1 // SPDX-License-Identifier: GPL-2.0
3 * linux/kernel/seccomp.c
5 * Copyright 2004-2005 Andrea Arcangeli <andrea@cpushare.com>
7 * Copyright (C) 2012 Google, Inc.
8 * Will Drewry <wad@chromium.org>
10 * This defines a simple but solid secure-computing facility.
12 * Mode 1 uses a fixed list of allowed system calls.
13 * Mode 2 allows user-defined system call filters in the form
14 * of Berkeley Packet Filters/Linux Socket Filters.
17 #include <linux/refcount.h>
18 #include <linux/audit.h>
19 #include <linux/compat.h>
20 #include <linux/coredump.h>
21 #include <linux/kmemleak.h>
22 #include <linux/nospec.h>
23 #include <linux/prctl.h>
24 #include <linux/sched.h>
25 #include <linux/sched/task_stack.h>
26 #include <linux/seccomp.h>
27 #include <linux/slab.h>
28 #include <linux/syscalls.h>
29 #include <linux/sysctl.h>
31 #ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
32 #include <asm/syscall.h>
35 #ifdef CONFIG_SECCOMP_FILTER
36 #include <linux/file.h>
37 #include <linux/filter.h>
38 #include <linux/pid.h>
39 #include <linux/ptrace.h>
40 #include <linux/security.h>
41 #include <linux/tracehook.h>
42 #include <linux/uaccess.h>
43 #include <linux/anon_inodes.h>
48 SECCOMP_NOTIFY_REPLIED,
51 struct seccomp_knotif {
52 /* The struct pid of the task whose filter triggered the notification */
53 struct task_struct *task;
55 /* The "cookie" for this request; this is unique for this filter. */
59 * The seccomp data. This pointer is valid the entire time this
60 * notification is active, since it comes from __seccomp_filter which
61 * eclipses the entire lifecycle here.
63 const struct seccomp_data *data;
66 * Notification states. When SECCOMP_RET_USER_NOTIF is returned, a
67 * struct seccomp_knotif is created and starts out in INIT. Once the
68 * handler reads the notification off of an FD, it transitions to SENT.
69 * If a signal is received the state transitions back to INIT and
70 * another message is sent. When the userspace handler replies, state
71 * transitions to REPLIED.
73 enum notify_state state;
75 /* The return values, only valid when in SECCOMP_NOTIFY_REPLIED */
79 /* Signals when this has entered SECCOMP_NOTIFY_REPLIED */
80 struct completion ready;
82 struct list_head list;
86 * struct notification - container for seccomp userspace notifications. Since
87 * most seccomp filters will not have notification listeners attached and this
88 * structure is fairly large, we store the notification-specific stuff in a
91 * @request: A semaphore that users of this notification can wait on for
92 * changes. Actual reads and writes are still controlled with
93 * filter->notify_lock.
94 * @next_id: The id of the next request.
95 * @notifications: A list of struct seccomp_knotif elements.
96 * @wqh: A wait queue for poll.
99 struct semaphore request;
101 struct list_head notifications;
102 wait_queue_head_t wqh;
106 * struct seccomp_filter - container for seccomp BPF programs
108 * @usage: reference count to manage the object lifetime.
109 * get/put helpers should be used when accessing an instance
110 * outside of a lifetime-guarded section. In general, this
111 * is only needed for handling filters shared across tasks.
112 * @log: true if all actions except for SECCOMP_RET_ALLOW should be logged
113 * @prev: points to a previously installed, or inherited, filter
114 * @prog: the BPF program to evaluate
115 * @notif: the struct that holds all notification related information
116 * @notify_lock: A lock for all notification-related accesses.
118 * seccomp_filter objects are organized in a tree linked via the @prev
119 * pointer. For any task, it appears to be a singly-linked list starting
120 * with current->seccomp.filter, the most recently attached or inherited filter.
121 * However, multiple filters may share a @prev node, by way of fork(), which
122 * results in a unidirectional tree existing in memory. This is similar to
123 * how namespaces work.
125 * seccomp_filter objects should never be modified after being attached
126 * to a task_struct (other than @usage).
128 struct seccomp_filter {
131 struct seccomp_filter *prev;
132 struct bpf_prog *prog;
133 struct notification *notif;
134 struct mutex notify_lock;
137 /* Limit any path through the tree to 256KB worth of instructions. */
138 #define MAX_INSNS_PER_PATH ((1 << 18) / sizeof(struct sock_filter))
141 * Endianness is explicitly ignored and left for BPF program authors to manage
142 * as per the specific architecture.
144 static void populate_seccomp_data(struct seccomp_data *sd)
146 struct task_struct *task = current;
147 struct pt_regs *regs = task_pt_regs(task);
148 unsigned long args[6];
150 sd->nr = syscall_get_nr(task, regs);
151 sd->arch = syscall_get_arch(task);
152 syscall_get_arguments(task, regs, args);
153 sd->args[0] = args[0];
154 sd->args[1] = args[1];
155 sd->args[2] = args[2];
156 sd->args[3] = args[3];
157 sd->args[4] = args[4];
158 sd->args[5] = args[5];
159 sd->instruction_pointer = KSTK_EIP(task);
163 * seccomp_check_filter - verify seccomp filter code
164 * @filter: filter to verify
165 * @flen: length of filter
167 * Takes a previously checked filter (by bpf_check_classic) and
168 * redirects all filter code that loads struct sk_buff data
169 * and related data through seccomp_bpf_load. It also
170 * enforces length and alignment checking of those loads.
172 * Returns 0 if the rule set is legal or -EINVAL if not.
