1 // SPDX-License-Identifier: GPL-2.0-only
3 * linux/kernel/signal.c
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * 1997-11-02 Modified for POSIX.1b signals by Richard Henderson
9 * 2003-06-02 Jim Houston - Concurrent Computer Corp.
10 * Changes to use preallocated sigqueue structures
11 * to allow signals to be sent reliably.
14 #include <linux/slab.h>
15 #include <linux/export.h>
16 #include <linux/init.h>
17 #include <linux/sched/mm.h>
18 #include <linux/sched/user.h>
19 #include <linux/sched/debug.h>
20 #include <linux/sched/task.h>
21 #include <linux/sched/task_stack.h>
22 #include <linux/sched/cputime.h>
23 #include <linux/file.h>
25 #include <linux/proc_fs.h>
26 #include <linux/tty.h>
27 #include <linux/binfmts.h>
28 #include <linux/coredump.h>
29 #include <linux/security.h>
30 #include <linux/syscalls.h>
31 #include <linux/ptrace.h>
32 #include <linux/signal.h>
33 #include <linux/signalfd.h>
34 #include <linux/ratelimit.h>
35 #include <linux/tracehook.h>
36 #include <linux/capability.h>
37 #include <linux/freezer.h>
38 #include <linux/pid_namespace.h>
39 #include <linux/nsproxy.h>
40 #include <linux/user_namespace.h>
41 #include <linux/uprobes.h>
42 #include <linux/compat.h>
43 #include <linux/cn_proc.h>
44 #include <linux/compiler.h>
45 #include <linux/posix-timers.h>
46 #include <linux/cgroup.h>
47 #include <linux/audit.h>
49 #define CREATE_TRACE_POINTS
50 #include <trace/events/signal.h>
52 #include <asm/param.h>
53 #include <linux/uaccess.h>
54 #include <asm/unistd.h>
55 #include <asm/siginfo.h>
56 #include <asm/cacheflush.h>
57 #include <asm/syscall.h> /* for syscall_get_* */
60 * SLAB caches for signal bits.
63 static struct kmem_cache *sigqueue_cachep;
65 int print_fatal_signals __read_mostly;
67 static void __user *sig_handler(struct task_struct *t, int sig)
69 return t->sighand->action[sig - 1].sa.sa_handler;
72 static inline bool sig_handler_ignored(void __user *handler, int sig)
74 /* Is it explicitly or implicitly ignored? */
75 return handler == SIG_IGN ||
76 (handler == SIG_DFL && sig_kernel_ignore(sig));
79 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
83 handler = sig_handler(t, sig);
85 /* SIGKILL and SIGSTOP may not be sent to the global init */
86 if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
89 if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
90 handler == SIG_DFL && !(force && sig_kernel_only(sig)))
93 /* Only allow kernel generated signals to this kthread */
94 if (unlikely((t->flags & PF_KTHREAD) &&
95 (handler == SIG_KTHREAD_KERNEL) && !force))
98 return sig_handler_ignored(handler, sig);
101 static bool sig_ignored(struct task_struct *t, int sig, bool force)
104 * Blocked signals are never ignored, since the
105 * signal handler may change by the time it is
108 if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
112 * Tracers may want to know about even ignored signal unless it
113 * is SIGKILL which can't be reported anyway but can be ignored
114 * by SIGNAL_UNKILLABLE task.
116 if (t->ptrace && sig != SIGKILL)
119 return sig_task_ignored(t, sig, force);
123 * Re-calculate pending state from the set of locally pending
124 * signals, globally pending signals, and blocked signals.
126 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
131 switch (_NSIG_WORDS) {
133 for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
134 ready |= signal->sig[i] &~ blocked->sig[i];
137 case 4: ready = signal->sig[3] &~ blocked->sig[3];
138 ready |= signal->sig[2] &~ blocked->sig[2];
139 ready |= signal->sig[1] &~ blocked->sig[1];
140 ready |= signal->sig[0] &~ blocked->sig[0];
143 case 2: ready = signal->sig[1] &~ blocked->sig[1];
144 ready |= signal->sig[0] &~ blocked->sig[0];
147 case 1: ready = signal->sig[0] &~ blocked->sig[0];
152 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
154 static bool recalc_sigpending_tsk(struct task_struct *t)
156 if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
157 PENDING(&t->pending, &t->blocked) ||
158 PENDING(&t->signal->shared_pending, &t->blocked) ||
159 cgroup_task_frozen(t)) {
160 set_tsk_thread_flag(t, TIF_SIGPENDING);
165 * We must never clear the flag in another thread, or in current
166 * when it's possible the current syscall is returning -ERESTART*.
167 * So we don't clear it here, and only callers who know they should do.
173 * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
174 * This is superfluous when called on current, the wakeup is a harmless no-op.
176 void recalc_sigpending_and_wake(struct task_struct *t)
178 if (recalc_sigpending_tsk(t))
179 signal_wake_up(t, 0);
182 void recalc_sigpending(void)
184 if (!recalc_sigpending_tsk(current) && !freezing(current))
185 clear_thread_flag(TIF_SIGPENDING);
188 EXPORT_SYMBOL(recalc_sigpending);
190 void calculate_sigpending(void)
192 /* Have any signals or users of TIF_SIGPENDING been delayed
195 spin_lock_irq(¤t->sighand->siglock);
196 set_tsk_thread_flag(current, TIF_SIGPENDING);
198 spin_unlock_irq(¤t->sighand->siglock);
201 /* Given the mask, find the first available signal that should be serviced. */
203 #define SYNCHRONOUS_MASK \
204 (sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
205 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
207 int next_signal(struct sigpending *pending, sigset_t *mask)
209 unsigned long i, *s, *m, x;
212 s = pending->signal.sig;
216 * Handle the first word specially: it contains the
217 * synchronous signals that need to be dequeued first.
221 if (x & SYNCHRONOUS_MASK)
222 x &= SYNCHRONOUS_MASK;
227 switch (_NSIG_WORDS) {
229 for (i = 1; i < _NSIG_WORDS; ++i) {
233 sig = ffz(~x) + i*_NSIG_BPW + 1;
242 sig = ffz(~x) + _NSIG_BPW + 1;
253 static inline void print_dropped_signal(int sig)
255 static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
257 if (!print_fatal_signals)
260 if (!__ratelimit(&ratelimit_state))
263 pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
264 current->comm, current->pid, sig);
268 * task_set_jobctl_pending - set jobctl pending bits
270 * @mask: pending bits to set
272 * Clear @mask from @task->jobctl. @mask must be subset of
273 * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
274 * %JOBCTL_TRAPPING. If stop signo is being set, the existing signo is
275 * cleared. If @task is already being killed or exiting, this function
279 * Must be called with @task->sighand->siglock held.
282 * %true if @mask is set, %false if made noop because @task was dying.
284 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
286 BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
287 JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
288 BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
290 if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
293 if (mask & JOBCTL_STOP_SIGMASK)
294 task->jobctl &= ~JOBCTL_STOP_SIGMASK;
296 task->jobctl |= mask;
301 * task_clear_jobctl_trapping - clear jobctl trapping bit
304 * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
305 * Clear it and wake up the ptracer. Note that we don't need any further
306 * locking. @task->siglock guarantees that @task->parent points to the
310 * Must be called with @task->sighand->siglock held.
312 void task_clear_jobctl_trapping(struct task_struct *task)
314 if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
315 task->jobctl &= ~JOBCTL_TRAPPING;
316 smp_mb(); /* advised by wake_up_bit() */
317 wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
322 * task_clear_jobctl_pending - clear jobctl pending bits
324 * @mask: pending bits to clear
326 * Clear @mask from @task->jobctl. @mask must be subset of
327 * %JOBCTL_PENDING_MASK. If %JOBCTL_STOP_PENDING is being cleared, other
328 * STOP bits are cleared together.
330 * If clearing of @mask leaves no stop or trap pending, this function calls
331 * task_clear_jobctl_trapping().
334 * Must be called with @task->sighand->siglock held.
336 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
338 BUG_ON(mask & ~JOBCTL_PENDING_MASK);
340 if (mask & JOBCTL_STOP_PENDING)
341 mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
343 task->jobctl &= ~mask;
345 if (!(task->jobctl & JOBCTL_PENDING_MASK))
346 task_clear_jobctl_trapping(task);
350 * task_participate_group_stop - participate in a group stop
351 * @task: task participating in a group stop
353 * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
354 * Group stop states are cleared and the group stop count is consumed if
355 * %JOBCTL_STOP_CONSUME was set. If the consumption completes the group
356 * stop, the appropriate `SIGNAL_*` flags are set.
359 * Must be called with @task->sighand->siglock held.
362 * %true if group stop completion should be notified to the parent, %false
365 static bool task_participate_group_stop(struct task_struct *task)
367 struct signal_struct *sig = task->signal;
368 bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
370 WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
372 task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
377 if (!WARN_ON_ONCE(sig->group_stop_count == 0))
378 sig->group_stop_count--;
381 * Tell the caller to notify completion iff we are entering into a
382 * fresh group stop. Read comment in do_signal_stop() for details.
384 if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
385 signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
391 void task_join_group_stop(struct task_struct *task)
393 unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
394 struct signal_struct *sig = current->signal;
396 if (sig->group_stop_count) {
397 sig->group_stop_count++;
398 mask |= JOBCTL_STOP_CONSUME;
399 } else if (!(sig->flags & SIGNAL_STOP_STOPPED))
402 /* Have the new thread join an on-going signal group stop */
403 task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
407 * allocate a new signal queue record
408 * - this may be called without locks if and only if t == current, otherwise an
409 * appropriate lock must be held to stop the target task from exiting
411 static struct sigqueue *
412 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t gfp_flags,
413 int override_rlimit, const unsigned int sigqueue_flags)
415 struct sigqueue *q = NULL;
416 struct ucounts *ucounts = NULL;
420 * Protect access to @t credentials. This can go away when all
421 * callers hold rcu read lock.
423 * NOTE! A pending signal will hold on to the user refcount,
424 * and we get/put the refcount only when the sigpending count
425 * changes from/to zero.
428 ucounts = task_ucounts(t);
429 sigpending = inc_rlimit_get_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
434 if (override_rlimit || likely(sigpending <= task_rlimit(t, RLIMIT_SIGPENDING))) {
435 q = kmem_cache_alloc(sigqueue_cachep, gfp_flags);
437 print_dropped_signal(sig);
440 if (unlikely(q == NULL)) {
441 dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
443 INIT_LIST_HEAD(&q->list);
444 q->flags = sigqueue_flags;
445 q->ucounts = ucounts;
450 static void __sigqueue_free(struct sigqueue *q)
452 if (q->flags & SIGQUEUE_PREALLOC)
455 dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING);
458 kmem_cache_free(sigqueue_cachep, q);
461 void flush_sigqueue(struct sigpending *queue)
465 sigemptyset(&queue->signal);
466 while (!list_empty(&queue->list)) {
467 q = list_entry(queue->list.next, struct sigqueue , list);
468 list_del_init(&q->list);
474 * Flush all pending signals for this kthread.
476 void flush_signals(struct task_struct *t)
480 spin_lock_irqsave(&t->sighand->siglock, flags);
481 clear_tsk_thread_flag(t, TIF_SIGPENDING);
482 flush_sigqueue(&t->pending);
483 flush_sigqueue(&t->signal->shared_pending);
484 spin_unlock_irqrestore(&t->sighand->siglock, flags);
486 EXPORT_SYMBOL(flush_signals);
488 #ifdef CONFIG_POSIX_TIMERS
489 static void __flush_itimer_signals(struct sigpending *pending)
491 sigset_t signal, retain;
492 struct sigqueue *q, *n;
494 signal = pending->signal;
495 sigemptyset(&retain);
497 list_for_each_entry_safe(q, n, &pending->list, list) {
498 int sig = q->info.si_signo;
500 if (likely(q->info.si_code != SI_TIMER)) {
501 sigaddset(&retain, sig);
503 sigdelset(&signal, sig);
504 list_del_init(&q->list);
509 sigorsets(&pending->signal, &signal, &retain);
512 void flush_itimer_signals(void)
514 struct task_struct *tsk = current;
517 spin_lock_irqsave(&tsk->sighand->siglock, flags);
518 __flush_itimer_signals(&tsk->pending);
519 __flush_itimer_signals(&tsk->signal->shared_pending);
520 spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
524 void ignore_signals(struct task_struct *t)
528 for (i = 0; i < _NSIG; ++i)
529 t->sighand->action[i].sa.sa_handler = SIG_IGN;
535 * Flush all handlers for a task.
539 flush_signal_handlers(struct task_struct *t, int force_default)
542 struct k_sigaction *ka = &t->sighand->action[0];
543 for (i = _NSIG ; i != 0 ; i--) {
544 if (force_default || ka->sa.sa_handler != SIG_IGN)
545 ka->sa.sa_handler = SIG_DFL;
547 #ifdef __ARCH_HAS_SA_RESTORER
548 ka->sa.sa_restorer = NULL;
550 sigemptyset(&ka->sa.sa_mask);
555 bool unhandled_signal(struct task_struct *tsk, int sig)
557 void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
558 if (is_global_init(tsk))
561 if (handler != SIG_IGN && handler != SIG_DFL)
564 /* if ptraced, let the tracer determine */
568 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
571 struct sigqueue *q, *first = NULL;
574 * Collect the siginfo appropriate to this signal. Check if
575 * there is another siginfo for the same signal.
577 list_for_each_entry(q, &list->list, list) {
578 if (q->info.si_signo == sig) {
585 sigdelset(&list->signal, sig);
589 list_del_init(&first->list);
590 copy_siginfo(info, &first->info);
593 (first->flags & SIGQUEUE_PREALLOC) &&
594 (info->si_code == SI_TIMER) &&
595 (info->si_sys_private);
597 __sigqueue_free(first);
600 * Ok, it wasn't in the queue. This must be
601 * a fast-pathed signal or we must have been
602 * out of queue space. So zero out the info.
