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 /* Not exposed in headers: strictly internal use only. */
32 #define SECCOMP_MODE_DEAD (SECCOMP_MODE_FILTER + 1)
34 #ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
35 #include <asm/syscall.h>
38 #ifdef CONFIG_SECCOMP_FILTER
39 #include <linux/file.h>
40 #include <linux/filter.h>
41 #include <linux/pid.h>
42 #include <linux/ptrace.h>
43 #include <linux/capability.h>
44 #include <linux/tracehook.h>
45 #include <linux/uaccess.h>
46 #include <linux/anon_inodes.h>
49 * When SECCOMP_IOCTL_NOTIF_ID_VALID was first introduced, it had the
50 * wrong direction flag in the ioctl number. This is the broken one,
51 * which the kernel needs to keep supporting until all userspaces stop
52 * using the wrong command number.
54 #define SECCOMP_IOCTL_NOTIF_ID_VALID_WRONG_DIR SECCOMP_IOR(2, __u64)
59 SECCOMP_NOTIFY_REPLIED,
62 struct seccomp_knotif {
63 /* The struct pid of the task whose filter triggered the notification */
64 struct task_struct *task;
66 /* The "cookie" for this request; this is unique for this filter. */
70 * The seccomp data. This pointer is valid the entire time this
71 * notification is active, since it comes from __seccomp_filter which
72 * eclipses the entire lifecycle here.
74 const struct seccomp_data *data;
77 * Notification states. When SECCOMP_RET_USER_NOTIF is returned, a
78 * struct seccomp_knotif is created and starts out in INIT. Once the
79 * handler reads the notification off of an FD, it transitions to SENT.
80 * If a signal is received the state transitions back to INIT and
81 * another message is sent. When the userspace handler replies, state
82 * transitions to REPLIED.
84 enum notify_state state;
86 /* The return values, only valid when in SECCOMP_NOTIFY_REPLIED */
90 /* Signals when this has entered SECCOMP_NOTIFY_REPLIED */
91 struct completion ready;
93 struct list_head list;
97 * struct notification - container for seccomp userspace notifications. Since
98 * most seccomp filters will not have notification listeners attached and this
99 * structure is fairly large, we store the notification-specific stuff in a
100 * separate structure.
102 * @request: A semaphore that users of this notification can wait on for
103 * changes. Actual reads and writes are still controlled with
104 * filter->notify_lock.
105 * @next_id: The id of the next request.
106 * @notifications: A list of struct seccomp_knotif elements.
107 * @wqh: A wait queue for poll.
109 struct notification {
110 struct semaphore request;
112 struct list_head notifications;
113 wait_queue_head_t wqh;
117 * struct seccomp_filter - container for seccomp BPF programs
119 * @usage: reference count to manage the object lifetime.
120 * get/put helpers should be used when accessing an instance
121 * outside of a lifetime-guarded section. In general, this
122 * is only needed for handling filters shared across tasks.
123 * @log: true if all actions except for SECCOMP_RET_ALLOW should be logged
124 * @prev: points to a previously installed, or inherited, filter
125 * @prog: the BPF program to evaluate
126 * @notif: the struct that holds all notification related information
127 * @notify_lock: A lock for all notification-related accesses.
129 * seccomp_filter objects are organized in a tree linked via the @prev
130 * pointer. For any task, it appears to be a singly-linked list starting
131 * with current->seccomp.filter, the most recently attached or inherited filter.
132 * However, multiple filters may share a @prev node, by way of fork(), which
133 * results in a unidirectional tree existing in memory. This is similar to
134 * how namespaces work.
136 * seccomp_filter objects should never be modified after being attached
137 * to a task_struct (other than @usage).
139 struct seccomp_filter {
142 struct seccomp_filter *prev;
143 struct bpf_prog *prog;
144 struct notification *notif;
145 struct mutex notify_lock;
148 /* Limit any path through the tree to 256KB worth of instructions. */
149 #define MAX_INSNS_PER_PATH ((1 << 18) / sizeof(struct sock_filter))
152 * Endianness is explicitly ignored and left for BPF program authors to manage
153 * as per the specific architecture.
155 static void populate_seccomp_data(struct seccomp_data *sd)
157 struct task_struct *task = current;
158 struct pt_regs *regs = task_pt_regs(task);
159 unsigned long args[6];
161 sd->nr = syscall_get_nr(task, regs);
162 sd->arch = syscall_get_arch(task);
163 syscall_get_arguments(task, regs, args);
164 sd->args[0] = args[0];
165 sd->args[1] = args[1];
166 sd->args[2] = args[2];
167 sd->args[3] = args[3];
168 sd->args[4] = args[4];
169 sd->args[5] = args[5];
170 sd->instruction_pointer = KSTK_EIP(task);
174 * seccomp_check_filter - verify seccomp filter code
175 * @filter: filter to verify
176 * @flen: length of filter
178 * Takes a previously checked filter (by bpf_check_classic) and
179 * redirects all filter code that loads struct sk_buff data
180 * and related data through seccomp_bpf_load. It also
181 * enforces length and alignment checking of those loads.
183 * Returns 0 if the rule set is legal or -EINVAL if not.
