xref: /openbmc/linux/kernel/signal.c (revision 1bcca2b1)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/kernel/signal.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  1997-11-02  Modified for POSIX.1b signals by Richard Henderson
8  *
9  *  2003-06-02  Jim Houston - Concurrent Computer Corp.
10  *		Changes to use preallocated sigqueue structures
11  *		to allow signals to be sent reliably.
12  */
13 
14 #include <linux/slab.h>
15 #include <linux/export.h>
16 #include <linux/init.h>
17 #include <linux/sched/mm.h>
18 #include <linux/sched/user.h>
19 #include <linux/sched/debug.h>
20 #include <linux/sched/task.h>
21 #include <linux/sched/task_stack.h>
22 #include <linux/sched/cputime.h>
23 #include <linux/file.h>
24 #include <linux/fs.h>
25 #include <linux/proc_fs.h>
26 #include <linux/tty.h>
27 #include <linux/binfmts.h>
28 #include <linux/coredump.h>
29 #include <linux/security.h>
30 #include <linux/syscalls.h>
31 #include <linux/ptrace.h>
32 #include <linux/signal.h>
33 #include <linux/signalfd.h>
34 #include <linux/ratelimit.h>
35 #include <linux/task_work.h>
36 #include <linux/capability.h>
37 #include <linux/freezer.h>
38 #include <linux/pid_namespace.h>
39 #include <linux/nsproxy.h>
40 #include <linux/user_namespace.h>
41 #include <linux/uprobes.h>
42 #include <linux/compat.h>
43 #include <linux/cn_proc.h>
44 #include <linux/compiler.h>
45 #include <linux/posix-timers.h>
46 #include <linux/cgroup.h>
47 #include <linux/audit.h>
48 
49 #define CREATE_TRACE_POINTS
50 #include <trace/events/signal.h>
51 
52 #include <asm/param.h>
53 #include <linux/uaccess.h>
54 #include <asm/unistd.h>
55 #include <asm/siginfo.h>
56 #include <asm/cacheflush.h>
57 #include <asm/syscall.h>	/* for syscall_get_* */
58 
59 /*
60  * SLAB caches for signal bits.
61  */
62 
63 static struct kmem_cache *sigqueue_cachep;
64 
65 int print_fatal_signals __read_mostly;
66 
67 static void __user *sig_handler(struct task_struct *t, int sig)
68 {
69 	return t->sighand->action[sig - 1].sa.sa_handler;
70 }
71 
72 static inline bool sig_handler_ignored(void __user *handler, int sig)
73 {
74 	/* Is it explicitly or implicitly ignored? */
75 	return handler == SIG_IGN ||
76 	       (handler == SIG_DFL && sig_kernel_ignore(sig));
77 }
78 
79 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
80 {
81 	void __user *handler;
82 
83 	handler = sig_handler(t, sig);
84 
85 	/* SIGKILL and SIGSTOP may not be sent to the global init */
86 	if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
87 		return true;
88 
89 	if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
90 	    handler == SIG_DFL && !(force && sig_kernel_only(sig)))
91 		return true;
92 
93 	/* Only allow kernel generated signals to this kthread */
94 	if (unlikely((t->flags & PF_KTHREAD) &&
95 		     (handler == SIG_KTHREAD_KERNEL) && !force))
96 		return true;
97 
98 	return sig_handler_ignored(handler, sig);
99 }
100 
101 static bool sig_ignored(struct task_struct *t, int sig, bool force)
102 {
103 	/*
104 	 * Blocked signals are never ignored, since the
105 	 * signal handler may change by the time it is
106 	 * unblocked.
107 	 */
108 	if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
109 		return false;
110 
111 	/*
112 	 * Tracers may want to know about even ignored signal unless it
113 	 * is SIGKILL which can't be reported anyway but can be ignored
114 	 * by SIGNAL_UNKILLABLE task.
115 	 */
116 	if (t->ptrace && sig != SIGKILL)
117 		return false;
118 
119 	return sig_task_ignored(t, sig, force);
120 }
121 
122 /*
123  * Re-calculate pending state from the set of locally pending
124  * signals, globally pending signals, and blocked signals.