605 info->si_signo = sig;
607 info->si_code = SI_USER;
613 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
614 kernel_siginfo_t *info, bool *resched_timer)
616 int sig = next_signal(pending, mask);
619 collect_signal(sig, pending, info, resched_timer);
624 * Dequeue a signal and return the element to the caller, which is
625 * expected to free it.
627 * All callers have to hold the siglock.
629 int dequeue_signal(struct task_struct *tsk, sigset_t *mask, kernel_siginfo_t *info)
631 bool resched_timer = false;
634 /* We only dequeue private signals from ourselves, we don't let
635 * signalfd steal them
637 signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
639 signr = __dequeue_signal(&tsk->signal->shared_pending,
640 mask, info, &resched_timer);
641 #ifdef CONFIG_POSIX_TIMERS
645 * itimers are process shared and we restart periodic
646 * itimers in the signal delivery path to prevent DoS
647 * attacks in the high resolution timer case. This is
648 * compliant with the old way of self-restarting
649 * itimers, as the SIGALRM is a legacy signal and only
650 * queued once. Changing the restart behaviour to
651 * restart the timer in the signal dequeue path is
652 * reducing the timer noise on heavy loaded !highres
655 if (unlikely(signr == SIGALRM)) {
656 struct hrtimer *tmr = &tsk->signal->real_timer;
658 if (!hrtimer_is_queued(tmr) &&
659 tsk->signal->it_real_incr != 0) {
660 hrtimer_forward(tmr, tmr->base->get_time(),
661 tsk->signal->it_real_incr);
662 hrtimer_restart(tmr);
672 if (unlikely(sig_kernel_stop(signr))) {
674 * Set a marker that we have dequeued a stop signal. Our
675 * caller might release the siglock and then the pending
676 * stop signal it is about to process is no longer in the
677 * pending bitmasks, but must still be cleared by a SIGCONT
678 * (and overruled by a SIGKILL). So those cases clear this
679 * shared flag after we've set it. Note that this flag may
680 * remain set after the signal we return is ignored or
681 * handled. That doesn't matter because its only purpose
682 * is to alert stop-signal processing code when another
683 * processor has come along and cleared the flag.
685 current->jobctl |= JOBCTL_STOP_DEQUEUED;
687 #ifdef CONFIG_POSIX_TIMERS
690 * Release the siglock to ensure proper locking order
691 * of timer locks outside of siglocks. Note, we leave
692 * irqs disabled here, since the posix-timers code is
693 * about to disable them again anyway.
695 spin_unlock(&tsk->sighand->siglock);
696 posixtimer_rearm(info);
697 spin_lock(&tsk->sighand->siglock);
699 /* Don't expose the si_sys_private value to userspace */
700 info->si_sys_private = 0;
705 EXPORT_SYMBOL_GPL(dequeue_signal);
707 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
709 struct task_struct *tsk = current;
710 struct sigpending *pending = &tsk->pending;
711 struct sigqueue *q, *sync = NULL;
714 * Might a synchronous signal be in the queue?
716 if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
720 * Return the first synchronous signal in the queue.
722 list_for_each_entry(q, &pending->list, list) {
723 /* Synchronous signals have a positive si_code */
724 if ((q->info.si_code > SI_USER) &&
725 (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
733 * Check if there is another siginfo for the same signal.
735 list_for_each_entry_continue(q, &pending->list, list) {
736 if (q->info.si_signo == sync->info.si_signo)
740 sigdelset(&pending->signal, sync->info.si_signo);
743 list_del_init(&sync->list);
744 copy_siginfo(info, &sync->info);
745 __sigqueue_free(sync);
746 return info->si_signo;
750 * Tell a process that it has a new active signal..
752 * NOTE! we rely on the previous spin_lock to
753 * lock interrupts for us! We can only be called with
754 * "siglock" held, and the local interrupt must
755 * have been disabled when that got acquired!
757 * No need to set need_resched since signal event passing
758 * goes through ->blocked
760 void signal_wake_up_state(struct task_struct *t, unsigned int state)
762 set_tsk_thread_flag(t, TIF_SIGPENDING);
764 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
765 * case. We don't check t->state here because there is a race with it
766 * executing another processor and just now entering stopped state.
767 * By using wake_up_state, we ensure the process will wake up and
768 * handle its death signal.
770 if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
775 * Remove signals in mask from the pending set and queue.
776 * Returns 1 if any signals were found.
778 * All callers must be holding the siglock.
780 static void flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
782 struct sigqueue *q, *n;
785 sigandsets(&m, mask, &s->signal);
786 if (sigisemptyset(&m))
789 sigandnsets(&s->signal, &s->signal, mask);
790 list_for_each_entry_safe(q, n, &s->list, list) {
791 if (sigismember(mask, q->info.si_signo)) {
792 list_del_init(&q->list);
798 static inline int is_si_special(const struct kernel_siginfo *info)
800 return info <= SEND_SIG_PRIV;
803 static inline bool si_fromuser(const struct kernel_siginfo *info)
805 return info == SEND_SIG_NOINFO ||
806 (!is_si_special(info) && SI_FROMUSER(info));
810 * called with RCU read lock from check_kill_permission()
812 static bool kill_ok_by_cred(struct task_struct *t)
814 const struct cred *cred = current_cred();
815 const struct cred *tcred = __task_cred(t);
817 return uid_eq(cred->euid, tcred->suid) ||
818 uid_eq(cred->euid, tcred->uid) ||
819 uid_eq(cred->uid, tcred->suid) ||
820 uid_eq(cred->uid, tcred->uid) ||
821 ns_capable(tcred->user_ns, CAP_KILL);
825 * Bad permissions for sending the signal
826 * - the caller must hold the RCU read lock
828 static int check_kill_permission(int sig, struct kernel_siginfo *info,
829 struct task_struct *t)
834 if (!valid_signal(sig))
837 if (!si_fromuser(info))
840 error = audit_signal_info(sig, t); /* Let audit system see the signal */
844 if (!same_thread_group(current, t) &&
845 !kill_ok_by_cred(t)) {
848 sid = task_session(t);
850 * We don't return the error if sid == NULL. The
851 * task was unhashed, the caller must notice this.
853 if (!sid || sid == task_session(current))
861 return security_task_kill(t, info, sig, NULL);
865 * ptrace_trap_notify - schedule trap to notify ptracer
866 * @t: tracee wanting to notify tracer
868 * This function schedules sticky ptrace trap which is cleared on the next
869 * TRAP_STOP to notify ptracer of an event. @t must have been seized by
872 * If @t is running, STOP trap will be taken. If trapped for STOP and
873 * ptracer is listening for events, tracee is woken up so that it can
874 * re-trap for the new event. If trapped otherwise, STOP trap will be
875 * eventually taken without returning to userland after the existing traps
876 * are finished by PTRACE_CONT.
879 * Must be called with @task->sighand->siglock held.
881 static void ptrace_trap_notify(struct task_struct *t)
883 WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
884 assert_spin_locked(&t->sighand->siglock);
886 task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
887 ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
891 * Handle magic process-wide effects of stop/continue signals. Unlike
892 * the signal actions, these happen immediately at signal-generation
893 * time regardless of blocking, ignoring, or handling. This does the
894 * actual continuing for SIGCONT, but not the actual stopping for stop
895 * signals. The process stop is done as a signal action for SIG_DFL.
897 * Returns true if the signal should be actually delivered, otherwise
898 * it should be dropped.
900 static bool prepare_signal(int sig, struct task_struct *p, bool force)
902 struct signal_struct *signal = p->signal;
903 struct task_struct *t;
906 if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
907 if (!(signal->flags & SIGNAL_GROUP_EXIT))
908 return sig == SIGKILL;
910 * The process is in the middle of dying, nothing to do.
912 } else if (sig_kernel_stop(sig)) {
914 * This is a stop signal. Remove SIGCONT from all queues.
916 siginitset(&flush, sigmask(SIGCONT));
917 flush_sigqueue_mask(&flush, &signal->shared_pending);
918 for_each_thread(p, t)
919 flush_sigqueue_mask(&flush, &t->pending);
920 } else if (sig == SIGCONT) {
923 * Remove all stop signals from all queues, wake all threads.
925 siginitset(&flush, SIG_KERNEL_STOP_MASK);
926 flush_sigqueue_mask(&flush, &signal->shared_pending);
927 for_each_thread(p, t) {
928 flush_sigqueue_mask(&flush, &t->pending);
929 task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
930 if (likely(!(t->ptrace & PT_SEIZED)))
931 wake_up_state(t, __TASK_STOPPED);
933 ptrace_trap_notify(t);
937 * Notify the parent with CLD_CONTINUED if we were stopped.
939 * If we were in the middle of a group stop, we pretend it
940 * was already finished, and then continued. Since SIGCHLD
941 * doesn't queue we report only CLD_STOPPED, as if the next
942 * CLD_CONTINUED was dropped.
945 if (signal->flags & SIGNAL_STOP_STOPPED)
946 why |= SIGNAL_CLD_CONTINUED;
947 else if (signal->group_stop_count)
948 why |= SIGNAL_CLD_STOPPED;
952 * The first thread which returns from do_signal_stop()
953 * will take ->siglock, notice SIGNAL_CLD_MASK, and
954 * notify its parent. See get_signal().
956 signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
957 signal->group_stop_count = 0;
958 signal->group_exit_code = 0;
962 return !sig_ignored(p, sig, force);
966 * Test if P wants to take SIG. After we've checked all threads with this,
967 * it's equivalent to finding no threads not blocking SIG. Any threads not
968 * blocking SIG were ruled out because they are not running and already
969 * have pending signals. Such threads will dequeue from the shared queue
970 * as soon as they're available, so putting the signal on the shared queue
971 * will be equivalent to sending it to one such thread.
973 static inline bool wants_signal(int sig, struct task_struct *p)
975 if (sigismember(&p->blocked, sig))
978 if (p->flags & PF_EXITING)
984 if (task_is_stopped_or_traced(p))
987 return task_curr(p) || !task_sigpending(p);
990 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
992 struct signal_struct *signal = p->signal;
993 struct task_struct *t;
996 * Now find a thread we can wake up to take the signal off the queue.
998 * If the main thread wants the signal, it gets first crack.
999 * Probably the least surprising to the average bear.
1001 if (wants_signal(sig, p))
1003 else if ((type == PIDTYPE_PID) || thread_group_empty(p))
1005 * There is just one thread and it does not need to be woken.
1006 * It will dequeue unblocked signals before it runs again.
1011 * Otherwise try to find a suitable thread.
1013 t = signal->curr_target;
1014 while (!wants_signal(sig, t)) {
1016 if (t == signal->curr_target)
1018 * No thread needs to be woken.
1019 * Any eligible threads will see
1020 * the signal in the queue soon.
1024 signal->curr_target = t;
1028 * Found a killable thread. If the signal will be fatal,
1029 * then start taking the whole group down immediately.
1031 if (sig_fatal(p, sig) &&
1032 !(signal->flags & SIGNAL_GROUP_EXIT) &&
1033 !sigismember(&t->real_blocked, sig) &&
1034 (sig == SIGKILL || !p->ptrace)) {
1036 * This signal will be fatal to the whole group.
1038 if (!sig_kernel_coredump(sig)) {
1040 * Start a group exit and wake everybody up.
1041 * This way we don't have other threads
1042 * running and doing things after a slower
1043 * thread has the fatal signal pending.
1045 signal->flags = SIGNAL_GROUP_EXIT;
1046 signal->group_exit_code = sig;
1047 signal->group_stop_count = 0;
1050 task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1051 sigaddset(&t->pending.signal, SIGKILL);
1052 signal_wake_up(t, 1);
1053 } while_each_thread(p, t);
1059 * The signal is already in the shared-pending queue.
1060 * Tell the chosen thread to wake up and dequeue it.
1062 signal_wake_up(t, sig == SIGKILL);
1066 static inline bool legacy_queue(struct sigpending *signals, int sig)
1068 return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1071 static int __send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1072 enum pid_type type, bool force)
1074 struct sigpending *pending;
1076 int override_rlimit;
1077 int ret = 0, result;
1079 assert_spin_locked(&t->sighand->siglock);
1081 result = TRACE_SIGNAL_IGNORED;
1082 if (!prepare_signal(sig, t, force))
1085 pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1087 * Short-circuit ignored signals and support queuing
1088 * exactly one non-rt signal, so that we can get more
1089 * detailed information about the cause of the signal.
1091 result = TRACE_SIGNAL_ALREADY_PENDING;
1092 if (legacy_queue(pending, sig))
1095 result = TRACE_SIGNAL_DELIVERED;
1097 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1099 if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1103 * Real-time signals must be queued if sent by sigqueue, or
1104 * some other real-time mechanism. It is implementation
1105 * defined whether kill() does so. We attempt to do so, on
1106 * the principle of least surprise, but since kill is not
1107 * allowed to fail with EAGAIN when low on memory we just
1108 * make sure at least one signal gets delivered and don't
1109 * pass on the info struct.
1112 override_rlimit = (is_si_special(info) || info->si_code >= 0);
1114 override_rlimit = 0;
1116 q = __sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit, 0);
1119 list_add_tail(&q->list, &pending->list);
1120 switch ((unsigned long) info) {
1121 case (unsigned long) SEND_SIG_NOINFO:
1122 clear_siginfo(&q->info);
1123 q->info.si_signo = sig;
1124 q->info.si_errno = 0;
1125 q->info.si_code = SI_USER;
1126 q->info.si_pid = task_tgid_nr_ns(current,
1127 task_active_pid_ns(t));
1130 from_kuid_munged(task_cred_xxx(t, user_ns),
1134 case (unsigned long) SEND_SIG_PRIV:
1135 clear_siginfo(&q->info);
1136 q->info.si_signo = sig;
1137 q->info.si_errno = 0;
1138 q->info.si_code = SI_KERNEL;
1143 copy_siginfo(&q->info, info);
1146 } else if (!is_si_special(info) &&
1147 sig >= SIGRTMIN && info->si_code != SI_USER) {
1149 * Queue overflow, abort. We may abort if the
1150 * signal was rt and sent by user using something
1151 * other than kill().
1153 result = TRACE_SIGNAL_OVERFLOW_FAIL;
1158 * This is a silent loss of information. We still
1159 * send the signal, but the *info bits are lost.
