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/atomic.h>
17 #include <linux/audit.h>
18 #include <linux/compat.h>
19 #include <linux/nospec.h>
20 #include <linux/prctl.h>
21 #include <linux/sched.h>
22 #include <linux/seccomp.h>
23 #include <linux/slab.h>
24 #include <linux/syscalls.h>
26 #ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER
27 #include <asm/syscall.h>
30 #ifdef CONFIG_SECCOMP_FILTER
31 #include <linux/filter.h>
32 #include <linux/pid.h>
33 #include <linux/ptrace.h>
34 #include <linux/security.h>
35 #include <linux/tracehook.h>
36 #include <linux/uaccess.h>
39 * struct seccomp_filter - container for seccomp BPF programs
41 * @usage: reference count to manage the object lifetime.
42 * get/put helpers should be used when accessing an instance
43 * outside of a lifetime-guarded section. In general, this
44 * is only needed for handling filters shared across tasks.
45 * @prev: points to a previously installed, or inherited, filter
46 * @len: the number of instructions in the program
47 * @insnsi: the BPF program instructions to evaluate
49 * seccomp_filter objects are organized in a tree linked via the @prev
50 * pointer. For any task, it appears to be a singly-linked list starting
51 * with current->seccomp.filter, the most recently attached or inherited filter.
52 * However, multiple filters may share a @prev node, by way of fork(), which
53 * results in a unidirectional tree existing in memory. This is similar to
54 * how namespaces work.
56 * seccomp_filter objects should never be modified after being attached
57 * to a task_struct (other than @usage).
59 struct seccomp_filter {
61 struct seccomp_filter *prev;
62 struct bpf_prog *prog;
65 /* Limit any path through the tree to 256KB worth of instructions. */
66 #define MAX_INSNS_PER_PATH ((1 << 18) / sizeof(struct sock_filter))
69 * Endianness is explicitly ignored and left for BPF program authors to manage
70 * as per the specific architecture.
72 static void populate_seccomp_data(struct seccomp_data *sd)
74 struct task_struct *task = current;
75 struct pt_regs *regs = task_pt_regs(task);
76 unsigned long args[6];
78 sd->nr = syscall_get_nr(task, regs);
79 sd->arch = syscall_get_arch();
80 syscall_get_arguments(task, regs, 0, 6, args);
81 sd->args[0] = args[0];
82 sd->args[1] = args[1];
83 sd->args[2] = args[2];
84 sd->args[3] = args[3];
85 sd->args[4] = args[4];
86 sd->args[5] = args[5];
87 sd->instruction_pointer = KSTK_EIP(task);
91 * seccomp_check_filter - verify seccomp filter code
92 * @filter: filter to verify
93 * @flen: length of filter
95 * Takes a previously checked filter (by bpf_check_classic) and
96 * redirects all filter code that loads struct sk_buff data
97 * and related data through seccomp_bpf_load. It also
98 * enforces length and alignment checking of those loads.
100 * Returns 0 if the rule set is legal or -EINVAL if not.
102 static int seccomp_check_filter(struct sock_filter *filter, unsigned int flen)
105 for (pc = 0; pc < flen; pc++) {
106 struct sock_filter *ftest = &filter[pc];
107 u16 code = ftest->code;
111 case BPF_LD | BPF_W | BPF_ABS:
112 ftest->code = BPF_LDX | BPF_W | BPF_ABS;
113 /* 32-bit aligned and not out of bounds. */
114 if (k >= sizeof(struct seccomp_data) || k & 3)
117 case BPF_LD | BPF_W | BPF_LEN:
118 ftest->code = BPF_LD | BPF_IMM;
119 ftest->k = sizeof(struct seccomp_data);
121 case BPF_LDX | BPF_W | BPF_LEN:
122 ftest->code = BPF_LDX | BPF_IMM;
