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)
220 * If a given speculation mitigation is opt-in (prctl()-controlled),
221 * select it, by disabling speculation (enabling mitigation).
223 static inline void spec_mitigate(struct task_struct *task,
226 int state = arch_prctl_spec_ctrl_get(task, which);
228 if (state > 0 && (state & PR_SPEC_PRCTL))
229 arch_prctl_spec_ctrl_set(task, which, PR_SPEC_FORCE_DISABLE);
232 static inline void seccomp_assign_mode(struct task_struct *task,
233 unsigned long seccomp_mode,
236 assert_spin_locked(&task->sighand->siglock);
238 task->seccomp.mode = seccomp_mode;
240 * Make sure TIF_SECCOMP cannot be set before the mode (and
243 smp_mb__before_atomic();
244 /* Assume default seccomp processes want spec flaw mitigation. */
245 if ((flags & SECCOMP_FILTER_FLAG_SPEC_ALLOW) == 0)
246 spec_mitigate(task, PR_SPEC_STORE_BYPASS);
247 set_tsk_thread_flag(task, TIF_SECCOMP);
250 #ifdef CONFIG_SECCOMP_FILTER
251 /* Returns 1 if the parent is an ancestor of the child. */
252 static int is_ancestor(struct seccomp_filter *parent,
253 struct seccomp_filter *child)
255 /* NULL is the root ancestor. */
258 for (; child; child = child->prev)
265 * seccomp_can_sync_threads: checks if all threads can be synchronized
267 * Expects sighand and cred_guard_mutex locks to be held.
269 * Returns 0 on success, -ve on error, or the pid of a thread which was
270 * either not in the correct seccomp mode or it did not have an ancestral
273 static inline pid_t seccomp_can_sync_threads(void)
275 struct task_struct *thread, *caller;
277 BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex));
278 assert_spin_locked(¤t->sighand->siglock);
280 /* Validate all threads being eligible for synchronization. */
282 for_each_thread(caller, thread) {
285 /* Skip current, since it is initiating the sync. */
286 if (thread == caller)
289 if (thread->seccomp.mode == SECCOMP_MODE_DISABLED ||
290 (thread->seccomp.mode == SECCOMP_MODE_FILTER &&
291 is_ancestor(thread->seccomp.filter,
292 caller->seccomp.filter)))
295 /* Return the first thread that cannot be synchronized. */
296 failed = task_pid_vnr(thread);
297 /* If the pid cannot be resolved, then return -ESRCH */
298 if (unlikely(WARN_ON(failed == 0)))
307 * seccomp_sync_threads: sets all threads to use current's filter
309 * Expects sighand and cred_guard_mutex locks to be held, and for
310 * seccomp_can_sync_threads() to have returned success already
311 * without dropping the locks.
314 static inline void seccomp_sync_threads(unsigned long flags)
316 struct task_struct *thread, *caller;
318 BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex));
319 assert_spin_locked(¤t->sighand->siglock);
321 /* Synchronize all threads. */
323 for_each_thread(caller, thread) {
324 /* Skip current, since it needs no changes. */
325 if (thread == caller)
328 /* Get a task reference for the new leaf node. */
329 get_seccomp_filter(caller);
331 * Drop the task reference to the shared ancestor since
332 * current's path will hold a reference. (This also
333 * allows a put before the assignment.)
335 put_seccomp_filter(thread);
336 smp_store_release(&thread->seccomp.filter,
337 caller->seccomp.filter);
340 * Don't let an unprivileged task work around
341 * the no_new_privs restriction by creating
342 * a thread that sets it up, enters seccomp,
345 if (task_no_new_privs(caller))
346 task_set_no_new_privs(thread);
349 * Opt the other thread into seccomp if needed.
350 * As threads are considered to be trust-realm
351 * equivalent (see ptrace_may_access), it is safe to
352 * allow one thread to transition the other.
354 if (thread->seccomp.mode == SECCOMP_MODE_DISABLED)
355 seccomp_assign_mode(thread, SECCOMP_MODE_FILTER,
361 * seccomp_prepare_filter: Prepares a seccomp filter for use.
362 * @fprog: BPF program to install
364 * Returns filter on success or an ERR_PTR on failure.
366 static struct seccomp_filter *seccomp_prepare_filter(struct sock_fprog *fprog)
368 struct seccomp_filter *sfilter;
370 const bool save_orig = IS_ENABLED(CONFIG_CHECKPOINT_RESTORE);
372 if (fprog->len == 0 || fprog->len > BPF_MAXINSNS)
373 return ERR_PTR(-EINVAL);
375 BUG_ON(INT_MAX / fprog->len < sizeof(struct sock_filter));
378 * Installing a seccomp filter requires that the task has
379 * CAP_SYS_ADMIN in its namespace or be running with no_new_privs.