1161 result = TRACE_SIGNAL_LOSE_INFO;
1165 signalfd_notify(t, sig);
1166 sigaddset(&pending->signal, sig);
1168 /* Let multiprocess signals appear after on-going forks */
1169 if (type > PIDTYPE_TGID) {
1170 struct multiprocess_signals *delayed;
1171 hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1172 sigset_t *signal = &delayed->signal;
1173 /* Can't queue both a stop and a continue signal */
1175 sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1176 else if (sig_kernel_stop(sig))
1177 sigdelset(signal, SIGCONT);
1178 sigaddset(signal, sig);
1182 complete_signal(sig, t, type);
1184 trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1188 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1191 switch (siginfo_layout(info->si_signo, info->si_code)) {
1200 case SIL_FAULT_TRAPNO:
1201 case SIL_FAULT_MCEERR:
1202 case SIL_FAULT_BNDERR:
1203 case SIL_FAULT_PKUERR:
1204 case SIL_FAULT_PERF_EVENT:
1212 static int send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1215 /* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1218 if (info == SEND_SIG_NOINFO) {
1219 /* Force if sent from an ancestor pid namespace */
1220 force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1221 } else if (info == SEND_SIG_PRIV) {
1222 /* Don't ignore kernel generated signals */
1224 } else if (has_si_pid_and_uid(info)) {
1225 /* SIGKILL and SIGSTOP is special or has ids */
1226 struct user_namespace *t_user_ns;
1229 t_user_ns = task_cred_xxx(t, user_ns);
1230 if (current_user_ns() != t_user_ns) {
1231 kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1232 info->si_uid = from_kuid_munged(t_user_ns, uid);
1236 /* A kernel generated signal? */
1237 force = (info->si_code == SI_KERNEL);
1239 /* From an ancestor pid namespace? */
1240 if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1245 return __send_signal(sig, info, t, type, force);
1248 static void print_fatal_signal(int signr)
1250 struct pt_regs *regs = signal_pt_regs();
1251 pr_info("potentially unexpected fatal signal %d.\n", signr);
1253 #if defined(__i386__) && !defined(__arch_um__)
1254 pr_info("code at %08lx: ", regs->ip);
1257 for (i = 0; i < 16; i++) {
1260 if (get_user(insn, (unsigned char *)(regs->ip + i)))
1262 pr_cont("%02x ", insn);
1272 static int __init setup_print_fatal_signals(char *str)
1274 get_option (&str, &print_fatal_signals);
1279 __setup("print-fatal-signals=", setup_print_fatal_signals);
1282 __group_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1284 return send_signal(sig, info, p, PIDTYPE_TGID);
1287 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1290 unsigned long flags;
1293 if (lock_task_sighand(p, &flags)) {
1294 ret = send_signal(sig, info, p, type);
1295 unlock_task_sighand(p, &flags);
1302 * Force a signal that the process can't ignore: if necessary
1303 * we unblock the signal and change any SIG_IGN to SIG_DFL.
1305 * Note: If we unblock the signal, we always reset it to SIG_DFL,
1306 * since we do not want to have a signal handler that was blocked
1307 * be invoked when user space had explicitly blocked it.
1309 * We don't want to have recursive SIGSEGV's etc, for example,
1310 * that is why we also clear SIGNAL_UNKILLABLE.
1313 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t, bool sigdfl)
1315 unsigned long int flags;
1316 int ret, blocked, ignored;
1317 struct k_sigaction *action;
1318 int sig = info->si_signo;
1320 spin_lock_irqsave(&t->sighand->siglock, flags);
1321 action = &t->sighand->action[sig-1];
1322 ignored = action->sa.sa_handler == SIG_IGN;
1323 blocked = sigismember(&t->blocked, sig);
1324 if (blocked || ignored || sigdfl) {
1325 action->sa.sa_handler = SIG_DFL;
1326 action->sa.sa_flags |= SA_IMMUTABLE;
1328 sigdelset(&t->blocked, sig);
1329 recalc_sigpending_and_wake(t);
1333 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1334 * debugging to leave init killable.
1336 if (action->sa.sa_handler == SIG_DFL && !t->ptrace)
1337 t->signal->flags &= ~SIGNAL_UNKILLABLE;
1338 ret = send_signal(sig, info, t, PIDTYPE_PID);
1339 spin_unlock_irqrestore(&t->sighand->siglock, flags);
1344 int force_sig_info(struct kernel_siginfo *info)
1346 return force_sig_info_to_task(info, current, false);
1350 * Nuke all other threads in the group.
1352 int zap_other_threads(struct task_struct *p)
1354 struct task_struct *t = p;
1357 p->signal->group_stop_count = 0;
1359 while_each_thread(p, t) {
1360 task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1363 /* Don't bother with already dead threads */
1366 sigaddset(&t->pending.signal, SIGKILL);
1367 signal_wake_up(t, 1);
1373 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1374 unsigned long *flags)
1376 struct sighand_struct *sighand;
1380 sighand = rcu_dereference(tsk->sighand);
1381 if (unlikely(sighand == NULL))
1385 * This sighand can be already freed and even reused, but
1386 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1387 * initializes ->siglock: this slab can't go away, it has
1388 * the same object type, ->siglock can't be reinitialized.
1390 * We need to ensure that tsk->sighand is still the same
1391 * after we take the lock, we can race with de_thread() or
1392 * __exit_signal(). In the latter case the next iteration
1393 * must see ->sighand == NULL.
1395 spin_lock_irqsave(&sighand->siglock, *flags);
1396 if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1398 spin_unlock_irqrestore(&sighand->siglock, *flags);
1405 #ifdef CONFIG_LOCKDEP
1406 void lockdep_assert_task_sighand_held(struct task_struct *task)
1408 struct sighand_struct *sighand;
1411 sighand = rcu_dereference(task->sighand);
1413 lockdep_assert_held(&sighand->siglock);
1421 * send signal info to all the members of a group
1423 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1424 struct task_struct *p, enum pid_type type)
1429 ret = check_kill_permission(sig, info, p);
1433 ret = do_send_sig_info(sig, info, p, type);
1439 * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1440 * control characters do (^C, ^Z etc)
1441 * - the caller must hold at least a readlock on tasklist_lock
1443 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1445 struct task_struct *p = NULL;
1446 int retval, success;
1450 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1451 int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1454 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1455 return success ? 0 : retval;
1458 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1461 struct task_struct *p;
1465 p = pid_task(pid, PIDTYPE_PID);
1467 error = group_send_sig_info(sig, info, p, PIDTYPE_TGID);
1469 if (likely(!p || error != -ESRCH))
1473 * The task was unhashed in between, try again. If it
1474 * is dead, pid_task() will return NULL, if we race with
1475 * de_thread() it will find the new leader.
1480 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1484 error = kill_pid_info(sig, info, find_vpid(pid));
1489 static inline bool kill_as_cred_perm(const struct cred *cred,
1490 struct task_struct *target)
1492 const struct cred *pcred = __task_cred(target);
1494 return uid_eq(cred->euid, pcred->suid) ||
1495 uid_eq(cred->euid, pcred->uid) ||
1496 uid_eq(cred->uid, pcred->suid) ||
1497 uid_eq(cred->uid, pcred->uid);
1501 * The usb asyncio usage of siginfo is wrong. The glibc support
1502 * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1503 * AKA after the generic fields:
1504 * kernel_pid_t si_pid;
1505 * kernel_uid32_t si_uid;
1506 * sigval_t si_value;
1508 * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1509 * after the generic fields is:
1510 * void __user *si_addr;
1512 * This is a practical problem when there is a 64bit big endian kernel
1513 * and a 32bit userspace. As the 32bit address will encoded in the low
1514 * 32bits of the pointer. Those low 32bits will be stored at higher
1515 * address than appear in a 32 bit pointer. So userspace will not
1516 * see the address it was expecting for it's completions.
1518 * There is nothing in the encoding that can allow
1519 * copy_siginfo_to_user32 to detect this confusion of formats, so
1520 * handle this by requiring the caller of kill_pid_usb_asyncio to
1521 * notice when this situration takes place and to store the 32bit
1522 * pointer in sival_int, instead of sival_addr of the sigval_t addr
1525 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1526 struct pid *pid, const struct cred *cred)
1528 struct kernel_siginfo info;
1529 struct task_struct *p;
1530 unsigned long flags;
1533 if (!valid_signal(sig))
1536 clear_siginfo(&info);
1537 info.si_signo = sig;
1538 info.si_errno = errno;
1539 info.si_code = SI_ASYNCIO;
1540 *((sigval_t *)&info.si_pid) = addr;
1543 p = pid_task(pid, PIDTYPE_PID);
1548 if (!kill_as_cred_perm(cred, p)) {
1552 ret = security_task_kill(p, &info, sig, cred);
1557 if (lock_task_sighand(p, &flags)) {
1558 ret = __send_signal(sig, &info, p, PIDTYPE_TGID, false);
1559 unlock_task_sighand(p, &flags);
1567 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1570 * kill_something_info() interprets pid in interesting ways just like kill(2).
1572 * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1573 * is probably wrong. Should make it like BSD or SYSV.
1576 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1581 return kill_proc_info(sig, info, pid);
1583 /* -INT_MIN is undefined. Exclude this case to avoid a UBSAN warning */
1587 read_lock(&tasklist_lock);
1589 ret = __kill_pgrp_info(sig, info,
1590 pid ? find_vpid(-pid) : task_pgrp(current));
1592 int retval = 0, count = 0;
1593 struct task_struct * p;
1595 for_each_process(p) {
1596 if (task_pid_vnr(p) > 1 &&
1597 !same_thread_group(p, current)) {
1598 int err = group_send_sig_info(sig, info, p,
1605 ret = count ? retval : -ESRCH;
1607 read_unlock(&tasklist_lock);
1613 * These are for backward compatibility with the rest of the kernel source.
1616 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1619 * Make sure legacy kernel users don't send in bad values
1620 * (normal paths check this in check_kill_permission).
1622 if (!valid_signal(sig))
1625 return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1627 EXPORT_SYMBOL(send_sig_info);
1629 #define __si_special(priv) \
1630 ((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1633 send_sig(int sig, struct task_struct *p, int priv)
1635 return send_sig_info(sig, __si_special(priv), p);
1637 EXPORT_SYMBOL(send_sig);
1639 void force_sig(int sig)
1641 struct kernel_siginfo info;
1643 clear_siginfo(&info);
1644 info.si_signo = sig;
1646 info.si_code = SI_KERNEL;
1649 force_sig_info(&info);
1651 EXPORT_SYMBOL(force_sig);
1654 * When things go south during signal handling, we
1655 * will force a SIGSEGV. And if the signal that caused
1656 * the problem was already a SIGSEGV, we'll want to
1657 * make sure we don't even try to deliver the signal..
1659 void force_sigsegv(int sig)
1661 struct task_struct *p = current;
1663 if (sig == SIGSEGV) {
1664 unsigned long flags;
1665 spin_lock_irqsave(&p->sighand->siglock, flags);
1666 p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1667 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1672 int force_sig_fault_to_task(int sig, int code, void __user *addr
1673 ___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1674 , struct task_struct *t)
1676 struct kernel_siginfo info;
1678 clear_siginfo(&info);
1679 info.si_signo = sig;
1681 info.si_code = code;
1682 info.si_addr = addr;
1685 info.si_flags = flags;
1688 return force_sig_info_to_task(&info, t, false);
1691 int force_sig_fault(int sig, int code, void __user *addr
1692 ___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr))
1694 return force_sig_fault_to_task(sig, code, addr
1695 ___ARCH_SI_IA64(imm, flags, isr), current);
1698 int send_sig_fault(int sig, int code, void __user *addr
1699 ___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1700 , struct task_struct *t)
1702 struct kernel_siginfo info;
1704 clear_siginfo(&info);
1705 info.si_signo = sig;
1707 info.si_code = code;
1708 info.si_addr = addr;
1711 info.si_flags = flags;
1714 return send_sig_info(info.si_signo, &info, t);
1717 int force_sig_mceerr(int code, void __user *addr, short lsb)
1719 struct kernel_siginfo info;
1721 WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1722 clear_siginfo(&info);
1723 info.si_signo = SIGBUS;
1725 info.si_code = code;
1726 info.si_addr = addr;
1727 info.si_addr_lsb = lsb;
1728 return force_sig_info(&info);
1731 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1733 struct kernel_siginfo info;
1735 WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1736 clear_siginfo(&info);
1737 info.si_signo = SIGBUS;
1739 info.si_code = code;
1740 info.si_addr = addr;
1741 info.si_addr_lsb = lsb;
1742 return send_sig_info(info.si_signo, &info, t);
1744 EXPORT_SYMBOL(send_sig_mceerr);
1746 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1748 struct kernel_siginfo info;
1750 clear_siginfo(&info);
1751 info.si_signo = SIGSEGV;
1753 info.si_code = SEGV_BNDERR;
1754 info.si_addr = addr;
1755 info.si_lower = lower;
1756 info.si_upper = upper;
1757 return force_sig_info(&info);
1761 int force_sig_pkuerr(void __user *addr, u32 pkey)
1763 struct kernel_siginfo info;
1765 clear_siginfo(&info);
1766 info.si_signo = SIGSEGV;
1768 info.si_code = SEGV_PKUERR;
1769 info.si_addr = addr;
1770 info.si_pkey = pkey;
1771 return force_sig_info(&info);
1775 int force_sig_perf(void __user *addr, u32 type, u64 sig_data)
1777 struct kernel_siginfo info;
1779 clear_siginfo(&info);
1780 info.si_signo = SIGTRAP;
1782 info.si_code = TRAP_PERF;
1783 info.si_addr = addr;
1784 info.si_perf_data = sig_data;
1785 info.si_perf_type = type;
1787 return force_sig_info(&info);
1791 * force_sig_seccomp - signals the task to allow in-process syscall emulation
1792 * @syscall: syscall number to send to userland
1793 * @reason: filter-supplied reason code to send to userland (via si_errno)
1795 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
1797 int force_sig_seccomp(int syscall, int reason, bool force_coredump)
1799 struct kernel_siginfo info;
1801 clear_siginfo(&info);
1802 info.si_signo = SIGSYS;
1803 info.si_code = SYS_SECCOMP;
1804 info.si_call_addr = (void __user *)KSTK_EIP(current);
1805 info.si_errno = reason;
1806 info.si_arch = syscall_get_arch(current);
1807 info.si_syscall = syscall;
1808 return force_sig_info_to_task(&info, current, force_coredump);
1811 /* For the crazy architectures that include trap information in
1812 * the errno field, instead of an actual errno value.
