xref: /openbmc/linux/kernel/signal.c (revision 6a87e0f0)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/kernel/signal.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  1997-11-02  Modified for POSIX.1b signals by Richard Henderson
8  *
9  *  2003-06-02  Jim Houston - Concurrent Computer Corp.
10  *		Changes to use preallocated sigqueue structures
11  *		to allow signals to be sent reliably.
12  */
13 
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>
24 #include <linux/fs.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/task_work.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>
48 #include <linux/sysctl.h>
49 
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/signal.h>
52 
53 #include <asm/param.h>
54 #include <linux/uaccess.h>
55 #include <asm/unistd.h>
56 #include <asm/siginfo.h>
57 #include <asm/cacheflush.h>
58 #include <asm/syscall.h>	/* for syscall_get_* */
59 
60 /*
61  * SLAB caches for signal bits.
62  */
63 
64 static struct kmem_cache *sigqueue_cachep;
65 
66 int print_fatal_signals __read_mostly;
67 
68 static void __user *sig_handler(struct task_struct *t, int sig)
69 {
70 	return t->sighand->action[sig - 1].sa.sa_handler;
71 }
72 
73 static inline bool sig_handler_ignored(void __user *handler, int sig)
74 {
75 	/* Is it explicitly or implicitly ignored? */
76 	return handler == SIG_IGN ||
77 	       (handler == SIG_DFL && sig_kernel_ignore(sig));
78 }
79 
80 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
81 {
82 	void __user *handler;
83 
84 	handler = sig_handler(t, sig);
85 
86 	/* SIGKILL and SIGSTOP may not be sent to the global init */
87 	if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
88 		return true;
89 
90 	if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
91 	    handler == SIG_DFL && !(force && sig_kernel_only(sig)))
92 		return true;
93 
94 	/* Only allow kernel generated signals to this kthread */
95 	if (unlikely((t->flags & PF_KTHREAD) &&
96 		     (handler == SIG_KTHREAD_KERNEL) && !force))
97 		return true;
98 
99 	return sig_handler_ignored(handler, sig);
100 }
101 
102 static bool sig_ignored(struct task_struct *t, int sig, bool force)
103 {
104 	/*
105 	 * Blocked signals are never ignored, since the
106 	 * signal handler may change by the time it is
107 	 * unblocked.
108 	 */
109 	if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
110 		return false;
111 
112 	/*
113 	 * Tracers may want to know about even ignored signal unless it
114 	 * is SIGKILL which can't be reported anyway but can be ignored
115 	 * by SIGNAL_UNKILLABLE task.
116 	 */
117 	if (t->ptrace && sig != SIGKILL)
118 		return false;
119 
120 	return sig_task_ignored(t, sig, force);
121 }
122 
123 /*
124  * Re-calculate pending state from the set of locally pending
125  * signals, globally pending signals, and blocked signals.
126  */
127 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
128 {
129 	unsigned long ready;
130 	long i;
131 
132 	switch (_NSIG_WORDS) {
133 	default:
134 		for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
135 			ready |= signal->sig[i] &~ blocked->sig[i];
136 		break;
137 
138 	case 4: ready  = signal->sig[3] &~ blocked->sig[3];
139 		ready |= signal->sig[2] &~ blocked->sig[2];
140 		ready |= signal->sig[1] &~ blocked->sig[1];
141 		ready |= signal->sig[0] &~ blocked->sig[0];
142 		break;
143 
144 	case 2: ready  = signal->sig[1] &~ blocked->sig[1];
145 		ready |= signal->sig[0] &~ blocked->sig[0];
146 		break;
147 
148 	case 1: ready  = signal->sig[0] &~ blocked->sig[0];
149 	}
150 	return ready !=	0;
151 }
152 
153 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
154 
155 static bool recalc_sigpending_tsk(struct task_struct *t)
156 {
157 	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
158 	    PENDING(&t->pending, &t->blocked) ||
159 	    PENDING(&t->signal->shared_pending, &t->blocked) ||
160 	    cgroup_task_frozen(t)) {
161 		set_tsk_thread_flag(t, TIF_SIGPENDING);
162 		return true;
163 	}
164 
165 	/*
166 	 * We must never clear the flag in another thread, or in current
167 	 * when it's possible the current syscall is returning -ERESTART*.
168 	 * So we don't clear it here, and only callers who know they should do.
169 	 */
170 	return false;
171 }
172 
173 /*
174  * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
175  * This is superfluous when called on current, the wakeup is a harmless no-op.
176  */
177 void recalc_sigpending_and_wake(struct task_struct *t)
178 {
179 	if (recalc_sigpending_tsk(t))
180 		signal_wake_up(t, 0);
181 }
182 
183 void recalc_sigpending(void)
184 {
185 	if (!recalc_sigpending_tsk(current) && !freezing(current))
186 		clear_thread_flag(TIF_SIGPENDING);
187 
188 }
189 EXPORT_SYMBOL(recalc_sigpending);
190 
191 void calculate_sigpending(void)
192 {
193 	/* Have any signals or users of TIF_SIGPENDING been delayed
194 	 * until after fork?
195 	 */
196 	spin_lock_irq(&current->sighand->siglock);
197 	set_tsk_thread_flag(current, TIF_SIGPENDING);
198 	recalc_sigpending();
199 	spin_unlock_irq(&current->sighand->siglock);
200 }
201 
202 /* Given the mask, find the first available signal that should be serviced. */
203 
204 #define SYNCHRONOUS_MASK \
205 	(sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
206 	 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
207 
208 int next_signal(struct sigpending *pending, sigset_t *mask)
209 {
210 	unsigned long i, *s, *m, x;
211 	int sig = 0;
212 
213 	s = pending->signal.sig;
214 	m = mask->sig;
215 
216 	/*
217 	 * Handle the first word specially: it contains the
218 	 * synchronous signals that need to be dequeued first.
219 	 */
220 	x = *s &~ *m;
221 	if (x) {
222 		if (x & SYNCHRONOUS_MASK)
223 			x &= SYNCHRONOUS_MASK;
224 		sig = ffz(~x) + 1;
225 		return sig;
226 	}
227 
228 	switch (_NSIG_WORDS) {
229 	default:
230 		for (i = 1; i < _NSIG_WORDS; ++i) {
231 			x = *++s &~ *++m;
232 			if (!x)
233 				continue;
234 			sig = ffz(~x) + i*_NSIG_BPW + 1;
235 			break;
236 		}
237 		break;
238 
239 	case 2:
240 		x = s[1] &~ m[1];
241 		if (!x)
242 			break;
243 		sig = ffz(~x) + _NSIG_BPW + 1;
244 		break;
245 
246 	case 1:
247 		/* Nothing to do */
248 		break;
249 	}
250 
251 	return sig;
252 }
253 
254 static inline void print_dropped_signal(int sig)
255 {
256 	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
257 
258 	if (!print_fatal_signals)
259 		return;
260 
261 	if (!__ratelimit(&ratelimit_state))
262 		return;
263 
264 	pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
265 				current->comm, current->pid, sig);
266 }
267 
268 /**
269  * task_set_jobctl_pending - set jobctl pending bits
270  * @task: target task
271  * @mask: pending bits to set
272  *
273  * Clear @mask from @task->jobctl.  @mask must be subset of
274  * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
275  * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
276  * cleared.  If @task is already being killed or exiting, this function
277  * becomes noop.
278  *
279  * CONTEXT:
280  * Must be called with @task->sighand->siglock held.
281  *
282  * RETURNS:
283  * %true if @mask is set, %false if made noop because @task was dying.
284  */
285 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
286 {
287 	BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
288 			JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
289 	BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
290 
291 	if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
292 		return false;
293 
294 	if (mask & JOBCTL_STOP_SIGMASK)
295 		task->jobctl &= ~JOBCTL_STOP_SIGMASK;
296 
297 	task->jobctl |= mask;
298 	return true;
299 }
300 
301 /**
302  * task_clear_jobctl_trapping - clear jobctl trapping bit
303  * @task: target task
304  *
305  * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
306  * Clear it and wake up the ptracer.  Note that we don't need any further
307  * locking.  @task->siglock guarantees that @task->parent points to the
308  * ptracer.
309  *
310  * CONTEXT:
311  * Must be called with @task->sighand->siglock held.
312  */
313 void task_clear_jobctl_trapping(struct task_struct *task)
314 {
315 	if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
316 		task->jobctl &= ~JOBCTL_TRAPPING;
317 		smp_mb();	/* advised by wake_up_bit() */
318 		wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
319 	}
320 }
321 
322 /**
323  * task_clear_jobctl_pending - clear jobctl pending bits
324  * @task: target task
325  * @mask: pending bits to clear
326  *
327  * Clear @mask from @task->jobctl.  @mask must be subset of
328  * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
329  * STOP bits are cleared together.
330  *
331  * If clearing of @mask leaves no stop or trap pending, this function calls
332  * task_clear_jobctl_trapping().
333  *
334  * CONTEXT:
335  * Must be called with @task->sighand->siglock held.
336  */
337 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
338 {
339 	BUG_ON(mask & ~JOBCTL_PENDING_MASK);
340 
341 	if (mask & JOBCTL_STOP_PENDING)
342 		mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
343 
344 	task->jobctl &= ~mask;
345 
346 	if (!(task->jobctl & JOBCTL_PENDING_MASK))
347 		task_clear_jobctl_trapping(task);
348 }
349 
350 /**
351  * task_participate_group_stop - participate in a group stop
352  * @task: task participating in a group stop
353  *
354  * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
355  * Group stop states are cleared and the group stop count is consumed if
356  * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
357  * stop, the appropriate `SIGNAL_*` flags are set.
358  *
359  * CONTEXT:
360  * Must be called with @task->sighand->siglock held.
361  *
362  * RETURNS:
363  * %true if group stop completion should be notified to the parent, %false
364  * otherwise.
365  */
366 static bool task_participate_group_stop(struct task_struct *task)
367 {
368 	struct signal_struct *sig = task->signal;
369 	bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
370 
371 	WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
372 
373 	task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
374 
375 	if (!consume)
376 		return false;
377 
378 	if (!WARN_ON_ONCE(sig->group_stop_count == 0))
379 		sig->group_stop_count--;
380 
381 	/*
382 	 * Tell the caller to notify completion iff we are entering into a
383 	 * fresh group stop.  Read comment in do_signal_stop() for details.
384 	 */
385 	if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
386 		signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
387 		return true;
388 	}
389 	return false;
390 }
391 
392 void task_join_group_stop(struct task_struct *task)
393 {
394 	unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
395 	struct signal_struct *sig = current->signal;
396 
397 	if (sig->group_stop_count) {
398 		sig->group_stop_count++;
399 		mask |= JOBCTL_STOP_CONSUME;
400 	} else if (!(sig->flags & SIGNAL_STOP_STOPPED))
401 		return;
402 
403 	/* Have the new thread join an on-going signal group stop */
404 	task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
405 }
406 
407 /*
408  * allocate a new signal queue record
409  * - this may be called without locks if and only if t == current, otherwise an
410  *   appropriate lock must be held to stop the target task from exiting
411  */
412 static struct sigqueue *
413 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t gfp_flags,
414 		 int override_rlimit, const unsigned int sigqueue_flags)
415 {
416 	struct sigqueue *q = NULL;
417 	struct ucounts *ucounts = NULL;
418 	long sigpending;
419 
420 	/*
421 	 * Protect access to @t credentials. This can go away when all
422 	 * callers hold rcu read lock.
423 	 *
424 	 * NOTE! A pending signal will hold on to the user refcount,
425 	 * and we get/put the refcount only when the sigpending count
426 	 * changes from/to zero.
427 	 */
428 	rcu_read_lock();
429 	ucounts = task_ucounts(t);
430 	sigpending = inc_rlimit_get_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
431 	rcu_read_unlock();
432 	if (!sigpending)
433 		return NULL;
434 
435 	if (override_rlimit || likely(sigpending <= task_rlimit(t, RLIMIT_SIGPENDING))) {
436 		q = kmem_cache_alloc(sigqueue_cachep, gfp_flags);
437 	} else {
438 		print_dropped_signal(sig);
439 	}
440 
441 	if (unlikely(q == NULL)) {
442 		dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
443 	} else {
444 		INIT_LIST_HEAD(&q->list);
445 		q->flags = sigqueue_flags;
446 		q->ucounts = ucounts;
447 	}
448 	return q;
449 }
450 
451 static void __sigqueue_free(struct sigqueue *q)
452 {
453 	if (q->flags & SIGQUEUE_PREALLOC)
454 		return;
455 	if (q->ucounts) {
456 		dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING);
457 		q->ucounts = NULL;
458 	}
459 	kmem_cache_free(sigqueue_cachep, q);
460 }
461 
462 void flush_sigqueue(struct sigpending *queue)
463 {
464 	struct sigqueue *q;
465 
466 	sigemptyset(&queue->signal);
467 	while (!list_empty(&queue->list)) {
468 		q = list_entry(queue->list.next, struct sigqueue , list);
469 		list_del_init(&q->list);
470 		__sigqueue_free(q);
471 	}
472 }
473 
474 /*
475  * Flush all pending signals for this kthread.
476  */
477 void flush_signals(struct task_struct *t)
478 {
479 	unsigned long flags;
480 
481 	spin_lock_irqsave(&t->sighand->siglock, flags);
482 	clear_tsk_thread_flag(t, TIF_SIGPENDING);
483 	flush_sigqueue(&t->pending);
484 	flush_sigqueue(&t->signal->shared_pending);
485 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
486 }
487 EXPORT_SYMBOL(flush_signals);
488 
489 #ifdef CONFIG_POSIX_TIMERS
490 static void __flush_itimer_signals(struct sigpending *pending)
491 {
492 	sigset_t signal, retain;
493 	struct sigqueue *q, *n;
494 
495 	signal = pending->signal;
496 	sigemptyset(&retain);
497 
498 	list_for_each_entry_safe(q, n, &pending->list, list) {
499 		int sig = q->info.si_signo;
500 
501 		if (likely(q->info.si_code != SI_TIMER)) {
502 			sigaddset(&retain, sig);
503 		} else {
504 			sigdelset(&signal, sig);
505 			list_del_init(&q->list);
506 			__sigqueue_free(q);
507 		}
508 	}
509 
510 	sigorsets(&pending->signal, &signal, &retain);
511 }
512 
513 void flush_itimer_signals(void)
514 {
515 	struct task_struct *tsk = current;
516 	unsigned long flags;
517 
518 	spin_lock_irqsave(&tsk->sighand->siglock, flags);
519 	__flush_itimer_signals(&tsk->pending);
520 	__flush_itimer_signals(&tsk->signal->shared_pending);
521 	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
522 }
523 #endif
524 
525 void ignore_signals(struct task_struct *t)
526 {
527 	int i;
528 
529 	for (i = 0; i < _NSIG; ++i)
530 		t->sighand->action[i].sa.sa_handler = SIG_IGN;
531 
532 	flush_signals(t);
533 }
534 
535 /*
536  * Flush all handlers for a task.
537  */
538 
539 void
540 flush_signal_handlers(struct task_struct *t, int force_default)
541 {
542 	int i;
543 	struct k_sigaction *ka = &t->sighand->action[0];
544 	for (i = _NSIG ; i != 0 ; i--) {
545 		if (force_default || ka->sa.sa_handler != SIG_IGN)
546 			ka->sa.sa_handler = SIG_DFL;
547 		ka->sa.sa_flags = 0;
548 #ifdef __ARCH_HAS_SA_RESTORER
549 		ka->sa.sa_restorer = NULL;
550 #endif
551 		sigemptyset(&ka->sa.sa_mask);
552 		ka++;
553 	}
554 }
555 
556 bool unhandled_signal(struct task_struct *tsk, int sig)
557 {
558 	void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
559 	if (is_global_init(tsk))
560 		return true;
561 
562 	if (handler != SIG_IGN && handler != SIG_DFL)
563 		return false;
564 
565 	/* If dying, we handle all new signals by ignoring them */
566 	if (fatal_signal_pending(tsk))
567 		return false;
568 
569 	/* if ptraced, let the tracer determine */
570 	return !tsk->ptrace;
571 }
572 
573 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
574 			   bool *resched_timer)
575 {
576 	struct sigqueue *q, *first = NULL;
577 
578 	/*
579 	 * Collect the siginfo appropriate to this signal.  Check if
580 	 * there is another siginfo for the same signal.
581 	*/
582 	list_for_each_entry(q, &list->list, list) {
583 		if (q->info.si_signo == sig) {
584 			if (first)
585 				goto still_pending;
586 			first = q;
587 		}
588 	}
589 
590 	sigdelset(&list->signal, sig);
591 
592 	if (first) {
593 still_pending:
594 		list_del_init(&first->list);
595 		copy_siginfo(info, &first->info);
596 
597 		*resched_timer =
598 			(first->flags & SIGQUEUE_PREALLOC) &&
599 			(info->si_code == SI_TIMER) &&
600 			(info->si_sys_private);
601 
602 		__sigqueue_free(first);
603 	} else {
604 		/*
605 		 * Ok, it wasn't in the queue.  This must be
606 		 * a fast-pathed signal or we must have been
607 		 * out of queue space.  So zero out the info.
608 		 */
609 		clear_siginfo(info);
610 		info->si_signo = sig;
611 		info->si_errno = 0;
612 		info->si_code = SI_USER;
613 		info->si_pid = 0;
614 		info->si_uid = 0;
615 	}
616 }
617 
618 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
619 			kernel_siginfo_t *info, bool *resched_timer)
620 {
621 	int sig = next_signal(pending, mask);
622 
623 	if (sig)
624 		collect_signal(sig, pending, info, resched_timer);
625 	return sig;
626 }
627 
628 /*
629  * Dequeue a signal and return the element to the caller, which is
630  * expected to free it.
631  *
632  * All callers have to hold the siglock.
633  */
634 int dequeue_signal(struct task_struct *tsk, sigset_t *mask,
635 		   kernel_siginfo_t *info, enum pid_type *type)
636 {
637 	bool resched_timer = false;
638 	int signr;
639 
640 	/* We only dequeue private signals from ourselves, we don't let
641 	 * signalfd steal them
642 	 */
643 	*type = PIDTYPE_PID;
644 	signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
645 	if (!signr) {
646 		*type = PIDTYPE_TGID;
647 		signr = __dequeue_signal(&tsk->signal->shared_pending,
648 					 mask, info, &resched_timer);
649 #ifdef CONFIG_POSIX_TIMERS
650 		/*
651 		 * itimer signal ?
652 		 *
653 		 * itimers are process shared and we restart periodic
654 		 * itimers in the signal delivery path to prevent DoS
655 		 * attacks in the high resolution timer case. This is
656 		 * compliant with the old way of self-restarting
657 		 * itimers, as the SIGALRM is a legacy signal and only
658 		 * queued once. Changing the restart behaviour to
659 		 * restart the timer in the signal dequeue path is
660 		 * reducing the timer noise on heavy loaded !highres
661 		 * systems too.
