xref: /openbmc/linux/kernel/signal.c (revision d15cb3da)
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/tracehook.h>
36 #include <linux/capability.h>
37 #include <linux/freezer.h>
38 #include <linux/pid_namespace.h>
39 #include <linux/nsproxy.h>
40 #include <linux/user_namespace.h>
41 #include <linux/uprobes.h>
42 #include <linux/compat.h>
43 #include <linux/cn_proc.h>
44 #include <linux/compiler.h>
45 #include <linux/posix-timers.h>
46 #include <linux/cgroup.h>
47 #include <linux/audit.h>
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  * If we actually decide not to stop at all because the tracer
2193  * is gone, we keep current->exit_code unless clear_code.
2194  */
2195 static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t *info)
2196 	__releases(&current->sighand->siglock)
2197 	__acquires(&current->sighand->siglock)
2198 {
2199 	bool gstop_done = false;
2200 
2201 	if (arch_ptrace_stop_needed()) {
2202 		/*
2203 		 * The arch code has something special to do before a
2204 		 * ptrace stop.  This is allowed to block, e.g. for faults
2205 		 * on user stack pages.  We can't keep the siglock while
2206 		 * calling arch_ptrace_stop, so we must release it now.
2207 		 * To preserve proper semantics, we must do this before
2208 		 * any signal bookkeeping like checking group_stop_count.
2209 		 */
2210 		spin_unlock_irq(&current->sighand->siglock);
2211 		arch_ptrace_stop();
2212 		spin_lock_irq(&current->sighand->siglock);
2213 	}
2214 
2215 	/*
2216 	 * schedule() will not sleep if there is a pending signal that
2217 	 * can awaken the task.
2218 	 */
2219 	set_special_state(TASK_TRACED);
2220 
2221 	/*
2222 	 * We're committing to trapping.  TRACED should be visible before
2223 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2224 	 * Also, transition to TRACED and updates to ->jobctl should be
2225 	 * atomic with respect to siglock and should be done after the arch
2226 	 * hook as siglock is released and regrabbed across it.
2227 	 *
2228 	 *     TRACER				    TRACEE
2229 	 *
2230 	 *     ptrace_attach()
2231 	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2232 	 *     do_wait()
2233 	 *       set_current_state()                smp_wmb();
2234 	 *       ptrace_do_wait()
2235 	 *         wait_task_stopped()
2236 	 *           task_stopped_code()
2237 	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2238 	 */
2239 	smp_wmb();
2240 
2241 	current->last_siginfo = info;
2242 	current->exit_code = exit_code;
2243 
2244 	/*
2245 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2246 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2247 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2248 	 * could be clear now.  We act as if SIGCONT is received after
2249 	 * TASK_TRACED is entered - ignore it.
2250 	 */
2251 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2252 		gstop_done = task_participate_group_stop(current);
2253 
2254 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2255 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2256 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2257 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2258 
2259 	/* entering a trap, clear TRAPPING */
2260 	task_clear_jobctl_trapping(current);
2261 
2262 	spin_unlock_irq(&current->sighand->siglock);
2263 	read_lock(&tasklist_lock);
2264 	if (likely(current->ptrace)) {
2265 		/*
2266 		 * Notify parents of the stop.
2267 		 *
2268 		 * While ptraced, there are two parents - the ptracer and
2269 		 * the real_parent of the group_leader.  The ptracer should
2270 		 * know about every stop while the real parent is only
2271 		 * interested in the completion of group stop.  The states
2272 		 * for the two don't interact with each other.  Notify
2273 		 * separately unless they're gonna be duplicates.
2274 		 */
2275 		do_notify_parent_cldstop(current, true, why);
2276 		if (gstop_done && ptrace_reparented(current))
2277 			do_notify_parent_cldstop(current, false, why);
2278 
2279 		/*
2280 		 * Don't want to allow preemption here, because
2281 		 * sys_ptrace() needs this task to be inactive.
2282 		 *
2283 		 * XXX: implement read_unlock_no_resched().
2284 		 */
2285 		preempt_disable();
2286 		read_unlock(&tasklist_lock);
2287 		cgroup_enter_frozen();
2288 		preempt_enable_no_resched();
2289 		freezable_schedule();
2290 		cgroup_leave_frozen(true);
2291 	} else {
2292 		/*
2293 		 * By the time we got the lock, our tracer went away.
2294 		 * Don't drop the lock yet, another tracer may come.
2295 		 *
2296 		 * If @gstop_done, the ptracer went away between group stop
2297 		 * completion and here.  During detach, it would have set
2298 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
2299 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
2300 		 * the real parent of the group stop completion is enough.
2301 		 */
2302 		if (gstop_done)
2303 			do_notify_parent_cldstop(current, false, why);
2304 
2305 		/* tasklist protects us from ptrace_freeze_traced() */
2306 		__set_current_state(TASK_RUNNING);
2307 		if (clear_code)
2308 			current->exit_code = 0;
2309 		read_unlock(&tasklist_lock);
2310 	}
2311 
2312 	/*
2313 	 * We are back.  Now reacquire the siglock before touching
2314 	 * last_siginfo, so that we are sure to have synchronized with
2315 	 * any signal-sending on another CPU that wants to examine it.
2316 	 */
2317 	spin_lock_irq(&current->sighand->siglock);
2318 	current->last_siginfo = NULL;
2319 
2320 	/* LISTENING can be set only during STOP traps, clear it */
2321 	current->jobctl &= ~JOBCTL_LISTENING;
2322 
2323 	/*
2324 	 * Queued signals ignored us while we were stopped for tracing.
2325 	 * So check for any that we should take before resuming user mode.
2326 	 * This sets TIF_SIGPENDING, but never clears it.
2327 	 */
2328 	recalc_sigpending_tsk(current);
2329 }
2330 
2331 static void ptrace_do_notify(int signr, int exit_code, int why)
2332 {
2333 	kernel_siginfo_t info;
2334 
2335 	clear_siginfo(&info);
2336 	info.si_signo = signr;
2337 	info.si_code = exit_code;
2338 	info.si_pid = task_pid_vnr(current);
2339 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2340 
2341 	/* Let the debugger run.  */
2342 	ptrace_stop(exit_code, why, 1, &info);
2343 }
2344 
2345 void ptrace_notify(int exit_code)
2346 {
2347 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2348 	if (unlikely(current->task_works))
2349 		task_work_run();
2350 
2351 	spin_lock_irq(&current->sighand->siglock);
2352 	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
2353 	spin_unlock_irq(&current->sighand->siglock);
2354 }
2355 
2356 /**
2357  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2358  * @signr: signr causing group stop if initiating
2359  *
2360  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2361  * and participate in it.  If already set, participate in the existing
2362  * group stop.  If participated in a group stop (and thus slept), %true is
2363  * returned with siglock released.
2364  *
2365  * If ptraced, this function doesn't handle stop itself.  Instead,
2366  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2367  * untouched.  The caller must ensure that INTERRUPT trap handling takes
2368  * places afterwards.
2369  *
2370  * CONTEXT:
2371  * Must be called with @current->sighand->siglock held, which is released
2372  * on %true return.
2373  *
2374  * RETURNS:
2375  * %false if group stop is already cancelled or ptrace trap is scheduled.
2376  * %true if participated in group stop.
2377  */
2378 static bool do_signal_stop(int signr)
2379 	__releases(&current->sighand->siglock)
2380 {
2381 	struct signal_struct *sig = current->signal;
2382 
2383 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2384 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2385 		struct task_struct *t;
2386 
2387 		/* signr will be recorded in task->jobctl for retries */
2388 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2389 
2390 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2391 		    unlikely(sig->flags & SIGNAL_GROUP_EXIT) ||
2392 		    unlikely(sig->group_exec_task))
2393 			return false;
2394 		/*
2395 		 * There is no group stop already in progress.  We must
2396 		 * initiate one now.
2397 		 *
2398 		 * While ptraced, a task may be resumed while group stop is
2399 		 * still in effect and then receive a stop signal and
2400 		 * initiate another group stop.  This deviates from the
2401 		 * usual behavior as two consecutive stop signals can't
2402 		 * cause two group stops when !ptraced.  That is why we
2403 		 * also check !task_is_stopped(t) below.
2404 		 *
2405 		 * The condition can be distinguished by testing whether
2406 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2407 		 * group_exit_code in such case.
2408 		 *
2409 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2410 		 * an intervening stop signal is required to cause two
2411 		 * continued events regardless of ptrace.
2412 		 */
2413 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2414 			sig->group_exit_code = signr;
2415 
2416 		sig->group_stop_count = 0;
2417 
2418 		if (task_set_jobctl_pending(current, signr | gstop))
2419 			sig->group_stop_count++;
2420 
2421 		t = current;
2422 		while_each_thread(current, t) {
2423 			/*
2424 			 * Setting state to TASK_STOPPED for a group
2425 			 * stop is always done with the siglock held,
2426 			 * so this check has no races.
2427 			 */
2428 			if (!task_is_stopped(t) &&
2429 			    task_set_jobctl_pending(t, signr | gstop)) {
2430 				sig->group_stop_count++;
2431 				if (likely(!(t->ptrace & PT_SEIZED)))
2432 					signal_wake_up(t, 0);
2433 				else
2434 					ptrace_trap_notify(t);
2435 			}
2436 		}
2437 	}
2438 
2439 	if (likely(!current->ptrace)) {
2440 		int notify = 0;
2441 
2442 		/*
2443 		 * If there are no other threads in the group, or if there
2444 		 * is a group stop in progress and we are the last to stop,
2445 		 * report to the parent.
2446 		 */
2447 		if (task_participate_group_stop(current))
2448 			notify = CLD_STOPPED;
2449 
2450 		set_special_state(TASK_STOPPED);
2451 		spin_unlock_irq(&current->sighand->siglock);
2452 
2453 		/*
2454 		 * Notify the parent of the group stop completion.  Because
2455 		 * we're not holding either the siglock or tasklist_lock
2456 		 * here, ptracer may attach inbetween; however, this is for
2457 		 * group stop and should always be delivered to the real
2458 		 * parent of the group leader.  The new ptracer will get
2459 		 * its notification when this task transitions into
2460 		 * TASK_TRACED.
2461 		 */
2462 		if (notify) {
2463 			read_lock(&tasklist_lock);
2464 			do_notify_parent_cldstop(current, false, notify);
2465 			read_unlock(&tasklist_lock);
2466 		}
2467 
2468 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2469 		cgroup_enter_frozen();
2470 		freezable_schedule();
2471 		return true;
2472 	} else {
2473 		/*
2474 		 * While ptraced, group stop is handled by STOP trap.
