xref: /linux/kernel/signal.c (revision 3f276cece4dd9e8bf199d9bf3901eef8ca904c2d)
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/mm.h>
26 #include <linux/proc_fs.h>
27 #include <linux/tty.h>
28 #include <linux/binfmts.h>
29 #include <linux/coredump.h>
30 #include <linux/security.h>
31 #include <linux/syscalls.h>
32 #include <linux/ptrace.h>
33 #include <linux/signal.h>
34 #include <linux/signalfd.h>
35 #include <linux/ratelimit.h>
36 #include <linux/task_work.h>
37 #include <linux/capability.h>
38 #include <linux/freezer.h>
39 #include <linux/pid_namespace.h>
40 #include <linux/nsproxy.h>
41 #include <linux/user_namespace.h>
42 #include <linux/uprobes.h>
43 #include <linux/compat.h>
44 #include <linux/cn_proc.h>
45 #include <linux/compiler.h>
46 #include <linux/posix-timers.h>
47 #include <linux/cgroup.h>
48 #include <linux/audit.h>
49 #include <linux/sysctl.h>
50 #include <uapi/linux/pidfd.h>
51 
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/signal.h>
54 
55 #include <asm/param.h>
56 #include <linux/uaccess.h>
57 #include <asm/unistd.h>
58 #include <asm/siginfo.h>
59 #include <asm/cacheflush.h>
60 #include <asm/syscall.h>	/* for syscall_get_* */
61 
62 #include "time/posix-timers.h"
63 
64 /*
65  * SLAB caches for signal bits.
66  */
67 
68 static struct kmem_cache *sigqueue_cachep;
69 
70 int print_fatal_signals __read_mostly;
71 
72 static void __user *sig_handler(struct task_struct *t, int sig)
73 {
74 	return t->sighand->action[sig - 1].sa.sa_handler;
75 }
76 
77 static inline bool sig_handler_ignored(void __user *handler, int sig)
78 {
79 	/* Is it explicitly or implicitly ignored? */
80 	return handler == SIG_IGN ||
81 	       (handler == SIG_DFL && sig_kernel_ignore(sig));
82 }
83 
84 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
85 {
86 	void __user *handler;
87 
88 	handler = sig_handler(t, sig);
89 
90 	/* SIGKILL and SIGSTOP may not be sent to the global init */
91 	if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
92 		return true;
93 
94 	if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
95 	    handler == SIG_DFL && !(force && sig_kernel_only(sig)))
96 		return true;
97 
98 	/* Only allow kernel generated signals to this kthread */
99 	if (unlikely((t->flags & PF_KTHREAD) &&
100 		     (handler == SIG_KTHREAD_KERNEL) && !force))
101 		return true;
102 
103 	return sig_handler_ignored(handler, sig);
104 }
105 
106 static bool sig_ignored(struct task_struct *t, int sig, bool force)
107 {
108 	/*
109 	 * Blocked signals are never ignored, since the
110 	 * signal handler may change by the time it is
111 	 * unblocked.
112 	 */
113 	if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
114 		return false;
115 
116 	/*
117 	 * Tracers may want to know about even ignored signal unless it
118 	 * is SIGKILL which can't be reported anyway but can be ignored
119 	 * by SIGNAL_UNKILLABLE task.
120 	 */
121 	if (t->ptrace && sig != SIGKILL)
122 		return false;
123 
124 	return sig_task_ignored(t, sig, force);
125 }
126 
127 /*
128  * Re-calculate pending state from the set of locally pending
129  * signals, globally pending signals, and blocked signals.
130  */
131 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
132 {
133 	unsigned long ready;
134 	long i;
135 
136 	switch (_NSIG_WORDS) {
137 	default:
138 		for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
139 			ready |= signal->sig[i] &~ blocked->sig[i];
140 		break;
141 
142 	case 4: ready  = signal->sig[3] &~ blocked->sig[3];
143 		ready |= signal->sig[2] &~ blocked->sig[2];
144 		ready |= signal->sig[1] &~ blocked->sig[1];
145 		ready |= signal->sig[0] &~ blocked->sig[0];
146 		break;
147 
148 	case 2: ready  = signal->sig[1] &~ blocked->sig[1];
149 		ready |= signal->sig[0] &~ blocked->sig[0];
150 		break;
151 
152 	case 1: ready  = signal->sig[0] &~ blocked->sig[0];
153 	}
154 	return ready !=	0;
155 }
156 
157 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
158 
159 static bool recalc_sigpending_tsk(struct task_struct *t)
160 {
161 	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
162 	    PENDING(&t->pending, &t->blocked) ||
163 	    PENDING(&t->signal->shared_pending, &t->blocked) ||
164 	    cgroup_task_frozen(t)) {
165 		set_tsk_thread_flag(t, TIF_SIGPENDING);
166 		return true;
167 	}
168 
169 	/*
170 	 * We must never clear the flag in another thread, or in current
171 	 * when it's possible the current syscall is returning -ERESTART*.
172 	 * So we don't clear it here, and only callers who know they should do.
173 	 */
174 	return false;
175 }
176 
177 void recalc_sigpending(void)
178 {
179 	if (!recalc_sigpending_tsk(current) && !freezing(current)) {
180 		if (unlikely(test_thread_flag(TIF_SIGPENDING)))
181 			clear_thread_flag(TIF_SIGPENDING);
182 	}
183 }
184 EXPORT_SYMBOL(recalc_sigpending);
185 
186 void calculate_sigpending(void)
187 {
188 	/* Have any signals or users of TIF_SIGPENDING been delayed
189 	 * until after fork?
190 	 */
191 	spin_lock_irq(&current->sighand->siglock);
192 	set_tsk_thread_flag(current, TIF_SIGPENDING);
193 	recalc_sigpending();
194 	spin_unlock_irq(&current->sighand->siglock);
195 }
196 
197 /* Given the mask, find the first available signal that should be serviced. */
198 
199 #define SYNCHRONOUS_MASK \
200 	(sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
201 	 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
202 
203 int next_signal(struct sigpending *pending, sigset_t *mask)
204 {
205 	unsigned long i, *s, *m, x;
206 	int sig = 0;
207 
208 	s = pending->signal.sig;
209 	m = mask->sig;
210 
211 	/*
212 	 * Handle the first word specially: it contains the
213 	 * synchronous signals that need to be dequeued first.
214 	 */
215 	x = *s &~ *m;
216 	if (x) {
217 		if (x & SYNCHRONOUS_MASK)
218 			x &= SYNCHRONOUS_MASK;
219 		sig = ffz(~x) + 1;
220 		return sig;
221 	}
222 
223 	switch (_NSIG_WORDS) {
224 	default:
225 		for (i = 1; i < _NSIG_WORDS; ++i) {
226 			x = *++s &~ *++m;
227 			if (!x)
228 				continue;
229 			sig = ffz(~x) + i*_NSIG_BPW + 1;
230 			break;
231 		}
232 		break;
233 
234 	case 2:
235 		x = s[1] &~ m[1];
236 		if (!x)
237 			break;
238 		sig = ffz(~x) + _NSIG_BPW + 1;
239 		break;
240 
241 	case 1:
242 		/* Nothing to do */
243 		break;
244 	}
245 
246 	return sig;
247 }
248 
249 static inline void print_dropped_signal(int sig)
250 {
251 	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
252 
253 	if (!print_fatal_signals)
254 		return;
255 
256 	if (!__ratelimit(&ratelimit_state))
257 		return;
258 
259 	pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
260 				current->comm, current->pid, sig);
261 }
262 
263 /**
264  * task_set_jobctl_pending - set jobctl pending bits
265  * @task: target task
266  * @mask: pending bits to set
267  *
268  * Clear @mask from @task->jobctl.  @mask must be subset of
269  * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
270  * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
271  * cleared.  If @task is already being killed or exiting, this function
272  * becomes noop.
273  *
274  * CONTEXT:
275  * Must be called with @task->sighand->siglock held.
276  *
277  * RETURNS:
278  * %true if @mask is set, %false if made noop because @task was dying.
279  */
280 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
281 {
282 	BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
283 			JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
284 	BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
285 
286 	if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
287 		return false;
288 
289 	if (mask & JOBCTL_STOP_SIGMASK)
290 		task->jobctl &= ~JOBCTL_STOP_SIGMASK;
291 
292 	task->jobctl |= mask;
293 	return true;
294 }
295 
296 /**
297  * task_clear_jobctl_trapping - clear jobctl trapping bit
298  * @task: target task
299  *
300  * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
301  * Clear it and wake up the ptracer.  Note that we don't need any further
302  * locking.  @task->siglock guarantees that @task->parent points to the
303  * ptracer.
304  *
305  * CONTEXT:
306  * Must be called with @task->sighand->siglock held.
307  */
308 void task_clear_jobctl_trapping(struct task_struct *task)
309 {
310 	if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
311 		task->jobctl &= ~JOBCTL_TRAPPING;
312 		smp_mb();	/* advised by wake_up_bit() */
313 		wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
314 	}
315 }
316 
317 /**
318  * task_clear_jobctl_pending - clear jobctl pending bits
319  * @task: target task
320  * @mask: pending bits to clear
321  *
322  * Clear @mask from @task->jobctl.  @mask must be subset of
323  * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
324  * STOP bits are cleared together.
325  *
326  * If clearing of @mask leaves no stop or trap pending, this function calls
327  * task_clear_jobctl_trapping().
328  *
329  * CONTEXT:
330  * Must be called with @task->sighand->siglock held.
331  */
332 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
333 {
334 	BUG_ON(mask & ~JOBCTL_PENDING_MASK);
335 
336 	if (mask & JOBCTL_STOP_PENDING)
337 		mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
338 
339 	task->jobctl &= ~mask;
340 
341 	if (!(task->jobctl & JOBCTL_PENDING_MASK))
342 		task_clear_jobctl_trapping(task);
343 }
344 
345 /**
346  * task_participate_group_stop - participate in a group stop
347  * @task: task participating in a group stop
348  *
349  * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
350  * Group stop states are cleared and the group stop count is consumed if
351  * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
352  * stop, the appropriate `SIGNAL_*` flags are set.
353  *
354  * CONTEXT:
355  * Must be called with @task->sighand->siglock held.
356  *
357  * RETURNS:
358  * %true if group stop completion should be notified to the parent, %false
359  * otherwise.
360  */
361 static bool task_participate_group_stop(struct task_struct *task)
362 {
363 	struct signal_struct *sig = task->signal;
364 	bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
365 
366 	WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
367 
368 	task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
369 
370 	if (!consume)
371 		return false;
372 
373 	if (!WARN_ON_ONCE(sig->group_stop_count == 0))
374 		sig->group_stop_count--;
375 
376 	/*
377 	 * Tell the caller to notify completion iff we are entering into a
378 	 * fresh group stop.  Read comment in do_signal_stop() for details.
379 	 */
380 	if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
381 		signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
382 		return true;
383 	}
384 	return false;
385 }
386 
387 void task_join_group_stop(struct task_struct *task)
388 {
389 	unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
390 	struct signal_struct *sig = current->signal;
391 
392 	if (sig->group_stop_count) {
393 		sig->group_stop_count++;
394 		mask |= JOBCTL_STOP_CONSUME;
395 	} else if (!(sig->flags & SIGNAL_STOP_STOPPED))
396 		return;
397 
398 	/* Have the new thread join an on-going signal group stop */
399 	task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
400 }
401 
402 static struct ucounts *sig_get_ucounts(struct task_struct *t, int sig,
403 				       int override_rlimit)
404 {
405 	struct ucounts *ucounts;
406 	long sigpending;
407 
408 	/*
409 	 * Protect access to @t credentials. This can go away when all
410 	 * callers hold rcu read lock.
411 	 *
412 	 * NOTE! A pending signal will hold on to the user refcount,
413 	 * and we get/put the refcount only when the sigpending count
414 	 * changes from/to zero.
415 	 */
416 	rcu_read_lock();
417 	ucounts = task_ucounts(t);
418 	sigpending = inc_rlimit_get_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING,
419 					    override_rlimit);
420 	rcu_read_unlock();
421 	if (!sigpending)
422 		return NULL;
423 
424 	if (unlikely(!override_rlimit && sigpending > task_rlimit(t, RLIMIT_SIGPENDING))) {
425 		dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
426 		print_dropped_signal(sig);
427 		return NULL;
428 	}
429 
430 	return ucounts;
431 }
432 
433 static void __sigqueue_init(struct sigqueue *q, struct ucounts *ucounts,
434 			    const unsigned int sigqueue_flags)
435 {
436 	INIT_LIST_HEAD(&q->list);
437 	q->flags = sigqueue_flags;
438 	q->ucounts = ucounts;
439 }
440 
441 /*
442  * allocate a new signal queue record
443  * - this may be called without locks if and only if t == current, otherwise an
444  *   appropriate lock must be held to stop the target task from exiting
445  */
446 static struct sigqueue *sigqueue_alloc(int sig, struct task_struct *t, gfp_t gfp_flags,
447 				       int override_rlimit)
448 {
449 	struct ucounts *ucounts = sig_get_ucounts(t, sig, override_rlimit);
450 	struct sigqueue *q;
451 
452 	if (!ucounts)
453 		return NULL;
454 
455 	q = kmem_cache_alloc(sigqueue_cachep, gfp_flags);
456 	if (!q) {
457 		dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
458 		return NULL;
459 	}
460 
461 	__sigqueue_init(q, ucounts, 0);
462 	return q;
463 }
464 
465 static void __sigqueue_free(struct sigqueue *q)
466 {
467 	if (q->flags & SIGQUEUE_PREALLOC) {
468 		posixtimer_sigqueue_putref(q);
469 		return;
470 	}
471 	if (q->ucounts) {
472 		dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING);
473 		q->ucounts = NULL;
474 	}
475 	kmem_cache_free(sigqueue_cachep, q);
476 }
477 
478 void flush_sigqueue(struct sigpending *queue)
479 {
480 	struct sigqueue *q;
481 
482 	sigemptyset(&queue->signal);
483 	while (!list_empty(&queue->list)) {
484 		q = list_entry(queue->list.next, struct sigqueue , list);
485 		list_del_init(&q->list);
486 		__sigqueue_free(q);
487 	}
488 }
489 
490 /*
491  * Flush all pending signals for this kthread.
492  */
493 void flush_signals(struct task_struct *t)
494 {
495 	unsigned long flags;
496 
497 	spin_lock_irqsave(&t->sighand->siglock, flags);
498 	clear_tsk_thread_flag(t, TIF_SIGPENDING);
499 	flush_sigqueue(&t->pending);
500 	flush_sigqueue(&t->signal->shared_pending);
501 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
502 }
503 EXPORT_SYMBOL(flush_signals);
504 
505 void ignore_signals(struct task_struct *t)
506 {
507 	int i;
508 
509 	for (i = 0; i < _NSIG; ++i)
510 		t->sighand->action[i].sa.sa_handler = SIG_IGN;
511 
512 	flush_signals(t);
513 }
514 
515 /*
516  * Flush all handlers for a task.
517  */
518 
519 void
520 flush_signal_handlers(struct task_struct *t, int force_default)
521 {
522 	int i;
523 	struct k_sigaction *ka = &t->sighand->action[0];
524 	for (i = _NSIG ; i != 0 ; i--) {
525 		if (force_default || ka->sa.sa_handler != SIG_IGN)
526 			ka->sa.sa_handler = SIG_DFL;
527 		ka->sa.sa_flags = 0;
528 #ifdef __ARCH_HAS_SA_RESTORER
529 		ka->sa.sa_restorer = NULL;
530 #endif
531 		sigemptyset(&ka->sa.sa_mask);
532 		ka++;
533 	}
534 }
535 
536 bool unhandled_signal(struct task_struct *tsk, int sig)
537 {
538 	void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
539 	if (is_global_init(tsk))
540 		return true;
541 
542 	if (handler != SIG_IGN && handler != SIG_DFL)
543 		return false;
544 
545 	/* If dying, we handle all new signals by ignoring them */
546 	if (fatal_signal_pending(tsk))
547 		return false;
548 
549 	/* if ptraced, let the tracer determine */
550 	return !tsk->ptrace;
551 }
552 
553 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
554 			   struct sigqueue **timer_sigq)
555 {
556 	struct sigqueue *q, *first = NULL;
557 
558 	/*
559 	 * Collect the siginfo appropriate to this signal.  Check if
560 	 * there is another siginfo for the same signal.
561 	*/
562 	list_for_each_entry(q, &list->list, list) {
563 		if (q->info.si_signo == sig) {
564 			if (first)
565 				goto still_pending;
566 			first = q;
567 		}
568 	}
569 
570 	sigdelset(&list->signal, sig);
571 
572 	if (first) {
573 still_pending:
574 		list_del_init(&first->list);
575 		copy_siginfo(info, &first->info);
576 
577 		/*
578 		 * posix-timer signals are preallocated and freed when the last
579 		 * reference count is dropped in posixtimer_deliver_signal() or
580 		 * immediately on timer deletion when the signal is not pending.
581 		 * Spare the extra round through __sigqueue_free() which is
582 		 * ignoring preallocated signals.
583 		 */
584 		if (unlikely((first->flags & SIGQUEUE_PREALLOC) && (info->si_code == SI_TIMER)))
585 			*timer_sigq = first;
586 		else
587 			__sigqueue_free(first);
588 	} else {
589 		/*
590 		 * Ok, it wasn't in the queue.  This must be
591 		 * a fast-pathed signal or we must have been
592 		 * out of queue space.  So zero out the info.
593 		 */
594 		clear_siginfo(info);
595 		info->si_signo = sig;
596 		info->si_errno = 0;
597 		info->si_code = SI_USER;
598 		info->si_pid = 0;
599 		info->si_uid = 0;
600 	}
601 }
602 
603 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
604 			    kernel_siginfo_t *info, struct sigqueue **timer_sigq)
605 {
606 	int sig = next_signal(pending, mask);
607 
608 	if (sig)
609 		collect_signal(sig, pending, info, timer_sigq);
610 	return sig;
611 }
612 
613 /*
614  * Try to dequeue a signal. If a deliverable signal is found fill in the
615  * caller provided siginfo and return the signal number. Otherwise return
616  * 0.
617  */
618 int dequeue_signal(sigset_t *mask, kernel_siginfo_t *info, enum pid_type *type)
619 {
620 	struct task_struct *tsk = current;
621 	struct sigqueue *timer_sigq;
622 	int signr;
623 
624 	lockdep_assert_held(&tsk->sighand->siglock);
625 
626 again:
627 	*type = PIDTYPE_PID;
628 	timer_sigq = NULL;
629 	signr = __dequeue_signal(&tsk->pending, mask, info, &timer_sigq);
630 	if (!signr) {
631 		*type = PIDTYPE_TGID;
632 		signr = __dequeue_signal(&tsk->signal->shared_pending,
633 					 mask, info, &timer_sigq);
634 
635 		if (unlikely(signr == SIGALRM))
636 			posixtimer_rearm_itimer(tsk);
637 	}
638 
639 	recalc_sigpending();
640 	if (!signr)
641 		return 0;
642 
643 	if (unlikely(sig_kernel_stop(signr))) {
644 		/*
645 		 * Set a marker that we have dequeued a stop signal.  Our
646 		 * caller might release the siglock and then the pending
647 		 * stop signal it is about to process is no longer in the
648 		 * pending bitmasks, but must still be cleared by a SIGCONT
649 		 * (and overruled by a SIGKILL).  So those cases clear this
650 		 * shared flag after we've set it.  Note that this flag may
651 		 * remain set after the signal we return is ignored or
652 		 * handled.  That doesn't matter because its only purpose
653 		 * is to alert stop-signal processing code when another
654 		 * processor has come along and cleared the flag.
655 		 */
656 		current->jobctl |= JOBCTL_STOP_DEQUEUED;
657 	}
658 
659 	if (IS_ENABLED(CONFIG_POSIX_TIMERS) && unlikely(timer_sigq)) {
660 		if (!posixtimer_deliver_signal(info, timer_sigq))
661 			goto again;
662 	}
663 
664 	return signr;
665 }
666 EXPORT_SYMBOL_GPL(dequeue_signal);
667 
668 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
669 {
670 	struct task_struct *tsk = current;
671 	struct sigpending *pending = &tsk->pending;
672 	struct sigqueue *q, *sync = NULL;
673 
674 	/*
675 	 * Might a synchronous signal be in the queue?
676 	 */
677 	if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
678 		return 0;
679 
680 	/*
681 	 * Return the first synchronous signal in the queue.
682 	 */
683 	list_for_each_entry(q, &pending->list, list) {
684 		/* Synchronous signals have a positive si_code */
685 		if ((q->info.si_code > SI_USER) &&
686 		    (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
687 			sync = q;
688 			goto next;
689 		}
690 	}
691 	return 0;
692 next:
693 	/*
694 	 * Check if there is another siginfo for the same signal.