662 		 */
663 		if (unlikely(signr == SIGALRM)) {
664 			struct hrtimer *tmr = &tsk->signal->real_timer;
665 
666 			if (!hrtimer_is_queued(tmr) &&
667 			    tsk->signal->it_real_incr != 0) {
668 				hrtimer_forward(tmr, tmr->base->get_time(),
669 						tsk->signal->it_real_incr);
670 				hrtimer_restart(tmr);
671 			}
672 		}
673 #endif
674 	}
675 
676 	recalc_sigpending();
677 	if (!signr)
678 		return 0;
679 
680 	if (unlikely(sig_kernel_stop(signr))) {
681 		/*
682 		 * Set a marker that we have dequeued a stop signal.  Our
683 		 * caller might release the siglock and then the pending
684 		 * stop signal it is about to process is no longer in the
685 		 * pending bitmasks, but must still be cleared by a SIGCONT
686 		 * (and overruled by a SIGKILL).  So those cases clear this
687 		 * shared flag after we've set it.  Note that this flag may
688 		 * remain set after the signal we return is ignored or
689 		 * handled.  That doesn't matter because its only purpose
690 		 * is to alert stop-signal processing code when another
691 		 * processor has come along and cleared the flag.
692 		 */
693 		current->jobctl |= JOBCTL_STOP_DEQUEUED;
694 	}
695 #ifdef CONFIG_POSIX_TIMERS
696 	if (resched_timer) {
697 		/*
698 		 * Release the siglock to ensure proper locking order
699 		 * of timer locks outside of siglocks.  Note, we leave
700 		 * irqs disabled here, since the posix-timers code is
701 		 * about to disable them again anyway.
702 		 */
703 		spin_unlock(&tsk->sighand->siglock);
704 		posixtimer_rearm(info);
705 		spin_lock(&tsk->sighand->siglock);
706 
707 		/* Don't expose the si_sys_private value to userspace */
708 		info->si_sys_private = 0;
709 	}
710 #endif
711 	return signr;
712 }
713 EXPORT_SYMBOL_GPL(dequeue_signal);
714 
715 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
716 {
717 	struct task_struct *tsk = current;
718 	struct sigpending *pending = &tsk->pending;
719 	struct sigqueue *q, *sync = NULL;
720 
721 	/*
722 	 * Might a synchronous signal be in the queue?
723 	 */
724 	if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
725 		return 0;
726 
727 	/*
728 	 * Return the first synchronous signal in the queue.
729 	 */
730 	list_for_each_entry(q, &pending->list, list) {
731 		/* Synchronous signals have a positive si_code */
732 		if ((q->info.si_code > SI_USER) &&
733 		    (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
734 			sync = q;
735 			goto next;
736 		}
737 	}
738 	return 0;
739 next:
740 	/*
741 	 * Check if there is another siginfo for the same signal.
742 	 */
743 	list_for_each_entry_continue(q, &pending->list, list) {
744 		if (q->info.si_signo == sync->info.si_signo)
745 			goto still_pending;
746 	}
747 
748 	sigdelset(&pending->signal, sync->info.si_signo);
749 	recalc_sigpending();
750 still_pending:
751 	list_del_init(&sync->list);
752 	copy_siginfo(info, &sync->info);
753 	__sigqueue_free(sync);
754 	return info->si_signo;
755 }
756 
757 /*
758  * Tell a process that it has a new active signal..
759  *
760  * NOTE! we rely on the previous spin_lock to
761  * lock interrupts for us! We can only be called with
762  * "siglock" held, and the local interrupt must
763  * have been disabled when that got acquired!
764  *
765  * No need to set need_resched since signal event passing
766  * goes through ->blocked
767  */
768 void signal_wake_up_state(struct task_struct *t, unsigned int state)
769 {
770 	lockdep_assert_held(&t->sighand->siglock);
771 
772 	set_tsk_thread_flag(t, TIF_SIGPENDING);
773 
774 	/*
775 	 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
776 	 * case. We don't check t->state here because there is a race with it
777 	 * executing another processor and just now entering stopped state.
778 	 * By using wake_up_state, we ensure the process will wake up and
779 	 * handle its death signal.
780 	 */
781 	if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
782 		kick_process(t);
783 }
784 
785 /*
786  * Remove signals in mask from the pending set and queue.
787  * Returns 1 if any signals were found.
788  *
789  * All callers must be holding the siglock.
790  */
791 static void flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
792 {
793 	struct sigqueue *q, *n;
794 	sigset_t m;
795 
796 	sigandsets(&m, mask, &s->signal);
797 	if (sigisemptyset(&m))
798 		return;
799 
800 	sigandnsets(&s->signal, &s->signal, mask);
801 	list_for_each_entry_safe(q, n, &s->list, list) {
802 		if (sigismember(mask, q->info.si_signo)) {
803 			list_del_init(&q->list);
804 			__sigqueue_free(q);
805 		}
806 	}
807 }
808 
809 static inline int is_si_special(const struct kernel_siginfo *info)
810 {
811 	return info <= SEND_SIG_PRIV;
812 }
813 
814 static inline bool si_fromuser(const struct kernel_siginfo *info)
815 {
816 	return info == SEND_SIG_NOINFO ||
817 		(!is_si_special(info) && SI_FROMUSER(info));
818 }
819 
820 /*
821  * called with RCU read lock from check_kill_permission()
822  */
823 static bool kill_ok_by_cred(struct task_struct *t)
824 {
825 	const struct cred *cred = current_cred();
826 	const struct cred *tcred = __task_cred(t);
827 
828 	return uid_eq(cred->euid, tcred->suid) ||
829 	       uid_eq(cred->euid, tcred->uid) ||
830 	       uid_eq(cred->uid, tcred->suid) ||
831 	       uid_eq(cred->uid, tcred->uid) ||
832 	       ns_capable(tcred->user_ns, CAP_KILL);
833 }
834 
835 /*
836  * Bad permissions for sending the signal
837  * - the caller must hold the RCU read lock
838  */
839 static int check_kill_permission(int sig, struct kernel_siginfo *info,
840 				 struct task_struct *t)
841 {
842 	struct pid *sid;
843 	int error;
844 
845 	if (!valid_signal(sig))
846 		return -EINVAL;
847 
848 	if (!si_fromuser(info))
849 		return 0;
850 
851 	error = audit_signal_info(sig, t); /* Let audit system see the signal */
852 	if (error)
853 		return error;
854 
855 	if (!same_thread_group(current, t) &&
856 	    !kill_ok_by_cred(t)) {
857 		switch (sig) {
858 		case SIGCONT:
859 			sid = task_session(t);
860 			/*
861 			 * We don't return the error if sid == NULL. The
862 			 * task was unhashed, the caller must notice this.
863 			 */
864 			if (!sid || sid == task_session(current))
865 				break;
866 			fallthrough;
867 		default:
868 			return -EPERM;
869 		}
870 	}
871 
872 	return security_task_kill(t, info, sig, NULL);
873 }
874 
875 /**
876  * ptrace_trap_notify - schedule trap to notify ptracer
877  * @t: tracee wanting to notify tracer
878  *
879  * This function schedules sticky ptrace trap which is cleared on the next
880  * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
881  * ptracer.
882  *
883  * If @t is running, STOP trap will be taken.  If trapped for STOP and
884  * ptracer is listening for events, tracee is woken up so that it can
885  * re-trap for the new event.  If trapped otherwise, STOP trap will be
886  * eventually taken without returning to userland after the existing traps
887  * are finished by PTRACE_CONT.
888  *
889  * CONTEXT:
890  * Must be called with @task->sighand->siglock held.
891  */
892 static void ptrace_trap_notify(struct task_struct *t)
893 {
894 	WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
895 	lockdep_assert_held(&t->sighand->siglock);
896 
897 	task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
898 	ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
899 }
900 
901 /*
902  * Handle magic process-wide effects of stop/continue signals. Unlike
903  * the signal actions, these happen immediately at signal-generation
904  * time regardless of blocking, ignoring, or handling.  This does the
905  * actual continuing for SIGCONT, but not the actual stopping for stop
906  * signals. The process stop is done as a signal action for SIG_DFL.
907  *
908  * Returns true if the signal should be actually delivered, otherwise
909  * it should be dropped.
910  */
911 static bool prepare_signal(int sig, struct task_struct *p, bool force)
912 {
913 	struct signal_struct *signal = p->signal;
914 	struct task_struct *t;
915 	sigset_t flush;
916 
917 	if (signal->flags & SIGNAL_GROUP_EXIT) {
918 		if (signal->core_state)
919 			return sig == SIGKILL;
920 		/*
921 		 * The process is in the middle of dying, drop the signal.
922 		 */
923 		return false;
924 	} else if (sig_kernel_stop(sig)) {
925 		/*
926 		 * This is a stop signal.  Remove SIGCONT from all queues.
927 		 */
928 		siginitset(&flush, sigmask(SIGCONT));
929 		flush_sigqueue_mask(&flush, &signal->shared_pending);
930 		for_each_thread(p, t)
931 			flush_sigqueue_mask(&flush, &t->pending);
932 	} else if (sig == SIGCONT) {
933 		unsigned int why;
934 		/*
935 		 * Remove all stop signals from all queues, wake all threads.
936 		 */
937 		siginitset(&flush, SIG_KERNEL_STOP_MASK);
938 		flush_sigqueue_mask(&flush, &signal->shared_pending);
939 		for_each_thread(p, t) {
940 			flush_sigqueue_mask(&flush, &t->pending);
941 			task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
942 			if (likely(!(t->ptrace & PT_SEIZED))) {
943 				t->jobctl &= ~JOBCTL_STOPPED;
944 				wake_up_state(t, __TASK_STOPPED);
945 			} else
946 				ptrace_trap_notify(t);
947 		}
948 
949 		/*
950 		 * Notify the parent with CLD_CONTINUED if we were stopped.
951 		 *
952 		 * If we were in the middle of a group stop, we pretend it
953 		 * was already finished, and then continued. Since SIGCHLD
954 		 * doesn't queue we report only CLD_STOPPED, as if the next
955 		 * CLD_CONTINUED was dropped.
956 		 */
957 		why = 0;
958 		if (signal->flags & SIGNAL_STOP_STOPPED)
959 			why |= SIGNAL_CLD_CONTINUED;
960 		else if (signal->group_stop_count)
961 			why |= SIGNAL_CLD_STOPPED;
962 
963 		if (why) {
964 			/*
965 			 * The first thread which returns from do_signal_stop()
966 			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
967 			 * notify its parent. See get_signal().
968 			 */
969 			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
970 			signal->group_stop_count = 0;
971 			signal->group_exit_code = 0;
972 		}
973 	}
974 
975 	return !sig_ignored(p, sig, force);
976 }
977 
978 /*
979  * Test if P wants to take SIG.  After we've checked all threads with this,
980  * it's equivalent to finding no threads not blocking SIG.  Any threads not
981  * blocking SIG were ruled out because they are not running and already
982  * have pending signals.  Such threads will dequeue from the shared queue
983  * as soon as they're available, so putting the signal on the shared queue
984  * will be equivalent to sending it to one such thread.
985  */
986 static inline bool wants_signal(int sig, struct task_struct *p)
987 {
988 	if (sigismember(&p->blocked, sig))
989 		return false;
990 
991 	if (p->flags & PF_EXITING)
992 		return false;
993 
994 	if (sig == SIGKILL)
995 		return true;
996 
997 	if (task_is_stopped_or_traced(p))
998 		return false;
999 
1000 	return task_curr(p) || !task_sigpending(p);
1001 }
1002 
1003 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
1004 {
1005 	struct signal_struct *signal = p->signal;
1006 	struct task_struct *t;
1007 
1008 	/*
1009 	 * Now find a thread we can wake up to take the signal off the queue.
1010 	 *
1011 	 * Try the suggested task first (may or may not be the main thread).
1012 	 */
1013 	if (wants_signal(sig, p))
1014 		t = p;
1015 	else if ((type == PIDTYPE_PID) || thread_group_empty(p))
1016 		/*
1017 		 * There is just one thread and it does not need to be woken.
1018 		 * It will dequeue unblocked signals before it runs again.
1019 		 */
1020 		return;
1021 	else {
1022 		/*
1023 		 * Otherwise try to find a suitable thread.
1024 		 */
1025 		t = signal->curr_target;
1026 		while (!wants_signal(sig, t)) {
1027 			t = next_thread(t);
1028 			if (t == signal->curr_target)
1029 				/*
1030 				 * No thread needs to be woken.
1031 				 * Any eligible threads will see
1032 				 * the signal in the queue soon.
1033 				 */
1034 				return;
1035 		}
1036 		signal->curr_target = t;
1037 	}
1038 
1039 	/*
1040 	 * Found a killable thread.  If the signal will be fatal,
1041 	 * then start taking the whole group down immediately.
1042 	 */
1043 	if (sig_fatal(p, sig) &&
1044 	    (signal->core_state || !(signal->flags & SIGNAL_GROUP_EXIT)) &&
1045 	    !sigismember(&t->real_blocked, sig) &&
1046 	    (sig == SIGKILL || !p->ptrace)) {
1047 		/*
1048 		 * This signal will be fatal to the whole group.
1049 		 */
1050 		if (!sig_kernel_coredump(sig)) {
1051 			/*
1052 			 * Start a group exit and wake everybody up.
1053 			 * This way we don't have other threads
1054 			 * running and doing things after a slower
1055 			 * thread has the fatal signal pending.
1056 			 */
1057 			signal->flags = SIGNAL_GROUP_EXIT;
1058 			signal->group_exit_code = sig;
1059 			signal->group_stop_count = 0;
1060 			t = p;
1061 			do {
1062 				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1063 				sigaddset(&t->pending.signal, SIGKILL);
1064 				signal_wake_up(t, 1);
1065 			} while_each_thread(p, t);
1066 			return;
1067 		}
1068 	}
1069 
1070 	/*
1071 	 * The signal is already in the shared-pending queue.
1072 	 * Tell the chosen thread to wake up and dequeue it.
1073 	 */
1074 	signal_wake_up(t, sig == SIGKILL);
1075 	return;
1076 }
1077 
1078 static inline bool legacy_queue(struct sigpending *signals, int sig)
1079 {
1080 	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1081 }
1082 
1083 static int __send_signal_locked(int sig, struct kernel_siginfo *info,
1084 				struct task_struct *t, enum pid_type type, bool force)
1085 {
1086 	struct sigpending *pending;
1087 	struct sigqueue *q;
1088 	int override_rlimit;
1089 	int ret = 0, result;
1090 
1091 	lockdep_assert_held(&t->sighand->siglock);
1092 
1093 	result = TRACE_SIGNAL_IGNORED;
1094 	if (!prepare_signal(sig, t, force))
1095 		goto ret;
1096 
1097 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1098 	/*
1099 	 * Short-circuit ignored signals and support queuing
1100 	 * exactly one non-rt signal, so that we can get more
1101 	 * detailed information about the cause of the signal.
1102 	 */
1103 	result = TRACE_SIGNAL_ALREADY_PENDING;
1104 	if (legacy_queue(pending, sig))
1105 		goto ret;
1106 
1107 	result = TRACE_SIGNAL_DELIVERED;
1108 	/*
1109 	 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1110 	 */
1111 	if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1112 		goto out_set;
1113 
1114 	/*
1115 	 * Real-time signals must be queued if sent by sigqueue, or
1116 	 * some other real-time mechanism.  It is implementation
1117 	 * defined whether kill() does so.  We attempt to do so, on
1118 	 * the principle of least surprise, but since kill is not
1119 	 * allowed to fail with EAGAIN when low on memory we just
1120 	 * make sure at least one signal gets delivered and don't
1121 	 * pass on the info struct.
1122 	 */
1123 	if (sig < SIGRTMIN)
1124 		override_rlimit = (is_si_special(info) || info->si_code >= 0);
1125 	else
1126 		override_rlimit = 0;
1127 
1128 	q = __sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit, 0);
1129 
1130 	if (q) {
1131 		list_add_tail(&q->list, &pending->list);
1132 		switch ((unsigned long) info) {
1133 		case (unsigned long) SEND_SIG_NOINFO:
1134 			clear_siginfo(&q->info);
1135 			q->info.si_signo = sig;
1136 			q->info.si_errno = 0;
1137 			q->info.si_code = SI_USER;
1138 			q->info.si_pid = task_tgid_nr_ns(current,
1139 							task_active_pid_ns(t));
1140 			rcu_read_lock();
1141 			q->info.si_uid =
1142 				from_kuid_munged(task_cred_xxx(t, user_ns),
1143 						 current_uid());
1144 			rcu_read_unlock();
1145 			break;
1146 		case (unsigned long) SEND_SIG_PRIV:
1147 			clear_siginfo(&q->info);
1148 			q->info.si_signo = sig;
1149 			q->info.si_errno = 0;
1150 			q->info.si_code = SI_KERNEL;
1151 			q->info.si_pid = 0;
1152 			q->info.si_uid = 0;
1153 			break;
1154 		default:
1155 			copy_siginfo(&q->info, info);
1156 			break;
1157 		}
1158 	} else if (!is_si_special(info) &&
1159 		   sig >= SIGRTMIN && info->si_code != SI_USER) {
1160 		/*
1161 		 * Queue overflow, abort.  We may abort if the
1162 		 * signal was rt and sent by user using something
1163 		 * other than kill().
1164 		 */
1165 		result = TRACE_SIGNAL_OVERFLOW_FAIL;
1166 		ret = -EAGAIN;
1167 		goto ret;
1168 	} else {
1169 		/*
1170 		 * This is a silent loss of information.  We still
1171 		 * send the signal, but the *info bits are lost.
1172 		 */
1173 		result = TRACE_SIGNAL_LOSE_INFO;
1174 	}
1175 
1176 out_set:
1177 	signalfd_notify(t, sig);
1178 	sigaddset(&pending->signal, sig);
1179 
1180 	/* Let multiprocess signals appear after on-going forks */
1181 	if (type > PIDTYPE_TGID) {
1182 		struct multiprocess_signals *delayed;
1183 		hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1184 			sigset_t *signal = &delayed->signal;
1185 			/* Can't queue both a stop and a continue signal */
1186 			if (sig == SIGCONT)
1187 				sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1188 			else if (sig_kernel_stop(sig))
1189 				sigdelset(signal, SIGCONT);
1190 			sigaddset(signal, sig);
1191 		}
1192 	}
1193 
1194 	complete_signal(sig, t, type);
1195 ret:
1196 	trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1197 	return ret;
1198 }
1199 
1200 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1201 {
1202 	bool ret = false;
1203 	switch (siginfo_layout(info->si_signo, info->si_code)) {
1204 	case SIL_KILL:
1205 	case SIL_CHLD:
1206 	case SIL_RT:
1207 		ret = true;
1208 		break;
1209 	case SIL_TIMER:
1210 	case SIL_POLL:
1211 	case SIL_FAULT:
1212 	case SIL_FAULT_TRAPNO:
1213 	case SIL_FAULT_MCEERR:
1214 	case SIL_FAULT_BNDERR:
1215 	case SIL_FAULT_PKUERR:
1216 	case SIL_FAULT_PERF_EVENT:
1217 	case SIL_SYS:
1218 		ret = false;
1219 		break;
1220 	}
1221 	return ret;
1222 }
1223 
1224 int send_signal_locked(int sig, struct kernel_siginfo *info,
1225 		       struct task_struct *t, enum pid_type type)
1226 {
1227 	/* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1228 	bool force = false;
1229 
1230 	if (info == SEND_SIG_NOINFO) {
1231 		/* Force if sent from an ancestor pid namespace */
1232 		force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1233 	} else if (info == SEND_SIG_PRIV) {
1234 		/* Don't ignore kernel generated signals */
1235 		force = true;
1236 	} else if (has_si_pid_and_uid(info)) {
1237 		/* SIGKILL and SIGSTOP is special or has ids */
1238 		struct user_namespace *t_user_ns;
1239 
1240 		rcu_read_lock();
1241 		t_user_ns = task_cred_xxx(t, user_ns);
1242 		if (current_user_ns() != t_user_ns) {
1243 			kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1244 			info->si_uid = from_kuid_munged(t_user_ns, uid);
1245 		}
1246 		rcu_read_unlock();
1247 
1248 		/* A kernel generated signal? */
1249 		force = (info->si_code == SI_KERNEL);
1250 
1251 		/* From an ancestor pid namespace? */
1252 		if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1253 			info->si_pid = 0;
1254 			force = true;
1255 		}
1256 	}
1257 	return __send_signal_locked(sig, info, t, type, force);
1258 }
1259 
1260 static void print_fatal_signal(int signr)
1261 {
1262 	struct pt_regs *regs = task_pt_regs(current);
1263 	pr_info("potentially unexpected fatal signal %d.\n", signr);
1264 
1265 #if defined(__i386__) && !defined(__arch_um__)
1266 	pr_info("code at %08lx: ", regs->ip);
1267 	{
1268 		int i;
1269 		for (i = 0; i < 16; i++) {
1270 			unsigned char insn;
1271 
1272 			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1273 				break;
1274 			pr_cont("%02x ", insn);
1275 		}
1276 	}
1277 	pr_cont("\n");
1278 #endif
1279 	preempt_disable();
1280 	show_regs(regs);
1281 	preempt_enable();
1282 }
1283 
1284 static int __init setup_print_fatal_signals(char *str)
1285 {
1286 	get_option (&str, &print_fatal_signals);
1287 
1288 	return 1;
1289 }
1290 
1291 __setup("print-fatal-signals=", setup_print_fatal_signals);
1292 
1293 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1294 			enum pid_type type)
1295 {
1296 	unsigned long flags;
1297 	int ret = -ESRCH;
1298 
1299 	if (lock_task_sighand(p, &flags)) {
1300 		ret = send_signal_locked(sig, info, p, type);
1301 		unlock_task_sighand(p, &flags);
1302 	}
1303 
1304 	return ret;
1305 }
1306 
1307 enum sig_handler {
1308 	HANDLER_CURRENT, /* If reachable use the current handler */
1309 	HANDLER_SIG_DFL, /* Always use SIG_DFL handler semantics */
1310 	HANDLER_EXIT,	 /* Only visible as the process exit code */
1311 };
1312 
1313 /*
1314  * Force a signal that the process can't ignore: if necessary
1315  * we unblock the signal and change any SIG_IGN to SIG_DFL.