2475 		 * Schedule it and let the caller deal with it.
2476 		 */
2477 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2478 		return false;
2479 	}
2480 }
2481 
2482 /**
2483  * do_jobctl_trap - take care of ptrace jobctl traps
2484  *
2485  * When PT_SEIZED, it's used for both group stop and explicit
2486  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2487  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2488  * the stop signal; otherwise, %SIGTRAP.
2489  *
2490  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2491  * number as exit_code and no siginfo.
2492  *
2493  * CONTEXT:
2494  * Must be called with @current->sighand->siglock held, which may be
2495  * released and re-acquired before returning with intervening sleep.
2496  */
2497 static void do_jobctl_trap(void)
2498 {
2499 	struct signal_struct *signal = current->signal;
2500 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2501 
2502 	if (current->ptrace & PT_SEIZED) {
2503 		if (!signal->group_stop_count &&
2504 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2505 			signr = SIGTRAP;
2506 		WARN_ON_ONCE(!signr);
2507 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2508 				 CLD_STOPPED);
2509 	} else {
2510 		WARN_ON_ONCE(!signr);
2511 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2512 		current->exit_code = 0;
2513 	}
2514 }
2515 
2516 /**
2517  * do_freezer_trap - handle the freezer jobctl trap
2518  *
2519  * Puts the task into frozen state, if only the task is not about to quit.
2520  * In this case it drops JOBCTL_TRAP_FREEZE.
2521  *
2522  * CONTEXT:
2523  * Must be called with @current->sighand->siglock held,
2524  * which is always released before returning.
2525  */
2526 static void do_freezer_trap(void)
2527 	__releases(&current->sighand->siglock)
2528 {
2529 	/*
2530 	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2531 	 * let's make another loop to give it a chance to be handled.
2532 	 * In any case, we'll return back.
2533 	 */
2534 	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2535 	     JOBCTL_TRAP_FREEZE) {
2536 		spin_unlock_irq(&current->sighand->siglock);
2537 		return;
2538 	}
2539 
2540 	/*
2541 	 * Now we're sure that there is no pending fatal signal and no
2542 	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2543 	 * immediately (if there is a non-fatal signal pending), and
2544 	 * put the task into sleep.
2545 	 */
2546 	__set_current_state(TASK_INTERRUPTIBLE);
2547 	clear_thread_flag(TIF_SIGPENDING);
2548 	spin_unlock_irq(&current->sighand->siglock);
2549 	cgroup_enter_frozen();
2550 	freezable_schedule();
2551 }
2552 
2553 static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
2554 {
2555 	/*
2556 	 * We do not check sig_kernel_stop(signr) but set this marker
2557 	 * unconditionally because we do not know whether debugger will
2558 	 * change signr. This flag has no meaning unless we are going
2559 	 * to stop after return from ptrace_stop(). In this case it will
2560 	 * be checked in do_signal_stop(), we should only stop if it was
2561 	 * not cleared by SIGCONT while we were sleeping. See also the
2562 	 * comment in dequeue_signal().
2563 	 */
2564 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2565 	ptrace_stop(signr, CLD_TRAPPED, 0, info);
2566 
2567 	/* We're back.  Did the debugger cancel the sig?  */
2568 	signr = current->exit_code;
2569 	if (signr == 0)
2570 		return signr;
2571 
2572 	current->exit_code = 0;
2573 
2574 	/*
2575 	 * Update the siginfo structure if the signal has
2576 	 * changed.  If the debugger wanted something
2577 	 * specific in the siginfo structure then it should
2578 	 * have updated *info via PTRACE_SETSIGINFO.
2579 	 */
2580 	if (signr != info->si_signo) {
2581 		clear_siginfo(info);
2582 		info->si_signo = signr;
2583 		info->si_errno = 0;
2584 		info->si_code = SI_USER;
2585 		rcu_read_lock();
2586 		info->si_pid = task_pid_vnr(current->parent);
2587 		info->si_uid = from_kuid_munged(current_user_ns(),
2588 						task_uid(current->parent));
2589 		rcu_read_unlock();
2590 	}
2591 
2592 	/* If the (new) signal is now blocked, requeue it.  */
2593 	if (sigismember(&current->blocked, signr) ||
2594 	    fatal_signal_pending(current)) {
2595 		send_signal(signr, info, current, type);
2596 		signr = 0;
2597 	}
2598 
2599 	return signr;
2600 }
2601 
2602 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2603 {
2604 	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2605 	case SIL_FAULT:
2606 	case SIL_FAULT_TRAPNO:
2607 	case SIL_FAULT_MCEERR:
2608 	case SIL_FAULT_BNDERR:
2609 	case SIL_FAULT_PKUERR:
2610 	case SIL_FAULT_PERF_EVENT:
2611 		ksig->info.si_addr = arch_untagged_si_addr(
2612 			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2613 		break;
2614 	case SIL_KILL:
2615 	case SIL_TIMER:
2616 	case SIL_POLL:
2617 	case SIL_CHLD:
2618 	case SIL_RT:
2619 	case SIL_SYS:
2620 		break;
2621 	}
2622 }
2623 
2624 bool get_signal(struct ksignal *ksig)
2625 {
2626 	struct sighand_struct *sighand = current->sighand;
2627 	struct signal_struct *signal = current->signal;
2628 	int signr;
2629 
2630 	if (unlikely(current->task_works))
2631 		task_work_run();
2632 
2633 	/*
2634 	 * For non-generic architectures, check for TIF_NOTIFY_SIGNAL so
2635 	 * that the arch handlers don't all have to do it. If we get here
2636 	 * without TIF_SIGPENDING, just exit after running signal work.
2637 	 */
2638 	if (!IS_ENABLED(CONFIG_GENERIC_ENTRY)) {
2639 		if (test_thread_flag(TIF_NOTIFY_SIGNAL))
2640 			tracehook_notify_signal();
2641 		if (!task_sigpending(current))
2642 			return false;
2643 	}
2644 
2645 	if (unlikely(uprobe_deny_signal()))
2646 		return false;
2647 
2648 	/*
2649 	 * Do this once, we can't return to user-mode if freezing() == T.
2650 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2651 	 * thus do not need another check after return.
2652 	 */
2653 	try_to_freeze();
2654 
2655 relock:
2656 	spin_lock_irq(&sighand->siglock);
2657 
2658 	/*
2659 	 * Every stopped thread goes here after wakeup. Check to see if
2660 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2661 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2662 	 */
2663 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2664 		int why;
2665 
2666 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2667 			why = CLD_CONTINUED;
2668 		else
2669 			why = CLD_STOPPED;
2670 
2671 		signal->flags &= ~SIGNAL_CLD_MASK;
2672 
2673 		spin_unlock_irq(&sighand->siglock);
2674 
2675 		/*
2676 		 * Notify the parent that we're continuing.  This event is
2677 		 * always per-process and doesn't make whole lot of sense
2678 		 * for ptracers, who shouldn't consume the state via
2679 		 * wait(2) either, but, for backward compatibility, notify
2680 		 * the ptracer of the group leader too unless it's gonna be
2681 		 * a duplicate.
2682 		 */
2683 		read_lock(&tasklist_lock);
2684 		do_notify_parent_cldstop(current, false, why);
2685 
2686 		if (ptrace_reparented(current->group_leader))
2687 			do_notify_parent_cldstop(current->group_leader,
2688 						true, why);
2689 		read_unlock(&tasklist_lock);
2690 
2691 		goto relock;
2692 	}
2693 
2694 	for (;;) {
2695 		struct k_sigaction *ka;
2696 		enum pid_type type;
2697 
2698 		/* Has this task already been marked for death? */
2699 		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
2700 		     signal->group_exec_task) {
2701 			ksig->info.si_signo = signr = SIGKILL;
2702 			sigdelset(&current->pending.signal, SIGKILL);
2703 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2704 				&sighand->action[SIGKILL - 1]);
2705 			recalc_sigpending();
2706 			goto fatal;
2707 		}
2708 
2709 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2710 		    do_signal_stop(0))
2711 			goto relock;
2712 
2713 		if (unlikely(current->jobctl &
2714 			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2715 			if (current->jobctl & JOBCTL_TRAP_MASK) {
2716 				do_jobctl_trap();
2717 				spin_unlock_irq(&sighand->siglock);
2718 			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2719 				do_freezer_trap();
2720 
2721 			goto relock;
2722 		}
2723 
2724 		/*
2725 		 * If the task is leaving the frozen state, let's update
2726 		 * cgroup counters and reset the frozen bit.
2727 		 */
2728 		if (unlikely(cgroup_task_frozen(current))) {
2729 			spin_unlock_irq(&sighand->siglock);
2730 			cgroup_leave_frozen(false);
2731 			goto relock;
2732 		}
2733 
2734 		/*
2735 		 * Signals generated by the execution of an instruction
2736 		 * need to be delivered before any other pending signals
2737 		 * so that the instruction pointer in the signal stack
2738 		 * frame points to the faulting instruction.
2739 		 */
2740 		type = PIDTYPE_PID;
2741 		signr = dequeue_synchronous_signal(&ksig->info);
2742 		if (!signr)
2743 			signr = dequeue_signal(current, &current->blocked,
2744 					       &ksig->info, &type);
2745 
2746 		if (!signr)
2747 			break; /* will return 0 */
2748 
2749 		if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2750 		    !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2751 			signr = ptrace_signal(signr, &ksig->info, type);
2752 			if (!signr)
2753 				continue;
2754 		}
2755 
2756 		ka = &sighand->action[signr-1];
2757 
2758 		/* Trace actually delivered signals. */
2759 		trace_signal_deliver(signr, &ksig->info, ka);
2760 
2761 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2762 			continue;
2763 		if (ka->sa.sa_handler != SIG_DFL) {
2764 			/* Run the handler.  */
2765 			ksig->ka = *ka;
2766 
2767 			if (ka->sa.sa_flags & SA_ONESHOT)
2768 				ka->sa.sa_handler = SIG_DFL;
2769 
2770 			break; /* will return non-zero "signr" value */
2771 		}
2772 
2773 		/*
2774 		 * Now we are doing the default action for this signal.
2775 		 */
2776 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2777 			continue;
2778 
2779 		/*
2780 		 * Global init gets no signals it doesn't want.
2781 		 * Container-init gets no signals it doesn't want from same
2782 		 * container.
2783 		 *
2784 		 * Note that if global/container-init sees a sig_kernel_only()
2785 		 * signal here, the signal must have been generated internally
2786 		 * or must have come from an ancestor namespace. In either
2787 		 * case, the signal cannot be dropped.
