xref: /openbmc/linux/kernel/signal.c (revision fbb6b31a)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/kernel/signal.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  1997-11-02  Modified for POSIX.1b signals by Richard Henderson
8  *
9  *  2003-06-02  Jim Houston - Concurrent Computer Corp.
10  *		Changes to use preallocated sigqueue structures
11  *		to allow signals to be sent reliably.
12  */
13 
14 #include <linux/slab.h>
15 #include <linux/export.h>
16 #include <linux/init.h>
17 #include <linux/sched/mm.h>
18 #include <linux/sched/user.h>
19 #include <linux/sched/debug.h>
20 #include <linux/sched/task.h>
21 #include <linux/sched/task_stack.h>
22 #include <linux/sched/cputime.h>
23 #include <linux/file.h>
24 #include <linux/fs.h>
25 #include <linux/proc_fs.h>
26 #include <linux/tty.h>
27 #include <linux/binfmts.h>
28 #include <linux/coredump.h>
29 #include <linux/security.h>
30 #include <linux/syscalls.h>
31 #include <linux/ptrace.h>
32 #include <linux/signal.h>
33 #include <linux/signalfd.h>
34 #include <linux/ratelimit.h>
35 #include <linux/task_work.h>
36 #include <linux/capability.h>
37 #include <linux/freezer.h>
38 #include <linux/pid_namespace.h>
39 #include <linux/nsproxy.h>
40 #include <linux/user_namespace.h>
41 #include <linux/uprobes.h>
42 #include <linux/compat.h>
43 #include <linux/cn_proc.h>
44 #include <linux/compiler.h>
45 #include <linux/posix-timers.h>
46 #include <linux/cgroup.h>
47 #include <linux/audit.h>
48 
49 #define CREATE_TRACE_POINTS
50 #include <trace/events/signal.h>
51 
52 #include <asm/param.h>
53 #include <linux/uaccess.h>
54 #include <asm/unistd.h>
55 #include <asm/siginfo.h>
56 #include <asm/cacheflush.h>
57 #include <asm/syscall.h>	/* for syscall_get_* */
58 
59 /*
60  * SLAB caches for signal bits.
61  */
62 
63 static struct kmem_cache *sigqueue_cachep;
64 
65 int print_fatal_signals __read_mostly;
66 
67 static void __user *sig_handler(struct task_struct *t, int sig)
68 {
69 	return t->sighand->action[sig - 1].sa.sa_handler;
70 }
71 
72 static inline bool sig_handler_ignored(void __user *handler, int sig)
73 {
74 	/* Is it explicitly or implicitly ignored? */
75 	return handler == SIG_IGN ||
76 	       (handler == SIG_DFL && sig_kernel_ignore(sig));
77 }
78 
79 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
80 {
81 	void __user *handler;
82 
83 	handler = sig_handler(t, sig);
84 
85 	/* SIGKILL and SIGSTOP may not be sent to the global init */
86 	if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
87 		return true;
88 
89 	if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
90 	    handler == SIG_DFL && !(force && sig_kernel_only(sig)))
91 		return true;
92 
93 	/* Only allow kernel generated signals to this kthread */
94 	if (unlikely((t->flags & PF_KTHREAD) &&
95 		     (handler == SIG_KTHREAD_KERNEL) && !force))
96 		return true;
97 
98 	return sig_handler_ignored(handler, sig);
99 }
100 
101 static bool sig_ignored(struct task_struct *t, int sig, bool force)
102 {
103 	/*
104 	 * Blocked signals are never ignored, since the
105 	 * signal handler may change by the time it is
106 	 * unblocked.
107 	 */
108 	if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
109 		return false;
110 
111 	/*
112 	 * Tracers may want to know about even ignored signal unless it
113 	 * is SIGKILL which can't be reported anyway but can be ignored
114 	 * by SIGNAL_UNKILLABLE task.
115 	 */
116 	if (t->ptrace && sig != SIGKILL)
117 		return false;
118 
119 	return sig_task_ignored(t, sig, force);
120 }
121 
122 /*
123  * Re-calculate pending state from the set of locally pending
124  * signals, globally pending signals, and blocked signals.
125  */
126 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
127 {
128 	unsigned long ready;
129 	long i;
130 
131 	switch (_NSIG_WORDS) {
132 	default:
133 		for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
134 			ready |= signal->sig[i] &~ blocked->sig[i];
135 		break;
136 
137 	case 4: ready  = signal->sig[3] &~ blocked->sig[3];
138 		ready |= signal->sig[2] &~ blocked->sig[2];
139 		ready |= signal->sig[1] &~ blocked->sig[1];
140 		ready |= signal->sig[0] &~ blocked->sig[0];
141 		break;
142 
143 	case 2: ready  = signal->sig[1] &~ blocked->sig[1];
144 		ready |= signal->sig[0] &~ blocked->sig[0];
145 		break;
146 
147 	case 1: ready  = signal->sig[0] &~ blocked->sig[0];
148 	}
149 	return ready !=	0;
150 }
151 
152 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
153 
154 static bool recalc_sigpending_tsk(struct task_struct *t)
155 {
156 	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
157 	    PENDING(&t->pending, &t->blocked) ||
158 	    PENDING(&t->signal->shared_pending, &t->blocked) ||
159 	    cgroup_task_frozen(t)) {
160 		set_tsk_thread_flag(t, TIF_SIGPENDING);
161 		return true;
162 	}
163 
164 	/*
165 	 * We must never clear the flag in another thread, or in current
166 	 * when it's possible the current syscall is returning -ERESTART*.
167 	 * So we don't clear it here, and only callers who know they should do.
168 	 */
169 	return false;
170 }
171 
172 /*
173  * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
174  * This is superfluous when called on current, the wakeup is a harmless no-op.
175  */
176 void recalc_sigpending_and_wake(struct task_struct *t)
177 {
178 	if (recalc_sigpending_tsk(t))
179 		signal_wake_up(t, 0);
180 }
181 
182 void recalc_sigpending(void)
183 {
184 	if (!recalc_sigpending_tsk(current) && !freezing(current))
185 		clear_thread_flag(TIF_SIGPENDING);
186 
187 }
188 EXPORT_SYMBOL(recalc_sigpending);
189 
190 void calculate_sigpending(void)
191 {
192 	/* Have any signals or users of TIF_SIGPENDING been delayed
193 	 * until after fork?
194 	 */
195 	spin_lock_irq(&current->sighand->siglock);
196 	set_tsk_thread_flag(current, TIF_SIGPENDING);
197 	recalc_sigpending();
198 	spin_unlock_irq(&current->sighand->siglock);
199 }
200 
201 /* Given the mask, find the first available signal that should be serviced. */
202 
203 #define SYNCHRONOUS_MASK \
204 	(sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
205 	 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
206 
207 int next_signal(struct sigpending *pending, sigset_t *mask)
208 {
209 	unsigned long i, *s, *m, x;
210 	int sig = 0;
211 
212 	s = pending->signal.sig;
213 	m = mask->sig;
214 
215 	/*
216 	 * Handle the first word specially: it contains the
217 	 * synchronous signals that need to be dequeued first.
218 	 */
219 	x = *s &~ *m;
220 	if (x) {
221 		if (x & SYNCHRONOUS_MASK)
222 			x &= SYNCHRONOUS_MASK;
223 		sig = ffz(~x) + 1;
224 		return sig;
225 	}
226 
227 	switch (_NSIG_WORDS) {
228 	default:
229 		for (i = 1; i < _NSIG_WORDS; ++i) {
230 			x = *++s &~ *++m;
231 			if (!x)
232 				continue;
233 			sig = ffz(~x) + i*_NSIG_BPW + 1;
234 			break;
235 		}
236 		break;
237 
238 	case 2:
239 		x = s[1] &~ m[1];
240 		if (!x)
241 			break;
242 		sig = ffz(~x) + _NSIG_BPW + 1;
243 		break;
244 
245 	case 1:
246 		/* Nothing to do */
247 		break;
248 	}
249 
250 	return sig;
251 }
252 
253 static inline void print_dropped_signal(int sig)
254 {
255 	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
256 
257 	if (!print_fatal_signals)
258 		return;
259 
260 	if (!__ratelimit(&ratelimit_state))
261 		return;
262 
263 	pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
264 				current->comm, current->pid, sig);
265 }
266 
267 /**
268  * task_set_jobctl_pending - set jobctl pending bits
269  * @task: target task
270  * @mask: pending bits to set
271  *
272  * Clear @mask from @task->jobctl.  @mask must be subset of
273  * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
274  * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
275  * cleared.  If @task is already being killed or exiting, this function
276  * becomes noop.
277  *
278  * CONTEXT:
279  * Must be called with @task->sighand->siglock held.
280  *
281  * RETURNS:
282  * %true if @mask is set, %false if made noop because @task was dying.
283  */
284 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
285 {
286 	BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
287 			JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
288 	BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
289 
290 	if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
291 		return false;
292 
293 	if (mask & JOBCTL_STOP_SIGMASK)
294 		task->jobctl &= ~JOBCTL_STOP_SIGMASK;
295 
296 	task->jobctl |= mask;
297 	return true;
298 }
299 
300 /**
301  * task_clear_jobctl_trapping - clear jobctl trapping bit
302  * @task: target task
303  *
304  * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
305  * Clear it and wake up the ptracer.  Note that we don't need any further
306  * locking.  @task->siglock guarantees that @task->parent points to the
307  * ptracer.
308  *
309  * CONTEXT:
310  * Must be called with @task->sighand->siglock held.
311  */
312 void task_clear_jobctl_trapping(struct task_struct *task)
313 {
314 	if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
315 		task->jobctl &= ~JOBCTL_TRAPPING;
316 		smp_mb();	/* advised by wake_up_bit() */
317 		wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
318 	}
319 }
320 
321 /**
322  * task_clear_jobctl_pending - clear jobctl pending bits
323  * @task: target task
324  * @mask: pending bits to clear
325  *
326  * Clear @mask from @task->jobctl.  @mask must be subset of
327  * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
328  * STOP bits are cleared together.
329  *
330  * If clearing of @mask leaves no stop or trap pending, this function calls
331  * task_clear_jobctl_trapping().
332  *
333  * CONTEXT:
334  * Must be called with @task->sighand->siglock held.
335  */
336 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
337 {
338 	BUG_ON(mask & ~JOBCTL_PENDING_MASK);
339 
340 	if (mask & JOBCTL_STOP_PENDING)
341 		mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
342 
343 	task->jobctl &= ~mask;
344 
345 	if (!(task->jobctl & JOBCTL_PENDING_MASK))
346 		task_clear_jobctl_trapping(task);
347 }
348 
349 /**
350  * task_participate_group_stop - participate in a group stop
351  * @task: task participating in a group stop
352  *
353  * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
354  * Group stop states are cleared and the group stop count is consumed if
355  * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
356  * stop, the appropriate `SIGNAL_*` flags are set.
357  *
358  * CONTEXT:
359  * Must be called with @task->sighand->siglock held.
360  *
361  * RETURNS:
362  * %true if group stop completion should be notified to the parent, %false
363  * otherwise.
364  */
365 static bool task_participate_group_stop(struct task_struct *task)
366 {
367 	struct signal_struct *sig = task->signal;
368 	bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
369 
370 	WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
371 
372 	task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
373 
374 	if (!consume)
375 		return false;
376 
377 	if (!WARN_ON_ONCE(sig->group_stop_count == 0))
378 		sig->group_stop_count--;
379 
380 	/*
381 	 * Tell the caller to notify completion iff we are entering into a
382 	 * fresh group stop.  Read comment in do_signal_stop() for details.
383 	 */
384 	if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
385 		signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
386 		return true;
387 	}
388 	return false;
389 }
390 
391 void task_join_group_stop(struct task_struct *task)
392 {
393 	unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
394 	struct signal_struct *sig = current->signal;
395 
396 	if (sig->group_stop_count) {
397 		sig->group_stop_count++;
398 		mask |= JOBCTL_STOP_CONSUME;
399 	} else if (!(sig->flags & SIGNAL_STOP_STOPPED))
400 		return;
401 
402 	/* Have the new thread join an on-going signal group stop */
403 	task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
404 }
405 
406 /*
407  * allocate a new signal queue record
408  * - this may be called without locks if and only if t == current, otherwise an
409  *   appropriate lock must be held to stop the target task from exiting
410  */
411 static struct sigqueue *
412 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t gfp_flags,
413 		 int override_rlimit, const unsigned int sigqueue_flags)
414 {
415 	struct sigqueue *q = NULL;
416 	struct ucounts *ucounts = NULL;
417 	long sigpending;
418 
419 	/*
420 	 * Protect access to @t credentials. This can go away when all
421 	 * callers hold rcu read lock.
422 	 *
423 	 * NOTE! A pending signal will hold on to the user refcount,
424 	 * and we get/put the refcount only when the sigpending count
425 	 * changes from/to zero.
426 	 */
427 	rcu_read_lock();
428 	ucounts = task_ucounts(t);
429 	sigpending = inc_rlimit_get_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
430 	rcu_read_unlock();
431 	if (!sigpending)
432 		return NULL;
433 
434 	if (override_rlimit || likely(sigpending <= task_rlimit(t, RLIMIT_SIGPENDING))) {
435 		q = kmem_cache_alloc(sigqueue_cachep, gfp_flags);
436 	} else {
437 		print_dropped_signal(sig);
438 	}
439 
440 	if (unlikely(q == NULL)) {
441 		dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
442 	} else {
443 		INIT_LIST_HEAD(&q->list);
444 		q->flags = sigqueue_flags;
445 		q->ucounts = ucounts;
446 	}
447 	return q;
448 }
449 
450 static void __sigqueue_free(struct sigqueue *q)
451 {
452 	if (q->flags & SIGQUEUE_PREALLOC)
453 		return;
454 	if (q->ucounts) {
455 		dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING);
456 		q->ucounts = NULL;
457 	}
458 	kmem_cache_free(sigqueue_cachep, q);
459 }
460 
461 void flush_sigqueue(struct sigpending *queue)
462 {
463 	struct sigqueue *q;
464 
465 	sigemptyset(&queue->signal);
466 	while (!list_empty(&queue->list)) {
467 		q = list_entry(queue->list.next, struct sigqueue , list);
468 		list_del_init(&q->list);
469 		__sigqueue_free(q);
470 	}
471 }
472 
473 /*
474  * Flush all pending signals for this kthread.
475  */
476 void flush_signals(struct task_struct *t)
477 {
478 	unsigned long flags;
479 
480 	spin_lock_irqsave(&t->sighand->siglock, flags);
481 	clear_tsk_thread_flag(t, TIF_SIGPENDING);
482 	flush_sigqueue(&t->pending);
483 	flush_sigqueue(&t->signal->shared_pending);
484 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
485 }
486 EXPORT_SYMBOL(flush_signals);
487 
488 #ifdef CONFIG_POSIX_TIMERS
489 static void __flush_itimer_signals(struct sigpending *pending)
490 {
491 	sigset_t signal, retain;
492 	struct sigqueue *q, *n;
493 
494 	signal = pending->signal;
495 	sigemptyset(&retain);
496 
497 	list_for_each_entry_safe(q, n, &pending->list, list) {
498 		int sig = q->info.si_signo;
499 
500 		if (likely(q->info.si_code != SI_TIMER)) {
501 			sigaddset(&retain, sig);
502 		} else {
503 			sigdelset(&signal, sig);
504 			list_del_init(&q->list);
505 			__sigqueue_free(q);
506 		}
507 	}
508 
509 	sigorsets(&pending->signal, &signal, &retain);
510 }
511 
512 void flush_itimer_signals(void)
513 {
514 	struct task_struct *tsk = current;
515 	unsigned long flags;
516 
517 	spin_lock_irqsave(&tsk->sighand->siglock, flags);
518 	__flush_itimer_signals(&tsk->pending);
519 	__flush_itimer_signals(&tsk->signal->shared_pending);
520 	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
521 }
522 #endif
523 
524 void ignore_signals(struct task_struct *t)
525 {
526 	int i;
527 
528 	for (i = 0; i < _NSIG; ++i)
529 		t->sighand->action[i].sa.sa_handler = SIG_IGN;
530 
531 	flush_signals(t);
532 }
533 
534 /*
535  * Flush all handlers for a task.
536  */
537 
538 void
539 flush_signal_handlers(struct task_struct *t, int force_default)
540 {
541 	int i;
542 	struct k_sigaction *ka = &t->sighand->action[0];
543 	for (i = _NSIG ; i != 0 ; i--) {
544 		if (force_default || ka->sa.sa_handler != SIG_IGN)
545 			ka->sa.sa_handler = SIG_DFL;
546 		ka->sa.sa_flags = 0;
547 #ifdef __ARCH_HAS_SA_RESTORER
548 		ka->sa.sa_restorer = NULL;
549 #endif
550 		sigemptyset(&ka->sa.sa_mask);
551 		ka++;
552 	}
553 }
554 
555 bool unhandled_signal(struct task_struct *tsk, int sig)
556 {
557 	void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
558 	if (is_global_init(tsk))
559 		return true;
560 
561 	if (handler != SIG_IGN && handler != SIG_DFL)
562 		return false;
563 
564 	/* if ptraced, let the tracer determine */
565 	return !tsk->ptrace;
566 }
567 
568 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
569 			   bool *resched_timer)
570 {
571 	struct sigqueue *q, *first = NULL;
572 
573 	/*
574 	 * Collect the siginfo appropriate to this signal.  Check if
575 	 * there is another siginfo for the same signal.
576 	*/
577 	list_for_each_entry(q, &list->list, list) {
578 		if (q->info.si_signo == sig) {
579 			if (first)
580 				goto still_pending;
581 			first = q;
582 		}
583 	}
584 
585 	sigdelset(&list->signal, sig);
586 
587 	if (first) {
588 still_pending:
589 		list_del_init(&first->list);
590 		copy_siginfo(info, &first->info);
591 
592 		*resched_timer =
593 			(first->flags & SIGQUEUE_PREALLOC) &&
594 			(info->si_code == SI_TIMER) &&
595 			(info->si_sys_private);
596 
597 		__sigqueue_free(first);
598 	} else {
599 		/*
600 		 * Ok, it wasn't in the queue.  This must be
601 		 * a fast-pathed signal or we must have been
602 		 * out of queue space.  So zero out the info.
603 		 */
604 		clear_siginfo(info);
605 		info->si_signo = sig;
606 		info->si_errno = 0;
607 		info->si_code = SI_USER;
608 		info->si_pid = 0;
609 		info->si_uid = 0;
610 	}
611 }
612 
613 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
614 			kernel_siginfo_t *info, bool *resched_timer)
615 {
616 	int sig = next_signal(pending, mask);
617 
618 	if (sig)
619 		collect_signal(sig, pending, info, resched_timer);
620 	return sig;
621 }
622 
623 /*
624  * Dequeue a signal and return the element to the caller, which is
625  * expected to free it.
626  *
627  * All callers have to hold the siglock.
628  */
629 int dequeue_signal(struct task_struct *tsk, sigset_t *mask,
630 		   kernel_siginfo_t *info, enum pid_type *type)
631 {
632 	bool resched_timer = false;
633 	int signr;
634 
635 	/* We only dequeue private signals from ourselves, we don't let
636 	 * signalfd steal them
637 	 */
638 	*type = PIDTYPE_PID;
639 	signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
640 	if (!signr) {
641 		*type = PIDTYPE_TGID;
642 		signr = __dequeue_signal(&tsk->signal->shared_pending,
643 					 mask, info, &resched_timer);
644 #ifdef CONFIG_POSIX_TIMERS
645 		/*
646 		 * itimer signal ?