1316  *
1317  * Note: If we unblock the signal, we always reset it to SIG_DFL,
1318  * since we do not want to have a signal handler that was blocked
1319  * be invoked when user space had explicitly blocked it.
1320  *
1321  * We don't want to have recursive SIGSEGV's etc, for example,
1322  * that is why we also clear SIGNAL_UNKILLABLE.
1323  */
1324 static int
1325 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t,
1326 	enum sig_handler handler)
1327 {
1328 	unsigned long int flags;
1329 	int ret, blocked, ignored;
1330 	struct k_sigaction *action;
1331 	int sig = info->si_signo;
1332 
1333 	spin_lock_irqsave(&t->sighand->siglock, flags);
1334 	action = &t->sighand->action[sig-1];
1335 	ignored = action->sa.sa_handler == SIG_IGN;
1336 	blocked = sigismember(&t->blocked, sig);
1337 	if (blocked || ignored || (handler != HANDLER_CURRENT)) {
1338 		action->sa.sa_handler = SIG_DFL;
1339 		if (handler == HANDLER_EXIT)
1340 			action->sa.sa_flags |= SA_IMMUTABLE;
1341 		if (blocked) {
1342 			sigdelset(&t->blocked, sig);
1343 			recalc_sigpending_and_wake(t);
1344 		}
1345 	}
1346 	/*
1347 	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1348 	 * debugging to leave init killable. But HANDLER_EXIT is always fatal.
1349 	 */
1350 	if (action->sa.sa_handler == SIG_DFL &&
1351 	    (!t->ptrace || (handler == HANDLER_EXIT)))
1352 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1353 	ret = send_signal_locked(sig, info, t, PIDTYPE_PID);
1354 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1355 
1356 	return ret;
1357 }
1358 
1359 int force_sig_info(struct kernel_siginfo *info)
1360 {
1361 	return force_sig_info_to_task(info, current, HANDLER_CURRENT);
1362 }
1363 
1364 /*
1365  * Nuke all other threads in the group.
1366  */
1367 int zap_other_threads(struct task_struct *p)
1368 {
1369 	struct task_struct *t = p;
1370 	int count = 0;
1371 
1372 	p->signal->group_stop_count = 0;
1373 
1374 	while_each_thread(p, t) {
1375 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1376 		/* Don't require de_thread to wait for the vhost_worker */
1377 		if ((t->flags & (PF_IO_WORKER | PF_USER_WORKER)) != PF_USER_WORKER)
1378 			count++;
1379 
1380 		/* Don't bother with already dead threads */
1381 		if (t->exit_state)
1382 			continue;
1383 		sigaddset(&t->pending.signal, SIGKILL);
1384 		signal_wake_up(t, 1);
1385 	}
1386 
1387 	return count;
1388 }
1389 
1390 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1391 					   unsigned long *flags)
1392 {
1393 	struct sighand_struct *sighand;
1394 
1395 	rcu_read_lock();
1396 	for (;;) {
1397 		sighand = rcu_dereference(tsk->sighand);
1398 		if (unlikely(sighand == NULL))
1399 			break;
1400 
1401 		/*
1402 		 * This sighand can be already freed and even reused, but
1403 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1404 		 * initializes ->siglock: this slab can't go away, it has
1405 		 * the same object type, ->siglock can't be reinitialized.
1406 		 *
1407 		 * We need to ensure that tsk->sighand is still the same
1408 		 * after we take the lock, we can race with de_thread() or
1409 		 * __exit_signal(). In the latter case the next iteration
1410 		 * must see ->sighand == NULL.
1411 		 */
1412 		spin_lock_irqsave(&sighand->siglock, *flags);
1413 		if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1414 			break;
1415 		spin_unlock_irqrestore(&sighand->siglock, *flags);
1416 	}
1417 	rcu_read_unlock();
1418 
1419 	return sighand;
1420 }
1421 
1422 #ifdef CONFIG_LOCKDEP
1423 void lockdep_assert_task_sighand_held(struct task_struct *task)
1424 {
1425 	struct sighand_struct *sighand;
1426 
1427 	rcu_read_lock();
1428 	sighand = rcu_dereference(task->sighand);
1429 	if (sighand)
1430 		lockdep_assert_held(&sighand->siglock);
1431 	else
1432 		WARN_ON_ONCE(1);
1433 	rcu_read_unlock();
1434 }
1435 #endif
1436 
1437 /*
1438  * send signal info to all the members of a group
1439  */
1440 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1441 			struct task_struct *p, enum pid_type type)
1442 {
1443 	int ret;
1444 
1445 	rcu_read_lock();
1446 	ret = check_kill_permission(sig, info, p);
1447 	rcu_read_unlock();
1448 
1449 	if (!ret && sig)
1450 		ret = do_send_sig_info(sig, info, p, type);
1451 
1452 	return ret;
1453 }
1454 
1455 /*
1456  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1457  * control characters do (^C, ^Z etc)
1458  * - the caller must hold at least a readlock on tasklist_lock
1459  */
1460 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1461 {
1462 	struct task_struct *p = NULL;
1463 	int retval, success;
1464 
1465 	success = 0;
1466 	retval = -ESRCH;
1467 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1468 		int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1469 		success |= !err;
1470 		retval = err;
1471 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1472 	return success ? 0 : retval;
1473 }
1474 
1475 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1476 {
1477 	int error = -ESRCH;
1478 	struct task_struct *p;
1479 
1480 	for (;;) {
1481 		rcu_read_lock();
1482 		p = pid_task(pid, PIDTYPE_PID);
1483 		if (p)
1484 			error = group_send_sig_info(sig, info, p, PIDTYPE_TGID);
1485 		rcu_read_unlock();
1486 		if (likely(!p || error != -ESRCH))
1487 			return error;
1488 
1489 		/*
1490 		 * The task was unhashed in between, try again.  If it
1491 		 * is dead, pid_task() will return NULL, if we race with
1492 		 * de_thread() it will find the new leader.
1493 		 */
1494 	}
1495 }
1496 
1497 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1498 {
1499 	int error;
1500 	rcu_read_lock();
1501 	error = kill_pid_info(sig, info, find_vpid(pid));
1502 	rcu_read_unlock();
1503 	return error;
1504 }
1505 
1506 static inline bool kill_as_cred_perm(const struct cred *cred,
1507 				     struct task_struct *target)
1508 {
1509 	const struct cred *pcred = __task_cred(target);
1510 
1511 	return uid_eq(cred->euid, pcred->suid) ||
1512 	       uid_eq(cred->euid, pcred->uid) ||
1513 	       uid_eq(cred->uid, pcred->suid) ||
1514 	       uid_eq(cred->uid, pcred->uid);
1515 }
1516 
1517 /*
1518  * The usb asyncio usage of siginfo is wrong.  The glibc support
1519  * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1520  * AKA after the generic fields:
1521  *	kernel_pid_t	si_pid;
1522  *	kernel_uid32_t	si_uid;
1523  *	sigval_t	si_value;
1524  *
1525  * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1526  * after the generic fields is:
1527  *	void __user 	*si_addr;
1528  *
1529  * This is a practical problem when there is a 64bit big endian kernel
1530  * and a 32bit userspace.  As the 32bit address will encoded in the low
1531  * 32bits of the pointer.  Those low 32bits will be stored at higher
1532  * address than appear in a 32 bit pointer.  So userspace will not
1533  * see the address it was expecting for it's completions.
1534  *
1535  * There is nothing in the encoding that can allow
1536  * copy_siginfo_to_user32 to detect this confusion of formats, so
1537  * handle this by requiring the caller of kill_pid_usb_asyncio to
1538  * notice when this situration takes place and to store the 32bit
1539  * pointer in sival_int, instead of sival_addr of the sigval_t addr
1540  * parameter.
1541  */
1542 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1543 			 struct pid *pid, const struct cred *cred)
1544 {
1545 	struct kernel_siginfo info;
1546 	struct task_struct *p;
1547 	unsigned long flags;
1548 	int ret = -EINVAL;
1549 
1550 	if (!valid_signal(sig))
1551 		return ret;
1552 
1553 	clear_siginfo(&info);
1554 	info.si_signo = sig;
1555 	info.si_errno = errno;
1556 	info.si_code = SI_ASYNCIO;
1557 	*((sigval_t *)&info.si_pid) = addr;
1558 
1559 	rcu_read_lock();
1560 	p = pid_task(pid, PIDTYPE_PID);
1561 	if (!p) {
1562 		ret = -ESRCH;
1563 		goto out_unlock;
1564 	}
1565 	if (!kill_as_cred_perm(cred, p)) {
1566 		ret = -EPERM;
1567 		goto out_unlock;
1568 	}
1569 	ret = security_task_kill(p, &info, sig, cred);
1570 	if (ret)
1571 		goto out_unlock;
1572 
1573 	if (sig) {
1574 		if (lock_task_sighand(p, &flags)) {
1575 			ret = __send_signal_locked(sig, &info, p, PIDTYPE_TGID, false);
1576 			unlock_task_sighand(p, &flags);
1577 		} else
1578 			ret = -ESRCH;
1579 	}
1580 out_unlock:
1581 	rcu_read_unlock();
1582 	return ret;
1583 }
1584 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1585 
1586 /*
1587  * kill_something_info() interprets pid in interesting ways just like kill(2).
1588  *
1589  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1590  * is probably wrong.  Should make it like BSD or SYSV.
1591  */
1592 
1593 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1594 {
1595 	int ret;
1596 
1597 	if (pid > 0)
1598 		return kill_proc_info(sig, info, pid);
1599 
1600 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1601 	if (pid == INT_MIN)
1602 		return -ESRCH;
1603 
1604 	read_lock(&tasklist_lock);
1605 	if (pid != -1) {
1606 		ret = __kill_pgrp_info(sig, info,
1607 				pid ? find_vpid(-pid) : task_pgrp(current));
1608 	} else {
1609 		int retval = 0, count = 0;
1610 		struct task_struct * p;
1611 
1612 		for_each_process(p) {
1613 			if (task_pid_vnr(p) > 1 &&
1614 					!same_thread_group(p, current)) {
1615 				int err = group_send_sig_info(sig, info, p,
1616 							      PIDTYPE_MAX);
1617 				++count;
1618 				if (err != -EPERM)
1619 					retval = err;
1620 			}
1621 		}
1622 		ret = count ? retval : -ESRCH;
1623 	}
1624 	read_unlock(&tasklist_lock);
1625 
1626 	return ret;
1627 }
1628 
1629 /*
1630  * These are for backward compatibility with the rest of the kernel source.
1631  */
1632 
1633 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1634 {
1635 	/*
1636 	 * Make sure legacy kernel users don't send in bad values
1637 	 * (normal paths check this in check_kill_permission).
1638 	 */
1639 	if (!valid_signal(sig))
1640 		return -EINVAL;
1641 
1642 	return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1643 }
1644 EXPORT_SYMBOL(send_sig_info);
1645 
1646 #define __si_special(priv) \
1647 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1648 
1649 int
1650 send_sig(int sig, struct task_struct *p, int priv)
1651 {
1652 	return send_sig_info(sig, __si_special(priv), p);
1653 }
1654 EXPORT_SYMBOL(send_sig);
1655 
1656 void force_sig(int sig)
1657 {
1658 	struct kernel_siginfo info;
1659 
1660 	clear_siginfo(&info);
1661 	info.si_signo = sig;
1662 	info.si_errno = 0;
1663 	info.si_code = SI_KERNEL;
1664 	info.si_pid = 0;
1665 	info.si_uid = 0;
1666 	force_sig_info(&info);
1667 }
1668 EXPORT_SYMBOL(force_sig);
1669 
1670 void force_fatal_sig(int sig)
1671 {
1672 	struct kernel_siginfo info;
1673 
1674 	clear_siginfo(&info);
1675 	info.si_signo = sig;
1676 	info.si_errno = 0;
1677 	info.si_code = SI_KERNEL;
1678 	info.si_pid = 0;
1679 	info.si_uid = 0;
1680 	force_sig_info_to_task(&info, current, HANDLER_SIG_DFL);
1681 }
1682 
1683 void force_exit_sig(int sig)
1684 {
1685 	struct kernel_siginfo info;
1686 
1687 	clear_siginfo(&info);
1688 	info.si_signo = sig;
1689 	info.si_errno = 0;
1690 	info.si_code = SI_KERNEL;
1691 	info.si_pid = 0;
1692 	info.si_uid = 0;
1693 	force_sig_info_to_task(&info, current, HANDLER_EXIT);
1694 }
1695 
1696 /*
1697  * When things go south during signal handling, we
1698  * will force a SIGSEGV. And if the signal that caused
1699  * the problem was already a SIGSEGV, we'll want to
1700  * make sure we don't even try to deliver the signal..
1701  */
1702 void force_sigsegv(int sig)
1703 {
1704 	if (sig == SIGSEGV)
1705 		force_fatal_sig(SIGSEGV);
1706 	else
1707 		force_sig(SIGSEGV);
1708 }
1709 
1710 int force_sig_fault_to_task(int sig, int code, void __user *addr
1711 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1712 	, struct task_struct *t)
1713 {
1714 	struct kernel_siginfo info;
1715 
1716 	clear_siginfo(&info);
1717 	info.si_signo = sig;
1718 	info.si_errno = 0;
1719 	info.si_code  = code;
1720 	info.si_addr  = addr;
1721 #ifdef __ia64__
1722 	info.si_imm = imm;
1723 	info.si_flags = flags;
1724 	info.si_isr = isr;
1725 #endif
1726 	return force_sig_info_to_task(&info, t, HANDLER_CURRENT);
1727 }
1728 
1729 int force_sig_fault(int sig, int code, void __user *addr
1730 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr))
1731 {
1732 	return force_sig_fault_to_task(sig, code, addr
1733 				       ___ARCH_SI_IA64(imm, flags, isr), current);
1734 }
1735 
1736 int send_sig_fault(int sig, int code, void __user *addr
1737 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1738 	, struct task_struct *t)
1739 {
1740 	struct kernel_siginfo info;
1741 
1742 	clear_siginfo(&info);
1743 	info.si_signo = sig;
1744 	info.si_errno = 0;
1745 	info.si_code  = code;
1746 	info.si_addr  = addr;
1747 #ifdef __ia64__
1748 	info.si_imm = imm;
1749 	info.si_flags = flags;
1750 	info.si_isr = isr;
1751 #endif
1752 	return send_sig_info(info.si_signo, &info, t);
1753 }
1754 
1755 int force_sig_mceerr(int code, void __user *addr, short lsb)
1756 {
1757 	struct kernel_siginfo info;
1758 
1759 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1760 	clear_siginfo(&info);
1761 	info.si_signo = SIGBUS;
1762 	info.si_errno = 0;
1763 	info.si_code = code;
1764 	info.si_addr = addr;
1765 	info.si_addr_lsb = lsb;
1766 	return force_sig_info(&info);
1767 }
1768 
1769 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1770 {
1771 	struct kernel_siginfo info;
1772 
1773 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1774 	clear_siginfo(&info);
1775 	info.si_signo = SIGBUS;
1776 	info.si_errno = 0;
1777 	info.si_code = code;
1778 	info.si_addr = addr;
1779 	info.si_addr_lsb = lsb;
1780 	return send_sig_info(info.si_signo, &info, t);
1781 }
1782 EXPORT_SYMBOL(send_sig_mceerr);
1783 
1784 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1785 {
1786 	struct kernel_siginfo info;
1787 
1788 	clear_siginfo(&info);
1789 	info.si_signo = SIGSEGV;
1790 	info.si_errno = 0;
1791 	info.si_code  = SEGV_BNDERR;
1792 	info.si_addr  = addr;
1793 	info.si_lower = lower;
1794 	info.si_upper = upper;
1795 	return force_sig_info(&info);
1796 }
1797 
1798 #ifdef SEGV_PKUERR
1799 int force_sig_pkuerr(void __user *addr, u32 pkey)
1800 {
1801 	struct kernel_siginfo info;
1802 
1803 	clear_siginfo(&info);
1804 	info.si_signo = SIGSEGV;
1805 	info.si_errno = 0;
1806 	info.si_code  = SEGV_PKUERR;
1807 	info.si_addr  = addr;
1808 	info.si_pkey  = pkey;
1809 	return force_sig_info(&info);
1810 }
1811 #endif
1812 
1813 int send_sig_perf(void __user *addr, u32 type, u64 sig_data)
1814 {
1815 	struct kernel_siginfo info;
1816 
1817 	clear_siginfo(&info);
1818 	info.si_signo     = SIGTRAP;
1819 	info.si_errno     = 0;
1820 	info.si_code      = TRAP_PERF;
1821 	info.si_addr      = addr;
1822 	info.si_perf_data = sig_data;
1823 	info.si_perf_type = type;
1824 
1825 	/*
1826 	 * Signals generated by perf events should not terminate the whole
1827 	 * process if SIGTRAP is blocked, however, delivering the signal
1828 	 * asynchronously is better than not delivering at all. But tell user
1829 	 * space if the signal was asynchronous, so it can clearly be
1830 	 * distinguished from normal synchronous ones.
1831 	 */
1832 	info.si_perf_flags = sigismember(&current->blocked, info.si_signo) ?