174 static int seccomp_check_filter(struct sock_filter *filter, unsigned int flen)
177 for (pc = 0; pc < flen; pc++) {
178 struct sock_filter *ftest = &filter[pc];
179 u16 code = ftest->code;
183 case BPF_LD | BPF_W | BPF_ABS:
184 ftest->code = BPF_LDX | BPF_W | BPF_ABS;
185 /* 32-bit aligned and not out of bounds. */
186 if (k >= sizeof(struct seccomp_data) || k & 3)
189 case BPF_LD | BPF_W | BPF_LEN:
190 ftest->code = BPF_LD | BPF_IMM;
191 ftest->k = sizeof(struct seccomp_data);
193 case BPF_LDX | BPF_W | BPF_LEN:
194 ftest->code = BPF_LDX | BPF_IMM;
195 ftest->k = sizeof(struct seccomp_data);
197 /* Explicitly include allowed calls. */
198 case BPF_RET | BPF_K:
199 case BPF_RET | BPF_A:
200 case BPF_ALU | BPF_ADD | BPF_K:
201 case BPF_ALU | BPF_ADD | BPF_X:
202 case BPF_ALU | BPF_SUB | BPF_K:
203 case BPF_ALU | BPF_SUB | BPF_X:
204 case BPF_ALU | BPF_MUL | BPF_K:
205 case BPF_ALU | BPF_MUL | BPF_X:
206 case BPF_ALU | BPF_DIV | BPF_K:
207 case BPF_ALU | BPF_DIV | BPF_X:
208 case BPF_ALU | BPF_AND | BPF_K:
209 case BPF_ALU | BPF_AND | BPF_X:
210 case BPF_ALU | BPF_OR | BPF_K:
211 case BPF_ALU | BPF_OR | BPF_X:
212 case BPF_ALU | BPF_XOR | BPF_K:
213 case BPF_ALU | BPF_XOR | BPF_X:
214 case BPF_ALU | BPF_LSH | BPF_K:
215 case BPF_ALU | BPF_LSH | BPF_X:
216 case BPF_ALU | BPF_RSH | BPF_K:
217 case BPF_ALU | BPF_RSH | BPF_X:
218 case BPF_ALU | BPF_NEG:
219 case BPF_LD | BPF_IMM:
220 case BPF_LDX | BPF_IMM:
221 case BPF_MISC | BPF_TAX:
222 case BPF_MISC | BPF_TXA:
223 case BPF_LD | BPF_MEM:
224 case BPF_LDX | BPF_MEM:
227 case BPF_JMP | BPF_JA:
228 case BPF_JMP | BPF_JEQ | BPF_K:
229 case BPF_JMP | BPF_JEQ | BPF_X:
230 case BPF_JMP | BPF_JGE | BPF_K:
231 case BPF_JMP | BPF_JGE | BPF_X:
232 case BPF_JMP | BPF_JGT | BPF_K:
233 case BPF_JMP | BPF_JGT | BPF_X:
234 case BPF_JMP | BPF_JSET | BPF_K:
235 case BPF_JMP | BPF_JSET | BPF_X:
245 * seccomp_run_filters - evaluates all seccomp filters against @sd
246 * @sd: optional seccomp data to be passed to filters
247 * @match: stores struct seccomp_filter that resulted in the return value,
248 * unless filter returned SECCOMP_RET_ALLOW, in which case it will
251 * Returns valid seccomp BPF response codes.
253 #define ACTION_ONLY(ret) ((s32)((ret) & (SECCOMP_RET_ACTION_FULL)))
254 static u32 seccomp_run_filters(const struct seccomp_data *sd,
255 struct seccomp_filter **match)
257 u32 ret = SECCOMP_RET_ALLOW;
258 /* Make sure cross-thread synced filter points somewhere sane. */
259 struct seccomp_filter *f =
260 READ_ONCE(current->seccomp.filter);
262 /* Ensure unexpected behavior doesn't result in failing open. */
263 if (WARN_ON(f == NULL))
264 return SECCOMP_RET_KILL_PROCESS;
267 * All filters in the list are evaluated and the lowest BPF return
268 * value always takes priority (ignoring the DATA).
271 for (; f; f = f->prev) {
272 u32 cur_ret = BPF_PROG_RUN(f->prog, sd);
274 if (ACTION_ONLY(cur_ret) < ACTION_ONLY(ret)) {
282 #endif /* CONFIG_SECCOMP_FILTER */
284 static inline bool seccomp_may_assign_mode(unsigned long seccomp_mode)
286 assert_spin_locked(¤t->sighand->siglock);
288 if (current->seccomp.mode && current->seccomp.mode != seccomp_mode)
294 void __weak arch_seccomp_spec_mitigate(struct task_struct *task) { }
296 static inline void seccomp_assign_mode(struct task_struct *task,
297 unsigned long seccomp_mode,
300 assert_spin_locked(&task->sighand->siglock);
302 task->seccomp.mode = seccomp_mode;
304 * Make sure TIF_SECCOMP cannot be set before the mode (and
307 smp_mb__before_atomic();
308 /* Assume default seccomp processes want spec flaw mitigation. */
309 if ((flags & SECCOMP_FILTER_FLAG_SPEC_ALLOW) == 0)
310 arch_seccomp_spec_mitigate(task);
311 set_tsk_thread_flag(task, TIF_SECCOMP);
314 #ifdef CONFIG_SECCOMP_FILTER
315 /* Returns 1 if the parent is an ancestor of the child. */
316 static int is_ancestor(struct seccomp_filter *parent,
317 struct seccomp_filter *child)
319 /* NULL is the root ancestor. */
322 for (; child; child = child->prev)
329 * seccomp_can_sync_threads: checks if all threads can be synchronized
331 * Expects sighand and cred_guard_mutex locks to be held.
333 * Returns 0 on success, -ve on error, or the pid of a thread which was
334 * either not in the correct seccomp mode or did not have an ancestral
337 static inline pid_t seccomp_can_sync_threads(void)
339 struct task_struct *thread, *caller;
341 BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex));
342 assert_spin_locked(¤t->sighand->siglock);
344 /* Validate all threads being eligible for synchronization. */
346 for_each_thread(caller, thread) {
349 /* Skip current, since it is initiating the sync. */
350 if (thread == caller)
353 if (thread->seccomp.mode == SECCOMP_MODE_DISABLED ||
354 (thread->seccomp.mode == SECCOMP_MODE_FILTER &&
355 is_ancestor(thread->seccomp.filter,
356 caller->seccomp.filter)))
359 /* Return the first thread that cannot be synchronized. */
360 failed = task_pid_vnr(thread);
361 /* If the pid cannot be resolved, then return -ESRCH */
362 if (WARN_ON(failed == 0))
371 * seccomp_sync_threads: sets all threads to use current's filter
373 * Expects sighand and cred_guard_mutex locks to be held, and for
374 * seccomp_can_sync_threads() to have returned success already
375 * without dropping the locks.