647 		 *
648 		 * itimers are process shared and we restart periodic
649 		 * itimers in the signal delivery path to prevent DoS
650 		 * attacks in the high resolution timer case. This is
651 		 * compliant with the old way of self-restarting
652 		 * itimers, as the SIGALRM is a legacy signal and only
653 		 * queued once. Changing the restart behaviour to
654 		 * restart the timer in the signal dequeue path is
655 		 * reducing the timer noise on heavy loaded !highres
656 		 * systems too.
657 		 */
658 		if (unlikely(signr == SIGALRM)) {
659 			struct hrtimer *tmr = &tsk->signal->real_timer;
660 
661 			if (!hrtimer_is_queued(tmr) &&
662 			    tsk->signal->it_real_incr != 0) {
663 				hrtimer_forward(tmr, tmr->base->get_time(),
664 						tsk->signal->it_real_incr);
665 				hrtimer_restart(tmr);
666 			}
667 		}
668 #endif
669 	}
670 
671 	recalc_sigpending();
672 	if (!signr)
673 		return 0;
674 
675 	if (unlikely(sig_kernel_stop(signr))) {
676 		/*
677 		 * Set a marker that we have dequeued a stop signal.  Our
678 		 * caller might release the siglock and then the pending
679 		 * stop signal it is about to process is no longer in the
680 		 * pending bitmasks, but must still be cleared by a SIGCONT
681 		 * (and overruled by a SIGKILL).  So those cases clear this
682 		 * shared flag after we've set it.  Note that this flag may
683 		 * remain set after the signal we return is ignored or
684 		 * handled.  That doesn't matter because its only purpose
685 		 * is to alert stop-signal processing code when another
686 		 * processor has come along and cleared the flag.
687 		 */
688 		current->jobctl |= JOBCTL_STOP_DEQUEUED;
689 	}
690 #ifdef CONFIG_POSIX_TIMERS
691 	if (resched_timer) {
692 		/*
693 		 * Release the siglock to ensure proper locking order
694 		 * of timer locks outside of siglocks.  Note, we leave
695 		 * irqs disabled here, since the posix-timers code is
696 		 * about to disable them again anyway.
697 		 */
698 		spin_unlock(&tsk->sighand->siglock);
699 		posixtimer_rearm(info);
700 		spin_lock(&tsk->sighand->siglock);
701 
702 		/* Don't expose the si_sys_private value to userspace */
703 		info->si_sys_private = 0;
704 	}
705 #endif
706 	return signr;
707 }
708 EXPORT_SYMBOL_GPL(dequeue_signal);
709 
710 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
711 {
712 	struct task_struct *tsk = current;
713 	struct sigpending *pending = &tsk->pending;
714 	struct sigqueue *q, *sync = NULL;
715 
716 	/*
717 	 * Might a synchronous signal be in the queue?
718 	 */
719 	if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
720 		return 0;
721 
722 	/*
723 	 * Return the first synchronous signal in the queue.
724 	 */
725 	list_for_each_entry(q, &pending->list, list) {
726 		/* Synchronous signals have a positive si_code */
727 		if ((q->info.si_code > SI_USER) &&
728 		    (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
729 			sync = q;
730 			goto next;
731 		}
732 	}
733 	return 0;
734 next:
735 	/*
736 	 * Check if there is another siginfo for the same signal.
737 	 */
738 	list_for_each_entry_continue(q, &pending->list, list) {
739 		if (q->info.si_signo == sync->info.si_signo)
740 			goto still_pending;
741 	}
742 
743 	sigdelset(&pending->signal, sync->info.si_signo);
744 	recalc_sigpending();
745 still_pending:
746 	list_del_init(&sync->list);
747 	copy_siginfo(info, &sync->info);
748 	__sigqueue_free(sync);
749 	return info->si_signo;
750 }
751 
752 /*
753  * Tell a process that it has a new active signal..
754  *
755  * NOTE! we rely on the previous spin_lock to
756  * lock interrupts for us! We can only be called with
757  * "siglock" held, and the local interrupt must
758  * have been disabled when that got acquired!
759  *
760  * No need to set need_resched since signal event passing
761  * goes through ->blocked
762  */
763 void signal_wake_up_state(struct task_struct *t, unsigned int state)
764 {
765 	set_tsk_thread_flag(t, TIF_SIGPENDING);
766 	/*
767 	 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
768 	 * case. We don't check t->state here because there is a race with it
769 	 * executing another processor and just now entering stopped state.
770 	 * By using wake_up_state, we ensure the process will wake up and
771 	 * handle its death signal.
772 	 */
773 	if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
774 		kick_process(t);
775 }
776 
777 /*
778  * Remove signals in mask from the pending set and queue.
779  * Returns 1 if any signals were found.
780  *
781  * All callers must be holding the siglock.
782  */
783 static void flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
784 {
785 	struct sigqueue *q, *n;
786 	sigset_t m;
787 
788 	sigandsets(&m, mask, &s->signal);
789 	if (sigisemptyset(&m))
790 		return;
791 
792 	sigandnsets(&s->signal, &s->signal, mask);
793 	list_for_each_entry_safe(q, n, &s->list, list) {
794 		if (sigismember(mask, q->info.si_signo)) {
795 			list_del_init(&q->list);
796 			__sigqueue_free(q);
797 		}
798 	}
799 }
800 
801 static inline int is_si_special(const struct kernel_siginfo *info)
802 {
803 	return info <= SEND_SIG_PRIV;
804 }
805 
806 static inline bool si_fromuser(const struct kernel_siginfo *info)
807 {
808 	return info == SEND_SIG_NOINFO ||
809 		(!is_si_special(info) && SI_FROMUSER(info));
810 }
811 
812 /*
813  * called with RCU read lock from check_kill_permission()
814  */
815 static bool kill_ok_by_cred(struct task_struct *t)
816 {
817 	const struct cred *cred = current_cred();
818 	const struct cred *tcred = __task_cred(t);
819 
820 	return uid_eq(cred->euid, tcred->suid) ||
821 	       uid_eq(cred->euid, tcred->uid) ||
822 	       uid_eq(cred->uid, tcred->suid) ||
823 	       uid_eq(cred->uid, tcred->uid) ||
824 	       ns_capable(tcred->user_ns, CAP_KILL);
825 }
826 
827 /*
828  * Bad permissions for sending the signal
829  * - the caller must hold the RCU read lock
830  */
831 static int check_kill_permission(int sig, struct kernel_siginfo *info,
832 				 struct task_struct *t)
833 {
834 	struct pid *sid;
835 	int error;
836 
837 	if (!valid_signal(sig))
838 		return -EINVAL;
839 
840 	if (!si_fromuser(info))
841 		return 0;
842 
843 	error = audit_signal_info(sig, t); /* Let audit system see the signal */
844 	if (error)
845 		return error;
846 
847 	if (!same_thread_group(current, t) &&
848 	    !kill_ok_by_cred(t)) {
849 		switch (sig) {
850 		case SIGCONT:
851 			sid = task_session(t);
852 			/*
853 			 * We don't return the error if sid == NULL. The
854 			 * task was unhashed, the caller must notice this.
855 			 */
856 			if (!sid || sid == task_session(current))
857 				break;
858 			fallthrough;
859 		default:
860 			return -EPERM;
861 		}
862 	}
863 
864 	return security_task_kill(t, info, sig, NULL);
865 }
866 
867 /**
868  * ptrace_trap_notify - schedule trap to notify ptracer
869  * @t: tracee wanting to notify tracer
870  *
871  * This function schedules sticky ptrace trap which is cleared on the next
872  * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
873  * ptracer.
874  *
875  * If @t is running, STOP trap will be taken.  If trapped for STOP and
876  * ptracer is listening for events, tracee is woken up so that it can
877  * re-trap for the new event.  If trapped otherwise, STOP trap will be
878  * eventually taken without returning to userland after the existing traps
879  * are finished by PTRACE_CONT.
880  *
881  * CONTEXT:
882  * Must be called with @task->sighand->siglock held.
883  */
884 static void ptrace_trap_notify(struct task_struct *t)
885 {
886 	WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
887 	assert_spin_locked(&t->sighand->siglock);
888 
889 	task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
890 	ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
891 }
892 
893 /*
894  * Handle magic process-wide effects of stop/continue signals. Unlike
895  * the signal actions, these happen immediately at signal-generation
896  * time regardless of blocking, ignoring, or handling.  This does the
897  * actual continuing for SIGCONT, but not the actual stopping for stop
898  * signals. The process stop is done as a signal action for SIG_DFL.
899  *
900  * Returns true if the signal should be actually delivered, otherwise
901  * it should be dropped.
902  */
903 static bool prepare_signal(int sig, struct task_struct *p, bool force)
904 {
905 	struct signal_struct *signal = p->signal;
906 	struct task_struct *t;
907 	sigset_t flush;
908 
909 	if (signal->flags & SIGNAL_GROUP_EXIT) {
910 		if (signal->core_state)
911 			return sig == SIGKILL;
912 		/*
913 		 * The process is in the middle of dying, nothing to do.
914 		 */
915 	} else if (sig_kernel_stop(sig)) {
916 		/*
917 		 * This is a stop signal.  Remove SIGCONT from all queues.
918 		 */
919 		siginitset(&flush, sigmask(SIGCONT));
920 		flush_sigqueue_mask(&flush, &signal->shared_pending);
921 		for_each_thread(p, t)
922 			flush_sigqueue_mask(&flush, &t->pending);
923 	} else if (sig == SIGCONT) {
924 		unsigned int why;
925 		/*
926 		 * Remove all stop signals from all queues, wake all threads.
927 		 */
928 		siginitset(&flush, SIG_KERNEL_STOP_MASK);
929 		flush_sigqueue_mask(&flush, &signal->shared_pending);
930 		for_each_thread(p, t) {
931 			flush_sigqueue_mask(&flush, &t->pending);
932 			task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
933 			if (likely(!(t->ptrace & PT_SEIZED)))
934 				wake_up_state(t, __TASK_STOPPED);
935 			else
936 				ptrace_trap_notify(t);
937 		}
938 
939 		/*
940 		 * Notify the parent with CLD_CONTINUED if we were stopped.
941 		 *
942 		 * If we were in the middle of a group stop, we pretend it
943 		 * was already finished, and then continued. Since SIGCHLD
944 		 * doesn't queue we report only CLD_STOPPED, as if the next
945 		 * CLD_CONTINUED was dropped.
946 		 */
947 		why = 0;
948 		if (signal->flags & SIGNAL_STOP_STOPPED)
949 			why |= SIGNAL_CLD_CONTINUED;
950 		else if (signal->group_stop_count)
951 			why |= SIGNAL_CLD_STOPPED;
952 
953 		if (why) {
954 			/*
955 			 * The first thread which returns from do_signal_stop()
956 			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
957 			 * notify its parent. See get_signal().
958 			 */
959 			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
960 			signal->group_stop_count = 0;
961 			signal->group_exit_code = 0;
962 		}
963 	}
964 
965 	return !sig_ignored(p, sig, force);
966 }
967 
968 /*
969  * Test if P wants to take SIG.  After we've checked all threads with this,
970  * it's equivalent to finding no threads not blocking SIG.  Any threads not
971  * blocking SIG were ruled out because they are not running and already
972  * have pending signals.  Such threads will dequeue from the shared queue
973  * as soon as they're available, so putting the signal on the shared queue
974  * will be equivalent to sending it to one such thread.
975  */
976 static inline bool wants_signal(int sig, struct task_struct *p)
977 {
978 	if (sigismember(&p->blocked, sig))
979 		return false;
980 
981 	if (p->flags & PF_EXITING)
982 		return false;
983 
984 	if (sig == SIGKILL)
985 		return true;
986 
987 	if (task_is_stopped_or_traced(p))
988 		return false;
989 
990 	return task_curr(p) || !task_sigpending(p);
991 }
992 
993 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
994 {
995 	struct signal_struct *signal = p->signal;
996 	struct task_struct *t;
997 
998 	/*
999 	 * Now find a thread we can wake up to take the signal off the queue.
1000 	 *
1001 	 * If the main thread wants the signal, it gets first crack.
1002 	 * Probably the least surprising to the average bear.
1003 	 */
1004 	if (wants_signal(sig, p))
1005 		t = p;
1006 	else if ((type == PIDTYPE_PID) || thread_group_empty(p))
1007 		/*
1008 		 * There is just one thread and it does not need to be woken.
1009 		 * It will dequeue unblocked signals before it runs again.
1010 		 */
1011 		return;
1012 	else {
1013 		/*
1014 		 * Otherwise try to find a suitable thread.
1015 		 */
1016 		t = signal->curr_target;
1017 		while (!wants_signal(sig, t)) {
1018 			t = next_thread(t);
1019 			if (t == signal->curr_target)
1020 				/*
1021 				 * No thread needs to be woken.
1022 				 * Any eligible threads will see
1023 				 * the signal in the queue soon.
1024 				 */
1025 				return;
1026 		}
1027 		signal->curr_target = t;
1028 	}
1029 
1030 	/*
1031 	 * Found a killable thread.  If the signal will be fatal,
1032 	 * then start taking the whole group down immediately.
1033 	 */
1034 	if (sig_fatal(p, sig) &&
1035 	    (signal->core_state || !(signal->flags & SIGNAL_GROUP_EXIT)) &&
1036 	    !sigismember(&t->real_blocked, sig) &&
1037 	    (sig == SIGKILL || !p->ptrace)) {
1038 		/*
1039 		 * This signal will be fatal to the whole group.
1040 		 */
1041 		if (!sig_kernel_coredump(sig)) {
1042 			/*
1043 			 * Start a group exit and wake everybody up.
1044 			 * This way we don't have other threads
1045 			 * running and doing things after a slower
1046 			 * thread has the fatal signal pending.
1047 			 */
1048 			signal->flags = SIGNAL_GROUP_EXIT;
1049 			signal->group_exit_code = sig;
1050 			signal->group_stop_count = 0;
1051 			t = p;
1052 			do {
1053 				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1054 				sigaddset(&t->pending.signal, SIGKILL);
1055 				signal_wake_up(t, 1);
1056 			} while_each_thread(p, t);
1057 			return;
1058 		}
1059 	}
1060 
1061 	/*
1062 	 * The signal is already in the shared-pending queue.
1063 	 * Tell the chosen thread to wake up and dequeue it.
1064 	 */
1065 	signal_wake_up(t, sig == SIGKILL);
1066 	return;
1067 }
1068 
1069 static inline bool legacy_queue(struct sigpending *signals, int sig)
1070 {
1071 	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1072 }
1073 
1074 static int __send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1075 			enum pid_type type, bool force)
1076 {
1077 	struct sigpending *pending;
1078 	struct sigqueue *q;
1079 	int override_rlimit;
1080 	int ret = 0, result;
1081 
1082 	assert_spin_locked(&t->sighand->siglock);
1083 
1084 	result = TRACE_SIGNAL_IGNORED;
1085 	if (!prepare_signal(sig, t, force))
1086 		goto ret;
1087 
1088 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1089 	/*
1090 	 * Short-circuit ignored signals and support queuing
1091 	 * exactly one non-rt signal, so that we can get more
1092 	 * detailed information about the cause of the signal.
1093 	 */
1094 	result = TRACE_SIGNAL_ALREADY_PENDING;
1095 	if (legacy_queue(pending, sig))
1096 		goto ret;
1097 
1098 	result = TRACE_SIGNAL_DELIVERED;
1099 	/*
1100 	 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1101 	 */
1102 	if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1103 		goto out_set;
1104 
1105 	/*
1106 	 * Real-time signals must be queued if sent by sigqueue, or
1107 	 * some other real-time mechanism.  It is implementation
1108 	 * defined whether kill() does so.  We attempt to do so, on
1109 	 * the principle of least surprise, but since kill is not
1110 	 * allowed to fail with EAGAIN when low on memory we just
1111 	 * make sure at least one signal gets delivered and don't
1112 	 * pass on the info struct.
1113 	 */
1114 	if (sig < SIGRTMIN)
1115 		override_rlimit = (is_si_special(info) || info->si_code >= 0);
1116 	else
1117 		override_rlimit = 0;
1118 
1119 	q = __sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit, 0);
1120 
1121 	if (q) {
1122 		list_add_tail(&q->list, &pending->list);
1123 		switch ((unsigned long) info) {
1124 		case (unsigned long) SEND_SIG_NOINFO:
1125 			clear_siginfo(&q->info);
1126 			q->info.si_signo = sig;
1127 			q->info.si_errno = 0;
1128 			q->info.si_code = SI_USER;
1129 			q->info.si_pid = task_tgid_nr_ns(current,
1130 							task_active_pid_ns(t));
1131 			rcu_read_lock();
1132 			q->info.si_uid =
1133 				from_kuid_munged(task_cred_xxx(t, user_ns),
1134 						 current_uid());
1135 			rcu_read_unlock();
1136 			break;
1137 		case (unsigned long) SEND_SIG_PRIV:
1138 			clear_siginfo(&q->info);
1139 			q->info.si_signo = sig;
1140 			q->info.si_errno = 0;
1141 			q->info.si_code = SI_KERNEL;
1142 			q->info.si_pid = 0;
1143 			q->info.si_uid = 0;
1144 			break;
1145 		default:
1146 			copy_siginfo(&q->info, info);
1147 			break;
1148 		}
1149 	} else if (!is_si_special(info) &&
1150 		   sig >= SIGRTMIN && info->si_code != SI_USER) {
1151 		/*
1152 		 * Queue overflow, abort.  We may abort if the
1153 		 * signal was rt and sent by user using something
1154 		 * other than kill().
1155 		 */
1156 		result = TRACE_SIGNAL_OVERFLOW_FAIL;
1157 		ret = -EAGAIN;
1158 		goto ret;
1159 	} else {
1160 		/*
1161 		 * This is a silent loss of information.  We still
1162 		 * send the signal, but the *info bits are lost.