2788 		 */
2789 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2790 				!sig_kernel_only(signr))
2791 			continue;
2792 
2793 		if (sig_kernel_stop(signr)) {
2794 			/*
2795 			 * The default action is to stop all threads in
2796 			 * the thread group.  The job control signals
2797 			 * do nothing in an orphaned pgrp, but SIGSTOP
2798 			 * always works.  Note that siglock needs to be
2799 			 * dropped during the call to is_orphaned_pgrp()
2800 			 * because of lock ordering with tasklist_lock.
2801 			 * This allows an intervening SIGCONT to be posted.
2802 			 * We need to check for that and bail out if necessary.
2803 			 */
2804 			if (signr != SIGSTOP) {
2805 				spin_unlock_irq(&sighand->siglock);
2806 
2807 				/* signals can be posted during this window */
2808 
2809 				if (is_current_pgrp_orphaned())
2810 					goto relock;
2811 
2812 				spin_lock_irq(&sighand->siglock);
2813 			}
2814 
2815 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2816 				/* It released the siglock.  */
2817 				goto relock;
2818 			}
2819 
2820 			/*
2821 			 * We didn't actually stop, due to a race
2822 			 * with SIGCONT or something like that.
2823 			 */
2824 			continue;
2825 		}
2826 
2827 	fatal:
2828 		spin_unlock_irq(&sighand->siglock);
2829 		if (unlikely(cgroup_task_frozen(current)))
2830 			cgroup_leave_frozen(true);
2831 
2832 		/*
2833 		 * Anything else is fatal, maybe with a core dump.
2834 		 */
2835 		current->flags |= PF_SIGNALED;
2836 
2837 		if (sig_kernel_coredump(signr)) {
2838 			if (print_fatal_signals)
2839 				print_fatal_signal(ksig->info.si_signo);
2840 			proc_coredump_connector(current);
2841 			/*
2842 			 * If it was able to dump core, this kills all
2843 			 * other threads in the group and synchronizes with
2844 			 * their demise.  If we lost the race with another
2845 			 * thread getting here, it set group_exit_code
2846 			 * first and our do_group_exit call below will use
2847 			 * that value and ignore the one we pass it.
2848 			 */
2849 			do_coredump(&ksig->info);
2850 		}
2851 
2852 		/*
2853 		 * PF_IO_WORKER threads will catch and exit on fatal signals
2854 		 * themselves. They have cleanup that must be performed, so
2855 		 * we cannot call do_exit() on their behalf.
2856 		 */
2857 		if (current->flags & PF_IO_WORKER)
2858 			goto out;
2859 
2860 		/*
2861 		 * Death signals, no core dump.
2862 		 */
2863 		do_group_exit(ksig->info.si_signo);
2864 		/* NOTREACHED */
2865 	}
2866 	spin_unlock_irq(&sighand->siglock);
2867 out:
2868 	ksig->sig = signr;
2869 
2870 	if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2871 		hide_si_addr_tag_bits(ksig);
2872 
2873 	return ksig->sig > 0;
2874 }
2875 
2876 /**
2877  * signal_delivered - called after signal delivery to update blocked signals
2878  * @ksig:		kernel signal struct
2879  * @stepping:		nonzero if debugger single-step or block-step in use
2880  *
2881  * This function should be called when a signal has successfully been
2882  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2883  * is always blocked), and the signal itself is blocked unless %SA_NODEFER
2884  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2885  */
2886 static void signal_delivered(struct ksignal *ksig, int stepping)
2887 {
2888 	sigset_t blocked;
2889 
2890 	/* A signal was successfully delivered, and the
2891 	   saved sigmask was stored on the signal frame,
2892 	   and will be restored by sigreturn.  So we can
2893 	   simply clear the restore sigmask flag.  */
2894 	clear_restore_sigmask();
2895 
2896 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2897 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2898 		sigaddset(&blocked, ksig->sig);
2899 	set_current_blocked(&blocked);
2900 	if (current->sas_ss_flags & SS_AUTODISARM)
2901 		sas_ss_reset(current);
2902 	tracehook_signal_handler(stepping);
2903 }
2904 
2905 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2906 {
2907 	if (failed)
2908 		force_sigsegv(ksig->sig);
2909 	else
2910 		signal_delivered(ksig, stepping);
2911 }
2912 
2913 /*
2914  * It could be that complete_signal() picked us to notify about the
2915  * group-wide signal. Other threads should be notified now to take
2916  * the shared signals in @which since we will not.
2917  */
2918 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2919 {
2920 	sigset_t retarget;
2921 	struct task_struct *t;
2922 
2923 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2924 	if (sigisemptyset(&retarget))
2925 		return;
2926 
2927 	t = tsk;
2928 	while_each_thread(tsk, t) {
2929 		if (t->flags & PF_EXITING)
2930 			continue;
2931 
2932 		if (!has_pending_signals(&retarget, &t->blocked))
2933 			continue;
2934 		/* Remove the signals this thread can handle. */
2935 		sigandsets(&retarget, &retarget, &t->blocked);
2936 
2937 		if (!task_sigpending(t))
2938 			signal_wake_up(t, 0);
2939 
2940 		if (sigisemptyset(&retarget))
2941 			break;
2942 	}
2943 }
2944 
2945 void exit_signals(struct task_struct *tsk)
2946 {
2947 	int group_stop = 0;
2948 	sigset_t unblocked;
2949 
2950 	/*
2951 	 * @tsk is about to have PF_EXITING set - lock out users which
2952 	 * expect stable threadgroup.
2953 	 */
2954 	cgroup_threadgroup_change_begin(tsk);
2955 
2956 	if (thread_group_empty(tsk) || (tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
2957 		tsk->flags |= PF_EXITING;
2958 		cgroup_threadgroup_change_end(tsk);
2959 		return;
2960 	}
2961 
2962 	spin_lock_irq(&tsk->sighand->siglock);
2963 	/*
2964 	 * From now this task is not visible for group-wide signals,
2965 	 * see wants_signal(), do_signal_stop().
2966 	 */
2967 	tsk->flags |= PF_EXITING;
2968 
2969 	cgroup_threadgroup_change_end(tsk);
2970 
2971 	if (!task_sigpending(tsk))
2972 		goto out;
2973 
2974 	unblocked = tsk->blocked;
2975 	signotset(&unblocked);
2976 	retarget_shared_pending(tsk, &unblocked);
2977 
2978 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2979 	    task_participate_group_stop(tsk))
2980 		group_stop = CLD_STOPPED;
2981 out:
2982 	spin_unlock_irq(&tsk->sighand->siglock);
2983 
2984 	/*
2985 	 * If group stop has completed, deliver the notification.  This
2986 	 * should always go to the real parent of the group leader.
2987 	 */
2988 	if (unlikely(group_stop)) {
2989 		read_lock(&tasklist_lock);
2990 		do_notify_parent_cldstop(tsk, false, group_stop);
2991 		read_unlock(&tasklist_lock);
2992 	}
2993 }
2994 
2995 /*
2996  * System call entry points.
2997  */
2998 
2999 /**
3000  *  sys_restart_syscall - restart a system call
3001  */
3002 SYSCALL_DEFINE0(restart_syscall)
3003 {
3004 	struct restart_block *restart = &current->restart_block;
3005 	return restart->fn(restart);
3006 }
3007 
3008 long do_no_restart_syscall(struct restart_block *param)
3009 {
3010 	return -EINTR;
3011 }
3012 
3013 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
3014 {
3015 	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3016 		sigset_t newblocked;
3017 		/* A set of now blocked but previously unblocked signals. */
3018 		sigandnsets(&newblocked, newset, &current->blocked);
3019 		retarget_shared_pending(tsk, &newblocked);
3020 	}
3021 	tsk->blocked = *newset;
3022 	recalc_sigpending();
3023 }
3024 
3025 /**
3026  * set_current_blocked - change current->blocked mask
3027  * @newset: new mask
3028  *
3029  * It is wrong to change ->blocked directly, this helper should be used
3030  * to ensure the process can't miss a shared signal we are going to block.
3031  */
3032 void set_current_blocked(sigset_t *newset)
3033 {
3034 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3035 	__set_current_blocked(newset);
3036 }
3037 
3038 void __set_current_blocked(const sigset_t *newset)
3039 {
3040 	struct task_struct *tsk = current;
3041 
3042 	/*
3043 	 * In case the signal mask hasn't changed, there is nothing we need
3044 	 * to do. The current->blocked shouldn't be modified by other task.
3045 	 */
3046 	if (sigequalsets(&tsk->blocked, newset))
3047 		return;
3048 
3049 	spin_lock_irq(&tsk->sighand->siglock);
3050 	__set_task_blocked(tsk, newset);
3051 	spin_unlock_irq(&tsk->sighand->siglock);
3052 }
3053 
3054 /*
3055  * This is also useful for kernel threads that want to temporarily
3056  * (or permanently) block certain signals.
3057  *
3058  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3059  * interface happily blocks "unblockable" signals like SIGKILL
3060  * and friends.
3061  */
3062 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3063 {
3064 	struct task_struct *tsk = current;
3065 	sigset_t newset;
3066 
3067 	/* Lockless, only current can change ->blocked, never from irq */
3068 	if (oldset)
3069 		*oldset = tsk->blocked;
3070 
3071 	switch (how) {
3072 	case SIG_BLOCK:
3073 		sigorsets(&newset, &tsk->blocked, set);
3074 		break;
3075 	case SIG_UNBLOCK:
3076 		sigandnsets(&newset, &tsk->blocked, set);
3077 		break;
3078 	case SIG_SETMASK:
3079 		newset = *set;
3080 		break;
3081 	default:
3082 		return -EINVAL;
3083 	}
3084 
3085 	__set_current_blocked(&newset);
3086 	return 0;
3087 }
3088 EXPORT_SYMBOL(sigprocmask);
3089 
3090 /*
3091  * The api helps set app-provided sigmasks.
3092  *
3093  * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3094  * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3095  *
3096  * Note that it does set_restore_sigmask() in advance, so it must be always
3097  * paired with restore_saved_sigmask_unless() before return from syscall.