1833 				     TRAP_PERF_FLAG_ASYNC :
1834 				     0;
1835 
1836 	return send_sig_info(info.si_signo, &info, current);
1837 }
1838 
1839 /**
1840  * force_sig_seccomp - signals the task to allow in-process syscall emulation
1841  * @syscall: syscall number to send to userland
1842  * @reason: filter-supplied reason code to send to userland (via si_errno)
1843  * @force_coredump: true to trigger a coredump
1844  *
1845  * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
1846  */
1847 int force_sig_seccomp(int syscall, int reason, bool force_coredump)
1848 {
1849 	struct kernel_siginfo info;
1850 
1851 	clear_siginfo(&info);
1852 	info.si_signo = SIGSYS;
1853 	info.si_code = SYS_SECCOMP;
1854 	info.si_call_addr = (void __user *)KSTK_EIP(current);
1855 	info.si_errno = reason;
1856 	info.si_arch = syscall_get_arch(current);
1857 	info.si_syscall = syscall;
1858 	return force_sig_info_to_task(&info, current,
1859 		force_coredump ? HANDLER_EXIT : HANDLER_CURRENT);
1860 }
1861 
1862 /* For the crazy architectures that include trap information in
1863  * the errno field, instead of an actual errno value.
1864  */
1865 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1866 {
1867 	struct kernel_siginfo info;
1868 
1869 	clear_siginfo(&info);
1870 	info.si_signo = SIGTRAP;
1871 	info.si_errno = errno;
1872 	info.si_code  = TRAP_HWBKPT;
1873 	info.si_addr  = addr;
1874 	return force_sig_info(&info);
1875 }
1876 
1877 /* For the rare architectures that include trap information using
1878  * si_trapno.
1879  */
1880 int force_sig_fault_trapno(int sig, int code, void __user *addr, int trapno)
1881 {
1882 	struct kernel_siginfo info;
1883 
1884 	clear_siginfo(&info);
1885 	info.si_signo = sig;
1886 	info.si_errno = 0;
1887 	info.si_code  = code;
1888 	info.si_addr  = addr;
1889 	info.si_trapno = trapno;
1890 	return force_sig_info(&info);
1891 }
1892 
1893 /* For the rare architectures that include trap information using
1894  * si_trapno.
1895  */
1896 int send_sig_fault_trapno(int sig, int code, void __user *addr, int trapno,
1897 			  struct task_struct *t)
1898 {
1899 	struct kernel_siginfo info;
1900 
1901 	clear_siginfo(&info);
1902 	info.si_signo = sig;
1903 	info.si_errno = 0;
1904 	info.si_code  = code;
1905 	info.si_addr  = addr;
1906 	info.si_trapno = trapno;
1907 	return send_sig_info(info.si_signo, &info, t);
1908 }
1909 
1910 int kill_pgrp(struct pid *pid, int sig, int priv)
1911 {
1912 	int ret;
1913 
1914 	read_lock(&tasklist_lock);
1915 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1916 	read_unlock(&tasklist_lock);
1917 
1918 	return ret;
1919 }
1920 EXPORT_SYMBOL(kill_pgrp);
1921 
1922 int kill_pid(struct pid *pid, int sig, int priv)
1923 {
1924 	return kill_pid_info(sig, __si_special(priv), pid);
1925 }
1926 EXPORT_SYMBOL(kill_pid);
1927 
1928 /*
1929  * These functions support sending signals using preallocated sigqueue
1930  * structures.  This is needed "because realtime applications cannot
1931  * afford to lose notifications of asynchronous events, like timer
1932  * expirations or I/O completions".  In the case of POSIX Timers
1933  * we allocate the sigqueue structure from the timer_create.  If this
1934  * allocation fails we are able to report the failure to the application
1935  * with an EAGAIN error.
1936  */
1937 struct sigqueue *sigqueue_alloc(void)
1938 {
1939 	return __sigqueue_alloc(-1, current, GFP_KERNEL, 0, SIGQUEUE_PREALLOC);
1940 }
1941 
1942 void sigqueue_free(struct sigqueue *q)
1943 {
1944 	unsigned long flags;
1945 	spinlock_t *lock = &current->sighand->siglock;
1946 
1947 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1948 	/*
1949 	 * We must hold ->siglock while testing q->list
1950 	 * to serialize with collect_signal() or with
1951 	 * __exit_signal()->flush_sigqueue().
1952 	 */
1953 	spin_lock_irqsave(lock, flags);
1954 	q->flags &= ~SIGQUEUE_PREALLOC;
1955 	/*
1956 	 * If it is queued it will be freed when dequeued,
1957 	 * like the "regular" sigqueue.
1958 	 */
1959 	if (!list_empty(&q->list))
1960 		q = NULL;
1961 	spin_unlock_irqrestore(lock, flags);
1962 
1963 	if (q)
1964 		__sigqueue_free(q);
1965 }
1966 
1967 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1968 {
1969 	int sig = q->info.si_signo;
1970 	struct sigpending *pending;
1971 	struct task_struct *t;
1972 	unsigned long flags;
1973 	int ret, result;
1974 
1975 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1976 
1977 	ret = -1;
1978 	rcu_read_lock();
1979 
1980 	/*
1981 	 * This function is used by POSIX timers to deliver a timer signal.
1982 	 * Where type is PIDTYPE_PID (such as for timers with SIGEV_THREAD_ID
1983 	 * set), the signal must be delivered to the specific thread (queues
1984 	 * into t->pending).
1985 	 *
1986 	 * Where type is not PIDTYPE_PID, signals must be delivered to the
1987 	 * process. In this case, prefer to deliver to current if it is in
1988 	 * the same thread group as the target process, which avoids
1989 	 * unnecessarily waking up a potentially idle task.
1990 	 */
1991 	t = pid_task(pid, type);
1992 	if (!t)
1993 		goto ret;
1994 	if (type != PIDTYPE_PID && same_thread_group(t, current))
1995 		t = current;
1996 	if (!likely(lock_task_sighand(t, &flags)))
1997 		goto ret;
1998 
1999 	ret = 1; /* the signal is ignored */
2000 	result = TRACE_SIGNAL_IGNORED;
2001 	if (!prepare_signal(sig, t, false))
2002 		goto out;
2003 
2004 	ret = 0;
2005 	if (unlikely(!list_empty(&q->list))) {
2006 		/*
2007 		 * If an SI_TIMER entry is already queue just increment
2008 		 * the overrun count.
2009 		 */
2010 		BUG_ON(q->info.si_code != SI_TIMER);
2011 		q->info.si_overrun++;
2012 		result = TRACE_SIGNAL_ALREADY_PENDING;
2013 		goto out;
2014 	}
2015 	q->info.si_overrun = 0;
2016 
2017 	signalfd_notify(t, sig);
2018 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
2019 	list_add_tail(&q->list, &pending->list);
2020 	sigaddset(&pending->signal, sig);
2021 	complete_signal(sig, t, type);
2022 	result = TRACE_SIGNAL_DELIVERED;
2023 out:
2024 	trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
2025 	unlock_task_sighand(t, &flags);
2026 ret:
2027 	rcu_read_unlock();
2028 	return ret;
2029 }
2030 
2031 static void do_notify_pidfd(struct task_struct *task)
2032 {
2033 	struct pid *pid;
2034 
2035 	WARN_ON(task->exit_state == 0);
2036 	pid = task_pid(task);
2037 	wake_up_all(&pid->wait_pidfd);
2038 }
2039 
2040 /*
2041  * Let a parent know about the death of a child.
2042  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
2043  *
2044  * Returns true if our parent ignored us and so we've switched to
2045  * self-reaping.
2046  */
2047 bool do_notify_parent(struct task_struct *tsk, int sig)
2048 {
2049 	struct kernel_siginfo info;
2050 	unsigned long flags;
2051 	struct sighand_struct *psig;
2052 	bool autoreap = false;
2053 	u64 utime, stime;
2054 
2055 	WARN_ON_ONCE(sig == -1);
2056 
2057 	/* do_notify_parent_cldstop should have been called instead.  */
2058 	WARN_ON_ONCE(task_is_stopped_or_traced(tsk));
2059 
2060 	WARN_ON_ONCE(!tsk->ptrace &&
2061 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
2062 
2063 	/* Wake up all pidfd waiters */
2064 	do_notify_pidfd(tsk);
2065 
2066 	if (sig != SIGCHLD) {
2067 		/*
2068 		 * This is only possible if parent == real_parent.
2069 		 * Check if it has changed security domain.
2070 		 */
2071 		if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
2072 			sig = SIGCHLD;
2073 	}
2074 
2075 	clear_siginfo(&info);
2076 	info.si_signo = sig;
2077 	info.si_errno = 0;
2078 	/*
2079 	 * We are under tasklist_lock here so our parent is tied to
2080 	 * us and cannot change.
2081 	 *
2082 	 * task_active_pid_ns will always return the same pid namespace
2083 	 * until a task passes through release_task.
2084 	 *
2085 	 * write_lock() currently calls preempt_disable() which is the
2086 	 * same as rcu_read_lock(), but according to Oleg, this is not
2087 	 * correct to rely on this
2088 	 */
2089 	rcu_read_lock();
2090 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
2091 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
2092 				       task_uid(tsk));
2093 	rcu_read_unlock();
2094 
2095 	task_cputime(tsk, &utime, &stime);
2096 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
2097 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
2098 
2099 	info.si_status = tsk->exit_code & 0x7f;
2100 	if (tsk->exit_code & 0x80)
2101 		info.si_code = CLD_DUMPED;
2102 	else if (tsk->exit_code & 0x7f)
2103 		info.si_code = CLD_KILLED;
2104 	else {
2105 		info.si_code = CLD_EXITED;
2106 		info.si_status = tsk->exit_code >> 8;
2107 	}
2108 
2109 	psig = tsk->parent->sighand;
2110 	spin_lock_irqsave(&psig->siglock, flags);
2111 	if (!tsk->ptrace && sig == SIGCHLD &&
2112 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
2113 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
2114 		/*
2115 		 * We are exiting and our parent doesn't care.  POSIX.1
2116 		 * defines special semantics for setting SIGCHLD to SIG_IGN
2117 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
2118 		 * automatically and not left for our parent's wait4 call.
2119 		 * Rather than having the parent do it as a magic kind of
2120 		 * signal handler, we just set this to tell do_exit that we
2121 		 * can be cleaned up without becoming a zombie.  Note that
2122 		 * we still call __wake_up_parent in this case, because a
2123 		 * blocked sys_wait4 might now return -ECHILD.
2124 		 *
2125 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
2126 		 * is implementation-defined: we do (if you don't want
2127 		 * it, just use SIG_IGN instead).
2128 		 */
2129 		autoreap = true;
2130 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2131 			sig = 0;
2132 	}
2133 	/*
2134 	 * Send with __send_signal as si_pid and si_uid are in the
2135 	 * parent's namespaces.
2136 	 */
2137 	if (valid_signal(sig) && sig)
2138 		__send_signal_locked(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2139 	__wake_up_parent(tsk, tsk->parent);
2140 	spin_unlock_irqrestore(&psig->siglock, flags);
2141 
2142 	return autoreap;
2143 }
2144 
2145 /**
2146  * do_notify_parent_cldstop - notify parent of stopped/continued state change
2147  * @tsk: task reporting the state change
2148  * @for_ptracer: the notification is for ptracer
2149  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2150  *
2151  * Notify @tsk's parent that the stopped/continued state has changed.  If
2152  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2153  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2154  *
2155  * CONTEXT:
2156  * Must be called with tasklist_lock at least read locked.
2157  */
2158 static void do_notify_parent_cldstop(struct task_struct *tsk,
2159 				     bool for_ptracer, int why)
2160 {
2161 	struct kernel_siginfo info;
2162 	unsigned long flags;
2163 	struct task_struct *parent;
2164 	struct sighand_struct *sighand;
2165 	u64 utime, stime;
2166 
2167 	if (for_ptracer) {
2168 		parent = tsk->parent;
2169 	} else {
2170 		tsk = tsk->group_leader;
2171 		parent = tsk->real_parent;
2172 	}
2173 
2174 	clear_siginfo(&info);
2175 	info.si_signo = SIGCHLD;
2176 	info.si_errno = 0;
2177 	/*
2178 	 * see comment in do_notify_parent() about the following 4 lines
2179 	 */
2180 	rcu_read_lock();
2181 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2182 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2183 	rcu_read_unlock();
2184 
2185 	task_cputime(tsk, &utime, &stime);
2186 	info.si_utime = nsec_to_clock_t(utime);
2187 	info.si_stime = nsec_to_clock_t(stime);
2188 
2189  	info.si_code = why;
2190  	switch (why) {
2191  	case CLD_CONTINUED:
2192  		info.si_status = SIGCONT;
2193  		break;
2194  	case CLD_STOPPED:
2195  		info.si_status = tsk->signal->group_exit_code & 0x7f;
2196  		break;
2197  	case CLD_TRAPPED:
2198  		info.si_status = tsk->exit_code & 0x7f;
2199  		break;
2200  	default:
2201  		BUG();
2202  	}
2203 
2204 	sighand = parent->sighand;
2205 	spin_lock_irqsave(&sighand->siglock, flags);
2206 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2207 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2208 		send_signal_locked(SIGCHLD, &info, parent, PIDTYPE_TGID);
2209 	/*
2210 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2211 	 */
2212 	__wake_up_parent(tsk, parent);
2213 	spin_unlock_irqrestore(&sighand->siglock, flags);
2214 }
2215 
2216 /*
2217  * This must be called with current->sighand->siglock held.
2218  *
2219  * This should be the path for all ptrace stops.
2220  * We always set current->last_siginfo while stopped here.
2221  * That makes it a way to test a stopped process for
2222  * being ptrace-stopped vs being job-control-stopped.
2223  *
2224  * Returns the signal the ptracer requested the code resume
2225  * with.  If the code did not stop because the tracer is gone,
2226  * the stop signal remains unchanged unless clear_code.
2227  */
2228 static int ptrace_stop(int exit_code, int why, unsigned long message,
2229 		       kernel_siginfo_t *info)
2230 	__releases(&current->sighand->siglock)
2231 	__acquires(&current->sighand->siglock)
2232 {
2233 	bool gstop_done = false;
2234 
2235 	if (arch_ptrace_stop_needed()) {
2236 		/*
2237 		 * The arch code has something special to do before a
2238 		 * ptrace stop.  This is allowed to block, e.g. for faults
2239 		 * on user stack pages.  We can't keep the siglock while
2240 		 * calling arch_ptrace_stop, so we must release it now.
2241 		 * To preserve proper semantics, we must do this before
2242 		 * any signal bookkeeping like checking group_stop_count.
2243 		 */
2244 		spin_unlock_irq(&current->sighand->siglock);
2245 		arch_ptrace_stop();
2246 		spin_lock_irq(&current->sighand->siglock);
2247 	}
2248 
2249 	/*
2250 	 * After this point ptrace_signal_wake_up or signal_wake_up
2251 	 * will clear TASK_TRACED if ptrace_unlink happens or a fatal
2252 	 * signal comes in.  Handle previous ptrace_unlinks and fatal
2253 	 * signals here to prevent ptrace_stop sleeping in schedule.
2254 	 */
2255 	if (!current->ptrace || __fatal_signal_pending(current))
2256 		return exit_code;
2257 
2258 	set_special_state(TASK_TRACED);
2259 	current->jobctl |= JOBCTL_TRACED;
2260 
2261 	/*
2262 	 * We're committing to trapping.  TRACED should be visible before
2263 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2264 	 * Also, transition to TRACED and updates to ->jobctl should be
2265 	 * atomic with respect to siglock and should be done after the arch
2266 	 * hook as siglock is released and regrabbed across it.
2267 	 *
2268 	 *     TRACER				    TRACEE
2269 	 *
2270 	 *     ptrace_attach()
2271 	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2272 	 *     do_wait()
2273 	 *       set_current_state()                smp_wmb();
2274 	 *       ptrace_do_wait()
2275 	 *         wait_task_stopped()
2276 	 *           task_stopped_code()
2277 	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2278 	 */
2279 	smp_wmb();
2280 
2281 	current->ptrace_message = message;
2282 	current->last_siginfo = info;
2283 	current->exit_code = exit_code;
2284 
2285 	/*
2286 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2287 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2288 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2289 	 * could be clear now.  We act as if SIGCONT is received after
2290 	 * TASK_TRACED is entered - ignore it.
2291 	 */
2292 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2293 		gstop_done = task_participate_group_stop(current);
2294 
2295 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2296 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2297 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2298 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2299 
2300 	/* entering a trap, clear TRAPPING */
2301 	task_clear_jobctl_trapping(current);
2302 
2303 	spin_unlock_irq(&current->sighand->siglock);
2304 	read_lock(&tasklist_lock);
2305 	/*
2306 	 * Notify parents of the stop.
2307 	 *
2308 	 * While ptraced, there are two parents - the ptracer and
2309 	 * the real_parent of the group_leader.  The ptracer should
2310 	 * know about every stop while the real parent is only
2311 	 * interested in the completion of group stop.  The states
2312 	 * for the two don't interact with each other.  Notify
2313 	 * separately unless they're gonna be duplicates.
2314 	 */
2315 	if (current->ptrace)
2316 		do_notify_parent_cldstop(current, true, why);
2317 	if (gstop_done && (!current->ptrace || ptrace_reparented(current)))
2318 		do_notify_parent_cldstop(current, false, why);
2319 
2320 	/*
2321 	 * Don't want to allow preemption here, because
2322 	 * sys_ptrace() needs this task to be inactive.
2323 	 *
2324 	 * XXX: implement read_unlock_no_resched().
2325 	 */
2326 	preempt_disable();
2327 	read_unlock(&tasklist_lock);
2328 	cgroup_enter_frozen();
2329 	preempt_enable_no_resched();
2330 	schedule();
2331 	cgroup_leave_frozen(true);
2332 
2333 	/*
2334 	 * We are back.  Now reacquire the siglock before touching
2335 	 * last_siginfo, so that we are sure to have synchronized with
2336 	 * any signal-sending on another CPU that wants to examine it.
2337 	 */
2338 	spin_lock_irq(&current->sighand->siglock);
2339 	exit_code = current->exit_code;
2340 	current->last_siginfo = NULL;
2341 	current->ptrace_message = 0;
2342 	current->exit_code = 0;
2343 
2344 	/* LISTENING can be set only during STOP traps, clear it */
2345 	current->jobctl &= ~(JOBCTL_LISTENING | JOBCTL_PTRACE_FROZEN);
2346 
2347 	/*
2348 	 * Queued signals ignored us while we were stopped for tracing.
2349 	 * So check for any that we should take before resuming user mode.
2350 	 * This sets TIF_SIGPENDING, but never clears it.
2351 	 */
2352 	recalc_sigpending_tsk(current);
2353 	return exit_code;
2354 }
2355 
2356 static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
2357 {
2358 	kernel_siginfo_t info;
2359 
2360 	clear_siginfo(&info);
2361 	info.si_signo = signr;
2362 	info.si_code = exit_code;
2363 	info.si_pid = task_pid_vnr(current);
2364 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2365 
2366 	/* Let the debugger run.  */
2367 	return ptrace_stop(exit_code, why, message, &info);
2368 }
2369 
2370 int ptrace_notify(int exit_code, unsigned long message)
2371 {
2372 	int signr;
2373 
2374 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2375 	if (unlikely(task_work_pending(current)))
2376 		task_work_run();
2377 
2378 	spin_lock_irq(&current->sighand->siglock);
2379 	signr = ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
2380 	spin_unlock_irq(&current->sighand->siglock);
2381 	return signr;
2382 }
2383 
2384 /**
2385  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2386  * @signr: signr causing group stop if initiating
2387  *
2388  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2389  * and participate in it.  If already set, participate in the existing
2390  * group stop.  If participated in a group stop (and thus slept), %true is
2391  * returned with siglock released.