1163 		 */
1164 		result = TRACE_SIGNAL_LOSE_INFO;
1165 	}
1166 
1167 out_set:
1168 	signalfd_notify(t, sig);
1169 	sigaddset(&pending->signal, sig);
1170 
1171 	/* Let multiprocess signals appear after on-going forks */
1172 	if (type > PIDTYPE_TGID) {
1173 		struct multiprocess_signals *delayed;
1174 		hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1175 			sigset_t *signal = &delayed->signal;
1176 			/* Can't queue both a stop and a continue signal */
1177 			if (sig == SIGCONT)
1178 				sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1179 			else if (sig_kernel_stop(sig))
1180 				sigdelset(signal, SIGCONT);
1181 			sigaddset(signal, sig);
1182 		}
1183 	}
1184 
1185 	complete_signal(sig, t, type);
1186 ret:
1187 	trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1188 	return ret;
1189 }
1190 
1191 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1192 {
1193 	bool ret = false;
1194 	switch (siginfo_layout(info->si_signo, info->si_code)) {
1195 	case SIL_KILL:
1196 	case SIL_CHLD:
1197 	case SIL_RT:
1198 		ret = true;
1199 		break;
1200 	case SIL_TIMER:
1201 	case SIL_POLL:
1202 	case SIL_FAULT:
1203 	case SIL_FAULT_TRAPNO:
1204 	case SIL_FAULT_MCEERR:
1205 	case SIL_FAULT_BNDERR:
1206 	case SIL_FAULT_PKUERR:
1207 	case SIL_FAULT_PERF_EVENT:
1208 	case SIL_SYS:
1209 		ret = false;
1210 		break;
1211 	}
1212 	return ret;
1213 }
1214 
1215 static int send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1216 			enum pid_type type)
1217 {
1218 	/* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1219 	bool force = false;
1220 
1221 	if (info == SEND_SIG_NOINFO) {
1222 		/* Force if sent from an ancestor pid namespace */
1223 		force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1224 	} else if (info == SEND_SIG_PRIV) {
1225 		/* Don't ignore kernel generated signals */
1226 		force = true;
1227 	} else if (has_si_pid_and_uid(info)) {
1228 		/* SIGKILL and SIGSTOP is special or has ids */
1229 		struct user_namespace *t_user_ns;
1230 
1231 		rcu_read_lock();
1232 		t_user_ns = task_cred_xxx(t, user_ns);
1233 		if (current_user_ns() != t_user_ns) {
1234 			kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1235 			info->si_uid = from_kuid_munged(t_user_ns, uid);
1236 		}
1237 		rcu_read_unlock();
1238 
1239 		/* A kernel generated signal? */
1240 		force = (info->si_code == SI_KERNEL);
1241 
1242 		/* From an ancestor pid namespace? */
1243 		if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1244 			info->si_pid = 0;
1245 			force = true;
1246 		}
1247 	}
1248 	return __send_signal(sig, info, t, type, force);
1249 }
1250 
1251 static void print_fatal_signal(int signr)
1252 {
1253 	struct pt_regs *regs = signal_pt_regs();
1254 	pr_info("potentially unexpected fatal signal %d.\n", signr);
1255 
1256 #if defined(__i386__) && !defined(__arch_um__)
1257 	pr_info("code at %08lx: ", regs->ip);
1258 	{
1259 		int i;
1260 		for (i = 0; i < 16; i++) {
1261 			unsigned char insn;
1262 
1263 			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1264 				break;
1265 			pr_cont("%02x ", insn);
1266 		}
1267 	}
1268 	pr_cont("\n");
1269 #endif
1270 	preempt_disable();
1271 	show_regs(regs);
1272 	preempt_enable();
1273 }
1274 
1275 static int __init setup_print_fatal_signals(char *str)
1276 {
1277 	get_option (&str, &print_fatal_signals);
1278 
1279 	return 1;
1280 }
1281 
1282 __setup("print-fatal-signals=", setup_print_fatal_signals);
1283 
1284 int
1285 __group_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1286 {
1287 	return send_signal(sig, info, p, PIDTYPE_TGID);
1288 }
1289 
1290 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1291 			enum pid_type type)
1292 {
1293 	unsigned long flags;
1294 	int ret = -ESRCH;
1295 
1296 	if (lock_task_sighand(p, &flags)) {
1297 		ret = send_signal(sig, info, p, type);
1298 		unlock_task_sighand(p, &flags);
1299 	}
1300 
1301 	return ret;
1302 }
1303 
1304 enum sig_handler {
1305 	HANDLER_CURRENT, /* If reachable use the current handler */
1306 	HANDLER_SIG_DFL, /* Always use SIG_DFL handler semantics */
1307 	HANDLER_EXIT,	 /* Only visible as the process exit code */
1308 };
1309 
1310 /*
1311  * Force a signal that the process can't ignore: if necessary
1312  * we unblock the signal and change any SIG_IGN to SIG_DFL.
1313  *
1314  * Note: If we unblock the signal, we always reset it to SIG_DFL,
1315  * since we do not want to have a signal handler that was blocked
1316  * be invoked when user space had explicitly blocked it.
1317  *
1318  * We don't want to have recursive SIGSEGV's etc, for example,
1319  * that is why we also clear SIGNAL_UNKILLABLE.
1320  */
1321 static int
1322 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t,
1323 	enum sig_handler handler)
1324 {
1325 	unsigned long int flags;
1326 	int ret, blocked, ignored;
1327 	struct k_sigaction *action;
1328 	int sig = info->si_signo;
1329 
1330 	spin_lock_irqsave(&t->sighand->siglock, flags);
1331 	action = &t->sighand->action[sig-1];
1332 	ignored = action->sa.sa_handler == SIG_IGN;
1333 	blocked = sigismember(&t->blocked, sig);
1334 	if (blocked || ignored || (handler != HANDLER_CURRENT)) {
1335 		action->sa.sa_handler = SIG_DFL;
1336 		if (handler == HANDLER_EXIT)
1337 			action->sa.sa_flags |= SA_IMMUTABLE;
1338 		if (blocked) {
1339 			sigdelset(&t->blocked, sig);
1340 			recalc_sigpending_and_wake(t);
1341 		}
1342 	}
1343 	/*
1344 	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1345 	 * debugging to leave init killable. But HANDLER_EXIT is always fatal.
1346 	 */
1347 	if (action->sa.sa_handler == SIG_DFL &&
1348 	    (!t->ptrace || (handler == HANDLER_EXIT)))
1349 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1350 	ret = send_signal(sig, info, t, PIDTYPE_PID);
1351 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1352 
1353 	return ret;
1354 }
1355 
1356 int force_sig_info(struct kernel_siginfo *info)
1357 {
1358 	return force_sig_info_to_task(info, current, HANDLER_CURRENT);
1359 }
1360 
1361 /*
1362  * Nuke all other threads in the group.
1363  */
1364 int zap_other_threads(struct task_struct *p)
1365 {
1366 	struct task_struct *t = p;
1367 	int count = 0;
1368 
1369 	p->signal->group_stop_count = 0;
1370 
1371 	while_each_thread(p, t) {
1372 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1373 		count++;
1374 
1375 		/* Don't bother with already dead threads */
1376 		if (t->exit_state)
1377 			continue;
1378 		sigaddset(&t->pending.signal, SIGKILL);
1379 		signal_wake_up(t, 1);
1380 	}
1381 
1382 	return count;
1383 }
1384 
1385 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1386 					   unsigned long *flags)
1387 {
1388 	struct sighand_struct *sighand;
1389 
1390 	rcu_read_lock();
1391 	for (;;) {
1392 		sighand = rcu_dereference(tsk->sighand);
1393 		if (unlikely(sighand == NULL))
1394 			break;
1395 
1396 		/*
1397 		 * This sighand can be already freed and even reused, but
1398 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1399 		 * initializes ->siglock: this slab can't go away, it has
1400 		 * the same object type, ->siglock can't be reinitialized.
1401 		 *
1402 		 * We need to ensure that tsk->sighand is still the same
1403 		 * after we take the lock, we can race with de_thread() or
1404 		 * __exit_signal(). In the latter case the next iteration
1405 		 * must see ->sighand == NULL.
1406 		 */
1407 		spin_lock_irqsave(&sighand->siglock, *flags);
1408 		if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1409 			break;
1410 		spin_unlock_irqrestore(&sighand->siglock, *flags);
1411 	}
1412 	rcu_read_unlock();
1413 
1414 	return sighand;
1415 }
1416 
1417 #ifdef CONFIG_LOCKDEP
1418 void lockdep_assert_task_sighand_held(struct task_struct *task)
1419 {
1420 	struct sighand_struct *sighand;
1421 
1422 	rcu_read_lock();
1423 	sighand = rcu_dereference(task->sighand);
1424 	if (sighand)
1425 		lockdep_assert_held(&sighand->siglock);
1426 	else
1427 		WARN_ON_ONCE(1);
1428 	rcu_read_unlock();
1429 }
1430 #endif
1431 
1432 /*
1433  * send signal info to all the members of a group
1434  */
1435 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1436 			struct task_struct *p, enum pid_type type)
1437 {
1438 	int ret;
1439 
1440 	rcu_read_lock();
1441 	ret = check_kill_permission(sig, info, p);
1442 	rcu_read_unlock();
1443 
1444 	if (!ret && sig)
1445 		ret = do_send_sig_info(sig, info, p, type);
1446 
1447 	return ret;
1448 }
1449 
1450 /*
1451  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1452  * control characters do (^C, ^Z etc)
1453  * - the caller must hold at least a readlock on tasklist_lock
1454  */
1455 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1456 {
1457 	struct task_struct *p = NULL;
1458 	int retval, success;
1459 
1460 	success = 0;
1461 	retval = -ESRCH;
1462 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1463 		int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1464 		success |= !err;
1465 		retval = err;
1466 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1467 	return success ? 0 : retval;
1468 }
1469 
1470 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1471 {
1472 	int error = -ESRCH;
1473 	struct task_struct *p;
1474 
1475 	for (;;) {
1476 		rcu_read_lock();
1477 		p = pid_task(pid, PIDTYPE_PID);
1478 		if (p)
1479 			error = group_send_sig_info(sig, info, p, PIDTYPE_TGID);
1480 		rcu_read_unlock();
1481 		if (likely(!p || error != -ESRCH))
1482 			return error;
1483 
1484 		/*
1485 		 * The task was unhashed in between, try again.  If it
1486 		 * is dead, pid_task() will return NULL, if we race with
1487 		 * de_thread() it will find the new leader.
1488 		 */
1489 	}
1490 }
1491 
1492 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1493 {
1494 	int error;
1495 	rcu_read_lock();
1496 	error = kill_pid_info(sig, info, find_vpid(pid));
1497 	rcu_read_unlock();
1498 	return error;
1499 }
1500 
1501 static inline bool kill_as_cred_perm(const struct cred *cred,
1502 				     struct task_struct *target)
1503 {
1504 	const struct cred *pcred = __task_cred(target);
1505 
1506 	return uid_eq(cred->euid, pcred->suid) ||
1507 	       uid_eq(cred->euid, pcred->uid) ||
1508 	       uid_eq(cred->uid, pcred->suid) ||
1509 	       uid_eq(cred->uid, pcred->uid);
1510 }
1511 
1512 /*
1513  * The usb asyncio usage of siginfo is wrong.  The glibc support
1514  * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1515  * AKA after the generic fields:
1516  *	kernel_pid_t	si_pid;
1517  *	kernel_uid32_t	si_uid;
1518  *	sigval_t	si_value;
1519  *
1520  * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1521  * after the generic fields is:
1522  *	void __user 	*si_addr;
1523  *
1524  * This is a practical problem when there is a 64bit big endian kernel
1525  * and a 32bit userspace.  As the 32bit address will encoded in the low
1526  * 32bits of the pointer.  Those low 32bits will be stored at higher
1527  * address than appear in a 32 bit pointer.  So userspace will not
1528  * see the address it was expecting for it's completions.
1529  *
1530  * There is nothing in the encoding that can allow
1531  * copy_siginfo_to_user32 to detect this confusion of formats, so
1532  * handle this by requiring the caller of kill_pid_usb_asyncio to
1533  * notice when this situration takes place and to store the 32bit
1534  * pointer in sival_int, instead of sival_addr of the sigval_t addr
1535  * parameter.
1536  */
1537 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1538 			 struct pid *pid, const struct cred *cred)
1539 {
1540 	struct kernel_siginfo info;
1541 	struct task_struct *p;
1542 	unsigned long flags;
1543 	int ret = -EINVAL;
1544 
1545 	if (!valid_signal(sig))
1546 		return ret;
1547 
1548 	clear_siginfo(&info);
1549 	info.si_signo = sig;
1550 	info.si_errno = errno;
1551 	info.si_code = SI_ASYNCIO;
1552 	*((sigval_t *)&info.si_pid) = addr;
1553 
1554 	rcu_read_lock();
1555 	p = pid_task(pid, PIDTYPE_PID);
1556 	if (!p) {
1557 		ret = -ESRCH;
1558 		goto out_unlock;
1559 	}
1560 	if (!kill_as_cred_perm(cred, p)) {
1561 		ret = -EPERM;
1562 		goto out_unlock;
1563 	}
1564 	ret = security_task_kill(p, &info, sig, cred);
1565 	if (ret)
1566 		goto out_unlock;
1567 
1568 	if (sig) {
1569 		if (lock_task_sighand(p, &flags)) {
1570 			ret = __send_signal(sig, &info, p, PIDTYPE_TGID, false);
1571 			unlock_task_sighand(p, &flags);
1572 		} else
1573 			ret = -ESRCH;
1574 	}
1575 out_unlock:
1576 	rcu_read_unlock();
1577 	return ret;
1578 }
1579 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1580 
1581 /*
1582  * kill_something_info() interprets pid in interesting ways just like kill(2).
1583  *
1584  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1585  * is probably wrong.  Should make it like BSD or SYSV.
1586  */
1587 
1588 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1589 {
1590 	int ret;
1591 
1592 	if (pid > 0)
1593 		return kill_proc_info(sig, info, pid);
1594 
1595 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1596 	if (pid == INT_MIN)
1597 		return -ESRCH;
1598 
1599 	read_lock(&tasklist_lock);
1600 	if (pid != -1) {
1601 		ret = __kill_pgrp_info(sig, info,
1602 				pid ? find_vpid(-pid) : task_pgrp(current));
1603 	} else {
1604 		int retval = 0, count = 0;
1605 		struct task_struct * p;
1606 
1607 		for_each_process(p) {
1608 			if (task_pid_vnr(p) > 1 &&
1609 					!same_thread_group(p, current)) {
1610 				int err = group_send_sig_info(sig, info, p,
1611 							      PIDTYPE_MAX);
1612 				++count;
1613 				if (err != -EPERM)
1614 					retval = err;
1615 			}
1616 		}
1617 		ret = count ? retval : -ESRCH;
1618 	}
1619 	read_unlock(&tasklist_lock);
1620 
1621 	return ret;
1622 }
1623 
1624 /*
1625  * These are for backward compatibility with the rest of the kernel source.
1626  */
1627 
1628 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1629 {
1630 	/*
1631 	 * Make sure legacy kernel users don't send in bad values
1632 	 * (normal paths check this in check_kill_permission).
1633 	 */
1634 	if (!valid_signal(sig))
1635 		return -EINVAL;
1636 
1637 	return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1638 }
1639 EXPORT_SYMBOL(send_sig_info);
1640 
1641 #define __si_special(priv) \
1642 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1643 
1644 int
1645 send_sig(int sig, struct task_struct *p, int priv)
1646 {
1647 	return send_sig_info(sig, __si_special(priv), p);
1648 }
1649 EXPORT_SYMBOL(send_sig);
1650 
1651 void force_sig(int sig)
1652 {
1653 	struct kernel_siginfo info;
1654 
1655 	clear_siginfo(&info);
1656 	info.si_signo = sig;
1657 	info.si_errno = 0;
1658 	info.si_code = SI_KERNEL;
1659 	info.si_pid = 0;
1660 	info.si_uid = 0;
1661 	force_sig_info(&info);
1662 }
1663 EXPORT_SYMBOL(force_sig);
1664 
1665 void force_fatal_sig(int sig)
1666 {
1667 	struct kernel_siginfo info;
1668 
1669 	clear_siginfo(&info);
1670 	info.si_signo = sig;
1671 	info.si_errno = 0;
1672 	info.si_code = SI_KERNEL;
1673 	info.si_pid = 0;
1674 	info.si_uid = 0;
1675 	force_sig_info_to_task(&info, current, HANDLER_SIG_DFL);
1676 }
1677 
1678 void force_exit_sig(int sig)
1679 {
1680 	struct kernel_siginfo info;
1681 
1682 	clear_siginfo(&info);
1683 	info.si_signo = sig;
1684 	info.si_errno = 0;
1685 	info.si_code = SI_KERNEL;
1686 	info.si_pid = 0;
1687 	info.si_uid = 0;
1688 	force_sig_info_to_task(&info, current, HANDLER_EXIT);
1689 }
1690 
1691 /*
1692  * When things go south during signal handling, we
1693  * will force a SIGSEGV. And if the signal that caused
1694  * the problem was already a SIGSEGV, we'll want to
1695  * make sure we don't even try to deliver the signal..
1696  */
1697 void force_sigsegv(int sig)
1698 {
1699 	if (sig == SIGSEGV)
1700 		force_fatal_sig(SIGSEGV);
1701 	else
1702 		force_sig(SIGSEGV);
1703 }
1704 
1705 int force_sig_fault_to_task(int sig, int code, void __user *addr
1706 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1707 	, struct task_struct *t)
1708 {
1709 	struct kernel_siginfo info;
1710 
1711 	clear_siginfo(&info);
1712 	info.si_signo = sig;
1713 	info.si_errno = 0;
1714 	info.si_code  = code;
1715 	info.si_addr  = addr;
1716 #ifdef __ia64__
1717 	info.si_imm = imm;
1718 	info.si_flags = flags;
1719 	info.si_isr = isr;
1720 #endif
1721 	return force_sig_info_to_task(&info, t, HANDLER_CURRENT);
1722 }
1723 
1724 int force_sig_fault(int sig, int code, void __user *addr
1725 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr))
1726 {
1727 	return force_sig_fault_to_task(sig, code, addr
1728 				       ___ARCH_SI_IA64(imm, flags, isr), current);
1729 }
1730 
1731 int send_sig_fault(int sig, int code, void __user *addr
1732 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1733 	, struct task_struct *t)
1734 {
1735 	struct kernel_siginfo info;
1736 
1737 	clear_siginfo(&info);
1738 	info.si_signo = sig;
1739 	info.si_errno = 0;
1740 	info.si_code  = code;
1741 	info.si_addr  = addr;
1742 #ifdef __ia64__
1743 	info.si_imm = imm;
1744 	info.si_flags = flags;
1745 	info.si_isr = isr;
1746 #endif
1747 	return send_sig_info(info.si_signo, &info, t);
1748 }
1749 
1750 int force_sig_mceerr(int code, void __user *addr, short lsb)
1751 {
1752 	struct kernel_siginfo info;
1753 
1754 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1755 	clear_siginfo(&info);
1756 	info.si_signo = SIGBUS;
1757 	info.si_errno = 0;
1758 	info.si_code = code;
1759 	info.si_addr = addr;
1760 	info.si_addr_lsb = lsb;
1761 	return force_sig_info(&info);
1762 }
1763 
1764 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1765 {
1766 	struct kernel_siginfo info;
1767 
1768 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1769 	clear_siginfo(&info);
1770 	info.si_signo = SIGBUS;
1771 	info.si_errno = 0;
1772 	info.si_code = code;
1773 	info.si_addr = addr;
1774 	info.si_addr_lsb = lsb;
1775 	return send_sig_info(info.si_signo, &info, t);
1776 }
1777 EXPORT_SYMBOL(send_sig_mceerr);
1778 
1779 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1780 {
1781 	struct kernel_siginfo info;
1782 
1783 	clear_siginfo(&info);
1784 	info.si_signo = SIGSEGV;
1785 	info.si_errno = 0;
1786 	info.si_code  = SEGV_BNDERR;
1787 	info.si_addr  = addr;
1788 	info.si_lower = lower;
1789 	info.si_upper = upper;
1790 	return force_sig_info(&info);
1791 }
1792 
1793 #ifdef SEGV_PKUERR
1794 int force_sig_pkuerr(void __user *addr, u32 pkey)
1795 {
1796 	struct kernel_siginfo info;
1797 
1798 	clear_siginfo(&info);
1799 	info.si_signo = SIGSEGV;
1800 	info.si_errno = 0;
1801 	info.si_code  = SEGV_PKUERR;
1802 	info.si_addr  = addr;
1803 	info.si_pkey  = pkey;
1804 	return force_sig_info(&info);
1805 }
1806 #endif
1807 
1808 int force_sig_perf(void __user *addr, u32 type, u64 sig_data)
1809 {
1810 	struct kernel_siginfo info;
1811 
1812 	clear_siginfo(&info);
1813 	info.si_signo     = SIGTRAP;
1814 	info.si_errno     = 0;
1815 	info.si_code      = TRAP_PERF;
1816 	info.si_addr      = addr;
1817 	info.si_perf_data = sig_data;
1818 	info.si_perf_type = type;
1819 
1820 	return force_sig_info(&info);
1821 }
1822 
1823 /**
1824  * force_sig_seccomp - signals the task to allow in-process syscall emulation
1825  * @syscall: syscall number to send to userland
1826  * @reason: filter-supplied reason code to send to userland (via si_errno)
1827  * @force_coredump: true to trigger a coredump
1828  *
1829  * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
1830  */
1831 int force_sig_seccomp(int syscall, int reason, bool force_coredump)
1832 {
1833 	struct kernel_siginfo info;
1834 
1835 	clear_siginfo(&info);
1836 	info.si_signo = SIGSYS;
1837 	info.si_code = SYS_SECCOMP;
1838 	info.si_call_addr = (void __user *)KSTK_EIP(current);
1839 	info.si_errno = reason;
1840 	info.si_arch = syscall_get_arch(current);
1841 	info.si_syscall = syscall;
1842 	return force_sig_info_to_task(&info, current,
1843 		force_coredump ? HANDLER_EXIT : HANDLER_CURRENT);
1844 }
1845 
1846 /* For the crazy architectures that include trap information in
1847  * the errno field, instead of an actual errno value.