3098  */
3099 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3100 {
3101 	sigset_t kmask;
3102 
3103 	if (!umask)
3104 		return 0;
3105 	if (sigsetsize != sizeof(sigset_t))
3106 		return -EINVAL;
3107 	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3108 		return -EFAULT;
3109 
3110 	set_restore_sigmask();
3111 	current->saved_sigmask = current->blocked;
3112 	set_current_blocked(&kmask);
3113 
3114 	return 0;
3115 }
3116 
3117 #ifdef CONFIG_COMPAT
3118 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3119 			    size_t sigsetsize)
3120 {
3121 	sigset_t kmask;
3122 
3123 	if (!umask)
3124 		return 0;
3125 	if (sigsetsize != sizeof(compat_sigset_t))
3126 		return -EINVAL;
3127 	if (get_compat_sigset(&kmask, umask))
3128 		return -EFAULT;
3129 
3130 	set_restore_sigmask();
3131 	current->saved_sigmask = current->blocked;
3132 	set_current_blocked(&kmask);
3133 
3134 	return 0;
3135 }
3136 #endif
3137 
3138 /**
3139  *  sys_rt_sigprocmask - change the list of currently blocked signals
3140  *  @how: whether to add, remove, or set signals
3141  *  @nset: stores pending signals
3142  *  @oset: previous value of signal mask if non-null
3143  *  @sigsetsize: size of sigset_t type
3144  */
3145 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3146 		sigset_t __user *, oset, size_t, sigsetsize)
3147 {
3148 	sigset_t old_set, new_set;
3149 	int error;
3150 
3151 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3152 	if (sigsetsize != sizeof(sigset_t))
3153 		return -EINVAL;
3154 
3155 	old_set = current->blocked;
3156 
3157 	if (nset) {
3158 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3159 			return -EFAULT;
3160 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3161 
3162 		error = sigprocmask(how, &new_set, NULL);
3163 		if (error)
3164 			return error;
3165 	}
3166 
3167 	if (oset) {
3168 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3169 			return -EFAULT;
3170 	}
3171 
3172 	return 0;
3173 }
3174 
3175 #ifdef CONFIG_COMPAT
3176 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3177 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3178 {
3179 	sigset_t old_set = current->blocked;
3180 
3181 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3182 	if (sigsetsize != sizeof(sigset_t))
3183 		return -EINVAL;
3184 
3185 	if (nset) {
3186 		sigset_t new_set;
3187 		int error;
3188 		if (get_compat_sigset(&new_set, nset))
3189 			return -EFAULT;
3190 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3191 
3192 		error = sigprocmask(how, &new_set, NULL);
3193 		if (error)
3194 			return error;
3195 	}
3196 	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3197 }
3198 #endif
3199 
3200 static void do_sigpending(sigset_t *set)
3201 {
3202 	spin_lock_irq(&current->sighand->siglock);
3203 	sigorsets(set, &current->pending.signal,
3204 		  &current->signal->shared_pending.signal);
3205 	spin_unlock_irq(&current->sighand->siglock);
3206 
3207 	/* Outside the lock because only this thread touches it.  */
3208 	sigandsets(set, &current->blocked, set);
3209 }
3210 
3211 /**
3212  *  sys_rt_sigpending - examine a pending signal that has been raised
3213  *			while blocked
3214  *  @uset: stores pending signals
3215  *  @sigsetsize: size of sigset_t type or larger
3216  */
3217 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3218 {
3219 	sigset_t set;
3220 
3221 	if (sigsetsize > sizeof(*uset))
3222 		return -EINVAL;
3223 
3224 	do_sigpending(&set);
3225 
3226 	if (copy_to_user(uset, &set, sigsetsize))
3227 		return -EFAULT;
3228 
3229 	return 0;
3230 }
3231 
3232 #ifdef CONFIG_COMPAT
3233 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3234 		compat_size_t, sigsetsize)
3235 {
3236 	sigset_t set;
3237 
3238 	if (sigsetsize > sizeof(*uset))
3239 		return -EINVAL;
3240 
3241 	do_sigpending(&set);
3242 
3243 	return put_compat_sigset(uset, &set, sigsetsize);
3244 }
3245 #endif
3246 
3247 static const struct {
3248 	unsigned char limit, layout;
3249 } sig_sicodes[] = {
3250 	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3251 	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3252 	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3253 	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3254 	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3255 #if defined(SIGEMT)
3256 	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3257 #endif
3258 	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3259 	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3260 	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3261 };
3262 
3263 static bool known_siginfo_layout(unsigned sig, int si_code)
3264 {
3265 	if (si_code == SI_KERNEL)
3266 		return true;
3267 	else if ((si_code > SI_USER)) {
3268 		if (sig_specific_sicodes(sig)) {
3269 			if (si_code <= sig_sicodes[sig].limit)
3270 				return true;
3271 		}
3272 		else if (si_code <= NSIGPOLL)
3273 			return true;
3274 	}
3275 	else if (si_code >= SI_DETHREAD)
3276 		return true;
3277 	else if (si_code == SI_ASYNCNL)
3278 		return true;
3279 	return false;
3280 }
3281 
3282 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3283 {
3284 	enum siginfo_layout layout = SIL_KILL;
3285 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3286 		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3287 		    (si_code <= sig_sicodes[sig].limit)) {
3288 			layout = sig_sicodes[sig].layout;
3289 			/* Handle the exceptions */
3290 			if ((sig == SIGBUS) &&
3291 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3292 				layout = SIL_FAULT_MCEERR;
3293 			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3294 				layout = SIL_FAULT_BNDERR;
3295 #ifdef SEGV_PKUERR
3296 			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3297 				layout = SIL_FAULT_PKUERR;
3298 #endif
3299 			else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3300 				layout = SIL_FAULT_PERF_EVENT;
3301 			else if (IS_ENABLED(CONFIG_SPARC) &&
3302 				 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3303 				layout = SIL_FAULT_TRAPNO;
3304 			else if (IS_ENABLED(CONFIG_ALPHA) &&
3305 				 ((sig == SIGFPE) ||
3306 				  ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3307 				layout = SIL_FAULT_TRAPNO;
3308 		}
3309 		else if (si_code <= NSIGPOLL)
3310 			layout = SIL_POLL;
3311 	} else {
3312 		if (si_code == SI_TIMER)
3313 			layout = SIL_TIMER;
3314 		else if (si_code == SI_SIGIO)
3315 			layout = SIL_POLL;
3316 		else if (si_code < 0)
3317 			layout = SIL_RT;
3318 	}
3319 	return layout;
3320 }
3321 
3322 static inline char __user *si_expansion(const siginfo_t __user *info)
3323 {
3324 	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3325 }
3326 
3327 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3328 {
3329 	char __user *expansion = si_expansion(to);
3330 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3331 		return -EFAULT;
3332 	if (clear_user(expansion, SI_EXPANSION_SIZE))
3333 		return -EFAULT;
3334 	return 0;
3335 }
3336 
3337 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3338 				       const siginfo_t __user *from)
3339 {
3340 	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3341 		char __user *expansion = si_expansion(from);
3342 		char buf[SI_EXPANSION_SIZE];
3343 		int i;
3344 		/*
3345 		 * An unknown si_code might need more than
3346 		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3347 		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3348 		 * will return this data to userspace exactly.
3349 		 */
3350 		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3351 			return -EFAULT;
3352 		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3353 			if (buf[i] != 0)
3354 				return -E2BIG;
3355 		}
3356 	}
3357 	return 0;
3358 }
3359 
3360 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3361 				    const siginfo_t __user *from)
3362 {
3363 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3364 		return -EFAULT;
3365 	to->si_signo = signo;
3366 	return post_copy_siginfo_from_user(to, from);
3367 }
3368 
3369 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3370 {
3371 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3372 		return -EFAULT;
3373 	return post_copy_siginfo_from_user(to, from);
3374 }
3375 
3376 #ifdef CONFIG_COMPAT
3377 /**
3378  * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3379  * @to: compat siginfo destination
3380  * @from: kernel siginfo source
3381  *
3382  * Note: This function does not work properly for the SIGCHLD on x32, but
3383  * fortunately it doesn't have to.  The only valid callers for this function are
3384  * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3385  * The latter does not care because SIGCHLD will never cause a coredump.