1814 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1816 struct kernel_siginfo info;
1818 clear_siginfo(&info);
1819 info.si_signo = SIGTRAP;
1820 info.si_errno = errno;
1821 info.si_code = TRAP_HWBKPT;
1822 info.si_addr = addr;
1823 return force_sig_info(&info);
1826 /* For the rare architectures that include trap information using
1829 int force_sig_fault_trapno(int sig, int code, void __user *addr, int trapno)
1831 struct kernel_siginfo info;
1833 clear_siginfo(&info);
1834 info.si_signo = sig;
1836 info.si_code = code;
1837 info.si_addr = addr;
1838 info.si_trapno = trapno;
1839 return force_sig_info(&info);
1842 /* For the rare architectures that include trap information using
1845 int send_sig_fault_trapno(int sig, int code, void __user *addr, int trapno,
1846 struct task_struct *t)
1848 struct kernel_siginfo info;
1850 clear_siginfo(&info);
1851 info.si_signo = sig;
1853 info.si_code = code;
1854 info.si_addr = addr;
1855 info.si_trapno = trapno;
1856 return send_sig_info(info.si_signo, &info, t);
1859 int kill_pgrp(struct pid *pid, int sig, int priv)
1863 read_lock(&tasklist_lock);
1864 ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1865 read_unlock(&tasklist_lock);
1869 EXPORT_SYMBOL(kill_pgrp);
1871 int kill_pid(struct pid *pid, int sig, int priv)
1873 return kill_pid_info(sig, __si_special(priv), pid);
1875 EXPORT_SYMBOL(kill_pid);
1878 * These functions support sending signals using preallocated sigqueue
1879 * structures. This is needed "because realtime applications cannot
1880 * afford to lose notifications of asynchronous events, like timer
1881 * expirations or I/O completions". In the case of POSIX Timers
1882 * we allocate the sigqueue structure from the timer_create. If this
1883 * allocation fails we are able to report the failure to the application
1884 * with an EAGAIN error.
1886 struct sigqueue *sigqueue_alloc(void)
1888 return __sigqueue_alloc(-1, current, GFP_KERNEL, 0, SIGQUEUE_PREALLOC);
1891 void sigqueue_free(struct sigqueue *q)
1893 unsigned long flags;
1894 spinlock_t *lock = ¤t->sighand->siglock;
1896 BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1898 * We must hold ->siglock while testing q->list
1899 * to serialize with collect_signal() or with
1900 * __exit_signal()->flush_sigqueue().
1902 spin_lock_irqsave(lock, flags);
1903 q->flags &= ~SIGQUEUE_PREALLOC;
1905 * If it is queued it will be freed when dequeued,
1906 * like the "regular" sigqueue.
1908 if (!list_empty(&q->list))
1910 spin_unlock_irqrestore(lock, flags);
1916 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1918 int sig = q->info.si_signo;
1919 struct sigpending *pending;
1920 struct task_struct *t;
1921 unsigned long flags;
1924 BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1928 t = pid_task(pid, type);
1929 if (!t || !likely(lock_task_sighand(t, &flags)))
1932 ret = 1; /* the signal is ignored */
1933 result = TRACE_SIGNAL_IGNORED;
1934 if (!prepare_signal(sig, t, false))
1938 if (unlikely(!list_empty(&q->list))) {
1940 * If an SI_TIMER entry is already queue just increment
1941 * the overrun count.
1943 BUG_ON(q->info.si_code != SI_TIMER);
1944 q->info.si_overrun++;
1945 result = TRACE_SIGNAL_ALREADY_PENDING;
1948 q->info.si_overrun = 0;
1950 signalfd_notify(t, sig);
1951 pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1952 list_add_tail(&q->list, &pending->list);
1953 sigaddset(&pending->signal, sig);
1954 complete_signal(sig, t, type);
1955 result = TRACE_SIGNAL_DELIVERED;
1957 trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
1958 unlock_task_sighand(t, &flags);
1964 static void do_notify_pidfd(struct task_struct *task)
1968 WARN_ON(task->exit_state == 0);
1969 pid = task_pid(task);
1970 wake_up_all(&pid->wait_pidfd);
1974 * Let a parent know about the death of a child.
1975 * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1977 * Returns true if our parent ignored us and so we've switched to
1980 bool do_notify_parent(struct task_struct *tsk, int sig)
1982 struct kernel_siginfo info;
1983 unsigned long flags;
1984 struct sighand_struct *psig;
1985 bool autoreap = false;
1990 /* do_notify_parent_cldstop should have been called instead. */
1991 BUG_ON(task_is_stopped_or_traced(tsk));
1993 BUG_ON(!tsk->ptrace &&
1994 (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1996 /* Wake up all pidfd waiters */
1997 do_notify_pidfd(tsk);
1999 if (sig != SIGCHLD) {
2001 * This is only possible if parent == real_parent.
2002 * Check if it has changed security domain.
2004 if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
2008 clear_siginfo(&info);
2009 info.si_signo = sig;
2012 * We are under tasklist_lock here so our parent is tied to
2013 * us and cannot change.
2015 * task_active_pid_ns will always return the same pid namespace
2016 * until a task passes through release_task.
2018 * write_lock() currently calls preempt_disable() which is the
2019 * same as rcu_read_lock(), but according to Oleg, this is not
2020 * correct to rely on this
2023 info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
2024 info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
2028 task_cputime(tsk, &utime, &stime);
2029 info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
2030 info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
2032 info.si_status = tsk->exit_code & 0x7f;
2033 if (tsk->exit_code & 0x80)
2034 info.si_code = CLD_DUMPED;
2035 else if (tsk->exit_code & 0x7f)
2036 info.si_code = CLD_KILLED;
2038 info.si_code = CLD_EXITED;
2039 info.si_status = tsk->exit_code >> 8;
2042 psig = tsk->parent->sighand;
2043 spin_lock_irqsave(&psig->siglock, flags);
2044 if (!tsk->ptrace && sig == SIGCHLD &&
2045 (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
2046 (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
2048 * We are exiting and our parent doesn't care. POSIX.1
2049 * defines special semantics for setting SIGCHLD to SIG_IGN
2050 * or setting the SA_NOCLDWAIT flag: we should be reaped
2051 * automatically and not left for our parent's wait4 call.
2052 * Rather than having the parent do it as a magic kind of
2053 * signal handler, we just set this to tell do_exit that we
2054 * can be cleaned up without becoming a zombie. Note that
2055 * we still call __wake_up_parent in this case, because a
2056 * blocked sys_wait4 might now return -ECHILD.
2058 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
2059 * is implementation-defined: we do (if you don't want
2060 * it, just use SIG_IGN instead).
2063 if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2067 * Send with __send_signal as si_pid and si_uid are in the
2068 * parent's namespaces.
2070 if (valid_signal(sig) && sig)
2071 __send_signal(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2072 __wake_up_parent(tsk, tsk->parent);
2073 spin_unlock_irqrestore(&psig->siglock, flags);
2079 * do_notify_parent_cldstop - notify parent of stopped/continued state change
2080 * @tsk: task reporting the state change
2081 * @for_ptracer: the notification is for ptracer
2082 * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2084 * Notify @tsk's parent that the stopped/continued state has changed. If
2085 * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2086 * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2089 * Must be called with tasklist_lock at least read locked.
2091 static void do_notify_parent_cldstop(struct task_struct *tsk,
2092 bool for_ptracer, int why)
2094 struct kernel_siginfo info;
2095 unsigned long flags;
2096 struct task_struct *parent;
2097 struct sighand_struct *sighand;
2101 parent = tsk->parent;
2103 tsk = tsk->group_leader;
2104 parent = tsk->real_parent;
2107 clear_siginfo(&info);
2108 info.si_signo = SIGCHLD;
2111 * see comment in do_notify_parent() about the following 4 lines
2114 info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2115 info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2118 task_cputime(tsk, &utime, &stime);
2119 info.si_utime = nsec_to_clock_t(utime);
2120 info.si_stime = nsec_to_clock_t(stime);
2125 info.si_status = SIGCONT;
2128 info.si_status = tsk->signal->group_exit_code & 0x7f;
2131 info.si_status = tsk->exit_code & 0x7f;
2137 sighand = parent->sighand;
2138 spin_lock_irqsave(&sighand->siglock, flags);
2139 if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2140 !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2141 __group_send_sig_info(SIGCHLD, &info, parent);
2143 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2145 __wake_up_parent(tsk, parent);
2146 spin_unlock_irqrestore(&sighand->siglock, flags);
2149 static inline bool may_ptrace_stop(void)
2151 if (!likely(current->ptrace))
2154 * Are we in the middle of do_coredump?
2155 * If so and our tracer is also part of the coredump stopping
2156 * is a deadlock situation, and pointless because our tracer
2157 * is dead so don't allow us to stop.
2158 * If SIGKILL was already sent before the caller unlocked
2159 * ->siglock we must see ->core_state != NULL. Otherwise it
2160 * is safe to enter schedule().
2162 * This is almost outdated, a task with the pending SIGKILL can't
2163 * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
2164 * after SIGKILL was already dequeued.
2166 if (unlikely(current->mm->core_state) &&
2167 unlikely(current->mm == current->parent->mm))
2175 * This must be called with current->sighand->siglock held.
2177 * This should be the path for all ptrace stops.
2178 * We always set current->last_siginfo while stopped here.
2179 * That makes it a way to test a stopped process for
2180 * being ptrace-stopped vs being job-control-stopped.
2182 * If we actually decide not to stop at all because the tracer
2183 * is gone, we keep current->exit_code unless clear_code.
2185 static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t *info)
2186 __releases(¤t->sighand->siglock)
2187 __acquires(¤t->sighand->siglock)
2189 bool gstop_done = false;
2191 if (arch_ptrace_stop_needed(exit_code, info)) {
2193 * The arch code has something special to do before a
2194 * ptrace stop. This is allowed to block, e.g. for faults
2195 * on user stack pages. We can't keep the siglock while
2196 * calling arch_ptrace_stop, so we must release it now.
2197 * To preserve proper semantics, we must do this before
2198 * any signal bookkeeping like checking group_stop_count.
2200 spin_unlock_irq(¤t->sighand->siglock);
2201 arch_ptrace_stop(exit_code, info);
2202 spin_lock_irq(¤t->sighand->siglock);
2206 * schedule() will not sleep if there is a pending signal that
2207 * can awaken the task.
2209 set_special_state(TASK_TRACED);
2212 * We're committing to trapping. TRACED should be visible before
2213 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2214 * Also, transition to TRACED and updates to ->jobctl should be
2215 * atomic with respect to siglock and should be done after the arch
2216 * hook as siglock is released and regrabbed across it.
2221 * [L] wait_on_bit(JOBCTL_TRAPPING) [S] set_special_state(TRACED)
2223 * set_current_state() smp_wmb();
2225 * wait_task_stopped()
2226 * task_stopped_code()
2227 * [L] task_is_traced() [S] task_clear_jobctl_trapping();
2231 current->last_siginfo = info;
2232 current->exit_code = exit_code;
2235 * If @why is CLD_STOPPED, we're trapping to participate in a group
2236 * stop. Do the bookkeeping. Note that if SIGCONT was delievered
2237 * across siglock relocks since INTERRUPT was scheduled, PENDING
2238 * could be clear now. We act as if SIGCONT is received after
2239 * TASK_TRACED is entered - ignore it.
2241 if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2242 gstop_done = task_participate_group_stop(current);
2244 /* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2245 task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2246 if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2247 task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2249 /* entering a trap, clear TRAPPING */
2250 task_clear_jobctl_trapping(current);
2252 spin_unlock_irq(¤t->sighand->siglock);
2253 read_lock(&tasklist_lock);
2254 if (may_ptrace_stop()) {
2256 * Notify parents of the stop.
2258 * While ptraced, there are two parents - the ptracer and
2259 * the real_parent of the group_leader. The ptracer should
2260 * know about every stop while the real parent is only
2261 * interested in the completion of group stop. The states
2262 * for the two don't interact with each other. Notify
2263 * separately unless they're gonna be duplicates.
2265 do_notify_parent_cldstop(current, true, why);
2266 if (gstop_done && ptrace_reparented(current))
2267 do_notify_parent_cldstop(current, false, why);
2270 * Don't want to allow preemption here, because
2271 * sys_ptrace() needs this task to be inactive.
2273 * XXX: implement read_unlock_no_resched().
2276 read_unlock(&tasklist_lock);
2277 cgroup_enter_frozen();
2278 preempt_enable_no_resched();
2279 freezable_schedule();
2280 cgroup_leave_frozen(true);
2283 * By the time we got the lock, our tracer went away.
2284 * Don't drop the lock yet, another tracer may come.
2286 * If @gstop_done, the ptracer went away between group stop
2287 * completion and here. During detach, it would have set
2288 * JOBCTL_STOP_PENDING on us and we'll re-enter
2289 * TASK_STOPPED in do_signal_stop() on return, so notifying
2290 * the real parent of the group stop completion is enough.
2293 do_notify_parent_cldstop(current, false, why);
2295 /* tasklist protects us from ptrace_freeze_traced() */
2296 __set_current_state(TASK_RUNNING);
2298 current->exit_code = 0;
2299 read_unlock(&tasklist_lock);
2303 * We are back. Now reacquire the siglock before touching
2304 * last_siginfo, so that we are sure to have synchronized with
2305 * any signal-sending on another CPU that wants to examine it.
2307 spin_lock_irq(¤t->sighand->siglock);
2308 current->last_siginfo = NULL;
2310 /* LISTENING can be set only during STOP traps, clear it */
2311 current->jobctl &= ~JOBCTL_LISTENING;
2314 * Queued signals ignored us while we were stopped for tracing.