1848  */
1849 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1850 {
1851 	struct kernel_siginfo info;
1852 
1853 	clear_siginfo(&info);
1854 	info.si_signo = SIGTRAP;
1855 	info.si_errno = errno;
1856 	info.si_code  = TRAP_HWBKPT;
1857 	info.si_addr  = addr;
1858 	return force_sig_info(&info);
1859 }
1860 
1861 /* For the rare architectures that include trap information using
1862  * si_trapno.
1863  */
1864 int force_sig_fault_trapno(int sig, int code, void __user *addr, int trapno)
1865 {
1866 	struct kernel_siginfo info;
1867 
1868 	clear_siginfo(&info);
1869 	info.si_signo = sig;
1870 	info.si_errno = 0;
1871 	info.si_code  = code;
1872 	info.si_addr  = addr;
1873 	info.si_trapno = trapno;
1874 	return force_sig_info(&info);
1875 }
1876 
1877 /* For the rare architectures that include trap information using
1878  * si_trapno.
1879  */
1880 int send_sig_fault_trapno(int sig, int code, void __user *addr, int trapno,
1881 			  struct task_struct *t)
1882 {
1883 	struct kernel_siginfo info;
1884 
1885 	clear_siginfo(&info);
1886 	info.si_signo = sig;
1887 	info.si_errno = 0;
1888 	info.si_code  = code;
1889 	info.si_addr  = addr;
1890 	info.si_trapno = trapno;
1891 	return send_sig_info(info.si_signo, &info, t);
1892 }
1893 
1894 int kill_pgrp(struct pid *pid, int sig, int priv)
1895 {
1896 	int ret;
1897 
1898 	read_lock(&tasklist_lock);
1899 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1900 	read_unlock(&tasklist_lock);
1901 
1902 	return ret;
1903 }
1904 EXPORT_SYMBOL(kill_pgrp);
1905 
1906 int kill_pid(struct pid *pid, int sig, int priv)
1907 {
1908 	return kill_pid_info(sig, __si_special(priv), pid);
1909 }
1910 EXPORT_SYMBOL(kill_pid);
1911 
1912 /*
1913  * These functions support sending signals using preallocated sigqueue
1914  * structures.  This is needed "because realtime applications cannot
1915  * afford to lose notifications of asynchronous events, like timer
1916  * expirations or I/O completions".  In the case of POSIX Timers
1917  * we allocate the sigqueue structure from the timer_create.  If this
1918  * allocation fails we are able to report the failure to the application
1919  * with an EAGAIN error.
1920  */
1921 struct sigqueue *sigqueue_alloc(void)
1922 {
1923 	return __sigqueue_alloc(-1, current, GFP_KERNEL, 0, SIGQUEUE_PREALLOC);
1924 }
1925 
1926 void sigqueue_free(struct sigqueue *q)
1927 {
1928 	unsigned long flags;
1929 	spinlock_t *lock = &current->sighand->siglock;
1930 
1931 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1932 	/*
1933 	 * We must hold ->siglock while testing q->list
1934 	 * to serialize with collect_signal() or with
1935 	 * __exit_signal()->flush_sigqueue().
1936 	 */
1937 	spin_lock_irqsave(lock, flags);
1938 	q->flags &= ~SIGQUEUE_PREALLOC;
1939 	/*
1940 	 * If it is queued it will be freed when dequeued,
1941 	 * like the "regular" sigqueue.
1942 	 */
1943 	if (!list_empty(&q->list))
1944 		q = NULL;
1945 	spin_unlock_irqrestore(lock, flags);
1946 
1947 	if (q)
1948 		__sigqueue_free(q);
1949 }
1950 
1951 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1952 {
1953 	int sig = q->info.si_signo;
1954 	struct sigpending *pending;
1955 	struct task_struct *t;
1956 	unsigned long flags;
1957 	int ret, result;
1958 
1959 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1960 
1961 	ret = -1;
1962 	rcu_read_lock();
1963 	t = pid_task(pid, type);
1964 	if (!t || !likely(lock_task_sighand(t, &flags)))
1965 		goto ret;
1966 
1967 	ret = 1; /* the signal is ignored */
1968 	result = TRACE_SIGNAL_IGNORED;
1969 	if (!prepare_signal(sig, t, false))
1970 		goto out;
1971 
1972 	ret = 0;
1973 	if (unlikely(!list_empty(&q->list))) {
1974 		/*
1975 		 * If an SI_TIMER entry is already queue just increment
1976 		 * the overrun count.
1977 		 */
1978 		BUG_ON(q->info.si_code != SI_TIMER);
1979 		q->info.si_overrun++;
1980 		result = TRACE_SIGNAL_ALREADY_PENDING;
1981 		goto out;
1982 	}
1983 	q->info.si_overrun = 0;
1984 
1985 	signalfd_notify(t, sig);
1986 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1987 	list_add_tail(&q->list, &pending->list);
1988 	sigaddset(&pending->signal, sig);
1989 	complete_signal(sig, t, type);
1990 	result = TRACE_SIGNAL_DELIVERED;
1991 out:
1992 	trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
1993 	unlock_task_sighand(t, &flags);
1994 ret:
1995 	rcu_read_unlock();
1996 	return ret;
1997 }
1998 
1999 static void do_notify_pidfd(struct task_struct *task)
2000 {
2001 	struct pid *pid;
2002 
2003 	WARN_ON(task->exit_state == 0);
2004 	pid = task_pid(task);
2005 	wake_up_all(&pid->wait_pidfd);
2006 }
2007 
2008 /*
2009  * Let a parent know about the death of a child.
2010  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
2011  *
2012  * Returns true if our parent ignored us and so we've switched to
2013  * self-reaping.
2014  */
2015 bool do_notify_parent(struct task_struct *tsk, int sig)
2016 {
2017 	struct kernel_siginfo info;
2018 	unsigned long flags;
2019 	struct sighand_struct *psig;
2020 	bool autoreap = false;
2021 	u64 utime, stime;
2022 
2023 	BUG_ON(sig == -1);
2024 
2025  	/* do_notify_parent_cldstop should have been called instead.  */
2026  	BUG_ON(task_is_stopped_or_traced(tsk));
2027 
2028 	BUG_ON(!tsk->ptrace &&
2029 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
2030 
2031 	/* Wake up all pidfd waiters */
2032 	do_notify_pidfd(tsk);
2033 
2034 	if (sig != SIGCHLD) {
2035 		/*
2036 		 * This is only possible if parent == real_parent.
2037 		 * Check if it has changed security domain.
2038 		 */
2039 		if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
2040 			sig = SIGCHLD;
2041 	}
2042 
2043 	clear_siginfo(&info);
2044 	info.si_signo = sig;
2045 	info.si_errno = 0;
2046 	/*
2047 	 * We are under tasklist_lock here so our parent is tied to
2048 	 * us and cannot change.
2049 	 *
2050 	 * task_active_pid_ns will always return the same pid namespace
2051 	 * until a task passes through release_task.
2052 	 *
2053 	 * write_lock() currently calls preempt_disable() which is the
2054 	 * same as rcu_read_lock(), but according to Oleg, this is not
2055 	 * correct to rely on this
2056 	 */
2057 	rcu_read_lock();
2058 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
2059 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
2060 				       task_uid(tsk));
2061 	rcu_read_unlock();
2062 
2063 	task_cputime(tsk, &utime, &stime);
2064 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
2065 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
2066 
2067 	info.si_status = tsk->exit_code & 0x7f;
2068 	if (tsk->exit_code & 0x80)
2069 		info.si_code = CLD_DUMPED;
2070 	else if (tsk->exit_code & 0x7f)
2071 		info.si_code = CLD_KILLED;
2072 	else {
2073 		info.si_code = CLD_EXITED;
2074 		info.si_status = tsk->exit_code >> 8;
2075 	}
2076 
2077 	psig = tsk->parent->sighand;
2078 	spin_lock_irqsave(&psig->siglock, flags);
2079 	if (!tsk->ptrace && sig == SIGCHLD &&
2080 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
2081 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
2082 		/*
2083 		 * We are exiting and our parent doesn't care.  POSIX.1
2084 		 * defines special semantics for setting SIGCHLD to SIG_IGN
2085 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
2086 		 * automatically and not left for our parent's wait4 call.
2087 		 * Rather than having the parent do it as a magic kind of
2088 		 * signal handler, we just set this to tell do_exit that we
2089 		 * can be cleaned up without becoming a zombie.  Note that
2090 		 * we still call __wake_up_parent in this case, because a
2091 		 * blocked sys_wait4 might now return -ECHILD.
2092 		 *
2093 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
2094 		 * is implementation-defined: we do (if you don't want
2095 		 * it, just use SIG_IGN instead).
2096 		 */
2097 		autoreap = true;
2098 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2099 			sig = 0;
2100 	}
2101 	/*
2102 	 * Send with __send_signal as si_pid and si_uid are in the
2103 	 * parent's namespaces.
2104 	 */
2105 	if (valid_signal(sig) && sig)
2106 		__send_signal(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2107 	__wake_up_parent(tsk, tsk->parent);
2108 	spin_unlock_irqrestore(&psig->siglock, flags);
2109 
2110 	return autoreap;
2111 }
2112 
2113 /**
2114  * do_notify_parent_cldstop - notify parent of stopped/continued state change
2115  * @tsk: task reporting the state change
2116  * @for_ptracer: the notification is for ptracer
2117  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2118  *
2119  * Notify @tsk's parent that the stopped/continued state has changed.  If
2120  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2121  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2122  *
2123  * CONTEXT:
2124  * Must be called with tasklist_lock at least read locked.
2125  */
2126 static void do_notify_parent_cldstop(struct task_struct *tsk,
2127 				     bool for_ptracer, int why)
2128 {
2129 	struct kernel_siginfo info;
2130 	unsigned long flags;
2131 	struct task_struct *parent;
2132 	struct sighand_struct *sighand;
2133 	u64 utime, stime;
2134 
2135 	if (for_ptracer) {
2136 		parent = tsk->parent;
2137 	} else {
2138 		tsk = tsk->group_leader;
2139 		parent = tsk->real_parent;
2140 	}
2141 
2142 	clear_siginfo(&info);
2143 	info.si_signo = SIGCHLD;
2144 	info.si_errno = 0;
2145 	/*
2146 	 * see comment in do_notify_parent() about the following 4 lines
2147 	 */
2148 	rcu_read_lock();
2149 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2150 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2151 	rcu_read_unlock();
2152 
2153 	task_cputime(tsk, &utime, &stime);
2154 	info.si_utime = nsec_to_clock_t(utime);
2155 	info.si_stime = nsec_to_clock_t(stime);
2156 
2157  	info.si_code = why;
2158  	switch (why) {
2159  	case CLD_CONTINUED:
2160  		info.si_status = SIGCONT;
2161  		break;
2162  	case CLD_STOPPED:
2163  		info.si_status = tsk->signal->group_exit_code & 0x7f;
2164  		break;
2165  	case CLD_TRAPPED:
2166  		info.si_status = tsk->exit_code & 0x7f;
2167  		break;
2168  	default:
2169  		BUG();
2170  	}
2171 
2172 	sighand = parent->sighand;
2173 	spin_lock_irqsave(&sighand->siglock, flags);
2174 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2175 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2176 		__group_send_sig_info(SIGCHLD, &info, parent);
2177 	/*
2178 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2179 	 */
2180 	__wake_up_parent(tsk, parent);
2181 	spin_unlock_irqrestore(&sighand->siglock, flags);
2182 }
2183 
2184 /*
2185  * This must be called with current->sighand->siglock held.
2186  *
2187  * This should be the path for all ptrace stops.
2188  * We always set current->last_siginfo while stopped here.
2189  * That makes it a way to test a stopped process for
2190  * being ptrace-stopped vs being job-control-stopped.
2191  *
2192  * Returns the signal the ptracer requested the code resume
2193  * with.  If the code did not stop because the tracer is gone,
2194  * the stop signal remains unchanged unless clear_code.
2195  */
2196 static int ptrace_stop(int exit_code, int why, int clear_code,
2197 			unsigned long message, kernel_siginfo_t *info)
2198 	__releases(&current->sighand->siglock)
2199 	__acquires(&current->sighand->siglock)
2200 {
2201 	bool gstop_done = false;
2202 	bool read_code = true;
2203 
2204 	if (arch_ptrace_stop_needed()) {
2205 		/*
2206 		 * The arch code has something special to do before a
2207 		 * ptrace stop.  This is allowed to block, e.g. for faults
2208 		 * on user stack pages.  We can't keep the siglock while
2209 		 * calling arch_ptrace_stop, so we must release it now.
2210 		 * To preserve proper semantics, we must do this before
2211 		 * any signal bookkeeping like checking group_stop_count.
2212 		 */
2213 		spin_unlock_irq(&current->sighand->siglock);
2214 		arch_ptrace_stop();
2215 		spin_lock_irq(&current->sighand->siglock);
2216 	}
2217 
2218 	/*
2219 	 * schedule() will not sleep if there is a pending signal that
2220 	 * can awaken the task.
2221 	 */
2222 	set_special_state(TASK_TRACED);
2223 
2224 	/*
2225 	 * We're committing to trapping.  TRACED should be visible before
2226 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2227 	 * Also, transition to TRACED and updates to ->jobctl should be
2228 	 * atomic with respect to siglock and should be done after the arch
2229 	 * hook as siglock is released and regrabbed across it.
2230 	 *
2231 	 *     TRACER				    TRACEE
2232 	 *
2233 	 *     ptrace_attach()
2234 	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2235 	 *     do_wait()
2236 	 *       set_current_state()                smp_wmb();
2237 	 *       ptrace_do_wait()
2238 	 *         wait_task_stopped()
2239 	 *           task_stopped_code()
2240 	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2241 	 */
2242 	smp_wmb();
2243 
2244 	current->ptrace_message = message;
2245 	current->last_siginfo = info;
2246 	current->exit_code = exit_code;
2247 
2248 	/*
2249 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2250 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2251 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2252 	 * could be clear now.  We act as if SIGCONT is received after
2253 	 * TASK_TRACED is entered - ignore it.
2254 	 */
2255 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2256 		gstop_done = task_participate_group_stop(current);
2257 
2258 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2259 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2260 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2261 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2262 
2263 	/* entering a trap, clear TRAPPING */
2264 	task_clear_jobctl_trapping(current);
2265 
2266 	spin_unlock_irq(&current->sighand->siglock);
2267 	read_lock(&tasklist_lock);
2268 	if (likely(current->ptrace)) {
2269 		/*
2270 		 * Notify parents of the stop.
2271 		 *
2272 		 * While ptraced, there are two parents - the ptracer and
2273 		 * the real_parent of the group_leader.  The ptracer should
2274 		 * know about every stop while the real parent is only
2275 		 * interested in the completion of group stop.  The states
2276 		 * for the two don't interact with each other.  Notify
2277 		 * separately unless they're gonna be duplicates.
2278 		 */
2279 		do_notify_parent_cldstop(current, true, why);
2280 		if (gstop_done && ptrace_reparented(current))
2281 			do_notify_parent_cldstop(current, false, why);
2282 
2283 		/*
2284 		 * Don't want to allow preemption here, because
2285 		 * sys_ptrace() needs this task to be inactive.
2286 		 *
2287 		 * XXX: implement read_unlock_no_resched().
2288 		 */
2289 		preempt_disable();
2290 		read_unlock(&tasklist_lock);
2291 		cgroup_enter_frozen();
2292 		preempt_enable_no_resched();
2293 		freezable_schedule();
2294 		cgroup_leave_frozen(true);
2295 	} else {
2296 		/*
2297 		 * By the time we got the lock, our tracer went away.
2298 		 * Don't drop the lock yet, another tracer may come.
2299 		 *
2300 		 * If @gstop_done, the ptracer went away between group stop
2301 		 * completion and here.  During detach, it would have set
2302 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
2303 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
2304 		 * the real parent of the group stop completion is enough.
2305 		 */
2306 		if (gstop_done)
2307 			do_notify_parent_cldstop(current, false, why);
2308 
2309 		/* tasklist protects us from ptrace_freeze_traced() */
2310 		__set_current_state(TASK_RUNNING);
2311 		read_code = false;
2312 		if (clear_code)
2313 			exit_code = 0;
2314 		read_unlock(&tasklist_lock);
2315 	}
2316 
2317 	/*
2318 	 * We are back.  Now reacquire the siglock before touching
2319 	 * last_siginfo, so that we are sure to have synchronized with
2320 	 * any signal-sending on another CPU that wants to examine it.
2321 	 */
2322 	spin_lock_irq(&current->sighand->siglock);
2323 	if (read_code)
2324 		exit_code = current->exit_code;
2325 	current->last_siginfo = NULL;
2326 	current->ptrace_message = 0;
2327 	current->exit_code = 0;
2328 
2329 	/* LISTENING can be set only during STOP traps, clear it */
2330 	current->jobctl &= ~JOBCTL_LISTENING;
2331 
2332 	/*
2333 	 * Queued signals ignored us while we were stopped for tracing.
2334 	 * So check for any that we should take before resuming user mode.
2335 	 * This sets TIF_SIGPENDING, but never clears it.