3386  */
3387 void copy_siginfo_to_external32(struct compat_siginfo *to,
3388 		const struct kernel_siginfo *from)
3389 {
3390 	memset(to, 0, sizeof(*to));
3391 
3392 	to->si_signo = from->si_signo;
3393 	to->si_errno = from->si_errno;
3394 	to->si_code  = from->si_code;
3395 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3396 	case SIL_KILL:
3397 		to->si_pid = from->si_pid;
3398 		to->si_uid = from->si_uid;
3399 		break;
3400 	case SIL_TIMER:
3401 		to->si_tid     = from->si_tid;
3402 		to->si_overrun = from->si_overrun;
3403 		to->si_int     = from->si_int;
3404 		break;
3405 	case SIL_POLL:
3406 		to->si_band = from->si_band;
3407 		to->si_fd   = from->si_fd;
3408 		break;
3409 	case SIL_FAULT:
3410 		to->si_addr = ptr_to_compat(from->si_addr);
3411 		break;
3412 	case SIL_FAULT_TRAPNO:
3413 		to->si_addr = ptr_to_compat(from->si_addr);
3414 		to->si_trapno = from->si_trapno;
3415 		break;
3416 	case SIL_FAULT_MCEERR:
3417 		to->si_addr = ptr_to_compat(from->si_addr);
3418 		to->si_addr_lsb = from->si_addr_lsb;
3419 		break;
3420 	case SIL_FAULT_BNDERR:
3421 		to->si_addr = ptr_to_compat(from->si_addr);
3422 		to->si_lower = ptr_to_compat(from->si_lower);
3423 		to->si_upper = ptr_to_compat(from->si_upper);
3424 		break;
3425 	case SIL_FAULT_PKUERR:
3426 		to->si_addr = ptr_to_compat(from->si_addr);
3427 		to->si_pkey = from->si_pkey;
3428 		break;
3429 	case SIL_FAULT_PERF_EVENT:
3430 		to->si_addr = ptr_to_compat(from->si_addr);
3431 		to->si_perf_data = from->si_perf_data;
3432 		to->si_perf_type = from->si_perf_type;
3433 		break;
3434 	case SIL_CHLD:
3435 		to->si_pid = from->si_pid;
3436 		to->si_uid = from->si_uid;
3437 		to->si_status = from->si_status;
3438 		to->si_utime = from->si_utime;
3439 		to->si_stime = from->si_stime;
3440 		break;
3441 	case SIL_RT:
3442 		to->si_pid = from->si_pid;
3443 		to->si_uid = from->si_uid;
3444 		to->si_int = from->si_int;
3445 		break;
3446 	case SIL_SYS:
3447 		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3448 		to->si_syscall   = from->si_syscall;
3449 		to->si_arch      = from->si_arch;
3450 		break;
3451 	}
3452 }
3453 
3454 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3455 			   const struct kernel_siginfo *from)
3456 {
3457 	struct compat_siginfo new;
3458 
3459 	copy_siginfo_to_external32(&new, from);
3460 	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3461 		return -EFAULT;
3462 	return 0;
3463 }
3464 
3465 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3466 					 const struct compat_siginfo *from)
3467 {
3468 	clear_siginfo(to);
3469 	to->si_signo = from->si_signo;
3470 	to->si_errno = from->si_errno;
3471 	to->si_code  = from->si_code;
3472 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3473 	case SIL_KILL:
3474 		to->si_pid = from->si_pid;
3475 		to->si_uid = from->si_uid;
3476 		break;
3477 	case SIL_TIMER:
3478 		to->si_tid     = from->si_tid;
3479 		to->si_overrun = from->si_overrun;
3480 		to->si_int     = from->si_int;
3481 		break;
3482 	case SIL_POLL:
3483 		to->si_band = from->si_band;
3484 		to->si_fd   = from->si_fd;
3485 		break;
3486 	case SIL_FAULT:
3487 		to->si_addr = compat_ptr(from->si_addr);
3488 		break;
3489 	case SIL_FAULT_TRAPNO:
3490 		to->si_addr = compat_ptr(from->si_addr);
3491 		to->si_trapno = from->si_trapno;
3492 		break;
3493 	case SIL_FAULT_MCEERR:
3494 		to->si_addr = compat_ptr(from->si_addr);
3495 		to->si_addr_lsb = from->si_addr_lsb;
3496 		break;
3497 	case SIL_FAULT_BNDERR:
3498 		to->si_addr = compat_ptr(from->si_addr);
3499 		to->si_lower = compat_ptr(from->si_lower);
3500 		to->si_upper = compat_ptr(from->si_upper);
3501 		break;
3502 	case SIL_FAULT_PKUERR:
3503 		to->si_addr = compat_ptr(from->si_addr);
3504 		to->si_pkey = from->si_pkey;
3505 		break;
3506 	case SIL_FAULT_PERF_EVENT:
3507 		to->si_addr = compat_ptr(from->si_addr);
3508 		to->si_perf_data = from->si_perf_data;
3509 		to->si_perf_type = from->si_perf_type;
3510 		break;
3511 	case SIL_CHLD:
3512 		to->si_pid    = from->si_pid;
3513 		to->si_uid    = from->si_uid;
3514 		to->si_status = from->si_status;
3515 #ifdef CONFIG_X86_X32_ABI
3516 		if (in_x32_syscall()) {
3517 			to->si_utime = from->_sifields._sigchld_x32._utime;
3518 			to->si_stime = from->_sifields._sigchld_x32._stime;
3519 		} else
3520 #endif
3521 		{
3522 			to->si_utime = from->si_utime;
3523 			to->si_stime = from->si_stime;
3524 		}
3525 		break;
3526 	case SIL_RT:
3527 		to->si_pid = from->si_pid;
3528 		to->si_uid = from->si_uid;
3529 		to->si_int = from->si_int;
3530 		break;
3531 	case SIL_SYS:
3532 		to->si_call_addr = compat_ptr(from->si_call_addr);
3533 		to->si_syscall   = from->si_syscall;
3534 		to->si_arch      = from->si_arch;
3535 		break;
3536 	}
3537 	return 0;
3538 }
3539 
3540 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3541 				      const struct compat_siginfo __user *ufrom)
3542 {
3543 	struct compat_siginfo from;
3544 
3545 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3546 		return -EFAULT;
3547 
3548 	from.si_signo = signo;
3549 	return post_copy_siginfo_from_user32(to, &from);
3550 }
3551 
3552 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3553 			     const struct compat_siginfo __user *ufrom)
3554 {
3555 	struct compat_siginfo from;
3556 
3557 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3558 		return -EFAULT;
3559 
3560 	return post_copy_siginfo_from_user32(to, &from);
3561 }
3562 #endif /* CONFIG_COMPAT */
3563 
3564 /**
3565  *  do_sigtimedwait - wait for queued signals specified in @which
3566  *  @which: queued signals to wait for
3567  *  @info: if non-null, the signal's siginfo is returned here
3568  *  @ts: upper bound on process time suspension
3569  */
3570 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3571 		    const struct timespec64 *ts)
3572 {
3573 	ktime_t *to = NULL, timeout = KTIME_MAX;
3574 	struct task_struct *tsk = current;
3575 	sigset_t mask = *which;
3576 	enum pid_type type;
3577 	int sig, ret = 0;
3578 
3579 	if (ts) {
3580 		if (!timespec64_valid(ts))
3581 			return -EINVAL;
3582 		timeout = timespec64_to_ktime(*ts);
3583 		to = &timeout;
3584 	}
3585 
3586 	/*
3587 	 * Invert the set of allowed signals to get those we want to block.
3588 	 */
3589 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3590 	signotset(&mask);
3591 
3592 	spin_lock_irq(&tsk->sighand->siglock);
3593 	sig = dequeue_signal(tsk, &mask, info, &type);
3594 	if (!sig && timeout) {
3595 		/*
3596 		 * None ready, temporarily unblock those we're interested
3597 		 * while we are sleeping in so that we'll be awakened when
3598 		 * they arrive. Unblocking is always fine, we can avoid
3599 		 * set_current_blocked().
3600 		 */
3601 		tsk->real_blocked = tsk->blocked;
3602 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3603 		recalc_sigpending();
3604 		spin_unlock_irq(&tsk->sighand->siglock);
3605 
3606 		__set_current_state(TASK_INTERRUPTIBLE);
3607 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3608 							 HRTIMER_MODE_REL);
3609 		spin_lock_irq(&tsk->sighand->siglock);
3610 		__set_task_blocked(tsk, &tsk->real_blocked);
3611 		sigemptyset(&tsk->real_blocked);
3612 		sig = dequeue_signal(tsk, &mask, info, &type);
3613 	}
3614 	spin_unlock_irq(&tsk->sighand->siglock);
3615 
3616 	if (sig)
3617 		return sig;
3618 	return ret ? -EINTR : -EAGAIN;
3619 }
3620 
3621 /**
3622  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3623  *			in @uthese
3624  *  @uthese: queued signals to wait for
3625  *  @uinfo: if non-null, the signal's siginfo is returned here
3626  *  @uts: upper bound on process time suspension
3627  *  @sigsetsize: size of sigset_t type
3628  */
3629 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3630 		siginfo_t __user *, uinfo,
3631 		const struct __kernel_timespec __user *, uts,
3632 		size_t, sigsetsize)
3633 {
3634 	sigset_t these;
3635 	struct timespec64 ts;
3636 	kernel_siginfo_t info;
3637 	int ret;
3638 
3639 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3640 	if (sigsetsize != sizeof(sigset_t))
3641 		return -EINVAL;
3642 
3643 	if (copy_from_user(&these, uthese, sizeof(these)))
3644 		return -EFAULT;
3645 
3646 	if (uts) {
3647 		if (get_timespec64(&ts, uts))
3648 			return -EFAULT;
3649 	}
3650 
3651 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3652 
3653 	if (ret > 0 && uinfo) {
3654 		if (copy_siginfo_to_user(uinfo, &info))
3655 			ret = -EFAULT;
3656 	}
3657 
3658 	return ret;
3659 }
3660 
3661 #ifdef CONFIG_COMPAT_32BIT_TIME
3662 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3663 		siginfo_t __user *, uinfo,
3664 		const struct old_timespec32 __user *, uts,
3665 		size_t, sigsetsize)
3666 {
3667 	sigset_t these;
3668 	struct timespec64 ts;
3669 	kernel_siginfo_t info;
3670 	int ret;
3671 
3672 	if (sigsetsize != sizeof(sigset_t))
3673 		return -EINVAL;
3674 
3675 	if (copy_from_user(&these, uthese, sizeof(these)))
3676 		return -EFAULT;
3677 
3678 	if (uts) {
3679 		if (get_old_timespec32(&ts, uts))
3680 			return -EFAULT;
3681 	}
3682 
3683 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3684 
3685 	if (ret > 0 && uinfo) {
3686 		if (copy_siginfo_to_user(uinfo, &info))
3687 			ret = -EFAULT;
3688 	}
3689 
3690 	return ret;
3691 }
3692 #endif
3693 
3694 #ifdef CONFIG_COMPAT
3695 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3696 		struct compat_siginfo __user *, uinfo,
3697 		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3698 {
3699 	sigset_t s;
3700 	struct timespec64 t;
3701 	kernel_siginfo_t info;
3702 	long ret;
3703 
3704 	if (sigsetsize != sizeof(sigset_t))
3705 		return -EINVAL;
3706 
3707 	if (get_compat_sigset(&s, uthese))
3708 		return -EFAULT;
3709 
3710 	if (uts) {
3711 		if (get_timespec64(&t, uts))
3712 			return -EFAULT;
3713 	}
3714 
3715 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3716 
3717 	if (ret > 0 && uinfo) {
3718 		if (copy_siginfo_to_user32(uinfo, &info))
3719 			ret = -EFAULT;
3720 	}
3721 
3722 	return ret;
3723 }
3724 
3725 #ifdef CONFIG_COMPAT_32BIT_TIME
3726 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3727 		struct compat_siginfo __user *, uinfo,
3728 		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3729 {
3730 	sigset_t s;
3731 	struct timespec64 t;
3732 	kernel_siginfo_t info;
3733 	long ret;
3734 
3735 	if (sigsetsize != sizeof(sigset_t))
3736 		return -EINVAL;
3737 
3738 	if (get_compat_sigset(&s, uthese))
3739 		return -EFAULT;
3740 
3741 	if (uts) {
3742 		if (get_old_timespec32(&t, uts))
3743 			return -EFAULT;
3744 	}
3745 
3746 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3747 
3748 	if (ret > 0 && uinfo) {
3749 		if (copy_siginfo_to_user32(uinfo, &info))
3750 			ret = -EFAULT;
3751 	}
3752 
3753 	return ret;
3754 }
3755 #endif
3756 #endif
3757 
3758 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3759 {
3760 	clear_siginfo(info);
3761 	info->si_signo = sig;
3762 	info->si_errno = 0;
3763 	info->si_code = SI_USER;
3764 	info->si_pid = task_tgid_vnr(current);
3765 	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3766 }
3767 
3768 /**
3769  *  sys_kill - send a signal to a process
3770  *  @pid: the PID of the process
3771  *  @sig: signal to be sent
3772  */
3773 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3774 {
3775 	struct kernel_siginfo info;
3776 
3777 	prepare_kill_siginfo(sig, &info);
3778 
3779 	return kill_something_info(sig, &info, pid);
3780 }
3781 
3782 /*
3783  * Verify that the signaler and signalee either are in the same pid namespace
3784  * or that the signaler's pid namespace is an ancestor of the signalee's pid
3785  * namespace.