2315 * So check for any that we should take before resuming user mode.
2316 * This sets TIF_SIGPENDING, but never clears it.
2318 recalc_sigpending_tsk(current);
2321 static void ptrace_do_notify(int signr, int exit_code, int why)
2323 kernel_siginfo_t info;
2325 clear_siginfo(&info);
2326 info.si_signo = signr;
2327 info.si_code = exit_code;
2328 info.si_pid = task_pid_vnr(current);
2329 info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2331 /* Let the debugger run. */
2332 ptrace_stop(exit_code, why, 1, &info);
2335 void ptrace_notify(int exit_code)
2337 BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2338 if (unlikely(current->task_works))
2341 spin_lock_irq(¤t->sighand->siglock);
2342 ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
2343 spin_unlock_irq(¤t->sighand->siglock);
2347 * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2348 * @signr: signr causing group stop if initiating
2350 * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2351 * and participate in it. If already set, participate in the existing
2352 * group stop. If participated in a group stop (and thus slept), %true is
2353 * returned with siglock released.
2355 * If ptraced, this function doesn't handle stop itself. Instead,
2356 * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2357 * untouched. The caller must ensure that INTERRUPT trap handling takes
2358 * places afterwards.
2361 * Must be called with @current->sighand->siglock held, which is released
2365 * %false if group stop is already cancelled or ptrace trap is scheduled.
2366 * %true if participated in group stop.
2368 static bool do_signal_stop(int signr)
2369 __releases(¤t->sighand->siglock)
2371 struct signal_struct *sig = current->signal;
2373 if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2374 unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2375 struct task_struct *t;
2377 /* signr will be recorded in task->jobctl for retries */
2378 WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2380 if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2381 unlikely(signal_group_exit(sig)))
2384 * There is no group stop already in progress. We must
2387 * While ptraced, a task may be resumed while group stop is
2388 * still in effect and then receive a stop signal and
2389 * initiate another group stop. This deviates from the
2390 * usual behavior as two consecutive stop signals can't
2391 * cause two group stops when !ptraced. That is why we
2392 * also check !task_is_stopped(t) below.
2394 * The condition can be distinguished by testing whether
2395 * SIGNAL_STOP_STOPPED is already set. Don't generate
2396 * group_exit_code in such case.
2398 * This is not necessary for SIGNAL_STOP_CONTINUED because
2399 * an intervening stop signal is required to cause two
2400 * continued events regardless of ptrace.
2402 if (!(sig->flags & SIGNAL_STOP_STOPPED))
2403 sig->group_exit_code = signr;
2405 sig->group_stop_count = 0;
2407 if (task_set_jobctl_pending(current, signr | gstop))
2408 sig->group_stop_count++;
2411 while_each_thread(current, t) {
2413 * Setting state to TASK_STOPPED for a group
2414 * stop is always done with the siglock held,
2415 * so this check has no races.
2417 if (!task_is_stopped(t) &&
2418 task_set_jobctl_pending(t, signr | gstop)) {
2419 sig->group_stop_count++;
2420 if (likely(!(t->ptrace & PT_SEIZED)))
2421 signal_wake_up(t, 0);
2423 ptrace_trap_notify(t);
2428 if (likely(!current->ptrace)) {
2432 * If there are no other threads in the group, or if there
2433 * is a group stop in progress and we are the last to stop,
2434 * report to the parent.
2436 if (task_participate_group_stop(current))
2437 notify = CLD_STOPPED;
2439 set_special_state(TASK_STOPPED);
2440 spin_unlock_irq(¤t->sighand->siglock);
2443 * Notify the parent of the group stop completion. Because
2444 * we're not holding either the siglock or tasklist_lock
2445 * here, ptracer may attach inbetween; however, this is for
2446 * group stop and should always be delivered to the real
2447 * parent of the group leader. The new ptracer will get
2448 * its notification when this task transitions into
2452 read_lock(&tasklist_lock);
2453 do_notify_parent_cldstop(current, false, notify);
2454 read_unlock(&tasklist_lock);
2457 /* Now we don't run again until woken by SIGCONT or SIGKILL */
2458 cgroup_enter_frozen();
2459 freezable_schedule();
2463 * While ptraced, group stop is handled by STOP trap.
2464 * Schedule it and let the caller deal with it.
2466 task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2472 * do_jobctl_trap - take care of ptrace jobctl traps
2474 * When PT_SEIZED, it's used for both group stop and explicit
2475 * SEIZE/INTERRUPT traps. Both generate PTRACE_EVENT_STOP trap with
2476 * accompanying siginfo. If stopped, lower eight bits of exit_code contain
2477 * the stop signal; otherwise, %SIGTRAP.
2479 * When !PT_SEIZED, it's used only for group stop trap with stop signal
2480 * number as exit_code and no siginfo.
2483 * Must be called with @current->sighand->siglock held, which may be
2484 * released and re-acquired before returning with intervening sleep.
2486 static void do_jobctl_trap(void)
2488 struct signal_struct *signal = current->signal;
2489 int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2491 if (current->ptrace & PT_SEIZED) {
2492 if (!signal->group_stop_count &&
2493 !(signal->flags & SIGNAL_STOP_STOPPED))
2495 WARN_ON_ONCE(!signr);
2496 ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2499 WARN_ON_ONCE(!signr);
2500 ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2501 current->exit_code = 0;
2506 * do_freezer_trap - handle the freezer jobctl trap
2508 * Puts the task into frozen state, if only the task is not about to quit.
2509 * In this case it drops JOBCTL_TRAP_FREEZE.
2512 * Must be called with @current->sighand->siglock held,
2513 * which is always released before returning.
2515 static void do_freezer_trap(void)
2516 __releases(¤t->sighand->siglock)
2519 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2520 * let's make another loop to give it a chance to be handled.
2521 * In any case, we'll return back.
2523 if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2524 JOBCTL_TRAP_FREEZE) {
2525 spin_unlock_irq(¤t->sighand->siglock);
2530 * Now we're sure that there is no pending fatal signal and no
2531 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2532 * immediately (if there is a non-fatal signal pending), and
2533 * put the task into sleep.
2535 __set_current_state(TASK_INTERRUPTIBLE);
2536 clear_thread_flag(TIF_SIGPENDING);
2537 spin_unlock_irq(¤t->sighand->siglock);
2538 cgroup_enter_frozen();
2539 freezable_schedule();
2542 static int ptrace_signal(int signr, kernel_siginfo_t *info)
2545 * We do not check sig_kernel_stop(signr) but set this marker
2546 * unconditionally because we do not know whether debugger will
2547 * change signr. This flag has no meaning unless we are going
2548 * to stop after return from ptrace_stop(). In this case it will
2549 * be checked in do_signal_stop(), we should only stop if it was
2550 * not cleared by SIGCONT while we were sleeping. See also the
2551 * comment in dequeue_signal().
2553 current->jobctl |= JOBCTL_STOP_DEQUEUED;
2554 ptrace_stop(signr, CLD_TRAPPED, 0, info);
2556 /* We're back. Did the debugger cancel the sig? */
2557 signr = current->exit_code;
2561 current->exit_code = 0;
2564 * Update the siginfo structure if the signal has
2565 * changed. If the debugger wanted something
2566 * specific in the siginfo structure then it should
2567 * have updated *info via PTRACE_SETSIGINFO.
2569 if (signr != info->si_signo) {
2570 clear_siginfo(info);
2571 info->si_signo = signr;
2573 info->si_code = SI_USER;
2575 info->si_pid = task_pid_vnr(current->parent);
2576 info->si_uid = from_kuid_munged(current_user_ns(),
2577 task_uid(current->parent));
2581 /* If the (new) signal is now blocked, requeue it. */
2582 if (sigismember(¤t->blocked, signr)) {
2583 send_signal(signr, info, current, PIDTYPE_PID);
2590 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2592 switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2594 case SIL_FAULT_TRAPNO:
2595 case SIL_FAULT_MCEERR:
2596 case SIL_FAULT_BNDERR:
2597 case SIL_FAULT_PKUERR:
2598 case SIL_FAULT_PERF_EVENT:
2599 ksig->info.si_addr = arch_untagged_si_addr(
2600 ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2612 bool get_signal(struct ksignal *ksig)
2614 struct sighand_struct *sighand = current->sighand;
2615 struct signal_struct *signal = current->signal;
2618 if (unlikely(current->task_works))
2622 * For non-generic architectures, check for TIF_NOTIFY_SIGNAL so
2623 * that the arch handlers don't all have to do it. If we get here
2624 * without TIF_SIGPENDING, just exit after running signal work.
2626 if (!IS_ENABLED(CONFIG_GENERIC_ENTRY)) {
2627 if (test_thread_flag(TIF_NOTIFY_SIGNAL))
2628 tracehook_notify_signal();
2629 if (!task_sigpending(current))
2633 if (unlikely(uprobe_deny_signal()))
2637 * Do this once, we can't return to user-mode if freezing() == T.
2638 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2639 * thus do not need another check after return.
2644 spin_lock_irq(&sighand->siglock);
2647 * Every stopped thread goes here after wakeup. Check to see if
2648 * we should notify the parent, prepare_signal(SIGCONT) encodes
2649 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2651 if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2654 if (signal->flags & SIGNAL_CLD_CONTINUED)
2655 why = CLD_CONTINUED;
2659 signal->flags &= ~SIGNAL_CLD_MASK;
2661 spin_unlock_irq(&sighand->siglock);
2664 * Notify the parent that we're continuing. This event is
2665 * always per-process and doesn't make whole lot of sense
2666 * for ptracers, who shouldn't consume the state via
2667 * wait(2) either, but, for backward compatibility, notify
2668 * the ptracer of the group leader too unless it's gonna be
2671 read_lock(&tasklist_lock);
2672 do_notify_parent_cldstop(current, false, why);
2674 if (ptrace_reparented(current->group_leader))
2675 do_notify_parent_cldstop(current->group_leader,
2677 read_unlock(&tasklist_lock);
2682 /* Has this task already been marked for death? */
2683 if (signal_group_exit(signal)) {
2684 ksig->info.si_signo = signr = SIGKILL;
2685 sigdelset(¤t->pending.signal, SIGKILL);
2686 trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2687 &sighand->action[SIGKILL - 1]);
2688 recalc_sigpending();
2693 struct k_sigaction *ka;
2695 if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2699 if (unlikely(current->jobctl &
2700 (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2701 if (current->jobctl & JOBCTL_TRAP_MASK) {
2703 spin_unlock_irq(&sighand->siglock);
2704 } else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2711 * If the task is leaving the frozen state, let's update
2712 * cgroup counters and reset the frozen bit.
2714 if (unlikely(cgroup_task_frozen(current))) {
2715 spin_unlock_irq(&sighand->siglock);
2716 cgroup_leave_frozen(false);
2721 * Signals generated by the execution of an instruction
2722 * need to be delivered before any other pending signals
2723 * so that the instruction pointer in the signal stack
2724 * frame points to the faulting instruction.
2726 signr = dequeue_synchronous_signal(&ksig->info);
2728 signr = dequeue_signal(current, ¤t->blocked, &ksig->info);
2731 break; /* will return 0 */
2733 if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2734 !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2735 signr = ptrace_signal(signr, &ksig->info);
2740 ka = &sighand->action[signr-1];
2742 /* Trace actually delivered signals. */
2743 trace_signal_deliver(signr, &ksig->info, ka);
2745 if (ka->sa.sa_handler == SIG_IGN) /* Do nothing. */
2747 if (ka->sa.sa_handler != SIG_DFL) {
2748 /* Run the handler. */
2751 if (ka->sa.sa_flags & SA_ONESHOT)
2752 ka->sa.sa_handler = SIG_DFL;
2754 break; /* will return non-zero "signr" value */
2758 * Now we are doing the default action for this signal.
2760 if (sig_kernel_ignore(signr)) /* Default is nothing. */
2764 * Global init gets no signals it doesn't want.
2765 * Container-init gets no signals it doesn't want from same
2768 * Note that if global/container-init sees a sig_kernel_only()
2769 * signal here, the signal must have been generated internally
2770 * or must have come from an ancestor namespace. In either
2771 * case, the signal cannot be dropped.
2773 if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2774 !sig_kernel_only(signr))
2777 if (sig_kernel_stop(signr)) {
2779 * The default action is to stop all threads in
2780 * the thread group. The job control signals
2781 * do nothing in an orphaned pgrp, but SIGSTOP
2782 * always works. Note that siglock needs to be
2783 * dropped during the call to is_orphaned_pgrp()
2784 * because of lock ordering with tasklist_lock.
2785 * This allows an intervening SIGCONT to be posted.
2786 * We need to check for that and bail out if necessary.
2788 if (signr != SIGSTOP) {
2789 spin_unlock_irq(&sighand->siglock);
2791 /* signals can be posted during this window */
2793 if (is_current_pgrp_orphaned())
2796 spin_lock_irq(&sighand->siglock);
2799 if (likely(do_signal_stop(ksig->info.si_signo))) {
2800 /* It released the siglock. */
2805 * We didn't actually stop, due to a race
2806 * with SIGCONT or something like that.
2812 spin_unlock_irq(&sighand->siglock);
2813 if (unlikely(cgroup_task_frozen(current)))
2814 cgroup_leave_frozen(true);
2817 * Anything else is fatal, maybe with a core dump.
2819 current->flags |= PF_SIGNALED;
2821 if (sig_kernel_coredump(signr)) {
2822 if (print_fatal_signals)
2823 print_fatal_signal(ksig->info.si_signo);
2824 proc_coredump_connector(current);
2826 * If it was able to dump core, this kills all
2827 * other threads in the group and synchronizes with
2828 * their demise. If we lost the race with another
2829 * thread getting here, it set group_exit_code
2830 * first and our do_group_exit call below will use
2831 * that value and ignore the one we pass it.
2833 do_coredump(&ksig->info);
2837 * PF_IO_WORKER threads will catch and exit on fatal signals
2838 * themselves. They have cleanup that must be performed, so
2839 * we cannot call do_exit() on their behalf.