185 static int seccomp_check_filter(struct sock_filter *filter, unsigned int flen)
188 for (pc = 0; pc < flen; pc++) {
189 struct sock_filter *ftest = &filter[pc];
190 u16 code = ftest->code;
194 case BPF_LD | BPF_W | BPF_ABS:
195 ftest->code = BPF_LDX | BPF_W | BPF_ABS;
196 /* 32-bit aligned and not out of bounds. */
197 if (k >= sizeof(struct seccomp_data) || k & 3)
200 case BPF_LD | BPF_W | BPF_LEN:
201 ftest->code = BPF_LD | BPF_IMM;
202 ftest->k = sizeof(struct seccomp_data);
204 case BPF_LDX | BPF_W | BPF_LEN:
205 ftest->code = BPF_LDX | BPF_IMM;
206 ftest->k = sizeof(struct seccomp_data);
208 /* Explicitly include allowed calls. */
209 case BPF_RET | BPF_K:
210 case BPF_RET | BPF_A:
211 case BPF_ALU | BPF_ADD | BPF_K:
212 case BPF_ALU | BPF_ADD | BPF_X:
213 case BPF_ALU | BPF_SUB | BPF_K:
214 case BPF_ALU | BPF_SUB | BPF_X:
215 case BPF_ALU | BPF_MUL | BPF_K:
216 case BPF_ALU | BPF_MUL | BPF_X:
217 case BPF_ALU | BPF_DIV | BPF_K:
218 case BPF_ALU | BPF_DIV | BPF_X:
219 case BPF_ALU | BPF_AND | BPF_K:
220 case BPF_ALU | BPF_AND | BPF_X:
221 case BPF_ALU | BPF_OR | BPF_K:
222 case BPF_ALU | BPF_OR | BPF_X:
223 case BPF_ALU | BPF_XOR | BPF_K:
224 case BPF_ALU | BPF_XOR | BPF_X:
225 case BPF_ALU | BPF_LSH | BPF_K:
226 case BPF_ALU | BPF_LSH | BPF_X:
227 case BPF_ALU | BPF_RSH | BPF_K:
228 case BPF_ALU | BPF_RSH | BPF_X:
229 case BPF_ALU | BPF_NEG:
230 case BPF_LD | BPF_IMM:
231 case BPF_LDX | BPF_IMM:
232 case BPF_MISC | BPF_TAX:
233 case BPF_MISC | BPF_TXA:
234 case BPF_LD | BPF_MEM:
235 case BPF_LDX | BPF_MEM:
238 case BPF_JMP | BPF_JA:
239 case BPF_JMP | BPF_JEQ | BPF_K:
240 case BPF_JMP | BPF_JEQ | BPF_X:
241 case BPF_JMP | BPF_JGE | BPF_K:
242 case BPF_JMP | BPF_JGE | BPF_X:
243 case BPF_JMP | BPF_JGT | BPF_K:
244 case BPF_JMP | BPF_JGT | BPF_X:
245 case BPF_JMP | BPF_JSET | BPF_K:
246 case BPF_JMP | BPF_JSET | BPF_X:
256 * seccomp_run_filters - evaluates all seccomp filters against @sd
257 * @sd: optional seccomp data to be passed to filters
258 * @match: stores struct seccomp_filter that resulted in the return value,
259 * unless filter returned SECCOMP_RET_ALLOW, in which case it will
262 * Returns valid seccomp BPF response codes.
264 #define ACTION_ONLY(ret) ((s32)((ret) & (SECCOMP_RET_ACTION_FULL)))
265 static u32 seccomp_run_filters(const struct seccomp_data *sd,
266 struct seccomp_filter **match)
268 u32 ret = SECCOMP_RET_ALLOW;
269 /* Make sure cross-thread synced filter points somewhere sane. */
270 struct seccomp_filter *f =
271 READ_ONCE(current->seccomp.filter);
273 /* Ensure unexpected behavior doesn't result in failing open. */
274 if (WARN_ON(f == NULL))
275 return SECCOMP_RET_KILL_PROCESS;
278 * All filters in the list are evaluated and the lowest BPF return
279 * value always takes priority (ignoring the DATA).
282 for (; f; f = f->prev) {
283 u32 cur_ret = BPF_PROG_RUN(f->prog, sd);
285 if (ACTION_ONLY(cur_ret) < ACTION_ONLY(ret)) {
293 #endif /* CONFIG_SECCOMP_FILTER */
295 static inline bool seccomp_may_assign_mode(unsigned long seccomp_mode)
297 assert_spin_locked(¤t->sighand->siglock);
299 if (current->seccomp.mode && current->seccomp.mode != seccomp_mode)
305 void __weak arch_seccomp_spec_mitigate(struct task_struct *task) { }
307 static inline void seccomp_assign_mode(struct task_struct *task,
308 unsigned long seccomp_mode,
311 assert_spin_locked(&task->sighand->siglock);
313 task->seccomp.mode = seccomp_mode;
315 * Make sure TIF_SECCOMP cannot be set before the mode (and
318 smp_mb__before_atomic();
319 /* Assume default seccomp processes want spec flaw mitigation. */
320 if ((flags & SECCOMP_FILTER_FLAG_SPEC_ALLOW) == 0)
321 arch_seccomp_spec_mitigate(task);
322 set_tsk_thread_flag(task, TIF_SECCOMP);
325 #ifdef CONFIG_SECCOMP_FILTER
326 /* Returns 1 if the parent is an ancestor of the child. */
327 static int is_ancestor(struct seccomp_filter *parent,
328 struct seccomp_filter *child)
330 /* NULL is the root ancestor. */
333 for (; child; child = child->prev)
340 * seccomp_can_sync_threads: checks if all threads can be synchronized
342 * Expects sighand and cred_guard_mutex locks to be held.
344 * Returns 0 on success, -ve on error, or the pid of a thread which was
345 * either not in the correct seccomp mode or did not have an ancestral
348 static inline pid_t seccomp_can_sync_threads(void)
350 struct task_struct *thread, *caller;
352 BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex));
353 assert_spin_locked(¤t->sighand->siglock);
355 /* Validate all threads being eligible for synchronization. */
357 for_each_thread(caller, thread) {
360 /* Skip current, since it is initiating the sync. */
361 if (thread == caller)
364 if (thread->seccomp.mode == SECCOMP_MODE_DISABLED ||
365 (thread->seccomp.mode == SECCOMP_MODE_FILTER &&
366 is_ancestor(thread->seccomp.filter,
367 caller->seccomp.filter)))
370 /* Return the first thread that cannot be synchronized. */
371 failed = task_pid_vnr(thread);
372 /* If the pid cannot be resolved, then return -ESRCH */
373 if (WARN_ON(failed == 0))
382 * seccomp_sync_threads: sets all threads to use current's filter
384 * Expects sighand and cred_guard_mutex locks to be held, and for
385 * seccomp_can_sync_threads() to have returned success already
386 * without dropping the locks.
389 static inline void seccomp_sync_threads(unsigned long flags)
391 struct task_struct *thread, *caller;
393 BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex));
394 assert_spin_locked(¤t->sighand->siglock);
396 /* Synchronize all threads. */
398 for_each_thread(caller, thread) {
399 /* Skip current, since it needs no changes. */
400 if (thread == caller)
403 /* Get a task reference for the new leaf node. */
404 get_seccomp_filter(caller);
406 * Drop the task reference to the shared ancestor since
407 * current's path will hold a reference. (This also
408 * allows a put before the assignment.)
410 put_seccomp_filter(thread);
411 smp_store_release(&thread->seccomp.filter,
412 caller->seccomp.filter);
415 * Don't let an unprivileged task work around
416 * the no_new_privs restriction by creating
417 * a thread that sets it up, enters seccomp,
420 if (task_no_new_privs(caller))
421 task_set_no_new_privs(thread);
424 * Opt the other thread into seccomp if needed.