2336 	 */
2337 	recalc_sigpending_tsk(current);
2338 	return exit_code;
2339 }
2340 
2341 static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
2342 {
2343 	kernel_siginfo_t info;
2344 
2345 	clear_siginfo(&info);
2346 	info.si_signo = signr;
2347 	info.si_code = exit_code;
2348 	info.si_pid = task_pid_vnr(current);
2349 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2350 
2351 	/* Let the debugger run.  */
2352 	return ptrace_stop(exit_code, why, 1, message, &info);
2353 }
2354 
2355 int ptrace_notify(int exit_code, unsigned long message)
2356 {
2357 	int signr;
2358 
2359 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2360 	if (unlikely(task_work_pending(current)))
2361 		task_work_run();
2362 
2363 	spin_lock_irq(&current->sighand->siglock);
2364 	signr = ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
2365 	spin_unlock_irq(&current->sighand->siglock);
2366 	return signr;
2367 }
2368 
2369 /**
2370  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2371  * @signr: signr causing group stop if initiating
2372  *
2373  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2374  * and participate in it.  If already set, participate in the existing
2375  * group stop.  If participated in a group stop (and thus slept), %true is
2376  * returned with siglock released.
2377  *
2378  * If ptraced, this function doesn't handle stop itself.  Instead,
2379  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2380  * untouched.  The caller must ensure that INTERRUPT trap handling takes
2381  * places afterwards.
2382  *
2383  * CONTEXT:
2384  * Must be called with @current->sighand->siglock held, which is released
2385  * on %true return.
2386  *
2387  * RETURNS:
2388  * %false if group stop is already cancelled or ptrace trap is scheduled.
2389  * %true if participated in group stop.
2390  */
2391 static bool do_signal_stop(int signr)
2392 	__releases(&current->sighand->siglock)
2393 {
2394 	struct signal_struct *sig = current->signal;
2395 
2396 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2397 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2398 		struct task_struct *t;
2399 
2400 		/* signr will be recorded in task->jobctl for retries */
2401 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2402 
2403 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2404 		    unlikely(sig->flags & SIGNAL_GROUP_EXIT) ||
2405 		    unlikely(sig->group_exec_task))
2406 			return false;
2407 		/*
2408 		 * There is no group stop already in progress.  We must
2409 		 * initiate one now.
2410 		 *
2411 		 * While ptraced, a task may be resumed while group stop is
2412 		 * still in effect and then receive a stop signal and
2413 		 * initiate another group stop.  This deviates from the
2414 		 * usual behavior as two consecutive stop signals can't
2415 		 * cause two group stops when !ptraced.  That is why we
2416 		 * also check !task_is_stopped(t) below.
2417 		 *
2418 		 * The condition can be distinguished by testing whether
2419 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2420 		 * group_exit_code in such case.
2421 		 *
2422 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2423 		 * an intervening stop signal is required to cause two
2424 		 * continued events regardless of ptrace.
2425 		 */
2426 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2427 			sig->group_exit_code = signr;
2428 
2429 		sig->group_stop_count = 0;
2430 
2431 		if (task_set_jobctl_pending(current, signr | gstop))
2432 			sig->group_stop_count++;
2433 
2434 		t = current;
2435 		while_each_thread(current, t) {
2436 			/*
2437 			 * Setting state to TASK_STOPPED for a group
2438 			 * stop is always done with the siglock held,
2439 			 * so this check has no races.
2440 			 */
2441 			if (!task_is_stopped(t) &&
2442 			    task_set_jobctl_pending(t, signr | gstop)) {
2443 				sig->group_stop_count++;
2444 				if (likely(!(t->ptrace & PT_SEIZED)))
2445 					signal_wake_up(t, 0);
2446 				else
2447 					ptrace_trap_notify(t);
2448 			}
2449 		}
2450 	}
2451 
2452 	if (likely(!current->ptrace)) {
2453 		int notify = 0;
2454 
2455 		/*
2456 		 * If there are no other threads in the group, or if there
2457 		 * is a group stop in progress and we are the last to stop,
2458 		 * report to the parent.
2459 		 */
2460 		if (task_participate_group_stop(current))
2461 			notify = CLD_STOPPED;
2462 
2463 		set_special_state(TASK_STOPPED);
2464 		spin_unlock_irq(&current->sighand->siglock);
2465 
2466 		/*
2467 		 * Notify the parent of the group stop completion.  Because
2468 		 * we're not holding either the siglock or tasklist_lock
2469 		 * here, ptracer may attach inbetween; however, this is for
2470 		 * group stop and should always be delivered to the real
2471 		 * parent of the group leader.  The new ptracer will get
2472 		 * its notification when this task transitions into
2473 		 * TASK_TRACED.
2474 		 */
2475 		if (notify) {
2476 			read_lock(&tasklist_lock);
2477 			do_notify_parent_cldstop(current, false, notify);
2478 			read_unlock(&tasklist_lock);
2479 		}
2480 
2481 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2482 		cgroup_enter_frozen();
2483 		freezable_schedule();
2484 		return true;
2485 	} else {
2486 		/*
2487 		 * While ptraced, group stop is handled by STOP trap.
2488 		 * Schedule it and let the caller deal with it.
2489 		 */
2490 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2491 		return false;
2492 	}
2493 }
2494 
2495 /**
2496  * do_jobctl_trap - take care of ptrace jobctl traps
2497  *
2498  * When PT_SEIZED, it's used for both group stop and explicit
2499  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2500  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2501  * the stop signal; otherwise, %SIGTRAP.
2502  *
2503  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2504  * number as exit_code and no siginfo.
2505  *
2506  * CONTEXT:
2507  * Must be called with @current->sighand->siglock held, which may be
2508  * released and re-acquired before returning with intervening sleep.
2509  */
2510 static void do_jobctl_trap(void)
2511 {
2512 	struct signal_struct *signal = current->signal;
2513 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2514 
2515 	if (current->ptrace & PT_SEIZED) {
2516 		if (!signal->group_stop_count &&
2517 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2518 			signr = SIGTRAP;
2519 		WARN_ON_ONCE(!signr);
2520 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2521 				 CLD_STOPPED, 0);
2522 	} else {
2523 		WARN_ON_ONCE(!signr);
2524 		ptrace_stop(signr, CLD_STOPPED, 0, 0, NULL);
2525 	}
2526 }
2527 
2528 /**
2529  * do_freezer_trap - handle the freezer jobctl trap
2530  *
2531  * Puts the task into frozen state, if only the task is not about to quit.
2532  * In this case it drops JOBCTL_TRAP_FREEZE.
2533  *
2534  * CONTEXT:
2535  * Must be called with @current->sighand->siglock held,
2536  * which is always released before returning.
2537  */
2538 static void do_freezer_trap(void)
2539 	__releases(&current->sighand->siglock)
2540 {
2541 	/*
2542 	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2543 	 * let's make another loop to give it a chance to be handled.
2544 	 * In any case, we'll return back.
2545 	 */
2546 	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2547 	     JOBCTL_TRAP_FREEZE) {
2548 		spin_unlock_irq(&current->sighand->siglock);
2549 		return;
2550 	}
2551 
2552 	/*
2553 	 * Now we're sure that there is no pending fatal signal and no
2554 	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2555 	 * immediately (if there is a non-fatal signal pending), and
2556 	 * put the task into sleep.
2557 	 */
2558 	__set_current_state(TASK_INTERRUPTIBLE);
2559 	clear_thread_flag(TIF_SIGPENDING);
2560 	spin_unlock_irq(&current->sighand->siglock);
2561 	cgroup_enter_frozen();
2562 	freezable_schedule();
2563 }
2564 
2565 static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
2566 {
2567 	/*
2568 	 * We do not check sig_kernel_stop(signr) but set this marker
2569 	 * unconditionally because we do not know whether debugger will
2570 	 * change signr. This flag has no meaning unless we are going
2571 	 * to stop after return from ptrace_stop(). In this case it will
2572 	 * be checked in do_signal_stop(), we should only stop if it was
2573 	 * not cleared by SIGCONT while we were sleeping. See also the
2574 	 * comment in dequeue_signal().
2575 	 */
2576 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2577 	signr = ptrace_stop(signr, CLD_TRAPPED, 0, 0, info);
2578 
2579 	/* We're back.  Did the debugger cancel the sig?  */
2580 	if (signr == 0)
2581 		return signr;
2582 
2583 	/*
2584 	 * Update the siginfo structure if the signal has
2585 	 * changed.  If the debugger wanted something
2586 	 * specific in the siginfo structure then it should
2587 	 * have updated *info via PTRACE_SETSIGINFO.
2588 	 */
2589 	if (signr != info->si_signo) {
2590 		clear_siginfo(info);
2591 		info->si_signo = signr;
2592 		info->si_errno = 0;
2593 		info->si_code = SI_USER;
2594 		rcu_read_lock();
2595 		info->si_pid = task_pid_vnr(current->parent);
2596 		info->si_uid = from_kuid_munged(current_user_ns(),
2597 						task_uid(current->parent));
2598 		rcu_read_unlock();
2599 	}
2600 
2601 	/* If the (new) signal is now blocked, requeue it.  */
2602 	if (sigismember(&current->blocked, signr) ||
2603 	    fatal_signal_pending(current)) {
2604 		send_signal(signr, info, current, type);
2605 		signr = 0;
2606 	}
2607 
2608 	return signr;
2609 }
2610 
2611 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2612 {
2613 	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2614 	case SIL_FAULT:
2615 	case SIL_FAULT_TRAPNO:
2616 	case SIL_FAULT_MCEERR:
2617 	case SIL_FAULT_BNDERR:
2618 	case SIL_FAULT_PKUERR:
2619 	case SIL_FAULT_PERF_EVENT:
2620 		ksig->info.si_addr = arch_untagged_si_addr(
2621 			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2622 		break;
2623 	case SIL_KILL:
2624 	case SIL_TIMER:
2625 	case SIL_POLL:
2626 	case SIL_CHLD:
2627 	case SIL_RT:
2628 	case SIL_SYS:
2629 		break;
2630 	}
2631 }
2632 
2633 bool get_signal(struct ksignal *ksig)
2634 {
2635 	struct sighand_struct *sighand = current->sighand;
2636 	struct signal_struct *signal = current->signal;
2637 	int signr;
2638 
2639 	clear_notify_signal();
2640 	if (unlikely(task_work_pending(current)))
2641 		task_work_run();
2642 
2643 	if (!task_sigpending(current))
2644 		return false;
2645 
2646 	if (unlikely(uprobe_deny_signal()))
2647 		return false;
2648 
2649 	/*
2650 	 * Do this once, we can't return to user-mode if freezing() == T.
2651 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2652 	 * thus do not need another check after return.
2653 	 */
2654 	try_to_freeze();
2655 
2656 relock:
2657 	spin_lock_irq(&sighand->siglock);
2658 
2659 	/*
2660 	 * Every stopped thread goes here after wakeup. Check to see if
2661 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2662 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2663 	 */
2664 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2665 		int why;
2666 
2667 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2668 			why = CLD_CONTINUED;
2669 		else
2670 			why = CLD_STOPPED;
2671 
2672 		signal->flags &= ~SIGNAL_CLD_MASK;
2673 
2674 		spin_unlock_irq(&sighand->siglock);
2675 
2676 		/*
2677 		 * Notify the parent that we're continuing.  This event is
2678 		 * always per-process and doesn't make whole lot of sense
2679 		 * for ptracers, who shouldn't consume the state via
2680 		 * wait(2) either, but, for backward compatibility, notify
2681 		 * the ptracer of the group leader too unless it's gonna be
2682 		 * a duplicate.
2683 		 */
2684 		read_lock(&tasklist_lock);
2685 		do_notify_parent_cldstop(current, false, why);
2686 
2687 		if (ptrace_reparented(current->group_leader))
2688 			do_notify_parent_cldstop(current->group_leader,
2689 						true, why);
2690 		read_unlock(&tasklist_lock);
2691 
2692 		goto relock;
2693 	}
2694 
2695 	for (;;) {
2696 		struct k_sigaction *ka;
2697 		enum pid_type type;
2698 
2699 		/* Has this task already been marked for death? */
2700 		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
2701 		     signal->group_exec_task) {
2702 			ksig->info.si_signo = signr = SIGKILL;
2703 			sigdelset(&current->pending.signal, SIGKILL);
2704 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2705 				&sighand->action[SIGKILL - 1]);
2706 			recalc_sigpending();
2707 			goto fatal;
2708 		}
2709 
2710 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2711 		    do_signal_stop(0))
2712 			goto relock;
2713 
2714 		if (unlikely(current->jobctl &
2715 			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2716 			if (current->jobctl & JOBCTL_TRAP_MASK) {
2717 				do_jobctl_trap();
2718 				spin_unlock_irq(&sighand->siglock);
2719 			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2720 				do_freezer_trap();
2721 
2722 			goto relock;
2723 		}
2724 
2725 		/*
2726 		 * If the task is leaving the frozen state, let's update
2727 		 * cgroup counters and reset the frozen bit.
2728 		 */
2729 		if (unlikely(cgroup_task_frozen(current))) {
2730 			spin_unlock_irq(&sighand->siglock);
2731 			cgroup_leave_frozen(false);
2732 			goto relock;
2733 		}
2734 
2735 		/*
2736 		 * Signals generated by the execution of an instruction
2737 		 * need to be delivered before any other pending signals
2738 		 * so that the instruction pointer in the signal stack
2739 		 * frame points to the faulting instruction.
2740 		 */
2741 		type = PIDTYPE_PID;
2742 		signr = dequeue_synchronous_signal(&ksig->info);
2743 		if (!signr)
2744 			signr = dequeue_signal(current, &current->blocked,
2745 					       &ksig->info, &type);
2746 
2747 		if (!signr)
2748 			break; /* will return 0 */
2749 
2750 		if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2751 		    !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2752 			signr = ptrace_signal(signr, &ksig->info, type);
2753 			if (!signr)
2754 				continue;
2755 		}
2756 
2757 		ka = &sighand->action[signr-1];
2758 
2759 		/* Trace actually delivered signals. */
2760 		trace_signal_deliver(signr, &ksig->info, ka);
2761 
2762 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2763 			continue;
2764 		if (ka->sa.sa_handler != SIG_DFL) {
2765 			/* Run the handler.  */
2766 			ksig->ka = *ka;
2767 
2768 			if (ka->sa.sa_flags & SA_ONESHOT)
2769 				ka->sa.sa_handler = SIG_DFL;
2770 
2771 			break; /* will return non-zero "signr" value */
2772 		}
2773 
2774 		/*
2775 		 * Now we are doing the default action for this signal.
2776 		 */
2777 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2778 			continue;
2779 
2780 		/*
2781 		 * Global init gets no signals it doesn't want.
2782 		 * Container-init gets no signals it doesn't want from same
2783 		 * container.
2784 		 *
2785 		 * Note that if global/container-init sees a sig_kernel_only()
2786 		 * signal here, the signal must have been generated internally
2787 		 * or must have come from an ancestor namespace. In either
2788 		 * case, the signal cannot be dropped.
2789 		 */
2790 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2791 				!sig_kernel_only(signr))
2792 			continue;
2793 
2794 		if (sig_kernel_stop(signr)) {
2795 			/*
2796 			 * The default action is to stop all threads in
2797 			 * the thread group.  The job control signals
2798 			 * do nothing in an orphaned pgrp, but SIGSTOP
2799 			 * always works.  Note that siglock needs to be
2800 			 * dropped during the call to is_orphaned_pgrp()
2801 			 * because of lock ordering with tasklist_lock.
2802 			 * This allows an intervening SIGCONT to be posted.
2803 			 * We need to check for that and bail out if necessary.
2804 			 */
2805 			if (signr != SIGSTOP) {
2806 				spin_unlock_irq(&sighand->siglock);
2807 
2808 				/* signals can be posted during this window */
2809 
2810 				if (is_current_pgrp_orphaned())
2811 					goto relock;
2812 
2813 				spin_lock_irq(&sighand->siglock);
2814 			}
2815 
2816 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2817 				/* It released the siglock.  */
2818 				goto relock;
2819 			}
2820 
2821 			/*
2822 			 * We didn't actually stop, due to a race
2823 			 * with SIGCONT or something like that.
2824 			 */
2825 			continue;
2826 		}
2827 
2828 	fatal:
2829 		spin_unlock_irq(&sighand->siglock);
2830 		if (unlikely(cgroup_task_frozen(current)))
2831 			cgroup_leave_frozen(true);
2832 
2833 		/*
2834 		 * Anything else is fatal, maybe with a core dump.
2835 		 */
2836 		current->flags |= PF_SIGNALED;
2837 
2838 		if (sig_kernel_coredump(signr)) {
2839 			if (print_fatal_signals)
2840 				print_fatal_signal(ksig->info.si_signo);
2841 			proc_coredump_connector(current);
2842 			/*
2843 			 * If it was able to dump core, this kills all
2844 			 * other threads in the group and synchronizes with
2845 			 * their demise.  If we lost the race with another
2846 			 * thread getting here, it set group_exit_code
2847 			 * first and our do_group_exit call below will use
2848 			 * that value and ignore the one we pass it.
2849 			 */
2850 			do_coredump(&ksig->info);
2851 		}
2852 
2853 		/*
2854 		 * PF_IO_WORKER threads will catch and exit on fatal signals
2855 		 * themselves. They have cleanup that must be performed, so
2856 		 * we cannot call do_exit() on their behalf.
2857 		 */
2858 		if (current->flags & PF_IO_WORKER)
2859 			goto out;
2860 
2861 		/*
2862 		 * Death signals, no core dump.
2863 		 */
2864 		do_group_exit(ksig->info.si_signo);
2865 		/* NOTREACHED */
2866 	}
2867 	spin_unlock_irq(&sighand->siglock);
2868 out:
2869 	ksig->sig = signr;
2870 
2871 	if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2872 		hide_si_addr_tag_bits(ksig);
2873 
2874 	return ksig->sig > 0;
2875 }
2876 
2877 /**
2878  * signal_delivered - called after signal delivery to update blocked signals
2879  * @ksig:		kernel signal struct
2880  * @stepping:		nonzero if debugger single-step or block-step in use
2881  *
2882  * This function should be called when a signal has successfully been
2883  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2884  * is always blocked), and the signal itself is blocked unless %SA_NODEFER
2885  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2886  */
2887 static void signal_delivered(struct ksignal *ksig, int stepping)
2888 {
2889 	sigset_t blocked;
2890 
2891 	/* A signal was successfully delivered, and the
2892 	   saved sigmask was stored on the signal frame,
2893 	   and will be restored by sigreturn.  So we can
2894 	   simply clear the restore sigmask flag.  */
2895 	clear_restore_sigmask();
2896 
2897 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2898 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2899 		sigaddset(&blocked, ksig->sig);
2900 	set_current_blocked(&blocked);
2901 	if (current->sas_ss_flags & SS_AUTODISARM)
2902 		sas_ss_reset(current);
2903 	if (stepping)
2904 		ptrace_notify(SIGTRAP, 0);
2905 }
2906 
2907 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2908 {
2909 	if (failed)
2910 		force_sigsegv(ksig->sig);
2911 	else
2912 		signal_delivered(ksig, stepping);
2913 }
2914 
2915 /*
2916  * It could be that complete_signal() picked us to notify about the
2917  * group-wide signal. Other threads should be notified now to take
2918  * the shared signals in @which since we will not.
2919  */
2920 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2921 {
2922 	sigset_t retarget;
2923 	struct task_struct *t;
2924 
2925 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2926 	if (sigisemptyset(&retarget))
2927 		return;
2928 
2929 	t = tsk;
2930 	while_each_thread(tsk, t) {
2931 		if (t->flags & PF_EXITING)
2932 			continue;
2933 
2934 		if (!has_pending_signals(&retarget, &t->blocked))
2935 			continue;
2936 		/* Remove the signals this thread can handle. */
2937 		sigandsets(&retarget, &retarget, &t->blocked);
2938 
2939 		if (!task_sigpending(t))
2940 			signal_wake_up(t, 0);
2941 
2942 		if (sigisemptyset(&retarget))
2943 			break;
2944 	}
2945 }
2946 
2947 void exit_signals(struct task_struct *tsk)
2948 {
2949 	int group_stop = 0;
2950 	sigset_t unblocked;
2951 
2952 	/*
2953 	 * @tsk is about to have PF_EXITING set - lock out users which
2954 	 * expect stable threadgroup.