2841 if (current->flags & PF_IO_WORKER)
2845 * Death signals, no core dump.
2847 do_group_exit(ksig->info.si_signo);
2850 spin_unlock_irq(&sighand->siglock);
2854 if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2855 hide_si_addr_tag_bits(ksig);
2857 return ksig->sig > 0;
2861 * signal_delivered -
2862 * @ksig: kernel signal struct
2863 * @stepping: nonzero if debugger single-step or block-step in use
2865 * This function should be called when a signal has successfully been
2866 * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2867 * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2868 * is set in @ksig->ka.sa.sa_flags. Tracing is notified.
2870 static void signal_delivered(struct ksignal *ksig, int stepping)
2874 /* A signal was successfully delivered, and the
2875 saved sigmask was stored on the signal frame,
2876 and will be restored by sigreturn. So we can
2877 simply clear the restore sigmask flag. */
2878 clear_restore_sigmask();
2880 sigorsets(&blocked, ¤t->blocked, &ksig->ka.sa.sa_mask);
2881 if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2882 sigaddset(&blocked, ksig->sig);
2883 set_current_blocked(&blocked);
2884 if (current->sas_ss_flags & SS_AUTODISARM)
2885 sas_ss_reset(current);
2886 tracehook_signal_handler(stepping);
2889 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2892 force_sigsegv(ksig->sig);
2894 signal_delivered(ksig, stepping);
2898 * It could be that complete_signal() picked us to notify about the
2899 * group-wide signal. Other threads should be notified now to take
2900 * the shared signals in @which since we will not.
2902 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2905 struct task_struct *t;
2907 sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2908 if (sigisemptyset(&retarget))
2912 while_each_thread(tsk, t) {
2913 if (t->flags & PF_EXITING)
2916 if (!has_pending_signals(&retarget, &t->blocked))
2918 /* Remove the signals this thread can handle. */
2919 sigandsets(&retarget, &retarget, &t->blocked);
2921 if (!task_sigpending(t))
2922 signal_wake_up(t, 0);
2924 if (sigisemptyset(&retarget))
2929 void exit_signals(struct task_struct *tsk)
2935 * @tsk is about to have PF_EXITING set - lock out users which
2936 * expect stable threadgroup.
2938 cgroup_threadgroup_change_begin(tsk);
2940 if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2941 tsk->flags |= PF_EXITING;
2942 cgroup_threadgroup_change_end(tsk);
2946 spin_lock_irq(&tsk->sighand->siglock);
2948 * From now this task is not visible for group-wide signals,
2949 * see wants_signal(), do_signal_stop().
2951 tsk->flags |= PF_EXITING;
2953 cgroup_threadgroup_change_end(tsk);
2955 if (!task_sigpending(tsk))
2958 unblocked = tsk->blocked;
2959 signotset(&unblocked);
2960 retarget_shared_pending(tsk, &unblocked);
2962 if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2963 task_participate_group_stop(tsk))
2964 group_stop = CLD_STOPPED;
2966 spin_unlock_irq(&tsk->sighand->siglock);
2969 * If group stop has completed, deliver the notification. This
2970 * should always go to the real parent of the group leader.
2972 if (unlikely(group_stop)) {
2973 read_lock(&tasklist_lock);
2974 do_notify_parent_cldstop(tsk, false, group_stop);
2975 read_unlock(&tasklist_lock);
2980 * System call entry points.
2984 * sys_restart_syscall - restart a system call
2986 SYSCALL_DEFINE0(restart_syscall)
2988 struct restart_block *restart = ¤t->restart_block;
2989 return restart->fn(restart);
2992 long do_no_restart_syscall(struct restart_block *param)
2997 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2999 if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3000 sigset_t newblocked;
3001 /* A set of now blocked but previously unblocked signals. */
3002 sigandnsets(&newblocked, newset, ¤t->blocked);
3003 retarget_shared_pending(tsk, &newblocked);
3005 tsk->blocked = *newset;
3006 recalc_sigpending();
3010 * set_current_blocked - change current->blocked mask
3013 * It is wrong to change ->blocked directly, this helper should be used
3014 * to ensure the process can't miss a shared signal we are going to block.
3016 void set_current_blocked(sigset_t *newset)
3018 sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3019 __set_current_blocked(newset);
3022 void __set_current_blocked(const sigset_t *newset)
3024 struct task_struct *tsk = current;
3027 * In case the signal mask hasn't changed, there is nothing we need
3028 * to do. The current->blocked shouldn't be modified by other task.
3030 if (sigequalsets(&tsk->blocked, newset))
3033 spin_lock_irq(&tsk->sighand->siglock);
3034 __set_task_blocked(tsk, newset);
3035 spin_unlock_irq(&tsk->sighand->siglock);
3039 * This is also useful for kernel threads that want to temporarily
3040 * (or permanently) block certain signals.
3042 * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3043 * interface happily blocks "unblockable" signals like SIGKILL
3046 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3048 struct task_struct *tsk = current;
3051 /* Lockless, only current can change ->blocked, never from irq */
3053 *oldset = tsk->blocked;
3057 sigorsets(&newset, &tsk->blocked, set);
3060 sigandnsets(&newset, &tsk->blocked, set);
3069 __set_current_blocked(&newset);
3072 EXPORT_SYMBOL(sigprocmask);
3075 * The api helps set app-provided sigmasks.
3077 * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3078 * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3080 * Note that it does set_restore_sigmask() in advance, so it must be always
3081 * paired with restore_saved_sigmask_unless() before return from syscall.
3083 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3089 if (sigsetsize != sizeof(sigset_t))
3091 if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3094 set_restore_sigmask();
3095 current->saved_sigmask = current->blocked;
3096 set_current_blocked(&kmask);
3101 #ifdef CONFIG_COMPAT
3102 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3109 if (sigsetsize != sizeof(compat_sigset_t))
3111 if (get_compat_sigset(&kmask, umask))
3114 set_restore_sigmask();
3115 current->saved_sigmask = current->blocked;
3116 set_current_blocked(&kmask);
3123 * sys_rt_sigprocmask - change the list of currently blocked signals
3124 * @how: whether to add, remove, or set signals
3125 * @nset: stores pending signals
3126 * @oset: previous value of signal mask if non-null
3127 * @sigsetsize: size of sigset_t type
3129 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3130 sigset_t __user *, oset, size_t, sigsetsize)
3132 sigset_t old_set, new_set;
3135 /* XXX: Don't preclude handling different sized sigset_t's. */
3136 if (sigsetsize != sizeof(sigset_t))
3139 old_set = current->blocked;
3142 if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3144 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3146 error = sigprocmask(how, &new_set, NULL);
3152 if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3159 #ifdef CONFIG_COMPAT
3160 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3161 compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3163 sigset_t old_set = current->blocked;
3165 /* XXX: Don't preclude handling different sized sigset_t's. */
3166 if (sigsetsize != sizeof(sigset_t))
3172 if (get_compat_sigset(&new_set, nset))
3174 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3176 error = sigprocmask(how, &new_set, NULL);
3180 return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3184 static void do_sigpending(sigset_t *set)
3186 spin_lock_irq(¤t->sighand->siglock);
3187 sigorsets(set, ¤t->pending.signal,
3188 ¤t->signal->shared_pending.signal);
3189 spin_unlock_irq(¤t->sighand->siglock);
3191 /* Outside the lock because only this thread touches it. */
3192 sigandsets(set, ¤t->blocked, set);
3196 * sys_rt_sigpending - examine a pending signal that has been raised
3198 * @uset: stores pending signals
3199 * @sigsetsize: size of sigset_t type or larger
3201 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3205 if (sigsetsize > sizeof(*uset))
3208 do_sigpending(&set);
3210 if (copy_to_user(uset, &set, sigsetsize))
3216 #ifdef CONFIG_COMPAT
3217 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3218 compat_size_t, sigsetsize)
3222 if (sigsetsize > sizeof(*uset))
3225 do_sigpending(&set);
3227 return put_compat_sigset(uset, &set, sigsetsize);
3231 static const struct {
3232 unsigned char limit, layout;
3234 [SIGILL] = { NSIGILL, SIL_FAULT },
3235 [SIGFPE] = { NSIGFPE, SIL_FAULT },
3236 [SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3237 [SIGBUS] = { NSIGBUS, SIL_FAULT },
3238 [SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3240 [SIGEMT] = { NSIGEMT, SIL_FAULT },
3242 [SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3243 [SIGPOLL] = { NSIGPOLL, SIL_POLL },
3244 [SIGSYS] = { NSIGSYS, SIL_SYS },
3247 static bool known_siginfo_layout(unsigned sig, int si_code)
3249 if (si_code == SI_KERNEL)
3251 else if ((si_code > SI_USER)) {
3252 if (sig_specific_sicodes(sig)) {
3253 if (si_code <= sig_sicodes[sig].limit)
3256 else if (si_code <= NSIGPOLL)
3259 else if (si_code >= SI_DETHREAD)
3261 else if (si_code == SI_ASYNCNL)
3266 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3268 enum siginfo_layout layout = SIL_KILL;
3269 if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3270 if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3271 (si_code <= sig_sicodes[sig].limit)) {
3272 layout = sig_sicodes[sig].layout;
3273 /* Handle the exceptions */
3274 if ((sig == SIGBUS) &&
3275 (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3276 layout = SIL_FAULT_MCEERR;
3277 else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3278 layout = SIL_FAULT_BNDERR;
3280 else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3281 layout = SIL_FAULT_PKUERR;
3283 else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3284 layout = SIL_FAULT_PERF_EVENT;
3285 else if (IS_ENABLED(CONFIG_SPARC) &&
3286 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3287 layout = SIL_FAULT_TRAPNO;
3288 else if (IS_ENABLED(CONFIG_ALPHA) &&
3290 ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3291 layout = SIL_FAULT_TRAPNO;
3293 else if (si_code <= NSIGPOLL)
3296 if (si_code == SI_TIMER)
3298 else if (si_code == SI_SIGIO)
3300 else if (si_code < 0)
3306 static inline char __user *si_expansion(const siginfo_t __user *info)
3308 return ((char __user *)info) + sizeof(struct kernel_siginfo);
3311 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3313 char __user *expansion = si_expansion(to);
3314 if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3316 if (clear_user(expansion, SI_EXPANSION_SIZE))
3321 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3322 const siginfo_t __user *from)
3324 if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3325 char __user *expansion = si_expansion(from);
3326 char buf[SI_EXPANSION_SIZE];
3329 * An unknown si_code might need more than
3330 * sizeof(struct kernel_siginfo) bytes. Verify all of the
3331 * extra bytes are 0. This guarantees copy_siginfo_to_user
3332 * will return this data to userspace exactly.
3334 if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3336 for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3344 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3345 const siginfo_t __user *from)
3347 if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3349 to->si_signo = signo;
3350 return post_copy_siginfo_from_user(to, from);
3353 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3355 if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3357 return post_copy_siginfo_from_user(to, from);
3360 #ifdef CONFIG_COMPAT
3362 * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3363 * @to: compat siginfo destination
3364 * @from: kernel siginfo source
3366 * Note: This function does not work properly for the SIGCHLD on x32, but
3367 * fortunately it doesn't have to. The only valid callers for this function are
3368 * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3369 * The latter does not care because SIGCHLD will never cause a coredump.