2392  *
2393  * If ptraced, this function doesn't handle stop itself.  Instead,
2394  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2395  * untouched.  The caller must ensure that INTERRUPT trap handling takes
2396  * places afterwards.
2397  *
2398  * CONTEXT:
2399  * Must be called with @current->sighand->siglock held, which is released
2400  * on %true return.
2401  *
2402  * RETURNS:
2403  * %false if group stop is already cancelled or ptrace trap is scheduled.
2404  * %true if participated in group stop.
2405  */
2406 static bool do_signal_stop(int signr)
2407 	__releases(&current->sighand->siglock)
2408 {
2409 	struct signal_struct *sig = current->signal;
2410 
2411 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2412 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2413 		struct task_struct *t;
2414 
2415 		/* signr will be recorded in task->jobctl for retries */
2416 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2417 
2418 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2419 		    unlikely(sig->flags & SIGNAL_GROUP_EXIT) ||
2420 		    unlikely(sig->group_exec_task))
2421 			return false;
2422 		/*
2423 		 * There is no group stop already in progress.  We must
2424 		 * initiate one now.
2425 		 *
2426 		 * While ptraced, a task may be resumed while group stop is
2427 		 * still in effect and then receive a stop signal and
2428 		 * initiate another group stop.  This deviates from the
2429 		 * usual behavior as two consecutive stop signals can't
2430 		 * cause two group stops when !ptraced.  That is why we
2431 		 * also check !task_is_stopped(t) below.
2432 		 *
2433 		 * The condition can be distinguished by testing whether
2434 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2435 		 * group_exit_code in such case.
2436 		 *
2437 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2438 		 * an intervening stop signal is required to cause two
2439 		 * continued events regardless of ptrace.
2440 		 */
2441 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2442 			sig->group_exit_code = signr;
2443 
2444 		sig->group_stop_count = 0;
2445 
2446 		if (task_set_jobctl_pending(current, signr | gstop))
2447 			sig->group_stop_count++;
2448 
2449 		t = current;
2450 		while_each_thread(current, t) {
2451 			/*
2452 			 * Setting state to TASK_STOPPED for a group
2453 			 * stop is always done with the siglock held,
2454 			 * so this check has no races.
2455 			 */
2456 			if (!task_is_stopped(t) &&
2457 			    task_set_jobctl_pending(t, signr | gstop)) {
2458 				sig->group_stop_count++;
2459 				if (likely(!(t->ptrace & PT_SEIZED)))
2460 					signal_wake_up(t, 0);
2461 				else
2462 					ptrace_trap_notify(t);
2463 			}
2464 		}
2465 	}
2466 
2467 	if (likely(!current->ptrace)) {
2468 		int notify = 0;
2469 
2470 		/*
2471 		 * If there are no other threads in the group, or if there
2472 		 * is a group stop in progress and we are the last to stop,
2473 		 * report to the parent.
2474 		 */
2475 		if (task_participate_group_stop(current))
2476 			notify = CLD_STOPPED;
2477 
2478 		current->jobctl |= JOBCTL_STOPPED;
2479 		set_special_state(TASK_STOPPED);
2480 		spin_unlock_irq(&current->sighand->siglock);
2481 
2482 		/*
2483 		 * Notify the parent of the group stop completion.  Because
2484 		 * we're not holding either the siglock or tasklist_lock
2485 		 * here, ptracer may attach inbetween; however, this is for
2486 		 * group stop and should always be delivered to the real
2487 		 * parent of the group leader.  The new ptracer will get
2488 		 * its notification when this task transitions into
2489 		 * TASK_TRACED.
2490 		 */
2491 		if (notify) {
2492 			read_lock(&tasklist_lock);
2493 			do_notify_parent_cldstop(current, false, notify);
2494 			read_unlock(&tasklist_lock);
2495 		}
2496 
2497 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2498 		cgroup_enter_frozen();
2499 		schedule();
2500 		return true;
2501 	} else {
2502 		/*
2503 		 * While ptraced, group stop is handled by STOP trap.
2504 		 * Schedule it and let the caller deal with it.
2505 		 */
2506 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2507 		return false;
2508 	}
2509 }
2510 
2511 /**
2512  * do_jobctl_trap - take care of ptrace jobctl traps
2513  *
2514  * When PT_SEIZED, it's used for both group stop and explicit
2515  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2516  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2517  * the stop signal; otherwise, %SIGTRAP.
2518  *
2519  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2520  * number as exit_code and no siginfo.
2521  *
2522  * CONTEXT:
2523  * Must be called with @current->sighand->siglock held, which may be
2524  * released and re-acquired before returning with intervening sleep.
2525  */
2526 static void do_jobctl_trap(void)
2527 {
2528 	struct signal_struct *signal = current->signal;
2529 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2530 
2531 	if (current->ptrace & PT_SEIZED) {
2532 		if (!signal->group_stop_count &&
2533 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2534 			signr = SIGTRAP;
2535 		WARN_ON_ONCE(!signr);
2536 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2537 				 CLD_STOPPED, 0);
2538 	} else {
2539 		WARN_ON_ONCE(!signr);
2540 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2541 	}
2542 }
2543 
2544 /**
2545  * do_freezer_trap - handle the freezer jobctl trap
2546  *
2547  * Puts the task into frozen state, if only the task is not about to quit.
2548  * In this case it drops JOBCTL_TRAP_FREEZE.
2549  *
2550  * CONTEXT:
2551  * Must be called with @current->sighand->siglock held,
2552  * which is always released before returning.
2553  */
2554 static void do_freezer_trap(void)
2555 	__releases(&current->sighand->siglock)
2556 {
2557 	/*
2558 	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2559 	 * let's make another loop to give it a chance to be handled.
2560 	 * In any case, we'll return back.
2561 	 */
2562 	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2563 	     JOBCTL_TRAP_FREEZE) {
2564 		spin_unlock_irq(&current->sighand->siglock);
2565 		return;
2566 	}
2567 
2568 	/*
2569 	 * Now we're sure that there is no pending fatal signal and no
2570 	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2571 	 * immediately (if there is a non-fatal signal pending), and
2572 	 * put the task into sleep.
2573 	 */
2574 	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
2575 	clear_thread_flag(TIF_SIGPENDING);
2576 	spin_unlock_irq(&current->sighand->siglock);
2577 	cgroup_enter_frozen();
2578 	schedule();
2579 }
2580 
2581 static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
2582 {
2583 	/*
2584 	 * We do not check sig_kernel_stop(signr) but set this marker
2585 	 * unconditionally because we do not know whether debugger will
2586 	 * change signr. This flag has no meaning unless we are going
2587 	 * to stop after return from ptrace_stop(). In this case it will
2588 	 * be checked in do_signal_stop(), we should only stop if it was
2589 	 * not cleared by SIGCONT while we were sleeping. See also the
2590 	 * comment in dequeue_signal().
2591 	 */
2592 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2593 	signr = ptrace_stop(signr, CLD_TRAPPED, 0, info);
2594 
2595 	/* We're back.  Did the debugger cancel the sig?  */
2596 	if (signr == 0)
2597 		return signr;
2598 
2599 	/*
2600 	 * Update the siginfo structure if the signal has
2601 	 * changed.  If the debugger wanted something
2602 	 * specific in the siginfo structure then it should
2603 	 * have updated *info via PTRACE_SETSIGINFO.
2604 	 */
2605 	if (signr != info->si_signo) {
2606 		clear_siginfo(info);
2607 		info->si_signo = signr;
2608 		info->si_errno = 0;
2609 		info->si_code = SI_USER;
2610 		rcu_read_lock();
2611 		info->si_pid = task_pid_vnr(current->parent);
2612 		info->si_uid = from_kuid_munged(current_user_ns(),
2613 						task_uid(current->parent));
2614 		rcu_read_unlock();
2615 	}
2616 
2617 	/* If the (new) signal is now blocked, requeue it.  */
2618 	if (sigismember(&current->blocked, signr) ||
2619 	    fatal_signal_pending(current)) {
2620 		send_signal_locked(signr, info, current, type);
2621 		signr = 0;
2622 	}
2623 
2624 	return signr;
2625 }
2626 
2627 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2628 {
2629 	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2630 	case SIL_FAULT:
2631 	case SIL_FAULT_TRAPNO:
2632 	case SIL_FAULT_MCEERR:
2633 	case SIL_FAULT_BNDERR:
2634 	case SIL_FAULT_PKUERR:
2635 	case SIL_FAULT_PERF_EVENT:
2636 		ksig->info.si_addr = arch_untagged_si_addr(
2637 			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2638 		break;
2639 	case SIL_KILL:
2640 	case SIL_TIMER:
2641 	case SIL_POLL:
2642 	case SIL_CHLD:
2643 	case SIL_RT:
2644 	case SIL_SYS:
2645 		break;
2646 	}
2647 }
2648 
2649 bool get_signal(struct ksignal *ksig)
2650 {
2651 	struct sighand_struct *sighand = current->sighand;
2652 	struct signal_struct *signal = current->signal;
2653 	int signr;
2654 
2655 	clear_notify_signal();
2656 	if (unlikely(task_work_pending(current)))
2657 		task_work_run();
2658 
2659 	if (!task_sigpending(current))
2660 		return false;
2661 
2662 	if (unlikely(uprobe_deny_signal()))
2663 		return false;
2664 
2665 	/*
2666 	 * Do this once, we can't return to user-mode if freezing() == T.
2667 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2668 	 * thus do not need another check after return.
2669 	 */
2670 	try_to_freeze();
2671 
2672 relock:
2673 	spin_lock_irq(&sighand->siglock);
2674 
2675 	/*
2676 	 * Every stopped thread goes here after wakeup. Check to see if
2677 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2678 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2679 	 */
2680 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2681 		int why;
2682 
2683 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2684 			why = CLD_CONTINUED;
2685 		else
2686 			why = CLD_STOPPED;
2687 
2688 		signal->flags &= ~SIGNAL_CLD_MASK;
2689 
2690 		spin_unlock_irq(&sighand->siglock);
2691 
2692 		/*
2693 		 * Notify the parent that we're continuing.  This event is
2694 		 * always per-process and doesn't make whole lot of sense
2695 		 * for ptracers, who shouldn't consume the state via
2696 		 * wait(2) either, but, for backward compatibility, notify
2697 		 * the ptracer of the group leader too unless it's gonna be
2698 		 * a duplicate.
2699 		 */
2700 		read_lock(&tasklist_lock);
2701 		do_notify_parent_cldstop(current, false, why);
2702 
2703 		if (ptrace_reparented(current->group_leader))
2704 			do_notify_parent_cldstop(current->group_leader,
2705 						true, why);
2706 		read_unlock(&tasklist_lock);
2707 
2708 		goto relock;
2709 	}
2710 
2711 	for (;;) {
2712 		struct k_sigaction *ka;
2713 		enum pid_type type;
2714 
2715 		/* Has this task already been marked for death? */
2716 		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
2717 		     signal->group_exec_task) {
2718 			clear_siginfo(&ksig->info);
2719 			ksig->info.si_signo = signr = SIGKILL;
2720 			sigdelset(&current->pending.signal, SIGKILL);
2721 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2722 				&sighand->action[SIGKILL - 1]);
2723 			recalc_sigpending();
2724 			goto fatal;
2725 		}
2726 
2727 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2728 		    do_signal_stop(0))
2729 			goto relock;
2730 
2731 		if (unlikely(current->jobctl &
2732 			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2733 			if (current->jobctl & JOBCTL_TRAP_MASK) {
2734 				do_jobctl_trap();
2735 				spin_unlock_irq(&sighand->siglock);
2736 			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2737 				do_freezer_trap();
2738 
2739 			goto relock;
2740 		}
2741 
2742 		/*
2743 		 * If the task is leaving the frozen state, let's update
2744 		 * cgroup counters and reset the frozen bit.
2745 		 */
2746 		if (unlikely(cgroup_task_frozen(current))) {
2747 			spin_unlock_irq(&sighand->siglock);
2748 			cgroup_leave_frozen(false);
2749 			goto relock;
2750 		}
2751 
2752 		/*
2753 		 * Signals generated by the execution of an instruction
2754 		 * need to be delivered before any other pending signals
2755 		 * so that the instruction pointer in the signal stack
2756 		 * frame points to the faulting instruction.
2757 		 */
2758 		type = PIDTYPE_PID;
2759 		signr = dequeue_synchronous_signal(&ksig->info);
2760 		if (!signr)
2761 			signr = dequeue_signal(current, &current->blocked,
2762 					       &ksig->info, &type);
2763 
2764 		if (!signr)
2765 			break; /* will return 0 */
2766 
2767 		if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2768 		    !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2769 			signr = ptrace_signal(signr, &ksig->info, type);
2770 			if (!signr)
2771 				continue;
2772 		}
2773 
2774 		ka = &sighand->action[signr-1];
2775 
2776 		/* Trace actually delivered signals. */
2777 		trace_signal_deliver(signr, &ksig->info, ka);
2778 
2779 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2780 			continue;
2781 		if (ka->sa.sa_handler != SIG_DFL) {
2782 			/* Run the handler.  */
2783 			ksig->ka = *ka;
2784 
2785 			if (ka->sa.sa_flags & SA_ONESHOT)
2786 				ka->sa.sa_handler = SIG_DFL;
2787 
2788 			break; /* will return non-zero "signr" value */
2789 		}
2790 
2791 		/*
2792 		 * Now we are doing the default action for this signal.
2793 		 */
2794 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2795 			continue;
2796 
2797 		/*
2798 		 * Global init gets no signals it doesn't want.
2799 		 * Container-init gets no signals it doesn't want from same
2800 		 * container.
2801 		 *
2802 		 * Note that if global/container-init sees a sig_kernel_only()
2803 		 * signal here, the signal must have been generated internally
2804 		 * or must have come from an ancestor namespace. In either
2805 		 * case, the signal cannot be dropped.
2806 		 */
2807 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2808 				!sig_kernel_only(signr))
2809 			continue;
2810 
2811 		if (sig_kernel_stop(signr)) {
2812 			/*
2813 			 * The default action is to stop all threads in
2814 			 * the thread group.  The job control signals
2815 			 * do nothing in an orphaned pgrp, but SIGSTOP
2816 			 * always works.  Note that siglock needs to be
2817 			 * dropped during the call to is_orphaned_pgrp()
2818 			 * because of lock ordering with tasklist_lock.
2819 			 * This allows an intervening SIGCONT to be posted.
2820 			 * We need to check for that and bail out if necessary.
2821 			 */
2822 			if (signr != SIGSTOP) {
2823 				spin_unlock_irq(&sighand->siglock);
2824 
2825 				/* signals can be posted during this window */
2826 
2827 				if (is_current_pgrp_orphaned())
2828 					goto relock;
2829 
2830 				spin_lock_irq(&sighand->siglock);
2831 			}
2832 
2833 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2834 				/* It released the siglock.  */
2835 				goto relock;
2836 			}
2837 
2838 			/*
2839 			 * We didn't actually stop, due to a race
2840 			 * with SIGCONT or something like that.
2841 			 */
2842 			continue;
2843 		}
2844 
2845 	fatal:
2846 		spin_unlock_irq(&sighand->siglock);
2847 		if (unlikely(cgroup_task_frozen(current)))
2848 			cgroup_leave_frozen(true);
2849 
2850 		/*
2851 		 * Anything else is fatal, maybe with a core dump.
2852 		 */
2853 		current->flags |= PF_SIGNALED;
2854 
2855 		if (sig_kernel_coredump(signr)) {
2856 			if (print_fatal_signals)
2857 				print_fatal_signal(ksig->info.si_signo);
2858 			proc_coredump_connector(current);
2859 			/*
2860 			 * If it was able to dump core, this kills all
2861 			 * other threads in the group and synchronizes with
2862 			 * their demise.  If we lost the race with another
2863 			 * thread getting here, it set group_exit_code
2864 			 * first and our do_group_exit call below will use
2865 			 * that value and ignore the one we pass it.
2866 			 */
2867 			do_coredump(&ksig->info);
2868 		}
2869 
2870 		/*
2871 		 * PF_USER_WORKER threads will catch and exit on fatal signals
2872 		 * themselves. They have cleanup that must be performed, so
2873 		 * we cannot call do_exit() on their behalf.
2874 		 */
2875 		if (current->flags & PF_USER_WORKER)
2876 			goto out;
2877 
2878 		/*
2879 		 * Death signals, no core dump.
2880 		 */
2881 		do_group_exit(ksig->info.si_signo);
2882 		/* NOTREACHED */
2883 	}
2884 	spin_unlock_irq(&sighand->siglock);
2885 out:
2886 	ksig->sig = signr;
2887 
2888 	if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2889 		hide_si_addr_tag_bits(ksig);
2890 
2891 	return ksig->sig > 0;
2892 }
2893 
2894 /**
2895  * signal_delivered - called after signal delivery to update blocked signals
2896  * @ksig:		kernel signal struct
2897  * @stepping:		nonzero if debugger single-step or block-step in use
2898  *
2899  * This function should be called when a signal has successfully been
2900  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2901  * is always blocked), and the signal itself is blocked unless %SA_NODEFER
2902  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2903  */
2904 static void signal_delivered(struct ksignal *ksig, int stepping)
2905 {
2906 	sigset_t blocked;
2907 
2908 	/* A signal was successfully delivered, and the
2909 	   saved sigmask was stored on the signal frame,
2910 	   and will be restored by sigreturn.  So we can
2911 	   simply clear the restore sigmask flag.  */
2912 	clear_restore_sigmask();
2913 
2914 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2915 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2916 		sigaddset(&blocked, ksig->sig);
2917 	set_current_blocked(&blocked);
2918 	if (current->sas_ss_flags & SS_AUTODISARM)
2919 		sas_ss_reset(current);
2920 	if (stepping)
2921 		ptrace_notify(SIGTRAP, 0);
2922 }
2923 
2924 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2925 {
2926 	if (failed)
2927 		force_sigsegv(ksig->sig);
2928 	else
2929 		signal_delivered(ksig, stepping);
2930 }
2931 
2932 /*
2933  * It could be that complete_signal() picked us to notify about the
2934  * group-wide signal. Other threads should be notified now to take
2935  * the shared signals in @which since we will not.
2936  */
2937 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2938 {
2939 	sigset_t retarget;
2940 	struct task_struct *t;
2941 
2942 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2943 	if (sigisemptyset(&retarget))
2944 		return;
2945 
2946 	t = tsk;
2947 	while_each_thread(tsk, t) {
2948 		if (t->flags & PF_EXITING)
2949 			continue;
2950 
2951 		if (!has_pending_signals(&retarget, &t->blocked))
2952 			continue;
2953 		/* Remove the signals this thread can handle. */
2954 		sigandsets(&retarget, &retarget, &t->blocked);
2955 
2956 		if (!task_sigpending(t))
2957 			signal_wake_up(t, 0);
2958 
2959 		if (sigisemptyset(&retarget))
2960 			break;
2961 	}
2962 }
2963 
2964 void exit_signals(struct task_struct *tsk)
2965 {
2966 	int group_stop = 0;
2967 	sigset_t unblocked;
2968 
2969 	/*
2970 	 * @tsk is about to have PF_EXITING set - lock out users which
2971 	 * expect stable threadgroup.