695 	 */
696 	list_for_each_entry_continue(q, &pending->list, list) {
697 		if (q->info.si_signo == sync->info.si_signo)
698 			goto still_pending;
699 	}
700 
701 	sigdelset(&pending->signal, sync->info.si_signo);
702 	recalc_sigpending();
703 still_pending:
704 	list_del_init(&sync->list);
705 	copy_siginfo(info, &sync->info);
706 	__sigqueue_free(sync);
707 	return info->si_signo;
708 }
709 
710 /*
711  * Tell a process that it has a new active signal..
712  *
713  * NOTE! we rely on the previous spin_lock to
714  * lock interrupts for us! We can only be called with
715  * "siglock" held, and the local interrupt must
716  * have been disabled when that got acquired!
717  *
718  * No need to set need_resched since signal event passing
719  * goes through ->blocked
720  */
721 void signal_wake_up_state(struct task_struct *t, unsigned int state)
722 {
723 	lockdep_assert_held(&t->sighand->siglock);
724 
725 	set_tsk_thread_flag(t, TIF_SIGPENDING);
726 
727 	/*
728 	 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
729 	 * case. We don't check t->state here because there is a race with it
730 	 * executing another processor and just now entering stopped state.
731 	 * By using wake_up_state, we ensure the process will wake up and
732 	 * handle its death signal.
733 	 */
734 	if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
735 		kick_process(t);
736 }
737 
738 static inline void posixtimer_sig_ignore(struct task_struct *tsk, struct sigqueue *q);
739 
740 static void sigqueue_free_ignored(struct task_struct *tsk, struct sigqueue *q)
741 {
742 	if (likely(!(q->flags & SIGQUEUE_PREALLOC) || q->info.si_code != SI_TIMER))
743 		__sigqueue_free(q);
744 	else
745 		posixtimer_sig_ignore(tsk, q);
746 }
747 
748 /* Remove signals in mask from the pending set and queue. */
749 static void flush_sigqueue_mask(struct task_struct *p, sigset_t *mask, struct sigpending *s)
750 {
751 	struct sigqueue *q, *n;
752 	sigset_t m;
753 
754 	lockdep_assert_held(&p->sighand->siglock);
755 
756 	sigandsets(&m, mask, &s->signal);
757 	if (sigisemptyset(&m))
758 		return;
759 
760 	sigandnsets(&s->signal, &s->signal, mask);
761 	list_for_each_entry_safe(q, n, &s->list, list) {
762 		if (sigismember(mask, q->info.si_signo)) {
763 			list_del_init(&q->list);
764 			sigqueue_free_ignored(p, q);
765 		}
766 	}
767 }
768 
769 static inline int is_si_special(const struct kernel_siginfo *info)
770 {
771 	return info <= SEND_SIG_PRIV;
772 }
773 
774 static inline bool si_fromuser(const struct kernel_siginfo *info)
775 {
776 	return info == SEND_SIG_NOINFO ||
777 		(!is_si_special(info) && SI_FROMUSER(info));
778 }
779 
780 /*
781  * called with RCU read lock from check_kill_permission()
782  */
783 static bool kill_ok_by_cred(struct task_struct *t)
784 {
785 	const struct cred *cred = current_cred();
786 	const struct cred *tcred = __task_cred(t);
787 
788 	return uid_eq(cred->euid, tcred->suid) ||
789 	       uid_eq(cred->euid, tcred->uid) ||
790 	       uid_eq(cred->uid, tcred->suid) ||
791 	       uid_eq(cred->uid, tcred->uid) ||
792 	       ns_capable(tcred->user_ns, CAP_KILL);
793 }
794 
795 /*
796  * Bad permissions for sending the signal
797  * - the caller must hold the RCU read lock
798  */
799 static int check_kill_permission(int sig, struct kernel_siginfo *info,
800 				 struct task_struct *t)
801 {
802 	struct pid *sid;
803 	int error;
804 
805 	if (!valid_signal(sig))
806 		return -EINVAL;
807 
808 	if (!si_fromuser(info))
809 		return 0;
810 
811 	error = audit_signal_info(sig, t); /* Let audit system see the signal */
812 	if (error)
813 		return error;
814 
815 	if (!same_thread_group(current, t) &&
816 	    !kill_ok_by_cred(t)) {
817 		switch (sig) {
818 		case SIGCONT:
819 			sid = task_session(t);
820 			/*
821 			 * We don't return the error if sid == NULL. The
822 			 * task was unhashed, the caller must notice this.
823 			 */
824 			if (!sid || sid == task_session(current))
825 				break;
826 			fallthrough;
827 		default:
828 			return -EPERM;
829 		}
830 	}
831 
832 	return security_task_kill(t, info, sig, NULL);
833 }
834 
835 /**
836  * ptrace_trap_notify - schedule trap to notify ptracer
837  * @t: tracee wanting to notify tracer
838  *
839  * This function schedules sticky ptrace trap which is cleared on the next
840  * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
841  * ptracer.
842  *
843  * If @t is running, STOP trap will be taken.  If trapped for STOP and
844  * ptracer is listening for events, tracee is woken up so that it can
845  * re-trap for the new event.  If trapped otherwise, STOP trap will be
846  * eventually taken without returning to userland after the existing traps
847  * are finished by PTRACE_CONT.
848  *
849  * CONTEXT:
850  * Must be called with @task->sighand->siglock held.
851  */
852 static void ptrace_trap_notify(struct task_struct *t)
853 {
854 	WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
855 	lockdep_assert_held(&t->sighand->siglock);
856 
857 	task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
858 	ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
859 }
860 
861 /*
862  * Handle magic process-wide effects of stop/continue signals. Unlike
863  * the signal actions, these happen immediately at signal-generation
864  * time regardless of blocking, ignoring, or handling.  This does the
865  * actual continuing for SIGCONT, but not the actual stopping for stop
866  * signals. The process stop is done as a signal action for SIG_DFL.
867  *
868  * Returns true if the signal should be actually delivered, otherwise
869  * it should be dropped.
870  */
871 static bool prepare_signal(int sig, struct task_struct *p, bool force)
872 {
873 	struct signal_struct *signal = p->signal;
874 	struct task_struct *t;
875 	sigset_t flush;
876 
877 	if (signal->flags & SIGNAL_GROUP_EXIT) {
878 		if (signal->core_state)
879 			return sig == SIGKILL;
880 		/*
881 		 * The process is in the middle of dying, drop the signal.
882 		 */
883 		return false;
884 	} else if (sig_kernel_stop(sig)) {
885 		/*
886 		 * This is a stop signal.  Remove SIGCONT from all queues.
887 		 */
888 		siginitset(&flush, sigmask(SIGCONT));
889 		flush_sigqueue_mask(p, &flush, &signal->shared_pending);
890 		for_each_thread(p, t)
891 			flush_sigqueue_mask(p, &flush, &t->pending);
892 	} else if (sig == SIGCONT) {
893 		unsigned int why;
894 		/*
895 		 * Remove all stop signals from all queues, wake all threads.
896 		 */
897 		siginitset(&flush, SIG_KERNEL_STOP_MASK);
898 		flush_sigqueue_mask(p, &flush, &signal->shared_pending);
899 		for_each_thread(p, t) {
900 			flush_sigqueue_mask(p, &flush, &t->pending);
901 			task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
902 			if (likely(!(t->ptrace & PT_SEIZED))) {
903 				t->jobctl &= ~JOBCTL_STOPPED;
904 				wake_up_state(t, __TASK_STOPPED);
905 			} else
906 				ptrace_trap_notify(t);
907 		}
908 
909 		/*
910 		 * Notify the parent with CLD_CONTINUED if we were stopped.
911 		 *
912 		 * If we were in the middle of a group stop, we pretend it
913 		 * was already finished, and then continued. Since SIGCHLD
914 		 * doesn't queue we report only CLD_STOPPED, as if the next
915 		 * CLD_CONTINUED was dropped.
916 		 */
917 		why = 0;
918 		if (signal->flags & SIGNAL_STOP_STOPPED)
919 			why |= SIGNAL_CLD_CONTINUED;
920 		else if (signal->group_stop_count)
921 			why |= SIGNAL_CLD_STOPPED;
922 
923 		if (why) {
924 			/*
925 			 * The first thread which returns from do_signal_stop()
926 			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
927 			 * notify its parent. See get_signal().
928 			 */
929 			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
930 			signal->group_stop_count = 0;
931 			signal->group_exit_code = 0;
932 		}
933 	}
934 
935 	return !sig_ignored(p, sig, force);
936 }
937 
938 /*
939  * Test if P wants to take SIG.  After we've checked all threads with this,
940  * it's equivalent to finding no threads not blocking SIG.  Any threads not
941  * blocking SIG were ruled out because they are not running and already
942  * have pending signals.  Such threads will dequeue from the shared queue
943  * as soon as they're available, so putting the signal on the shared queue
944  * will be equivalent to sending it to one such thread.
945  */
946 static inline bool wants_signal(int sig, struct task_struct *p)
947 {
948 	if (sigismember(&p->blocked, sig))
949 		return false;
950 
951 	if (p->flags & PF_EXITING)
952 		return false;
953 
954 	if (sig == SIGKILL)
955 		return true;
956 
957 	if (task_is_stopped_or_traced(p))
958 		return false;
959 
960 	return task_curr(p) || !task_sigpending(p);
961 }
962 
963 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
964 {
965 	struct signal_struct *signal = p->signal;
966 	struct task_struct *t;
967 
968 	/*
969 	 * Now find a thread we can wake up to take the signal off the queue.
970 	 *
971 	 * Try the suggested task first (may or may not be the main thread).
972 	 */
973 	if (wants_signal(sig, p))
974 		t = p;
975 	else if ((type == PIDTYPE_PID) || thread_group_empty(p))
976 		/*
977 		 * There is just one thread and it does not need to be woken.
978 		 * It will dequeue unblocked signals before it runs again.
979 		 */
980 		return;
981 	else {
982 		/*
983 		 * Otherwise try to find a suitable thread.
984 		 */
985 		t = signal->curr_target;
986 		while (!wants_signal(sig, t)) {
987 			t = next_thread(t);
988 			if (t == signal->curr_target)
989 				/*
990 				 * No thread needs to be woken.
991 				 * Any eligible threads will see
992 				 * the signal in the queue soon.
993 				 */
994 				return;
995 		}
996 		signal->curr_target = t;
997 	}
998 
999 	/*
1000 	 * Found a killable thread.  If the signal will be fatal,
1001 	 * then start taking the whole group down immediately.
1002 	 */
1003 	if (sig_fatal(p, sig) && !sigismember(&t->real_blocked, sig) &&
1004 	    (sig == SIGKILL || !p->ptrace)) {
1005 		/*
1006 		 * This signal will be fatal to the whole group.
1007 		 */
1008 		if (!sig_kernel_coredump(sig)) {
1009 			/*
1010 			 * Start a group exit and wake everybody up.
1011 			 * This way we don't have other threads
1012 			 * running and doing things after a slower
1013 			 * thread has the fatal signal pending.
1014 			 */
1015 			signal->flags = SIGNAL_GROUP_EXIT;
1016 			signal->group_exit_code = sig;
1017 			signal->group_stop_count = 0;
1018 			__for_each_thread(signal, t) {
1019 				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1020 				sigaddset(&t->pending.signal, SIGKILL);
1021 				signal_wake_up(t, 1);
1022 			}
1023 			return;
1024 		}
1025 	}
1026 
1027 	/*
1028 	 * The signal is already in the shared-pending queue.
1029 	 * Tell the chosen thread to wake up and dequeue it.
1030 	 */
1031 	signal_wake_up(t, sig == SIGKILL);
1032 	return;
1033 }
1034 
1035 static inline bool legacy_queue(struct sigpending *signals, int sig)
1036 {
1037 	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1038 }
1039 
1040 static int __send_signal_locked(int sig, struct kernel_siginfo *info,
1041 				struct task_struct *t, enum pid_type type, bool force)
1042 {
1043 	struct sigpending *pending;
1044 	struct sigqueue *q;
1045 	int override_rlimit;
1046 	int ret = 0, result;
1047 
1048 	lockdep_assert_held(&t->sighand->siglock);
1049 
1050 	result = TRACE_SIGNAL_IGNORED;
1051 	if (!prepare_signal(sig, t, force))
1052 		goto ret;
1053 
1054 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1055 	/*
1056 	 * Short-circuit ignored signals and support queuing
1057 	 * exactly one non-rt signal, so that we can get more
1058 	 * detailed information about the cause of the signal.
1059 	 */
1060 	result = TRACE_SIGNAL_ALREADY_PENDING;
1061 	if (legacy_queue(pending, sig))
1062 		goto ret;
1063 
1064 	result = TRACE_SIGNAL_DELIVERED;
1065 	/*
1066 	 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1067 	 */
1068 	if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1069 		goto out_set;
1070 
1071 	/*
1072 	 * Real-time signals must be queued if sent by sigqueue, or
1073 	 * some other real-time mechanism.  It is implementation
1074 	 * defined whether kill() does so.  We attempt to do so, on
1075 	 * the principle of least surprise, but since kill is not
1076 	 * allowed to fail with EAGAIN when low on memory we just
1077 	 * make sure at least one signal gets delivered and don't
1078 	 * pass on the info struct.
1079 	 */
1080 	if (sig < SIGRTMIN)
1081 		override_rlimit = (is_si_special(info) || info->si_code >= 0);
1082 	else
1083 		override_rlimit = 0;
1084 
1085 	q = sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit);
1086 
1087 	if (q) {
1088 		list_add_tail(&q->list, &pending->list);
1089 		switch ((unsigned long) info) {
1090 		case (unsigned long) SEND_SIG_NOINFO:
1091 			clear_siginfo(&q->info);
1092 			q->info.si_signo = sig;
1093 			q->info.si_errno = 0;
1094 			q->info.si_code = SI_USER;
1095 			q->info.si_pid = task_tgid_nr_ns(current,
1096 							task_active_pid_ns(t));
1097 			rcu_read_lock();
1098 			q->info.si_uid =
1099 				from_kuid_munged(task_cred_xxx(t, user_ns),
1100 						 current_uid());
1101 			rcu_read_unlock();
1102 			break;
1103 		case (unsigned long) SEND_SIG_PRIV:
1104 			clear_siginfo(&q->info);
1105 			q->info.si_signo = sig;
1106 			q->info.si_errno = 0;
1107 			q->info.si_code = SI_KERNEL;
1108 			q->info.si_pid = 0;
1109 			q->info.si_uid = 0;
1110 			break;
1111 		default:
1112 			copy_siginfo(&q->info, info);
1113 			break;
1114 		}
1115 	} else if (!is_si_special(info) &&
1116 		   sig >= SIGRTMIN && info->si_code != SI_USER) {
1117 		/*
1118 		 * Queue overflow, abort.  We may abort if the
1119 		 * signal was rt and sent by user using something
1120 		 * other than kill().
1121 		 */
1122 		result = TRACE_SIGNAL_OVERFLOW_FAIL;
1123 		ret = -EAGAIN;
1124 		goto ret;
1125 	} else {
1126 		/*
1127 		 * This is a silent loss of information.  We still
1128 		 * send the signal, but the *info bits are lost.
1129 		 */
1130 		result = TRACE_SIGNAL_LOSE_INFO;
1131 	}
1132 
1133 out_set:
1134 	signalfd_notify(t, sig);
1135 	sigaddset(&pending->signal, sig);
1136 
1137 	/* Let multiprocess signals appear after on-going forks */
1138 	if (type > PIDTYPE_TGID) {
1139 		struct multiprocess_signals *delayed;
1140 		hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1141 			sigset_t *signal = &delayed->signal;
1142 			/* Can't queue both a stop and a continue signal */
1143 			if (sig == SIGCONT)
1144 				sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1145 			else if (sig_kernel_stop(sig))
1146 				sigdelset(signal, SIGCONT);
1147 			sigaddset(signal, sig);
1148 		}
1149 	}
1150 
1151 	complete_signal(sig, t, type);
1152 ret:
1153 	trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1154 	return ret;
1155 }
1156 
1157 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1158 {
1159 	bool ret = false;
1160 	switch (siginfo_layout(info->si_signo, info->si_code)) {
1161 	case SIL_KILL:
1162 	case SIL_CHLD:
1163 	case SIL_RT:
1164 		ret = true;
1165 		break;
1166 	case SIL_TIMER:
1167 	case SIL_POLL:
1168 	case SIL_FAULT:
1169 	case SIL_FAULT_TRAPNO:
1170 	case SIL_FAULT_MCEERR:
1171 	case SIL_FAULT_BNDERR:
1172 	case SIL_FAULT_PKUERR:
1173 	case SIL_FAULT_PERF_EVENT:
1174 	case SIL_SYS:
1175 		ret = false;
1176 		break;
1177 	}
1178 	return ret;
1179 }
1180 
1181 int send_signal_locked(int sig, struct kernel_siginfo *info,
1182 		       struct task_struct *t, enum pid_type type)
1183 {
1184 	/* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1185 	bool force = false;
1186 
1187 	if (info == SEND_SIG_NOINFO) {
1188 		/* Force if sent from an ancestor pid namespace */
1189 		force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1190 	} else if (info == SEND_SIG_PRIV) {
1191 		/* Don't ignore kernel generated signals */
1192 		force = true;
1193 	} else if (has_si_pid_and_uid(info)) {
1194 		/* SIGKILL and SIGSTOP is special or has ids */
1195 		struct user_namespace *t_user_ns;
1196 
1197 		rcu_read_lock();
1198 		t_user_ns = task_cred_xxx(t, user_ns);
1199 		if (current_user_ns() != t_user_ns) {
1200 			kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1201 			info->si_uid = from_kuid_munged(t_user_ns, uid);
1202 		}
1203 		rcu_read_unlock();
1204 
1205 		/* A kernel generated signal? */
1206 		force = (info->si_code == SI_KERNEL);
1207 
1208 		/* From an ancestor pid namespace? */
1209 		if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1210 			info->si_pid = 0;
1211 			force = true;
1212 		}
1213 	}
1214 	return __send_signal_locked(sig, info, t, type, force);
1215 }
1216 
1217 static void print_fatal_signal(int signr)
1218 {
1219 	struct pt_regs *regs = task_pt_regs(current);
1220 	struct file *exe_file;
1221 
1222 	exe_file = get_task_exe_file(current);
1223 	if (exe_file) {
1224 		pr_info("%pD: %s: potentially unexpected fatal signal %d.\n",
1225 			exe_file, current->comm, signr);
1226 		fput(exe_file);
1227 	} else {
1228 		pr_info("%s: potentially unexpected fatal signal %d.\n",
1229 			current->comm, signr);
1230 	}
1231 
1232 #if defined(__i386__) && !defined(__arch_um__)
1233 	pr_info("code at %08lx: ", regs->ip);
1234 	{
1235 		int i;
1236 		for (i = 0; i < 16; i++) {
1237 			unsigned char insn;
1238 
1239 			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1240 				break;
1241 			pr_cont("%02x ", insn);
1242 		}
1243 	}
1244 	pr_cont("\n");
1245 #endif
1246 	preempt_disable();
1247 	show_regs(regs);
1248 	preempt_enable();
1249 }
1250 
1251 static int __init setup_print_fatal_signals(char *str)
1252 {
1253 	get_option (&str, &print_fatal_signals);
1254 
1255 	return 1;
1256 }
1257 
1258 __setup("print-fatal-signals=", setup_print_fatal_signals);
1259 
1260 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1261 			enum pid_type type)
1262 {
1263 	unsigned long flags;
1264 	int ret = -ESRCH;
1265 
1266 	if (lock_task_sighand(p, &flags)) {
1267 		ret = send_signal_locked(sig, info, p, type);
1268 		unlock_task_sighand(p, &flags);
1269 	}
1270 
1271 	return ret;
1272 }
1273 
1274 enum sig_handler {
1275 	HANDLER_CURRENT, /* If reachable use the current handler */
1276 	HANDLER_SIG_DFL, /* Always use SIG_DFL handler semantics */
1277 	HANDLER_EXIT,	 /* Only visible as the process exit code */
1278 };
1279 
1280 /*
1281  * Force a signal that the process can't ignore: if necessary
1282  * we unblock the signal and change any SIG_IGN to SIG_DFL.
1283  *
1284  * Note: If we unblock the signal, we always reset it to SIG_DFL,
1285  * since we do not want to have a signal handler that was blocked
1286  * be invoked when user space had explicitly blocked it.
1287  *
1288  * We don't want to have recursive SIGSEGV's etc, for example,
1289  * that is why we also clear SIGNAL_UNKILLABLE.