378 static inline void seccomp_sync_threads(unsigned long flags)
380 struct task_struct *thread, *caller;
382 BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex));
383 assert_spin_locked(¤t->sighand->siglock);
385 /* Synchronize all threads. */
387 for_each_thread(caller, thread) {
388 /* Skip current, since it needs no changes. */
389 if (thread == caller)
392 /* Get a task reference for the new leaf node. */
393 get_seccomp_filter(caller);
395 * Drop the task reference to the shared ancestor since
396 * current's path will hold a reference. (This also
397 * allows a put before the assignment.)
399 put_seccomp_filter(thread);
400 smp_store_release(&thread->seccomp.filter,
401 caller->seccomp.filter);
404 * Don't let an unprivileged task work around
405 * the no_new_privs restriction by creating
406 * a thread that sets it up, enters seccomp,
409 if (task_no_new_privs(caller))
410 task_set_no_new_privs(thread);
413 * Opt the other thread into seccomp if needed.
414 * As threads are considered to be trust-realm
415 * equivalent (see ptrace_may_access), it is safe to
416 * allow one thread to transition the other.
418 if (thread->seccomp.mode == SECCOMP_MODE_DISABLED)
419 seccomp_assign_mode(thread, SECCOMP_MODE_FILTER,
425 * seccomp_prepare_filter: Prepares a seccomp filter for use.
426 * @fprog: BPF program to install
428 * Returns filter on success or an ERR_PTR on failure.
430 static struct seccomp_filter *seccomp_prepare_filter(struct sock_fprog *fprog)
432 struct seccomp_filter *sfilter;
434 const bool save_orig = IS_ENABLED(CONFIG_CHECKPOINT_RESTORE);
436 if (fprog->len == 0 || fprog->len > BPF_MAXINSNS)
437 return ERR_PTR(-EINVAL);
439 BUG_ON(INT_MAX / fprog->len < sizeof(struct sock_filter));
442 * Installing a seccomp filter requires that the task has
443 * CAP_SYS_ADMIN in its namespace or be running with no_new_privs.
444 * This avoids scenarios where unprivileged tasks can affect the
445 * behavior of privileged children.
447 if (!task_no_new_privs(current) &&
448 security_capable(current_cred(), current_user_ns(),
449 CAP_SYS_ADMIN, CAP_OPT_NOAUDIT) != 0)
450 return ERR_PTR(-EACCES);
452 /* Allocate a new seccomp_filter */
453 sfilter = kzalloc(sizeof(*sfilter), GFP_KERNEL | __GFP_NOWARN);
455 return ERR_PTR(-ENOMEM);
457 mutex_init(&sfilter->notify_lock);
458 ret = bpf_prog_create_from_user(&sfilter->prog, fprog,
459 seccomp_check_filter, save_orig);
465 refcount_set(&sfilter->usage, 1);
471 * seccomp_prepare_user_filter - prepares a user-supplied sock_fprog
472 * @user_filter: pointer to the user data containing a sock_fprog.
474 * Returns 0 on success and non-zero otherwise.
476 static struct seccomp_filter *
477 seccomp_prepare_user_filter(const char __user *user_filter)
479 struct sock_fprog fprog;
480 struct seccomp_filter *filter = ERR_PTR(-EFAULT);
483 if (in_compat_syscall()) {
484 struct compat_sock_fprog fprog32;
485 if (copy_from_user(&fprog32, user_filter, sizeof(fprog32)))
487 fprog.len = fprog32.len;
488 fprog.filter = compat_ptr(fprog32.filter);
489 } else /* falls through to the if below. */
491 if (copy_from_user(&fprog, user_filter, sizeof(fprog)))
493 filter = seccomp_prepare_filter(&fprog);
499 * seccomp_attach_filter: validate and attach filter
500 * @flags: flags to change filter behavior
501 * @filter: seccomp filter to add to the current process
503 * Caller must be holding current->sighand->siglock lock.
505 * Returns 0 on success, -ve on error, or
506 * - in TSYNC mode: the pid of a thread which was either not in the correct
507 * seccomp mode or did not have an ancestral seccomp filter
508 * - in NEW_LISTENER mode: the fd of the new listener
510 static long seccomp_attach_filter(unsigned int flags,
511 struct seccomp_filter *filter)
513 unsigned long total_insns;
514 struct seccomp_filter *walker;
516 assert_spin_locked(¤t->sighand->siglock);
518 /* Validate resulting filter length. */
519 total_insns = filter->prog->len;
520 for (walker = current->seccomp.filter; walker; walker = walker->prev)
521 total_insns += walker->prog->len + 4; /* 4 instr penalty */
522 if (total_insns > MAX_INSNS_PER_PATH)
525 /* If thread sync has been requested, check that it is possible. */
526 if (flags & SECCOMP_FILTER_FLAG_TSYNC) {
529 ret = seccomp_can_sync_threads();
534 /* Set log flag, if present. */
535 if (flags & SECCOMP_FILTER_FLAG_LOG)
539 * If there is an existing filter, make it the prev and don't drop its
542 filter->prev = current->seccomp.filter;
543 current->seccomp.filter = filter;
545 /* Now that the new filter is in place, synchronize to all threads. */
546 if (flags & SECCOMP_FILTER_FLAG_TSYNC)
547 seccomp_sync_threads(flags);
552 static void __get_seccomp_filter(struct seccomp_filter *filter)
554 refcount_inc(&filter->usage);
557 /* get_seccomp_filter - increments the reference count of the filter on @tsk */
558 void get_seccomp_filter(struct task_struct *tsk)
560 struct seccomp_filter *orig = tsk->seccomp.filter;
563 __get_seccomp_filter(orig);
566 static inline void seccomp_filter_free(struct seccomp_filter *filter)
569 bpf_prog_destroy(filter->prog);
574 static void __put_seccomp_filter(struct seccomp_filter *orig)
576 /* Clean up single-reference branches iteratively. */
577 while (orig && refcount_dec_and_test(&orig->usage)) {
578 struct seccomp_filter *freeme = orig;
580 seccomp_filter_free(freeme);