2972 	 */
2973 	cgroup_threadgroup_change_begin(tsk);
2974 
2975 	if (thread_group_empty(tsk) || (tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
2976 		sched_mm_cid_exit_signals(tsk);
2977 		tsk->flags |= PF_EXITING;
2978 		cgroup_threadgroup_change_end(tsk);
2979 		return;
2980 	}
2981 
2982 	spin_lock_irq(&tsk->sighand->siglock);
2983 	/*
2984 	 * From now this task is not visible for group-wide signals,
2985 	 * see wants_signal(), do_signal_stop().
2986 	 */
2987 	sched_mm_cid_exit_signals(tsk);
2988 	tsk->flags |= PF_EXITING;
2989 
2990 	cgroup_threadgroup_change_end(tsk);
2991 
2992 	if (!task_sigpending(tsk))
2993 		goto out;
2994 
2995 	unblocked = tsk->blocked;
2996 	signotset(&unblocked);
2997 	retarget_shared_pending(tsk, &unblocked);
2998 
2999 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
3000 	    task_participate_group_stop(tsk))
3001 		group_stop = CLD_STOPPED;
3002 out:
3003 	spin_unlock_irq(&tsk->sighand->siglock);
3004 
3005 	/*
3006 	 * If group stop has completed, deliver the notification.  This
3007 	 * should always go to the real parent of the group leader.
3008 	 */
3009 	if (unlikely(group_stop)) {
3010 		read_lock(&tasklist_lock);
3011 		do_notify_parent_cldstop(tsk, false, group_stop);
3012 		read_unlock(&tasklist_lock);
3013 	}
3014 }
3015 
3016 /*
3017  * System call entry points.
3018  */
3019 
3020 /**
3021  *  sys_restart_syscall - restart a system call
3022  */
3023 SYSCALL_DEFINE0(restart_syscall)
3024 {
3025 	struct restart_block *restart = &current->restart_block;
3026 	return restart->fn(restart);
3027 }
3028 
3029 long do_no_restart_syscall(struct restart_block *param)
3030 {
3031 	return -EINTR;
3032 }
3033 
3034 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
3035 {
3036 	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3037 		sigset_t newblocked;
3038 		/* A set of now blocked but previously unblocked signals. */
3039 		sigandnsets(&newblocked, newset, &current->blocked);
3040 		retarget_shared_pending(tsk, &newblocked);
3041 	}
3042 	tsk->blocked = *newset;
3043 	recalc_sigpending();
3044 }
3045 
3046 /**
3047  * set_current_blocked - change current->blocked mask
3048  * @newset: new mask
3049  *
3050  * It is wrong to change ->blocked directly, this helper should be used
3051  * to ensure the process can't miss a shared signal we are going to block.
3052  */
3053 void set_current_blocked(sigset_t *newset)
3054 {
3055 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3056 	__set_current_blocked(newset);
3057 }
3058 
3059 void __set_current_blocked(const sigset_t *newset)
3060 {
3061 	struct task_struct *tsk = current;
3062 
3063 	/*
3064 	 * In case the signal mask hasn't changed, there is nothing we need
3065 	 * to do. The current->blocked shouldn't be modified by other task.
3066 	 */
3067 	if (sigequalsets(&tsk->blocked, newset))
3068 		return;
3069 
3070 	spin_lock_irq(&tsk->sighand->siglock);
3071 	__set_task_blocked(tsk, newset);
3072 	spin_unlock_irq(&tsk->sighand->siglock);
3073 }
3074 
3075 /*
3076  * This is also useful for kernel threads that want to temporarily
3077  * (or permanently) block certain signals.
3078  *
3079  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3080  * interface happily blocks "unblockable" signals like SIGKILL
3081  * and friends.
3082  */
3083 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3084 {
3085 	struct task_struct *tsk = current;
3086 	sigset_t newset;
3087 
3088 	/* Lockless, only current can change ->blocked, never from irq */
3089 	if (oldset)
3090 		*oldset = tsk->blocked;
3091 
3092 	switch (how) {
3093 	case SIG_BLOCK:
3094 		sigorsets(&newset, &tsk->blocked, set);
3095 		break;
3096 	case SIG_UNBLOCK:
3097 		sigandnsets(&newset, &tsk->blocked, set);
3098 		break;
3099 	case SIG_SETMASK:
3100 		newset = *set;
3101 		break;
3102 	default:
3103 		return -EINVAL;
3104 	}
3105 
3106 	__set_current_blocked(&newset);
3107 	return 0;
3108 }
3109 EXPORT_SYMBOL(sigprocmask);
3110 
3111 /*
3112  * The api helps set app-provided sigmasks.
3113  *
3114  * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3115  * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3116  *
3117  * Note that it does set_restore_sigmask() in advance, so it must be always
3118  * paired with restore_saved_sigmask_unless() before return from syscall.
3119  */
3120 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3121 {
3122 	sigset_t kmask;
3123 
3124 	if (!umask)
3125 		return 0;
3126 	if (sigsetsize != sizeof(sigset_t))
3127 		return -EINVAL;
3128 	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3129 		return -EFAULT;
3130 
3131 	set_restore_sigmask();
3132 	current->saved_sigmask = current->blocked;
3133 	set_current_blocked(&kmask);
3134 
3135 	return 0;
3136 }
3137 
3138 #ifdef CONFIG_COMPAT
3139 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3140 			    size_t sigsetsize)
3141 {
3142 	sigset_t kmask;
3143 
3144 	if (!umask)
3145 		return 0;
3146 	if (sigsetsize != sizeof(compat_sigset_t))
3147 		return -EINVAL;
3148 	if (get_compat_sigset(&kmask, umask))
3149 		return -EFAULT;
3150 
3151 	set_restore_sigmask();
3152 	current->saved_sigmask = current->blocked;
3153 	set_current_blocked(&kmask);
3154 
3155 	return 0;
3156 }
3157 #endif
3158 
3159 /**
3160  *  sys_rt_sigprocmask - change the list of currently blocked signals
3161  *  @how: whether to add, remove, or set signals
3162  *  @nset: stores pending signals
3163  *  @oset: previous value of signal mask if non-null
3164  *  @sigsetsize: size of sigset_t type
3165  */
3166 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3167 		sigset_t __user *, oset, size_t, sigsetsize)
3168 {
3169 	sigset_t old_set, new_set;
3170 	int error;
3171 
3172 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3173 	if (sigsetsize != sizeof(sigset_t))
3174 		return -EINVAL;
3175 
3176 	old_set = current->blocked;
3177 
3178 	if (nset) {
3179 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3180 			return -EFAULT;
3181 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3182 
3183 		error = sigprocmask(how, &new_set, NULL);
3184 		if (error)
3185 			return error;
3186 	}
3187 
3188 	if (oset) {
3189 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3190 			return -EFAULT;
3191 	}
3192 
3193 	return 0;
3194 }
3195 
3196 #ifdef CONFIG_COMPAT
3197 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3198 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3199 {
3200 	sigset_t old_set = current->blocked;
3201 
3202 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3203 	if (sigsetsize != sizeof(sigset_t))
3204 		return -EINVAL;
3205 
3206 	if (nset) {
3207 		sigset_t new_set;
3208 		int error;
3209 		if (get_compat_sigset(&new_set, nset))
3210 			return -EFAULT;
3211 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3212 
3213 		error = sigprocmask(how, &new_set, NULL);
3214 		if (error)
3215 			return error;
3216 	}
3217 	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3218 }
3219 #endif
3220 
3221 static void do_sigpending(sigset_t *set)
3222 {
3223 	spin_lock_irq(&current->sighand->siglock);
3224 	sigorsets(set, &current->pending.signal,
3225 		  &current->signal->shared_pending.signal);
3226 	spin_unlock_irq(&current->sighand->siglock);
3227 
3228 	/* Outside the lock because only this thread touches it.  */
3229 	sigandsets(set, &current->blocked, set);
3230 }
3231 
3232 /**
3233  *  sys_rt_sigpending - examine a pending signal that has been raised
3234  *			while blocked
3235  *  @uset: stores pending signals
3236  *  @sigsetsize: size of sigset_t type or larger
3237  */
3238 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3239 {
3240 	sigset_t set;
3241 
3242 	if (sigsetsize > sizeof(*uset))
3243 		return -EINVAL;
3244 
3245 	do_sigpending(&set);
3246 
3247 	if (copy_to_user(uset, &set, sigsetsize))
3248 		return -EFAULT;
3249 
3250 	return 0;
3251 }
3252 
3253 #ifdef CONFIG_COMPAT
3254 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3255 		compat_size_t, sigsetsize)
3256 {
3257 	sigset_t set;
3258 
3259 	if (sigsetsize > sizeof(*uset))
3260 		return -EINVAL;
3261 
3262 	do_sigpending(&set);
3263 
3264 	return put_compat_sigset(uset, &set, sigsetsize);
3265 }
3266 #endif
3267 
3268 static const struct {
3269 	unsigned char limit, layout;
3270 } sig_sicodes[] = {
3271 	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3272 	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3273 	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3274 	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3275 	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3276 #if defined(SIGEMT)
3277 	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3278 #endif
3279 	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3280 	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3281 	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3282 };
3283 
3284 static bool known_siginfo_layout(unsigned sig, int si_code)
3285 {
3286 	if (si_code == SI_KERNEL)
3287 		return true;
3288 	else if ((si_code > SI_USER)) {
3289 		if (sig_specific_sicodes(sig)) {
3290 			if (si_code <= sig_sicodes[sig].limit)
3291 				return true;
3292 		}
3293 		else if (si_code <= NSIGPOLL)
3294 			return true;
3295 	}
3296 	else if (si_code >= SI_DETHREAD)
3297 		return true;
3298 	else if (si_code == SI_ASYNCNL)
3299 		return true;
3300 	return false;
3301 }
3302 
3303 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3304 {
3305 	enum siginfo_layout layout = SIL_KILL;
3306 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3307 		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3308 		    (si_code <= sig_sicodes[sig].limit)) {
3309 			layout = sig_sicodes[sig].layout;
3310 			/* Handle the exceptions */
3311 			if ((sig == SIGBUS) &&
3312 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3313 				layout = SIL_FAULT_MCEERR;
3314 			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3315 				layout = SIL_FAULT_BNDERR;
3316 #ifdef SEGV_PKUERR
3317 			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3318 				layout = SIL_FAULT_PKUERR;
3319 #endif
3320 			else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3321 				layout = SIL_FAULT_PERF_EVENT;
3322 			else if (IS_ENABLED(CONFIG_SPARC) &&
3323 				 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3324 				layout = SIL_FAULT_TRAPNO;
3325 			else if (IS_ENABLED(CONFIG_ALPHA) &&
3326 				 ((sig == SIGFPE) ||
3327 				  ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3328 				layout = SIL_FAULT_TRAPNO;
3329 		}
3330 		else if (si_code <= NSIGPOLL)
3331 			layout = SIL_POLL;
3332 	} else {
3333 		if (si_code == SI_TIMER)
3334 			layout = SIL_TIMER;
3335 		else if (si_code == SI_SIGIO)
3336 			layout = SIL_POLL;
3337 		else if (si_code < 0)
3338 			layout = SIL_RT;
3339 	}
3340 	return layout;
3341 }
3342 
3343 static inline char __user *si_expansion(const siginfo_t __user *info)
3344 {
3345 	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3346 }
3347 
3348 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3349 {
3350 	char __user *expansion = si_expansion(to);
3351 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3352 		return -EFAULT;
3353 	if (clear_user(expansion, SI_EXPANSION_SIZE))
3354 		return -EFAULT;
3355 	return 0;
3356 }
3357 
3358 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3359 				       const siginfo_t __user *from)
3360 {
3361 	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3362 		char __user *expansion = si_expansion(from);
3363 		char buf[SI_EXPANSION_SIZE];
3364 		int i;
3365 		/*
3366 		 * An unknown si_code might need more than
3367 		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3368 		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3369 		 * will return this data to userspace exactly.
3370 		 */
3371 		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3372 			return -EFAULT;
3373 		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3374 			if (buf[i] != 0)
3375 				return -E2BIG;
3376 		}
3377 	}
3378 	return 0;
3379 }
3380 
3381 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3382 				    const siginfo_t __user *from)
3383 {
3384 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3385 		return -EFAULT;
3386 	to->si_signo = signo;
3387 	return post_copy_siginfo_from_user(to, from);
3388 }
3389 
3390 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3391 {
3392 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3393 		return -EFAULT;
3394 	return post_copy_siginfo_from_user(to, from);
3395 }
3396 
3397 #ifdef CONFIG_COMPAT
3398 /**
3399  * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3400  * @to: compat siginfo destination
3401  * @from: kernel siginfo source
3402  *
3403  * Note: This function does not work properly for the SIGCHLD on x32, but
3404  * fortunately it doesn't have to.  The only valid callers for this function are
3405  * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3406  * The latter does not care because SIGCHLD will never cause a coredump.
3407  */
3408 void copy_siginfo_to_external32(struct compat_siginfo *to,
3409 		const struct kernel_siginfo *from)
3410 {
3411 	memset(to, 0, sizeof(*to));
3412 
3413 	to->si_signo = from->si_signo;
3414 	to->si_errno = from->si_errno;
3415 	to->si_code  = from->si_code;
3416 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3417 	case SIL_KILL:
3418 		to->si_pid = from->si_pid;
3419 		to->si_uid = from->si_uid;
3420 		break;
3421 	case SIL_TIMER:
3422 		to->si_tid     = from->si_tid;
3423 		to->si_overrun = from->si_overrun;
3424 		to->si_int     = from->si_int;
3425 		break;
3426 	case SIL_POLL:
3427 		to->si_band = from->si_band;
3428 		to->si_fd   = from->si_fd;
3429 		break;
3430 	case SIL_FAULT:
3431 		to->si_addr = ptr_to_compat(from->si_addr);
3432 		break;
3433 	case SIL_FAULT_TRAPNO:
3434 		to->si_addr = ptr_to_compat(from->si_addr);
3435 		to->si_trapno = from->si_trapno;
3436 		break;
3437 	case SIL_FAULT_MCEERR:
3438 		to->si_addr = ptr_to_compat(from->si_addr);
3439 		to->si_addr_lsb = from->si_addr_lsb;
3440 		break;
3441 	case SIL_FAULT_BNDERR:
3442 		to->si_addr = ptr_to_compat(from->si_addr);
3443 		to->si_lower = ptr_to_compat(from->si_lower);
3444 		to->si_upper = ptr_to_compat(from->si_upper);
3445 		break;
3446 	case SIL_FAULT_PKUERR:
3447 		to->si_addr = ptr_to_compat(from->si_addr);
3448 		to->si_pkey = from->si_pkey;
3449 		break;
3450 	case SIL_FAULT_PERF_EVENT:
3451 		to->si_addr = ptr_to_compat(from->si_addr);
3452 		to->si_perf_data = from->si_perf_data;
3453 		to->si_perf_type = from->si_perf_type;
3454 		to->si_perf_flags = from->si_perf_flags;
3455 		break;
3456 	case SIL_CHLD:
3457 		to->si_pid = from->si_pid;
3458 		to->si_uid = from->si_uid;
3459 		to->si_status = from->si_status;
3460 		to->si_utime = from->si_utime;
3461 		to->si_stime = from->si_stime;
3462 		break;
3463 	case SIL_RT:
3464 		to->si_pid = from->si_pid;
3465 		to->si_uid = from->si_uid;
3466 		to->si_int = from->si_int;
3467 		break;
3468 	case SIL_SYS:
3469 		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3470 		to->si_syscall   = from->si_syscall;
3471 		to->si_arch      = from->si_arch;
3472 		break;
3473 	}
3474 }
3475 
3476 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3477 			   const struct kernel_siginfo *from)
3478 {
3479 	struct compat_siginfo new;
3480 
3481 	copy_siginfo_to_external32(&new, from);
3482 	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3483 		return -EFAULT;
3484 	return 0;
3485 }
3486 
3487 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3488 					 const struct compat_siginfo *from)
3489 {
3490 	clear_siginfo(to);
3491 	to->si_signo = from->si_signo;
3492 	to->si_errno = from->si_errno;
3493 	to->si_code  = from->si_code;
3494 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3495 	case SIL_KILL:
3496 		to->si_pid = from->si_pid;
3497 		to->si_uid = from->si_uid;
3498 		break;
3499 	case SIL_TIMER:
3500 		to->si_tid     = from->si_tid;
3501 		to->si_overrun = from->si_overrun;
3502 		to->si_int     = from->si_int;
3503 		break;
3504 	case SIL_POLL:
3505 		to->si_band = from->si_band;
3506 		to->si_fd   = from->si_fd;
3507 		break;
3508 	case SIL_FAULT:
3509 		to->si_addr = compat_ptr(from->si_addr);
3510 		break;
3511 	case SIL_FAULT_TRAPNO:
3512 		to->si_addr = compat_ptr(from->si_addr);
3513 		to->si_trapno = from->si_trapno;
3514 		break;
3515 	case SIL_FAULT_MCEERR:
3516 		to->si_addr = compat_ptr(from->si_addr);
3517 		to->si_addr_lsb = from->si_addr_lsb;
3518 		break;
3519 	case SIL_FAULT_BNDERR:
3520 		to->si_addr = compat_ptr(from->si_addr);
3521 		to->si_lower = compat_ptr(from->si_lower);
3522 		to->si_upper = compat_ptr(from->si_upper);
3523 		break;
3524 	case SIL_FAULT_PKUERR:
3525 		to->si_addr = compat_ptr(from->si_addr);
3526 		to->si_pkey = from->si_pkey;
3527 		break;
3528 	case SIL_FAULT_PERF_EVENT:
3529 		to->si_addr = compat_ptr(from->si_addr);
3530 		to->si_perf_data = from->si_perf_data;
3531 		to->si_perf_type = from->si_perf_type;
3532 		to->si_perf_flags = from->si_perf_flags;
3533 		break;
3534 	case SIL_CHLD:
3535 		to->si_pid    = from->si_pid;
3536 		to->si_uid    = from->si_uid;
3537 		to->si_status = from->si_status;
3538 #ifdef CONFIG_X86_X32_ABI
3539 		if (in_x32_syscall()) {
3540 			to->si_utime = from->_sifields._sigchld_x32._utime;
3541 			to->si_stime = from->_sifields._sigchld_x32._stime;
3542 		} else
3543 #endif
3544 		{
3545 			to->si_utime = from->si_utime;
3546 			to->si_stime = from->si_stime;
3547 		}
3548 		break;
3549 	case SIL_RT:
3550 		to->si_pid = from->si_pid;
3551 		to->si_uid = from->si_uid;
3552 		to->si_int = from->si_int;
3553 		break;
3554 	case SIL_SYS:
3555 		to->si_call_addr = compat_ptr(from->si_call_addr);
3556 		to->si_syscall   = from->si_syscall;
3557 		to->si_arch      = from->si_arch;
3558 		break;
3559 	}
3560 	return 0;
3561 }
3562 
3563 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3564 				      const struct compat_siginfo __user *ufrom)
3565 {
3566 	struct compat_siginfo from;
3567 
3568 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3569 		return -EFAULT;
3570 
3571 	from.si_signo = signo;
3572 	return post_copy_siginfo_from_user32(to, &from);
3573 }
3574 
3575 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3576 			     const struct compat_siginfo __user *ufrom)
3577 {
3578 	struct compat_siginfo from;
3579 
3580 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3581 		return -EFAULT;
3582 
3583 	return post_copy_siginfo_from_user32(to, &from);
3584 }
3585 #endif /* CONFIG_COMPAT */
3586 
3587 /**
3588  *  do_sigtimedwait - wait for queued signals specified in @which
3589  *  @which: queued signals to wait for
3590  *  @info: if non-null, the signal's siginfo is returned here
3591  *  @ts: upper bound on process time suspension
3592  */
3593 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3594 		    const struct timespec64 *ts)
3595 {
3596 	ktime_t *to = NULL, timeout = KTIME_MAX;
3597 	struct task_struct *tsk = current;
3598 	sigset_t mask = *which;
3599 	enum pid_type type;
3600 	int sig, ret = 0;
3601 
3602 	if (ts) {
3603 		if (!timespec64_valid(ts))
3604 			return -EINVAL;
3605 		timeout = timespec64_to_ktime(*ts);
3606 		to = &timeout;
3607 	}
3608 
3609 	/*
3610 	 * Invert the set of allowed signals to get those we want to block.