2955 	 */
2956 	cgroup_threadgroup_change_begin(tsk);
2957 
2958 	if (thread_group_empty(tsk) || (tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
2959 		tsk->flags |= PF_EXITING;
2960 		cgroup_threadgroup_change_end(tsk);
2961 		return;
2962 	}
2963 
2964 	spin_lock_irq(&tsk->sighand->siglock);
2965 	/*
2966 	 * From now this task is not visible for group-wide signals,
2967 	 * see wants_signal(), do_signal_stop().
2968 	 */
2969 	tsk->flags |= PF_EXITING;
2970 
2971 	cgroup_threadgroup_change_end(tsk);
2972 
2973 	if (!task_sigpending(tsk))
2974 		goto out;
2975 
2976 	unblocked = tsk->blocked;
2977 	signotset(&unblocked);
2978 	retarget_shared_pending(tsk, &unblocked);
2979 
2980 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2981 	    task_participate_group_stop(tsk))
2982 		group_stop = CLD_STOPPED;
2983 out:
2984 	spin_unlock_irq(&tsk->sighand->siglock);
2985 
2986 	/*
2987 	 * If group stop has completed, deliver the notification.  This
2988 	 * should always go to the real parent of the group leader.
2989 	 */
2990 	if (unlikely(group_stop)) {
2991 		read_lock(&tasklist_lock);
2992 		do_notify_parent_cldstop(tsk, false, group_stop);
2993 		read_unlock(&tasklist_lock);
2994 	}
2995 }
2996 
2997 /*
2998  * System call entry points.
2999  */
3000 
3001 /**
3002  *  sys_restart_syscall - restart a system call
3003  */
3004 SYSCALL_DEFINE0(restart_syscall)
3005 {
3006 	struct restart_block *restart = &current->restart_block;
3007 	return restart->fn(restart);
3008 }
3009 
3010 long do_no_restart_syscall(struct restart_block *param)
3011 {
3012 	return -EINTR;
3013 }
3014 
3015 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
3016 {
3017 	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3018 		sigset_t newblocked;
3019 		/* A set of now blocked but previously unblocked signals. */
3020 		sigandnsets(&newblocked, newset, &current->blocked);
3021 		retarget_shared_pending(tsk, &newblocked);
3022 	}
3023 	tsk->blocked = *newset;
3024 	recalc_sigpending();
3025 }
3026 
3027 /**
3028  * set_current_blocked - change current->blocked mask
3029  * @newset: new mask
3030  *
3031  * It is wrong to change ->blocked directly, this helper should be used
3032  * to ensure the process can't miss a shared signal we are going to block.
3033  */
3034 void set_current_blocked(sigset_t *newset)
3035 {
3036 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3037 	__set_current_blocked(newset);
3038 }
3039 
3040 void __set_current_blocked(const sigset_t *newset)
3041 {
3042 	struct task_struct *tsk = current;
3043 
3044 	/*
3045 	 * In case the signal mask hasn't changed, there is nothing we need
3046 	 * to do. The current->blocked shouldn't be modified by other task.
3047 	 */
3048 	if (sigequalsets(&tsk->blocked, newset))
3049 		return;
3050 
3051 	spin_lock_irq(&tsk->sighand->siglock);
3052 	__set_task_blocked(tsk, newset);
3053 	spin_unlock_irq(&tsk->sighand->siglock);
3054 }
3055 
3056 /*
3057  * This is also useful for kernel threads that want to temporarily
3058  * (or permanently) block certain signals.
3059  *
3060  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3061  * interface happily blocks "unblockable" signals like SIGKILL
3062  * and friends.
3063  */
3064 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3065 {
3066 	struct task_struct *tsk = current;
3067 	sigset_t newset;
3068 
3069 	/* Lockless, only current can change ->blocked, never from irq */
3070 	if (oldset)
3071 		*oldset = tsk->blocked;
3072 
3073 	switch (how) {
3074 	case SIG_BLOCK:
3075 		sigorsets(&newset, &tsk->blocked, set);
3076 		break;
3077 	case SIG_UNBLOCK:
3078 		sigandnsets(&newset, &tsk->blocked, set);
3079 		break;
3080 	case SIG_SETMASK:
3081 		newset = *set;
3082 		break;
3083 	default:
3084 		return -EINVAL;
3085 	}
3086 
3087 	__set_current_blocked(&newset);
3088 	return 0;
3089 }
3090 EXPORT_SYMBOL(sigprocmask);
3091 
3092 /*
3093  * The api helps set app-provided sigmasks.
3094  *
3095  * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3096  * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3097  *
3098  * Note that it does set_restore_sigmask() in advance, so it must be always
3099  * paired with restore_saved_sigmask_unless() before return from syscall.
3100  */
3101 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3102 {
3103 	sigset_t kmask;
3104 
3105 	if (!umask)
3106 		return 0;
3107 	if (sigsetsize != sizeof(sigset_t))
3108 		return -EINVAL;
3109 	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3110 		return -EFAULT;
3111 
3112 	set_restore_sigmask();
3113 	current->saved_sigmask = current->blocked;
3114 	set_current_blocked(&kmask);
3115 
3116 	return 0;
3117 }
3118 
3119 #ifdef CONFIG_COMPAT
3120 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3121 			    size_t sigsetsize)
3122 {
3123 	sigset_t kmask;
3124 
3125 	if (!umask)
3126 		return 0;
3127 	if (sigsetsize != sizeof(compat_sigset_t))
3128 		return -EINVAL;
3129 	if (get_compat_sigset(&kmask, umask))
3130 		return -EFAULT;
3131 
3132 	set_restore_sigmask();
3133 	current->saved_sigmask = current->blocked;
3134 	set_current_blocked(&kmask);
3135 
3136 	return 0;
3137 }
3138 #endif
3139 
3140 /**
3141  *  sys_rt_sigprocmask - change the list of currently blocked signals
3142  *  @how: whether to add, remove, or set signals
3143  *  @nset: stores pending signals
3144  *  @oset: previous value of signal mask if non-null
3145  *  @sigsetsize: size of sigset_t type
3146  */
3147 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3148 		sigset_t __user *, oset, size_t, sigsetsize)
3149 {
3150 	sigset_t old_set, new_set;
3151 	int error;
3152 
3153 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3154 	if (sigsetsize != sizeof(sigset_t))
3155 		return -EINVAL;
3156 
3157 	old_set = current->blocked;
3158 
3159 	if (nset) {
3160 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3161 			return -EFAULT;
3162 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3163 
3164 		error = sigprocmask(how, &new_set, NULL);
3165 		if (error)
3166 			return error;
3167 	}
3168 
3169 	if (oset) {
3170 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3171 			return -EFAULT;
3172 	}
3173 
3174 	return 0;
3175 }
3176 
3177 #ifdef CONFIG_COMPAT
3178 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3179 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3180 {
3181 	sigset_t old_set = current->blocked;
3182 
3183 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3184 	if (sigsetsize != sizeof(sigset_t))
3185 		return -EINVAL;
3186 
3187 	if (nset) {
3188 		sigset_t new_set;
3189 		int error;
3190 		if (get_compat_sigset(&new_set, nset))
3191 			return -EFAULT;
3192 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3193 
3194 		error = sigprocmask(how, &new_set, NULL);
3195 		if (error)
3196 			return error;
3197 	}
3198 	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3199 }
3200 #endif
3201 
3202 static void do_sigpending(sigset_t *set)
3203 {
3204 	spin_lock_irq(&current->sighand->siglock);
3205 	sigorsets(set, &current->pending.signal,
3206 		  &current->signal->shared_pending.signal);
3207 	spin_unlock_irq(&current->sighand->siglock);
3208 
3209 	/* Outside the lock because only this thread touches it.  */
3210 	sigandsets(set, &current->blocked, set);
3211 }
3212 
3213 /**
3214  *  sys_rt_sigpending - examine a pending signal that has been raised
3215  *			while blocked
3216  *  @uset: stores pending signals
3217  *  @sigsetsize: size of sigset_t type or larger
3218  */
3219 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3220 {
3221 	sigset_t set;
3222 
3223 	if (sigsetsize > sizeof(*uset))
3224 		return -EINVAL;
3225 
3226 	do_sigpending(&set);
3227 
3228 	if (copy_to_user(uset, &set, sigsetsize))
3229 		return -EFAULT;
3230 
3231 	return 0;
3232 }
3233 
3234 #ifdef CONFIG_COMPAT
3235 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3236 		compat_size_t, sigsetsize)
3237 {
3238 	sigset_t set;
3239 
3240 	if (sigsetsize > sizeof(*uset))
3241 		return -EINVAL;
3242 
3243 	do_sigpending(&set);
3244 
3245 	return put_compat_sigset(uset, &set, sigsetsize);
3246 }
3247 #endif
3248 
3249 static const struct {
3250 	unsigned char limit, layout;
3251 } sig_sicodes[] = {
3252 	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3253 	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3254 	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3255 	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3256 	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3257 #if defined(SIGEMT)
3258 	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3259 #endif
3260 	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3261 	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3262 	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3263 };
3264 
3265 static bool known_siginfo_layout(unsigned sig, int si_code)
3266 {
3267 	if (si_code == SI_KERNEL)
3268 		return true;
3269 	else if ((si_code > SI_USER)) {
3270 		if (sig_specific_sicodes(sig)) {
3271 			if (si_code <= sig_sicodes[sig].limit)
3272 				return true;
3273 		}
3274 		else if (si_code <= NSIGPOLL)
3275 			return true;
3276 	}
3277 	else if (si_code >= SI_DETHREAD)
3278 		return true;
3279 	else if (si_code == SI_ASYNCNL)
3280 		return true;
3281 	return false;
3282 }
3283 
3284 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3285 {
3286 	enum siginfo_layout layout = SIL_KILL;
3287 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3288 		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3289 		    (si_code <= sig_sicodes[sig].limit)) {
3290 			layout = sig_sicodes[sig].layout;
3291 			/* Handle the exceptions */
3292 			if ((sig == SIGBUS) &&
3293 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3294 				layout = SIL_FAULT_MCEERR;
3295 			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3296 				layout = SIL_FAULT_BNDERR;
3297 #ifdef SEGV_PKUERR
3298 			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3299 				layout = SIL_FAULT_PKUERR;
3300 #endif
3301 			else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3302 				layout = SIL_FAULT_PERF_EVENT;
3303 			else if (IS_ENABLED(CONFIG_SPARC) &&
3304 				 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3305 				layout = SIL_FAULT_TRAPNO;
3306 			else if (IS_ENABLED(CONFIG_ALPHA) &&
3307 				 ((sig == SIGFPE) ||
3308 				  ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3309 				layout = SIL_FAULT_TRAPNO;
3310 		}
3311 		else if (si_code <= NSIGPOLL)
3312 			layout = SIL_POLL;
3313 	} else {
3314 		if (si_code == SI_TIMER)
3315 			layout = SIL_TIMER;
3316 		else if (si_code == SI_SIGIO)
3317 			layout = SIL_POLL;
3318 		else if (si_code < 0)
3319 			layout = SIL_RT;
3320 	}
3321 	return layout;
3322 }
3323 
3324 static inline char __user *si_expansion(const siginfo_t __user *info)
3325 {
3326 	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3327 }
3328 
3329 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3330 {
3331 	char __user *expansion = si_expansion(to);
3332 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3333 		return -EFAULT;
3334 	if (clear_user(expansion, SI_EXPANSION_SIZE))
3335 		return -EFAULT;
3336 	return 0;
3337 }
3338 
3339 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3340 				       const siginfo_t __user *from)
3341 {
3342 	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3343 		char __user *expansion = si_expansion(from);
3344 		char buf[SI_EXPANSION_SIZE];
3345 		int i;
3346 		/*
3347 		 * An unknown si_code might need more than
3348 		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3349 		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3350 		 * will return this data to userspace exactly.
3351 		 */
3352 		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3353 			return -EFAULT;
3354 		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3355 			if (buf[i] != 0)
3356 				return -E2BIG;
3357 		}
3358 	}
3359 	return 0;
3360 }
3361 
3362 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3363 				    const siginfo_t __user *from)
3364 {
3365 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3366 		return -EFAULT;
3367 	to->si_signo = signo;
3368 	return post_copy_siginfo_from_user(to, from);
3369 }
3370 
3371 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3372 {
3373 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3374 		return -EFAULT;
3375 	return post_copy_siginfo_from_user(to, from);
3376 }
3377 
3378 #ifdef CONFIG_COMPAT
3379 /**
3380  * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3381  * @to: compat siginfo destination
3382  * @from: kernel siginfo source
3383  *
3384  * Note: This function does not work properly for the SIGCHLD on x32, but
3385  * fortunately it doesn't have to.  The only valid callers for this function are
3386  * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3387  * The latter does not care because SIGCHLD will never cause a coredump.
3388  */
3389 void copy_siginfo_to_external32(struct compat_siginfo *to,
3390 		const struct kernel_siginfo *from)
3391 {
3392 	memset(to, 0, sizeof(*to));
3393 
3394 	to->si_signo = from->si_signo;
3395 	to->si_errno = from->si_errno;
3396 	to->si_code  = from->si_code;
3397 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3398 	case SIL_KILL:
3399 		to->si_pid = from->si_pid;
3400 		to->si_uid = from->si_uid;
3401 		break;
3402 	case SIL_TIMER:
3403 		to->si_tid     = from->si_tid;
3404 		to->si_overrun = from->si_overrun;
3405 		to->si_int     = from->si_int;
3406 		break;
3407 	case SIL_POLL:
3408 		to->si_band = from->si_band;
3409 		to->si_fd   = from->si_fd;
3410 		break;
3411 	case SIL_FAULT:
3412 		to->si_addr = ptr_to_compat(from->si_addr);
3413 		break;
3414 	case SIL_FAULT_TRAPNO:
3415 		to->si_addr = ptr_to_compat(from->si_addr);
3416 		to->si_trapno = from->si_trapno;
3417 		break;
3418 	case SIL_FAULT_MCEERR:
3419 		to->si_addr = ptr_to_compat(from->si_addr);
3420 		to->si_addr_lsb = from->si_addr_lsb;
3421 		break;
3422 	case SIL_FAULT_BNDERR:
3423 		to->si_addr = ptr_to_compat(from->si_addr);
3424 		to->si_lower = ptr_to_compat(from->si_lower);
3425 		to->si_upper = ptr_to_compat(from->si_upper);
3426 		break;
3427 	case SIL_FAULT_PKUERR:
3428 		to->si_addr = ptr_to_compat(from->si_addr);
3429 		to->si_pkey = from->si_pkey;
3430 		break;
3431 	case SIL_FAULT_PERF_EVENT:
3432 		to->si_addr = ptr_to_compat(from->si_addr);
3433 		to->si_perf_data = from->si_perf_data;
3434 		to->si_perf_type = from->si_perf_type;
3435 		break;
3436 	case SIL_CHLD:
3437 		to->si_pid = from->si_pid;
3438 		to->si_uid = from->si_uid;
3439 		to->si_status = from->si_status;
3440 		to->si_utime = from->si_utime;
3441 		to->si_stime = from->si_stime;
3442 		break;
3443 	case SIL_RT:
3444 		to->si_pid = from->si_pid;
3445 		to->si_uid = from->si_uid;
3446 		to->si_int = from->si_int;
3447 		break;
3448 	case SIL_SYS:
3449 		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3450 		to->si_syscall   = from->si_syscall;
3451 		to->si_arch      = from->si_arch;
3452 		break;
3453 	}
3454 }
3455 
3456 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3457 			   const struct kernel_siginfo *from)
3458 {
3459 	struct compat_siginfo new;
3460 
3461 	copy_siginfo_to_external32(&new, from);
3462 	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3463 		return -EFAULT;
3464 	return 0;
3465 }
3466 
3467 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3468 					 const struct compat_siginfo *from)
3469 {
3470 	clear_siginfo(to);
3471 	to->si_signo = from->si_signo;
3472 	to->si_errno = from->si_errno;
3473 	to->si_code  = from->si_code;
3474 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3475 	case SIL_KILL:
3476 		to->si_pid = from->si_pid;
3477 		to->si_uid = from->si_uid;
3478 		break;
3479 	case SIL_TIMER:
3480 		to->si_tid     = from->si_tid;
3481 		to->si_overrun = from->si_overrun;
3482 		to->si_int     = from->si_int;
3483 		break;
3484 	case SIL_POLL:
3485 		to->si_band = from->si_band;
3486 		to->si_fd   = from->si_fd;
3487 		break;
3488 	case SIL_FAULT:
3489 		to->si_addr = compat_ptr(from->si_addr);
3490 		break;
3491 	case SIL_FAULT_TRAPNO:
3492 		to->si_addr = compat_ptr(from->si_addr);
3493 		to->si_trapno = from->si_trapno;
3494 		break;
3495 	case SIL_FAULT_MCEERR:
3496 		to->si_addr = compat_ptr(from->si_addr);
3497 		to->si_addr_lsb = from->si_addr_lsb;
3498 		break;
3499 	case SIL_FAULT_BNDERR:
3500 		to->si_addr = compat_ptr(from->si_addr);
3501 		to->si_lower = compat_ptr(from->si_lower);
3502 		to->si_upper = compat_ptr(from->si_upper);
3503 		break;
3504 	case SIL_FAULT_PKUERR:
3505 		to->si_addr = compat_ptr(from->si_addr);
3506 		to->si_pkey = from->si_pkey;
3507 		break;
3508 	case SIL_FAULT_PERF_EVENT:
3509 		to->si_addr = compat_ptr(from->si_addr);
3510 		to->si_perf_data = from->si_perf_data;
3511 		to->si_perf_type = from->si_perf_type;
3512 		break;
3513 	case SIL_CHLD:
3514 		to->si_pid    = from->si_pid;
3515 		to->si_uid    = from->si_uid;
3516 		to->si_status = from->si_status;
3517 #ifdef CONFIG_X86_X32_ABI
3518 		if (in_x32_syscall()) {
3519 			to->si_utime = from->_sifields._sigchld_x32._utime;
3520 			to->si_stime = from->_sifields._sigchld_x32._stime;
3521 		} else
3522 #endif
3523 		{
3524 			to->si_utime = from->si_utime;
3525 			to->si_stime = from->si_stime;
3526 		}
3527 		break;
3528 	case SIL_RT:
3529 		to->si_pid = from->si_pid;
3530 		to->si_uid = from->si_uid;
3531 		to->si_int = from->si_int;
3532 		break;
3533 	case SIL_SYS:
3534 		to->si_call_addr = compat_ptr(from->si_call_addr);
3535 		to->si_syscall   = from->si_syscall;
3536 		to->si_arch      = from->si_arch;
3537 		break;
3538 	}
3539 	return 0;
3540 }
3541 
3542 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3543 				      const struct compat_siginfo __user *ufrom)
3544 {
3545 	struct compat_siginfo from;
3546 
3547 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3548 		return -EFAULT;
3549 
3550 	from.si_signo = signo;
3551 	return post_copy_siginfo_from_user32(to, &from);
3552 }
3553 
3554 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3555 			     const struct compat_siginfo __user *ufrom)
3556 {
3557 	struct compat_siginfo from;
3558 
3559 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3560 		return -EFAULT;
3561 
3562 	return post_copy_siginfo_from_user32(to, &from);
3563 }
3564 #endif /* CONFIG_COMPAT */
3565 
3566 /**
3567  *  do_sigtimedwait - wait for queued signals specified in @which
3568  *  @which: queued signals to wait for
3569  *  @info: if non-null, the signal's siginfo is returned here
3570  *  @ts: upper bound on process time suspension
3571  */
3572 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3573 		    const struct timespec64 *ts)
3574 {
3575 	ktime_t *to = NULL, timeout = KTIME_MAX;
3576 	struct task_struct *tsk = current;
3577 	sigset_t mask = *which;
3578 	enum pid_type type;
3579 	int sig, ret = 0;
3580 
3581 	if (ts) {
3582 		if (!timespec64_valid(ts))
3583 			return -EINVAL;
3584 		timeout = timespec64_to_ktime(*ts);
3585 		to = &timeout;
3586 	}
3587 
3588 	/*
3589 	 * Invert the set of allowed signals to get those we want to block.