380 * This avoids scenarios where unprivileged tasks can affect the
381 * behavior of privileged children.
383 if (!task_no_new_privs(current) &&
384 security_capable_noaudit(current_cred(), current_user_ns(),
386 return ERR_PTR(-EACCES);
388 /* Allocate a new seccomp_filter */
389 sfilter = kzalloc(sizeof(*sfilter), GFP_KERNEL | __GFP_NOWARN);
391 return ERR_PTR(-ENOMEM);
393 ret = bpf_prog_create_from_user(&sfilter->prog, fprog,
394 seccomp_check_filter, save_orig);
400 atomic_set(&sfilter->usage, 1);
406 * seccomp_prepare_user_filter - prepares a user-supplied sock_fprog
407 * @user_filter: pointer to the user data containing a sock_fprog.
409 * Returns 0 on success and non-zero otherwise.
411 static struct seccomp_filter *
412 seccomp_prepare_user_filter(const char __user *user_filter)
414 struct sock_fprog fprog;
415 struct seccomp_filter *filter = ERR_PTR(-EFAULT);
418 if (in_compat_syscall()) {
419 struct compat_sock_fprog fprog32;
420 if (copy_from_user(&fprog32, user_filter, sizeof(fprog32)))
422 fprog.len = fprog32.len;
423 fprog.filter = compat_ptr(fprog32.filter);
424 } else /* falls through to the if below. */
426 if (copy_from_user(&fprog, user_filter, sizeof(fprog)))
428 filter = seccomp_prepare_filter(&fprog);
434 * seccomp_attach_filter: validate and attach filter
435 * @flags: flags to change filter behavior
436 * @filter: seccomp filter to add to the current process
438 * Caller must be holding current->sighand->siglock lock.
440 * Returns 0 on success, -ve on error.
442 static long seccomp_attach_filter(unsigned int flags,
443 struct seccomp_filter *filter)
445 unsigned long total_insns;
446 struct seccomp_filter *walker;
448 assert_spin_locked(¤t->sighand->siglock);
450 /* Validate resulting filter length. */
451 total_insns = filter->prog->len;
452 for (walker = current->seccomp.filter; walker; walker = walker->prev)
453 total_insns += walker->prog->len + 4; /* 4 instr penalty */
454 if (total_insns > MAX_INSNS_PER_PATH)
457 /* If thread sync has been requested, check that it is possible. */
458 if (flags & SECCOMP_FILTER_FLAG_TSYNC) {
461 ret = seccomp_can_sync_threads();
467 * If there is an existing filter, make it the prev and don't drop its
470 filter->prev = current->seccomp.filter;
471 current->seccomp.filter = filter;
473 /* Now that the new filter is in place, synchronize to all threads. */
474 if (flags & SECCOMP_FILTER_FLAG_TSYNC)
475 seccomp_sync_threads(flags);
480 void __get_seccomp_filter(struct seccomp_filter *filter)
482 /* Reference count is bounded by the number of total processes. */
483 atomic_inc(&filter->usage);
486 /* get_seccomp_filter - increments the reference count of the filter on @tsk */
487 void get_seccomp_filter(struct task_struct *tsk)
489 struct seccomp_filter *orig = tsk->seccomp.filter;
492 __get_seccomp_filter(orig);
495 static inline void seccomp_filter_free(struct seccomp_filter *filter)
498 bpf_prog_destroy(filter->prog);
503 static void __put_seccomp_filter(struct seccomp_filter *orig)
505 /* Clean up single-reference branches iteratively. */
506 while (orig && atomic_dec_and_test(&orig->usage)) {
507 struct seccomp_filter *freeme = orig;
509 seccomp_filter_free(freeme);
513 /* put_seccomp_filter - decrements the ref count of tsk->seccomp.filter */
514 void put_seccomp_filter(struct task_struct *tsk)
516 __put_seccomp_filter(tsk->seccomp.filter);
520 * seccomp_send_sigsys - signals the task to allow in-process syscall emulation
521 * @syscall: syscall number to send to userland
522 * @reason: filter-supplied reason code to send to userland (via si_errno)
524 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
526 static void seccomp_send_sigsys(int syscall, int reason)
529 memset(&info, 0, sizeof(info));
530 info.si_signo = SIGSYS;
531 info.si_code = SYS_SECCOMP;
532 info.si_call_addr = (void __user *)KSTK_EIP(current);
533 info.si_errno = reason;
534 info.si_arch = syscall_get_arch();
535 info.si_syscall = syscall;
536 force_sig_info(SIGSYS, &info, current);
538 #endif /* CONFIG_SECCOMP_FILTER */
541 * Secure computing mode 1 allows only read/write/exit/sigreturn.