584 /* put_seccomp_filter - decrements the ref count of tsk->seccomp.filter */
585 void put_seccomp_filter(struct task_struct *tsk)
587 __put_seccomp_filter(tsk->seccomp.filter);
590 static void seccomp_init_siginfo(kernel_siginfo_t *info, int syscall, int reason)
593 info->si_signo = SIGSYS;
594 info->si_code = SYS_SECCOMP;
595 info->si_call_addr = (void __user *)KSTK_EIP(current);
596 info->si_errno = reason;
597 info->si_arch = syscall_get_arch(current);
598 info->si_syscall = syscall;
602 * seccomp_send_sigsys - signals the task to allow in-process syscall emulation
603 * @syscall: syscall number to send to userland
604 * @reason: filter-supplied reason code to send to userland (via si_errno)
606 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
608 static void seccomp_send_sigsys(int syscall, int reason)
610 struct kernel_siginfo info;
611 seccomp_init_siginfo(&info, syscall, reason);
612 force_sig_info(&info);
614 #endif /* CONFIG_SECCOMP_FILTER */
616 /* For use with seccomp_actions_logged */
617 #define SECCOMP_LOG_KILL_PROCESS (1 << 0)
618 #define SECCOMP_LOG_KILL_THREAD (1 << 1)
619 #define SECCOMP_LOG_TRAP (1 << 2)
620 #define SECCOMP_LOG_ERRNO (1 << 3)
621 #define SECCOMP_LOG_TRACE (1 << 4)
622 #define SECCOMP_LOG_LOG (1 << 5)
623 #define SECCOMP_LOG_ALLOW (1 << 6)
624 #define SECCOMP_LOG_USER_NOTIF (1 << 7)
626 static u32 seccomp_actions_logged = SECCOMP_LOG_KILL_PROCESS |
627 SECCOMP_LOG_KILL_THREAD |
630 SECCOMP_LOG_USER_NOTIF |
634 static inline void seccomp_log(unsigned long syscall, long signr, u32 action,
640 case SECCOMP_RET_ALLOW:
642 case SECCOMP_RET_TRAP:
643 log = requested && seccomp_actions_logged & SECCOMP_LOG_TRAP;
645 case SECCOMP_RET_ERRNO:
646 log = requested && seccomp_actions_logged & SECCOMP_LOG_ERRNO;
648 case SECCOMP_RET_TRACE:
649 log = requested && seccomp_actions_logged & SECCOMP_LOG_TRACE;
651 case SECCOMP_RET_USER_NOTIF:
652 log = requested && seccomp_actions_logged & SECCOMP_LOG_USER_NOTIF;
654 case SECCOMP_RET_LOG:
655 log = seccomp_actions_logged & SECCOMP_LOG_LOG;
657 case SECCOMP_RET_KILL_THREAD:
658 log = seccomp_actions_logged & SECCOMP_LOG_KILL_THREAD;
660 case SECCOMP_RET_KILL_PROCESS:
662 log = seccomp_actions_logged & SECCOMP_LOG_KILL_PROCESS;
666 * Emit an audit message when the action is RET_KILL_*, RET_LOG, or the
667 * FILTER_FLAG_LOG bit was set. The admin has the ability to silence
668 * any action from being logged by removing the action name from the
669 * seccomp_actions_logged sysctl.
674 audit_seccomp(syscall, signr, action);
678 * Secure computing mode 1 allows only read/write/exit/sigreturn.
679 * To be fully secure this must be combined with rlimit
680 * to limit the stack allocations too.
682 static const int mode1_syscalls[] = {
683 __NR_seccomp_read, __NR_seccomp_write, __NR_seccomp_exit, __NR_seccomp_sigreturn,
684 0, /* null terminated */
687 static void __secure_computing_strict(int this_syscall)
689 const int *syscall_whitelist = mode1_syscalls;
691 if (in_compat_syscall())
692 syscall_whitelist = get_compat_mode1_syscalls();
695 if (*syscall_whitelist == this_syscall)
697 } while (*++syscall_whitelist);
702 seccomp_log(this_syscall, SIGKILL, SECCOMP_RET_KILL_THREAD, true);
706 #ifndef CONFIG_HAVE_ARCH_SECCOMP_FILTER
707 void secure_computing_strict(int this_syscall)
709 int mode = current->seccomp.mode;
711 if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) &&
712 unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
715 if (mode == SECCOMP_MODE_DISABLED)
717 else if (mode == SECCOMP_MODE_STRICT)
718 __secure_computing_strict(this_syscall);
724 #ifdef CONFIG_SECCOMP_FILTER
725 static u64 seccomp_next_notify_id(struct seccomp_filter *filter)
728 * Note: overflow is ok here, the id just needs to be unique per
731 lockdep_assert_held(&filter->notify_lock);
732 return filter->notif->next_id++;
735 static void seccomp_do_user_notification(int this_syscall,
736 struct seccomp_filter *match,
737 const struct seccomp_data *sd)
741 struct seccomp_knotif n = {};
743 mutex_lock(&match->notify_lock);
749 n.state = SECCOMP_NOTIFY_INIT;
751 n.id = seccomp_next_notify_id(match);
752 init_completion(&n.ready);
753 list_add(&n.list, &match->notif->notifications);
755 up(&match->notif->request);
756 wake_up_poll(&match->notif->wqh, EPOLLIN | EPOLLRDNORM);
757 mutex_unlock(&match->notify_lock);
760 * This is where we wait for a reply from userspace.
762 err = wait_for_completion_interruptible(&n.ready);
763 mutex_lock(&match->notify_lock);
770 * Note that it's possible the listener died in between the time when
771 * we were notified of a respons (or a signal) and when we were able to
772 * re-acquire the lock, so only delete from the list if the
773 * notification actually exists.
775 * Also note that this test is only valid because there's no way to
776 * *reattach* to a notifier right now. If one is added, we'll need to
777 * keep track of the notif itself and make sure they match here.
782 mutex_unlock(&match->notify_lock);
783 syscall_set_return_value(current, task_pt_regs(current),
787 static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
788 const bool recheck_after_trace)
790 u32 filter_ret, action;
791 struct seccomp_filter *match = NULL;
793 struct seccomp_data sd_local;
796 * Make sure that any changes to mode from another thread have
797 * been seen after TIF_SECCOMP was seen.