1290  */
1291 static int
1292 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t,
1293 	enum sig_handler handler)
1294 {
1295 	unsigned long int flags;
1296 	int ret, blocked, ignored;
1297 	struct k_sigaction *action;
1298 	int sig = info->si_signo;
1299 
1300 	spin_lock_irqsave(&t->sighand->siglock, flags);
1301 	action = &t->sighand->action[sig-1];
1302 	ignored = action->sa.sa_handler == SIG_IGN;
1303 	blocked = sigismember(&t->blocked, sig);
1304 	if (blocked || ignored || (handler != HANDLER_CURRENT)) {
1305 		action->sa.sa_handler = SIG_DFL;
1306 		if (handler == HANDLER_EXIT)
1307 			action->sa.sa_flags |= SA_IMMUTABLE;
1308 		if (blocked)
1309 			sigdelset(&t->blocked, sig);
1310 	}
1311 	/*
1312 	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1313 	 * debugging to leave init killable. But HANDLER_EXIT is always fatal.
1314 	 */
1315 	if (action->sa.sa_handler == SIG_DFL &&
1316 	    (!t->ptrace || (handler == HANDLER_EXIT)))
1317 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1318 	ret = send_signal_locked(sig, info, t, PIDTYPE_PID);
1319 	/* This can happen if the signal was already pending and blocked */
1320 	if (!task_sigpending(t))
1321 		signal_wake_up(t, 0);
1322 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1323 
1324 	return ret;
1325 }
1326 
1327 int force_sig_info(struct kernel_siginfo *info)
1328 {
1329 	return force_sig_info_to_task(info, current, HANDLER_CURRENT);
1330 }
1331 
1332 /*
1333  * Nuke all other threads in the group.
1334  */
1335 int zap_other_threads(struct task_struct *p)
1336 {
1337 	struct task_struct *t;
1338 	int count = 0;
1339 
1340 	p->signal->group_stop_count = 0;
1341 
1342 	for_other_threads(p, t) {
1343 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1344 		count++;
1345 
1346 		/* Don't bother with already dead threads */
1347 		if (t->exit_state)
1348 			continue;
1349 		sigaddset(&t->pending.signal, SIGKILL);
1350 		signal_wake_up(t, 1);
1351 	}
1352 
1353 	return count;
1354 }
1355 
1356 struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
1357 					 unsigned long *flags)
1358 {
1359 	struct sighand_struct *sighand;
1360 
1361 	rcu_read_lock();
1362 	for (;;) {
1363 		sighand = rcu_dereference(tsk->sighand);
1364 		if (unlikely(sighand == NULL))
1365 			break;
1366 
1367 		/*
1368 		 * This sighand can be already freed and even reused, but
1369 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1370 		 * initializes ->siglock: this slab can't go away, it has
1371 		 * the same object type, ->siglock can't be reinitialized.
1372 		 *
1373 		 * We need to ensure that tsk->sighand is still the same
1374 		 * after we take the lock, we can race with de_thread() or
1375 		 * __exit_signal(). In the latter case the next iteration
1376 		 * must see ->sighand == NULL.
1377 		 */
1378 		spin_lock_irqsave(&sighand->siglock, *flags);
1379 		if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1380 			break;
1381 		spin_unlock_irqrestore(&sighand->siglock, *flags);
1382 	}
1383 	rcu_read_unlock();
1384 
1385 	return sighand;
1386 }
1387 
1388 #ifdef CONFIG_LOCKDEP
1389 void lockdep_assert_task_sighand_held(struct task_struct *task)
1390 {
1391 	struct sighand_struct *sighand;
1392 
1393 	rcu_read_lock();
1394 	sighand = rcu_dereference(task->sighand);
1395 	if (sighand)
1396 		lockdep_assert_held(&sighand->siglock);
1397 	else
1398 		WARN_ON_ONCE(1);
1399 	rcu_read_unlock();
1400 }
1401 #endif
1402 
1403 /*
1404  * send signal info to all the members of a thread group or to the
1405  * individual thread if type == PIDTYPE_PID.
1406  */
1407 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1408 			struct task_struct *p, enum pid_type type)
1409 {
1410 	int ret;
1411 
1412 	rcu_read_lock();
1413 	ret = check_kill_permission(sig, info, p);
1414 	rcu_read_unlock();
1415 
1416 	if (!ret && sig)
1417 		ret = do_send_sig_info(sig, info, p, type);
1418 
1419 	return ret;
1420 }
1421 
1422 /*
1423  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1424  * control characters do (^C, ^Z etc)
1425  * - the caller must hold at least a readlock on tasklist_lock
1426  */
1427 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1428 {
1429 	struct task_struct *p = NULL;
1430 	int ret = -ESRCH;
1431 
1432 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1433 		int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1434 		/*
1435 		 * If group_send_sig_info() succeeds at least once ret
1436 		 * becomes 0 and after that the code below has no effect.
1437 		 * Otherwise we return the last err or -ESRCH if this
1438 		 * process group is empty.
1439 		 */
1440 		if (ret)
1441 			ret = err;
1442 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1443 
1444 	return ret;
1445 }
1446 
1447 static int kill_pid_info_type(int sig, struct kernel_siginfo *info,
1448 				struct pid *pid, enum pid_type type)
1449 {
1450 	int error = -ESRCH;
1451 	struct task_struct *p;
1452 
1453 	for (;;) {
1454 		rcu_read_lock();
1455 		p = pid_task(pid, PIDTYPE_PID);
1456 		if (p)
1457 			error = group_send_sig_info(sig, info, p, type);
1458 		rcu_read_unlock();
1459 		if (likely(!p || error != -ESRCH))
1460 			return error;
1461 		/*
1462 		 * The task was unhashed in between, try again.  If it
1463 		 * is dead, pid_task() will return NULL, if we race with
1464 		 * de_thread() it will find the new leader.
1465 		 */
1466 	}
1467 }
1468 
1469 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1470 {
1471 	return kill_pid_info_type(sig, info, pid, PIDTYPE_TGID);
1472 }
1473 
1474 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1475 {
1476 	int error;
1477 	rcu_read_lock();
1478 	error = kill_pid_info(sig, info, find_vpid(pid));
1479 	rcu_read_unlock();
1480 	return error;
1481 }
1482 
1483 static inline bool kill_as_cred_perm(const struct cred *cred,
1484 				     struct task_struct *target)
1485 {
1486 	const struct cred *pcred = __task_cred(target);
1487 
1488 	return uid_eq(cred->euid, pcred->suid) ||
1489 	       uid_eq(cred->euid, pcred->uid) ||
1490 	       uid_eq(cred->uid, pcred->suid) ||
1491 	       uid_eq(cred->uid, pcred->uid);
1492 }
1493 
1494 /*
1495  * The usb asyncio usage of siginfo is wrong.  The glibc support
1496  * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1497  * AKA after the generic fields:
1498  *	kernel_pid_t	si_pid;
1499  *	kernel_uid32_t	si_uid;
1500  *	sigval_t	si_value;
1501  *
1502  * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1503  * after the generic fields is:
1504  *	void __user 	*si_addr;
1505  *
1506  * This is a practical problem when there is a 64bit big endian kernel
1507  * and a 32bit userspace.  As the 32bit address will encoded in the low
1508  * 32bits of the pointer.  Those low 32bits will be stored at higher
1509  * address than appear in a 32 bit pointer.  So userspace will not
1510  * see the address it was expecting for it's completions.
1511  *
1512  * There is nothing in the encoding that can allow
1513  * copy_siginfo_to_user32 to detect this confusion of formats, so
1514  * handle this by requiring the caller of kill_pid_usb_asyncio to
1515  * notice when this situration takes place and to store the 32bit
1516  * pointer in sival_int, instead of sival_addr of the sigval_t addr
1517  * parameter.
1518  */
1519 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1520 			 struct pid *pid, const struct cred *cred)
1521 {
1522 	struct kernel_siginfo info;
1523 	struct task_struct *p;
1524 	unsigned long flags;
1525 	int ret = -EINVAL;
1526 
1527 	if (!valid_signal(sig))
1528 		return ret;
1529 
1530 	clear_siginfo(&info);
1531 	info.si_signo = sig;
1532 	info.si_errno = errno;
1533 	info.si_code = SI_ASYNCIO;
1534 	*((sigval_t *)&info.si_pid) = addr;
1535 
1536 	rcu_read_lock();
1537 	p = pid_task(pid, PIDTYPE_PID);
1538 	if (!p) {
1539 		ret = -ESRCH;
1540 		goto out_unlock;
1541 	}
1542 	if (!kill_as_cred_perm(cred, p)) {
1543 		ret = -EPERM;
1544 		goto out_unlock;
1545 	}
1546 	ret = security_task_kill(p, &info, sig, cred);
1547 	if (ret)
1548 		goto out_unlock;
1549 
1550 	if (sig) {
1551 		if (lock_task_sighand(p, &flags)) {
1552 			ret = __send_signal_locked(sig, &info, p, PIDTYPE_TGID, false);
1553 			unlock_task_sighand(p, &flags);
1554 		} else
1555 			ret = -ESRCH;
1556 	}
1557 out_unlock:
1558 	rcu_read_unlock();
1559 	return ret;
1560 }
1561 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1562 
1563 /*
1564  * kill_something_info() interprets pid in interesting ways just like kill(2).
1565  *
1566  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1567  * is probably wrong.  Should make it like BSD or SYSV.
1568  */
1569 
1570 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1571 {
1572 	int ret;
1573 
1574 	if (pid > 0)
1575 		return kill_proc_info(sig, info, pid);
1576 
1577 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1578 	if (pid == INT_MIN)
1579 		return -ESRCH;
1580 
1581 	read_lock(&tasklist_lock);
1582 	if (pid != -1) {
1583 		ret = __kill_pgrp_info(sig, info,
1584 				pid ? find_vpid(-pid) : task_pgrp(current));
1585 	} else {
1586 		int retval = 0, count = 0;
1587 		struct task_struct * p;
1588 
1589 		for_each_process(p) {
1590 			if (task_pid_vnr(p) > 1 &&
1591 					!same_thread_group(p, current)) {
1592 				int err = group_send_sig_info(sig, info, p,
1593 							      PIDTYPE_MAX);
1594 				++count;
1595 				if (err != -EPERM)
1596 					retval = err;
1597 			}
1598 		}
1599 		ret = count ? retval : -ESRCH;
1600 	}
1601 	read_unlock(&tasklist_lock);
1602 
1603 	return ret;
1604 }
1605 
1606 /*
1607  * These are for backward compatibility with the rest of the kernel source.
1608  */
1609 
1610 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1611 {
1612 	/*
1613 	 * Make sure legacy kernel users don't send in bad values
1614 	 * (normal paths check this in check_kill_permission).
1615 	 */
1616 	if (!valid_signal(sig))
1617 		return -EINVAL;
1618 
1619 	return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1620 }
1621 EXPORT_SYMBOL(send_sig_info);
1622 
1623 #define __si_special(priv) \
1624 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1625 
1626 int
1627 send_sig(int sig, struct task_struct *p, int priv)
1628 {
1629 	return send_sig_info(sig, __si_special(priv), p);
1630 }
1631 EXPORT_SYMBOL(send_sig);
1632 
1633 void force_sig(int sig)
1634 {
1635 	struct kernel_siginfo info;
1636 
1637 	clear_siginfo(&info);
1638 	info.si_signo = sig;
1639 	info.si_errno = 0;
1640 	info.si_code = SI_KERNEL;
1641 	info.si_pid = 0;
1642 	info.si_uid = 0;
1643 	force_sig_info(&info);
1644 }
1645 EXPORT_SYMBOL(force_sig);
1646 
1647 void force_fatal_sig(int sig)
1648 {
1649 	struct kernel_siginfo info;
1650 
1651 	clear_siginfo(&info);
1652 	info.si_signo = sig;
1653 	info.si_errno = 0;
1654 	info.si_code = SI_KERNEL;
1655 	info.si_pid = 0;
1656 	info.si_uid = 0;
1657 	force_sig_info_to_task(&info, current, HANDLER_SIG_DFL);
1658 }
1659 
1660 void force_exit_sig(int sig)
1661 {
1662 	struct kernel_siginfo info;
1663 
1664 	clear_siginfo(&info);
1665 	info.si_signo = sig;
1666 	info.si_errno = 0;
1667 	info.si_code = SI_KERNEL;
1668 	info.si_pid = 0;
1669 	info.si_uid = 0;
1670 	force_sig_info_to_task(&info, current, HANDLER_EXIT);
1671 }
1672 
1673 /*
1674  * When things go south during signal handling, we
1675  * will force a SIGSEGV. And if the signal that caused
1676  * the problem was already a SIGSEGV, we'll want to
1677  * make sure we don't even try to deliver the signal..
1678  */
1679 void force_sigsegv(int sig)
1680 {
1681 	if (sig == SIGSEGV)
1682 		force_fatal_sig(SIGSEGV);
1683 	else
1684 		force_sig(SIGSEGV);
1685 }
1686 
1687 int force_sig_fault_to_task(int sig, int code, void __user *addr,
1688 			    struct task_struct *t)
1689 {
1690 	struct kernel_siginfo info;
1691 
1692 	clear_siginfo(&info);
1693 	info.si_signo = sig;
1694 	info.si_errno = 0;
1695 	info.si_code  = code;
1696 	info.si_addr  = addr;
1697 	return force_sig_info_to_task(&info, t, HANDLER_CURRENT);
1698 }
1699 
1700 int force_sig_fault(int sig, int code, void __user *addr)
1701 {
1702 	return force_sig_fault_to_task(sig, code, addr, current);
1703 }
1704 
1705 int send_sig_fault(int sig, int code, void __user *addr, struct task_struct *t)
1706 {
1707 	struct kernel_siginfo info;
1708 
1709 	clear_siginfo(&info);
1710 	info.si_signo = sig;
1711 	info.si_errno = 0;
1712 	info.si_code  = code;
1713 	info.si_addr  = addr;
1714 	return send_sig_info(info.si_signo, &info, t);
1715 }
1716 
1717 int force_sig_mceerr(int code, void __user *addr, short lsb)
1718 {
1719 	struct kernel_siginfo info;
1720 
1721 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1722 	clear_siginfo(&info);
1723 	info.si_signo = SIGBUS;
1724 	info.si_errno = 0;
1725 	info.si_code = code;
1726 	info.si_addr = addr;
1727 	info.si_addr_lsb = lsb;
1728 	return force_sig_info(&info);
1729 }
1730 
1731 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1732 {
1733 	struct kernel_siginfo info;
1734 
1735 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1736 	clear_siginfo(&info);
1737 	info.si_signo = SIGBUS;
1738 	info.si_errno = 0;
1739 	info.si_code = code;
1740 	info.si_addr = addr;
1741 	info.si_addr_lsb = lsb;
1742 	return send_sig_info(info.si_signo, &info, t);
1743 }
1744 EXPORT_SYMBOL(send_sig_mceerr);
1745 
1746 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1747 {
1748 	struct kernel_siginfo info;
1749 
1750 	clear_siginfo(&info);
1751 	info.si_signo = SIGSEGV;
1752 	info.si_errno = 0;
1753 	info.si_code  = SEGV_BNDERR;
1754 	info.si_addr  = addr;
1755 	info.si_lower = lower;
1756 	info.si_upper = upper;
1757 	return force_sig_info(&info);
1758 }
1759 
1760 #ifdef SEGV_PKUERR
1761 int force_sig_pkuerr(void __user *addr, u32 pkey)
1762 {
1763 	struct kernel_siginfo info;
1764 
1765 	clear_siginfo(&info);
1766 	info.si_signo = SIGSEGV;
1767 	info.si_errno = 0;
1768 	info.si_code  = SEGV_PKUERR;
1769 	info.si_addr  = addr;
1770 	info.si_pkey  = pkey;
1771 	return force_sig_info(&info);
1772 }
1773 #endif
1774 
1775 int send_sig_perf(void __user *addr, u32 type, u64 sig_data)
1776 {
1777 	struct kernel_siginfo info;
1778 
1779 	clear_siginfo(&info);
1780 	info.si_signo     = SIGTRAP;
1781 	info.si_errno     = 0;
1782 	info.si_code      = TRAP_PERF;
1783 	info.si_addr      = addr;
1784 	info.si_perf_data = sig_data;
1785 	info.si_perf_type = type;
1786 
1787 	/*
1788 	 * Signals generated by perf events should not terminate the whole
1789 	 * process if SIGTRAP is blocked, however, delivering the signal
1790 	 * asynchronously is better than not delivering at all. But tell user
1791 	 * space if the signal was asynchronous, so it can clearly be
1792 	 * distinguished from normal synchronous ones.
1793 	 */
1794 	info.si_perf_flags = sigismember(&current->blocked, info.si_signo) ?
1795 				     TRAP_PERF_FLAG_ASYNC :
1796 				     0;
1797 
1798 	return send_sig_info(info.si_signo, &info, current);
1799 }
1800 
1801 /**
1802  * force_sig_seccomp - signals the task to allow in-process syscall emulation
1803  * @syscall: syscall number to send to userland
1804  * @reason: filter-supplied reason code to send to userland (via si_errno)
1805  * @force_coredump: true to trigger a coredump
1806  *
1807  * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
1808  */
1809 int force_sig_seccomp(int syscall, int reason, bool force_coredump)
1810 {
1811 	struct kernel_siginfo info;
1812 
1813 	clear_siginfo(&info);
1814 	info.si_signo = SIGSYS;
1815 	info.si_code = SYS_SECCOMP;
1816 	info.si_call_addr = (void __user *)KSTK_EIP(current);
1817 	info.si_errno = reason;
1818 	info.si_arch = syscall_get_arch(current);
1819 	info.si_syscall = syscall;
1820 	return force_sig_info_to_task(&info, current,
1821 		force_coredump ? HANDLER_EXIT : HANDLER_CURRENT);
1822 }
1823 
1824 /* For the crazy architectures that include trap information in
1825  * the errno field, instead of an actual errno value.
1826  */
1827 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1828 {
1829 	struct kernel_siginfo info;
1830 
1831 	clear_siginfo(&info);
1832 	info.si_signo = SIGTRAP;
1833 	info.si_errno = errno;
1834 	info.si_code  = TRAP_HWBKPT;
1835 	info.si_addr  = addr;
1836 	return force_sig_info(&info);
1837 }
1838 
1839 /* For the rare architectures that include trap information using
1840  * si_trapno.
1841  */
1842 int force_sig_fault_trapno(int sig, int code, void __user *addr, int trapno)
1843 {
1844 	struct kernel_siginfo info;
1845 
1846 	clear_siginfo(&info);
1847 	info.si_signo = sig;
1848 	info.si_errno = 0;
1849 	info.si_code  = code;
1850 	info.si_addr  = addr;
1851 	info.si_trapno = trapno;
1852 	return force_sig_info(&info);
1853 }
1854 
1855 /* For the rare architectures that include trap information using
1856  * si_trapno.
1857  */
1858 int send_sig_fault_trapno(int sig, int code, void __user *addr, int trapno,
1859 			  struct task_struct *t)
1860 {
1861 	struct kernel_siginfo info;
1862 
1863 	clear_siginfo(&info);
1864 	info.si_signo = sig;
1865 	info.si_errno = 0;
1866 	info.si_code  = code;
1867 	info.si_addr  = addr;
1868 	info.si_trapno = trapno;
1869 	return send_sig_info(info.si_signo, &info, t);
1870 }
1871 
1872 static int kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1873 {
1874 	int ret;
1875 	read_lock(&tasklist_lock);
1876 	ret = __kill_pgrp_info(sig, info, pgrp);
1877 	read_unlock(&tasklist_lock);
1878 	return ret;
1879 }
1880 
1881 int kill_pgrp(struct pid *pid, int sig, int priv)
1882 {
1883 	return kill_pgrp_info(sig, __si_special(priv), pid);
1884 }
1885 EXPORT_SYMBOL(kill_pgrp);
1886 
1887 int kill_pid(struct pid *pid, int sig, int priv)
1888 {
1889 	return kill_pid_info(sig, __si_special(priv), pid);
1890 }
1891 EXPORT_SYMBOL(kill_pid);
1892 
1893 #ifdef CONFIG_POSIX_TIMERS
1894 /*
1895  * These functions handle POSIX timer signals. POSIX timers use
1896  * preallocated sigqueue structs for sending signals.