3590 	 */
3591 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3592 	signotset(&mask);
3593 
3594 	spin_lock_irq(&tsk->sighand->siglock);
3595 	sig = dequeue_signal(tsk, &mask, info, &type);
3596 	if (!sig && timeout) {
3597 		/*
3598 		 * None ready, temporarily unblock those we're interested
3599 		 * while we are sleeping in so that we'll be awakened when
3600 		 * they arrive. Unblocking is always fine, we can avoid
3601 		 * set_current_blocked().
3602 		 */
3603 		tsk->real_blocked = tsk->blocked;
3604 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3605 		recalc_sigpending();
3606 		spin_unlock_irq(&tsk->sighand->siglock);
3607 
3608 		__set_current_state(TASK_INTERRUPTIBLE);
3609 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3610 							 HRTIMER_MODE_REL);
3611 		spin_lock_irq(&tsk->sighand->siglock);
3612 		__set_task_blocked(tsk, &tsk->real_blocked);
3613 		sigemptyset(&tsk->real_blocked);
3614 		sig = dequeue_signal(tsk, &mask, info, &type);
3615 	}
3616 	spin_unlock_irq(&tsk->sighand->siglock);
3617 
3618 	if (sig)
3619 		return sig;
3620 	return ret ? -EINTR : -EAGAIN;
3621 }
3622 
3623 /**
3624  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3625  *			in @uthese
3626  *  @uthese: queued signals to wait for
3627  *  @uinfo: if non-null, the signal's siginfo is returned here
3628  *  @uts: upper bound on process time suspension
3629  *  @sigsetsize: size of sigset_t type
3630  */
3631 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3632 		siginfo_t __user *, uinfo,
3633 		const struct __kernel_timespec __user *, uts,
3634 		size_t, sigsetsize)
3635 {
3636 	sigset_t these;
3637 	struct timespec64 ts;
3638 	kernel_siginfo_t info;
3639 	int ret;
3640 
3641 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3642 	if (sigsetsize != sizeof(sigset_t))
3643 		return -EINVAL;
3644 
3645 	if (copy_from_user(&these, uthese, sizeof(these)))
3646 		return -EFAULT;
3647 
3648 	if (uts) {
3649 		if (get_timespec64(&ts, uts))
3650 			return -EFAULT;
3651 	}
3652 
3653 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3654 
3655 	if (ret > 0 && uinfo) {
3656 		if (copy_siginfo_to_user(uinfo, &info))
3657 			ret = -EFAULT;
3658 	}
3659 
3660 	return ret;
3661 }
3662 
3663 #ifdef CONFIG_COMPAT_32BIT_TIME
3664 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3665 		siginfo_t __user *, uinfo,
3666 		const struct old_timespec32 __user *, uts,
3667 		size_t, sigsetsize)
3668 {
3669 	sigset_t these;
3670 	struct timespec64 ts;
3671 	kernel_siginfo_t info;
3672 	int ret;
3673 
3674 	if (sigsetsize != sizeof(sigset_t))
3675 		return -EINVAL;
3676 
3677 	if (copy_from_user(&these, uthese, sizeof(these)))
3678 		return -EFAULT;
3679 
3680 	if (uts) {
3681 		if (get_old_timespec32(&ts, uts))
3682 			return -EFAULT;
3683 	}
3684 
3685 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3686 
3687 	if (ret > 0 && uinfo) {
3688 		if (copy_siginfo_to_user(uinfo, &info))
3689 			ret = -EFAULT;
3690 	}
3691 
3692 	return ret;
3693 }
3694 #endif
3695 
3696 #ifdef CONFIG_COMPAT
3697 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3698 		struct compat_siginfo __user *, uinfo,
3699 		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3700 {
3701 	sigset_t s;
3702 	struct timespec64 t;
3703 	kernel_siginfo_t info;
3704 	long ret;
3705 
3706 	if (sigsetsize != sizeof(sigset_t))
3707 		return -EINVAL;
3708 
3709 	if (get_compat_sigset(&s, uthese))
3710 		return -EFAULT;
3711 
3712 	if (uts) {
3713 		if (get_timespec64(&t, uts))
3714 			return -EFAULT;
3715 	}
3716 
3717 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3718 
3719 	if (ret > 0 && uinfo) {
3720 		if (copy_siginfo_to_user32(uinfo, &info))
3721 			ret = -EFAULT;
3722 	}
3723 
3724 	return ret;
3725 }
3726 
3727 #ifdef CONFIG_COMPAT_32BIT_TIME
3728 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3729 		struct compat_siginfo __user *, uinfo,
3730 		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3731 {
3732 	sigset_t s;
3733 	struct timespec64 t;
3734 	kernel_siginfo_t info;
3735 	long ret;
3736 
3737 	if (sigsetsize != sizeof(sigset_t))
3738 		return -EINVAL;
3739 
3740 	if (get_compat_sigset(&s, uthese))
3741 		return -EFAULT;
3742 
3743 	if (uts) {
3744 		if (get_old_timespec32(&t, uts))
3745 			return -EFAULT;
3746 	}
3747 
3748 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3749 
3750 	if (ret > 0 && uinfo) {
3751 		if (copy_siginfo_to_user32(uinfo, &info))
3752 			ret = -EFAULT;
3753 	}
3754 
3755 	return ret;
3756 }
3757 #endif
3758 #endif
3759 
3760 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3761 {
3762 	clear_siginfo(info);
3763 	info->si_signo = sig;
3764 	info->si_errno = 0;
3765 	info->si_code = SI_USER;
3766 	info->si_pid = task_tgid_vnr(current);
3767 	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3768 }
3769 
3770 /**
3771  *  sys_kill - send a signal to a process
3772  *  @pid: the PID of the process
3773  *  @sig: signal to be sent
3774  */
3775 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3776 {
3777 	struct kernel_siginfo info;
3778 
3779 	prepare_kill_siginfo(sig, &info);
3780 
3781 	return kill_something_info(sig, &info, pid);
3782 }
3783 
3784 /*
3785  * Verify that the signaler and signalee either are in the same pid namespace
3786  * or that the signaler's pid namespace is an ancestor of the signalee's pid
3787  * namespace.
3788  */
3789 static bool access_pidfd_pidns(struct pid *pid)
3790 {
3791 	struct pid_namespace *active = task_active_pid_ns(current);
3792 	struct pid_namespace *p = ns_of_pid(pid);
3793 
3794 	for (;;) {
3795 		if (!p)
3796 			return false;
3797 		if (p == active)
3798 			break;
3799 		p = p->parent;
3800 	}
3801 
3802 	return true;
3803 }
3804 
3805 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3806 		siginfo_t __user *info)
3807 {
3808 #ifdef CONFIG_COMPAT
3809 	/*
3810 	 * Avoid hooking up compat syscalls and instead handle necessary
3811 	 * conversions here. Note, this is a stop-gap measure and should not be
3812 	 * considered a generic solution.
3813 	 */
3814 	if (in_compat_syscall())
3815 		return copy_siginfo_from_user32(
3816 			kinfo, (struct compat_siginfo __user *)info);
3817 #endif
3818 	return copy_siginfo_from_user(kinfo, info);
3819 }
3820 
3821 static struct pid *pidfd_to_pid(const struct file *file)
3822 {
3823 	struct pid *pid;
3824 
3825 	pid = pidfd_pid(file);
3826 	if (!IS_ERR(pid))
3827 		return pid;
3828 
3829 	return tgid_pidfd_to_pid(file);
3830 }
3831 
3832 /**
3833  * sys_pidfd_send_signal - Signal a process through a pidfd
3834  * @pidfd:  file descriptor of the process
3835  * @sig:    signal to send
3836  * @info:   signal info
3837  * @flags:  future flags
3838  *
3839  * The syscall currently only signals via PIDTYPE_PID which covers
3840  * kill(<positive-pid>, <signal>. It does not signal threads or process
3841  * groups.
3842  * In order to extend the syscall to threads and process groups the @flags
3843  * argument should be used. In essence, the @flags argument will determine
3844  * what is signaled and not the file descriptor itself. Put in other words,
3845  * grouping is a property of the flags argument not a property of the file
3846  * descriptor.
3847  *
3848  * Return: 0 on success, negative errno on failure
3849  */
3850 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3851 		siginfo_t __user *, info, unsigned int, flags)
3852 {
3853 	int ret;
3854 	struct fd f;
3855 	struct pid *pid;
3856 	kernel_siginfo_t kinfo;
3857 
3858 	/* Enforce flags be set to 0 until we add an extension. */
3859 	if (flags)
3860 		return -EINVAL;
3861 
3862 	f = fdget(pidfd);
3863 	if (!f.file)
3864 		return -EBADF;
3865 
3866 	/* Is this a pidfd? */
3867 	pid = pidfd_to_pid(f.file);
3868 	if (IS_ERR(pid)) {
3869 		ret = PTR_ERR(pid);
3870 		goto err;
3871 	}
3872 
3873 	ret = -EINVAL;
3874 	if (!access_pidfd_pidns(pid))
3875 		goto err;
3876 
3877 	if (info) {
3878 		ret = copy_siginfo_from_user_any(&kinfo, info);
3879 		if (unlikely(ret))
3880 			goto err;
3881 
3882 		ret = -EINVAL;
3883 		if (unlikely(sig != kinfo.si_signo))
3884 			goto err;
3885 
3886 		/* Only allow sending arbitrary signals to yourself. */
3887 		ret = -EPERM;
3888 		if ((task_pid(current) != pid) &&
3889 		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3890 			goto err;
3891 	} else {
3892 		prepare_kill_siginfo(sig, &kinfo);
3893 	}
3894 
3895 	ret = kill_pid_info(sig, &kinfo, pid);
3896 
3897 err:
3898 	fdput(f);
3899 	return ret;
3900 }
3901 
3902 static int
3903 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3904 {
3905 	struct task_struct *p;
3906 	int error = -ESRCH;
3907 
3908 	rcu_read_lock();
3909 	p = find_task_by_vpid(pid);
3910 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3911 		error = check_kill_permission(sig, info, p);
3912 		/*
3913 		 * The null signal is a permissions and process existence
3914 		 * probe.  No signal is actually delivered.
3915 		 */
3916 		if (!error && sig) {
3917 			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3918 			/*
3919 			 * If lock_task_sighand() failed we pretend the task
3920 			 * dies after receiving the signal. The window is tiny,
3921 			 * and the signal is private anyway.
3922 			 */
3923 			if (unlikely(error == -ESRCH))
3924 				error = 0;
3925 		}
3926 	}
3927 	rcu_read_unlock();
3928 
3929 	return error;
3930 }
3931 
3932 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3933 {
3934 	struct kernel_siginfo info;
3935 
3936 	clear_siginfo(&info);
3937 	info.si_signo = sig;
3938 	info.si_errno = 0;
3939 	info.si_code = SI_TKILL;
3940 	info.si_pid = task_tgid_vnr(current);
3941 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3942 
3943 	return do_send_specific(tgid, pid, sig, &info);
3944 }
3945 
3946 /**
3947  *  sys_tgkill - send signal to one specific thread
3948  *  @tgid: the thread group ID of the thread
3949  *  @pid: the PID of the thread
3950  *  @sig: signal to be sent
3951  *
3952  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3953  *  exists but it's not belonging to the target process anymore. This
3954  *  method solves the problem of threads exiting and PIDs getting reused.
3955  */
3956 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3957 {
3958 	/* This is only valid for single tasks */
3959 	if (pid <= 0 || tgid <= 0)
3960 		return -EINVAL;
3961 
3962 	return do_tkill(tgid, pid, sig);
3963 }
3964 
3965 /**
3966  *  sys_tkill - send signal to one specific task
3967  *  @pid: the PID of the task
3968  *  @sig: signal to be sent
3969  *
3970  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3971  */
3972 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3973 {
3974 	/* This is only valid for single tasks */
3975 	if (pid <= 0)
3976 		return -EINVAL;
3977 
3978 	return do_tkill(0, pid, sig);
3979 }
3980 
3981 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
3982 {
3983 	/* Not even root can pretend to send signals from the kernel.
3984 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3985 	 */
3986 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3987 	    (task_pid_vnr(current) != pid))
3988 		return -EPERM;
3989 
3990 	/* POSIX.1b doesn't mention process groups.  */
3991 	return kill_proc_info(sig, info, pid);
3992 }
3993 
3994 /**
3995  *  sys_rt_sigqueueinfo - send signal information to a signal
3996  *  @pid: the PID of the thread
3997  *  @sig: signal to be sent
3998  *  @uinfo: signal info to be sent
3999  */
4000 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
4001 		siginfo_t __user *, uinfo)
4002 {
4003 	kernel_siginfo_t info;
4004 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4005 	if (unlikely(ret))
4006 		return ret;
4007 	return do_rt_sigqueueinfo(pid, sig, &info);
4008 }
4009 
4010 #ifdef CONFIG_COMPAT
4011 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
4012 			compat_pid_t, pid,
4013 			int, sig,
4014 			struct compat_siginfo __user *, uinfo)
4015 {
4016 	kernel_siginfo_t info;
4017 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4018 	if (unlikely(ret))
4019 		return ret;
4020 	return do_rt_sigqueueinfo(pid, sig, &info);
4021 }
4022 #endif
4023 
4024 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4025 {
4026 	/* This is only valid for single tasks */
4027 	if (pid <= 0 || tgid <= 0)
4028 		return -EINVAL;
4029 
4030 	/* Not even root can pretend to send signals from the kernel.
4031 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4032 	 */
4033 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4034 	    (task_pid_vnr(current) != pid))
4035 		return -EPERM;
4036 
4037 	return do_send_specific(tgid, pid, sig, info);
4038 }
4039 
4040 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4041 		siginfo_t __user *, uinfo)
4042 {
4043 	kernel_siginfo_t info;
4044 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4045 	if (unlikely(ret))
4046 		return ret;
4047 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4048 }
4049 
4050 #ifdef CONFIG_COMPAT
4051 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4052 			compat_pid_t, tgid,
4053 			compat_pid_t, pid,
4054 			int, sig,
4055 			struct compat_siginfo __user *, uinfo)
4056 {
4057 	kernel_siginfo_t info;
4058 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4059 	if (unlikely(ret))
4060 		return ret;
4061 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4062 }
4063 #endif
4064 
4065 /*
4066  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4067  */
4068 void kernel_sigaction(int sig, __sighandler_t action)
4069 {
4070 	spin_lock_irq(&current->sighand->siglock);
4071 	current->sighand->action[sig - 1].sa.sa_handler = action;
4072 	if (action == SIG_IGN) {
4073 		sigset_t mask;
4074 
4075 		sigemptyset(&mask);
4076 		sigaddset(&mask, sig);
4077 
4078 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
4079 		flush_sigqueue_mask(&mask, &current->pending);
4080 		recalc_sigpending();
4081 	}
4082 	spin_unlock_irq(&current->sighand->siglock);
4083 }
4084 EXPORT_SYMBOL(kernel_sigaction);
4085 
4086 void __weak sigaction_compat_abi(struct k_sigaction *act,
4087 		struct k_sigaction *oact)
4088 {
4089 }
4090 
4091 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4092 {
4093 	struct task_struct *p = current, *t;
4094 	struct k_sigaction *k;
4095 	sigset_t mask;
4096 
4097 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4098 		return -EINVAL;
4099 
4100 	k = &p->sighand->action[sig-1];
4101 
4102 	spin_lock_irq(&p->sighand->siglock);
4103 	if (k->sa.sa_flags & SA_IMMUTABLE) {
4104 		spin_unlock_irq(&p->sighand->siglock);
4105 		return -EINVAL;
4106 	}
4107 	if (oact)
4108 		*oact = *k;
4109 
4110 	/*
4111 	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4112 	 * e.g. by having an architecture use the bit in their uapi.
4113 	 */
4114 	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4115 
4116 	/*
4117 	 * Clear unknown flag bits in order to allow userspace to detect missing
4118 	 * support for flag bits and to allow the kernel to use non-uapi bits
4119 	 * internally.
4120 	 */
4121 	if (act)
4122 		act->sa.sa_flags &= UAPI_SA_FLAGS;
4123 	if (oact)
4124 		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4125 
4126 	sigaction_compat_abi(act, oact);
4127 
4128 	if (act) {
4129 		sigdelsetmask(&act->sa.sa_mask,
4130 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4131 		*k = *act;
4132 		/*
4133 		 * POSIX 3.3.1.3:
4134 		 *  "Setting a signal action to SIG_IGN for a signal that is
4135 		 *   pending shall cause the pending signal to be discarded,
4136 		 *   whether or not it is blocked."
4137 		 *
4138 		 *  "Setting a signal action to SIG_DFL for a signal that is
4139 		 *   pending and whose default action is to ignore the signal
4140 		 *   (for example, SIGCHLD), shall cause the pending signal to
4141 		 *   be discarded, whether or not it is blocked"
4142 		 */
4143 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4144 			sigemptyset(&mask);
4145 			sigaddset(&mask, sig);
4146 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4147 			for_each_thread(p, t)
4148 				flush_sigqueue_mask(&mask, &t->pending);
4149 		}
4150 	}
4151 
4152 	spin_unlock_irq(&p->sighand->siglock);
4153 	return 0;
4154 }
4155 
4156 #ifdef CONFIG_DYNAMIC_SIGFRAME
4157 static inline void sigaltstack_lock(void)
4158 	__acquires(&current->sighand->siglock)
4159 {
4160 	spin_lock_irq(&current->sighand->siglock);
4161 }
4162 
4163 static inline void sigaltstack_unlock(void)
4164 	__releases(&current->sighand->siglock)
4165 {
4166 	spin_unlock_irq(&current->sighand->siglock);
4167 }
4168 #else
4169 static inline void sigaltstack_lock(void) { }
4170 static inline void sigaltstack_unlock(void) { }
4171 #endif
4172 
4173 static int
4174 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4175 		size_t min_ss_size)
4176 {
4177 	struct task_struct *t = current;
4178 	int ret = 0;
4179 
4180 	if (oss) {
4181 		memset(oss, 0, sizeof(stack_t));
4182 		oss->ss_sp = (void __user *) t->sas_ss_sp;
4183 		oss->ss_size = t->sas_ss_size;
4184 		oss->ss_flags = sas_ss_flags(sp) |
4185 			(current->sas_ss_flags & SS_FLAG_BITS);
4186 	}
4187 
4188 	if (ss) {
4189 		void __user *ss_sp = ss->ss_sp;
4190 		size_t ss_size = ss->ss_size;
4191 		unsigned ss_flags = ss->ss_flags;
4192 		int ss_mode;
4193 
4194 		if (unlikely(on_sig_stack(sp)))
4195 			return -EPERM;
4196 
4197 		ss_mode = ss_flags & ~SS_FLAG_BITS;
4198 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4199 				ss_mode != 0))
4200 			return -EINVAL;
4201 
4202 		/*
4203 		 * Return before taking any locks if no actual
4204 		 * sigaltstack changes were requested.