425 * As threads are considered to be trust-realm
426 * equivalent (see ptrace_may_access), it is safe to
427 * allow one thread to transition the other.
429 if (thread->seccomp.mode == SECCOMP_MODE_DISABLED)
430 seccomp_assign_mode(thread, SECCOMP_MODE_FILTER,
436 * seccomp_prepare_filter: Prepares a seccomp filter for use.
437 * @fprog: BPF program to install
439 * Returns filter on success or an ERR_PTR on failure.
441 static struct seccomp_filter *seccomp_prepare_filter(struct sock_fprog *fprog)
443 struct seccomp_filter *sfilter;
445 const bool save_orig = IS_ENABLED(CONFIG_CHECKPOINT_RESTORE);
447 if (fprog->len == 0 || fprog->len > BPF_MAXINSNS)
448 return ERR_PTR(-EINVAL);
450 BUG_ON(INT_MAX / fprog->len < sizeof(struct sock_filter));
453 * Installing a seccomp filter requires that the task has
454 * CAP_SYS_ADMIN in its namespace or be running with no_new_privs.
455 * This avoids scenarios where unprivileged tasks can affect the
456 * behavior of privileged children.
458 if (!task_no_new_privs(current) &&
459 !ns_capable_noaudit(current_user_ns(), CAP_SYS_ADMIN))
460 return ERR_PTR(-EACCES);
462 /* Allocate a new seccomp_filter */
463 sfilter = kzalloc(sizeof(*sfilter), GFP_KERNEL | __GFP_NOWARN);
465 return ERR_PTR(-ENOMEM);
467 mutex_init(&sfilter->notify_lock);
468 ret = bpf_prog_create_from_user(&sfilter->prog, fprog,
469 seccomp_check_filter, save_orig);
475 refcount_set(&sfilter->usage, 1);
481 * seccomp_prepare_user_filter - prepares a user-supplied sock_fprog
482 * @user_filter: pointer to the user data containing a sock_fprog.
484 * Returns 0 on success and non-zero otherwise.
486 static struct seccomp_filter *
487 seccomp_prepare_user_filter(const char __user *user_filter)
489 struct sock_fprog fprog;
490 struct seccomp_filter *filter = ERR_PTR(-EFAULT);
493 if (in_compat_syscall()) {
494 struct compat_sock_fprog fprog32;
495 if (copy_from_user(&fprog32, user_filter, sizeof(fprog32)))
497 fprog.len = fprog32.len;
498 fprog.filter = compat_ptr(fprog32.filter);
499 } else /* falls through to the if below. */
501 if (copy_from_user(&fprog, user_filter, sizeof(fprog)))
503 filter = seccomp_prepare_filter(&fprog);
509 * seccomp_attach_filter: validate and attach filter
510 * @flags: flags to change filter behavior
511 * @filter: seccomp filter to add to the current process
513 * Caller must be holding current->sighand->siglock lock.
515 * Returns 0 on success, -ve on error, or
516 * - in TSYNC mode: the pid of a thread which was either not in the correct
517 * seccomp mode or did not have an ancestral seccomp filter
518 * - in NEW_LISTENER mode: the fd of the new listener
520 static long seccomp_attach_filter(unsigned int flags,
521 struct seccomp_filter *filter)
523 unsigned long total_insns;
524 struct seccomp_filter *walker;
526 assert_spin_locked(¤t->sighand->siglock);
528 /* Validate resulting filter length. */
529 total_insns = filter->prog->len;
530 for (walker = current->seccomp.filter; walker; walker = walker->prev)
531 total_insns += walker->prog->len + 4; /* 4 instr penalty */
532 if (total_insns > MAX_INSNS_PER_PATH)
535 /* If thread sync has been requested, check that it is possible. */
536 if (flags & SECCOMP_FILTER_FLAG_TSYNC) {
539 ret = seccomp_can_sync_threads();
544 /* Set log flag, if present. */
545 if (flags & SECCOMP_FILTER_FLAG_LOG)
549 * If there is an existing filter, make it the prev and don't drop its
552 filter->prev = current->seccomp.filter;
553 current->seccomp.filter = filter;
555 /* Now that the new filter is in place, synchronize to all threads. */
556 if (flags & SECCOMP_FILTER_FLAG_TSYNC)
557 seccomp_sync_threads(flags);
562 static void __get_seccomp_filter(struct seccomp_filter *filter)
564 refcount_inc(&filter->usage);
567 /* get_seccomp_filter - increments the reference count of the filter on @tsk */
568 void get_seccomp_filter(struct task_struct *tsk)
570 struct seccomp_filter *orig = tsk->seccomp.filter;
573 __get_seccomp_filter(orig);
576 static inline void seccomp_filter_free(struct seccomp_filter *filter)
579 bpf_prog_destroy(filter->prog);
584 static void __put_seccomp_filter(struct seccomp_filter *orig)
586 /* Clean up single-reference branches iteratively. */
587 while (orig && refcount_dec_and_test(&orig->usage)) {
588 struct seccomp_filter *freeme = orig;
590 seccomp_filter_free(freeme);
594 /* put_seccomp_filter - decrements the ref count of tsk->seccomp.filter */
595 void put_seccomp_filter(struct task_struct *tsk)
597 __put_seccomp_filter(tsk->seccomp.filter);
600 static void seccomp_init_siginfo(kernel_siginfo_t *info, int syscall, int reason)
603 info->si_signo = SIGSYS;
604 info->si_code = SYS_SECCOMP;
605 info->si_call_addr = (void __user *)KSTK_EIP(current);
606 info->si_errno = reason;
607 info->si_arch = syscall_get_arch(current);
608 info->si_syscall = syscall;
612 * seccomp_send_sigsys - signals the task to allow in-process syscall emulation
613 * @syscall: syscall number to send to userland
614 * @reason: filter-supplied reason code to send to userland (via si_errno)
616 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
618 static void seccomp_send_sigsys(int syscall, int reason)
620 struct kernel_siginfo info;
621 seccomp_init_siginfo(&info, syscall, reason);
622 force_sig_info(&info);
624 #endif /* CONFIG_SECCOMP_FILTER */
626 /* For use with seccomp_actions_logged */
627 #define SECCOMP_LOG_KILL_PROCESS (1 << 0)
628 #define SECCOMP_LOG_KILL_THREAD (1 << 1)
629 #define SECCOMP_LOG_TRAP (1 << 2)
630 #define SECCOMP_LOG_ERRNO (1 << 3)
631 #define SECCOMP_LOG_TRACE (1 << 4)
632 #define SECCOMP_LOG_LOG (1 << 5)
633 #define SECCOMP_LOG_ALLOW (1 << 6)
634 #define SECCOMP_LOG_USER_NOTIF (1 << 7)
636 static u32 seccomp_actions_logged = SECCOMP_LOG_KILL_PROCESS |
637 SECCOMP_LOG_KILL_THREAD |