1897  */
1898 static void __flush_itimer_signals(struct sigpending *pending)
1899 {
1900 	sigset_t signal, retain;
1901 	struct sigqueue *q, *n;
1902 
1903 	signal = pending->signal;
1904 	sigemptyset(&retain);
1905 
1906 	list_for_each_entry_safe(q, n, &pending->list, list) {
1907 		int sig = q->info.si_signo;
1908 
1909 		if (likely(q->info.si_code != SI_TIMER)) {
1910 			sigaddset(&retain, sig);
1911 		} else {
1912 			sigdelset(&signal, sig);
1913 			list_del_init(&q->list);
1914 			__sigqueue_free(q);
1915 		}
1916 	}
1917 
1918 	sigorsets(&pending->signal, &signal, &retain);
1919 }
1920 
1921 void flush_itimer_signals(void)
1922 {
1923 	struct task_struct *tsk = current;
1924 
1925 	guard(spinlock_irqsave)(&tsk->sighand->siglock);
1926 	__flush_itimer_signals(&tsk->pending);
1927 	__flush_itimer_signals(&tsk->signal->shared_pending);
1928 }
1929 
1930 bool posixtimer_init_sigqueue(struct sigqueue *q)
1931 {
1932 	struct ucounts *ucounts = sig_get_ucounts(current, -1, 0);
1933 
1934 	if (!ucounts)
1935 		return false;
1936 	clear_siginfo(&q->info);
1937 	__sigqueue_init(q, ucounts, SIGQUEUE_PREALLOC);
1938 	return true;
1939 }
1940 
1941 static void posixtimer_queue_sigqueue(struct sigqueue *q, struct task_struct *t, enum pid_type type)
1942 {
1943 	struct sigpending *pending;
1944 	int sig = q->info.si_signo;
1945 
1946 	signalfd_notify(t, sig);
1947 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1948 	list_add_tail(&q->list, &pending->list);
1949 	sigaddset(&pending->signal, sig);
1950 	complete_signal(sig, t, type);
1951 }
1952 
1953 /*
1954  * This function is used by POSIX timers to deliver a timer signal.
1955  * Where type is PIDTYPE_PID (such as for timers with SIGEV_THREAD_ID
1956  * set), the signal must be delivered to the specific thread (queues
1957  * into t->pending).
1958  *
1959  * Where type is not PIDTYPE_PID, signals must be delivered to the
1960  * process. In this case, prefer to deliver to current if it is in
1961  * the same thread group as the target process and its sighand is
1962  * stable, which avoids unnecessarily waking up a potentially idle task.
1963  */
1964 static inline struct task_struct *posixtimer_get_target(struct k_itimer *tmr)
1965 {
1966 	struct task_struct *t = pid_task(tmr->it_pid, tmr->it_pid_type);
1967 
1968 	if (t && tmr->it_pid_type != PIDTYPE_PID &&
1969 	    same_thread_group(t, current) && !current->exit_state)
1970 		t = current;
1971 	return t;
1972 }
1973 
1974 void posixtimer_send_sigqueue(struct k_itimer *tmr)
1975 {
1976 	struct sigqueue *q = &tmr->sigq;
1977 	int sig = q->info.si_signo;
1978 	struct task_struct *t;
1979 	unsigned long flags;
1980 	int result;
1981 
1982 	guard(rcu)();
1983 
1984 	t = posixtimer_get_target(tmr);
1985 	if (!t)
1986 		return;
1987 
1988 	if (!likely(lock_task_sighand(t, &flags)))
1989 		return;
1990 
1991 	/*
1992 	 * Update @tmr::sigqueue_seq for posix timer signals with sighand
1993 	 * locked to prevent a race against dequeue_signal().
1994 	 */
1995 	tmr->it_sigqueue_seq = tmr->it_signal_seq;
1996 
1997 	/*
1998 	 * Set the signal delivery status under sighand lock, so that the
1999 	 * ignored signal handling can distinguish between a periodic and a
2000 	 * non-periodic timer.
2001 	 */
2002 	tmr->it_sig_periodic = tmr->it_status == POSIX_TIMER_REQUEUE_PENDING;
2003 
2004 	if (!prepare_signal(sig, t, false)) {
2005 		result = TRACE_SIGNAL_IGNORED;
2006 
2007 		if (!list_empty(&q->list)) {
2008 			/*
2009 			 * The signal was ignored and blocked. The timer
2010 			 * expiry queued it because blocked signals are
2011 			 * queued independent of the ignored state.
2012 			 *
2013 			 * The unblocking set SIGPENDING, but the signal
2014 			 * was not yet dequeued from the pending list.
2015 			 * So prepare_signal() sees unblocked and ignored,
2016 			 * which ends up here. Leave it queued like a
2017 			 * regular signal.
2018 			 *
2019 			 * The same happens when the task group is exiting
2020 			 * and the signal is already queued.
2021 			 * prepare_signal() treats SIGNAL_GROUP_EXIT as
2022 			 * ignored independent of its queued state. This
2023 			 * gets cleaned up in __exit_signal().
2024 			 */
2025 			goto out;
2026 		}
2027 
2028 		/* Periodic timers with SIG_IGN are queued on the ignored list */
2029 		if (tmr->it_sig_periodic) {
2030 			/*
2031 			 * Already queued means the timer was rearmed after
2032 			 * the previous expiry got it on the ignore list.
2033 			 * Nothing to do for that case.
2034 			 */
2035 			if (hlist_unhashed(&tmr->ignored_list)) {
2036 				/*
2037 				 * Take a signal reference and queue it on
2038 				 * the ignored list.
2039 				 */
2040 				posixtimer_sigqueue_getref(q);
2041 				posixtimer_sig_ignore(t, q);
2042 			}
2043 		} else if (!hlist_unhashed(&tmr->ignored_list)) {
2044 			/*
2045 			 * Covers the case where a timer was periodic and
2046 			 * then the signal was ignored. Later it was rearmed
2047 			 * as oneshot timer. The previous signal is invalid
2048 			 * now, and this oneshot signal has to be dropped.
2049 			 * Remove it from the ignored list and drop the
2050 			 * reference count as the signal is not longer
2051 			 * queued.
2052 			 */
2053 			hlist_del_init(&tmr->ignored_list);
2054 			posixtimer_putref(tmr);
2055 		}
2056 		goto out;
2057 	}
2058 
2059 	if (unlikely(!list_empty(&q->list))) {
2060 		/* This holds a reference count already */
2061 		result = TRACE_SIGNAL_ALREADY_PENDING;
2062 		goto out;
2063 	}
2064 
2065 	/*
2066 	 * If the signal is on the ignore list, it got blocked after it was
2067 	 * ignored earlier. But nothing lifted the ignore. Move it back to
2068 	 * the pending list to be consistent with the regular signal
2069 	 * handling. This already holds a reference count.
2070 	 *
2071 	 * If it's not on the ignore list acquire a reference count.
2072 	 */
2073 	if (likely(hlist_unhashed(&tmr->ignored_list)))
2074 		posixtimer_sigqueue_getref(q);
2075 	else
2076 		hlist_del_init(&tmr->ignored_list);
2077 
2078 	posixtimer_queue_sigqueue(q, t, tmr->it_pid_type);
2079 	result = TRACE_SIGNAL_DELIVERED;
2080 out:
2081 	trace_signal_generate(sig, &q->info, t, tmr->it_pid_type != PIDTYPE_PID, result);
2082 	unlock_task_sighand(t, &flags);
2083 }
2084 
2085 static inline void posixtimer_sig_ignore(struct task_struct *tsk, struct sigqueue *q)
2086 {
2087 	struct k_itimer *tmr = container_of(q, struct k_itimer, sigq);
2088 
2089 	/*
2090 	 * If the timer is marked deleted already or the signal originates
2091 	 * from a non-periodic timer, then just drop the reference
2092 	 * count. Otherwise queue it on the ignored list.
2093 	 */
2094 	if (posixtimer_valid(tmr) && tmr->it_sig_periodic)
2095 		hlist_add_head(&tmr->ignored_list, &tsk->signal->ignored_posix_timers);
2096 	else
2097 		posixtimer_putref(tmr);
2098 }
2099 
2100 static void posixtimer_sig_unignore(struct task_struct *tsk, int sig)
2101 {
2102 	struct hlist_head *head = &tsk->signal->ignored_posix_timers;
2103 	struct hlist_node *tmp;
2104 	struct k_itimer *tmr;
2105 
2106 	if (likely(hlist_empty(head)))
2107 		return;
2108 
2109 	/*
2110 	 * Rearming a timer with sighand lock held is not possible due to
2111 	 * lock ordering vs. tmr::it_lock. Just stick the sigqueue back and
2112 	 * let the signal delivery path deal with it whether it needs to be
2113 	 * rearmed or not. This cannot be decided here w/o dropping sighand
2114 	 * lock and creating a loop retry horror show.
2115 	 */
2116 	hlist_for_each_entry_safe(tmr, tmp , head, ignored_list) {
2117 		struct task_struct *target;
2118 
2119 		/*
2120 		 * tmr::sigq.info.si_signo is immutable, so accessing it
2121 		 * without holding tmr::it_lock is safe.
2122 		 */
2123 		if (tmr->sigq.info.si_signo != sig)
2124 			continue;
2125 
2126 		hlist_del_init(&tmr->ignored_list);
2127 
2128 		/* This should never happen and leaks a reference count */
2129 		if (WARN_ON_ONCE(!list_empty(&tmr->sigq.list)))
2130 			continue;
2131 
2132 		/*
2133 		 * Get the target for the signal. If target is a thread and
2134 		 * has exited by now, drop the reference count.
2135 		 */
2136 		guard(rcu)();
2137 		target = posixtimer_get_target(tmr);
2138 		if (target)
2139 			posixtimer_queue_sigqueue(&tmr->sigq, target, tmr->it_pid_type);
2140 		else
2141 			posixtimer_putref(tmr);
2142 	}
2143 }
2144 #else /* CONFIG_POSIX_TIMERS */
2145 static inline void posixtimer_sig_ignore(struct task_struct *tsk, struct sigqueue *q) { }
2146 static inline void posixtimer_sig_unignore(struct task_struct *tsk, int sig) { }
2147 #endif /* !CONFIG_POSIX_TIMERS */
2148 
2149 void do_notify_pidfd(struct task_struct *task)
2150 {
2151 	struct pid *pid = task_pid(task);
2152 
2153 	WARN_ON(task->exit_state == 0);
2154 
2155 	__wake_up(&pid->wait_pidfd, TASK_NORMAL, 0,
2156 			poll_to_key(EPOLLIN | EPOLLRDNORM));
2157 }
2158 
2159 /*
2160  * Let a parent know about the death of a child.
2161  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
2162  *
2163  * Returns true if our parent ignored us and so we've switched to
2164  * self-reaping.
2165  */
2166 bool do_notify_parent(struct task_struct *tsk, int sig)
2167 {
2168 	struct kernel_siginfo info;
2169 	unsigned long flags;
2170 	struct sighand_struct *psig;
2171 	bool autoreap = false;
2172 	u64 utime, stime;
2173 
2174 	if (WARN_ON_ONCE(!valid_signal(sig)))
2175 		return false;
2176 
2177 	/* do_notify_parent_cldstop should have been called instead.  */
2178 	WARN_ON_ONCE(task_is_stopped_or_traced(tsk));
2179 
2180 	WARN_ON_ONCE(!tsk->ptrace && !thread_group_empty(tsk));
2181 
2182 	/* ptraced, or group-leader without sub-threads */
2183 	do_notify_pidfd(tsk);
2184 
2185 	if (sig != SIGCHLD) {
2186 		/*
2187 		 * This is only possible if parent == real_parent.
2188 		 * Check if it has changed security domain.
2189 		 */
2190 		if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
2191 			sig = SIGCHLD;
2192 	}
2193 
2194 	clear_siginfo(&info);
2195 	info.si_signo = sig;
2196 	info.si_errno = 0;
2197 	/*
2198 	 * We are under tasklist_lock here so our parent is tied to
2199 	 * us and cannot change.
2200 	 *
2201 	 * task_active_pid_ns will always return the same pid namespace
2202 	 * until a task passes through release_task.
2203 	 *
2204 	 * write_lock() currently calls preempt_disable() which is the
2205 	 * same as rcu_read_lock(), but according to Oleg, this is not
2206 	 * correct to rely on this
2207 	 */
2208 	rcu_read_lock();
2209 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
2210 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
2211 				       task_uid(tsk));
2212 	rcu_read_unlock();
2213 
2214 	task_cputime(tsk, &utime, &stime);
2215 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
2216 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
2217 
2218 	info.si_status = tsk->exit_code & 0x7f;
2219 	if (tsk->exit_code & 0x80)
2220 		info.si_code = CLD_DUMPED;
2221 	else if (tsk->exit_code & 0x7f)
2222 		info.si_code = CLD_KILLED;
2223 	else {
2224 		info.si_code = CLD_EXITED;
2225 		info.si_status = tsk->exit_code >> 8;
2226 	}
2227 
2228 	psig = tsk->parent->sighand;
2229 	spin_lock_irqsave(&psig->siglock, flags);
2230 	if (!tsk->ptrace && sig == SIGCHLD &&
2231 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
2232 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
2233 		/*
2234 		 * We are exiting and our parent doesn't care.  POSIX.1
2235 		 * defines special semantics for setting SIGCHLD to SIG_IGN
2236 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
2237 		 * automatically and not left for our parent's wait4 call.
2238 		 * Rather than having the parent do it as a magic kind of
2239 		 * signal handler, we just set this to tell do_exit that we
2240 		 * can be cleaned up without becoming a zombie.  Note that
2241 		 * we still call __wake_up_parent in this case, because a
2242 		 * blocked sys_wait4 might now return -ECHILD.
2243 		 *
2244 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
2245 		 * is implementation-defined: we do (if you don't want
2246 		 * it, just use SIG_IGN instead).
2247 		 */
2248 		autoreap = true;
2249 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2250 			sig = 0;
2251 	}
2252 	/*
2253 	 * Send with __send_signal as si_pid and si_uid are in the
2254 	 * parent's namespaces.
2255 	 */
2256 	if (sig)
2257 		__send_signal_locked(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2258 	__wake_up_parent(tsk, tsk->parent);
2259 	spin_unlock_irqrestore(&psig->siglock, flags);
2260 
2261 	return autoreap;
2262 }
2263 
2264 /**
2265  * do_notify_parent_cldstop - notify parent of stopped/continued state change
2266  * @tsk: task reporting the state change
2267  * @for_ptracer: the notification is for ptracer
2268  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2269  *
2270  * Notify @tsk's parent that the stopped/continued state has changed.  If
2271  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2272  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2273  *
2274  * CONTEXT:
2275  * Must be called with tasklist_lock at least read locked.
2276  */
2277 static void do_notify_parent_cldstop(struct task_struct *tsk,
2278 				     bool for_ptracer, int why)
2279 {
2280 	struct kernel_siginfo info;
2281 	unsigned long flags;
2282 	struct task_struct *parent;
2283 	struct sighand_struct *sighand;
2284 	u64 utime, stime;
2285 
2286 	if (for_ptracer) {
2287 		parent = tsk->parent;
2288 	} else {
2289 		tsk = tsk->group_leader;
2290 		parent = tsk->real_parent;
2291 	}
2292 
2293 	clear_siginfo(&info);
2294 	info.si_signo = SIGCHLD;
2295 	info.si_errno = 0;
2296 	/*
2297 	 * see comment in do_notify_parent() about the following 4 lines
2298 	 */
2299 	rcu_read_lock();
2300 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2301 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2302 	rcu_read_unlock();
2303 
2304 	task_cputime(tsk, &utime, &stime);
2305 	info.si_utime = nsec_to_clock_t(utime);
2306 	info.si_stime = nsec_to_clock_t(stime);
2307 
2308  	info.si_code = why;
2309  	switch (why) {
2310  	case CLD_CONTINUED:
2311  		info.si_status = SIGCONT;
2312  		break;
2313  	case CLD_STOPPED:
2314  		info.si_status = tsk->signal->group_exit_code & 0x7f;
2315  		break;
2316  	case CLD_TRAPPED:
2317  		info.si_status = tsk->exit_code & 0x7f;
2318  		break;
2319  	default:
2320  		BUG();
2321  	}
2322 
2323 	sighand = parent->sighand;
2324 	spin_lock_irqsave(&sighand->siglock, flags);
2325 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2326 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2327 		send_signal_locked(SIGCHLD, &info, parent, PIDTYPE_TGID);
2328 	/*
2329 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2330 	 */
2331 	__wake_up_parent(tsk, parent);
2332 	spin_unlock_irqrestore(&sighand->siglock, flags);
2333 }
2334 
2335 /*
2336  * This must be called with current->sighand->siglock held.
2337  *
2338  * This should be the path for all ptrace stops.
2339  * We always set current->last_siginfo while stopped here.
2340  * That makes it a way to test a stopped process for
2341  * being ptrace-stopped vs being job-control-stopped.
2342  *
2343  * Returns the signal the ptracer requested the code resume
2344  * with.  If the code did not stop because the tracer is gone,
2345  * the stop signal remains unchanged unless clear_code.
2346  */
2347 static int ptrace_stop(int exit_code, int why, unsigned long message,
2348 		       kernel_siginfo_t *info)
2349 	__releases(&current->sighand->siglock)
2350 	__acquires(&current->sighand->siglock)
2351 {
2352 	bool gstop_done = false;
2353 
2354 	if (arch_ptrace_stop_needed()) {
2355 		/*
2356 		 * The arch code has something special to do before a
2357 		 * ptrace stop.  This is allowed to block, e.g. for faults
2358 		 * on user stack pages.  We can't keep the siglock while
2359 		 * calling arch_ptrace_stop, so we must release it now.
2360 		 * To preserve proper semantics, we must do this before
2361 		 * any signal bookkeeping like checking group_stop_count.
2362 		 */
2363 		spin_unlock_irq(&current->sighand->siglock);
2364 		arch_ptrace_stop();
2365 		spin_lock_irq(&current->sighand->siglock);
2366 	}
2367 
2368 	/*
2369 	 * After this point ptrace_signal_wake_up or signal_wake_up
2370 	 * will clear TASK_TRACED if ptrace_unlink happens or a fatal
2371 	 * signal comes in.  Handle previous ptrace_unlinks and fatal
2372 	 * signals here to prevent ptrace_stop sleeping in schedule.
2373 	 */
2374 	if (!current->ptrace || __fatal_signal_pending(current))
2375 		return exit_code;
2376 
2377 	set_special_state(TASK_TRACED);
2378 	current->jobctl |= JOBCTL_TRACED;
2379 
2380 	/*
2381 	 * We're committing to trapping.  TRACED should be visible before
2382 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2383 	 * Also, transition to TRACED and updates to ->jobctl should be
2384 	 * atomic with respect to siglock and should be done after the arch
2385 	 * hook as siglock is released and regrabbed across it.
2386 	 *
2387 	 *     TRACER				    TRACEE
2388 	 *
2389 	 *     ptrace_attach()
2390 	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2391 	 *     do_wait()
2392 	 *       set_current_state()                smp_wmb();
2393 	 *       ptrace_do_wait()
2394 	 *         wait_task_stopped()
2395 	 *           task_stopped_code()
2396 	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2397 	 */
2398 	smp_wmb();
2399 
2400 	current->ptrace_message = message;
2401 	current->last_siginfo = info;
2402 	current->exit_code = exit_code;
2403 
2404 	/*
2405 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2406 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2407 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2408 	 * could be clear now.  We act as if SIGCONT is received after
2409 	 * TASK_TRACED is entered - ignore it.
2410 	 */
2411 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2412 		gstop_done = task_participate_group_stop(current);
2413 
2414 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2415 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2416 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2417 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2418 
2419 	/* entering a trap, clear TRAPPING */
2420 	task_clear_jobctl_trapping(current);
2421 
2422 	spin_unlock_irq(&current->sighand->siglock);
2423 	read_lock(&tasklist_lock);
2424 	/*
2425 	 * Notify parents of the stop.
2426 	 *
2427 	 * While ptraced, there are two parents - the ptracer and
2428 	 * the real_parent of the group_leader.  The ptracer should
2429 	 * know about every stop while the real parent is only
2430 	 * interested in the completion of group stop.  The states
2431 	 * for the two don't interact with each other.  Notify
2432 	 * separately unless they're gonna be duplicates.
2433 	 */
2434 	if (current->ptrace)
2435 		do_notify_parent_cldstop(current, true, why);
2436 	if (gstop_done && (!current->ptrace || ptrace_reparented(current)))
2437 		do_notify_parent_cldstop(current, false, why);
2438 
2439 	/*
2440 	 * The previous do_notify_parent_cldstop() invocation woke ptracer.
2441 	 * One a PREEMPTION kernel this can result in preemption requirement
2442 	 * which will be fulfilled after read_unlock() and the ptracer will be
2443 	 * put on the CPU.
2444 	 * The ptracer is in wait_task_inactive(, __TASK_TRACED) waiting for
2445 	 * this task wait in schedule(). If this task gets preempted then it
2446 	 * remains enqueued on the runqueue. The ptracer will observe this and
2447 	 * then sleep for a delay of one HZ tick. In the meantime this task
2448 	 * gets scheduled, enters schedule() and will wait for the ptracer.
2449 	 *
2450 	 * This preemption point is not bad from a correctness point of
2451 	 * view but extends the runtime by one HZ tick time due to the
2452 	 * ptracer's sleep.  The preempt-disable section ensures that there
2453 	 * will be no preemption between unlock and schedule() and so
2454 	 * improving the performance since the ptracer will observe that
2455 	 * the tracee is scheduled out once it gets on the CPU.
2456 	 *
2457 	 * On PREEMPT_RT locking tasklist_lock does not disable preemption.