125  */
126 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
127 {
128 	unsigned long ready;
129 	long i;
130 
131 	switch (_NSIG_WORDS) {
132 	default:
133 		for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
134 			ready |= signal->sig[i] &~ blocked->sig[i];
135 		break;
136 
137 	case 4: ready  = signal->sig[3] &~ blocked->sig[3];
138 		ready |= signal->sig[2] &~ blocked->sig[2];
139 		ready |= signal->sig[1] &~ blocked->sig[1];
140 		ready |= signal->sig[0] &~ blocked->sig[0];
141 		break;
142 
143 	case 2: ready  = signal->sig[1] &~ blocked->sig[1];
144 		ready |= signal->sig[0] &~ blocked->sig[0];
145 		break;
146 
147 	case 1: ready  = signal->sig[0] &~ blocked->sig[0];
148 	}
149 	return ready !=	0;
150 }
151 
152 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
153 
154 static bool recalc_sigpending_tsk(struct task_struct *t)
155 {
156 	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
157 	    PENDING(&t->pending, &t->blocked) ||
158 	    PENDING(&t->signal->shared_pending, &t->blocked) ||
159 	    cgroup_task_frozen(t)) {
160 		set_tsk_thread_flag(t, TIF_SIGPENDING);
161 		return true;
162 	}
163 
164 	/*
165 	 * We must never clear the flag in another thread, or in current
166 	 * when it's possible the current syscall is returning -ERESTART*.
167 	 * So we don't clear it here, and only callers who know they should do.
168 	 */
169 	return false;
170 }
171 
172 /*
173  * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
174  * This is superfluous when called on current, the wakeup is a harmless no-op.
175  */
176 void recalc_sigpending_and_wake(struct task_struct *t)
177 {
178 	if (recalc_sigpending_tsk(t))
179 		signal_wake_up(t, 0);
180 }
181 
182 void recalc_sigpending(void)
183 {
184 	if (!recalc_sigpending_tsk(current) && !freezing(current))
185 		clear_thread_flag(TIF_SIGPENDING);
186 
187 }
188 EXPORT_SYMBOL(recalc_sigpending);
189 
190 void calculate_sigpending(void)
191 {
192 	/* Have any signals or users of TIF_SIGPENDING been delayed
193 	 * until after fork?
194 	 */
195 	spin_lock_irq(&current->sighand->siglock);
196 	set_tsk_thread_flag(current, TIF_SIGPENDING);
197 	recalc_sigpending();
198 	spin_unlock_irq(&current->sighand->siglock);
199 }
200 
201 /* Given the mask, find the first available signal that should be serviced. */
202 
203 #define SYNCHRONOUS_MASK \
204 	(sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
205 	 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
206 
207 int next_signal(struct sigpending *pending, sigset_t *mask)
208 {
209 	unsigned long i, *s, *m, x;
210 	int sig = 0;
211 
212 	s = pending->signal.sig;
213 	m = mask->sig;
214 
215 	/*
216 	 * Handle the first word specially: it contains the
217 	 * synchronous signals that need to be dequeued first.
218 	 */
219 	x = *s &~ *m;
220 	if (x) {
221 		if (x & SYNCHRONOUS_MASK)
222 			x &= SYNCHRONOUS_MASK;
223 		sig = ffz(~x) + 1;
224 		return sig;
225 	}
226 
227 	switch (_NSIG_WORDS) {
228 	default:
229 		for (i = 1; i < _NSIG_WORDS; ++i) {
230 			x = *++s &~ *++m;
231 			if (!x)
232 				continue;
233 			sig = ffz(~x) + i*_NSIG_BPW + 1;
234 			break;
235 		}
236 		break;
237 
238 	case 2:
239 		x = s[1] &~ m[1];
240 		if (!x)
241 			break;
242 		sig = ffz(~x) + _NSIG_BPW + 1;
243 		break;
244 
245 	case 1:
246 		/* Nothing to do */
247 		break;
248 	}
249 
250 	return sig;
251 }
252 
253 static inline void print_dropped_signal(int sig)
254 {
255 	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
256 
257 	if (!print_fatal_signals)
258 		return;
259 
260 	if (!__ratelimit(&ratelimit_state))
261 		return;
262 
263 	pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
264 				current->comm, current->pid, sig);
265 }
266 
267 /**
268  * task_set_jobctl_pending - set jobctl pending bits
269  * @task: target task
270  * @mask: pending bits to set
271  *
272  * Clear @mask from @task->jobctl.  @mask must be subset of
273  * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
274  * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
275  * cleared.  If @task is already being killed or exiting, this function
276  * becomes noop.
277  *
278  * CONTEXT:
279  * Must be called with @task->sighand->siglock held.
280  *
281  * RETURNS:
282  * %true if @mask is set, %false if made noop because @task was dying.
283  */
284 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
285 {
286 	BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
287 			JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
288 	BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
289 
290 	if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
291 		return false;
292 
293 	if (mask & JOBCTL_STOP_SIGMASK)
294 		task->jobctl &= ~JOBCTL_STOP_SIGMASK;
295 
296 	task->jobctl |= mask;
297 	return true;
298 }
299 
300 /**
301  * task_clear_jobctl_trapping - clear jobctl trapping bit
302  * @task: target task
303  *
304  * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
305  * Clear it and wake up the ptracer.  Note that we don't need any further
306  * locking.  @task->siglock guarantees that @task->parent points to the
307  * ptracer.