3371 void copy_siginfo_to_external32(struct compat_siginfo *to,
3372 const struct kernel_siginfo *from)
3374 memset(to, 0, sizeof(*to));
3376 to->si_signo = from->si_signo;
3377 to->si_errno = from->si_errno;
3378 to->si_code = from->si_code;
3379 switch(siginfo_layout(from->si_signo, from->si_code)) {
3381 to->si_pid = from->si_pid;
3382 to->si_uid = from->si_uid;
3385 to->si_tid = from->si_tid;
3386 to->si_overrun = from->si_overrun;
3387 to->si_int = from->si_int;
3390 to->si_band = from->si_band;
3391 to->si_fd = from->si_fd;
3394 to->si_addr = ptr_to_compat(from->si_addr);
3396 case SIL_FAULT_TRAPNO:
3397 to->si_addr = ptr_to_compat(from->si_addr);
3398 to->si_trapno = from->si_trapno;
3400 case SIL_FAULT_MCEERR:
3401 to->si_addr = ptr_to_compat(from->si_addr);
3402 to->si_addr_lsb = from->si_addr_lsb;
3404 case SIL_FAULT_BNDERR:
3405 to->si_addr = ptr_to_compat(from->si_addr);
3406 to->si_lower = ptr_to_compat(from->si_lower);
3407 to->si_upper = ptr_to_compat(from->si_upper);
3409 case SIL_FAULT_PKUERR:
3410 to->si_addr = ptr_to_compat(from->si_addr);
3411 to->si_pkey = from->si_pkey;
3413 case SIL_FAULT_PERF_EVENT:
3414 to->si_addr = ptr_to_compat(from->si_addr);
3415 to->si_perf_data = from->si_perf_data;
3416 to->si_perf_type = from->si_perf_type;
3419 to->si_pid = from->si_pid;
3420 to->si_uid = from->si_uid;
3421 to->si_status = from->si_status;
3422 to->si_utime = from->si_utime;
3423 to->si_stime = from->si_stime;
3426 to->si_pid = from->si_pid;
3427 to->si_uid = from->si_uid;
3428 to->si_int = from->si_int;
3431 to->si_call_addr = ptr_to_compat(from->si_call_addr);
3432 to->si_syscall = from->si_syscall;
3433 to->si_arch = from->si_arch;
3438 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3439 const struct kernel_siginfo *from)
3441 struct compat_siginfo new;
3443 copy_siginfo_to_external32(&new, from);
3444 if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3449 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3450 const struct compat_siginfo *from)
3453 to->si_signo = from->si_signo;
3454 to->si_errno = from->si_errno;
3455 to->si_code = from->si_code;
3456 switch(siginfo_layout(from->si_signo, from->si_code)) {
3458 to->si_pid = from->si_pid;
3459 to->si_uid = from->si_uid;
3462 to->si_tid = from->si_tid;
3463 to->si_overrun = from->si_overrun;
3464 to->si_int = from->si_int;
3467 to->si_band = from->si_band;
3468 to->si_fd = from->si_fd;
3471 to->si_addr = compat_ptr(from->si_addr);
3473 case SIL_FAULT_TRAPNO:
3474 to->si_addr = compat_ptr(from->si_addr);
3475 to->si_trapno = from->si_trapno;
3477 case SIL_FAULT_MCEERR:
3478 to->si_addr = compat_ptr(from->si_addr);
3479 to->si_addr_lsb = from->si_addr_lsb;
3481 case SIL_FAULT_BNDERR:
3482 to->si_addr = compat_ptr(from->si_addr);
3483 to->si_lower = compat_ptr(from->si_lower);
3484 to->si_upper = compat_ptr(from->si_upper);
3486 case SIL_FAULT_PKUERR:
3487 to->si_addr = compat_ptr(from->si_addr);
3488 to->si_pkey = from->si_pkey;
3490 case SIL_FAULT_PERF_EVENT:
3491 to->si_addr = compat_ptr(from->si_addr);
3492 to->si_perf_data = from->si_perf_data;
3493 to->si_perf_type = from->si_perf_type;
3496 to->si_pid = from->si_pid;
3497 to->si_uid = from->si_uid;
3498 to->si_status = from->si_status;
3499 #ifdef CONFIG_X86_X32_ABI
3500 if (in_x32_syscall()) {
3501 to->si_utime = from->_sifields._sigchld_x32._utime;
3502 to->si_stime = from->_sifields._sigchld_x32._stime;
3506 to->si_utime = from->si_utime;
3507 to->si_stime = from->si_stime;
3511 to->si_pid = from->si_pid;
3512 to->si_uid = from->si_uid;
3513 to->si_int = from->si_int;
3516 to->si_call_addr = compat_ptr(from->si_call_addr);
3517 to->si_syscall = from->si_syscall;
3518 to->si_arch = from->si_arch;
3524 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3525 const struct compat_siginfo __user *ufrom)
3527 struct compat_siginfo from;
3529 if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3532 from.si_signo = signo;
3533 return post_copy_siginfo_from_user32(to, &from);
3536 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3537 const struct compat_siginfo __user *ufrom)
3539 struct compat_siginfo from;
3541 if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3544 return post_copy_siginfo_from_user32(to, &from);
3546 #endif /* CONFIG_COMPAT */
3549 * do_sigtimedwait - wait for queued signals specified in @which
3550 * @which: queued signals to wait for
3551 * @info: if non-null, the signal's siginfo is returned here
3552 * @ts: upper bound on process time suspension
3554 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3555 const struct timespec64 *ts)
3557 ktime_t *to = NULL, timeout = KTIME_MAX;
3558 struct task_struct *tsk = current;
3559 sigset_t mask = *which;
3563 if (!timespec64_valid(ts))
3565 timeout = timespec64_to_ktime(*ts);
3570 * Invert the set of allowed signals to get those we want to block.
3572 sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3575 spin_lock_irq(&tsk->sighand->siglock);
3576 sig = dequeue_signal(tsk, &mask, info);
3577 if (!sig && timeout) {
3579 * None ready, temporarily unblock those we're interested
3580 * while we are sleeping in so that we'll be awakened when
3581 * they arrive. Unblocking is always fine, we can avoid
3582 * set_current_blocked().
3584 tsk->real_blocked = tsk->blocked;
3585 sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3586 recalc_sigpending();
3587 spin_unlock_irq(&tsk->sighand->siglock);
3589 __set_current_state(TASK_INTERRUPTIBLE);
3590 ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3592 spin_lock_irq(&tsk->sighand->siglock);
3593 __set_task_blocked(tsk, &tsk->real_blocked);
3594 sigemptyset(&tsk->real_blocked);
3595 sig = dequeue_signal(tsk, &mask, info);
3597 spin_unlock_irq(&tsk->sighand->siglock);
3601 return ret ? -EINTR : -EAGAIN;
3605 * sys_rt_sigtimedwait - synchronously wait for queued signals specified
3607 * @uthese: queued signals to wait for
3608 * @uinfo: if non-null, the signal's siginfo is returned here
3609 * @uts: upper bound on process time suspension
3610 * @sigsetsize: size of sigset_t type
3612 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3613 siginfo_t __user *, uinfo,
3614 const struct __kernel_timespec __user *, uts,
3618 struct timespec64 ts;
3619 kernel_siginfo_t info;
3622 /* XXX: Don't preclude handling different sized sigset_t's. */
3623 if (sigsetsize != sizeof(sigset_t))
3626 if (copy_from_user(&these, uthese, sizeof(these)))
3630 if (get_timespec64(&ts, uts))
3634 ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3636 if (ret > 0 && uinfo) {
3637 if (copy_siginfo_to_user(uinfo, &info))
3644 #ifdef CONFIG_COMPAT_32BIT_TIME
3645 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3646 siginfo_t __user *, uinfo,
3647 const struct old_timespec32 __user *, uts,
3651 struct timespec64 ts;
3652 kernel_siginfo_t info;
3655 if (sigsetsize != sizeof(sigset_t))
3658 if (copy_from_user(&these, uthese, sizeof(these)))
3662 if (get_old_timespec32(&ts, uts))
3666 ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3668 if (ret > 0 && uinfo) {
3669 if (copy_siginfo_to_user(uinfo, &info))
3677 #ifdef CONFIG_COMPAT
3678 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3679 struct compat_siginfo __user *, uinfo,
3680 struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3683 struct timespec64 t;
3684 kernel_siginfo_t info;
3687 if (sigsetsize != sizeof(sigset_t))
3690 if (get_compat_sigset(&s, uthese))
3694 if (get_timespec64(&t, uts))
3698 ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3700 if (ret > 0 && uinfo) {
3701 if (copy_siginfo_to_user32(uinfo, &info))
3708 #ifdef CONFIG_COMPAT_32BIT_TIME
3709 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3710 struct compat_siginfo __user *, uinfo,
3711 struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3714 struct timespec64 t;
3715 kernel_siginfo_t info;
3718 if (sigsetsize != sizeof(sigset_t))
3721 if (get_compat_sigset(&s, uthese))
3725 if (get_old_timespec32(&t, uts))
3729 ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3731 if (ret > 0 && uinfo) {
3732 if (copy_siginfo_to_user32(uinfo, &info))
3741 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3743 clear_siginfo(info);
3744 info->si_signo = sig;
3746 info->si_code = SI_USER;
3747 info->si_pid = task_tgid_vnr(current);
3748 info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3752 * sys_kill - send a signal to a process
3753 * @pid: the PID of the process
3754 * @sig: signal to be sent
3756 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3758 struct kernel_siginfo info;
3760 prepare_kill_siginfo(sig, &info);
3762 return kill_something_info(sig, &info, pid);
3766 * Verify that the signaler and signalee either are in the same pid namespace
3767 * or that the signaler's pid namespace is an ancestor of the signalee's pid
3770 static bool access_pidfd_pidns(struct pid *pid)
3772 struct pid_namespace *active = task_active_pid_ns(current);
3773 struct pid_namespace *p = ns_of_pid(pid);
3786 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3787 siginfo_t __user *info)
3789 #ifdef CONFIG_COMPAT
3791 * Avoid hooking up compat syscalls and instead handle necessary
3792 * conversions here. Note, this is a stop-gap measure and should not be
3793 * considered a generic solution.
3795 if (in_compat_syscall())
3796 return copy_siginfo_from_user32(
3797 kinfo, (struct compat_siginfo __user *)info);
3799 return copy_siginfo_from_user(kinfo, info);
3802 static struct pid *pidfd_to_pid(const struct file *file)
3806 pid = pidfd_pid(file);
3810 return tgid_pidfd_to_pid(file);
3814 * sys_pidfd_send_signal - Signal a process through a pidfd
3815 * @pidfd: file descriptor of the process
3816 * @sig: signal to send
3817 * @info: signal info
3818 * @flags: future flags
3820 * The syscall currently only signals via PIDTYPE_PID which covers
3821 * kill(<positive-pid>, <signal>. It does not signal threads or process
3823 * In order to extend the syscall to threads and process groups the @flags
3824 * argument should be used. In essence, the @flags argument will determine
3825 * what is signaled and not the file descriptor itself. Put in other words,
3826 * grouping is a property of the flags argument not a property of the file
3829 * Return: 0 on success, negative errno on failure
3831 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3832 siginfo_t __user *, info, unsigned int, flags)
3837 kernel_siginfo_t kinfo;
3839 /* Enforce flags be set to 0 until we add an extension. */
3847 /* Is this a pidfd? */
3848 pid = pidfd_to_pid(f.file);
3855 if (!access_pidfd_pidns(pid))
3859 ret = copy_siginfo_from_user_any(&kinfo, info);
3864 if (unlikely(sig != kinfo.si_signo))
3867 /* Only allow sending arbitrary signals to yourself. */
3869 if ((task_pid(current) != pid) &&
3870 (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3873 prepare_kill_siginfo(sig, &kinfo);
3876 ret = kill_pid_info(sig, &kinfo, pid);
3884 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3886 struct task_struct *p;
3890 p = find_task_by_vpid(pid);
3891 if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3892 error = check_kill_permission(sig, info, p);
3894 * The null signal is a permissions and process existence
3895 * probe. No signal is actually delivered.
3897 if (!error && sig) {
3898 error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3900 * If lock_task_sighand() failed we pretend the task
3901 * dies after receiving the signal. The window is tiny,
3902 * and the signal is private anyway.
3904 if (unlikely(error == -ESRCH))
3913 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3915 struct kernel_siginfo info;
3917 clear_siginfo(&info);
3918 info.si_signo = sig;
3920 info.si_code = SI_TKILL;
3921 info.si_pid = task_tgid_vnr(current);
3922 info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3924 return do_send_specific(tgid, pid, sig, &info);
3928 * sys_tgkill - send signal to one specific thread
3929 * @tgid: the thread group ID of the thread
3930 * @pid: the PID of the thread
3931 * @sig: signal to be sent
3933 * This syscall also checks the @tgid and returns -ESRCH even if the PID
3934 * exists but it's not belonging to the target process anymore. This
3935 * method solves the problem of threads exiting and PIDs getting reused.
3937 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3939 /* This is only valid for single tasks */
3940 if (pid <= 0 || tgid <= 0)
3943 return do_tkill(tgid, pid, sig);
3947 * sys_tkill - send signal to one specific task
3948 * @pid: the PID of the task
3949 * @sig: signal to be sent
3951 * Send a signal to only one task, even if it's a CLONE_THREAD task.
3953 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3955 /* This is only valid for single tasks */
3959 return do_tkill(0, pid, sig);
3962 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
3964 /* Not even root can pretend to send signals from the kernel.
3965 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3967 if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3968 (task_pid_vnr(current) != pid))
3971 /* POSIX.1b doesn't mention process groups. */
3972 return kill_proc_info(sig, info, pid);
3976 * sys_rt_sigqueueinfo - send signal information to a signal
3977 * @pid: the PID of the thread
3978 * @sig: signal to be sent
3979 * @uinfo: signal info to be sent
3981 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
3982 siginfo_t __user *, uinfo)
3984 kernel_siginfo_t info;
3985 int ret = __copy_siginfo_from_user(sig, &info, uinfo);
3988 return do_rt_sigqueueinfo(pid, sig, &info);
3991 #ifdef CONFIG_COMPAT
3992 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
3995 struct compat_siginfo __user *, uinfo)
3997 kernel_siginfo_t info;
3998 int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4001 return do_rt_sigqueueinfo(pid, sig, &info);
4005 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4007 /* This is only valid for single tasks */
4008 if (pid <= 0 || tgid <= 0)
4011 /* Not even root can pretend to send signals from the kernel.
4012 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4014 if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4015 (task_pid_vnr(current) != pid))
4018 return do_send_specific(tgid, pid, sig, info);
4021 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4022 siginfo_t __user *, uinfo)
4024 kernel_siginfo_t info;
4025 int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4028 return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4031 #ifdef CONFIG_COMPAT
4032 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4036 struct compat_siginfo __user *, uinfo)
4038 kernel_siginfo_t info;
4039 int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4042 return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4047 * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4049 void kernel_sigaction(int sig, __sighandler_t action)
4051 spin_lock_irq(¤t->sighand->siglock);
4052 current->sighand->action[sig - 1].sa.sa_handler = action;
4053 if (action == SIG_IGN) {
4057 sigaddset(&mask, sig);
4059 flush_sigqueue_mask(&mask, ¤t->signal->shared_pending);
4060 flush_sigqueue_mask(&mask, ¤t->pending);
4061 recalc_sigpending();
4063 spin_unlock_irq(¤t->sighand->siglock);
4065 EXPORT_SYMBOL(kernel_sigaction);
4067 void __weak sigaction_compat_abi(struct k_sigaction *act,
4068 struct k_sigaction *oact)
4072 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4074 struct task_struct *p = current, *t;
4075 struct k_sigaction *k;
4078 if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4081 k = &p->sighand->action[sig-1];
4083 spin_lock_irq(&p->sighand->siglock);
4084 if (k->sa.sa_flags & SA_IMMUTABLE) {
4085 spin_unlock_irq(&p->sighand->siglock);
4092 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4093 * e.g. by having an architecture use the bit in their uapi.
4095 BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4098 * Clear unknown flag bits in order to allow userspace to detect missing
4099 * support for flag bits and to allow the kernel to use non-uapi bits
4103 act->sa.sa_flags &= UAPI_SA_FLAGS;
4105 oact->sa.sa_flags &= UAPI_SA_FLAGS;
4107 sigaction_compat_abi(act, oact);
4110 sigdelsetmask(&act->sa.sa_mask,
4111 sigmask(SIGKILL) | sigmask(SIGSTOP));
4115 * "Setting a signal action to SIG_IGN for a signal that is
4116 * pending shall cause the pending signal to be discarded,
4117 * whether or not it is blocked."