2458 	 * Therefore the task can be preempted after do_notify_parent_cldstop()
2459 	 * before unlocking tasklist_lock so there is no benefit in doing this.
2460 	 *
2461 	 * In fact disabling preemption is harmful on PREEMPT_RT because
2462 	 * the spinlock_t in cgroup_enter_frozen() must not be acquired
2463 	 * with preemption disabled due to the 'sleeping' spinlock
2464 	 * substitution of RT.
2465 	 */
2466 	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
2467 		preempt_disable();
2468 	read_unlock(&tasklist_lock);
2469 	cgroup_enter_frozen();
2470 	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
2471 		preempt_enable_no_resched();
2472 	schedule();
2473 	cgroup_leave_frozen(true);
2474 
2475 	/*
2476 	 * We are back.  Now reacquire the siglock before touching
2477 	 * last_siginfo, so that we are sure to have synchronized with
2478 	 * any signal-sending on another CPU that wants to examine it.
2479 	 */
2480 	spin_lock_irq(&current->sighand->siglock);
2481 	exit_code = current->exit_code;
2482 	current->last_siginfo = NULL;
2483 	current->ptrace_message = 0;
2484 	current->exit_code = 0;
2485 
2486 	/* LISTENING can be set only during STOP traps, clear it */
2487 	current->jobctl &= ~(JOBCTL_LISTENING | JOBCTL_PTRACE_FROZEN);
2488 
2489 	/*
2490 	 * Queued signals ignored us while we were stopped for tracing.
2491 	 * So check for any that we should take before resuming user mode.
2492 	 * This sets TIF_SIGPENDING, but never clears it.
2493 	 */
2494 	recalc_sigpending_tsk(current);
2495 	return exit_code;
2496 }
2497 
2498 static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
2499 {
2500 	kernel_siginfo_t info;
2501 
2502 	clear_siginfo(&info);
2503 	info.si_signo = signr;
2504 	info.si_code = exit_code;
2505 	info.si_pid = task_pid_vnr(current);
2506 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2507 
2508 	/* Let the debugger run.  */
2509 	return ptrace_stop(exit_code, why, message, &info);
2510 }
2511 
2512 int ptrace_notify(int exit_code, unsigned long message)
2513 {
2514 	int signr;
2515 
2516 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2517 	if (unlikely(task_work_pending(current)))
2518 		task_work_run();
2519 
2520 	spin_lock_irq(&current->sighand->siglock);
2521 	signr = ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
2522 	spin_unlock_irq(&current->sighand->siglock);
2523 	return signr;
2524 }
2525 
2526 /**
2527  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2528  * @signr: signr causing group stop if initiating
2529  *
2530  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2531  * and participate in it.  If already set, participate in the existing
2532  * group stop.  If participated in a group stop (and thus slept), %true is
2533  * returned with siglock released.
2534  *
2535  * If ptraced, this function doesn't handle stop itself.  Instead,
2536  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2537  * untouched.  The caller must ensure that INTERRUPT trap handling takes
2538  * places afterwards.
2539  *
2540  * CONTEXT:
2541  * Must be called with @current->sighand->siglock held, which is released
2542  * on %true return.
2543  *
2544  * RETURNS:
2545  * %false if group stop is already cancelled or ptrace trap is scheduled.
2546  * %true if participated in group stop.
2547  */
2548 static bool do_signal_stop(int signr)
2549 	__releases(&current->sighand->siglock)
2550 {
2551 	struct signal_struct *sig = current->signal;
2552 
2553 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2554 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2555 		struct task_struct *t;
2556 
2557 		/* signr will be recorded in task->jobctl for retries */
2558 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2559 
2560 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2561 		    unlikely(sig->flags & SIGNAL_GROUP_EXIT) ||
2562 		    unlikely(sig->group_exec_task))
2563 			return false;
2564 		/*
2565 		 * There is no group stop already in progress.  We must
2566 		 * initiate one now.
2567 		 *
2568 		 * While ptraced, a task may be resumed while group stop is
2569 		 * still in effect and then receive a stop signal and
2570 		 * initiate another group stop.  This deviates from the
2571 		 * usual behavior as two consecutive stop signals can't
2572 		 * cause two group stops when !ptraced.  That is why we
2573 		 * also check !task_is_stopped(t) below.
2574 		 *
2575 		 * The condition can be distinguished by testing whether
2576 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2577 		 * group_exit_code in such case.
2578 		 *
2579 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2580 		 * an intervening stop signal is required to cause two
2581 		 * continued events regardless of ptrace.
2582 		 */
2583 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2584 			sig->group_exit_code = signr;
2585 
2586 		sig->group_stop_count = 0;
2587 		if (task_set_jobctl_pending(current, signr | gstop))
2588 			sig->group_stop_count++;
2589 
2590 		for_other_threads(current, t) {
2591 			/*
2592 			 * Setting state to TASK_STOPPED for a group
2593 			 * stop is always done with the siglock held,
2594 			 * so this check has no races.
2595 			 */
2596 			if (!task_is_stopped(t) &&
2597 			    task_set_jobctl_pending(t, signr | gstop)) {
2598 				sig->group_stop_count++;
2599 				if (likely(!(t->ptrace & PT_SEIZED)))
2600 					signal_wake_up(t, 0);
2601 				else
2602 					ptrace_trap_notify(t);
2603 			}
2604 		}
2605 	}
2606 
2607 	if (likely(!current->ptrace)) {
2608 		int notify = 0;
2609 
2610 		/*
2611 		 * If there are no other threads in the group, or if there
2612 		 * is a group stop in progress and we are the last to stop,
2613 		 * report to the parent.
2614 		 */
2615 		if (task_participate_group_stop(current))
2616 			notify = CLD_STOPPED;
2617 
2618 		current->jobctl |= JOBCTL_STOPPED;
2619 		set_special_state(TASK_STOPPED);
2620 		spin_unlock_irq(&current->sighand->siglock);
2621 
2622 		/*
2623 		 * Notify the parent of the group stop completion.  Because
2624 		 * we're not holding either the siglock or tasklist_lock
2625 		 * here, ptracer may attach inbetween; however, this is for
2626 		 * group stop and should always be delivered to the real
2627 		 * parent of the group leader.  The new ptracer will get
2628 		 * its notification when this task transitions into
2629 		 * TASK_TRACED.
2630 		 */
2631 		if (notify) {
2632 			read_lock(&tasklist_lock);
2633 			do_notify_parent_cldstop(current, false, notify);
2634 			read_unlock(&tasklist_lock);
2635 		}
2636 
2637 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2638 		cgroup_enter_frozen();
2639 		schedule();
2640 		return true;
2641 	} else {
2642 		/*
2643 		 * While ptraced, group stop is handled by STOP trap.
2644 		 * Schedule it and let the caller deal with it.
2645 		 */
2646 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2647 		return false;
2648 	}
2649 }
2650 
2651 /**
2652  * do_jobctl_trap - take care of ptrace jobctl traps
2653  *
2654  * When PT_SEIZED, it's used for both group stop and explicit
2655  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2656  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2657  * the stop signal; otherwise, %SIGTRAP.
2658  *
2659  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2660  * number as exit_code and no siginfo.
2661  *
2662  * CONTEXT:
2663  * Must be called with @current->sighand->siglock held, which may be
2664  * released and re-acquired before returning with intervening sleep.
2665  */
2666 static void do_jobctl_trap(void)
2667 {
2668 	struct signal_struct *signal = current->signal;
2669 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2670 
2671 	if (current->ptrace & PT_SEIZED) {
2672 		if (!signal->group_stop_count &&
2673 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2674 			signr = SIGTRAP;
2675 		WARN_ON_ONCE(!signr);
2676 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2677 				 CLD_STOPPED, 0);
2678 	} else {
2679 		WARN_ON_ONCE(!signr);
2680 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2681 	}
2682 }
2683 
2684 /**
2685  * do_freezer_trap - handle the freezer jobctl trap
2686  *
2687  * Puts the task into frozen state, if only the task is not about to quit.
2688  * In this case it drops JOBCTL_TRAP_FREEZE.
2689  *
2690  * CONTEXT:
2691  * Must be called with @current->sighand->siglock held,
2692  * which is always released before returning.
2693  */
2694 static void do_freezer_trap(void)
2695 	__releases(&current->sighand->siglock)
2696 {
2697 	/*
2698 	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2699 	 * let's make another loop to give it a chance to be handled.
2700 	 * In any case, we'll return back.
2701 	 */
2702 	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2703 	     JOBCTL_TRAP_FREEZE) {
2704 		spin_unlock_irq(&current->sighand->siglock);
2705 		return;
2706 	}
2707 
2708 	/*
2709 	 * Now we're sure that there is no pending fatal signal and no
2710 	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2711 	 * immediately (if there is a non-fatal signal pending), and
2712 	 * put the task into sleep.
2713 	 */
2714 	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
2715 	clear_thread_flag(TIF_SIGPENDING);
2716 	spin_unlock_irq(&current->sighand->siglock);
2717 	cgroup_enter_frozen();
2718 	schedule();
2719 
2720 	/*
2721 	 * We could've been woken by task_work, run it to clear
2722 	 * TIF_NOTIFY_SIGNAL. The caller will retry if necessary.
2723 	 */
2724 	clear_notify_signal();
2725 	if (unlikely(task_work_pending(current)))
2726 		task_work_run();
2727 }
2728 
2729 static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
2730 {
2731 	/*
2732 	 * We do not check sig_kernel_stop(signr) but set this marker
2733 	 * unconditionally because we do not know whether debugger will
2734 	 * change signr. This flag has no meaning unless we are going
2735 	 * to stop after return from ptrace_stop(). In this case it will
2736 	 * be checked in do_signal_stop(), we should only stop if it was
2737 	 * not cleared by SIGCONT while we were sleeping. See also the
2738 	 * comment in dequeue_signal().
2739 	 */
2740 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2741 	signr = ptrace_stop(signr, CLD_TRAPPED, 0, info);
2742 
2743 	/* We're back.  Did the debugger cancel the sig?  */
2744 	if (signr == 0)
2745 		return signr;
2746 
2747 	/*
2748 	 * Update the siginfo structure if the signal has
2749 	 * changed.  If the debugger wanted something
2750 	 * specific in the siginfo structure then it should
2751 	 * have updated *info via PTRACE_SETSIGINFO.
2752 	 */
2753 	if (signr != info->si_signo) {
2754 		clear_siginfo(info);
2755 		info->si_signo = signr;
2756 		info->si_errno = 0;
2757 		info->si_code = SI_USER;
2758 		rcu_read_lock();
2759 		info->si_pid = task_pid_vnr(current->parent);
2760 		info->si_uid = from_kuid_munged(current_user_ns(),
2761 						task_uid(current->parent));
2762 		rcu_read_unlock();
2763 	}
2764 
2765 	/* If the (new) signal is now blocked, requeue it.  */
2766 	if (sigismember(&current->blocked, signr) ||
2767 	    fatal_signal_pending(current)) {
2768 		send_signal_locked(signr, info, current, type);
2769 		signr = 0;
2770 	}
2771 
2772 	return signr;
2773 }
2774 
2775 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2776 {
2777 	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2778 	case SIL_FAULT:
2779 	case SIL_FAULT_TRAPNO:
2780 	case SIL_FAULT_MCEERR:
2781 	case SIL_FAULT_BNDERR:
2782 	case SIL_FAULT_PKUERR:
2783 	case SIL_FAULT_PERF_EVENT:
2784 		ksig->info.si_addr = arch_untagged_si_addr(
2785 			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2786 		break;
2787 	case SIL_KILL:
2788 	case SIL_TIMER:
2789 	case SIL_POLL:
2790 	case SIL_CHLD:
2791 	case SIL_RT:
2792 	case SIL_SYS:
2793 		break;
2794 	}
2795 }
2796 
2797 bool get_signal(struct ksignal *ksig)
2798 {
2799 	struct sighand_struct *sighand = current->sighand;
2800 	struct signal_struct *signal = current->signal;
2801 	int signr;
2802 
2803 	clear_notify_signal();
2804 	if (unlikely(task_work_pending(current)))
2805 		task_work_run();
2806 
2807 	if (!task_sigpending(current))
2808 		return false;
2809 
2810 	if (unlikely(uprobe_deny_signal()))
2811 		return false;
2812 
2813 	/*
2814 	 * Do this once, we can't return to user-mode if freezing() == T.
2815 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2816 	 * thus do not need another check after return.
2817 	 */
2818 	try_to_freeze();
2819 
2820 relock:
2821 	spin_lock_irq(&sighand->siglock);
2822 
2823 	/*
2824 	 * Every stopped thread goes here after wakeup. Check to see if
2825 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2826 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2827 	 */
2828 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2829 		int why;
2830 
2831 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2832 			why = CLD_CONTINUED;
2833 		else
2834 			why = CLD_STOPPED;
2835 
2836 		signal->flags &= ~SIGNAL_CLD_MASK;
2837 
2838 		spin_unlock_irq(&sighand->siglock);
2839 
2840 		/*
2841 		 * Notify the parent that we're continuing.  This event is
2842 		 * always per-process and doesn't make whole lot of sense
2843 		 * for ptracers, who shouldn't consume the state via
2844 		 * wait(2) either, but, for backward compatibility, notify
2845 		 * the ptracer of the group leader too unless it's gonna be
2846 		 * a duplicate.
2847 		 */
2848 		read_lock(&tasklist_lock);
2849 		do_notify_parent_cldstop(current, false, why);
2850 
2851 		if (ptrace_reparented(current->group_leader))
2852 			do_notify_parent_cldstop(current->group_leader,
2853 						true, why);
2854 		read_unlock(&tasklist_lock);
2855 
2856 		goto relock;
2857 	}
2858 
2859 	for (;;) {
2860 		struct k_sigaction *ka;
2861 		enum pid_type type;
2862 
2863 		/* Has this task already been marked for death? */
2864 		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
2865 		     signal->group_exec_task) {
2866 			signr = SIGKILL;
2867 			sigdelset(&current->pending.signal, SIGKILL);
2868 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2869 					     &sighand->action[SIGKILL-1]);
2870 			recalc_sigpending();
2871 			/*
2872 			 * implies do_group_exit() or return to PF_USER_WORKER,
2873 			 * no need to initialize ksig->info/etc.
2874 			 */
2875 			goto fatal;
2876 		}
2877 
2878 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2879 		    do_signal_stop(0))
2880 			goto relock;
2881 
2882 		if (unlikely(current->jobctl &
2883 			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2884 			if (current->jobctl & JOBCTL_TRAP_MASK) {
2885 				do_jobctl_trap();
2886 				spin_unlock_irq(&sighand->siglock);
2887 			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2888 				do_freezer_trap();
2889 
2890 			goto relock;
2891 		}
2892 
2893 		/*
2894 		 * If the task is leaving the frozen state, let's update
2895 		 * cgroup counters and reset the frozen bit.
2896 		 */
2897 		if (unlikely(cgroup_task_frozen(current))) {
2898 			spin_unlock_irq(&sighand->siglock);
2899 			cgroup_leave_frozen(false);
2900 			goto relock;
2901 		}
2902 
2903 		/*
2904 		 * Signals generated by the execution of an instruction
2905 		 * need to be delivered before any other pending signals
2906 		 * so that the instruction pointer in the signal stack
2907 		 * frame points to the faulting instruction.
2908 		 */
2909 		type = PIDTYPE_PID;
2910 		signr = dequeue_synchronous_signal(&ksig->info);
2911 		if (!signr)
2912 			signr = dequeue_signal(&current->blocked, &ksig->info, &type);
2913 
2914 		if (!signr)
2915 			break; /* will return 0 */
2916 
2917 		if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2918 		    !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2919 			signr = ptrace_signal(signr, &ksig->info, type);
2920 			if (!signr)
2921 				continue;
2922 		}
2923 
2924 		ka = &sighand->action[signr-1];
2925 
2926 		/* Trace actually delivered signals. */
2927 		trace_signal_deliver(signr, &ksig->info, ka);
2928 
2929 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2930 			continue;
2931 		if (ka->sa.sa_handler != SIG_DFL) {
2932 			/* Run the handler.  */
2933 			ksig->ka = *ka;
2934 
2935 			if (ka->sa.sa_flags & SA_ONESHOT)
2936 				ka->sa.sa_handler = SIG_DFL;
2937 
2938 			break; /* will return non-zero "signr" value */
2939 		}
2940 
2941 		/*
2942 		 * Now we are doing the default action for this signal.
2943 		 */
2944 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2945 			continue;
2946 
2947 		/*
2948 		 * Global init gets no signals it doesn't want.
2949 		 * Container-init gets no signals it doesn't want from same
2950 		 * container.
2951 		 *
2952 		 * Note that if global/container-init sees a sig_kernel_only()
2953 		 * signal here, the signal must have been generated internally
2954 		 * or must have come from an ancestor namespace. In either
2955 		 * case, the signal cannot be dropped.
2956 		 */
2957 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2958 				!sig_kernel_only(signr))
2959 			continue;
2960 
2961 		if (sig_kernel_stop(signr)) {
2962 			/*
2963 			 * The default action is to stop all threads in
2964 			 * the thread group.  The job control signals
2965 			 * do nothing in an orphaned pgrp, but SIGSTOP
2966 			 * always works.  Note that siglock needs to be
2967 			 * dropped during the call to is_orphaned_pgrp()
2968 			 * because of lock ordering with tasklist_lock.
2969 			 * This allows an intervening SIGCONT to be posted.
2970 			 * We need to check for that and bail out if necessary.
2971 			 */
2972 			if (signr != SIGSTOP) {
2973 				spin_unlock_irq(&sighand->siglock);
2974 
2975 				/* signals can be posted during this window */
2976 
2977 				if (is_current_pgrp_orphaned())
2978 					goto relock;
2979 
2980 				spin_lock_irq(&sighand->siglock);
2981 			}
2982 
2983 			if (likely(do_signal_stop(signr))) {
2984 				/* It released the siglock.  */
2985 				goto relock;
2986 			}
2987 
2988 			/*
2989 			 * We didn't actually stop, due to a race
2990 			 * with SIGCONT or something like that.
2991 			 */
2992 			continue;
2993 		}
2994 
2995 	fatal:
2996 		spin_unlock_irq(&sighand->siglock);
2997 		if (unlikely(cgroup_task_frozen(current)))
2998 			cgroup_leave_frozen(true);
2999 
3000 		/*
3001 		 * Anything else is fatal, maybe with a core dump.
3002 		 */
3003 		current->flags |= PF_SIGNALED;
3004 
3005 		if (sig_kernel_coredump(signr)) {
3006 			if (print_fatal_signals)
3007 				print_fatal_signal(signr);
3008 			proc_coredump_connector(current);
3009 			/*
3010 			 * If it was able to dump core, this kills all
3011 			 * other threads in the group and synchronizes with
3012 			 * their demise.  If we lost the race with another
3013 			 * thread getting here, it set group_exit_code
3014 			 * first and our do_group_exit call below will use
3015 			 * that value and ignore the one we pass it.
3016 			 */
3017 			vfs_coredump(&ksig->info);
3018 		}
3019 
3020 		/*
3021 		 * PF_USER_WORKER threads will catch and exit on fatal signals
3022 		 * themselves. They have cleanup that must be performed, so we
3023 		 * cannot call do_exit() on their behalf. Note that ksig won't
3024 		 * be properly initialized, PF_USER_WORKER's shouldn't use it.
3025 		 */
3026 		if (current->flags & PF_USER_WORKER)
3027 			goto out;
3028 
3029 		/*
3030 		 * Death signals, no core dump.
3031 		 */
3032 		do_group_exit(signr);
3033 		/* NOTREACHED */
3034 	}
3035 	spin_unlock_irq(&sighand->siglock);
3036 
3037 	ksig->sig = signr;
3038 
3039 	if (signr && !(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
3040 		hide_si_addr_tag_bits(ksig);
3041 out:
3042 	return signr > 0;
3043 }
3044 
3045 /**
3046  * signal_delivered - called after signal delivery to update blocked signals
3047  * @ksig:		kernel signal struct
3048  * @stepping:		nonzero if debugger single-step or block-step in use
3049  *
3050  * This function should be called when a signal has successfully been
3051  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
3052  * is always blocked), and the signal itself is blocked unless %SA_NODEFER
3053  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
3054  */
3055 static void signal_delivered(struct ksignal *ksig, int stepping)
3056 {
3057 	sigset_t blocked;
3058 
3059 	/* A signal was successfully delivered, and the
3060 	   saved sigmask was stored on the signal frame,
3061 	   and will be restored by sigreturn.  So we can
3062 	   simply clear the restore sigmask flag.  */
3063 	clear_restore_sigmask();
3064 
3065 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
3066 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
3067 		sigaddset(&blocked, ksig->sig);
3068 	set_current_blocked(&blocked);
3069 	if (current->sas_ss_flags & SS_AUTODISARM)
3070 		sas_ss_reset(current);
3071 	if (stepping)
3072 		ptrace_notify(SIGTRAP, 0);
3073 }
3074 
3075 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
3076 {
3077 	if (failed)
3078 		force_sigsegv(ksig->sig);
3079 	else
3080 		signal_delivered(ksig, stepping);
3081 }
3082 
3083 /*
3084  * It could be that complete_signal() picked us to notify about the
3085  * group-wide signal. Other threads should be notified now to take
3086  * the shared signals in @which since we will not.