542 * To be fully secure this must be combined with rlimit
543 * to limit the stack allocations too.
545 static const int mode1_syscalls[] = {
546 __NR_seccomp_read, __NR_seccomp_write, __NR_seccomp_exit, __NR_seccomp_sigreturn,
547 0, /* null terminated */
550 static void __secure_computing_strict(int this_syscall)
552 const int *syscall_whitelist = mode1_syscalls;
554 if (in_compat_syscall())
555 syscall_whitelist = get_compat_mode1_syscalls();
558 if (*syscall_whitelist == this_syscall)
560 } while (*++syscall_whitelist);
565 audit_seccomp(this_syscall, SIGKILL, SECCOMP_RET_KILL);
569 #ifndef CONFIG_HAVE_ARCH_SECCOMP_FILTER
570 void secure_computing_strict(int this_syscall)
572 int mode = current->seccomp.mode;
574 if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) &&
575 unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
578 if (mode == SECCOMP_MODE_DISABLED)
580 else if (mode == SECCOMP_MODE_STRICT)
581 __secure_computing_strict(this_syscall);
587 #ifdef CONFIG_SECCOMP_FILTER
588 static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
589 const bool recheck_after_trace)
591 u32 filter_ret, action;
595 * Make sure that any changes to mode from another thread have
596 * been seen after TIF_SECCOMP was seen.
600 filter_ret = seccomp_run_filters(sd);
601 data = filter_ret & SECCOMP_RET_DATA;
602 action = filter_ret & SECCOMP_RET_ACTION;
605 case SECCOMP_RET_ERRNO:
606 /* Set low-order bits as an errno, capped at MAX_ERRNO. */
607 if (data > MAX_ERRNO)
609 syscall_set_return_value(current, task_pt_regs(current),
613 case SECCOMP_RET_TRAP:
614 /* Show the handler the original registers. */
615 syscall_rollback(current, task_pt_regs(current));
616 /* Let the filter pass back 16 bits of data. */
617 seccomp_send_sigsys(this_syscall, data);
620 case SECCOMP_RET_TRACE:
621 /* We've been put in this state by the ptracer already. */
622 if (recheck_after_trace)
625 /* ENOSYS these calls if there is no tracer attached. */
626 if (!ptrace_event_enabled(current, PTRACE_EVENT_SECCOMP)) {
627 syscall_set_return_value(current,
628 task_pt_regs(current),
633 /* Allow the BPF to provide the event message */
634 ptrace_event(PTRACE_EVENT_SECCOMP, data);
636 * The delivery of a fatal signal during event
637 * notification may silently skip tracer notification,
638 * which could leave us with a potentially unmodified
639 * syscall that the tracer would have liked to have
640 * changed. Since the process is about to die, we just
641 * force the syscall to be skipped and let the signal
642 * kill the process and correctly handle any tracer exit
645 if (fatal_signal_pending(current))
647 /* Check if the tracer forced the syscall to be skipped. */
648 this_syscall = syscall_get_nr(current, task_pt_regs(current));
649 if (this_syscall < 0)
653 * Recheck the syscall, since it may have changed. This
654 * intentionally uses a NULL struct seccomp_data to force
655 * a reload of all registers. This does not goto skip since
656 * a skip would have already been reported.
658 if (__seccomp_filter(this_syscall, NULL, true))
663 case SECCOMP_RET_ALLOW:
666 case SECCOMP_RET_KILL:
668 audit_seccomp(this_syscall, SIGSYS, action);
675 audit_seccomp(this_syscall, 0, action);
679 static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd,
680 const bool recheck_after_trace)
686 int __secure_computing(const struct seccomp_data *sd)
688 int mode = current->seccomp.mode;
691 if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) &&
692 unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
695 this_syscall = sd ? sd->nr :
696 syscall_get_nr(current, task_pt_regs(current));
699 case SECCOMP_MODE_STRICT:
700 __secure_computing_strict(this_syscall); /* may call do_exit */
702 case SECCOMP_MODE_FILTER:
703 return __seccomp_filter(this_syscall, sd, false);
708 #endif /* CONFIG_HAVE_ARCH_SECCOMP_FILTER */
710 long prctl_get_seccomp(void)
712 return current->seccomp.mode;
716 * seccomp_set_mode_strict: internal function for setting strict seccomp
718 * Once current->seccomp.mode is non-zero, it may not be changed.
720 * Returns 0 on success or -EINVAL on failure.
722 static long seccomp_set_mode_strict(void)
724 const unsigned long seccomp_mode = SECCOMP_MODE_STRICT;
727 spin_lock_irq(¤t->sighand->siglock);
729 if (!seccomp_may_assign_mode(seccomp_mode))
735 seccomp_assign_mode(current, seccomp_mode, 0);
739 spin_unlock_irq(¤t->sighand->siglock);
744 #ifdef CONFIG_SECCOMP_FILTER
746 * seccomp_set_mode_filter: internal function for setting seccomp filter
747 * @flags: flags to change filter behavior
748 * @filter: struct sock_fprog containing filter
750 * This function may be called repeatedly to install additional filters.