802 populate_seccomp_data(&sd_local);
806 filter_ret = seccomp_run_filters(sd, &match);
807 data = filter_ret & SECCOMP_RET_DATA;
808 action = filter_ret & SECCOMP_RET_ACTION_FULL;
811 case SECCOMP_RET_ERRNO:
812 /* Set low-order bits as an errno, capped at MAX_ERRNO. */
813 if (data > MAX_ERRNO)
815 syscall_set_return_value(current, task_pt_regs(current),
819 case SECCOMP_RET_TRAP:
820 /* Show the handler the original registers. */
821 syscall_rollback(current, task_pt_regs(current));
822 /* Let the filter pass back 16 bits of data. */
823 seccomp_send_sigsys(this_syscall, data);
826 case SECCOMP_RET_TRACE:
827 /* We've been put in this state by the ptracer already. */
828 if (recheck_after_trace)
831 /* ENOSYS these calls if there is no tracer attached. */
832 if (!ptrace_event_enabled(current, PTRACE_EVENT_SECCOMP)) {
833 syscall_set_return_value(current,
834 task_pt_regs(current),
839 /* Allow the BPF to provide the event message */
840 ptrace_event(PTRACE_EVENT_SECCOMP, data);
842 * The delivery of a fatal signal during event
843 * notification may silently skip tracer notification,
844 * which could leave us with a potentially unmodified
845 * syscall that the tracer would have liked to have
846 * changed. Since the process is about to die, we just
847 * force the syscall to be skipped and let the signal
848 * kill the process and correctly handle any tracer exit
851 if (fatal_signal_pending(current))
853 /* Check if the tracer forced the syscall to be skipped. */
854 this_syscall = syscall_get_nr(current, task_pt_regs(current));
855 if (this_syscall < 0)
859 * Recheck the syscall, since it may have changed. This
860 * intentionally uses a NULL struct seccomp_data to force
861 * a reload of all registers. This does not goto skip since
862 * a skip would have already been reported.
864 if (__seccomp_filter(this_syscall, NULL, true))
869 case SECCOMP_RET_USER_NOTIF:
870 seccomp_do_user_notification(this_syscall, match, sd);
873 case SECCOMP_RET_LOG:
874 seccomp_log(this_syscall, 0, action, true);
877 case SECCOMP_RET_ALLOW:
879 * Note that the "match" filter will always be NULL for
880 * this action since SECCOMP_RET_ALLOW is the starting
881 * state in seccomp_run_filters().
885 case SECCOMP_RET_KILL_THREAD:
886 case SECCOMP_RET_KILL_PROCESS:
888 seccomp_log(this_syscall, SIGSYS, action, true);
889 /* Dump core only if this is the last remaining thread. */
890 if (action == SECCOMP_RET_KILL_PROCESS ||
891 get_nr_threads(current) == 1) {
892 kernel_siginfo_t info;
894 /* Show the original registers in the dump. */
895 syscall_rollback(current, task_pt_regs(current));
896 /* Trigger a manual coredump since do_exit skips it. */
897 seccomp_init_siginfo(&info, this_syscall, data);
900 if (action == SECCOMP_RET_KILL_PROCESS)
901 do_group_exit(SIGSYS);
909 seccomp_log(this_syscall, 0, action, match ? match->log : false);
913 static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
914 const bool recheck_after_trace)
920 int __secure_computing(const struct seccomp_data *sd)
922 int mode = current->seccomp.mode;
925 if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) &&
926 unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
929 this_syscall = sd ? sd->nr :
930 syscall_get_nr(current, task_pt_regs(current));
933 case SECCOMP_MODE_STRICT:
934 __secure_computing_strict(this_syscall); /* may call do_exit */
936 case SECCOMP_MODE_FILTER:
937 return __seccomp_filter(this_syscall, sd, false);
942 #endif /* CONFIG_HAVE_ARCH_SECCOMP_FILTER */
944 long prctl_get_seccomp(void)
946 return current->seccomp.mode;
950 * seccomp_set_mode_strict: internal function for setting strict seccomp
952 * Once current->seccomp.mode is non-zero, it may not be changed.
954 * Returns 0 on success or -EINVAL on failure.
956 static long seccomp_set_mode_strict(void)
958 const unsigned long seccomp_mode = SECCOMP_MODE_STRICT;
961 spin_lock_irq(¤t->sighand->siglock);
963 if (!seccomp_may_assign_mode(seccomp_mode))
969 seccomp_assign_mode(current, seccomp_mode, 0);
973 spin_unlock_irq(¤t->sighand->siglock);
978 #ifdef CONFIG_SECCOMP_FILTER
979 static int seccomp_notify_release(struct inode *inode, struct file *file)
981 struct seccomp_filter *filter = file->private_data;
982 struct seccomp_knotif *knotif;
987 mutex_lock(&filter->notify_lock);
990 * If this file is being closed because e.g. the task who owned it
991 * died, let's wake everyone up who was waiting on us.
993 list_for_each_entry(knotif, &filter->notif->notifications, list) {
994 if (knotif->state == SECCOMP_NOTIFY_REPLIED)
997 knotif->state = SECCOMP_NOTIFY_REPLIED;
998 knotif->error = -ENOSYS;
1001 complete(&knotif->ready);
1004 kfree(filter->notif);
1005 filter->notif = NULL;
1006 mutex_unlock(&filter->notify_lock);
1007 __put_seccomp_filter(filter);
1011 static long seccomp_notify_recv(struct seccomp_filter *filter,
1014 struct seccomp_knotif *knotif = NULL, *cur;
1015 struct seccomp_notif unotif;
1018 /* Verify that we're not given garbage to keep struct extensible. */
1019 ret = check_zeroed_user(buf, sizeof(unotif));
1025 memset(&unotif, 0, sizeof(unotif));
1027 ret = down_interruptible(&filter->notif->request);
1031 mutex_lock(&filter->notify_lock);
1032 list_for_each_entry(cur, &filter->notif->notifications, list) {
1033 if (cur->state == SECCOMP_NOTIFY_INIT) {
1040 * If we didn't find a notification, it could be that the task was
1041 * interrupted by a fatal signal between the time we were woken and
1042 * when we were able to acquire the rw lock.
1049 unotif.id = knotif->id;
1050 unotif.pid = task_pid_vnr(knotif->task);
1051 unotif.data = *(knotif->data);
1053 knotif->state = SECCOMP_NOTIFY_SENT;
1054 wake_up_poll(&filter->notif->wqh, EPOLLOUT | EPOLLWRNORM);
1057 mutex_unlock(&filter->notify_lock);
1059 if (ret == 0 && copy_to_user(buf, &unotif, sizeof(unotif))) {
1063 * Userspace screwed up. To make sure that we keep this
1064 * notification alive, let's reset it back to INIT. It
1065 * may have died when we released the lock, so we need to make
1066 * sure it's still around.