3786  */
3787 static bool access_pidfd_pidns(struct pid *pid)
3788 {
3789 	struct pid_namespace *active = task_active_pid_ns(current);
3790 	struct pid_namespace *p = ns_of_pid(pid);
3791 
3792 	for (;;) {
3793 		if (!p)
3794 			return false;
3795 		if (p == active)
3796 			break;
3797 		p = p->parent;
3798 	}
3799 
3800 	return true;
3801 }
3802 
3803 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3804 		siginfo_t __user *info)
3805 {
3806 #ifdef CONFIG_COMPAT
3807 	/*
3808 	 * Avoid hooking up compat syscalls and instead handle necessary
3809 	 * conversions here. Note, this is a stop-gap measure and should not be
3810 	 * considered a generic solution.
3811 	 */
3812 	if (in_compat_syscall())
3813 		return copy_siginfo_from_user32(
3814 			kinfo, (struct compat_siginfo __user *)info);
3815 #endif
3816 	return copy_siginfo_from_user(kinfo, info);
3817 }
3818 
3819 static struct pid *pidfd_to_pid(const struct file *file)
3820 {
3821 	struct pid *pid;
3822 
3823 	pid = pidfd_pid(file);
3824 	if (!IS_ERR(pid))
3825 		return pid;
3826 
3827 	return tgid_pidfd_to_pid(file);
3828 }
3829 
3830 /**
3831  * sys_pidfd_send_signal - Signal a process through a pidfd
3832  * @pidfd:  file descriptor of the process
3833  * @sig:    signal to send
3834  * @info:   signal info
3835  * @flags:  future flags
3836  *
3837  * The syscall currently only signals via PIDTYPE_PID which covers
3838  * kill(<positive-pid>, <signal>. It does not signal threads or process
3839  * groups.
3840  * In order to extend the syscall to threads and process groups the @flags
3841  * argument should be used. In essence, the @flags argument will determine
3842  * what is signaled and not the file descriptor itself. Put in other words,
3843  * grouping is a property of the flags argument not a property of the file
3844  * descriptor.
3845  *
3846  * Return: 0 on success, negative errno on failure
3847  */
3848 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3849 		siginfo_t __user *, info, unsigned int, flags)
3850 {
3851 	int ret;
3852 	struct fd f;
3853 	struct pid *pid;
3854 	kernel_siginfo_t kinfo;
3855 
3856 	/* Enforce flags be set to 0 until we add an extension. */
3857 	if (flags)
3858 		return -EINVAL;
3859 
3860 	f = fdget(pidfd);
3861 	if (!f.file)
3862 		return -EBADF;
3863 
3864 	/* Is this a pidfd? */
3865 	pid = pidfd_to_pid(f.file);
3866 	if (IS_ERR(pid)) {
3867 		ret = PTR_ERR(pid);
3868 		goto err;
3869 	}
3870 
3871 	ret = -EINVAL;
3872 	if (!access_pidfd_pidns(pid))
3873 		goto err;
3874 
3875 	if (info) {
3876 		ret = copy_siginfo_from_user_any(&kinfo, info);
3877 		if (unlikely(ret))
3878 			goto err;
3879 
3880 		ret = -EINVAL;
3881 		if (unlikely(sig != kinfo.si_signo))
3882 			goto err;
3883 
3884 		/* Only allow sending arbitrary signals to yourself. */
3885 		ret = -EPERM;
3886 		if ((task_pid(current) != pid) &&
3887 		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3888 			goto err;
3889 	} else {
3890 		prepare_kill_siginfo(sig, &kinfo);
3891 	}
3892 
3893 	ret = kill_pid_info(sig, &kinfo, pid);
3894 
3895 err:
3896 	fdput(f);
3897 	return ret;
3898 }
3899 
3900 static int
3901 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3902 {
3903 	struct task_struct *p;
3904 	int error = -ESRCH;
3905 
3906 	rcu_read_lock();
3907 	p = find_task_by_vpid(pid);
3908 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3909 		error = check_kill_permission(sig, info, p);
3910 		/*
3911 		 * The null signal is a permissions and process existence
3912 		 * probe.  No signal is actually delivered.
3913 		 */
3914 		if (!error && sig) {
3915 			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3916 			/*
3917 			 * If lock_task_sighand() failed we pretend the task
3918 			 * dies after receiving the signal. The window is tiny,
3919 			 * and the signal is private anyway.
3920 			 */
3921 			if (unlikely(error == -ESRCH))
3922 				error = 0;
3923 		}
3924 	}
3925 	rcu_read_unlock();
3926 
3927 	return error;
3928 }
3929 
3930 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3931 {
3932 	struct kernel_siginfo info;
3933 
3934 	clear_siginfo(&info);
3935 	info.si_signo = sig;
3936 	info.si_errno = 0;
3937 	info.si_code = SI_TKILL;
3938 	info.si_pid = task_tgid_vnr(current);
3939 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3940 
3941 	return do_send_specific(tgid, pid, sig, &info);
3942 }
3943 
3944 /**
3945  *  sys_tgkill - send signal to one specific thread
3946  *  @tgid: the thread group ID of the thread
3947  *  @pid: the PID of the thread
3948  *  @sig: signal to be sent
3949  *
3950  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3951  *  exists but it's not belonging to the target process anymore. This
3952  *  method solves the problem of threads exiting and PIDs getting reused.
3953  */
3954 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3955 {
3956 	/* This is only valid for single tasks */
3957 	if (pid <= 0 || tgid <= 0)
3958 		return -EINVAL;
3959 
3960 	return do_tkill(tgid, pid, sig);
3961 }
3962 
3963 /**
3964  *  sys_tkill - send signal to one specific task
3965  *  @pid: the PID of the task
3966  *  @sig: signal to be sent
3967  *
3968  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3969  */
3970 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3971 {
3972 	/* This is only valid for single tasks */
3973 	if (pid <= 0)
3974 		return -EINVAL;
3975 
3976 	return do_tkill(0, pid, sig);
3977 }
3978 
3979 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
3980 {
3981 	/* Not even root can pretend to send signals from the kernel.
3982 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3983 	 */
3984 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3985 	    (task_pid_vnr(current) != pid))
3986 		return -EPERM;
3987 
3988 	/* POSIX.1b doesn't mention process groups.  */
3989 	return kill_proc_info(sig, info, pid);
3990 }
3991 
3992 /**
3993  *  sys_rt_sigqueueinfo - send signal information to a signal
3994  *  @pid: the PID of the thread
3995  *  @sig: signal to be sent
3996  *  @uinfo: signal info to be sent
3997  */
3998 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
3999 		siginfo_t __user *, uinfo)
4000 {
4001 	kernel_siginfo_t info;
4002 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4003 	if (unlikely(ret))
4004 		return ret;
4005 	return do_rt_sigqueueinfo(pid, sig, &info);
4006 }
4007 
4008 #ifdef CONFIG_COMPAT
4009 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
4010 			compat_pid_t, pid,
4011 			int, sig,
4012 			struct compat_siginfo __user *, uinfo)
4013 {
4014 	kernel_siginfo_t info;
4015 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4016 	if (unlikely(ret))
4017 		return ret;
4018 	return do_rt_sigqueueinfo(pid, sig, &info);
4019 }
4020 #endif
4021 
4022 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4023 {
4024 	/* This is only valid for single tasks */
4025 	if (pid <= 0 || tgid <= 0)
4026 		return -EINVAL;
4027 
4028 	/* Not even root can pretend to send signals from the kernel.
4029 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4030 	 */
4031 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4032 	    (task_pid_vnr(current) != pid))
4033 		return -EPERM;
4034 
4035 	return do_send_specific(tgid, pid, sig, info);
4036 }
4037 
4038 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4039 		siginfo_t __user *, uinfo)
4040 {
4041 	kernel_siginfo_t info;
4042 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4043 	if (unlikely(ret))
4044 		return ret;
4045 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4046 }
4047 
4048 #ifdef CONFIG_COMPAT
4049 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4050 			compat_pid_t, tgid,
4051 			compat_pid_t, pid,
4052 			int, sig,
4053 			struct compat_siginfo __user *, uinfo)
4054 {
4055 	kernel_siginfo_t info;
4056 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4057 	if (unlikely(ret))
4058 		return ret;
4059 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4060 }
4061 #endif
4062 
4063 /*
4064  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4065  */
4066 void kernel_sigaction(int sig, __sighandler_t action)
4067 {
4068 	spin_lock_irq(&current->sighand->siglock);
4069 	current->sighand->action[sig - 1].sa.sa_handler = action;
4070 	if (action == SIG_IGN) {
4071 		sigset_t mask;
4072 
4073 		sigemptyset(&mask);
4074 		sigaddset(&mask, sig);
4075 
4076 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
4077 		flush_sigqueue_mask(&mask, &current->pending);
4078 		recalc_sigpending();
4079 	}
4080 	spin_unlock_irq(&current->sighand->siglock);
4081 }
4082 EXPORT_SYMBOL(kernel_sigaction);
4083 
4084 void __weak sigaction_compat_abi(struct k_sigaction *act,
4085 		struct k_sigaction *oact)
4086 {
4087 }
4088 
4089 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4090 {
4091 	struct task_struct *p = current, *t;
4092 	struct k_sigaction *k;
4093 	sigset_t mask;
4094 
4095 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4096 		return -EINVAL;
4097 
4098 	k = &p->sighand->action[sig-1];
4099 
4100 	spin_lock_irq(&p->sighand->siglock);
4101 	if (k->sa.sa_flags & SA_IMMUTABLE) {
4102 		spin_unlock_irq(&p->sighand->siglock);
4103 		return -EINVAL;
4104 	}
4105 	if (oact)
4106 		*oact = *k;
4107 
4108 	/*
4109 	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4110 	 * e.g. by having an architecture use the bit in their uapi.
4111 	 */
4112 	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4113 
4114 	/*
4115 	 * Clear unknown flag bits in order to allow userspace to detect missing
4116 	 * support for flag bits and to allow the kernel to use non-uapi bits
4117 	 * internally.
4118 	 */
4119 	if (act)
4120 		act->sa.sa_flags &= UAPI_SA_FLAGS;
4121 	if (oact)
4122 		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4123 
4124 	sigaction_compat_abi(act, oact);
4125 
4126 	if (act) {
4127 		sigdelsetmask(&act->sa.sa_mask,
4128 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4129 		*k = *act;
4130 		/*
4131 		 * POSIX 3.3.1.3:
4132 		 *  "Setting a signal action to SIG_IGN for a signal that is
4133 		 *   pending shall cause the pending signal to be discarded,
4134 		 *   whether or not it is blocked."