640 SECCOMP_LOG_USER_NOTIF |
644 static inline void seccomp_log(unsigned long syscall, long signr, u32 action,
650 case SECCOMP_RET_ALLOW:
652 case SECCOMP_RET_TRAP:
653 log = requested && seccomp_actions_logged & SECCOMP_LOG_TRAP;
655 case SECCOMP_RET_ERRNO:
656 log = requested && seccomp_actions_logged & SECCOMP_LOG_ERRNO;
658 case SECCOMP_RET_TRACE:
659 log = requested && seccomp_actions_logged & SECCOMP_LOG_TRACE;
661 case SECCOMP_RET_USER_NOTIF:
662 log = requested && seccomp_actions_logged & SECCOMP_LOG_USER_NOTIF;
664 case SECCOMP_RET_LOG:
665 log = seccomp_actions_logged & SECCOMP_LOG_LOG;
667 case SECCOMP_RET_KILL_THREAD:
668 log = seccomp_actions_logged & SECCOMP_LOG_KILL_THREAD;
670 case SECCOMP_RET_KILL_PROCESS:
672 log = seccomp_actions_logged & SECCOMP_LOG_KILL_PROCESS;
676 * Emit an audit message when the action is RET_KILL_*, RET_LOG, or the
677 * FILTER_FLAG_LOG bit was set. The admin has the ability to silence
678 * any action from being logged by removing the action name from the
679 * seccomp_actions_logged sysctl.
684 audit_seccomp(syscall, signr, action);
688 * Secure computing mode 1 allows only read/write/exit/sigreturn.
689 * To be fully secure this must be combined with rlimit
690 * to limit the stack allocations too.
692 static const int mode1_syscalls[] = {
693 __NR_seccomp_read, __NR_seccomp_write, __NR_seccomp_exit, __NR_seccomp_sigreturn,
694 0, /* null terminated */
697 static void __secure_computing_strict(int this_syscall)
699 const int *syscall_whitelist = mode1_syscalls;
701 if (in_compat_syscall())
702 syscall_whitelist = get_compat_mode1_syscalls();
705 if (*syscall_whitelist == this_syscall)
707 } while (*++syscall_whitelist);
712 current->seccomp.mode = SECCOMP_MODE_DEAD;
713 seccomp_log(this_syscall, SIGKILL, SECCOMP_RET_KILL_THREAD, true);
717 #ifndef CONFIG_HAVE_ARCH_SECCOMP_FILTER
718 void secure_computing_strict(int this_syscall)
720 int mode = current->seccomp.mode;
722 if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) &&
723 unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
726 if (mode == SECCOMP_MODE_DISABLED)
728 else if (mode == SECCOMP_MODE_STRICT)
729 __secure_computing_strict(this_syscall);
735 #ifdef CONFIG_SECCOMP_FILTER
736 static u64 seccomp_next_notify_id(struct seccomp_filter *filter)
739 * Note: overflow is ok here, the id just needs to be unique per
742 lockdep_assert_held(&filter->notify_lock);
743 return filter->notif->next_id++;
746 static void seccomp_do_user_notification(int this_syscall,
747 struct seccomp_filter *match,
748 const struct seccomp_data *sd)
752 struct seccomp_knotif n = {};
754 mutex_lock(&match->notify_lock);
760 n.state = SECCOMP_NOTIFY_INIT;
762 n.id = seccomp_next_notify_id(match);
763 init_completion(&n.ready);
764 list_add(&n.list, &match->notif->notifications);
766 up(&match->notif->request);
767 wake_up_poll(&match->notif->wqh, EPOLLIN | EPOLLRDNORM);
768 mutex_unlock(&match->notify_lock);
771 * This is where we wait for a reply from userspace.
773 err = wait_for_completion_interruptible(&n.ready);
774 mutex_lock(&match->notify_lock);
781 * Note that it's possible the listener died in between the time when
782 * we were notified of a respons (or a signal) and when we were able to
783 * re-acquire the lock, so only delete from the list if the
784 * notification actually exists.
786 * Also note that this test is only valid because there's no way to
787 * *reattach* to a notifier right now. If one is added, we'll need to
788 * keep track of the notif itself and make sure they match here.
793 mutex_unlock(&match->notify_lock);
794 syscall_set_return_value(current, task_pt_regs(current),
798 static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
799 const bool recheck_after_trace)
801 u32 filter_ret, action;
802 struct seccomp_filter *match = NULL;
804 struct seccomp_data sd_local;
807 * Make sure that any changes to mode from another thread have
808 * been seen after TIF_SECCOMP was seen.
813 populate_seccomp_data(&sd_local);
817 filter_ret = seccomp_run_filters(sd, &match);
818 data = filter_ret & SECCOMP_RET_DATA;
819 action = filter_ret & SECCOMP_RET_ACTION_FULL;
822 case SECCOMP_RET_ERRNO:
823 /* Set low-order bits as an errno, capped at MAX_ERRNO. */
824 if (data > MAX_ERRNO)
826 syscall_set_return_value(current, task_pt_regs(current),
830 case SECCOMP_RET_TRAP:
831 /* Show the handler the original registers. */
832 syscall_rollback(current, task_pt_regs(current));
833 /* Let the filter pass back 16 bits of data. */
834 seccomp_send_sigsys(this_syscall, data);
837 case SECCOMP_RET_TRACE:
838 /* We've been put in this state by the ptracer already. */
839 if (recheck_after_trace)
842 /* ENOSYS these calls if there is no tracer attached. */
843 if (!ptrace_event_enabled(current, PTRACE_EVENT_SECCOMP)) {
844 syscall_set_return_value(current,
845 task_pt_regs(current),
850 /* Allow the BPF to provide the event message */
851 ptrace_event(PTRACE_EVENT_SECCOMP, data);
853 * The delivery of a fatal signal during event
854 * notification may silently skip tracer notification,
855 * which could leave us with a potentially unmodified
856 * syscall that the tracer would have liked to have
857 * changed. Since the process is about to die, we just
858 * force the syscall to be skipped and let the signal
859 * kill the process and correctly handle any tracer exit
862 if (fatal_signal_pending(current))
864 /* Check if the tracer forced the syscall to be skipped. */
865 this_syscall = syscall_get_nr(current, task_pt_regs(current));
866 if (this_syscall < 0)
870 * Recheck the syscall, since it may have changed. This
871 * intentionally uses a NULL struct seccomp_data to force
872 * a reload of all registers. This does not goto skip since
873 * a skip would have already been reported.