308  *
309  * CONTEXT:
310  * Must be called with @task->sighand->siglock held.
311  */
312 void task_clear_jobctl_trapping(struct task_struct *task)
313 {
314 	if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
315 		task->jobctl &= ~JOBCTL_TRAPPING;
316 		smp_mb();	/* advised by wake_up_bit() */
317 		wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
318 	}
319 }
320 
321 /**
322  * task_clear_jobctl_pending - clear jobctl pending bits
323  * @task: target task
324  * @mask: pending bits to clear
325  *
326  * Clear @mask from @task->jobctl.  @mask must be subset of
327  * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
328  * STOP bits are cleared together.
329  *
330  * If clearing of @mask leaves no stop or trap pending, this function calls
331  * task_clear_jobctl_trapping().
332  *
333  * CONTEXT:
334  * Must be called with @task->sighand->siglock held.
335  */
336 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
337 {
338 	BUG_ON(mask & ~JOBCTL_PENDING_MASK);
339 
340 	if (mask & JOBCTL_STOP_PENDING)
341 		mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
342 
343 	task->jobctl &= ~mask;
344 
345 	if (!(task->jobctl & JOBCTL_PENDING_MASK))
346 		task_clear_jobctl_trapping(task);
347 }
348 
349 /**
350  * task_participate_group_stop - participate in a group stop
351  * @task: task participating in a group stop
352  *
353  * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
354  * Group stop states are cleared and the group stop count is consumed if
355  * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
356  * stop, the appropriate `SIGNAL_*` flags are set.
357  *
358  * CONTEXT:
359  * Must be called with @task->sighand->siglock held.
360  *
361  * RETURNS:
362  * %true if group stop completion should be notified to the parent, %false
363  * otherwise.
364  */
365 static bool task_participate_group_stop(struct task_struct *task)
366 {
367 	struct signal_struct *sig = task->signal;
368 	bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
369 
370 	WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
371 
372 	task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
373 
374 	if (!consume)
375 		return false;
376 
377 	if (!WARN_ON_ONCE(sig->group_stop_count == 0))
378 		sig->group_stop_count--;
379 
380 	/*
381 	 * Tell the caller to notify completion iff we are entering into a
382 	 * fresh group stop.  Read comment in do_signal_stop() for details.
383 	 */
384 	if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
385 		signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
386 		return true;
387 	}
388 	return false;
389 }
390 
391 void task_join_group_stop(struct task_struct *task)
392 {
393 	unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
394 	struct signal_struct *sig = current->signal;
395 
396 	if (sig->group_stop_count) {
397 		sig->group_stop_count++;
398 		mask |= JOBCTL_STOP_CONSUME;
399 	} else if (!(sig->flags & SIGNAL_STOP_STOPPED))
400 		return;
401 
402 	/* Have the new thread join an on-going signal group stop */
403 	task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
404 }
405 
406 /*
407  * allocate a new signal queue record
408  * - this may be called without locks if and only if t == current, otherwise an
409  *   appropriate lock must be held to stop the target task from exiting
410  */
411 static struct sigqueue *
412 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t gfp_flags,
413 		 int override_rlimit, const unsigned int sigqueue_flags)
414 {
415 	struct sigqueue *q = NULL;
416 	struct ucounts *ucounts = NULL;
417 	long sigpending;
418 
419 	/*
420 	 * Protect access to @t credentials. This can go away when all
421 	 * callers hold rcu read lock.
422 	 *
423 	 * NOTE! A pending signal will hold on to the user refcount,
424 	 * and we get/put the refcount only when the sigpending count
425 	 * changes from/to zero.
426 	 */
427 	rcu_read_lock();
428 	ucounts = task_ucounts(t);
429 	sigpending = inc_rlimit_get_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
430 	rcu_read_unlock();
431 	if (!sigpending)
432 		return NULL;
433 
434 	if (override_rlimit || likely(sigpending <= task_rlimit(t, RLIMIT_SIGPENDING))) {
435 		q = kmem_cache_alloc(sigqueue_cachep, gfp_flags);
436 	} else {
437 		print_dropped_signal(sig);
438 	}
439 
440 	if (unlikely(q == NULL)) {
441 		dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
442 	} else {
443 		INIT_LIST_HEAD(&q->list);
444 		q->flags = sigqueue_flags;
445 		q->ucounts = ucounts;
446 	}
447 	return q;
448 }
449 
450 static void __sigqueue_free(struct sigqueue *q)
451 {
452 	if (q->flags & SIGQUEUE_PREALLOC)
453 		return;
454 	if (q->ucounts) {
455 		dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING);
456 		q->ucounts = NULL;
457 	}
458 	kmem_cache_free(sigqueue_cachep, q);
459 }
460 
461 void flush_sigqueue(struct sigpending *queue)
462 {
463 	struct sigqueue *q;
464 
465 	sigemptyset(&queue->signal);
466 	while (!list_empty(&queue->list)) {
467 		q = list_entry(queue->list.next, struct sigqueue , list);
468 		list_del_init(&q->list);
469 		__sigqueue_free(q);
470 	}
471 }
472 
473 /*
474  * Flush all pending signals for this kthread.
