns 0 on success or -EINVAL on failure.
866 long prctl_set_seccomp(unsigned long seccomp_mode
, char __user
*filter
)
871 switch (seccomp_mode
) {
872 case SECCOMP_MODE_STRICT
:
873 op
= SECCOMP_SET_MODE_STRICT
;
875 * Setting strict mode through prctl always ignored filter,
876 * so make sure it is always NULL here to pass the internal
877 * check in do_seccomp().
881 case SECCOMP_MODE_FILTER
:
882 op
= SECCOMP_SET_MODE_FILTER
;
889 /* prctl interface doesn't have flags, so they are always zero. */
890 return do_seccomp(op
, 0, uargs
);
893 #if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
894 long seccomp_get_filter(struct task_struct
*task
, unsigned long filter_off
,
897 struct seccomp_filter
*filter
;
898 struct sock_fprog_kern
*fprog
;
900 unsigned long count
= 0;
902 if (!capable(CAP_SYS_ADMIN
) ||
903 current
->seccomp
.mode
!= SECCOMP_MODE_DISABLED
) {
907 spin_lock_irq(&task
->sighand
->siglock
);
908 if (task
->seccomp
.mode
!= SECCOMP_MODE_FILTER
) {
913 filter
= task
->seccomp
.filter
;
915 filter
= filter
->prev
;
919 if (filter_off
>= count
) {
925 filter
= task
->seccomp
.filter
;
926 while (filter
&& count
> 1) {
927 filter
= filter
->prev
;
931 if (WARN_ON(count
!= 1 || !filter
)) {
932 /* The filter tree shouldn't shrink while we're using it. */
937 fprog
= filter
->prog
->orig_prog
;
939 /* This must be a new non-cBPF filter, since we save
940 * every cBPF filter's orig_prog above when
941 * CONFIG_CHECKPOINT_RESTORE is enabled.
951 get_seccomp_filter(task
);
952 spin_unlock_irq(&task
->sighand
->siglock
);
954 if (copy_to_user(data
, fprog
->filter
, bpf_classic_proglen(fprog
)))
957 put_seccomp_filter(task
);
961 spin_unlock_irq(&task
->sighand
->siglock
);
968 /* Human readable action names for friendly sysctl interaction */
969 #define SECCOMP_RET_KILL_NAME "kill"
970 #define SECCOMP_RET_TRAP_NAME "trap"
971 #define SECCOMP_RET_ERRNO_NAME "errno"
972 #define SECCOMP_RET_TRACE_NAME "trace"
973 #define SECCOMP_RET_ALLOW_NAME "allow"
975 static const char seccomp_actions_avail
[] = SECCOMP_RET_KILL_NAME
" "
976 SECCOMP_RET_TRAP_NAME
" "
977 SECCOMP_RET_ERRNO_NAME
" "
978 SECCOMP_RET_TRACE_NAME
" "
979 SECCOMP_RET_ALLOW_NAME
;
981 static struct ctl_path seccomp_sysctl_path
[] = {
982 { .procname
= "kernel", },
983 { .procname
= "seccomp", },
987 static struct ctl_table seccomp_sysctl_table
[] = {
989 .procname
= "actions_avail",
990 .data
= (void *) &seccomp_actions_avail
,
991 .maxlen
= sizeof(seccomp_actions_avail
),
993 .proc_handler
= proc_dostring
,
998 static int __init
seccomp_sysctl_init(void)
1000 struct ctl_table_header
*hdr
;
1002 hdr
= register_sysctl_paths(seccomp_sysctl_path
, seccomp_sysctl_table
);
1004 pr_warn("seccomp: sysctl registration failed\n");
1006 kmemleak_not_leak(hdr
);
1011 device_initcall(seccomp_sysctl_init
)
1013 #endif /* CONFIG_SYSCTL */