751 * Every filter successfully installed will be evaluated (in reverse order)
752 * for each system call the task makes.
754 * Once current->seccomp.mode is non-zero, it may not be changed.
756 * Returns 0 on success or -EINVAL on failure.
758 static long seccomp_set_mode_filter(unsigned int flags,
759 const char __user *filter)
761 const unsigned long seccomp_mode = SECCOMP_MODE_FILTER;
762 struct seccomp_filter *prepared = NULL;
765 /* Validate flags. */
766 if (flags & ~SECCOMP_FILTER_FLAG_MASK)
769 /* Prepare the new filter before holding any locks. */
770 prepared = seccomp_prepare_user_filter(filter);
771 if (IS_ERR(prepared))
772 return PTR_ERR(prepared);
775 * Make sure we cannot change seccomp or nnp state via TSYNC
776 * while another thread is in the middle of calling exec.
778 if (flags & SECCOMP_FILTER_FLAG_TSYNC &&
779 mutex_lock_killable(¤t->signal->cred_guard_mutex))
782 spin_lock_irq(¤t->sighand->siglock);
784 if (!seccomp_may_assign_mode(seccomp_mode))
787 ret = seccomp_attach_filter(flags, prepared);
790 /* Do not free the successfully attached filter. */
793 seccomp_assign_mode(current, seccomp_mode, flags);
795 spin_unlock_irq(¤t->sighand->siglock);
796 if (flags & SECCOMP_FILTER_FLAG_TSYNC)
797 mutex_unlock(¤t->signal->cred_guard_mutex);
799 seccomp_filter_free(prepared);
803 static inline long seccomp_set_mode_filter(unsigned int flags,
804 const char __user *filter)
810 /* Common entry point for both prctl and syscall. */
811 static long do_seccomp(unsigned int op, unsigned int flags,
812 const char __user *uargs)
815 case SECCOMP_SET_MODE_STRICT:
816 if (flags != 0 || uargs != NULL)
818 return seccomp_set_mode_strict();
819 case SECCOMP_SET_MODE_FILTER:
820 return seccomp_set_mode_filter(flags, uargs);
826 SYSCALL_DEFINE3(seccomp, unsigned int, op, unsigned int, flags,
827 const char __user *, uargs)
829 return do_seccomp(op, flags, uargs);
833 * prctl_set_seccomp: configures current->seccomp.mode
834 * @seccomp_mode: requested mode to use
835 * @filter: optional struct sock_fprog for use with SECCOMP_MODE_FILTER
837 * Returns 0 on success or -EINVAL on failure.
839 long prctl_set_seccomp(unsigned long seccomp_mode, char __user *filter)
844 switch (seccomp_mode) {
845 case SECCOMP_MODE_STRICT:
846 op = SECCOMP_SET_MODE_STRICT;
848 * Setting strict mode through prctl always ignored filter,
849 * so make sure it is always NULL here to pass the internal
850 * check in do_seccomp().
854 case SECCOMP_MODE_FILTER:
855 op = SECCOMP_SET_MODE_FILTER;
862 /* prctl interface doesn't have flags, so they are always zero. */
863 return do_seccomp(op, 0, uargs);
866 #if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
867 long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
870 struct seccomp_filter *filter;
871 struct sock_fprog_kern *fprog;
873 unsigned long count = 0;
875 if (!capable(CAP_SYS_ADMIN) ||
876 current->seccomp.mode != SECCOMP_MODE_DISABLED) {
880 spin_lock_irq(&task->sighand->siglock);
881 if (task->seccomp.mode != SECCOMP_MODE_FILTER) {
886 filter = task->seccomp.filter;
888 filter = filter->prev;
892 if (filter_off >= count) {
898 filter = task->seccomp.filter;
899 while (filter && count > 1) {
900 filter = filter->prev;
904 if (WARN_ON(count != 1 || !filter)) {
905 /* The filter tree shouldn't shrink while we're using it. */
910 fprog = filter->prog->orig_prog;
912 /* This must be a new non-cBPF filter, since we save
913 * every cBPF filter's orig_prog above when
914 * CONFIG_CHECKPOINT_RESTORE is enabled.
924 __get_seccomp_filter(filter);
925 spin_unlock_irq(&task->sighand->siglock);
927 if (copy_to_user(data, fprog->filter, bpf_classic_proglen(fprog)))
930 __put_seccomp_filter(filter);
934 spin_unlock_irq(&task->sighand->siglock);