1069 mutex_lock(&filter->notify_lock);
1070 list_for_each_entry(cur, &filter->notif->notifications, list) {
1071 if (cur->id == unotif.id) {
1078 knotif->state = SECCOMP_NOTIFY_INIT;
1079 up(&filter->notif->request);
1081 mutex_unlock(&filter->notify_lock);
1087 static long seccomp_notify_send(struct seccomp_filter *filter,
1090 struct seccomp_notif_resp resp = {};
1091 struct seccomp_knotif *knotif = NULL, *cur;
1094 if (copy_from_user(&resp, buf, sizeof(resp)))
1100 ret = mutex_lock_interruptible(&filter->notify_lock);
1104 list_for_each_entry(cur, &filter->notif->notifications, list) {
1105 if (cur->id == resp.id) {
1116 /* Allow exactly one reply. */
1117 if (knotif->state != SECCOMP_NOTIFY_SENT) {
1123 knotif->state = SECCOMP_NOTIFY_REPLIED;
1124 knotif->error = resp.error;
1125 knotif->val = resp.val;
1126 complete(&knotif->ready);
1128 mutex_unlock(&filter->notify_lock);
1132 static long seccomp_notify_id_valid(struct seccomp_filter *filter,
1135 struct seccomp_knotif *knotif = NULL;
1139 if (copy_from_user(&id, buf, sizeof(id)))
1142 ret = mutex_lock_interruptible(&filter->notify_lock);
1147 list_for_each_entry(knotif, &filter->notif->notifications, list) {
1148 if (knotif->id == id) {
1149 if (knotif->state == SECCOMP_NOTIFY_SENT)
1156 mutex_unlock(&filter->notify_lock);
1160 static long seccomp_notify_ioctl(struct file *file, unsigned int cmd,
1163 struct seccomp_filter *filter = file->private_data;
1164 void __user *buf = (void __user *)arg;
1167 case SECCOMP_IOCTL_NOTIF_RECV:
1168 return seccomp_notify_recv(filter, buf);
1169 case SECCOMP_IOCTL_NOTIF_SEND:
1170 return seccomp_notify_send(filter, buf);
1171 case SECCOMP_IOCTL_NOTIF_ID_VALID:
1172 return seccomp_notify_id_valid(filter, buf);
1178 static __poll_t seccomp_notify_poll(struct file *file,
1179 struct poll_table_struct *poll_tab)
1181 struct seccomp_filter *filter = file->private_data;
1183 struct seccomp_knotif *cur;
1185 poll_wait(file, &filter->notif->wqh, poll_tab);
1187 if (mutex_lock_interruptible(&filter->notify_lock) < 0)
1190 list_for_each_entry(cur, &filter->notif->notifications, list) {
1191 if (cur->state == SECCOMP_NOTIFY_INIT)
1192 ret |= EPOLLIN | EPOLLRDNORM;
1193 if (cur->state == SECCOMP_NOTIFY_SENT)
1194 ret |= EPOLLOUT | EPOLLWRNORM;
1195 if ((ret & EPOLLIN) && (ret & EPOLLOUT))
1199 mutex_unlock(&filter->notify_lock);
1204 static const struct file_operations seccomp_notify_ops = {
1205 .poll = seccomp_notify_poll,
1206 .release = seccomp_notify_release,
1207 .unlocked_ioctl = seccomp_notify_ioctl,
1210 static struct file *init_listener(struct seccomp_filter *filter)
1212 struct file *ret = ERR_PTR(-EBUSY);
1213 struct seccomp_filter *cur;
1215 for (cur = current->seccomp.filter; cur; cur = cur->prev) {
1220 ret = ERR_PTR(-ENOMEM);
1221 filter->notif = kzalloc(sizeof(*(filter->notif)), GFP_KERNEL);
1225 sema_init(&filter->notif->request, 0);
1226 filter->notif->next_id = get_random_u64();
1227 INIT_LIST_HEAD(&filter->notif->notifications);
1228 init_waitqueue_head(&filter->notif->wqh);
1230 ret = anon_inode_getfile("seccomp notify", &seccomp_notify_ops,
1235 /* The file has a reference to it now */
1236 __get_seccomp_filter(filter);
1240 kfree(filter->notif);
1246 * seccomp_set_mode_filter: internal function for setting seccomp filter
1247 * @flags: flags to change filter behavior
1248 * @filter: struct sock_fprog containing filter
1250 * This function may be called repeatedly to install additional filters.
1251 * Every filter successfully installed will be evaluated (in reverse order)
1252 * for each system call the task makes.
1254 * Once current->seccomp.mode is non-zero, it may not be changed.
1256 * Returns 0 on success or -EINVAL on failure.
1258 static long seccomp_set_mode_filter(unsigned int flags,
1259 const char __user *filter)
1261 const unsigned long seccomp_mode = SECCOMP_MODE_FILTER;
1262 struct seccomp_filter *prepared = NULL;
1265 struct file *listener_f = NULL;
1267 /* Validate flags. */
1268 if (flags & ~SECCOMP_FILTER_FLAG_MASK)
1272 * In the successful case, NEW_LISTENER returns the new listener fd.
1273 * But in the failure case, TSYNC returns the thread that died. If you
1274 * combine these two flags, there's no way to tell whether something
1275 * succeeded or failed. So, let's disallow this combination.
1277 if ((flags & SECCOMP_FILTER_FLAG_TSYNC) &&
1278 (flags & SECCOMP_FILTER_FLAG_NEW_LISTENER))
1281 /* Prepare the new filter before holding any locks. */
1282 prepared = seccomp_prepare_user_filter(filter);
1283 if (IS_ERR(prepared))
1284 return PTR_ERR(prepared);
1286 if (flags & SECCOMP_FILTER_FLAG_NEW_LISTENER) {
1287 listener = get_unused_fd_flags(O_CLOEXEC);
1293 listener_f = init_listener(prepared);
1294 if (IS_ERR(listener_f)) {
1295 put_unused_fd(listener);
1296 ret = PTR_ERR(listener_f);
1302 * Make sure we cannot change seccomp or nnp state via TSYNC
1303 * while another thread is in the middle of calling exec.