4119 * "Setting a signal action to SIG_DFL for a signal that is
4120 * pending and whose default action is to ignore the signal
4121 * (for example, SIGCHLD), shall cause the pending signal to
4122 * be discarded, whether or not it is blocked"
4124 if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4126 sigaddset(&mask, sig);
4127 flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4128 for_each_thread(p, t)
4129 flush_sigqueue_mask(&mask, &t->pending);
4133 spin_unlock_irq(&p->sighand->siglock);
4138 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4141 struct task_struct *t = current;
4144 memset(oss, 0, sizeof(stack_t));
4145 oss->ss_sp = (void __user *) t->sas_ss_sp;
4146 oss->ss_size = t->sas_ss_size;
4147 oss->ss_flags = sas_ss_flags(sp) |
4148 (current->sas_ss_flags & SS_FLAG_BITS);
4152 void __user *ss_sp = ss->ss_sp;
4153 size_t ss_size = ss->ss_size;
4154 unsigned ss_flags = ss->ss_flags;
4157 if (unlikely(on_sig_stack(sp)))
4160 ss_mode = ss_flags & ~SS_FLAG_BITS;
4161 if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4165 if (ss_mode == SS_DISABLE) {
4169 if (unlikely(ss_size < min_ss_size))
4173 t->sas_ss_sp = (unsigned long) ss_sp;
4174 t->sas_ss_size = ss_size;
4175 t->sas_ss_flags = ss_flags;
4180 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4184 if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4186 err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4187 current_user_stack_pointer(),
4189 if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4194 int restore_altstack(const stack_t __user *uss)
4197 if (copy_from_user(&new, uss, sizeof(stack_t)))
4199 (void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4201 /* squash all but EFAULT for now */
4205 int __save_altstack(stack_t __user *uss, unsigned long sp)
4207 struct task_struct *t = current;
4208 int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4209 __put_user(t->sas_ss_flags, &uss->ss_flags) |
4210 __put_user(t->sas_ss_size, &uss->ss_size);
4214 #ifdef CONFIG_COMPAT
4215 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4216 compat_stack_t __user *uoss_ptr)
4222 compat_stack_t uss32;
4223 if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4225 uss.ss_sp = compat_ptr(uss32.ss_sp);
4226 uss.ss_flags = uss32.ss_flags;
4227 uss.ss_size = uss32.ss_size;
4229 ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4230 compat_user_stack_pointer(),
4231 COMPAT_MINSIGSTKSZ);
4232 if (ret >= 0 && uoss_ptr) {
4234 memset(&old, 0, sizeof(old));
4235 old.ss_sp = ptr_to_compat(uoss.ss_sp);
4236 old.ss_flags = uoss.ss_flags;
4237 old.ss_size = uoss.ss_size;
4238 if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4244 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4245 const compat_stack_t __user *, uss_ptr,
4246 compat_stack_t __user *, uoss_ptr)
4248 return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4251 int compat_restore_altstack(const compat_stack_t __user *uss)
4253 int err = do_compat_sigaltstack(uss, NULL);
4254 /* squash all but -EFAULT for now */
4255 return err == -EFAULT ? err : 0;
4258 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4261 struct task_struct *t = current;
4262 err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4264 __put_user(t->sas_ss_flags, &uss->ss_flags) |
4265 __put_user(t->sas_ss_size, &uss->ss_size);
4270 #ifdef __ARCH_WANT_SYS_SIGPENDING
4273 * sys_sigpending - examine pending signals
4274 * @uset: where mask of pending signal is returned
4276 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4280 if (sizeof(old_sigset_t) > sizeof(*uset))
4283 do_sigpending(&set);
4285 if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4291 #ifdef CONFIG_COMPAT
4292 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4296 do_sigpending(&set);
4298 return put_user(set.sig[0], set32);
4304 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4306 * sys_sigprocmask - examine and change blocked signals
4307 * @how: whether to add, remove, or set signals
4308 * @nset: signals to add or remove (if non-null)
4309 * @oset: previous value of signal mask if non-null
4311 * Some platforms have their own version with special arguments;
4312 * others support only sys_rt_sigprocmask.
4315 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4316 old_sigset_t __user *, oset)
4318 old_sigset_t old_set, new_set;
4319 sigset_t new_blocked;
4321 old_set = current->blocked.sig[0];
4324 if (copy_from_user(&new_set, nset, sizeof(*nset)))
4327 new_blocked = current->blocked;
4331 sigaddsetmask(&new_blocked, new_set);
4334 sigdelsetmask(&new_blocked, new_set);
4337 new_blocked.sig[0] = new_set;
4343 set_current_blocked(&new_blocked);
4347 if (copy_to_user(oset, &old_set, sizeof(*oset)))
4353 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4355 #ifndef CONFIG_ODD_RT_SIGACTION
4357 * sys_rt_sigaction - alter an action taken by a process
4358 * @sig: signal to be sent
4359 * @act: new sigaction
4360 * @oact: used to save the previous sigaction
4361 * @sigsetsize: size of sigset_t type
4363 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4364 const struct sigaction __user *, act,
4365 struct sigaction __user *, oact,
4368 struct k_sigaction new_sa, old_sa;
4371 /* XXX: Don't preclude handling different sized sigset_t's. */
4372 if (sigsetsize != sizeof(sigset_t))
4375 if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4378 ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4382 if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4387 #ifdef CONFIG_COMPAT
4388 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4389 const struct compat_sigaction __user *, act,
4390 struct compat_sigaction __user *, oact,
4391 compat_size_t, sigsetsize)
4393 struct k_sigaction new_ka, old_ka;
4394 #ifdef __ARCH_HAS_SA_RESTORER
4395 compat_uptr_t restorer;
4399 /* XXX: Don't preclude handling different sized sigset_t's. */
4400 if (sigsetsize != sizeof(compat_sigset_t))
4404 compat_uptr_t handler;
4405 ret = get_user(handler, &act->sa_handler);
4406 new_ka.sa.sa_handler = compat_ptr(handler);
4407 #ifdef __ARCH_HAS_SA_RESTORER
4408 ret |= get_user(restorer, &act->sa_restorer);
4409 new_ka.sa.sa_restorer = compat_ptr(restorer);
4411 ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4412 ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4417 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4419 ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4421 ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4422 sizeof(oact->sa_mask));
4423 ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4424 #ifdef __ARCH_HAS_SA_RESTORER
4425 ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4426 &oact->sa_restorer);
4432 #endif /* !CONFIG_ODD_RT_SIGACTION */
4434 #ifdef CONFIG_OLD_SIGACTION
4435 SYSCALL_DEFINE3(sigaction, int, sig,
4436 const struct old_sigaction __user *, act,
4437 struct old_sigaction __user *, oact)
4439 struct k_sigaction new_ka, old_ka;
4444 if (!access_ok(act, sizeof(*act)) ||
4445 __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4446 __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4447 __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4448 __get_user(mask, &act->sa_mask))
4450 #ifdef __ARCH_HAS_KA_RESTORER
4451 new_ka.ka_restorer = NULL;
4453 siginitset(&new_ka.sa.sa_mask, mask);
4456 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4459 if (!access_ok(oact, sizeof(*oact)) ||
4460 __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4461 __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4462 __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4463 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4470 #ifdef CONFIG_COMPAT_OLD_SIGACTION
4471 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4472 const struct compat_old_sigaction __user *, act,
4473 struct compat_old_sigaction __user *, oact)
4475 struct k_sigaction new_ka, old_ka;
4477 compat_old_sigset_t mask;
4478 compat_uptr_t handler, restorer;
4481 if (!access_ok(act, sizeof(*act)) ||
4482 __get_user(handler, &act->sa_handler) ||
4483 __get_user(restorer, &act->sa_restorer) ||
4484 __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4485 __get_user(mask, &act->sa_mask))
4488 #ifdef __ARCH_HAS_KA_RESTORER
4489 new_ka.ka_restorer = NULL;
4491 new_ka.sa.sa_handler = compat_ptr(handler);
4492 new_ka.sa.sa_restorer = compat_ptr(restorer);
4493 siginitset(&new_ka.sa.sa_mask, mask);
4496 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4499 if (!access_ok(oact, sizeof(*oact)) ||
4500 __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4501 &oact->sa_handler) ||
4502 __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4503 &oact->sa_restorer) ||
4504 __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4505 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4512 #ifdef CONFIG_SGETMASK_SYSCALL
4515 * For backwards compatibility. Functionality superseded by sigprocmask.
4517 SYSCALL_DEFINE0(sgetmask)
4520 return current->blocked.sig[0];
4523 SYSCALL_DEFINE1(ssetmask, int, newmask)
4525 int old = current->blocked.sig[0];
4528 siginitset(&newset, newmask);
4529 set_current_blocked(&newset);
4533 #endif /* CONFIG_SGETMASK_SYSCALL */
4535 #ifdef __ARCH_WANT_SYS_SIGNAL
4537 * For backwards compatibility. Functionality superseded by sigaction.
4539 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4541 struct k_sigaction new_sa, old_sa;
4544 new_sa.sa.sa_handler = handler;
4545 new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4546 sigemptyset(&new_sa.sa.sa_mask);
4548 ret = do_sigaction(sig, &new_sa, &old_sa);
4550 return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4552 #endif /* __ARCH_WANT_SYS_SIGNAL */
4554 #ifdef __ARCH_WANT_SYS_PAUSE
4556 SYSCALL_DEFINE0(pause)
4558 while (!signal_pending(current)) {
4559 __set_current_state(TASK_INTERRUPTIBLE);
4562 return -ERESTARTNOHAND;
4567 static int sigsuspend(sigset_t *set)
4569 current->saved_sigmask = current->blocked;
4570 set_current_blocked(set);
4572 while (!signal_pending(current)) {
4573 __set_current_state(TASK_INTERRUPTIBLE);
4576 set_restore_sigmask();
4577 return -ERESTARTNOHAND;
4581 * sys_rt_sigsuspend - replace the signal mask for a value with the
4582 * @unewset value until a signal is received
4583 * @unewset: new signal mask value
4584 * @sigsetsize: size of sigset_t type
4586 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4590 /* XXX: Don't preclude handling different sized sigset_t's. */
4591 if (sigsetsize != sizeof(sigset_t))
4594 if (copy_from_user(&newset, unewset, sizeof(newset)))
4596 return sigsuspend(&newset);
4599 #ifdef CONFIG_COMPAT
4600 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4604 /* XXX: Don't preclude handling different sized sigset_t's. */
4605 if (sigsetsize != sizeof(sigset_t))
4608 if (get_compat_sigset(&newset, unewset))
4610 return sigsuspend(&newset);
4614 #ifdef CONFIG_OLD_SIGSUSPEND
4615 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4618 siginitset(&blocked, mask);
4619 return sigsuspend(&blocked);
4622 #ifdef CONFIG_OLD_SIGSUSPEND3
4623 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4626 siginitset(&blocked, mask);
4627 return sigsuspend(&blocked);
4631 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4636 static inline void siginfo_buildtime_checks(void)
4638 BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4640 /* Verify the offsets in the two siginfos match */
4641 #define CHECK_OFFSET(field) \
4642 BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4645 CHECK_OFFSET(si_pid);
4646 CHECK_OFFSET(si_uid);
4649 CHECK_OFFSET(si_tid);
4650 CHECK_OFFSET(si_overrun);
4651 CHECK_OFFSET(si_value);
4654 CHECK_OFFSET(si_pid);
4655 CHECK_OFFSET(si_uid);
4656 CHECK_OFFSET(si_value);
4659 CHECK_OFFSET(si_pid);
4660 CHECK_OFFSET(si_uid);
4661 CHECK_OFFSET(si_status);
4662 CHECK_OFFSET(si_utime);
4663 CHECK_OFFSET(si_stime);
4666 CHECK_OFFSET(si_addr);
4667 CHECK_OFFSET(si_trapno);
4668 CHECK_OFFSET(si_addr_lsb);
4669 CHECK_OFFSET(si_lower);
4670 CHECK_OFFSET(si_upper);
4671 CHECK_OFFSET(si_pkey);
4672 CHECK_OFFSET(si_perf_data);
4673 CHECK_OFFSET(si_perf_type);
4676 CHECK_OFFSET(si_band);
4677 CHECK_OFFSET(si_fd);
4680 CHECK_OFFSET(si_call_addr);
4681 CHECK_OFFSET(si_syscall);
4682 CHECK_OFFSET(si_arch);
4686 BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4687 offsetof(struct siginfo, si_addr));
4688 if (sizeof(int) == sizeof(void __user *)) {
4689 BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4690 sizeof(void __user *));
4692 BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4693 sizeof_field(struct siginfo, si_uid)) !=
4694 sizeof(void __user *));
4695 BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4696 offsetof(struct siginfo, si_uid));
4698 #ifdef CONFIG_COMPAT
4699 BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4700 offsetof(struct compat_siginfo, si_addr));
4701 BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4702 sizeof(compat_uptr_t));
4703 BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4704 sizeof_field(struct siginfo, si_pid));
4708 void __init signals_init(void)
4710 siginfo_buildtime_checks();
4712 sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
4715 #ifdef CONFIG_KGDB_KDB
4716 #include <linux/kdb.h>
4718 * kdb_send_sig - Allows kdb to send signals without exposing
4719 * signal internals. This function checks if the required locks are
4720 * available before calling the main signal code, to avoid kdb
4723 void kdb_send_sig(struct task_struct *t, int sig)
4725 static struct task_struct *kdb_prev_t;
4727 if (!spin_trylock(&t->sighand->siglock)) {
4728 kdb_printf("Can't do kill command now.\n"
4729 "The sigmask lock is held somewhere else in "
4730 "kernel, try again later\n");
4733 new_t = kdb_prev_t != t;
4735 if (!task_is_running(t) && new_t) {
4736 spin_unlock(&t->sighand->siglock);
4737 kdb_printf("Process is not RUNNING, sending a signal from "
4738 "kdb risks deadlock\n"
4739 "on the run queue locks. "
4740 "The signal has _not_ been sent.\n"
4741 "Reissue the kill command if you want to risk "
4745 ret = send_signal(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4746 spin_unlock(&t->sighand->siglock);
4748 kdb_printf("Fail to deliver Signal %d to process %d.\n",
4751 kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4753 #endif /* CONFIG_KGDB_KDB */