875 if (__seccomp_filter(this_syscall, NULL, true))
880 case SECCOMP_RET_USER_NOTIF:
881 seccomp_do_user_notification(this_syscall, match, sd);
884 case SECCOMP_RET_LOG:
885 seccomp_log(this_syscall, 0, action, true);
888 case SECCOMP_RET_ALLOW:
890 * Note that the "match" filter will always be NULL for
891 * this action since SECCOMP_RET_ALLOW is the starting
892 * state in seccomp_run_filters().
896 case SECCOMP_RET_KILL_THREAD:
897 case SECCOMP_RET_KILL_PROCESS:
899 current->seccomp.mode = SECCOMP_MODE_DEAD;
900 seccomp_log(this_syscall, SIGSYS, action, true);
901 /* Dump core only if this is the last remaining thread. */
902 if (action == SECCOMP_RET_KILL_PROCESS ||
903 get_nr_threads(current) == 1) {
904 kernel_siginfo_t info;
906 /* Show the original registers in the dump. */
907 syscall_rollback(current, task_pt_regs(current));
908 /* Trigger a manual coredump since do_exit skips it. */
909 seccomp_init_siginfo(&info, this_syscall, data);
912 if (action == SECCOMP_RET_KILL_PROCESS)
913 do_group_exit(SIGSYS);
921 seccomp_log(this_syscall, 0, action, match ? match->log : false);
925 static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
926 const bool recheck_after_trace)
934 int __secure_computing(const struct seccomp_data *sd)
936 int mode = current->seccomp.mode;
939 if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) &&
940 unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
943 this_syscall = sd ? sd->nr :
944 syscall_get_nr(current, task_pt_regs(current));
947 case SECCOMP_MODE_STRICT:
948 __secure_computing_strict(this_syscall); /* may call do_exit */
950 case SECCOMP_MODE_FILTER:
951 return __seccomp_filter(this_syscall, sd, false);
952 /* Surviving SECCOMP_RET_KILL_* must be proactively impossible. */
953 case SECCOMP_MODE_DEAD:
961 #endif /* CONFIG_HAVE_ARCH_SECCOMP_FILTER */
963 long prctl_get_seccomp(void)
965 return current->seccomp.mode;
969 * seccomp_set_mode_strict: internal function for setting strict seccomp
971 * Once current->seccomp.mode is non-zero, it may not be changed.
973 * Returns 0 on success or -EINVAL on failure.
975 static long seccomp_set_mode_strict(void)
977 const unsigned long seccomp_mode = SECCOMP_MODE_STRICT;
980 spin_lock_irq(¤t->sighand->siglock);
982 if (!seccomp_may_assign_mode(seccomp_mode))
988 seccomp_assign_mode(current, seccomp_mode, 0);
992 spin_unlock_irq(¤t->sighand->siglock);
997 #ifdef CONFIG_SECCOMP_FILTER
998 static int seccomp_notify_release(struct inode *inode, struct file *file)
1000 struct seccomp_filter *filter = file->private_data;
1001 struct seccomp_knotif *knotif;
1006 mutex_lock(&filter->notify_lock);
1009 * If this file is being closed because e.g. the task who owned it
1010 * died, let's wake everyone up who was waiting on us.
1012 list_for_each_entry(knotif, &filter->notif->notifications, list) {
1013 if (knotif->state == SECCOMP_NOTIFY_REPLIED)
1016 knotif->state = SECCOMP_NOTIFY_REPLIED;
1017 knotif->error = -ENOSYS;
1020 complete(&knotif->ready);
1023 kfree(filter->notif);
1024 filter->notif = NULL;
1025 mutex_unlock(&filter->notify_lock);
1026 __put_seccomp_filter(filter);
1030 static long seccomp_notify_recv(struct seccomp_filter *filter,
1033 struct seccomp_knotif *knotif = NULL, *cur;
1034 struct seccomp_notif unotif;
1037 /* Verify that we're not given garbage to keep struct extensible. */
1038 ret = check_zeroed_user(buf, sizeof(unotif));
1044 memset(&unotif, 0, sizeof(unotif));
1046 ret = down_interruptible(&filter->notif->request);
1050 mutex_lock(&filter->notify_lock);
1051 list_for_each_entry(cur, &filter->notif->notifications, list) {
1052 if (cur->state == SECCOMP_NOTIFY_INIT) {
1059 * If we didn't find a notification, it could be that the task was
1060 * interrupted by a fatal signal between the time we were woken and
1061 * when we were able to acquire the rw lock.
1068 unotif.id = knotif->id;
1069 unotif.pid = task_pid_vnr(knotif->task);
1070 unotif.data = *(knotif->data);
1072 knotif->state = SECCOMP_NOTIFY_SENT;
1073 wake_up_poll(&filter->notif->wqh, EPOLLOUT | EPOLLWRNORM);
1076 mutex_unlock(&filter->notify_lock);
1078 if (ret == 0 && copy_to_user(buf, &unotif, sizeof(unotif))) {
1082 * Userspace screwed up. To make sure that we keep this
1083 * notification alive, let's reset it back to INIT. It
1084 * may have died when we released the lock, so we need to make
1085 * sure it's still around.