123 ftest->k = sizeof(struct seccomp_data);
125 /* Explicitly include allowed calls. */
126 case BPF_RET | BPF_K:
127 case BPF_RET | BPF_A:
128 case BPF_ALU | BPF_ADD | BPF_K:
129 case BPF_ALU | BPF_ADD | BPF_X:
130 case BPF_ALU | BPF_SUB | BPF_K:
131 case BPF_ALU | BPF_SUB | BPF_X:
132 case BPF_ALU | BPF_MUL | BPF_K:
133 case BPF_ALU | BPF_MUL | BPF_X:
134 case BPF_ALU | BPF_DIV | BPF_K:
135 case BPF_ALU | BPF_DIV | BPF_X:
136 case BPF_ALU | BPF_AND | BPF_K:
137 case BPF_ALU | BPF_AND | BPF_X:
138 case BPF_ALU | BPF_OR | BPF_K:
139 case BPF_ALU | BPF_OR | BPF_X:
140 case BPF_ALU | BPF_XOR | BPF_K:
141 case BPF_ALU | BPF_XOR | BPF_X:
142 case BPF_ALU | BPF_LSH | BPF_K:
143 case BPF_ALU | BPF_LSH | BPF_X:
144 case BPF_ALU | BPF_RSH | BPF_K:
145 case BPF_ALU | BPF_RSH | BPF_X:
146 case BPF_ALU | BPF_NEG:
147 case BPF_LD | BPF_IMM:
148 case BPF_LDX | BPF_IMM:
149 case BPF_MISC | BPF_TAX:
150 case BPF_MISC | BPF_TXA:
151 case BPF_LD | BPF_MEM:
152 case BPF_LDX | BPF_MEM:
155 case BPF_JMP | BPF_JA:
156 case BPF_JMP | BPF_JEQ | BPF_K:
157 case BPF_JMP | BPF_JEQ | BPF_X:
158 case BPF_JMP | BPF_JGE | BPF_K:
159 case BPF_JMP | BPF_JGE | BPF_X:
160 case BPF_JMP | BPF_JGT | BPF_K:
161 case BPF_JMP | BPF_JGT | BPF_X:
162 case BPF_JMP | BPF_JSET | BPF_K:
163 case BPF_JMP | BPF_JSET | BPF_X:
173 * seccomp_run_filters - evaluates all seccomp filters against @syscall
174 * @syscall: number of the current system call
176 * Returns valid seccomp BPF response codes.
178 static u32 seccomp_run_filters(const struct seccomp_data *sd)
180 struct seccomp_data sd_local;
181 u32 ret = SECCOMP_RET_ALLOW;
182 /* Make sure cross-thread synced filter points somewhere sane. */
183 struct seccomp_filter *f =
184 lockless_dereference(current->seccomp.filter);
186 /* Ensure unexpected behavior doesn't result in failing open. */
187 if (unlikely(WARN_ON(f == NULL)))
188 return SECCOMP_RET_KILL;
191 populate_seccomp_data(&sd_local);
196 * All filters in the list are evaluated and the lowest BPF return
197 * value always takes priority (ignoring the DATA).
199 for (; f; f = f->prev) {
200 u32 cur_ret = BPF_PROG_RUN(f->prog, (void *)sd);
202 if ((cur_ret & SECCOMP_RET_ACTION) < (ret & SECCOMP_RET_ACTION))
207 #endif /* CONFIG_SECCOMP_FILTER */
209 static inline bool seccomp_may_assign_mode(unsigned long seccomp_mode)
211 assert_spin_locked(¤t->sighand->siglock);
213 if (current->seccomp.mode && current->seccomp.mode != seccomp_mode)
219 void __weak arch_seccomp_spec_mitigate(struct task_struct *task) { }
221 static inline void seccomp_assign_mode(struct task_struct *task,
222 unsigned long seccomp_mode,
225 assert_spin_locked(&task->sighand->siglock);
227 task->seccomp.mode = seccomp_mode;
229 * Make sure TIF_SECCOMP cannot be set before the mode (and
232 smp_mb__before_atomic();
233 /* Assume default seccomp processes want spec flaw mitigation. */
234 if ((flags & SECCOMP_FILTER_FLAG_SPEC_ALLOW) == 0)
235 arch_seccomp_spec_mitigate(task);
236 set_tsk_thread_flag(task, TIF_SECCOMP);
239 #ifdef CONFIG_SECCOMP_FILTER
240 /* Returns 1 if the parent is an ancestor of the child. */
241 static int is_ancestor(struct seccomp_filter *parent,
242 struct seccomp_filter *child)
244 /* NULL is the root ancestor. */
247 for (; child; child = child->prev)
254 * seccomp_can_sync_threads: checks if all threads can be synchronized
256 * Expects sighand and cred_guard_mutex locks to be held.