4135 		 *
4136 		 *  "Setting a signal action to SIG_DFL for a signal that is
4137 		 *   pending and whose default action is to ignore the signal
4138 		 *   (for example, SIGCHLD), shall cause the pending signal to
4139 		 *   be discarded, whether or not it is blocked"
4140 		 */
4141 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4142 			sigemptyset(&mask);
4143 			sigaddset(&mask, sig);
4144 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4145 			for_each_thread(p, t)
4146 				flush_sigqueue_mask(&mask, &t->pending);
4147 		}
4148 	}
4149 
4150 	spin_unlock_irq(&p->sighand->siglock);
4151 	return 0;
4152 }
4153 
4154 #ifdef CONFIG_DYNAMIC_SIGFRAME
4155 static inline void sigaltstack_lock(void)
4156 	__acquires(&current->sighand->siglock)
4157 {
4158 	spin_lock_irq(&current->sighand->siglock);
4159 }
4160 
4161 static inline void sigaltstack_unlock(void)
4162 	__releases(&current->sighand->siglock)
4163 {
4164 	spin_unlock_irq(&current->sighand->siglock);
4165 }
4166 #else
4167 static inline void sigaltstack_lock(void) { }
4168 static inline void sigaltstack_unlock(void) { }
4169 #endif
4170 
4171 static int
4172 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4173 		size_t min_ss_size)
4174 {
4175 	struct task_struct *t = current;
4176 	int ret = 0;
4177 
4178 	if (oss) {
4179 		memset(oss, 0, sizeof(stack_t));
4180 		oss->ss_sp = (void __user *) t->sas_ss_sp;
4181 		oss->ss_size = t->sas_ss_size;
4182 		oss->ss_flags = sas_ss_flags(sp) |
4183 			(current->sas_ss_flags & SS_FLAG_BITS);
4184 	}
4185 
4186 	if (ss) {
4187 		void __user *ss_sp = ss->ss_sp;
4188 		size_t ss_size = ss->ss_size;
4189 		unsigned ss_flags = ss->ss_flags;
4190 		int ss_mode;
4191 
4192 		if (unlikely(on_sig_stack(sp)))
4193 			return -EPERM;
4194 
4195 		ss_mode = ss_flags & ~SS_FLAG_BITS;
4196 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4197 				ss_mode != 0))
4198 			return -EINVAL;
4199 
4200 		/*
4201 		 * Return before taking any locks if no actual
4202 		 * sigaltstack changes were requested.
4203 		 */
4204 		if (t->sas_ss_sp == (unsigned long)ss_sp &&
4205 		    t->sas_ss_size == ss_size &&
4206 		    t->sas_ss_flags == ss_flags)
4207 			return 0;
4208 
4209 		sigaltstack_lock();
4210 		if (ss_mode == SS_DISABLE) {
4211 			ss_size = 0;
4212 			ss_sp = NULL;
4213 		} else {
4214 			if (unlikely(ss_size < min_ss_size))
4215 				ret = -ENOMEM;
4216 			if (!sigaltstack_size_valid(ss_size))
4217 				ret = -ENOMEM;
4218 		}
4219 		if (!ret) {
4220 			t->sas_ss_sp = (unsigned long) ss_sp;
4221 			t->sas_ss_size = ss_size;
4222 			t->sas_ss_flags = ss_flags;
4223 		}
4224 		sigaltstack_unlock();
4225 	}
4226 	return ret;
4227 }
4228 
4229 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4230 {
4231 	stack_t new, old;
4232 	int err;
4233 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4234 		return -EFAULT;
4235 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4236 			      current_user_stack_pointer(),
4237 			      MINSIGSTKSZ);
4238 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4239 		err = -EFAULT;
4240 	return err;
4241 }
4242 
4243 int restore_altstack(const stack_t __user *uss)
4244 {
4245 	stack_t new;
4246 	if (copy_from_user(&new, uss, sizeof(stack_t)))
4247 		return -EFAULT;
4248 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4249 			     MINSIGSTKSZ);
4250 	/* squash all but EFAULT for now */
4251 	return 0;
4252 }
4253 
4254 int __save_altstack(stack_t __user *uss, unsigned long sp)
4255 {
4256 	struct task_struct *t = current;
4257 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4258 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4259 		__put_user(t->sas_ss_size, &uss->ss_size);
4260 	return err;
4261 }
4262 
4263 #ifdef CONFIG_COMPAT
4264 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4265 				 compat_stack_t __user *uoss_ptr)
4266 {
4267 	stack_t uss, uoss;
4268 	int ret;
4269 
4270 	if (uss_ptr) {
4271 		compat_stack_t uss32;
4272 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4273 			return -EFAULT;
4274 		uss.ss_sp = compat_ptr(uss32.ss_sp);
4275 		uss.ss_flags = uss32.ss_flags;
4276 		uss.ss_size = uss32.ss_size;
4277 	}
4278 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4279 			     compat_user_stack_pointer(),
4280 			     COMPAT_MINSIGSTKSZ);
4281 	if (ret >= 0 && uoss_ptr)  {
4282 		compat_stack_t old;
4283 		memset(&old, 0, sizeof(old));
4284 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4285 		old.ss_flags = uoss.ss_flags;
4286 		old.ss_size = uoss.ss_size;
4287 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4288 			ret = -EFAULT;
4289 	}
4290 	return ret;
4291 }
4292 
4293 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4294 			const compat_stack_t __user *, uss_ptr,
4295 			compat_stack_t __user *, uoss_ptr)
4296 {
4297 	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4298 }
4299 
4300 int compat_restore_altstack(const compat_stack_t __user *uss)
4301 {
4302 	int err = do_compat_sigaltstack(uss, NULL);
4303 	/* squash all but -EFAULT for now */
4304 	return err == -EFAULT ? err : 0;
4305 }
4306 
4307 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4308 {
4309 	int err;
4310 	struct task_struct *t = current;
4311 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4312 			 &uss->ss_sp) |
4313 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4314 		__put_user(t->sas_ss_size, &uss->ss_size);
4315 	return err;
4316 }
4317 #endif
4318 
4319 #ifdef __ARCH_WANT_SYS_SIGPENDING
4320 
4321 /**
4322  *  sys_sigpending - examine pending signals
4323  *  @uset: where mask of pending signal is returned
4324  */
4325 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4326 {
4327 	sigset_t set;
4328 
4329 	if (sizeof(old_sigset_t) > sizeof(*uset))
4330 		return -EINVAL;
4331 
4332 	do_sigpending(&set);
4333 
4334 	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4335 		return -EFAULT;
4336 
4337 	return 0;
4338 }
4339 
4340 #ifdef CONFIG_COMPAT
4341 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4342 {
4343 	sigset_t set;
4344 
4345 	do_sigpending(&set);
4346 
4347 	return put_user(set.sig[0], set32);
4348 }
4349 #endif
4350 
4351 #endif
4352 
4353 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4354 /**
4355  *  sys_sigprocmask - examine and change blocked signals
4356  *  @how: whether to add, remove, or set signals
4357  *  @nset: signals to add or remove (if non-null)
4358  *  @oset: previous value of signal mask if non-null
4359  *
4360  * Some platforms have their own version with special arguments;
4361  * others support only sys_rt_sigprocmask.
4362  */
4363 
4364 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4365 		old_sigset_t __user *, oset)
4366 {
4367 	old_sigset_t old_set, new_set;
4368 	sigset_t new_blocked;
4369 
4370 	old_set = current->blocked.sig[0];
4371 
4372 	if (nset) {
4373 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4374 			return -EFAULT;
4375 
4376 		new_blocked = current->blocked;
4377 
4378 		switch (how) {
4379 		case SIG_BLOCK:
4380 			sigaddsetmask(&new_blocked, new_set);
4381 			break;
4382 		case SIG_UNBLOCK:
4383 			sigdelsetmask(&new_blocked, new_set);
4384 			break;
4385 		case SIG_SETMASK:
4386 			new_blocked.sig[0] = new_set;
4387 			break;
4388 		default:
4389 			return -EINVAL;
4390 		}
4391 
4392 		set_current_blocked(&new_blocked);
4393 	}
4394 
4395 	if (oset) {
4396 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4397 			return -EFAULT;
4398 	}
4399 
4400 	return 0;
4401 }
4402 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4403 
4404 #ifndef CONFIG_ODD_RT_SIGACTION
4405 /**
4406  *  sys_rt_sigaction - alter an action taken by a process
4407  *  @sig: signal to be sent
4408  *  @act: new sigaction
4409  *  @oact: used to save the previous sigaction
4410  *  @sigsetsize: size of sigset_t type
4411  */
4412 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4413 		const struct sigaction __user *, act,
4414 		struct sigaction __user *, oact,
4415 		size_t, sigsetsize)
4416 {
4417 	struct k_sigaction new_sa, old_sa;
4418 	int ret;
4419 
4420 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4421 	if (sigsetsize != sizeof(sigset_t))
4422 		return -EINVAL;
4423 
4424 	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4425 		return -EFAULT;
4426 
4427 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4428 	if (ret)
4429 		return ret;
4430 
4431 	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4432 		return -EFAULT;
4433 
4434 	return 0;
4435 }
4436 #ifdef CONFIG_COMPAT
4437 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4438 		const struct compat_sigaction __user *, act,
4439 		struct compat_sigaction __user *, oact,
4440 		compat_size_t, sigsetsize)
4441 {
4442 	struct k_sigaction new_ka, old_ka;
4443 #ifdef __ARCH_HAS_SA_RESTORER
4444 	compat_uptr_t restorer;
4445 #endif
4446 	int ret;
4447 
4448 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4449 	if (sigsetsize != sizeof(compat_sigset_t))
4450 		return -EINVAL;
4451 
4452 	if (act) {
4453 		compat_uptr_t handler;
4454 		ret = get_user(handler, &act->sa_handler);
4455 		new_ka.sa.sa_handler = compat_ptr(handler);
4456 #ifdef __ARCH_HAS_SA_RESTORER
4457 		ret |= get_user(restorer, &act->sa_restorer);
4458 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4459 #endif
4460 		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4461 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4462 		if (ret)
4463 			return -EFAULT;
4464 	}
4465 
4466 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4467 	if (!ret && oact) {
4468 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4469 			       &oact->sa_handler);
4470 		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4471 					 sizeof(oact->sa_mask));
4472 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4473 #ifdef __ARCH_HAS_SA_RESTORER
4474 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4475 				&oact->sa_restorer);
4476 #endif
4477 	}
4478 	return ret;
4479 }
4480 #endif
4481 #endif /* !CONFIG_ODD_RT_SIGACTION */
4482 
4483 #ifdef CONFIG_OLD_SIGACTION
4484 SYSCALL_DEFINE3(sigaction, int, sig,
4485 		const struct old_sigaction __user *, act,
4486 	        struct old_sigaction __user *, oact)
4487 {
4488 	struct k_sigaction new_ka, old_ka;
4489 	int ret;
4490 
4491 	if (act) {
4492 		old_sigset_t mask;
4493 		if (!access_ok(act, sizeof(*act)) ||
4494 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4495 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4496 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4497 		    __get_user(mask, &act->sa_mask))
4498 			return -EFAULT;
4499 #ifdef __ARCH_HAS_KA_RESTORER
4500 		new_ka.ka_restorer = NULL;
4501 #endif
4502 		siginitset(&new_ka.sa.sa_mask, mask);
4503 	}
4504 
4505 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4506 
4507 	if (!ret && oact) {
4508 		if (!access_ok(oact, sizeof(*oact)) ||
4509 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4510 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4511 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4512 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4513 			return -EFAULT;
4514 	}
4515 
4516 	return ret;
4517 }
4518 #endif
4519 #ifdef CONFIG_COMPAT_OLD_SIGACTION
4520 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4521 		const struct compat_old_sigaction __user *, act,
4522 	        struct compat_old_sigaction __user *, oact)
4523 {
4524 	struct k_sigaction new_ka, old_ka;
4525 	int ret;
4526 	compat_old_sigset_t mask;
4527 	compat_uptr_t handler, restorer;
4528 
4529 	if (act) {
4530 		if (!access_ok(act, sizeof(*act)) ||
4531 		    __get_user(handler, &act->sa_handler) ||
4532 		    __get_user(restorer, &act->sa_restorer) ||
4533 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4534 		    __get_user(mask, &act->sa_mask))
4535 			return -EFAULT;
4536 
4537 #ifdef __ARCH_HAS_KA_RESTORER
4538 		new_ka.ka_restorer = NULL;
4539 #endif
4540 		new_ka.sa.sa_handler = compat_ptr(handler);
4541 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4542 		siginitset(&new_ka.sa.sa_mask, mask);
4543 	}
4544 
4545 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4546 
4547 	if (!ret && oact) {
4548 		if (!access_ok(oact, sizeof(*oact)) ||
4549 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4550 			       &oact->sa_handler) ||
4551 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4552 			       &oact->sa_restorer) ||
4553 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4554 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4555 			return -EFAULT;
4556 	}
4557 	return ret;
4558 }
4559 #endif
4560 
4561 #ifdef CONFIG_SGETMASK_SYSCALL
4562 
4563 /*
4564  * For backwards compatibility.  Functionality superseded by sigprocmask.