1088 mutex_lock(&filter->notify_lock);
1089 list_for_each_entry(cur, &filter->notif->notifications, list) {
1090 if (cur->id == unotif.id) {
1097 knotif->state = SECCOMP_NOTIFY_INIT;
1098 up(&filter->notif->request);
1100 mutex_unlock(&filter->notify_lock);
1106 static long seccomp_notify_send(struct seccomp_filter *filter,
1109 struct seccomp_notif_resp resp = {};
1110 struct seccomp_knotif *knotif = NULL, *cur;
1113 if (copy_from_user(&resp, buf, sizeof(resp)))
1119 ret = mutex_lock_interruptible(&filter->notify_lock);
1123 list_for_each_entry(cur, &filter->notif->notifications, list) {
1124 if (cur->id == resp.id) {
1135 /* Allow exactly one reply. */
1136 if (knotif->state != SECCOMP_NOTIFY_SENT) {
1142 knotif->state = SECCOMP_NOTIFY_REPLIED;
1143 knotif->error = resp.error;
1144 knotif->val = resp.val;
1145 complete(&knotif->ready);
1147 mutex_unlock(&filter->notify_lock);
1151 static long seccomp_notify_id_valid(struct seccomp_filter *filter,
1154 struct seccomp_knotif *knotif = NULL;
1158 if (copy_from_user(&id, buf, sizeof(id)))
1161 ret = mutex_lock_interruptible(&filter->notify_lock);
1166 list_for_each_entry(knotif, &filter->notif->notifications, list) {
1167 if (knotif->id == id) {
1168 if (knotif->state == SECCOMP_NOTIFY_SENT)
1175 mutex_unlock(&filter->notify_lock);
1179 static long seccomp_notify_ioctl(struct file *file, unsigned int cmd,
1182 struct seccomp_filter *filter = file->private_data;
1183 void __user *buf = (void __user *)arg;
1186 case SECCOMP_IOCTL_NOTIF_RECV:
1187 return seccomp_notify_recv(filter, buf);
1188 case SECCOMP_IOCTL_NOTIF_SEND:
1189 return seccomp_notify_send(filter, buf);
1190 case SECCOMP_IOCTL_NOTIF_ID_VALID_WRONG_DIR:
1191 case SECCOMP_IOCTL_NOTIF_ID_VALID:
1192 return seccomp_notify_id_valid(filter, buf);
1198 static __poll_t seccomp_notify_poll(struct file *file,
1199 struct poll_table_struct *poll_tab)
1201 struct seccomp_filter *filter = file->private_data;
1203 struct seccomp_knotif *cur;
1205 poll_wait(file, &filter->notif->wqh, poll_tab);
1207 if (mutex_lock_interruptible(&filter->notify_lock) < 0)
1210 list_for_each_entry(cur, &filter->notif->notifications, list) {
1211 if (cur->state == SECCOMP_NOTIFY_INIT)
1212 ret |= EPOLLIN | EPOLLRDNORM;
1213 if (cur->state == SECCOMP_NOTIFY_SENT)
1214 ret |= EPOLLOUT | EPOLLWRNORM;
1215 if ((ret & EPOLLIN) && (ret & EPOLLOUT))
1219 mutex_unlock(&filter->notify_lock);
1224 static const struct file_operations seccomp_notify_ops = {
1225 .poll = seccomp_notify_poll,
1226 .release = seccomp_notify_release,
1227 .unlocked_ioctl = seccomp_notify_ioctl,
1228 .compat_ioctl = seccomp_notify_ioctl,
1231 static struct file *init_listener(struct seccomp_filter *filter)
1235 ret = ERR_PTR(-ENOMEM);
1236 filter->notif = kzalloc(sizeof(*(filter->notif)), GFP_KERNEL);
1240 sema_init(&filter->notif->request, 0);
1241 filter->notif->next_id = get_random_u64();
1242 INIT_LIST_HEAD(&filter->notif->notifications);
1243 init_waitqueue_head(&filter->notif->wqh);
1245 ret = anon_inode_getfile("seccomp notify", &seccomp_notify_ops,
1250 /* The file has a reference to it now */
1251 __get_seccomp_filter(filter);
1255 kfree(filter->notif);
1261 * Does @new_child have a listener while an ancestor also has a listener?
1262 * If so, we'll want to reject this filter.
1263 * This only has to be tested for the current process, even in the TSYNC case,
1264 * because TSYNC installs @child with the same parent on all threads.
1265 * Note that @new_child is not hooked up to its parent at this point yet, so
1266 * we use current->seccomp.filter.
1268 static bool has_duplicate_listener(struct seccomp_filter *new_child)
1270 struct seccomp_filter *cur;
1272 /* must be protected against concurrent TSYNC */
1273 lockdep_assert_held(¤t->sighand->siglock);
1275 if (!new_child->notif)
1277 for (cur = current->seccomp.filter; cur; cur = cur->prev) {
1286 * seccomp_set_mode_filter: internal function for setting seccomp filter
1287 * @flags: flags to change filter behavior
1288 * @filter: struct sock_fprog containing filter
1290 * This function may be called repeatedly to install additional filters.
1291 * Every filter successfully installed will be evaluated (in reverse order)
1292 * for each system call the task makes.
1294 * Once current->seccomp.mode is non-zero, it may not be changed.
1296 * Returns 0 on success or -EINVAL on failure.
1298 static long seccomp_set_mode_filter(unsigned int flags,
1299 const char __user *filter)
1301 const unsigned long seccomp_mode = SECCOMP_MODE_FILTER;
1302 struct seccomp_filter *prepared = NULL;
1305 struct file *listener_f = NULL;
1307 /* Validate flags. */
1308 if (flags & ~SECCOMP_FILTER_FLAG_MASK)
1312 * In the successful case, NEW_LISTENER returns the new listener fd.
1313 * But in the failure case, TSYNC returns the thread that died. If you
1314 * combine these two flags, there's no way to tell whether something
1315 * succeeded or failed. So, let's disallow this combination.
1317 if ((flags & SECCOMP_FILTER_FLAG_TSYNC) &&
1318 (flags & SECCOMP_FILTER_FLAG_NEW_LISTENER))
1321 /* Prepare the new filter before holding any locks. */
1322 prepared = seccomp_prepare_user_filter(filter);
1323 if (IS_ERR(prepared))
1324 return PTR_ERR(prepared);
1326 if (flags & SECCOMP_FILTER_FLAG_NEW_LISTENER) {
1327 listener = get_unused_fd_flags(O_CLOEXEC);
1333 listener_f = init_listener(prepared);
1334 if (IS_ERR(listener_f)) {
1335 put_unused_fd(listener);
1336 ret = PTR_ERR(listener_f);
1342 * Make sure we cannot change seccomp or nnp state via TSYNC
1343 * while another thread is in the middle of calling exec.