1305 if (flags & SECCOMP_FILTER_FLAG_TSYNC &&
1306 mutex_lock_killable(¤t->signal->cred_guard_mutex))
1309 spin_lock_irq(¤t->sighand->siglock);
1311 if (!seccomp_may_assign_mode(seccomp_mode))
1314 ret = seccomp_attach_filter(flags, prepared);
1317 /* Do not free the successfully attached filter. */
1320 seccomp_assign_mode(current, seccomp_mode, flags);
1322 spin_unlock_irq(¤t->sighand->siglock);
1323 if (flags & SECCOMP_FILTER_FLAG_TSYNC)
1324 mutex_unlock(¤t->signal->cred_guard_mutex);
1326 if (flags & SECCOMP_FILTER_FLAG_NEW_LISTENER) {
1328 listener_f->private_data = NULL;
1330 put_unused_fd(listener);
1332 fd_install(listener, listener_f);
1337 seccomp_filter_free(prepared);
1341 static inline long seccomp_set_mode_filter(unsigned int flags,
1342 const char __user *filter)
1348 static long seccomp_get_action_avail(const char __user *uaction)
1352 if (copy_from_user(&action, uaction, sizeof(action)))
1356 case SECCOMP_RET_KILL_PROCESS:
1357 case SECCOMP_RET_KILL_THREAD:
1358 case SECCOMP_RET_TRAP:
1359 case SECCOMP_RET_ERRNO:
1360 case SECCOMP_RET_USER_NOTIF:
1361 case SECCOMP_RET_TRACE:
1362 case SECCOMP_RET_LOG:
1363 case SECCOMP_RET_ALLOW:
1372 static long seccomp_get_notif_sizes(void __user *usizes)
1374 struct seccomp_notif_sizes sizes = {
1375 .seccomp_notif = sizeof(struct seccomp_notif),
1376 .seccomp_notif_resp = sizeof(struct seccomp_notif_resp),
1377 .seccomp_data = sizeof(struct seccomp_data),
1380 if (copy_to_user(usizes, &sizes, sizeof(sizes)))
1386 /* Common entry point for both prctl and syscall. */
1387 static long do_seccomp(unsigned int op, unsigned int flags,
1391 case SECCOMP_SET_MODE_STRICT:
1392 if (flags != 0 || uargs != NULL)
1394 return seccomp_set_mode_strict();
1395 case SECCOMP_SET_MODE_FILTER:
1396 return seccomp_set_mode_filter(flags, uargs);
1397 case SECCOMP_GET_ACTION_AVAIL:
1401 return seccomp_get_action_avail(uargs);
1402 case SECCOMP_GET_NOTIF_SIZES:
1406 return seccomp_get_notif_sizes(uargs);
1412 SYSCALL_DEFINE3(seccomp, unsigned int, op, unsigned int, flags,
1413 void __user *, uargs)
1415 return do_seccomp(op, flags, uargs);
1419 * prctl_set_seccomp: configures current->seccomp.mode
1420 * @seccomp_mode: requested mode to use
1421 * @filter: optional struct sock_fprog for use with SECCOMP_MODE_FILTER
1423 * Returns 0 on success or -EINVAL on failure.
1425 long prctl_set_seccomp(unsigned long seccomp_mode, void __user *filter)
1430 switch (seccomp_mode) {
1431 case SECCOMP_MODE_STRICT:
1432 op = SECCOMP_SET_MODE_STRICT;
1434 * Setting strict mode through prctl always ignored filter,
1435 * so make sure it is always NULL here to pass the internal
1436 * check in do_seccomp().
1440 case SECCOMP_MODE_FILTER:
1441 op = SECCOMP_SET_MODE_FILTER;
1448 /* prctl interface doesn't have flags, so they are always zero. */
1449 return do_seccomp(op, 0, uargs);
1452 #if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
1453 static struct seccomp_filter *get_nth_filter(struct task_struct *task,
1454 unsigned long filter_off)
1456 struct seccomp_filter *orig, *filter;
1457 unsigned long count;
1460 * Note: this is only correct because the caller should be the (ptrace)
1461 * tracer of the task, otherwise lock_task_sighand is needed.
1463 spin_lock_irq(&task->sighand->siglock);
1465 if (task->seccomp.mode != SECCOMP_MODE_FILTER) {
1466 spin_unlock_irq(&task->sighand->siglock);
1467 return ERR_PTR(-EINVAL);
1470 orig = task->seccomp.filter;
1471 __get_seccomp_filter(orig);
1472 spin_unlock_irq(&task->sighand->siglock);
1475 for (filter = orig; filter; filter = filter->prev)
1478 if (filter_off >= count) {
1479 filter = ERR_PTR(-ENOENT);
1483 count -= filter_off;
1484 for (filter = orig; filter && count > 1; filter = filter->prev)
1487 if (WARN_ON(count != 1 || !filter)) {
1488 filter = ERR_PTR(-ENOENT);
1492 __get_seccomp_filter(filter);
1495 __put_seccomp_filter(orig);
1499 long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
1502 struct seccomp_filter *filter;
1503 struct sock_fprog_kern *fprog;
1506 if (!capable(CAP_SYS_ADMIN) ||
1507 current->seccomp.mode != SECCOMP_MODE_DISABLED) {
1511 filter = get_nth_filter(task, filter_off);
1513 return PTR_ERR(filter);
1515 fprog = filter->prog->orig_prog;
1517 /* This must be a new non-cBPF filter, since we save
1518 * every cBPF filter's orig_prog above when
1519 * CONFIG_CHECKPOINT_RESTORE is enabled.