4205 		 */
4206 		if (t->sas_ss_sp == (unsigned long)ss_sp &&
4207 		    t->sas_ss_size == ss_size &&
4208 		    t->sas_ss_flags == ss_flags)
4209 			return 0;
4210 
4211 		sigaltstack_lock();
4212 		if (ss_mode == SS_DISABLE) {
4213 			ss_size = 0;
4214 			ss_sp = NULL;
4215 		} else {
4216 			if (unlikely(ss_size < min_ss_size))
4217 				ret = -ENOMEM;
4218 			if (!sigaltstack_size_valid(ss_size))
4219 				ret = -ENOMEM;
4220 		}
4221 		if (!ret) {
4222 			t->sas_ss_sp = (unsigned long) ss_sp;
4223 			t->sas_ss_size = ss_size;
4224 			t->sas_ss_flags = ss_flags;
4225 		}
4226 		sigaltstack_unlock();
4227 	}
4228 	return ret;
4229 }
4230 
4231 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4232 {
4233 	stack_t new, old;
4234 	int err;
4235 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4236 		return -EFAULT;
4237 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4238 			      current_user_stack_pointer(),
4239 			      MINSIGSTKSZ);
4240 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4241 		err = -EFAULT;
4242 	return err;
4243 }
4244 
4245 int restore_altstack(const stack_t __user *uss)
4246 {
4247 	stack_t new;
4248 	if (copy_from_user(&new, uss, sizeof(stack_t)))
4249 		return -EFAULT;
4250 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4251 			     MINSIGSTKSZ);
4252 	/* squash all but EFAULT for now */
4253 	return 0;
4254 }
4255 
4256 int __save_altstack(stack_t __user *uss, unsigned long sp)
4257 {
4258 	struct task_struct *t = current;
4259 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4260 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4261 		__put_user(t->sas_ss_size, &uss->ss_size);
4262 	return err;
4263 }
4264 
4265 #ifdef CONFIG_COMPAT
4266 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4267 				 compat_stack_t __user *uoss_ptr)
4268 {
4269 	stack_t uss, uoss;
4270 	int ret;
4271 
4272 	if (uss_ptr) {
4273 		compat_stack_t uss32;
4274 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4275 			return -EFAULT;
4276 		uss.ss_sp = compat_ptr(uss32.ss_sp);
4277 		uss.ss_flags = uss32.ss_flags;
4278 		uss.ss_size = uss32.ss_size;
4279 	}
4280 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4281 			     compat_user_stack_pointer(),
4282 			     COMPAT_MINSIGSTKSZ);
4283 	if (ret >= 0 && uoss_ptr)  {
4284 		compat_stack_t old;
4285 		memset(&old, 0, sizeof(old));
4286 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4287 		old.ss_flags = uoss.ss_flags;
4288 		old.ss_size = uoss.ss_size;
4289 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4290 			ret = -EFAULT;
4291 	}
4292 	return ret;
4293 }
4294 
4295 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4296 			const compat_stack_t __user *, uss_ptr,
4297 			compat_stack_t __user *, uoss_ptr)
4298 {
4299 	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4300 }
4301 
4302 int compat_restore_altstack(const compat_stack_t __user *uss)
4303 {
4304 	int err = do_compat_sigaltstack(uss, NULL);
4305 	/* squash all but -EFAULT for now */
4306 	return err == -EFAULT ? err : 0;
4307 }
4308 
4309 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4310 {
4311 	int err;
4312 	struct task_struct *t = current;
4313 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4314 			 &uss->ss_sp) |
4315 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4316 		__put_user(t->sas_ss_size, &uss->ss_size);
4317 	return err;
4318 }
4319 #endif
4320 
4321 #ifdef __ARCH_WANT_SYS_SIGPENDING
4322 
4323 /**
4324  *  sys_sigpending - examine pending signals
4325  *  @uset: where mask of pending signal is returned
4326  */
4327 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4328 {
4329 	sigset_t set;
4330 
4331 	if (sizeof(old_sigset_t) > sizeof(*uset))
4332 		return -EINVAL;
4333 
4334 	do_sigpending(&set);
4335 
4336 	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4337 		return -EFAULT;
4338 
4339 	return 0;
4340 }
4341 
4342 #ifdef CONFIG_COMPAT
4343 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4344 {
4345 	sigset_t set;
4346 
4347 	do_sigpending(&set);
4348 
4349 	return put_user(set.sig[0], set32);
4350 }
4351 #endif
4352 
4353 #endif
4354 
4355 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4356 /**
4357  *  sys_sigprocmask - examine and change blocked signals
4358  *  @how: whether to add, remove, or set signals
4359  *  @nset: signals to add or remove (if non-null)
4360  *  @oset: previous value of signal mask if non-null
4361  *
4362  * Some platforms have their own version with special arguments;
4363  * others support only sys_rt_sigprocmask.
4364  */
4365 
4366 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4367 		old_sigset_t __user *, oset)
4368 {
4369 	old_sigset_t old_set, new_set;
4370 	sigset_t new_blocked;
4371 
4372 	old_set = current->blocked.sig[0];
4373 
4374 	if (nset) {
4375 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4376 			return -EFAULT;
4377 
4378 		new_blocked = current->blocked;
4379 
4380 		switch (how) {
4381 		case SIG_BLOCK:
4382 			sigaddsetmask(&new_blocked, new_set);
4383 			break;
4384 		case SIG_UNBLOCK:
4385 			sigdelsetmask(&new_blocked, new_set);
4386 			break;
4387 		case SIG_SETMASK:
4388 			new_blocked.sig[0] = new_set;
4389 			break;
4390 		default:
4391 			return -EINVAL;
4392 		}
4393 
4394 		set_current_blocked(&new_blocked);
4395 	}
4396 
4397 	if (oset) {
4398 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4399 			return -EFAULT;
4400 	}
4401 
4402 	return 0;
4403 }
4404 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4405 
4406 #ifndef CONFIG_ODD_RT_SIGACTION
4407 /**
4408  *  sys_rt_sigaction - alter an action taken by a process
4409  *  @sig: signal to be sent
4410  *  @act: new sigaction
4411  *  @oact: used to save the previous sigaction
4412  *  @sigsetsize: size of sigset_t type
4413  */
4414 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4415 		const struct sigaction __user *, act,
4416 		struct sigaction __user *, oact,
4417 		size_t, sigsetsize)
4418 {
4419 	struct k_sigaction new_sa, old_sa;
4420 	int ret;
4421 
4422 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4423 	if (sigsetsize != sizeof(sigset_t))
4424 		return -EINVAL;
4425 
4426 	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4427 		return -EFAULT;
4428 
4429 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4430 	if (ret)
4431 		return ret;
4432 
4433 	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4434 		return -EFAULT;
4435 
4436 	return 0;
4437 }
4438 #ifdef CONFIG_COMPAT
4439 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4440 		const struct compat_sigaction __user *, act,
4441 		struct compat_sigaction __user *, oact,
4442 		compat_size_t, sigsetsize)
4443 {
4444 	struct k_sigaction new_ka, old_ka;
4445 #ifdef __ARCH_HAS_SA_RESTORER
4446 	compat_uptr_t restorer;
4447 #endif
4448 	int ret;
4449 
4450 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4451 	if (sigsetsize != sizeof(compat_sigset_t))
4452 		return -EINVAL;
4453 
4454 	if (act) {
4455 		compat_uptr_t handler;
4456 		ret = get_user(handler, &act->sa_handler);
4457 		new_ka.sa.sa_handler = compat_ptr(handler);
4458 #ifdef __ARCH_HAS_SA_RESTORER
4459 		ret |= get_user(restorer, &act->sa_restorer);
4460 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4461 #endif
4462 		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4463 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4464 		if (ret)
4465 			return -EFAULT;
4466 	}
4467 
4468 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4469 	if (!ret && oact) {
4470 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4471 			       &oact->sa_handler);
4472 		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4473 					 sizeof(oact->sa_mask));
4474 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4475 #ifdef __ARCH_HAS_SA_RESTORER
4476 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4477 				&oact->sa_restorer);
4478 #endif
4479 	}
4480 	return ret;
4481 }
4482 #endif
4483 #endif /* !CONFIG_ODD_RT_SIGACTION */
4484 
4485 #ifdef CONFIG_OLD_SIGACTION
4486 SYSCALL_DEFINE3(sigaction, int, sig,
4487 		const struct old_sigaction __user *, act,
4488 	        struct old_sigaction __user *, oact)
4489 {
4490 	struct k_sigaction new_ka, old_ka;
4491 	int ret;
4492 
4493 	if (act) {
4494 		old_sigset_t mask;
4495 		if (!access_ok(act, sizeof(*act)) ||
4496 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4497 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4498 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4499 		    __get_user(mask, &act->sa_mask))
4500 			return -EFAULT;
4501 #ifdef __ARCH_HAS_KA_RESTORER
4502 		new_ka.ka_restorer = NULL;
4503 #endif
4504 		siginitset(&new_ka.sa.sa_mask, mask);
4505 	}
4506 
4507 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4508 
4509 	if (!ret && oact) {
4510 		if (!access_ok(oact, sizeof(*oact)) ||
4511 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4512 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4513 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4514 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4515 			return -EFAULT;
4516 	}
4517 
4518 	return ret;
4519 }
4520 #endif
4521 #ifdef CONFIG_COMPAT_OLD_SIGACTION
4522 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4523 		const struct compat_old_sigaction __user *, act,
4524 	        struct compat_old_sigaction __user *, oact)
4525 {
4526 	struct k_sigaction new_ka, old_ka;
4527 	int ret;
4528 	compat_old_sigset_t mask;
4529 	compat_uptr_t handler, restorer;
4530 
4531 	if (act) {
4532 		if (!access_ok(act, sizeof(*act)) ||
4533 		    __get_user(handler, &act->sa_handler) ||
4534 		    __get_user(restorer, &act->sa_restorer) ||
4535 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4536 		    __get_user(mask, &act->sa_mask))
4537 			return -EFAULT;
4538 
4539 #ifdef __ARCH_HAS_KA_RESTORER
4540 		new_ka.ka_restorer = NULL;
4541 #endif
4542 		new_ka.sa.sa_handler = compat_ptr(handler);
4543 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4544 		siginitset(&new_ka.sa.sa_mask, mask);
4545 	}
4546 
4547 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4548 
4549 	if (!ret && oact) {
4550 		if (!access_ok(oact, sizeof(*oact)) ||
4551 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4552 			       &oact->sa_handler) ||
4553 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4554 			       &oact->sa_restorer) ||
4555 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4556 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4557 			return -EFAULT;
4558 	}
4559 	return ret;
4560 }
4561 #endif
4562 
4563 #ifdef CONFIG_SGETMASK_SYSCALL
4564 
4565 /*
4566  * For backwards compatibility.  Functionality superseded by sigprocmask.
4567  */
4568 SYSCALL_DEFINE0(sgetmask)
4569 {
4570 	/* SMP safe */
4571 	return current->blocked.sig[0];
4572 }
4573 
4574 SYSCALL_DEFINE1(ssetmask, int, newmask)
4575 {
4576 	int old = current->blocked.sig[0];
4577 	sigset_t newset;
4578 
4579 	siginitset(&newset, newmask);
4580 	set_current_blocked(&newset);
4581 
4582 	return old;
4583 }
4584 #endif /* CONFIG_SGETMASK_SYSCALL */
4585 
4586 #ifdef __ARCH_WANT_SYS_SIGNAL
4587 /*
4588  * For backwards compatibility.  Functionality superseded by sigaction.
4589  */
4590 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4591 {
4592 	struct k_sigaction new_sa, old_sa;
4593 	int ret;
4594 
4595 	new_sa.sa.sa_handler = handler;
4596 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4597 	sigemptyset(&new_sa.sa.sa_mask);
4598 
4599 	ret = do_sigaction(sig, &new_sa, &old_sa);
4600 
4601 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4602 }
4603 #endif /* __ARCH_WANT_SYS_SIGNAL */
4604 
4605 #ifdef __ARCH_WANT_SYS_PAUSE
4606 
4607 SYSCALL_DEFINE0(pause)
4608 {
4609 	while (!signal_pending(current)) {
4610 		__set_current_state(TASK_INTERRUPTIBLE);
4611 		schedule();
4612 	}
4613 	return -ERESTARTNOHAND;
4614 }
4615 
4616 #endif
4617 
4618 static int sigsuspend(sigset_t *set)
4619 {
4620 	current->saved_sigmask = current->blocked;
4621 	set_current_blocked(set);
4622 
4623 	while (!signal_pending(current)) {
4624 		__set_current_state(TASK_INTERRUPTIBLE);
4625 		schedule();
4626 	}
4627 	set_restore_sigmask();
4628 	return -ERESTARTNOHAND;
4629 }
4630 
4631 /**
4632  *  sys_rt_sigsuspend - replace the signal mask for a value with the
4633  *	@unewset value until a signal is received
4634  *  @unewset: new signal mask value
4635  *  @sigsetsize: size of sigset_t type
4636  */
4637 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4638 {
4639 	sigset_t newset;
4640 
4641 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4642 	if (sigsetsize != sizeof(sigset_t))
4643 		return -EINVAL;
4644 
4645 	if (copy_from_user(&newset, unewset, sizeof(newset)))
4646 		return -EFAULT;
4647 	return sigsuspend(&newset);
4648 }
4649 
4650 #ifdef CONFIG_COMPAT
4651 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4652 {
4653 	sigset_t newset;
4654 
4655 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4656 	if (sigsetsize != sizeof(sigset_t))
4657 		return -EINVAL;
4658 
4659 	if (get_compat_sigset(&newset, unewset))
4660 		return -EFAULT;
4661 	return sigsuspend(&newset);
4662 }
4663 #endif
4664 
4665 #ifdef CONFIG_OLD_SIGSUSPEND
4666 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4667 {
4668 	sigset_t blocked;
4669 	siginitset(&blocked, mask);
4670 	return sigsuspend(&blocked);
4671 }
4672 #endif
4673 #ifdef CONFIG_OLD_SIGSUSPEND3
4674 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4675 {
4676 	sigset_t blocked;
4677 	siginitset(&blocked, mask);
4678 	return sigsuspend(&blocked);
4679 }
4680 #endif
4681 
4682 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4683 {
4684 	return NULL;
4685 }
4686 
4687 static inline void siginfo_buildtime_checks(void)
4688 {
4689 	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4690 
4691 	/* Verify the offsets in the two siginfos match */
4692 #define CHECK_OFFSET(field) \
4693 	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4694 
4695 	/* kill */
4696 	CHECK_OFFSET(si_pid);
4697 	CHECK_OFFSET(si_uid);
4698 
4699 	/* timer */
4700 	CHECK_OFFSET(si_tid);
4701 	CHECK_OFFSET(si_overrun);
4702 	CHECK_OFFSET(si_value);
4703 
4704 	/* rt */
4705 	CHECK_OFFSET(si_pid);
4706 	CHECK_OFFSET(si_uid);
4707 	CHECK_OFFSET(si_value);
4708 
4709 	/* sigchld */
4710 	CHECK_OFFSET(si_pid);
4711 	CHECK_OFFSET(si_uid);
4712 	CHECK_OFFSET(si_status);
4713 	CHECK_OFFSET(si_utime);
4714 	CHECK_OFFSET(si_stime);
4715 
4716 	/* sigfault */
4717 	CHECK_OFFSET(si_addr);
4718 	CHECK_OFFSET(si_trapno);
4719 	CHECK_OFFSET(si_addr_lsb);
4720 	CHECK_OFFSET(si_lower);
4721 	CHECK_OFFSET(si_upper);
4722 	CHECK_OFFSET(si_pkey);
4723 	CHECK_OFFSET(si_perf_data);
4724 	CHECK_OFFSET(si_perf_type);
4725 
4726 	/* sigpoll */
4727 	CHECK_OFFSET(si_band);
4728 	CHECK_OFFSET(si_fd);
4729 
4730 	/* sigsys */
4731 	CHECK_OFFSET(si_call_addr);
4732 	CHECK_OFFSET(si_syscall);
4733 	CHECK_OFFSET(si_arch);
4734 #undef CHECK_OFFSET
4735 
4736 	/* usb asyncio */
4737 	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4738 		     offsetof(struct siginfo, si_addr));
4739 	if (sizeof(int) == sizeof(void __user *)) {
4740 		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4741 			     sizeof(void __user *));
4742 	} else {
4743 		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4744 			      sizeof_field(struct siginfo, si_uid)) !=
4745 			     sizeof(void __user *));
4746 		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4747 			     offsetof(struct siginfo, si_uid));
4748 	}
4749 #ifdef CONFIG_COMPAT
4750 	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4751 		     offsetof(struct compat_siginfo, si_addr));
4752 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4753 		     sizeof(compat_uptr_t));
4754 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4755 		     sizeof_field(struct siginfo, si_pid));
4756 #endif
4757 }
4758 
4759 void __init signals_init(void)
4760 {
4761 	siginfo_buildtime_checks();
4762 
4763 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
4764 }
4765 
4766 #ifdef CONFIG_KGDB_KDB
4767 #include <linux/kdb.h>
4768 /*
4769  * kdb_send_sig - Allows kdb to send signals without exposing
4770  * signal internals.  This function checks if the required locks are
4771  * available before calling the main signal code, to avoid kdb
4772  * deadlocks.
4773  */
4774 void kdb_send_sig(struct task_struct *t, int sig)
4775 {
4776 	static struct task_struct *kdb_prev_t;
4777 	int new_t, ret;
4778 	if (!spin_trylock(&t->sighand->siglock)) {
4779 		kdb_printf("Can't do kill command now.\n"
4780 			   "The sigmask lock is held somewhere else in "
4781 			   "kernel, try again later\n");
4782 		return;
4783 	}
4784 	new_t = kdb_prev_t != t;
4785 	kdb_prev_t = t;
4786 	if (!task_is_running(t) && new_t) {
4787 		spin_unlock(&t->sighand->siglock);
4788 		kdb_printf("Process is not RUNNING, sending a signal from "
4789 			   "kdb risks deadlock\n"
4790 			   "on the run queue locks. "
4791 			   "The signal has _not_ been sent.\n"
4792 			   "Reissue the kill command if you want to risk "
4793 			   "the deadlock.\n");
4794 		return;
4795 	}
4796 	ret = send_signal(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4797 	spin_unlock(&t->sighand->siglock);
4798 	if (ret)
4799 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
4800 			   sig, t->pid);
4801 	else
4802 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4803 }
4804 #endif	/* CONFIG_KGDB_KDB */
4805