1345 if (flags & SECCOMP_FILTER_FLAG_TSYNC &&
1346 mutex_lock_killable(¤t->signal->cred_guard_mutex))
1349 spin_lock_irq(¤t->sighand->siglock);
1351 if (!seccomp_may_assign_mode(seccomp_mode))
1354 if (has_duplicate_listener(prepared)) {
1359 ret = seccomp_attach_filter(flags, prepared);
1362 /* Do not free the successfully attached filter. */
1365 seccomp_assign_mode(current, seccomp_mode, flags);
1367 spin_unlock_irq(¤t->sighand->siglock);
1368 if (flags & SECCOMP_FILTER_FLAG_TSYNC)
1369 mutex_unlock(¤t->signal->cred_guard_mutex);
1371 if (flags & SECCOMP_FILTER_FLAG_NEW_LISTENER) {
1373 listener_f->private_data = NULL;
1375 put_unused_fd(listener);
1377 fd_install(listener, listener_f);
1382 seccomp_filter_free(prepared);
1386 static inline long seccomp_set_mode_filter(unsigned int flags,
1387 const char __user *filter)
1393 static long seccomp_get_action_avail(const char __user *uaction)
1397 if (copy_from_user(&action, uaction, sizeof(action)))
1401 case SECCOMP_RET_KILL_PROCESS:
1402 case SECCOMP_RET_KILL_THREAD:
1403 case SECCOMP_RET_TRAP:
1404 case SECCOMP_RET_ERRNO:
1405 case SECCOMP_RET_USER_NOTIF:
1406 case SECCOMP_RET_TRACE:
1407 case SECCOMP_RET_LOG:
1408 case SECCOMP_RET_ALLOW:
1417 static long seccomp_get_notif_sizes(void __user *usizes)
1419 struct seccomp_notif_sizes sizes = {
1420 .seccomp_notif = sizeof(struct seccomp_notif),
1421 .seccomp_notif_resp = sizeof(struct seccomp_notif_resp),
1422 .seccomp_data = sizeof(struct seccomp_data),
1425 if (copy_to_user(usizes, &sizes, sizeof(sizes)))
1431 /* Common entry point for both prctl and syscall. */
1432 static long do_seccomp(unsigned int op, unsigned int flags,
1436 case SECCOMP_SET_MODE_STRICT:
1437 if (flags != 0 || uargs != NULL)
1439 return seccomp_set_mode_strict();
1440 case SECCOMP_SET_MODE_FILTER:
1441 return seccomp_set_mode_filter(flags, uargs);
1442 case SECCOMP_GET_ACTION_AVAIL:
1446 return seccomp_get_action_avail(uargs);
1447 case SECCOMP_GET_NOTIF_SIZES:
1451 return seccomp_get_notif_sizes(uargs);
1457 SYSCALL_DEFINE3(seccomp, unsigned int, op, unsigned int, flags,
1458 void __user *, uargs)
1460 return do_seccomp(op, flags, uargs);
1464 * prctl_set_seccomp: configures current->seccomp.mode
1465 * @seccomp_mode: requested mode to use
1466 * @filter: optional struct sock_fprog for use with SECCOMP_MODE_FILTER
1468 * Returns 0 on success or -EINVAL on failure.
1470 long prctl_set_seccomp(unsigned long seccomp_mode, void __user *filter)
1475 switch (seccomp_mode) {
1476 case SECCOMP_MODE_STRICT:
1477 op = SECCOMP_SET_MODE_STRICT;
1479 * Setting strict mode through prctl always ignored filter,
1480 * so make sure it is always NULL here to pass the internal
1481 * check in do_seccomp().
1485 case SECCOMP_MODE_FILTER:
1486 op = SECCOMP_SET_MODE_FILTER;
1493 /* prctl interface doesn't have flags, so they are always zero. */
1494 return do_seccomp(op, 0, uargs);
1497 #if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
1498 static struct seccomp_filter *get_nth_filter(struct task_struct *task,
1499 unsigned long filter_off)
1501 struct seccomp_filter *orig, *filter;
1502 unsigned long count;
1505 * Note: this is only correct because the caller should be the (ptrace)
1506 * tracer of the task, otherwise lock_task_sighand is needed.
1508 spin_lock_irq(&task->sighand->siglock);
1510 if (task->seccomp.mode != SECCOMP_MODE_FILTER) {
1511 spin_unlock_irq(&task->sighand->siglock);
1512 return ERR_PTR(-EINVAL);
1515 orig = task->seccomp.filter;
1516 __get_seccomp_filter(orig);
1517 spin_unlock_irq(&task->sighand->siglock);
1520 for (filter = orig; filter; filter = filter->prev)
1523 if (filter_off >= count) {
1524 filter = ERR_PTR(-ENOENT);
1528 count -= filter_off;
1529 for (filter = orig; filter && count > 1; filter = filter->prev)
1532 if (WARN_ON(count != 1 || !filter)) {
1533 filter = ERR_PTR(-ENOENT);
1537 __get_seccomp_filter(filter);
1540 __put_seccomp_filter(orig);
1544 long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
1547 struct seccomp_filter *filter;
1548 struct sock_fprog_kern *fprog;
1551 if (!capable(CAP_SYS_ADMIN) ||
1552 current->seccomp.mode != SECCOMP_MODE_DISABLED) {
1556 filter = get_nth_filter(task, filter_off);
1558 return PTR_ERR(filter);
1560 fprog = filter->prog->orig_prog;
1562 /* This must be a new non-cBPF filter, since we save
1563 * every cBPF filter's orig_prog above when
1564 * CONFIG_CHECKPOINT_RESTORE is enabled.