3087  */
3088 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
3089 {
3090 	sigset_t retarget;
3091 	struct task_struct *t;
3092 
3093 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
3094 	if (sigisemptyset(&retarget))
3095 		return;
3096 
3097 	for_other_threads(tsk, t) {
3098 		if (t->flags & PF_EXITING)
3099 			continue;
3100 
3101 		if (!has_pending_signals(&retarget, &t->blocked))
3102 			continue;
3103 		/* Remove the signals this thread can handle. */
3104 		sigandsets(&retarget, &retarget, &t->blocked);
3105 
3106 		if (!task_sigpending(t))
3107 			signal_wake_up(t, 0);
3108 
3109 		if (sigisemptyset(&retarget))
3110 			break;
3111 	}
3112 }
3113 
3114 void exit_signals(struct task_struct *tsk)
3115 {
3116 	int group_stop = 0;
3117 	sigset_t unblocked;
3118 
3119 	/*
3120 	 * @tsk is about to have PF_EXITING set - lock out users which
3121 	 * expect stable threadgroup.
3122 	 */
3123 	cgroup_threadgroup_change_begin(tsk);
3124 
3125 	if (thread_group_empty(tsk) || (tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
3126 		tsk->flags |= PF_EXITING;
3127 		cgroup_threadgroup_change_end(tsk);
3128 		return;
3129 	}
3130 
3131 	spin_lock_irq(&tsk->sighand->siglock);
3132 	/*
3133 	 * From now this task is not visible for group-wide signals,
3134 	 * see wants_signal(), do_signal_stop().
3135 	 */
3136 	tsk->flags |= PF_EXITING;
3137 
3138 	cgroup_threadgroup_change_end(tsk);
3139 
3140 	if (!task_sigpending(tsk))
3141 		goto out;
3142 
3143 	unblocked = tsk->blocked;
3144 	signotset(&unblocked);
3145 	retarget_shared_pending(tsk, &unblocked);
3146 
3147 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
3148 	    task_participate_group_stop(tsk))
3149 		group_stop = CLD_STOPPED;
3150 out:
3151 	spin_unlock_irq(&tsk->sighand->siglock);
3152 
3153 	/*
3154 	 * If group stop has completed, deliver the notification.  This
3155 	 * should always go to the real parent of the group leader.
3156 	 */
3157 	if (unlikely(group_stop)) {
3158 		read_lock(&tasklist_lock);
3159 		do_notify_parent_cldstop(tsk, false, group_stop);
3160 		read_unlock(&tasklist_lock);
3161 	}
3162 }
3163 
3164 /*
3165  * System call entry points.
3166  */
3167 
3168 /**
3169  *  sys_restart_syscall - restart a system call
3170  */
3171 SYSCALL_DEFINE0(restart_syscall)
3172 {
3173 	struct restart_block *restart = &current->restart_block;
3174 	return restart->fn(restart);
3175 }
3176 
3177 long do_no_restart_syscall(struct restart_block *param)
3178 {
3179 	return -EINTR;
3180 }
3181 
3182 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
3183 {
3184 	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3185 		sigset_t newblocked;
3186 		/* A set of now blocked but previously unblocked signals. */
3187 		sigandnsets(&newblocked, newset, &current->blocked);
3188 		retarget_shared_pending(tsk, &newblocked);
3189 	}
3190 	tsk->blocked = *newset;
3191 	recalc_sigpending();
3192 }
3193 
3194 /**
3195  * set_current_blocked - change current->blocked mask
3196  * @newset: new mask
3197  *
3198  * It is wrong to change ->blocked directly, this helper should be used
3199  * to ensure the process can't miss a shared signal we are going to block.
3200  */
3201 void set_current_blocked(sigset_t *newset)
3202 {
3203 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3204 	__set_current_blocked(newset);
3205 }
3206 
3207 void __set_current_blocked(const sigset_t *newset)
3208 {
3209 	struct task_struct *tsk = current;
3210 
3211 	/*
3212 	 * In case the signal mask hasn't changed, there is nothing we need
3213 	 * to do. The current->blocked shouldn't be modified by other task.
3214 	 */
3215 	if (sigequalsets(&tsk->blocked, newset))
3216 		return;
3217 
3218 	spin_lock_irq(&tsk->sighand->siglock);
3219 	__set_task_blocked(tsk, newset);
3220 	spin_unlock_irq(&tsk->sighand->siglock);
3221 }
3222 
3223 /*
3224  * This is also useful for kernel threads that want to temporarily
3225  * (or permanently) block certain signals.
3226  *
3227  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3228  * interface happily blocks "unblockable" signals like SIGKILL
3229  * and friends.
3230  */
3231 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3232 {
3233 	struct task_struct *tsk = current;
3234 	sigset_t newset;
3235 
3236 	/* Lockless, only current can change ->blocked, never from irq */
3237 	if (oldset)
3238 		*oldset = tsk->blocked;
3239 
3240 	switch (how) {
3241 	case SIG_BLOCK:
3242 		sigorsets(&newset, &tsk->blocked, set);
3243 		break;
3244 	case SIG_UNBLOCK:
3245 		sigandnsets(&newset, &tsk->blocked, set);
3246 		break;
3247 	case SIG_SETMASK:
3248 		newset = *set;
3249 		break;
3250 	default:
3251 		return -EINVAL;
3252 	}
3253 
3254 	__set_current_blocked(&newset);
3255 	return 0;
3256 }
3257 EXPORT_SYMBOL(sigprocmask);
3258 
3259 /*
3260  * The api helps set app-provided sigmasks.
3261  *
3262  * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3263  * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3264  *
3265  * Note that it does set_restore_sigmask() in advance, so it must be always
3266  * paired with restore_saved_sigmask_unless() before return from syscall.
3267  */
3268 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3269 {
3270 	sigset_t kmask;
3271 
3272 	if (!umask)
3273 		return 0;
3274 	if (sigsetsize != sizeof(sigset_t))
3275 		return -EINVAL;
3276 	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3277 		return -EFAULT;
3278 
3279 	set_restore_sigmask();
3280 	current->saved_sigmask = current->blocked;
3281 	set_current_blocked(&kmask);
3282 
3283 	return 0;
3284 }
3285 
3286 #ifdef CONFIG_COMPAT
3287 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3288 			    size_t sigsetsize)
3289 {
3290 	sigset_t kmask;
3291 
3292 	if (!umask)
3293 		return 0;
3294 	if (sigsetsize != sizeof(compat_sigset_t))
3295 		return -EINVAL;
3296 	if (get_compat_sigset(&kmask, umask))
3297 		return -EFAULT;
3298 
3299 	set_restore_sigmask();
3300 	current->saved_sigmask = current->blocked;
3301 	set_current_blocked(&kmask);
3302 
3303 	return 0;
3304 }
3305 #endif
3306 
3307 /**
3308  *  sys_rt_sigprocmask - change the list of currently blocked signals
3309  *  @how: whether to add, remove, or set signals
3310  *  @nset: stores pending signals
3311  *  @oset: previous value of signal mask if non-null
3312  *  @sigsetsize: size of sigset_t type
3313  */
3314 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3315 		sigset_t __user *, oset, size_t, sigsetsize)
3316 {
3317 	sigset_t old_set, new_set;
3318 	int error;
3319 
3320 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3321 	if (sigsetsize != sizeof(sigset_t))
3322 		return -EINVAL;
3323 
3324 	old_set = current->blocked;
3325 
3326 	if (nset) {
3327 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3328 			return -EFAULT;
3329 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3330 
3331 		error = sigprocmask(how, &new_set, NULL);
3332 		if (error)
3333 			return error;
3334 	}
3335 
3336 	if (oset) {
3337 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3338 			return -EFAULT;
3339 	}
3340 
3341 	return 0;
3342 }
3343 
3344 #ifdef CONFIG_COMPAT
3345 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3346 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3347 {
3348 	sigset_t old_set = current->blocked;
3349 
3350 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3351 	if (sigsetsize != sizeof(sigset_t))
3352 		return -EINVAL;
3353 
3354 	if (nset) {
3355 		sigset_t new_set;
3356 		int error;
3357 		if (get_compat_sigset(&new_set, nset))
3358 			return -EFAULT;
3359 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3360 
3361 		error = sigprocmask(how, &new_set, NULL);
3362 		if (error)
3363 			return error;
3364 	}
3365 	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3366 }
3367 #endif
3368 
3369 static void do_sigpending(sigset_t *set)
3370 {
3371 	spin_lock_irq(&current->sighand->siglock);
3372 	sigorsets(set, &current->pending.signal,
3373 		  &current->signal->shared_pending.signal);
3374 	spin_unlock_irq(&current->sighand->siglock);
3375 
3376 	/* Outside the lock because only this thread touches it.  */
3377 	sigandsets(set, &current->blocked, set);
3378 }
3379 
3380 /**
3381  *  sys_rt_sigpending - examine a pending signal that has been raised
3382  *			while blocked
3383  *  @uset: stores pending signals
3384  *  @sigsetsize: size of sigset_t type or larger
3385  */
3386 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3387 {
3388 	sigset_t set;
3389 
3390 	if (sigsetsize > sizeof(*uset))
3391 		return -EINVAL;
3392 
3393 	do_sigpending(&set);
3394 
3395 	if (copy_to_user(uset, &set, sigsetsize))
3396 		return -EFAULT;
3397 
3398 	return 0;
3399 }
3400 
3401 #ifdef CONFIG_COMPAT
3402 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3403 		compat_size_t, sigsetsize)
3404 {
3405 	sigset_t set;
3406 
3407 	if (sigsetsize > sizeof(*uset))
3408 		return -EINVAL;
3409 
3410 	do_sigpending(&set);
3411 
3412 	return put_compat_sigset(uset, &set, sigsetsize);
3413 }
3414 #endif
3415 
3416 static const struct {
3417 	unsigned char limit, layout;
3418 } sig_sicodes[] = {
3419 	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3420 	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3421 	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3422 	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3423 	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3424 #if defined(SIGEMT)
3425 	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3426 #endif
3427 	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3428 	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3429 	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3430 };
3431 
3432 static bool known_siginfo_layout(unsigned sig, int si_code)
3433 {
3434 	if (si_code == SI_KERNEL)
3435 		return true;
3436 	else if ((si_code > SI_USER)) {
3437 		if (sig_specific_sicodes(sig)) {
3438 			if (si_code <= sig_sicodes[sig].limit)
3439 				return true;
3440 		}
3441 		else if (si_code <= NSIGPOLL)
3442 			return true;
3443 	}
3444 	else if (si_code >= SI_DETHREAD)
3445 		return true;
3446 	else if (si_code == SI_ASYNCNL)
3447 		return true;
3448 	return false;
3449 }
3450 
3451 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3452 {
3453 	enum siginfo_layout layout = SIL_KILL;
3454 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3455 		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3456 		    (si_code <= sig_sicodes[sig].limit)) {
3457 			layout = sig_sicodes[sig].layout;
3458 			/* Handle the exceptions */
3459 			if ((sig == SIGBUS) &&
3460 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3461 				layout = SIL_FAULT_MCEERR;
3462 			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3463 				layout = SIL_FAULT_BNDERR;
3464 #ifdef SEGV_PKUERR
3465 			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3466 				layout = SIL_FAULT_PKUERR;
3467 #endif
3468 			else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3469 				layout = SIL_FAULT_PERF_EVENT;
3470 			else if (IS_ENABLED(CONFIG_SPARC) &&
3471 				 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3472 				layout = SIL_FAULT_TRAPNO;
3473 			else if (IS_ENABLED(CONFIG_ALPHA) &&
3474 				 ((sig == SIGFPE) ||
3475 				  ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3476 				layout = SIL_FAULT_TRAPNO;
3477 		}
3478 		else if (si_code <= NSIGPOLL)
3479 			layout = SIL_POLL;
3480 	} else {
3481 		if (si_code == SI_TIMER)
3482 			layout = SIL_TIMER;
3483 		else if (si_code == SI_SIGIO)
3484 			layout = SIL_POLL;
3485 		else if (si_code < 0)
3486 			layout = SIL_RT;
3487 	}
3488 	return layout;
3489 }
3490 
3491 static inline char __user *si_expansion(const siginfo_t __user *info)
3492 {
3493 	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3494 }
3495 
3496 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3497 {
3498 	char __user *expansion = si_expansion(to);
3499 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3500 		return -EFAULT;
3501 	if (clear_user(expansion, SI_EXPANSION_SIZE))
3502 		return -EFAULT;
3503 	return 0;
3504 }
3505 
3506 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3507 				       const siginfo_t __user *from)
3508 {
3509 	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3510 		char __user *expansion = si_expansion(from);
3511 		char buf[SI_EXPANSION_SIZE];
3512 		int i;
3513 		/*
3514 		 * An unknown si_code might need more than
3515 		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3516 		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3517 		 * will return this data to userspace exactly.
3518 		 */
3519 		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3520 			return -EFAULT;
3521 		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3522 			if (buf[i] != 0)
3523 				return -E2BIG;
3524 		}
3525 	}
3526 	return 0;
3527 }
3528 
3529 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3530 				    const siginfo_t __user *from)
3531 {
3532 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3533 		return -EFAULT;
3534 	to->si_signo = signo;
3535 	return post_copy_siginfo_from_user(to, from);
3536 }
3537 
3538 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3539 {
3540 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3541 		return -EFAULT;
3542 	return post_copy_siginfo_from_user(to, from);
3543 }
3544 
3545 #ifdef CONFIG_COMPAT
3546 /**
3547  * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3548  * @to: compat siginfo destination
3549  * @from: kernel siginfo source
3550  *
3551  * Note: This function does not work properly for the SIGCHLD on x32, but
3552  * fortunately it doesn't have to.  The only valid callers for this function are
3553  * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3554  * The latter does not care because SIGCHLD will never cause a coredump.
3555  */
3556 void copy_siginfo_to_external32(struct compat_siginfo *to,
3557 		const struct kernel_siginfo *from)
3558 {
3559 	memset(to, 0, sizeof(*to));
3560 
3561 	to->si_signo = from->si_signo;
3562 	to->si_errno = from->si_errno;
3563 	to->si_code  = from->si_code;
3564 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3565 	case SIL_KILL:
3566 		to->si_pid = from->si_pid;
3567 		to->si_uid = from->si_uid;
3568 		break;
3569 	case SIL_TIMER:
3570 		to->si_tid     = from->si_tid;
3571 		to->si_overrun = from->si_overrun;
3572 		to->si_int     = from->si_int;
3573 		break;
3574 	case SIL_POLL:
3575 		to->si_band = from->si_band;
3576 		to->si_fd   = from->si_fd;
3577 		break;
3578 	case SIL_FAULT:
3579 		to->si_addr = ptr_to_compat(from->si_addr);
3580 		break;
3581 	case SIL_FAULT_TRAPNO:
3582 		to->si_addr = ptr_to_compat(from->si_addr);
3583 		to->si_trapno = from->si_trapno;
3584 		break;
3585 	case SIL_FAULT_MCEERR:
3586 		to->si_addr = ptr_to_compat(from->si_addr);
3587 		to->si_addr_lsb = from->si_addr_lsb;
3588 		break;
3589 	case SIL_FAULT_BNDERR:
3590 		to->si_addr = ptr_to_compat(from->si_addr);
3591 		to->si_lower = ptr_to_compat(from->si_lower);
3592 		to->si_upper = ptr_to_compat(from->si_upper);
3593 		break;
3594 	case SIL_FAULT_PKUERR:
3595 		to->si_addr = ptr_to_compat(from->si_addr);
3596 		to->si_pkey = from->si_pkey;
3597 		break;
3598 	case SIL_FAULT_PERF_EVENT:
3599 		to->si_addr = ptr_to_compat(from->si_addr);
3600 		to->si_perf_data = from->si_perf_data;
3601 		to->si_perf_type = from->si_perf_type;
3602 		to->si_perf_flags = from->si_perf_flags;
3603 		break;
3604 	case SIL_CHLD:
3605 		to->si_pid = from->si_pid;
3606 		to->si_uid = from->si_uid;
3607 		to->si_status = from->si_status;
3608 		to->si_utime = from->si_utime;
3609 		to->si_stime = from->si_stime;
3610 		break;
3611 	case SIL_RT:
3612 		to->si_pid = from->si_pid;
3613 		to->si_uid = from->si_uid;
3614 		to->si_int = from->si_int;
3615 		break;
3616 	case SIL_SYS:
3617 		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3618 		to->si_syscall   = from->si_syscall;
3619 		to->si_arch      = from->si_arch;
3620 		break;
3621 	}
3622 }
3623 
3624 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3625 			   const struct kernel_siginfo *from)
3626 {
3627 	struct compat_siginfo new;
3628 
3629 	copy_siginfo_to_external32(&new, from);
3630 	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3631 		return -EFAULT;
3632 	return 0;
3633 }
3634 
3635 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3636 					 const struct compat_siginfo *from)
3637 {
3638 	clear_siginfo(to);
3639 	to->si_signo = from->si_signo;
3640 	to->si_errno = from->si_errno;
3641 	to->si_code  = from->si_code;
3642 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3643 	case SIL_KILL:
3644 		to->si_pid = from->si_pid;
3645 		to->si_uid = from->si_uid;
3646 		break;
3647 	case SIL_TIMER:
3648 		to->si_tid     = from->si_tid;
3649 		to->si_overrun = from->si_overrun;
3650 		to->si_int     = from->si_int;
3651 		break;
3652 	case SIL_POLL:
3653 		to->si_band = from->si_band;
3654 		to->si_fd   = from->si_fd;
3655 		break;
3656 	case SIL_FAULT:
3657 		to->si_addr = compat_ptr(from->si_addr);
3658 		break;
3659 	case SIL_FAULT_TRAPNO:
3660 		to->si_addr = compat_ptr(from->si_addr);
3661 		to->si_trapno = from->si_trapno;
3662 		break;
3663 	case SIL_FAULT_MCEERR:
3664 		to->si_addr = compat_ptr(from->si_addr);
3665 		to->si_addr_lsb = from->si_addr_lsb;
3666 		break;
3667 	case SIL_FAULT_BNDERR:
3668 		to->si_addr = compat_ptr(from->si_addr);
3669 		to->si_lower = compat_ptr(from->si_lower);
3670 		to->si_upper = compat_ptr(from->si_upper);
3671 		break;
3672 	case SIL_FAULT_PKUERR:
3673 		to->si_addr = compat_ptr(from->si_addr);
3674 		to->si_pkey = from->si_pkey;
3675 		break;
3676 	case SIL_FAULT_PERF_EVENT:
3677 		to->si_addr = compat_ptr(from->si_addr);
3678 		to->si_perf_data = from->si_perf_data;
3679 		to->si_perf_type = from->si_perf_type;
3680 		to->si_perf_flags = from->si_perf_flags;
3681 		break;
3682 	case SIL_CHLD:
3683 		to->si_pid    = from->si_pid;
3684 		to->si_uid    = from->si_uid;
3685 		to->si_status = from->si_status;
3686 #ifdef CONFIG_X86_X32_ABI
3687 		if (in_x32_syscall()) {
3688 			to->si_utime = from->_sifields._sigchld_x32._utime;
3689 			to->si_stime = from->_sifields._sigchld_x32._stime;
3690 		} else
3691 #endif
3692 		{
3693 			to->si_utime = from->si_utime;
3694 			to->si_stime = from->si_stime;
3695 		}
3696 		break;
3697 	case SIL_RT:
3698 		to->si_pid = from->si_pid;
3699 		to->si_uid = from->si_uid;
3700 		to->si_int = from->si_int;
3701 		break;
3702 	case SIL_SYS:
3703 		to->si_call_addr = compat_ptr(from->si_call_addr);
3704 		to->si_syscall   = from->si_syscall;
3705 		to->si_arch      = from->si_arch;
3706 		break;
3707 	}
3708 	return 0;
3709 }
3710 
3711 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3712 				      const struct compat_siginfo __user *ufrom)
3713 {
3714 	struct compat_siginfo from;
3715 
3716 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3717 		return -EFAULT;
3718 
3719 	from.si_signo = signo;
3720 	return post_copy_siginfo_from_user32(to, &from);
3721 }
3722 
3723 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3724 			     const struct compat_siginfo __user *ufrom)
3725 {
3726 	struct compat_siginfo from;
3727 
3728 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3729 		return -EFAULT;
3730 
3731 	return post_copy_siginfo_from_user32(to, &from);
3732 }
3733 #endif /* CONFIG_COMPAT */
3734 
3735 /**
3736  *  do_sigtimedwait - wait for queued signals specified in @which
3737  *  @which: queued signals to wait for
3738  *  @info: if non-null, the signal's siginfo is returned here
3739  *  @ts: upper bound on process time suspension
3740  */
3741 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3742 		    const struct timespec64 *ts)
3743 {
3744 	ktime_t *to = NULL, timeout = KTIME_MAX;
3745 	struct task_struct *tsk = current;
3746 	sigset_t mask = *which;
3747 	enum pid_type type;
3748 	int sig, ret = 0;
3749 
3750 	if (ts) {
3751 		if (!timespec64_valid(ts))
3752 			return -EINVAL;
3753 		timeout = timespec64_to_ktime(*ts);
3754 		to = &timeout;
3755 	}
3756 
3757 	/*
3758 	 * Invert the set of allowed signals to get those we want to block.