3611 	 */
3612 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3613 	signotset(&mask);
3614 
3615 	spin_lock_irq(&tsk->sighand->siglock);
3616 	sig = dequeue_signal(tsk, &mask, info, &type);
3617 	if (!sig && timeout) {
3618 		/*
3619 		 * None ready, temporarily unblock those we're interested
3620 		 * while we are sleeping in so that we'll be awakened when
3621 		 * they arrive. Unblocking is always fine, we can avoid
3622 		 * set_current_blocked().
3623 		 */
3624 		tsk->real_blocked = tsk->blocked;
3625 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3626 		recalc_sigpending();
3627 		spin_unlock_irq(&tsk->sighand->siglock);
3628 
3629 		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
3630 		ret = schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3631 					       HRTIMER_MODE_REL);
3632 		spin_lock_irq(&tsk->sighand->siglock);
3633 		__set_task_blocked(tsk, &tsk->real_blocked);
3634 		sigemptyset(&tsk->real_blocked);
3635 		sig = dequeue_signal(tsk, &mask, info, &type);
3636 	}
3637 	spin_unlock_irq(&tsk->sighand->siglock);
3638 
3639 	if (sig)
3640 		return sig;
3641 	return ret ? -EINTR : -EAGAIN;
3642 }
3643 
3644 /**
3645  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3646  *			in @uthese
3647  *  @uthese: queued signals to wait for
3648  *  @uinfo: if non-null, the signal's siginfo is returned here
3649  *  @uts: upper bound on process time suspension
3650  *  @sigsetsize: size of sigset_t type
3651  */
3652 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3653 		siginfo_t __user *, uinfo,
3654 		const struct __kernel_timespec __user *, uts,
3655 		size_t, sigsetsize)
3656 {
3657 	sigset_t these;
3658 	struct timespec64 ts;
3659 	kernel_siginfo_t info;
3660 	int ret;
3661 
3662 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3663 	if (sigsetsize != sizeof(sigset_t))
3664 		return -EINVAL;
3665 
3666 	if (copy_from_user(&these, uthese, sizeof(these)))
3667 		return -EFAULT;
3668 
3669 	if (uts) {
3670 		if (get_timespec64(&ts, uts))
3671 			return -EFAULT;
3672 	}
3673 
3674 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3675 
3676 	if (ret > 0 && uinfo) {
3677 		if (copy_siginfo_to_user(uinfo, &info))
3678 			ret = -EFAULT;
3679 	}
3680 
3681 	return ret;
3682 }
3683 
3684 #ifdef CONFIG_COMPAT_32BIT_TIME
3685 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3686 		siginfo_t __user *, uinfo,
3687 		const struct old_timespec32 __user *, uts,
3688 		size_t, sigsetsize)
3689 {
3690 	sigset_t these;
3691 	struct timespec64 ts;
3692 	kernel_siginfo_t info;
3693 	int ret;
3694 
3695 	if (sigsetsize != sizeof(sigset_t))
3696 		return -EINVAL;
3697 
3698 	if (copy_from_user(&these, uthese, sizeof(these)))
3699 		return -EFAULT;
3700 
3701 	if (uts) {
3702 		if (get_old_timespec32(&ts, uts))
3703 			return -EFAULT;
3704 	}
3705 
3706 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3707 
3708 	if (ret > 0 && uinfo) {
3709 		if (copy_siginfo_to_user(uinfo, &info))
3710 			ret = -EFAULT;
3711 	}
3712 
3713 	return ret;
3714 }
3715 #endif
3716 
3717 #ifdef CONFIG_COMPAT
3718 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3719 		struct compat_siginfo __user *, uinfo,
3720 		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3721 {
3722 	sigset_t s;
3723 	struct timespec64 t;
3724 	kernel_siginfo_t info;
3725 	long ret;
3726 
3727 	if (sigsetsize != sizeof(sigset_t))
3728 		return -EINVAL;
3729 
3730 	if (get_compat_sigset(&s, uthese))
3731 		return -EFAULT;
3732 
3733 	if (uts) {
3734 		if (get_timespec64(&t, uts))
3735 			return -EFAULT;
3736 	}
3737 
3738 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3739 
3740 	if (ret > 0 && uinfo) {
3741 		if (copy_siginfo_to_user32(uinfo, &info))
3742 			ret = -EFAULT;
3743 	}
3744 
3745 	return ret;
3746 }
3747 
3748 #ifdef CONFIG_COMPAT_32BIT_TIME
3749 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3750 		struct compat_siginfo __user *, uinfo,
3751 		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3752 {
3753 	sigset_t s;
3754 	struct timespec64 t;
3755 	kernel_siginfo_t info;
3756 	long ret;
3757 
3758 	if (sigsetsize != sizeof(sigset_t))
3759 		return -EINVAL;
3760 
3761 	if (get_compat_sigset(&s, uthese))
3762 		return -EFAULT;
3763 
3764 	if (uts) {
3765 		if (get_old_timespec32(&t, uts))
3766 			return -EFAULT;
3767 	}
3768 
3769 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3770 
3771 	if (ret > 0 && uinfo) {
3772 		if (copy_siginfo_to_user32(uinfo, &info))
3773 			ret = -EFAULT;
3774 	}
3775 
3776 	return ret;
3777 }
3778 #endif
3779 #endif
3780 
3781 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3782 {
3783 	clear_siginfo(info);
3784 	info->si_signo = sig;
3785 	info->si_errno = 0;
3786 	info->si_code = SI_USER;
3787 	info->si_pid = task_tgid_vnr(current);
3788 	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3789 }
3790 
3791 /**
3792  *  sys_kill - send a signal to a process
3793  *  @pid: the PID of the process
3794  *  @sig: signal to be sent
3795  */
3796 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3797 {
3798 	struct kernel_siginfo info;
3799 
3800 	prepare_kill_siginfo(sig, &info);
3801 
3802 	return kill_something_info(sig, &info, pid);
3803 }
3804 
3805 /*
3806  * Verify that the signaler and signalee either are in the same pid namespace
3807  * or that the signaler's pid namespace is an ancestor of the signalee's pid
3808  * namespace.
3809  */
3810 static bool access_pidfd_pidns(struct pid *pid)
3811 {
3812 	struct pid_namespace *active = task_active_pid_ns(current);
3813 	struct pid_namespace *p = ns_of_pid(pid);
3814 
3815 	for (;;) {
3816 		if (!p)
3817 			return false;
3818 		if (p == active)
3819 			break;
3820 		p = p->parent;
3821 	}
3822 
3823 	return true;
3824 }
3825 
3826 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3827 		siginfo_t __user *info)
3828 {
3829 #ifdef CONFIG_COMPAT
3830 	/*
3831 	 * Avoid hooking up compat syscalls and instead handle necessary
3832 	 * conversions here. Note, this is a stop-gap measure and should not be
3833 	 * considered a generic solution.
3834 	 */
3835 	if (in_compat_syscall())
3836 		return copy_siginfo_from_user32(
3837 			kinfo, (struct compat_siginfo __user *)info);
3838 #endif
3839 	return copy_siginfo_from_user(kinfo, info);
3840 }
3841 
3842 static struct pid *pidfd_to_pid(const struct file *file)
3843 {
3844 	struct pid *pid;
3845 
3846 	pid = pidfd_pid(file);
3847 	if (!IS_ERR(pid))
3848 		return pid;
3849 
3850 	return tgid_pidfd_to_pid(file);
3851 }
3852 
3853 /**
3854  * sys_pidfd_send_signal - Signal a process through a pidfd
3855  * @pidfd:  file descriptor of the process
3856  * @sig:    signal to send
3857  * @info:   signal info
3858  * @flags:  future flags
3859  *
3860  * The syscall currently only signals via PIDTYPE_PID which covers
3861  * kill(<positive-pid>, <signal>. It does not signal threads or process
3862  * groups.
3863  * In order to extend the syscall to threads and process groups the @flags
3864  * argument should be used. In essence, the @flags argument will determine
3865  * what is signaled and not the file descriptor itself. Put in other words,
3866  * grouping is a property of the flags argument not a property of the file
3867  * descriptor.
3868  *
3869  * Return: 0 on success, negative errno on failure
3870  */
3871 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3872 		siginfo_t __user *, info, unsigned int, flags)
3873 {
3874 	int ret;
3875 	struct fd f;
3876 	struct pid *pid;
3877 	kernel_siginfo_t kinfo;
3878 
3879 	/* Enforce flags be set to 0 until we add an extension. */
3880 	if (flags)
3881 		return -EINVAL;
3882 
3883 	f = fdget(pidfd);
3884 	if (!f.file)
3885 		return -EBADF;
3886 
3887 	/* Is this a pidfd? */
3888 	pid = pidfd_to_pid(f.file);
3889 	if (IS_ERR(pid)) {
3890 		ret = PTR_ERR(pid);
3891 		goto err;
3892 	}
3893 
3894 	ret = -EINVAL;
3895 	if (!access_pidfd_pidns(pid))
3896 		goto err;
3897 
3898 	if (info) {
3899 		ret = copy_siginfo_from_user_any(&kinfo, info);
3900 		if (unlikely(ret))
3901 			goto err;
3902 
3903 		ret = -EINVAL;
3904 		if (unlikely(sig != kinfo.si_signo))
3905 			goto err;
3906 
3907 		/* Only allow sending arbitrary signals to yourself. */
3908 		ret = -EPERM;
3909 		if ((task_pid(current) != pid) &&
3910 		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3911 			goto err;
3912 	} else {
3913 		prepare_kill_siginfo(sig, &kinfo);
3914 	}
3915 
3916 	ret = kill_pid_info(sig, &kinfo, pid);
3917 
3918 err:
3919 	fdput(f);
3920 	return ret;
3921 }
3922 
3923 static int
3924 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3925 {
3926 	struct task_struct *p;
3927 	int error = -ESRCH;
3928 
3929 	rcu_read_lock();
3930 	p = find_task_by_vpid(pid);
3931 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3932 		error = check_kill_permission(sig, info, p);
3933 		/*
3934 		 * The null signal is a permissions and process existence
3935 		 * probe.  No signal is actually delivered.
3936 		 */
3937 		if (!error && sig) {
3938 			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3939 			/*
3940 			 * If lock_task_sighand() failed we pretend the task
3941 			 * dies after receiving the signal. The window is tiny,
3942 			 * and the signal is private anyway.
3943 			 */
3944 			if (unlikely(error == -ESRCH))
3945 				error = 0;
3946 		}
3947 	}
3948 	rcu_read_unlock();
3949 
3950 	return error;
3951 }
3952 
3953 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3954 {
3955 	struct kernel_siginfo info;
3956 
3957 	clear_siginfo(&info);
3958 	info.si_signo = sig;
3959 	info.si_errno = 0;
3960 	info.si_code = SI_TKILL;
3961 	info.si_pid = task_tgid_vnr(current);
3962 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3963 
3964 	return do_send_specific(tgid, pid, sig, &info);
3965 }
3966 
3967 /**
3968  *  sys_tgkill - send signal to one specific thread
3969  *  @tgid: the thread group ID of the thread
3970  *  @pid: the PID of the thread
3971  *  @sig: signal to be sent
3972  *
3973  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3974  *  exists but it's not belonging to the target process anymore. This
3975  *  method solves the problem of threads exiting and PIDs getting reused.
3976  */
3977 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3978 {
3979 	/* This is only valid for single tasks */
3980 	if (pid <= 0 || tgid <= 0)
3981 		return -EINVAL;
3982 
3983 	return do_tkill(tgid, pid, sig);
3984 }
3985 
3986 /**
3987  *  sys_tkill - send signal to one specific task
3988  *  @pid: the PID of the task
3989  *  @sig: signal to be sent
3990  *
3991  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3992  */
3993 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3994 {
3995 	/* This is only valid for single tasks */
3996 	if (pid <= 0)
3997 		return -EINVAL;
3998 
3999 	return do_tkill(0, pid, sig);
4000 }
4001 
4002 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
4003 {
4004 	/* Not even root can pretend to send signals from the kernel.
4005 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4006 	 */
4007 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4008 	    (task_pid_vnr(current) != pid))
4009 		return -EPERM;
4010 
4011 	/* POSIX.1b doesn't mention process groups.  */
4012 	return kill_proc_info(sig, info, pid);
4013 }
4014 
4015 /**
4016  *  sys_rt_sigqueueinfo - send signal information to a signal
4017  *  @pid: the PID of the thread
4018  *  @sig: signal to be sent
4019  *  @uinfo: signal info to be sent
4020  */
4021 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
4022 		siginfo_t __user *, uinfo)
4023 {
4024 	kernel_siginfo_t info;
4025 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4026 	if (unlikely(ret))
4027 		return ret;
4028 	return do_rt_sigqueueinfo(pid, sig, &info);
4029 }
4030 
4031 #ifdef CONFIG_COMPAT
4032 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
4033 			compat_pid_t, pid,
4034 			int, sig,
4035 			struct compat_siginfo __user *, uinfo)
4036 {
4037 	kernel_siginfo_t info;
4038 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4039 	if (unlikely(ret))
4040 		return ret;
4041 	return do_rt_sigqueueinfo(pid, sig, &info);
4042 }
4043 #endif
4044 
4045 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4046 {
4047 	/* This is only valid for single tasks */
4048 	if (pid <= 0 || tgid <= 0)
4049 		return -EINVAL;
4050 
4051 	/* Not even root can pretend to send signals from the kernel.
4052 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4053 	 */
4054 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4055 	    (task_pid_vnr(current) != pid))
4056 		return -EPERM;
4057 
4058 	return do_send_specific(tgid, pid, sig, info);
4059 }
4060 
4061 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4062 		siginfo_t __user *, uinfo)
4063 {
4064 	kernel_siginfo_t info;
4065 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4066 	if (unlikely(ret))
4067 		return ret;
4068 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4069 }
4070 
4071 #ifdef CONFIG_COMPAT
4072 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4073 			compat_pid_t, tgid,
4074 			compat_pid_t, pid,
4075 			int, sig,
4076 			struct compat_siginfo __user *, uinfo)
4077 {
4078 	kernel_siginfo_t info;
4079 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4080 	if (unlikely(ret))
4081 		return ret;
4082 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4083 }
4084 #endif
4085 
4086 /*
4087  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4088  */
4089 void kernel_sigaction(int sig, __sighandler_t action)
4090 {
4091 	spin_lock_irq(&current->sighand->siglock);
4092 	current->sighand->action[sig - 1].sa.sa_handler = action;
4093 	if (action == SIG_IGN) {
4094 		sigset_t mask;
4095 
4096 		sigemptyset(&mask);
4097 		sigaddset(&mask, sig);
4098 
4099 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
4100 		flush_sigqueue_mask(&mask, &current->pending);
4101 		recalc_sigpending();
4102 	}
4103 	spin_unlock_irq(&current->sighand->siglock);
4104 }
4105 EXPORT_SYMBOL(kernel_sigaction);
4106 
4107 void __weak sigaction_compat_abi(struct k_sigaction *act,
4108 		struct k_sigaction *oact)
4109 {
4110 }
4111 
4112 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4113 {
4114 	struct task_struct *p = current, *t;
4115 	struct k_sigaction *k;
4116 	sigset_t mask;
4117 
4118 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4119 		return -EINVAL;
4120 
4121 	k = &p->sighand->action[sig-1];
4122 
4123 	spin_lock_irq(&p->sighand->siglock);
4124 	if (k->sa.sa_flags & SA_IMMUTABLE) {
4125 		spin_unlock_irq(&p->sighand->siglock);
4126 		return -EINVAL;
4127 	}
4128 	if (oact)
4129 		*oact = *k;
4130 
4131 	/*
4132 	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4133 	 * e.g. by having an architecture use the bit in their uapi.
4134 	 */
4135 	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4136 
4137 	/*
4138 	 * Clear unknown flag bits in order to allow userspace to detect missing
4139 	 * support for flag bits and to allow the kernel to use non-uapi bits
4140 	 * internally.
4141 	 */
4142 	if (act)
4143 		act->sa.sa_flags &= UAPI_SA_FLAGS;
4144 	if (oact)
4145 		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4146 
4147 	sigaction_compat_abi(act, oact);
4148 
4149 	if (act) {
4150 		sigdelsetmask(&act->sa.sa_mask,
4151 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4152 		*k = *act;
4153 		/*
4154 		 * POSIX 3.3.1.3:
4155 		 *  "Setting a signal action to SIG_IGN for a signal that is
4156 		 *   pending shall cause the pending signal to be discarded,
4157 		 *   whether or not it is blocked."
4158 		 *
4159 		 *  "Setting a signal action to SIG_DFL for a signal that is
4160 		 *   pending and whose default action is to ignore the signal
4161 		 *   (for example, SIGCHLD), shall cause the pending signal to
4162 		 *   be discarded, whether or not it is blocked"
4163 		 */
4164 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4165 			sigemptyset(&mask);
4166 			sigaddset(&mask, sig);
4167 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4168 			for_each_thread(p, t)
4169 				flush_sigqueue_mask(&mask, &t->pending);
4170 		}
4171 	}
4172 
4173 	spin_unlock_irq(&p->sighand->siglock);
4174 	return 0;
4175 }
4176 
4177 #ifdef CONFIG_DYNAMIC_SIGFRAME
4178 static inline void sigaltstack_lock(void)
4179 	__acquires(&current->sighand->siglock)
4180 {
4181 	spin_lock_irq(&current->sighand->siglock);
4182 }
4183 
4184 static inline void sigaltstack_unlock(void)
4185 	__releases(&current->sighand->siglock)
4186 {
4187 	spin_unlock_irq(&current->sighand->siglock);
4188 }
4189 #else
4190 static inline void sigaltstack_lock(void) { }
4191 static inline void sigaltstack_unlock(void) { }
4192 #endif
4193 
4194 static int
4195 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4196 		size_t min_ss_size)
4197 {
4198 	struct task_struct *t = current;
4199 	int ret = 0;
4200 
4201 	if (oss) {
4202 		memset(oss, 0, sizeof(stack_t));
4203 		oss->ss_sp = (void __user *) t->sas_ss_sp;
4204 		oss->ss_size = t->sas_ss_size;
4205 		oss->ss_flags = sas_ss_flags(sp) |
4206 			(current->sas_ss_flags & SS_FLAG_BITS);
4207 	}
4208 
4209 	if (ss) {
4210 		void __user *ss_sp = ss->ss_sp;
4211 		size_t ss_size = ss->ss_size;
4212 		unsigned ss_flags = ss->ss_flags;
4213 		int ss_mode;
4214 
4215 		if (unlikely(on_sig_stack(sp)))
4216 			return -EPERM;
4217 
4218 		ss_mode = ss_flags & ~SS_FLAG_BITS;
4219 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4220 				ss_mode != 0))
4221 			return -EINVAL;
4222 
4223 		/*
4224 		 * Return before taking any locks if no actual
4225 		 * sigaltstack changes were requested.