1529 if (copy_to_user(data, fprog->filter, bpf_classic_proglen(fprog)))
1533 __put_seccomp_filter(filter);
1537 long seccomp_get_metadata(struct task_struct *task,
1538 unsigned long size, void __user *data)
1541 struct seccomp_filter *filter;
1542 struct seccomp_metadata kmd = {};
1544 if (!capable(CAP_SYS_ADMIN) ||
1545 current->seccomp.mode != SECCOMP_MODE_DISABLED) {
1549 size = min_t(unsigned long, size, sizeof(kmd));
1551 if (size < sizeof(kmd.filter_off))
1554 if (copy_from_user(&kmd.filter_off, data, sizeof(kmd.filter_off)))
1557 filter = get_nth_filter(task, kmd.filter_off);
1559 return PTR_ERR(filter);
1562 kmd.flags |= SECCOMP_FILTER_FLAG_LOG;
1565 if (copy_to_user(data, &kmd, size))
1568 __put_seccomp_filter(filter);
1573 #ifdef CONFIG_SYSCTL
1575 /* Human readable action names for friendly sysctl interaction */
1576 #define SECCOMP_RET_KILL_PROCESS_NAME "kill_process"
1577 #define SECCOMP_RET_KILL_THREAD_NAME "kill_thread"
1578 #define SECCOMP_RET_TRAP_NAME "trap"
1579 #define SECCOMP_RET_ERRNO_NAME "errno"
1580 #define SECCOMP_RET_USER_NOTIF_NAME "user_notif"
1581 #define SECCOMP_RET_TRACE_NAME "trace"
1582 #define SECCOMP_RET_LOG_NAME "log"
1583 #define SECCOMP_RET_ALLOW_NAME "allow"
1585 static const char seccomp_actions_avail[] =
1586 SECCOMP_RET_KILL_PROCESS_NAME " "
1587 SECCOMP_RET_KILL_THREAD_NAME " "
1588 SECCOMP_RET_TRAP_NAME " "
1589 SECCOMP_RET_ERRNO_NAME " "
1590 SECCOMP_RET_USER_NOTIF_NAME " "
1591 SECCOMP_RET_TRACE_NAME " "
1592 SECCOMP_RET_LOG_NAME " "
1593 SECCOMP_RET_ALLOW_NAME;
1595 struct seccomp_log_name {
1600 static const struct seccomp_log_name seccomp_log_names[] = {
1601 { SECCOMP_LOG_KILL_PROCESS, SECCOMP_RET_KILL_PROCESS_NAME },
1602 { SECCOMP_LOG_KILL_THREAD, SECCOMP_RET_KILL_THREAD_NAME },
1603 { SECCOMP_LOG_TRAP, SECCOMP_RET_TRAP_NAME },
1604 { SECCOMP_LOG_ERRNO, SECCOMP_RET_ERRNO_NAME },
1605 { SECCOMP_LOG_USER_NOTIF, SECCOMP_RET_USER_NOTIF_NAME },
1606 { SECCOMP_LOG_TRACE, SECCOMP_RET_TRACE_NAME },
1607 { SECCOMP_LOG_LOG, SECCOMP_RET_LOG_NAME },
1608 { SECCOMP_LOG_ALLOW, SECCOMP_RET_ALLOW_NAME },
1612 static bool seccomp_names_from_actions_logged(char *names, size_t size,
1616 const struct seccomp_log_name *cur;
1617 bool append_sep = false;
1619 for (cur = seccomp_log_names; cur->name && size; cur++) {
1622 if (!(actions_logged & cur->log))
1626 ret = strscpy(names, sep, size);
1635 ret = strscpy(names, cur->name, size);
1646 static bool seccomp_action_logged_from_name(u32 *action_logged,
1649 const struct seccomp_log_name *cur;
1651 for (cur = seccomp_log_names; cur->name; cur++) {
1652 if (!strcmp(cur->name, name)) {
1653 *action_logged = cur->log;
1661 static bool seccomp_actions_logged_from_names(u32 *actions_logged, char *names)
1665 *actions_logged = 0;
1666 while ((name = strsep(&names, " ")) && *name) {
1667 u32 action_logged = 0;
1669 if (!seccomp_action_logged_from_name(&action_logged, name))
1672 *actions_logged |= action_logged;
1678 static int read_actions_logged(struct ctl_table *ro_table, void __user *buffer,
1679 size_t *lenp, loff_t *ppos)
1681 char names[sizeof(seccomp_actions_avail)];
1682 struct ctl_table table;
1684 memset(names, 0, sizeof(names));
1686 if (!seccomp_names_from_actions_logged(names, sizeof(names),
1687 seccomp_actions_logged, " "))
1692 table.maxlen = sizeof(names);
1693 return proc_dostring(&table, 0, buffer, lenp, ppos);
1696 static int write_actions_logged(struct ctl_table *ro_table, void __user *buffer,
1697 size_t *lenp, loff_t *ppos, u32 *actions_logged)
1699 char names[sizeof(seccomp_actions_avail)];
1700 struct ctl_table table;
1703 if (!capable(CAP_SYS_ADMIN))
1706 memset(names, 0, sizeof(names));
1710 table.maxlen = sizeof(names);
1711 ret = proc_dostring(&table, 1, buffer, lenp, ppos);
1715 if (!seccomp_actions_logged_from_names(actions_logged, table.data))
1718 if (*actions_logged & SECCOMP_LOG_ALLOW)
1721 seccomp_actions_logged = *actions_logged;
1725 static void audit_actions_logged(u32 actions_logged, u32 old_actions_logged,
1728 char names[sizeof(seccomp_actions_avail)];
1729 char old_names[sizeof(seccomp_actions_avail)];
1730 const char *new = names;
1731 const char *old = old_names;
1736 memset(names, 0, sizeof(names));
1737 memset(old_names, 0, sizeof(old_names));
1741 else if (!actions_logged)
1743 else if (!seccomp_names_from_actions_logged(names, sizeof(names),
1744 actions_logged, ","))
1747 if (!old_actions_logged)
1749 else if (!seccomp_names_from_actions_logged(old_names,
1751 old_actions_logged, ","))
1754 return audit_seccomp_actions_logged(new, old, !ret);
1757 static int seccomp_actions_logged_handler(struct ctl_table *ro_table, int write,
1758 void __user *buffer, size_t *lenp,
1764 u32 actions_logged = 0;
1765 u32 old_actions_logged = seccomp_actions_logged;
1767 ret = write_actions_logged(ro_table, buffer, lenp, ppos,
1769 audit_actions_logged(actions_logged, old_actions_logged, ret);
1771 ret = read_actions_logged(ro_table, buffer, lenp, ppos);
1776 static struct ctl_path seccomp_sysctl_path[] = {
1777 { .procname = "kernel", },
1778 { .procname = "seccomp", },
1782 static struct ctl_table seccomp_sysctl_table[] = {
1784 .procname = "actions_avail",
1785 .data = (void *) &seccomp_actions_avail,
1786 .maxlen = sizeof(seccomp_actions_avail),
1788 .proc_handler = proc_dostring,
1791 .procname = "actions_logged",
1793 .proc_handler = seccomp_actions_logged_handler,
1798 static int __init seccomp_sysctl_init(void)
1800 struct ctl_table_header *hdr;
1802 hdr = register_sysctl_paths(seccomp_sysctl_path, seccomp_sysctl_table);
1804 pr_warn("seccomp: sysctl registration failed\n");
1806 kmemleak_not_leak(hdr);
1811 device_initcall(seccomp_sysctl_init)
1813 #endif /* CONFIG_SYSCTL */