3759 	 */
3760 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3761 	signotset(&mask);
3762 
3763 	spin_lock_irq(&tsk->sighand->siglock);
3764 	sig = dequeue_signal(&mask, info, &type);
3765 	if (!sig && timeout) {
3766 		/*
3767 		 * None ready, temporarily unblock those we're interested
3768 		 * while we are sleeping in so that we'll be awakened when
3769 		 * they arrive. Unblocking is always fine, we can avoid
3770 		 * set_current_blocked().
3771 		 */
3772 		tsk->real_blocked = tsk->blocked;
3773 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3774 		recalc_sigpending();
3775 		spin_unlock_irq(&tsk->sighand->siglock);
3776 
3777 		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
3778 		ret = schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3779 					       HRTIMER_MODE_REL);
3780 		spin_lock_irq(&tsk->sighand->siglock);
3781 		__set_task_blocked(tsk, &tsk->real_blocked);
3782 		sigemptyset(&tsk->real_blocked);
3783 		sig = dequeue_signal(&mask, info, &type);
3784 	}
3785 	spin_unlock_irq(&tsk->sighand->siglock);
3786 
3787 	if (sig)
3788 		return sig;
3789 	return ret ? -EINTR : -EAGAIN;
3790 }
3791 
3792 /**
3793  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3794  *			in @uthese
3795  *  @uthese: queued signals to wait for
3796  *  @uinfo: if non-null, the signal's siginfo is returned here
3797  *  @uts: upper bound on process time suspension
3798  *  @sigsetsize: size of sigset_t type
3799  */
3800 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3801 		siginfo_t __user *, uinfo,
3802 		const struct __kernel_timespec __user *, uts,
3803 		size_t, sigsetsize)
3804 {
3805 	sigset_t these;
3806 	struct timespec64 ts;
3807 	kernel_siginfo_t info;
3808 	int ret;
3809 
3810 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3811 	if (sigsetsize != sizeof(sigset_t))
3812 		return -EINVAL;
3813 
3814 	if (copy_from_user(&these, uthese, sizeof(these)))
3815 		return -EFAULT;
3816 
3817 	if (uts) {
3818 		if (get_timespec64(&ts, uts))
3819 			return -EFAULT;
3820 	}
3821 
3822 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3823 
3824 	if (ret > 0 && uinfo) {
3825 		if (copy_siginfo_to_user(uinfo, &info))
3826 			ret = -EFAULT;
3827 	}
3828 
3829 	return ret;
3830 }
3831 
3832 #ifdef CONFIG_COMPAT_32BIT_TIME
3833 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3834 		siginfo_t __user *, uinfo,
3835 		const struct old_timespec32 __user *, uts,
3836 		size_t, sigsetsize)
3837 {
3838 	sigset_t these;
3839 	struct timespec64 ts;
3840 	kernel_siginfo_t info;
3841 	int ret;
3842 
3843 	if (sigsetsize != sizeof(sigset_t))
3844 		return -EINVAL;
3845 
3846 	if (copy_from_user(&these, uthese, sizeof(these)))
3847 		return -EFAULT;
3848 
3849 	if (uts) {
3850 		if (get_old_timespec32(&ts, uts))
3851 			return -EFAULT;
3852 	}
3853 
3854 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3855 
3856 	if (ret > 0 && uinfo) {
3857 		if (copy_siginfo_to_user(uinfo, &info))
3858 			ret = -EFAULT;
3859 	}
3860 
3861 	return ret;
3862 }
3863 #endif
3864 
3865 #ifdef CONFIG_COMPAT
3866 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3867 		struct compat_siginfo __user *, uinfo,
3868 		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3869 {
3870 	sigset_t s;
3871 	struct timespec64 t;
3872 	kernel_siginfo_t info;
3873 	long ret;
3874 
3875 	if (sigsetsize != sizeof(sigset_t))
3876 		return -EINVAL;
3877 
3878 	if (get_compat_sigset(&s, uthese))
3879 		return -EFAULT;
3880 
3881 	if (uts) {
3882 		if (get_timespec64(&t, uts))
3883 			return -EFAULT;
3884 	}
3885 
3886 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3887 
3888 	if (ret > 0 && uinfo) {
3889 		if (copy_siginfo_to_user32(uinfo, &info))
3890 			ret = -EFAULT;
3891 	}
3892 
3893 	return ret;
3894 }
3895 
3896 #ifdef CONFIG_COMPAT_32BIT_TIME
3897 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3898 		struct compat_siginfo __user *, uinfo,
3899 		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3900 {
3901 	sigset_t s;
3902 	struct timespec64 t;
3903 	kernel_siginfo_t info;
3904 	long ret;
3905 
3906 	if (sigsetsize != sizeof(sigset_t))
3907 		return -EINVAL;
3908 
3909 	if (get_compat_sigset(&s, uthese))
3910 		return -EFAULT;
3911 
3912 	if (uts) {
3913 		if (get_old_timespec32(&t, uts))
3914 			return -EFAULT;
3915 	}
3916 
3917 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3918 
3919 	if (ret > 0 && uinfo) {
3920 		if (copy_siginfo_to_user32(uinfo, &info))
3921 			ret = -EFAULT;
3922 	}
3923 
3924 	return ret;
3925 }
3926 #endif
3927 #endif
3928 
3929 static void prepare_kill_siginfo(int sig, struct kernel_siginfo *info,
3930 				 enum pid_type type)
3931 {
3932 	clear_siginfo(info);
3933 	info->si_signo = sig;
3934 	info->si_errno = 0;
3935 	info->si_code = (type == PIDTYPE_PID) ? SI_TKILL : SI_USER;
3936 	info->si_pid = task_tgid_vnr(current);
3937 	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3938 }
3939 
3940 /**
3941  *  sys_kill - send a signal to a process
3942  *  @pid: the PID of the process
3943  *  @sig: signal to be sent
3944  */
3945 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3946 {
3947 	struct kernel_siginfo info;
3948 
3949 	prepare_kill_siginfo(sig, &info, PIDTYPE_TGID);
3950 
3951 	return kill_something_info(sig, &info, pid);
3952 }
3953 
3954 /*
3955  * Verify that the signaler and signalee either are in the same pid namespace
3956  * or that the signaler's pid namespace is an ancestor of the signalee's pid
3957  * namespace.
3958  */
3959 static bool access_pidfd_pidns(struct pid *pid)
3960 {
3961 	struct pid_namespace *active = task_active_pid_ns(current);
3962 	struct pid_namespace *p = ns_of_pid(pid);
3963 
3964 	for (;;) {
3965 		if (!p)
3966 			return false;
3967 		if (p == active)
3968 			break;
3969 		p = p->parent;
3970 	}
3971 
3972 	return true;
3973 }
3974 
3975 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3976 		siginfo_t __user *info)
3977 {
3978 #ifdef CONFIG_COMPAT
3979 	/*
3980 	 * Avoid hooking up compat syscalls and instead handle necessary
3981 	 * conversions here. Note, this is a stop-gap measure and should not be
3982 	 * considered a generic solution.
3983 	 */
3984 	if (in_compat_syscall())
3985 		return copy_siginfo_from_user32(
3986 			kinfo, (struct compat_siginfo __user *)info);
3987 #endif
3988 	return copy_siginfo_from_user(kinfo, info);
3989 }
3990 
3991 static struct pid *pidfd_to_pid(const struct file *file)
3992 {
3993 	struct pid *pid;
3994 
3995 	pid = pidfd_pid(file);
3996 	if (!IS_ERR(pid))
3997 		return pid;
3998 
3999 	return tgid_pidfd_to_pid(file);
4000 }
4001 
4002 #define PIDFD_SEND_SIGNAL_FLAGS                            \
4003 	(PIDFD_SIGNAL_THREAD | PIDFD_SIGNAL_THREAD_GROUP | \
4004 	 PIDFD_SIGNAL_PROCESS_GROUP)
4005 
4006 static int do_pidfd_send_signal(struct pid *pid, int sig, enum pid_type type,
4007 				siginfo_t __user *info, unsigned int flags)
4008 {
4009 	kernel_siginfo_t kinfo;
4010 
4011 	switch (flags) {
4012 	case PIDFD_SIGNAL_THREAD:
4013 		type = PIDTYPE_PID;
4014 		break;
4015 	case PIDFD_SIGNAL_THREAD_GROUP:
4016 		type = PIDTYPE_TGID;
4017 		break;
4018 	case PIDFD_SIGNAL_PROCESS_GROUP:
4019 		type = PIDTYPE_PGID;
4020 		break;
4021 	}
4022 
4023 	if (info) {
4024 		int ret;
4025 
4026 		ret = copy_siginfo_from_user_any(&kinfo, info);
4027 		if (unlikely(ret))
4028 			return ret;
4029 
4030 		if (unlikely(sig != kinfo.si_signo))
4031 			return -EINVAL;
4032 
4033 		/* Only allow sending arbitrary signals to yourself. */
4034 		if ((task_pid(current) != pid || type > PIDTYPE_TGID) &&
4035 		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
4036 			return -EPERM;
4037 	} else {
4038 		prepare_kill_siginfo(sig, &kinfo, type);
4039 	}
4040 
4041 	if (type == PIDTYPE_PGID)
4042 		return kill_pgrp_info(sig, &kinfo, pid);
4043 
4044 	return kill_pid_info_type(sig, &kinfo, pid, type);
4045 }
4046 
4047 /**
4048  * sys_pidfd_send_signal - Signal a process through a pidfd
4049  * @pidfd:  file descriptor of the process
4050  * @sig:    signal to send
4051  * @info:   signal info
4052  * @flags:  future flags
4053  *
4054  * Send the signal to the thread group or to the individual thread depending
4055  * on PIDFD_THREAD.
4056  * In the future extension to @flags may be used to override the default scope
4057  * of @pidfd.
4058  *
4059  * Return: 0 on success, negative errno on failure
4060  */
4061 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
4062 		siginfo_t __user *, info, unsigned int, flags)
4063 {
4064 	struct pid *pid;
4065 	enum pid_type type;
4066 	int ret;
4067 
4068 	/* Enforce flags be set to 0 until we add an extension. */
4069 	if (flags & ~PIDFD_SEND_SIGNAL_FLAGS)
4070 		return -EINVAL;
4071 
4072 	/* Ensure that only a single signal scope determining flag is set. */
4073 	if (hweight32(flags & PIDFD_SEND_SIGNAL_FLAGS) > 1)
4074 		return -EINVAL;
4075 
4076 	switch (pidfd) {
4077 	case PIDFD_SELF_THREAD:
4078 		pid = get_task_pid(current, PIDTYPE_PID);
4079 		type = PIDTYPE_PID;
4080 		break;
4081 	case PIDFD_SELF_THREAD_GROUP:
4082 		pid = get_task_pid(current, PIDTYPE_TGID);
4083 		type = PIDTYPE_TGID;
4084 		break;
4085 	default: {
4086 		CLASS(fd, f)(pidfd);
4087 		if (fd_empty(f))
4088 			return -EBADF;
4089 
4090 		/* Is this a pidfd? */
4091 		pid = pidfd_to_pid(fd_file(f));
4092 		if (IS_ERR(pid))
4093 			return PTR_ERR(pid);
4094 
4095 		if (!access_pidfd_pidns(pid))
4096 			return -EINVAL;
4097 
4098 		/* Infer scope from the type of pidfd. */
4099 		if (fd_file(f)->f_flags & PIDFD_THREAD)
4100 			type = PIDTYPE_PID;
4101 		else
4102 			type = PIDTYPE_TGID;
4103 
4104 		return do_pidfd_send_signal(pid, sig, type, info, flags);
4105 	}
4106 	}
4107 
4108 	ret = do_pidfd_send_signal(pid, sig, type, info, flags);
4109 	put_pid(pid);
4110 
4111 	return ret;
4112 }
4113 
4114 static int
4115 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
4116 {
4117 	struct task_struct *p;
4118 	int error = -ESRCH;
4119 
4120 	rcu_read_lock();
4121 	p = find_task_by_vpid(pid);
4122 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
4123 		error = check_kill_permission(sig, info, p);
4124 		/*
4125 		 * The null signal is a permissions and process existence
4126 		 * probe.  No signal is actually delivered.
4127 		 */
4128 		if (!error && sig) {
4129 			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
4130 			/*
4131 			 * If lock_task_sighand() failed we pretend the task
4132 			 * dies after receiving the signal. The window is tiny,
4133 			 * and the signal is private anyway.
4134 			 */
4135 			if (unlikely(error == -ESRCH))
4136 				error = 0;
4137 		}
4138 	}
4139 	rcu_read_unlock();
4140 
4141 	return error;
4142 }
4143 
4144 static int do_tkill(pid_t tgid, pid_t pid, int sig)
4145 {
4146 	struct kernel_siginfo info;
4147 
4148 	prepare_kill_siginfo(sig, &info, PIDTYPE_PID);
4149 
4150 	return do_send_specific(tgid, pid, sig, &info);
4151 }
4152 
4153 /**
4154  *  sys_tgkill - send signal to one specific thread
4155  *  @tgid: the thread group ID of the thread
4156  *  @pid: the PID of the thread
4157  *  @sig: signal to be sent
4158  *
4159  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
4160  *  exists but it's not belonging to the target process anymore. This
4161  *  method solves the problem of threads exiting and PIDs getting reused.
4162  */
4163 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
4164 {
4165 	/* This is only valid for single tasks */
4166 	if (pid <= 0 || tgid <= 0)
4167 		return -EINVAL;
4168 
4169 	return do_tkill(tgid, pid, sig);
4170 }
4171 
4172 /**
4173  *  sys_tkill - send signal to one specific task
4174  *  @pid: the PID of the task
4175  *  @sig: signal to be sent
4176  *
4177  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
4178  */
4179 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
4180 {
4181 	/* This is only valid for single tasks */
4182 	if (pid <= 0)
4183 		return -EINVAL;
4184 
4185 	return do_tkill(0, pid, sig);
4186 }
4187 
4188 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
4189 {
4190 	/* Not even root can pretend to send signals from the kernel.
4191 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4192 	 */
4193 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4194 	    (task_pid_vnr(current) != pid))
4195 		return -EPERM;
4196 
4197 	/* POSIX.1b doesn't mention process groups.  */
4198 	return kill_proc_info(sig, info, pid);
4199 }
4200 
4201 /**
4202  *  sys_rt_sigqueueinfo - send signal information to a signal
4203  *  @pid: the PID of the thread
4204  *  @sig: signal to be sent
4205  *  @uinfo: signal info to be sent
4206  */
4207 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
4208 		siginfo_t __user *, uinfo)
4209 {
4210 	kernel_siginfo_t info;
4211 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4212 	if (unlikely(ret))
4213 		return ret;
4214 	return do_rt_sigqueueinfo(pid, sig, &info);
4215 }
4216 
4217 #ifdef CONFIG_COMPAT
4218 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
4219 			compat_pid_t, pid,
4220 			int, sig,
4221 			struct compat_siginfo __user *, uinfo)
4222 {
4223 	kernel_siginfo_t info;
4224 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4225 	if (unlikely(ret))
4226 		return ret;
4227 	return do_rt_sigqueueinfo(pid, sig, &info);
4228 }
4229 #endif
4230 
4231 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4232 {
4233 	/* This is only valid for single tasks */
4234 	if (pid <= 0 || tgid <= 0)
4235 		return -EINVAL;
4236 
4237 	/* Not even root can pretend to send signals from the kernel.
4238 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4239 	 */
4240 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4241 	    (task_pid_vnr(current) != pid))
4242 		return -EPERM;
4243 
4244 	return do_send_specific(tgid, pid, sig, info);
4245 }
4246 
4247 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4248 		siginfo_t __user *, uinfo)
4249 {
4250 	kernel_siginfo_t info;
4251 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4252 	if (unlikely(ret))
4253 		return ret;
4254 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4255 }
4256 
4257 #ifdef CONFIG_COMPAT
4258 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4259 			compat_pid_t, tgid,
4260 			compat_pid_t, pid,
4261 			int, sig,
4262 			struct compat_siginfo __user *, uinfo)
4263 {
4264 	kernel_siginfo_t info;
4265 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4266 	if (unlikely(ret))
4267 		return ret;
4268 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4269 }
4270 #endif
4271 
4272 /*
4273  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4274  */
4275 void kernel_sigaction(int sig, __sighandler_t action)
4276 {
4277 	spin_lock_irq(&current->sighand->siglock);
4278 	current->sighand->action[sig - 1].sa.sa_handler = action;
4279 	if (action == SIG_IGN) {
4280 		sigset_t mask;
4281 
4282 		sigemptyset(&mask);
4283 		sigaddset(&mask, sig);
4284 
4285 		flush_sigqueue_mask(current, &mask, &current->signal->shared_pending);
4286 		flush_sigqueue_mask(current, &mask, &current->pending);
4287 		recalc_sigpending();
4288 	}
4289 	spin_unlock_irq(&current->sighand->siglock);
4290 }
4291 EXPORT_SYMBOL(kernel_sigaction);
4292 
4293 void __weak sigaction_compat_abi(struct k_sigaction *act,
4294 		struct k_sigaction *oact)
4295 {
4296 }
4297 
4298 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4299 {
4300 	struct task_struct *p = current, *t;
4301 	struct k_sigaction *k;
4302 	sigset_t mask;
4303 
4304 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4305 		return -EINVAL;
4306 
4307 	k = &p->sighand->action[sig-1];
4308 
4309 	spin_lock_irq(&p->sighand->siglock);
4310 	if (k->sa.sa_flags & SA_IMMUTABLE) {
4311 		spin_unlock_irq(&p->sighand->siglock);
4312 		return -EINVAL;
4313 	}
4314 	if (oact)
4315 		*oact = *k;
4316 
4317 	/*
4318 	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4319 	 * e.g. by having an architecture use the bit in their uapi.
4320 	 */
4321 	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4322 
4323 	/*
4324 	 * Clear unknown flag bits in order to allow userspace to detect missing
4325 	 * support for flag bits and to allow the kernel to use non-uapi bits
4326 	 * internally.
4327 	 */
4328 	if (act)
4329 		act->sa.sa_flags &= UAPI_SA_FLAGS;
4330 	if (oact)
4331 		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4332 
4333 	sigaction_compat_abi(act, oact);
4334 
4335 	if (act) {
4336 		bool was_ignored = k->sa.sa_handler == SIG_IGN;
4337 
4338 		sigdelsetmask(&act->sa.sa_mask,
4339 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4340 		*k = *act;
4341 		/*
4342 		 * POSIX 3.3.1.3:
4343 		 *  "Setting a signal action to SIG_IGN for a signal that is
4344 		 *   pending shall cause the pending signal to be discarded,
4345 		 *   whether or not it is blocked."
4346 		 *
4347 		 *  "Setting a signal action to SIG_DFL for a signal that is
4348 		 *   pending and whose default action is to ignore the signal
4349 		 *   (for example, SIGCHLD), shall cause the pending signal to
4350 		 *   be discarded, whether or not it is blocked"
4351 		 */
4352 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4353 			sigemptyset(&mask);
4354 			sigaddset(&mask, sig);
4355 			flush_sigqueue_mask(p, &mask, &p->signal->shared_pending);
4356 			for_each_thread(p, t)
4357 				flush_sigqueue_mask(p, &mask, &t->pending);
4358 		} else if (was_ignored) {
4359 			posixtimer_sig_unignore(p, sig);
4360 		}
4361 	}
4362 
4363 	spin_unlock_irq(&p->sighand->siglock);
4364 	return 0;
4365 }
4366 
4367 #ifdef CONFIG_DYNAMIC_SIGFRAME
4368 static inline void sigaltstack_lock(void)
4369 	__acquires(&current->sighand->siglock)
4370 {
4371 	spin_lock_irq(&current->sighand->siglock);
4372 }
4373 
4374 static inline void sigaltstack_unlock(void)
4375 	__releases(&current->sighand->siglock)
4376 {
4377 	spin_unlock_irq(&current->sighand->siglock);
4378 }
4379 #else
4380 static inline void sigaltstack_lock(void) { }
4381 static inline void sigaltstack_unlock(void) { }
4382 #endif
4383 
4384 static int
4385 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4386 		size_t min_ss_size)
4387 {
4388 	struct task_struct *t = current;
4389 	int ret = 0;
4390 
4391 	if (oss) {
4392 		memset(oss, 0, sizeof(stack_t));
4393 		oss->ss_sp = (void __user *) t->sas_ss_sp;
4394 		oss->ss_size = t->sas_ss_size;
4395 		oss->ss_flags = sas_ss_flags(sp) |
4396 			(current->sas_ss_flags & SS_FLAG_BITS);
4397 	}
4398 
4399 	if (ss) {
4400 		void __user *ss_sp = ss->ss_sp;
4401 		size_t ss_size = ss->ss_size;
4402 		unsigned ss_flags = ss->ss_flags;
4403 		int ss_mode;
4404 
4405 		if (unlikely(on_sig_stack(sp)))
4406 			return -EPERM;
4407 
4408 		ss_mode = ss_flags & ~SS_FLAG_BITS;
4409 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4410 				ss_mode != 0))
4411 			return -EINVAL;
4412 
4413 		/*
4414 		 * Return before taking any locks if no actual
4415 		 * sigaltstack changes were requested.