4226 		 */
4227 		if (t->sas_ss_sp == (unsigned long)ss_sp &&
4228 		    t->sas_ss_size == ss_size &&
4229 		    t->sas_ss_flags == ss_flags)
4230 			return 0;
4231 
4232 		sigaltstack_lock();
4233 		if (ss_mode == SS_DISABLE) {
4234 			ss_size = 0;
4235 			ss_sp = NULL;
4236 		} else {
4237 			if (unlikely(ss_size < min_ss_size))
4238 				ret = -ENOMEM;
4239 			if (!sigaltstack_size_valid(ss_size))
4240 				ret = -ENOMEM;
4241 		}
4242 		if (!ret) {
4243 			t->sas_ss_sp = (unsigned long) ss_sp;
4244 			t->sas_ss_size = ss_size;
4245 			t->sas_ss_flags = ss_flags;
4246 		}
4247 		sigaltstack_unlock();
4248 	}
4249 	return ret;
4250 }
4251 
4252 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4253 {
4254 	stack_t new, old;
4255 	int err;
4256 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4257 		return -EFAULT;
4258 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4259 			      current_user_stack_pointer(),
4260 			      MINSIGSTKSZ);
4261 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4262 		err = -EFAULT;
4263 	return err;
4264 }
4265 
4266 int restore_altstack(const stack_t __user *uss)
4267 {
4268 	stack_t new;
4269 	if (copy_from_user(&new, uss, sizeof(stack_t)))
4270 		return -EFAULT;
4271 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4272 			     MINSIGSTKSZ);
4273 	/* squash all but EFAULT for now */
4274 	return 0;
4275 }
4276 
4277 int __save_altstack(stack_t __user *uss, unsigned long sp)
4278 {
4279 	struct task_struct *t = current;
4280 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4281 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4282 		__put_user(t->sas_ss_size, &uss->ss_size);
4283 	return err;
4284 }
4285 
4286 #ifdef CONFIG_COMPAT
4287 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4288 				 compat_stack_t __user *uoss_ptr)
4289 {
4290 	stack_t uss, uoss;
4291 	int ret;
4292 
4293 	if (uss_ptr) {
4294 		compat_stack_t uss32;
4295 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4296 			return -EFAULT;
4297 		uss.ss_sp = compat_ptr(uss32.ss_sp);
4298 		uss.ss_flags = uss32.ss_flags;
4299 		uss.ss_size = uss32.ss_size;
4300 	}
4301 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4302 			     compat_user_stack_pointer(),
4303 			     COMPAT_MINSIGSTKSZ);
4304 	if (ret >= 0 && uoss_ptr)  {
4305 		compat_stack_t old;
4306 		memset(&old, 0, sizeof(old));
4307 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4308 		old.ss_flags = uoss.ss_flags;
4309 		old.ss_size = uoss.ss_size;
4310 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4311 			ret = -EFAULT;
4312 	}
4313 	return ret;
4314 }
4315 
4316 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4317 			const compat_stack_t __user *, uss_ptr,
4318 			compat_stack_t __user *, uoss_ptr)
4319 {
4320 	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4321 }
4322 
4323 int compat_restore_altstack(const compat_stack_t __user *uss)
4324 {
4325 	int err = do_compat_sigaltstack(uss, NULL);
4326 	/* squash all but -EFAULT for now */
4327 	return err == -EFAULT ? err : 0;
4328 }
4329 
4330 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4331 {
4332 	int err;
4333 	struct task_struct *t = current;
4334 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4335 			 &uss->ss_sp) |
4336 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4337 		__put_user(t->sas_ss_size, &uss->ss_size);
4338 	return err;
4339 }
4340 #endif
4341 
4342 #ifdef __ARCH_WANT_SYS_SIGPENDING
4343 
4344 /**
4345  *  sys_sigpending - examine pending signals
4346  *  @uset: where mask of pending signal is returned
4347  */
4348 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4349 {
4350 	sigset_t set;
4351 
4352 	if (sizeof(old_sigset_t) > sizeof(*uset))
4353 		return -EINVAL;
4354 
4355 	do_sigpending(&set);
4356 
4357 	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4358 		return -EFAULT;
4359 
4360 	return 0;
4361 }
4362 
4363 #ifdef CONFIG_COMPAT
4364 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4365 {
4366 	sigset_t set;
4367 
4368 	do_sigpending(&set);
4369 
4370 	return put_user(set.sig[0], set32);
4371 }
4372 #endif
4373 
4374 #endif
4375 
4376 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4377 /**
4378  *  sys_sigprocmask - examine and change blocked signals
4379  *  @how: whether to add, remove, or set signals
4380  *  @nset: signals to add or remove (if non-null)
4381  *  @oset: previous value of signal mask if non-null
4382  *
4383  * Some platforms have their own version with special arguments;
4384  * others support only sys_rt_sigprocmask.
4385  */
4386 
4387 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4388 		old_sigset_t __user *, oset)
4389 {
4390 	old_sigset_t old_set, new_set;
4391 	sigset_t new_blocked;
4392 
4393 	old_set = current->blocked.sig[0];
4394 
4395 	if (nset) {
4396 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4397 			return -EFAULT;
4398 
4399 		new_blocked = current->blocked;
4400 
4401 		switch (how) {
4402 		case SIG_BLOCK:
4403 			sigaddsetmask(&new_blocked, new_set);
4404 			break;
4405 		case SIG_UNBLOCK:
4406 			sigdelsetmask(&new_blocked, new_set);
4407 			break;
4408 		case SIG_SETMASK:
4409 			new_blocked.sig[0] = new_set;
4410 			break;
4411 		default:
4412 			return -EINVAL;
4413 		}
4414 
4415 		set_current_blocked(&new_blocked);
4416 	}
4417 
4418 	if (oset) {
4419 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4420 			return -EFAULT;
4421 	}
4422 
4423 	return 0;
4424 }
4425 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4426 
4427 #ifndef CONFIG_ODD_RT_SIGACTION
4428 /**
4429  *  sys_rt_sigaction - alter an action taken by a process
4430  *  @sig: signal to be sent
4431  *  @act: new sigaction
4432  *  @oact: used to save the previous sigaction
4433  *  @sigsetsize: size of sigset_t type
4434  */
4435 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4436 		const struct sigaction __user *, act,
4437 		struct sigaction __user *, oact,
4438 		size_t, sigsetsize)
4439 {
4440 	struct k_sigaction new_sa, old_sa;
4441 	int ret;
4442 
4443 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4444 	if (sigsetsize != sizeof(sigset_t))
4445 		return -EINVAL;
4446 
4447 	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4448 		return -EFAULT;
4449 
4450 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4451 	if (ret)
4452 		return ret;
4453 
4454 	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4455 		return -EFAULT;
4456 
4457 	return 0;
4458 }
4459 #ifdef CONFIG_COMPAT
4460 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4461 		const struct compat_sigaction __user *, act,
4462 		struct compat_sigaction __user *, oact,
4463 		compat_size_t, sigsetsize)
4464 {
4465 	struct k_sigaction new_ka, old_ka;
4466 #ifdef __ARCH_HAS_SA_RESTORER
4467 	compat_uptr_t restorer;
4468 #endif
4469 	int ret;
4470 
4471 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4472 	if (sigsetsize != sizeof(compat_sigset_t))
4473 		return -EINVAL;
4474 
4475 	if (act) {
4476 		compat_uptr_t handler;
4477 		ret = get_user(handler, &act->sa_handler);
4478 		new_ka.sa.sa_handler = compat_ptr(handler);
4479 #ifdef __ARCH_HAS_SA_RESTORER
4480 		ret |= get_user(restorer, &act->sa_restorer);
4481 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4482 #endif
4483 		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4484 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4485 		if (ret)
4486 			return -EFAULT;
4487 	}
4488 
4489 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4490 	if (!ret && oact) {
4491 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4492 			       &oact->sa_handler);
4493 		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4494 					 sizeof(oact->sa_mask));
4495 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4496 #ifdef __ARCH_HAS_SA_RESTORER
4497 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4498 				&oact->sa_restorer);
4499 #endif
4500 	}
4501 	return ret;
4502 }
4503 #endif
4504 #endif /* !CONFIG_ODD_RT_SIGACTION */
4505 
4506 #ifdef CONFIG_OLD_SIGACTION
4507 SYSCALL_DEFINE3(sigaction, int, sig,
4508 		const struct old_sigaction __user *, act,
4509 	        struct old_sigaction __user *, oact)
4510 {
4511 	struct k_sigaction new_ka, old_ka;
4512 	int ret;
4513 
4514 	if (act) {
4515 		old_sigset_t mask;
4516 		if (!access_ok(act, sizeof(*act)) ||
4517 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4518 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4519 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4520 		    __get_user(mask, &act->sa_mask))
4521 			return -EFAULT;
4522 #ifdef __ARCH_HAS_KA_RESTORER
4523 		new_ka.ka_restorer = NULL;
4524 #endif
4525 		siginitset(&new_ka.sa.sa_mask, mask);
4526 	}
4527 
4528 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4529 
4530 	if (!ret && oact) {
4531 		if (!access_ok(oact, sizeof(*oact)) ||
4532 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4533 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4534 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4535 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4536 			return -EFAULT;
4537 	}
4538 
4539 	return ret;
4540 }
4541 #endif
4542 #ifdef CONFIG_COMPAT_OLD_SIGACTION
4543 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4544 		const struct compat_old_sigaction __user *, act,
4545 	        struct compat_old_sigaction __user *, oact)
4546 {
4547 	struct k_sigaction new_ka, old_ka;
4548 	int ret;
4549 	compat_old_sigset_t mask;
4550 	compat_uptr_t handler, restorer;
4551 
4552 	if (act) {
4553 		if (!access_ok(act, sizeof(*act)) ||
4554 		    __get_user(handler, &act->sa_handler) ||
4555 		    __get_user(restorer, &act->sa_restorer) ||
4556 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4557 		    __get_user(mask, &act->sa_mask))
4558 			return -EFAULT;
4559 
4560 #ifdef __ARCH_HAS_KA_RESTORER
4561 		new_ka.ka_restorer = NULL;
4562 #endif
4563 		new_ka.sa.sa_handler = compat_ptr(handler);
4564 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4565 		siginitset(&new_ka.sa.sa_mask, mask);
4566 	}
4567 
4568 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4569 
4570 	if (!ret && oact) {
4571 		if (!access_ok(oact, sizeof(*oact)) ||
4572 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4573 			       &oact->sa_handler) ||
4574 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4575 			       &oact->sa_restorer) ||
4576 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4577 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4578 			return -EFAULT;
4579 	}
4580 	return ret;
4581 }
4582 #endif
4583 
4584 #ifdef CONFIG_SGETMASK_SYSCALL
4585 
4586 /*
4587  * For backwards compatibility.  Functionality superseded by sigprocmask.
4588  */
4589 SYSCALL_DEFINE0(sgetmask)
4590 {
4591 	/* SMP safe */
4592 	return current->blocked.sig[0];
4593 }
4594 
4595 SYSCALL_DEFINE1(ssetmask, int, newmask)
4596 {
4597 	int old = current->blocked.sig[0];
4598 	sigset_t newset;
4599 
4600 	siginitset(&newset, newmask);
4601 	set_current_blocked(&newset);
4602 
4603 	return old;
4604 }
4605 #endif /* CONFIG_SGETMASK_SYSCALL */
4606 
4607 #ifdef __ARCH_WANT_SYS_SIGNAL
4608 /*
4609  * For backwards compatibility.  Functionality superseded by sigaction.
4610  */
4611 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4612 {
4613 	struct k_sigaction new_sa, old_sa;
4614 	int ret;
4615 
4616 	new_sa.sa.sa_handler = handler;
4617 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4618 	sigemptyset(&new_sa.sa.sa_mask);
4619 
4620 	ret = do_sigaction(sig, &new_sa, &old_sa);
4621 
4622 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4623 }
4624 #endif /* __ARCH_WANT_SYS_SIGNAL */
4625 
4626 #ifdef __ARCH_WANT_SYS_PAUSE
4627 
4628 SYSCALL_DEFINE0(pause)
4629 {
4630 	while (!signal_pending(current)) {
4631 		__set_current_state(TASK_INTERRUPTIBLE);
4632 		schedule();
4633 	}
4634 	return -ERESTARTNOHAND;
4635 }
4636 
4637 #endif
4638 
4639 static int sigsuspend(sigset_t *set)
4640 {
4641 	current->saved_sigmask = current->blocked;
4642 	set_current_blocked(set);
4643 
4644 	while (!signal_pending(current)) {
4645 		__set_current_state(TASK_INTERRUPTIBLE);
4646 		schedule();
4647 	}
4648 	set_restore_sigmask();
4649 	return -ERESTARTNOHAND;
4650 }
4651 
4652 /**
4653  *  sys_rt_sigsuspend - replace the signal mask for a value with the
4654  *	@unewset value until a signal is received
4655  *  @unewset: new signal mask value
4656  *  @sigsetsize: size of sigset_t type
4657  */
4658 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4659 {
4660 	sigset_t newset;
4661 
4662 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4663 	if (sigsetsize != sizeof(sigset_t))
4664 		return -EINVAL;
4665 
4666 	if (copy_from_user(&newset, unewset, sizeof(newset)))
4667 		return -EFAULT;
4668 	return sigsuspend(&newset);
4669 }
4670 
4671 #ifdef CONFIG_COMPAT
4672 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4673 {
4674 	sigset_t newset;
4675 
4676 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4677 	if (sigsetsize != sizeof(sigset_t))
4678 		return -EINVAL;
4679 
4680 	if (get_compat_sigset(&newset, unewset))
4681 		return -EFAULT;
4682 	return sigsuspend(&newset);
4683 }
4684 #endif
4685 
4686 #ifdef CONFIG_OLD_SIGSUSPEND
4687 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4688 {
4689 	sigset_t blocked;
4690 	siginitset(&blocked, mask);
4691 	return sigsuspend(&blocked);
4692 }
4693 #endif
4694 #ifdef CONFIG_OLD_SIGSUSPEND3
4695 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4696 {
4697 	sigset_t blocked;
4698 	siginitset(&blocked, mask);
4699 	return sigsuspend(&blocked);
4700 }
4701 #endif
4702 
4703 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4704 {
4705 	return NULL;
4706 }
4707 
4708 static inline void siginfo_buildtime_checks(void)
4709 {
4710 	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4711 
4712 	/* Verify the offsets in the two siginfos match */
4713 #define CHECK_OFFSET(field) \
4714 	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4715 
4716 	/* kill */
4717 	CHECK_OFFSET(si_pid);
4718 	CHECK_OFFSET(si_uid);
4719 
4720 	/* timer */
4721 	CHECK_OFFSET(si_tid);
4722 	CHECK_OFFSET(si_overrun);
4723 	CHECK_OFFSET(si_value);
4724 
4725 	/* rt */
4726 	CHECK_OFFSET(si_pid);
4727 	CHECK_OFFSET(si_uid);
4728 	CHECK_OFFSET(si_value);
4729 
4730 	/* sigchld */
4731 	CHECK_OFFSET(si_pid);
4732 	CHECK_OFFSET(si_uid);
4733 	CHECK_OFFSET(si_status);
4734 	CHECK_OFFSET(si_utime);
4735 	CHECK_OFFSET(si_stime);
4736 
4737 	/* sigfault */
4738 	CHECK_OFFSET(si_addr);
4739 	CHECK_OFFSET(si_trapno);
4740 	CHECK_OFFSET(si_addr_lsb);
4741 	CHECK_OFFSET(si_lower);
4742 	CHECK_OFFSET(si_upper);
4743 	CHECK_OFFSET(si_pkey);
4744 	CHECK_OFFSET(si_perf_data);
4745 	CHECK_OFFSET(si_perf_type);
4746 	CHECK_OFFSET(si_perf_flags);
4747 
4748 	/* sigpoll */
4749 	CHECK_OFFSET(si_band);
4750 	CHECK_OFFSET(si_fd);
4751 
4752 	/* sigsys */
4753 	CHECK_OFFSET(si_call_addr);
4754 	CHECK_OFFSET(si_syscall);
4755 	CHECK_OFFSET(si_arch);
4756 #undef CHECK_OFFSET
4757 
4758 	/* usb asyncio */
4759 	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4760 		     offsetof(struct siginfo, si_addr));
4761 	if (sizeof(int) == sizeof(void __user *)) {
4762 		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4763 			     sizeof(void __user *));
4764 	} else {
4765 		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4766 			      sizeof_field(struct siginfo, si_uid)) !=
4767 			     sizeof(void __user *));
4768 		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4769 			     offsetof(struct siginfo, si_uid));
4770 	}
4771 #ifdef CONFIG_COMPAT
4772 	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4773 		     offsetof(struct compat_siginfo, si_addr));
4774 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4775 		     sizeof(compat_uptr_t));
4776 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4777 		     sizeof_field(struct siginfo, si_pid));
4778 #endif
4779 }
4780 
4781 #if defined(CONFIG_SYSCTL)
4782 static struct ctl_table signal_debug_table[] = {
4783 #ifdef CONFIG_SYSCTL_EXCEPTION_TRACE
4784 	{
4785 		.procname	= "exception-trace",
4786 		.data		= &show_unhandled_signals,
4787 		.maxlen		= sizeof(int),
4788 		.mode		= 0644,
4789 		.proc_handler	= proc_dointvec
4790 	},
4791 #endif
4792 	{ }
4793 };
4794 
4795 static int __init init_signal_sysctls(void)
4796 {
4797 	register_sysctl_init("debug", signal_debug_table);
4798 	return 0;
4799 }
4800 early_initcall(init_signal_sysctls);
4801 #endif /* CONFIG_SYSCTL */
4802 
4803 void __init signals_init(void)
4804 {
4805 	siginfo_buildtime_checks();
4806 
4807 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
4808 }
4809 
4810 #ifdef CONFIG_KGDB_KDB
4811 #include <linux/kdb.h>
4812 /*
4813  * kdb_send_sig - Allows kdb to send signals without exposing
4814  * signal internals.  This function checks if the required locks are
4815  * available before calling the main signal code, to avoid kdb
4816  * deadlocks.
4817  */
4818 void kdb_send_sig(struct task_struct *t, int sig)
4819 {
4820 	static struct task_struct *kdb_prev_t;
4821 	int new_t, ret;
4822 	if (!spin_trylock(&t->sighand->siglock)) {
4823 		kdb_printf("Can't do kill command now.\n"
4824 			   "The sigmask lock is held somewhere else in "
4825 			   "kernel, try again later\n");
4826 		return;
4827 	}
4828 	new_t = kdb_prev_t != t;
4829 	kdb_prev_t = t;
4830 	if (!task_is_running(t) && new_t) {
4831 		spin_unlock(&t->sighand->siglock);
4832 		kdb_printf("Process is not RUNNING, sending a signal from "
4833 			   "kdb risks deadlock\n"
4834 			   "on the run queue locks. "
4835 			   "The signal has _not_ been sent.\n"
4836 			   "Reissue the kill command if you want to risk "
4837 			   "the deadlock.\n");
4838 		return;
4839 	}
4840 	ret = send_signal_locked(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4841 	spin_unlock(&t->sighand->siglock);
4842 	if (ret)
4843 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
4844 			   sig, t->pid);
4845 	else
4846 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4847 }
4848 #endif	/* CONFIG_KGDB_KDB */
4849