1574 if (copy_to_user(data, fprog->filter, bpf_classic_proglen(fprog)))
1578 __put_seccomp_filter(filter);
1582 long seccomp_get_metadata(struct task_struct *task,
1583 unsigned long size, void __user *data)
1586 struct seccomp_filter *filter;
1587 struct seccomp_metadata kmd = {};
1589 if (!capable(CAP_SYS_ADMIN) ||
1590 current->seccomp.mode != SECCOMP_MODE_DISABLED) {
1594 size = min_t(unsigned long, size, sizeof(kmd));
1596 if (size < sizeof(kmd.filter_off))
1599 if (copy_from_user(&kmd.filter_off, data, sizeof(kmd.filter_off)))
1602 filter = get_nth_filter(task, kmd.filter_off);
1604 return PTR_ERR(filter);
1607 kmd.flags |= SECCOMP_FILTER_FLAG_LOG;
1610 if (copy_to_user(data, &kmd, size))
1613 __put_seccomp_filter(filter);
1618 #ifdef CONFIG_SYSCTL
1620 /* Human readable action names for friendly sysctl interaction */
1621 #define SECCOMP_RET_KILL_PROCESS_NAME "kill_process"
1622 #define SECCOMP_RET_KILL_THREAD_NAME "kill_thread"
1623 #define SECCOMP_RET_TRAP_NAME "trap"
1624 #define SECCOMP_RET_ERRNO_NAME "errno"
1625 #define SECCOMP_RET_USER_NOTIF_NAME "user_notif"
1626 #define SECCOMP_RET_TRACE_NAME "trace"
1627 #define SECCOMP_RET_LOG_NAME "log"
1628 #define SECCOMP_RET_ALLOW_NAME "allow"
1630 static const char seccomp_actions_avail[] =
1631 SECCOMP_RET_KILL_PROCESS_NAME " "
1632 SECCOMP_RET_KILL_THREAD_NAME " "
1633 SECCOMP_RET_TRAP_NAME " "
1634 SECCOMP_RET_ERRNO_NAME " "
1635 SECCOMP_RET_USER_NOTIF_NAME " "
1636 SECCOMP_RET_TRACE_NAME " "
1637 SECCOMP_RET_LOG_NAME " "
1638 SECCOMP_RET_ALLOW_NAME;
1640 struct seccomp_log_name {
1645 static const struct seccomp_log_name seccomp_log_names[] = {
1646 { SECCOMP_LOG_KILL_PROCESS, SECCOMP_RET_KILL_PROCESS_NAME },
1647 { SECCOMP_LOG_KILL_THREAD, SECCOMP_RET_KILL_THREAD_NAME },
1648 { SECCOMP_LOG_TRAP, SECCOMP_RET_TRAP_NAME },
1649 { SECCOMP_LOG_ERRNO, SECCOMP_RET_ERRNO_NAME },
1650 { SECCOMP_LOG_USER_NOTIF, SECCOMP_RET_USER_NOTIF_NAME },
1651 { SECCOMP_LOG_TRACE, SECCOMP_RET_TRACE_NAME },
1652 { SECCOMP_LOG_LOG, SECCOMP_RET_LOG_NAME },
1653 { SECCOMP_LOG_ALLOW, SECCOMP_RET_ALLOW_NAME },
1657 static bool seccomp_names_from_actions_logged(char *names, size_t size,
1661 const struct seccomp_log_name *cur;
1662 bool append_sep = false;
1664 for (cur = seccomp_log_names; cur->name && size; cur++) {
1667 if (!(actions_logged & cur->log))
1671 ret = strscpy(names, sep, size);
1680 ret = strscpy(names, cur->name, size);
1691 static bool seccomp_action_logged_from_name(u32 *action_logged,
1694 const struct seccomp_log_name *cur;
1696 for (cur = seccomp_log_names; cur->name; cur++) {
1697 if (!strcmp(cur->name, name)) {
1698 *action_logged = cur->log;
1706 static bool seccomp_actions_logged_from_names(u32 *actions_logged, char *names)
1710 *actions_logged = 0;
1711 while ((name = strsep(&names, " ")) && *name) {
1712 u32 action_logged = 0;
1714 if (!seccomp_action_logged_from_name(&action_logged, name))
1717 *actions_logged |= action_logged;
1723 static int read_actions_logged(struct ctl_table *ro_table, void __user *buffer,
1724 size_t *lenp, loff_t *ppos)
1726 char names[sizeof(seccomp_actions_avail)];
1727 struct ctl_table table;
1729 memset(names, 0, sizeof(names));
1731 if (!seccomp_names_from_actions_logged(names, sizeof(names),
1732 seccomp_actions_logged, " "))
1737 table.maxlen = sizeof(names);
1738 return proc_dostring(&table, 0, buffer, lenp, ppos);
1741 static int write_actions_logged(struct ctl_table *ro_table, void __user *buffer,
1742 size_t *lenp, loff_t *ppos, u32 *actions_logged)
1744 char names[sizeof(seccomp_actions_avail)];
1745 struct ctl_table table;
1748 if (!capable(CAP_SYS_ADMIN))
1751 memset(names, 0, sizeof(names));
1755 table.maxlen = sizeof(names);
1756 ret = proc_dostring(&table, 1, buffer, lenp, ppos);
1760 if (!seccomp_actions_logged_from_names(actions_logged, table.data))
1763 if (*actions_logged & SECCOMP_LOG_ALLOW)
1766 seccomp_actions_logged = *actions_logged;
1770 static void audit_actions_logged(u32 actions_logged, u32 old_actions_logged,
1773 char names[sizeof(seccomp_actions_avail)];
1774 char old_names[sizeof(seccomp_actions_avail)];
1775 const char *new = names;
1776 const char *old = old_names;
1781 memset(names, 0, sizeof(names));
1782 memset(old_names, 0, sizeof(old_names));
1786 else if (!actions_logged)
1788 else if (!seccomp_names_from_actions_logged(names, sizeof(names),
1789 actions_logged, ","))
1792 if (!old_actions_logged)
1794 else if (!seccomp_names_from_actions_logged(old_names,
1796 old_actions_logged, ","))
1799 return audit_seccomp_actions_logged(new, old, !ret);
1802 static int seccomp_actions_logged_handler(struct ctl_table *ro_table, int write,
1803 void __user *buffer, size_t *lenp,
1809 u32 actions_logged = 0;
1810 u32 old_actions_logged = seccomp_actions_logged;
1812 ret = write_actions_logged(ro_table, buffer, lenp, ppos,
1814 audit_actions_logged(actions_logged, old_actions_logged, ret);
1816 ret = read_actions_logged(ro_table, buffer, lenp, ppos);
1821 static struct ctl_path seccomp_sysctl_path[] = {
1822 { .procname = "kernel", },
1823 { .procname = "seccomp", },
1827 static struct ctl_table seccomp_sysctl_table[] = {
1829 .procname = "actions_avail",
1830 .data = (void *) &seccomp_actions_avail,
1831 .maxlen = sizeof(seccomp_actions_avail),
1833 .proc_handler = proc_dostring,
1836 .procname = "actions_logged",
1838 .proc_handler = seccomp_actions_logged_handler,
1843 static int __init seccomp_sysctl_init(void)
1845 struct ctl_table_header *hdr;
1847 hdr = register_sysctl_paths(seccomp_sysctl_path, seccomp_sysctl_table);
1849 pr_warn("seccomp: sysctl registration failed\n");
1851 kmemleak_not_leak(hdr);
1856 device_initcall(seccomp_sysctl_init)
1858 #endif /* CONFIG_SYSCTL */