4565  */
4566 SYSCALL_DEFINE0(sgetmask)
4567 {
4568 	/* SMP safe */
4569 	return current->blocked.sig[0];
4570 }
4571 
4572 SYSCALL_DEFINE1(ssetmask, int, newmask)
4573 {
4574 	int old = current->blocked.sig[0];
4575 	sigset_t newset;
4576 
4577 	siginitset(&newset, newmask);
4578 	set_current_blocked(&newset);
4579 
4580 	return old;
4581 }
4582 #endif /* CONFIG_SGETMASK_SYSCALL */
4583 
4584 #ifdef __ARCH_WANT_SYS_SIGNAL
4585 /*
4586  * For backwards compatibility.  Functionality superseded by sigaction.
4587  */
4588 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4589 {
4590 	struct k_sigaction new_sa, old_sa;
4591 	int ret;
4592 
4593 	new_sa.sa.sa_handler = handler;
4594 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4595 	sigemptyset(&new_sa.sa.sa_mask);
4596 
4597 	ret = do_sigaction(sig, &new_sa, &old_sa);
4598 
4599 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4600 }
4601 #endif /* __ARCH_WANT_SYS_SIGNAL */
4602 
4603 #ifdef __ARCH_WANT_SYS_PAUSE
4604 
4605 SYSCALL_DEFINE0(pause)
4606 {
4607 	while (!signal_pending(current)) {
4608 		__set_current_state(TASK_INTERRUPTIBLE);
4609 		schedule();
4610 	}
4611 	return -ERESTARTNOHAND;
4612 }
4613 
4614 #endif
4615 
4616 static int sigsuspend(sigset_t *set)
4617 {
4618 	current->saved_sigmask = current->blocked;
4619 	set_current_blocked(set);
4620 
4621 	while (!signal_pending(current)) {
4622 		__set_current_state(TASK_INTERRUPTIBLE);
4623 		schedule();
4624 	}
4625 	set_restore_sigmask();
4626 	return -ERESTARTNOHAND;
4627 }
4628 
4629 /**
4630  *  sys_rt_sigsuspend - replace the signal mask for a value with the
4631  *	@unewset value until a signal is received
4632  *  @unewset: new signal mask value
4633  *  @sigsetsize: size of sigset_t type
4634  */
4635 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4636 {
4637 	sigset_t newset;
4638 
4639 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4640 	if (sigsetsize != sizeof(sigset_t))
4641 		return -EINVAL;
4642 
4643 	if (copy_from_user(&newset, unewset, sizeof(newset)))
4644 		return -EFAULT;
4645 	return sigsuspend(&newset);
4646 }
4647 
4648 #ifdef CONFIG_COMPAT
4649 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4650 {
4651 	sigset_t newset;
4652 
4653 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4654 	if (sigsetsize != sizeof(sigset_t))
4655 		return -EINVAL;
4656 
4657 	if (get_compat_sigset(&newset, unewset))
4658 		return -EFAULT;
4659 	return sigsuspend(&newset);
4660 }
4661 #endif
4662 
4663 #ifdef CONFIG_OLD_SIGSUSPEND
4664 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4665 {
4666 	sigset_t blocked;
4667 	siginitset(&blocked, mask);
4668 	return sigsuspend(&blocked);
4669 }
4670 #endif
4671 #ifdef CONFIG_OLD_SIGSUSPEND3
4672 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4673 {
4674 	sigset_t blocked;
4675 	siginitset(&blocked, mask);
4676 	return sigsuspend(&blocked);
4677 }
4678 #endif
4679 
4680 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4681 {
4682 	return NULL;
4683 }
4684 
4685 static inline void siginfo_buildtime_checks(void)
4686 {
4687 	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4688 
4689 	/* Verify the offsets in the two siginfos match */
4690 #define CHECK_OFFSET(field) \
4691 	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4692 
4693 	/* kill */
4694 	CHECK_OFFSET(si_pid);
4695 	CHECK_OFFSET(si_uid);
4696 
4697 	/* timer */
4698 	CHECK_OFFSET(si_tid);
4699 	CHECK_OFFSET(si_overrun);
4700 	CHECK_OFFSET(si_value);
4701 
4702 	/* rt */
4703 	CHECK_OFFSET(si_pid);
4704 	CHECK_OFFSET(si_uid);
4705 	CHECK_OFFSET(si_value);
4706 
4707 	/* sigchld */
4708 	CHECK_OFFSET(si_pid);
4709 	CHECK_OFFSET(si_uid);
4710 	CHECK_OFFSET(si_status);
4711 	CHECK_OFFSET(si_utime);
4712 	CHECK_OFFSET(si_stime);
4713 
4714 	/* sigfault */
4715 	CHECK_OFFSET(si_addr);
4716 	CHECK_OFFSET(si_trapno);
4717 	CHECK_OFFSET(si_addr_lsb);
4718 	CHECK_OFFSET(si_lower);
4719 	CHECK_OFFSET(si_upper);
4720 	CHECK_OFFSET(si_pkey);
4721 	CHECK_OFFSET(si_perf_data);
4722 	CHECK_OFFSET(si_perf_type);
4723 
4724 	/* sigpoll */
4725 	CHECK_OFFSET(si_band);
4726 	CHECK_OFFSET(si_fd);
4727 
4728 	/* sigsys */
4729 	CHECK_OFFSET(si_call_addr);
4730 	CHECK_OFFSET(si_syscall);
4731 	CHECK_OFFSET(si_arch);
4732 #undef CHECK_OFFSET
4733 
4734 	/* usb asyncio */
4735 	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4736 		     offsetof(struct siginfo, si_addr));
4737 	if (sizeof(int) == sizeof(void __user *)) {
4738 		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4739 			     sizeof(void __user *));
4740 	} else {
4741 		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4742 			      sizeof_field(struct siginfo, si_uid)) !=
4743 			     sizeof(void __user *));
4744 		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4745 			     offsetof(struct siginfo, si_uid));
4746 	}
4747 #ifdef CONFIG_COMPAT
4748 	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4749 		     offsetof(struct compat_siginfo, si_addr));
4750 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4751 		     sizeof(compat_uptr_t));
4752 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4753 		     sizeof_field(struct siginfo, si_pid));
4754 #endif
4755 }
4756 
4757 void __init signals_init(void)
4758 {
4759 	siginfo_buildtime_checks();
4760 
4761 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
4762 }
4763 
4764 #ifdef CONFIG_KGDB_KDB
4765 #include <linux/kdb.h>
4766 /*
4767  * kdb_send_sig - Allows kdb to send signals without exposing
4768  * signal internals.  This function checks if the required locks are
4769  * available before calling the main signal code, to avoid kdb
4770  * deadlocks.
4771  */
4772 void kdb_send_sig(struct task_struct *t, int sig)
4773 {
4774 	static struct task_struct *kdb_prev_t;
4775 	int new_t, ret;
4776 	if (!spin_trylock(&t->sighand->siglock)) {
4777 		kdb_printf("Can't do kill command now.\n"
4778 			   "The sigmask lock is held somewhere else in "
4779 			   "kernel, try again later\n");
4780 		return;
4781 	}
4782 	new_t = kdb_prev_t != t;
4783 	kdb_prev_t = t;
4784 	if (!task_is_running(t) && new_t) {
4785 		spin_unlock(&t->sighand->siglock);
4786 		kdb_printf("Process is not RUNNING, sending a signal from "
4787 			   "kdb risks deadlock\n"
4788 			   "on the run queue locks. "
4789 			   "The signal has _not_ been sent.\n"
4790 			   "Reissue the kill command if you want to risk "
4791 			   "the deadlock.\n");
4792 		return;
4793 	}
4794 	ret = send_signal(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4795 	spin_unlock(&t->sighand->siglock);
4796 	if (ret)
4797 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
4798 			   sig, t->pid);
4799 	else
4800 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4801 }
4802 #endif	/* CONFIG_KGDB_KDB */
4803