258 * Returns 0 on success, -ve on error, or the pid of a thread which was
259 * either not in the correct seccomp mode or it did not have an ancestral
262 static inline pid_t seccomp_can_sync_threads(void)
264 struct task_struct *thread, *caller;
266 BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex));
267 assert_spin_locked(¤t->sighand->siglock);
269 /* Validate all threads being eligible for synchronization. */
271 for_each_thread(caller, thread) {
274 /* Skip current, since it is initiating the sync. */
275 if (thread == caller)
278 if (thread->seccomp.mode == SECCOMP_MODE_DISABLED ||
279 (thread->seccomp.mode == SECCOMP_MODE_FILTER &&
280 is_ancestor(thread->seccomp.filter,
281 caller->seccomp.filter)))
284 /* Return the first thread that cannot be synchronized. */
285 failed = task_pid_vnr(thread);
286 /* If the pid cannot be resolved, then return -ESRCH */
287 if (unlikely(WARN_ON(failed == 0)))
296 * seccomp_sync_threads: sets all threads to use current's filter
298 * Expects sighand and cred_guard_mutex locks to be held, and for
299 * seccomp_can_sync_threads() to have returned success already
300 * without dropping the locks.
303 static inline void seccomp_sync_threads(unsigned long flags)
305 struct task_struct *thread, *caller;
307 BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex));
308 assert_spin_locked(¤t->sighand->siglock);
310 /* Synchronize all threads. */
312 for_each_thread(caller, thread) {
313 /* Skip current, since it needs no changes. */
314 if (thread == caller)
317 /* Get a task reference for the new leaf node. */
318 get_seccomp_filter(caller);
320 * Drop the task reference to the shared ancestor since
321 * current's path will hold a reference. (This also
322 * allows a put before the assignment.)
324 put_seccomp_filter(thread);
325 smp_store_release(&thread->seccomp.filter,
326 caller->seccomp.filter);
329 * Don't let an unprivileged task work around
330 * the no_new_privs restriction by creating
331 * a thread that sets it up, enters seccomp,
334 if (task_no_new_privs(caller))
335 task_set_no_new_privs(thread);
338 * Opt the other thread into seccomp if needed.
339 * As threads are considered to be trust-realm
340 * equivalent (see ptrace_may_access), it is safe to
341 * allow one thread to transition the other.
343 if (thread->seccomp.mode == SECCOMP_MODE_DISABLED)
344 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 = config_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 atomic_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 (is_compat_task()) {
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(flags);
469 void __get_seccomp_filter(struct seccomp_filter *filter)
471 /* Reference count is bounded by the number of total processes. */
472 atomic_inc(&filter->usage);
475 /* get_seccomp_filter - increments the reference count of the filter on @tsk */
476 void get_seccomp_filter(struct task_struct *tsk)
478 struct seccomp_filter *orig = tsk->seccomp.filter;
481 __get_seccomp_filter(orig);
484 static inline void seccomp_filter_free(struct seccomp_filter *filter)
487 bpf_prog_destroy(filter->prog);
492 static void __put_seccomp_filter(struct seccomp_filter *orig)
494 /* Clean up single-reference branches iteratively. */
495 while (orig && atomic_dec_and_test(&orig->usage)) {
496 struct seccomp_filter *freeme = orig;
498 seccomp_filter_free(freeme);
502 /* put_seccomp_filter - decrements the ref count of tsk->seccomp.filter */
503 void put_seccomp_filter(struct task_struct *tsk)
505 __put_seccomp_filter(tsk->seccomp.filter);
509 * seccomp_send_sigsys - signals the task to allow in-process syscall emulation
510 * @syscall: syscall number to send to userland
511 * @reason: filter-supplied reason code to send to userland (via si_errno)
513 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
515 static void seccomp_send_sigsys(int syscall, int reason)
518 memset(&info, 0, sizeof(info));
519 info.si_signo = SIGSYS;
520 info.si_code = SYS_SECCOMP;
521 info.si_call_addr = (void __user *)KSTK_EIP(current);
522 info.si_errno = reason;
523 info.si_arch = syscall_get_arch();
524 info.si_syscall = syscall;
525 force_sig_info(SIGSYS, &info, current);
527 #endif /* CONFIG_SECCOMP_FILTER */
530 * Secure computing mode 1 allows only read/write/exit/sigreturn.
531 * To be fully secure this must be combined with rlimit
532 * to limit the stack allocations too.