4416 		 */
4417 		if (t->sas_ss_sp == (unsigned long)ss_sp &&
4418 		    t->sas_ss_size == ss_size &&
4419 		    t->sas_ss_flags == ss_flags)
4420 			return 0;
4421 
4422 		sigaltstack_lock();
4423 		if (ss_mode == SS_DISABLE) {
4424 			ss_size = 0;
4425 			ss_sp = NULL;
4426 		} else {
4427 			if (unlikely(ss_size < min_ss_size))
4428 				ret = -ENOMEM;
4429 			if (!sigaltstack_size_valid(ss_size))
4430 				ret = -ENOMEM;
4431 		}
4432 		if (!ret) {
4433 			t->sas_ss_sp = (unsigned long) ss_sp;
4434 			t->sas_ss_size = ss_size;
4435 			t->sas_ss_flags = ss_flags;
4436 		}
4437 		sigaltstack_unlock();
4438 	}
4439 	return ret;
4440 }
4441 
4442 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4443 {
4444 	stack_t new, old;
4445 	int err;
4446 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4447 		return -EFAULT;
4448 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4449 			      current_user_stack_pointer(),
4450 			      MINSIGSTKSZ);
4451 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4452 		err = -EFAULT;
4453 	return err;
4454 }
4455 
4456 int restore_altstack(const stack_t __user *uss)
4457 {
4458 	stack_t new;
4459 	if (copy_from_user(&new, uss, sizeof(stack_t)))
4460 		return -EFAULT;
4461 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4462 			     MINSIGSTKSZ);
4463 	/* squash all but EFAULT for now */
4464 	return 0;
4465 }
4466 
4467 int __save_altstack(stack_t __user *uss, unsigned long sp)
4468 {
4469 	struct task_struct *t = current;
4470 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4471 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4472 		__put_user(t->sas_ss_size, &uss->ss_size);
4473 	return err;
4474 }
4475 
4476 #ifdef CONFIG_COMPAT
4477 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4478 				 compat_stack_t __user *uoss_ptr)
4479 {
4480 	stack_t uss, uoss;
4481 	int ret;
4482 
4483 	if (uss_ptr) {
4484 		compat_stack_t uss32;
4485 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4486 			return -EFAULT;
4487 		uss.ss_sp = compat_ptr(uss32.ss_sp);
4488 		uss.ss_flags = uss32.ss_flags;
4489 		uss.ss_size = uss32.ss_size;
4490 	}
4491 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4492 			     compat_user_stack_pointer(),
4493 			     COMPAT_MINSIGSTKSZ);
4494 	if (ret >= 0 && uoss_ptr)  {
4495 		compat_stack_t old;
4496 		memset(&old, 0, sizeof(old));
4497 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4498 		old.ss_flags = uoss.ss_flags;
4499 		old.ss_size = uoss.ss_size;
4500 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4501 			ret = -EFAULT;
4502 	}
4503 	return ret;
4504 }
4505 
4506 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4507 			const compat_stack_t __user *, uss_ptr,
4508 			compat_stack_t __user *, uoss_ptr)
4509 {
4510 	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4511 }
4512 
4513 int compat_restore_altstack(const compat_stack_t __user *uss)
4514 {
4515 	int err = do_compat_sigaltstack(uss, NULL);
4516 	/* squash all but -EFAULT for now */
4517 	return err == -EFAULT ? err : 0;
4518 }
4519 
4520 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4521 {
4522 	int err;
4523 	struct task_struct *t = current;
4524 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4525 			 &uss->ss_sp) |
4526 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4527 		__put_user(t->sas_ss_size, &uss->ss_size);
4528 	return err;
4529 }
4530 #endif
4531 
4532 #ifdef __ARCH_WANT_SYS_SIGPENDING
4533 
4534 /**
4535  *  sys_sigpending - examine pending signals
4536  *  @uset: where mask of pending signal is returned
4537  */
4538 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4539 {
4540 	sigset_t set;
4541 
4542 	if (sizeof(old_sigset_t) > sizeof(*uset))
4543 		return -EINVAL;
4544 
4545 	do_sigpending(&set);
4546 
4547 	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4548 		return -EFAULT;
4549 
4550 	return 0;
4551 }
4552 
4553 #ifdef CONFIG_COMPAT
4554 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4555 {
4556 	sigset_t set;
4557 
4558 	do_sigpending(&set);
4559 
4560 	return put_user(set.sig[0], set32);
4561 }
4562 #endif
4563 
4564 #endif
4565 
4566 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4567 /**
4568  *  sys_sigprocmask - examine and change blocked signals
4569  *  @how: whether to add, remove, or set signals
4570  *  @nset: signals to add or remove (if non-null)
4571  *  @oset: previous value of signal mask if non-null
4572  *
4573  * Some platforms have their own version with special arguments;
4574  * others support only sys_rt_sigprocmask.
4575  */
4576 
4577 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4578 		old_sigset_t __user *, oset)
4579 {
4580 	old_sigset_t old_set, new_set;
4581 	sigset_t new_blocked;
4582 
4583 	old_set = current->blocked.sig[0];
4584 
4585 	if (nset) {
4586 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4587 			return -EFAULT;
4588 
4589 		new_blocked = current->blocked;
4590 
4591 		switch (how) {
4592 		case SIG_BLOCK:
4593 			sigaddsetmask(&new_blocked, new_set);
4594 			break;
4595 		case SIG_UNBLOCK:
4596 			sigdelsetmask(&new_blocked, new_set);
4597 			break;
4598 		case SIG_SETMASK:
4599 			new_blocked.sig[0] = new_set;
4600 			break;
4601 		default:
4602 			return -EINVAL;
4603 		}
4604 
4605 		set_current_blocked(&new_blocked);
4606 	}
4607 
4608 	if (oset) {
4609 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4610 			return -EFAULT;
4611 	}
4612 
4613 	return 0;
4614 }
4615 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4616 
4617 #ifndef CONFIG_ODD_RT_SIGACTION
4618 /**
4619  *  sys_rt_sigaction - alter an action taken by a process
4620  *  @sig: signal to be sent
4621  *  @act: new sigaction
4622  *  @oact: used to save the previous sigaction
4623  *  @sigsetsize: size of sigset_t type
4624  */
4625 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4626 		const struct sigaction __user *, act,
4627 		struct sigaction __user *, oact,
4628 		size_t, sigsetsize)
4629 {
4630 	struct k_sigaction new_sa, old_sa;
4631 	int ret;
4632 
4633 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4634 	if (sigsetsize != sizeof(sigset_t))
4635 		return -EINVAL;
4636 
4637 	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4638 		return -EFAULT;
4639 
4640 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4641 	if (ret)
4642 		return ret;
4643 
4644 	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4645 		return -EFAULT;
4646 
4647 	return 0;
4648 }
4649 #ifdef CONFIG_COMPAT
4650 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4651 		const struct compat_sigaction __user *, act,
4652 		struct compat_sigaction __user *, oact,
4653 		compat_size_t, sigsetsize)
4654 {
4655 	struct k_sigaction new_ka, old_ka;
4656 #ifdef __ARCH_HAS_SA_RESTORER
4657 	compat_uptr_t restorer;
4658 #endif
4659 	int ret;
4660 
4661 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4662 	if (sigsetsize != sizeof(compat_sigset_t))
4663 		return -EINVAL;
4664 
4665 	if (act) {
4666 		compat_uptr_t handler;
4667 		ret = get_user(handler, &act->sa_handler);
4668 		new_ka.sa.sa_handler = compat_ptr(handler);
4669 #ifdef __ARCH_HAS_SA_RESTORER
4670 		ret |= get_user(restorer, &act->sa_restorer);
4671 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4672 #endif
4673 		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4674 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4675 		if (ret)
4676 			return -EFAULT;
4677 	}
4678 
4679 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4680 	if (!ret && oact) {
4681 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4682 			       &oact->sa_handler);
4683 		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4684 					 sizeof(oact->sa_mask));
4685 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4686 #ifdef __ARCH_HAS_SA_RESTORER
4687 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4688 				&oact->sa_restorer);
4689 #endif
4690 	}
4691 	return ret;
4692 }
4693 #endif
4694 #endif /* !CONFIG_ODD_RT_SIGACTION */
4695 
4696 #ifdef CONFIG_OLD_SIGACTION
4697 SYSCALL_DEFINE3(sigaction, int, sig,
4698 		const struct old_sigaction __user *, act,
4699 	        struct old_sigaction __user *, oact)
4700 {
4701 	struct k_sigaction new_ka, old_ka;
4702 	int ret;
4703 
4704 	if (act) {
4705 		old_sigset_t mask;
4706 		if (!access_ok(act, sizeof(*act)) ||
4707 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4708 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4709 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4710 		    __get_user(mask, &act->sa_mask))
4711 			return -EFAULT;
4712 #ifdef __ARCH_HAS_KA_RESTORER
4713 		new_ka.ka_restorer = NULL;
4714 #endif
4715 		siginitset(&new_ka.sa.sa_mask, mask);
4716 	}
4717 
4718 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4719 
4720 	if (!ret && oact) {
4721 		if (!access_ok(oact, sizeof(*oact)) ||
4722 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4723 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4724 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4725 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4726 			return -EFAULT;
4727 	}
4728 
4729 	return ret;
4730 }
4731 #endif
4732 #ifdef CONFIG_COMPAT_OLD_SIGACTION
4733 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4734 		const struct compat_old_sigaction __user *, act,
4735 	        struct compat_old_sigaction __user *, oact)
4736 {
4737 	struct k_sigaction new_ka, old_ka;
4738 	int ret;
4739 	compat_old_sigset_t mask;
4740 	compat_uptr_t handler, restorer;
4741 
4742 	if (act) {
4743 		if (!access_ok(act, sizeof(*act)) ||
4744 		    __get_user(handler, &act->sa_handler) ||
4745 		    __get_user(restorer, &act->sa_restorer) ||
4746 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4747 		    __get_user(mask, &act->sa_mask))
4748 			return -EFAULT;
4749 
4750 #ifdef __ARCH_HAS_KA_RESTORER
4751 		new_ka.ka_restorer = NULL;
4752 #endif
4753 		new_ka.sa.sa_handler = compat_ptr(handler);
4754 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4755 		siginitset(&new_ka.sa.sa_mask, mask);
4756 	}
4757 
4758 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4759 
4760 	if (!ret && oact) {
4761 		if (!access_ok(oact, sizeof(*oact)) ||
4762 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4763 			       &oact->sa_handler) ||
4764 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4765 			       &oact->sa_restorer) ||
4766 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4767 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4768 			return -EFAULT;
4769 	}
4770 	return ret;
4771 }
4772 #endif
4773 
4774 #ifdef CONFIG_SGETMASK_SYSCALL
4775 
4776 /*
4777  * For backwards compatibility.  Functionality superseded by sigprocmask.
4778  */
4779 SYSCALL_DEFINE0(sgetmask)
4780 {
4781 	/* SMP safe */
4782 	return current->blocked.sig[0];
4783 }
4784 
4785 SYSCALL_DEFINE1(ssetmask, int, newmask)
4786 {
4787 	int old = current->blocked.sig[0];
4788 	sigset_t newset;
4789 
4790 	siginitset(&newset, newmask);
4791 	set_current_blocked(&newset);
4792 
4793 	return old;
4794 }
4795 #endif /* CONFIG_SGETMASK_SYSCALL */
4796 
4797 #ifdef __ARCH_WANT_SYS_SIGNAL
4798 /*
4799  * For backwards compatibility.  Functionality superseded by sigaction.
4800  */
4801 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4802 {
4803 	struct k_sigaction new_sa, old_sa;
4804 	int ret;
4805 
4806 	new_sa.sa.sa_handler = handler;
4807 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4808 	sigemptyset(&new_sa.sa.sa_mask);
4809 
4810 	ret = do_sigaction(sig, &new_sa, &old_sa);
4811 
4812 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4813 }
4814 #endif /* __ARCH_WANT_SYS_SIGNAL */
4815 
4816 #ifdef __ARCH_WANT_SYS_PAUSE
4817 
4818 SYSCALL_DEFINE0(pause)
4819 {
4820 	while (!signal_pending(current)) {
4821 		__set_current_state(TASK_INTERRUPTIBLE);
4822 		schedule();
4823 	}
4824 	return -ERESTARTNOHAND;
4825 }
4826 
4827 #endif
4828 
4829 static int sigsuspend(sigset_t *set)
4830 {
4831 	current->saved_sigmask = current->blocked;
4832 	set_current_blocked(set);
4833 
4834 	while (!signal_pending(current)) {
4835 		__set_current_state(TASK_INTERRUPTIBLE);
4836 		schedule();
4837 	}
4838 	set_restore_sigmask();
4839 	return -ERESTARTNOHAND;
4840 }
4841 
4842 /**
4843  *  sys_rt_sigsuspend - replace the signal mask for a value with the
4844  *	@unewset value until a signal is received
4845  *  @unewset: new signal mask value
4846  *  @sigsetsize: size of sigset_t type
4847  */
4848 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4849 {
4850 	sigset_t newset;
4851 
4852 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4853 	if (sigsetsize != sizeof(sigset_t))
4854 		return -EINVAL;
4855 
4856 	if (copy_from_user(&newset, unewset, sizeof(newset)))
4857 		return -EFAULT;
4858 	return sigsuspend(&newset);
4859 }
4860 
4861 #ifdef CONFIG_COMPAT
4862 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4863 {
4864 	sigset_t newset;
4865 
4866 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4867 	if (sigsetsize != sizeof(sigset_t))
4868 		return -EINVAL;
4869 
4870 	if (get_compat_sigset(&newset, unewset))
4871 		return -EFAULT;
4872 	return sigsuspend(&newset);
4873 }
4874 #endif
4875 
4876 #ifdef CONFIG_OLD_SIGSUSPEND
4877 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4878 {
4879 	sigset_t blocked;
4880 	siginitset(&blocked, mask);
4881 	return sigsuspend(&blocked);
4882 }
4883 #endif
4884 #ifdef CONFIG_OLD_SIGSUSPEND3
4885 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4886 {
4887 	sigset_t blocked;
4888 	siginitset(&blocked, mask);
4889 	return sigsuspend(&blocked);
4890 }
4891 #endif
4892 
4893 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4894 {
4895 	return NULL;
4896 }
4897 
4898 static inline void siginfo_buildtime_checks(void)
4899 {
4900 	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4901 
4902 	/* Verify the offsets in the two siginfos match */
4903 #define CHECK_OFFSET(field) \
4904 	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4905 
4906 	/* kill */
4907 	CHECK_OFFSET(si_pid);
4908 	CHECK_OFFSET(si_uid);
4909 
4910 	/* timer */
4911 	CHECK_OFFSET(si_tid);
4912 	CHECK_OFFSET(si_overrun);
4913 	CHECK_OFFSET(si_value);
4914 
4915 	/* rt */
4916 	CHECK_OFFSET(si_pid);
4917 	CHECK_OFFSET(si_uid);
4918 	CHECK_OFFSET(si_value);
4919 
4920 	/* sigchld */
4921 	CHECK_OFFSET(si_pid);
4922 	CHECK_OFFSET(si_uid);
4923 	CHECK_OFFSET(si_status);
4924 	CHECK_OFFSET(si_utime);
4925 	CHECK_OFFSET(si_stime);
4926 
4927 	/* sigfault */
4928 	CHECK_OFFSET(si_addr);
4929 	CHECK_OFFSET(si_trapno);
4930 	CHECK_OFFSET(si_addr_lsb);
4931 	CHECK_OFFSET(si_lower);
4932 	CHECK_OFFSET(si_upper);
4933 	CHECK_OFFSET(si_pkey);
4934 	CHECK_OFFSET(si_perf_data);
4935 	CHECK_OFFSET(si_perf_type);
4936 	CHECK_OFFSET(si_perf_flags);
4937 
4938 	/* sigpoll */
4939 	CHECK_OFFSET(si_band);
4940 	CHECK_OFFSET(si_fd);
4941 
4942 	/* sigsys */
4943 	CHECK_OFFSET(si_call_addr);
4944 	CHECK_OFFSET(si_syscall);
4945 	CHECK_OFFSET(si_arch);
4946 #undef CHECK_OFFSET
4947 
4948 	/* usb asyncio */
4949 	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4950 		     offsetof(struct siginfo, si_addr));
4951 	if (sizeof(int) == sizeof(void __user *)) {
4952 		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4953 			     sizeof(void __user *));
4954 	} else {
4955 		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4956 			      sizeof_field(struct siginfo, si_uid)) !=
4957 			     sizeof(void __user *));
4958 		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4959 			     offsetof(struct siginfo, si_uid));
4960 	}
4961 #ifdef CONFIG_COMPAT
4962 	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4963 		     offsetof(struct compat_siginfo, si_addr));
4964 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4965 		     sizeof(compat_uptr_t));
4966 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4967 		     sizeof_field(struct siginfo, si_pid));
4968 #endif
4969 }
4970 
4971 #if defined(CONFIG_SYSCTL)
4972 static const struct ctl_table signal_debug_table[] = {
4973 #ifdef CONFIG_SYSCTL_EXCEPTION_TRACE
4974 	{
4975 		.procname	= "exception-trace",
4976 		.data		= &show_unhandled_signals,
4977 		.maxlen		= sizeof(int),
4978 		.mode		= 0644,
4979 		.proc_handler	= proc_dointvec
4980 	},
4981 #endif
4982 };
4983 
4984 static const struct ctl_table signal_table[] = {
4985 	{
4986 		.procname	= "print-fatal-signals",
4987 		.data		= &print_fatal_signals,
4988 		.maxlen		= sizeof(int),
4989 		.mode		= 0644,
4990 		.proc_handler	= proc_dointvec,
4991 	},
4992 };
4993 
4994 static int __init init_signal_sysctls(void)
4995 {
4996 	register_sysctl_init("debug", signal_debug_table);
4997 	register_sysctl_init("kernel", signal_table);
4998 	return 0;
4999 }
5000 early_initcall(init_signal_sysctls);
5001 #endif /* CONFIG_SYSCTL */
5002 
5003 void __init signals_init(void)
5004 {
5005 	siginfo_buildtime_checks();
5006 
5007 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
5008 }
5009 
5010 #ifdef CONFIG_KGDB_KDB
5011 #include <linux/kdb.h>
5012 /*
5013  * kdb_send_sig - Allows kdb to send signals without exposing
5014  * signal internals.  This function checks if the required locks are
5015  * available before calling the main signal code, to avoid kdb
5016  * deadlocks.
5017  */
5018 void kdb_send_sig(struct task_struct *t, int sig)
5019 {
5020 	static struct task_struct *kdb_prev_t;
5021 	int new_t, ret;
5022 	if (!spin_trylock(&t->sighand->siglock)) {
5023 		kdb_printf("Can't do kill command now.\n"
5024 			   "The sigmask lock is held somewhere else in "
5025 			   "kernel, try again later\n");
5026 		return;
5027 	}
5028 	new_t = kdb_prev_t != t;
5029 	kdb_prev_t = t;
5030 	if (!task_is_running(t) && new_t) {
5031 		spin_unlock(&t->sighand->siglock);
5032 		kdb_printf("Process is not RUNNING, sending a signal from "
5033 			   "kdb risks deadlock\n"
5034 			   "on the run queue locks. "
5035 			   "The signal has _not_ been sent.\n"
5036 			   "Reissue the kill command if you want to risk "
5037 			   "the deadlock.\n");
5038 		return;
5039 	}
5040 	ret = send_signal_locked(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
5041 	spin_unlock(&t->sighand->siglock);
5042 	if (ret)
5043 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
5044 			   sig, t->pid);
5045 	else
5046 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
5047 }
5048 #endif	/* CONFIG_KGDB_KDB */
5049