534 static int mode1_syscalls[] = {
535 __NR_seccomp_read, __NR_seccomp_write, __NR_seccomp_exit, __NR_seccomp_sigreturn,
536 0, /* null terminated */
540 static int mode1_syscalls_32[] = {
541 __NR_seccomp_read_32, __NR_seccomp_write_32, __NR_seccomp_exit_32, __NR_seccomp_sigreturn_32,
542 0, /* null terminated */
546 static void __secure_computing_strict(int this_syscall)
548 int *syscall_whitelist = mode1_syscalls;
550 if (is_compat_task())
551 syscall_whitelist = mode1_syscalls_32;
554 if (*syscall_whitelist == this_syscall)
556 } while (*++syscall_whitelist);
561 audit_seccomp(this_syscall, SIGKILL, SECCOMP_RET_KILL);
565 #ifndef CONFIG_HAVE_ARCH_SECCOMP_FILTER
566 void secure_computing_strict(int this_syscall)
568 int mode = current->seccomp.mode;
570 if (config_enabled(CONFIG_CHECKPOINT_RESTORE) &&
571 unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
574 if (mode == SECCOMP_MODE_DISABLED)
576 else if (mode == SECCOMP_MODE_STRICT)
577 __secure_computing_strict(this_syscall);
583 #ifdef CONFIG_SECCOMP_FILTER
584 static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
585 const bool recheck_after_trace)
587 u32 filter_ret, action;
591 * Make sure that any changes to mode from another thread have
592 * been seen after TIF_SECCOMP was seen.
596 filter_ret = seccomp_run_filters(sd);
597 data = filter_ret & SECCOMP_RET_DATA;
598 action = filter_ret & SECCOMP_RET_ACTION;
601 case SECCOMP_RET_ERRNO:
602 /* Set low-order bits as an errno, capped at MAX_ERRNO. */
603 if (data > MAX_ERRNO)
605 syscall_set_return_value(current, task_pt_regs(current),
609 case SECCOMP_RET_TRAP:
610 /* Show the handler the original registers. */
611 syscall_rollback(current, task_pt_regs(current));
612 /* Let the filter pass back 16 bits of data. */
613 seccomp_send_sigsys(this_syscall, data);
616 case SECCOMP_RET_TRACE:
617 /* We've been put in this state by the ptracer already. */
618 if (recheck_after_trace)
621 /* ENOSYS these calls if there is no tracer attached. */
622 if (!ptrace_event_enabled(current, PTRACE_EVENT_SECCOMP)) {
623 syscall_set_return_value(current,
624 task_pt_regs(current),
629 /* Allow the BPF to provide the event message */
630 ptrace_event(PTRACE_EVENT_SECCOMP, data);
632 * The delivery of a fatal signal during event
633 * notification may silently skip tracer notification,
634 * which could leave us with a potentially unmodified
635 * syscall that the tracer would have liked to have
636 * changed. Since the process is about to die, we just
637 * force the syscall to be skipped and let the signal
638 * kill the process and correctly handle any tracer exit
641 if (fatal_signal_pending(current))
643 /* Check if the tracer forced the syscall to be skipped. */
644 this_syscall = syscall_get_nr(current, task_pt_regs(current));
645 if (this_syscall < 0)
649 * Recheck the syscall, since it may have changed. This
650 * intentionally uses a NULL struct seccomp_data to force
651 * a reload of all registers. This does not goto skip since
652 * a skip would have already been reported.
654 if (__seccomp_filter(this_syscall, NULL, true))
659 case SECCOMP_RET_ALLOW:
662 case SECCOMP_RET_KILL:
664 audit_seccomp(this_syscall, SIGSYS, action);
671 audit_seccomp(this_syscall, 0, action);
675 static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
676 const bool recheck_after_trace)
682 int __secure_computing(const struct seccomp_data *sd)
684 int mode = current->seccomp.mode;
687 if (config_enabled(CONFIG_CHECKPOINT_RESTORE) &&
688 unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
691 this_syscall = sd ? sd->nr :
692 syscall_get_nr(current, task_pt_regs(current));
695 case SECCOMP_MODE_STRICT:
696 __secure_computing_strict(this_syscall); /* may call do_exit */
698 case SECCOMP_MODE_FILTER:
699 return __seccomp_filter(this_syscall, sd, false);
704 #endif /* CONFIG_HAVE_ARCH_SECCOMP_FILTER */
706 long prctl_get_seccomp(void)
708 return current->seccomp.mode;
712 * seccomp_set_mode_strict: internal function for setting strict seccomp
714 * Once current->seccomp.mode is non-zero, it may not be changed.
716 * Returns 0 on success or -EINVAL on failure.
718 static long seccomp_set_mode_strict(void)
720 const unsigned long seccomp_mode = SECCOMP_MODE_STRICT;
723 spin_lock_irq(¤t->sighand->siglock);
725 if (!seccomp_may_assign_mode(seccomp_mode))
731 seccomp_assign_mode(current, seccomp_mode, 0);
735 spin_unlock_irq(¤t->sighand->siglock);
740 #ifdef CONFIG_SECCOMP_FILTER
742 * seccomp_set_mode_filter: internal function for setting seccomp filter
743 * @flags: flags to change filter behavior
744 * @filter: struct sock_fprog containing filter
746 * This function may be called repeatedly to install additional filters.
747 * Every filter successfully installed will be evaluated (in reverse order)
748 * for each system call the task makes.
750 * Once current->seccomp.mode is non-zero, it may not be changed.
752 * Returns 0 on success or -EINVAL on failure.
754 static long seccomp_set_mode_filter(unsigned int flags,
755 const char __user *filter)
757 const unsigned long seccomp_mode = SECCOMP_MODE_FILTER;
758 struct seccomp_filter *prepared = NULL;
761 /* Validate flags. */
762 if (flags & ~SECCOMP_FILTER_FLAG_MASK)
765 /* Prepare the new filter before holding any locks. */
766 prepared = seccomp_prepare_user_filter(filter);
767 if (IS_ERR(prepared))
768 return PTR_ERR(prepared);
771 * Make sure we cannot change seccomp or nnp state via TSYNC
772 * while another thread is in the middle of calling exec.
774 if (flags & SECCOMP_FILTER_FLAG_TSYNC &&
775 mutex_lock_killable(¤t->signal->cred_guard_mutex))
778 spin_lock_irq(¤t->sighand->siglock);
780 if (!seccomp_may_assign_mode(seccomp_mode))
783 ret = seccomp_attach_filter(flags, prepared);
786 /* Do not free the successfully attached filter. */
789 seccomp_assign_mode(current, seccomp_mode, flags);
791 spin_unlock_irq(¤t->sighand->siglock);
792 if (flags & SECCOMP_FILTER_FLAG_TSYNC)
793 mutex_unlock(¤t->signal->cred_guard_mutex);
795 seccomp_filter_free(prepared);
799 static inline long seccomp_set_mode_filter(unsigned int flags,
800 const char __user *filter)
806 /* Common entry point for both prctl and syscall. */
807 static long do_seccomp(unsigned int op, unsigned int flags,
808 const char __user *uargs)
811 case SECCOMP_SET_MODE_STRICT:
812 if (flags != 0 || uargs != NULL)
814 return seccomp_set_mode_strict();
815 case SECCOMP_SET_MODE_FILTER:
816 return seccomp_set_mode_filter(flags, uargs);
822 SYSCALL_DEFINE3(seccomp, unsigned int, op, unsigned int, flags,
823 const char __user *, uargs)
825 return do_seccomp(op, flags, uargs);
829 * prctl_set_seccomp: configures current->seccomp.mode
830 * @seccomp_mode: requested mode to use
831 * @filter: optional struct sock_fprog for use with SECCOMP_MODE_FILTER
833 * Returns 0 on success or -EINVAL on failure.
835 long prctl_set_seccomp(unsigned long seccomp_mode, char __user *filter)
840 switch (seccomp_mode) {
841 case SECCOMP_MODE_STRICT:
842 op = SECCOMP_SET_MODE_STRICT;
844 * Setting strict mode through prctl always ignored filter,
845 * so make sure it is always NULL here to pass the internal
846 * check in do_seccomp().
850 case SECCOMP_MODE_FILTER:
851 op = SECCOMP_SET_MODE_FILTER;
858 /* prctl interface doesn't have flags, so they are always zero. */
859 return do_seccomp(op, 0, uargs);
862 #if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
863 long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
866 struct seccomp_filter *filter;
867 struct sock_fprog_kern *fprog;
869 unsigned long count = 0;
871 if (!capable(CAP_SYS_ADMIN) ||
872 current->seccomp.mode != SECCOMP_MODE_DISABLED) {
876 spin_lock_irq(&task->sighand->siglock);
877 if (task->seccomp.mode != SECCOMP_MODE_FILTER) {
882 filter = task->seccomp.filter;
884 filter = filter->prev;
888 if (filter_off >= count) {
894 filter = task->seccomp.filter;
895 while (filter && count > 1) {
896 filter = filter->prev;
900 if (WARN_ON(count != 1 || !filter)) {
901 /* The filter tree shouldn't shrink while we're using it. */
906 fprog = filter->prog->orig_prog;
908 /* This must be a new non-cBPF filter, since we save every
909 * every cBPF filter's orig_prog above when
910 * CONFIG_CHECKPOINT_RESTORE is enabled.
920 __get_seccomp_filter(filter);
921 spin_unlock_irq(&task->sighand->siglock);
923 if (copy_to_user(data, fprog->filter, bpf_classic_proglen(fprog)))
926 __put_seccomp_filter(filter);
930 spin_unlock_irq(&task->sighand->siglock);