475  */
476 void flush_signals(struct task_struct *t)
477 {
478 	unsigned long flags;
479 
480 	spin_lock_irqsave(&t->sighand->siglock, flags);
481 	clear_tsk_thread_flag(t, TIF_SIGPENDING);
482 	flush_sigqueue(&t->pending);
483 	flush_sigqueue(&t->signal->shared_pending);
484 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
485 }
486 EXPORT_SYMBOL(flush_signals);
487 
488 #ifdef CONFIG_POSIX_TIMERS
489 static void __flush_itimer_signals(struct sigpending *pending)
490 {
491 	sigset_t signal, retain;
492 	struct sigqueue *q, *n;
493 
494 	signal = pending->signal;
495 	sigemptyset(&retain);
496 
497 	list_for_each_entry_safe(q, n, &pending->list, list) {
498 		int sig = q->info.si_signo;
499 
500 		if (likely(q->info.si_code != SI_TIMER)) {
501 			sigaddset(&retain, sig);
502 		} else {
503 			sigdelset(&signal, sig);
504 			list_del_init(&q->list);
505 			__sigqueue_free(q);
506 		}
507 	}
508 
509 	sigorsets(&pending->signal, &signal, &retain);
510 }
511 
512 void flush_itimer_signals(void)
513 {
514 	struct task_struct *tsk = current;
515 	unsigned long flags;
516 
517 	spin_lock_irqsave(&tsk->sighand->siglock, flags);
518 	__flush_itimer_signals(&tsk->pending);
519 	__flush_itimer_signals(&tsk->signal->shared_pending);
520 	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
521 }
522 #endif
523 
524 void ignore_signals(struct task_struct *t)
525 {
526 	int i;
527 
528 	for (i = 0; i < _NSIG; ++i)
529 		t->sighand->action[i].sa.sa_handler = SIG_IGN;
530 
531 	flush_signals(t);
532 }
533 
534 /*
535  * Flush all handlers for a task.
536  */
537 
538 void
539 flush_signal_handlers(struct task_struct *t, int force_default)
540 {
541 	int i;
542 	struct k_sigaction *ka = &t->sighand->action[0];
543 	for (i = _NSIG ; i != 0 ; i--) {
544 		if (force_default || ka->sa.sa_handler != SIG_IGN)
545 			ka->sa.sa_handler = SIG_DFL;
546 		ka->sa.sa_flags = 0;
547 #ifdef __ARCH_HAS_SA_RESTORER
548 		ka->sa.sa_restorer = NULL;
549 #endif
550 		sigemptyset(&ka->sa.sa_mask);
551 		ka++;
552 	}
553 }
554 
555 bool unhandled_signal(struct task_struct *tsk, int sig)
556 {
557 	void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
558 	if (is_global_init(tsk))
559 		return true;
560 
561 	if (handler != SIG_IGN && handler != SIG_DFL)
562 		return false;
563 
564 	/* if ptraced, let the tracer determine */
565 	return !tsk->ptrace;
566 }
567 
568 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
569 			   bool *resched_timer)
570 {
571 	struct sigqueue *q, *first = NULL;
572 
573 	/*
574 	 * Collect the siginfo appropriate to this signal.  Check if
575 	 * there is another siginfo for the same signal.
576 	*/
577 	list_for_each_entry(q, &list->list, list) {
578 		if (q->info.si_signo == sig) {
579 			if (first)
580 				goto still_pending;
581 			first = q;
582 		}
583 	}
584 
585 	sigdelset(&list->signal, sig);
586 
587 	if (first) {
588 still_pending:
589 		list_del_init(&first->list);
590 		copy_siginfo(info, &first->info);
591 
592 		*resched_timer =
593 			(first->flags & SIGQUEUE_PREALLOC) &&
594 			(info->si_code == SI_TIMER) &&
595 			(info->si_sys_private);
596 
597 		__sigqueue_free(first);
598 	} else {
599 		/*
600 		 * Ok, it wasn't in the queue.  This must be
601 		 * a fast-pathed signal or we must have been
602 		 * out of queue space.  So zero out the info.
603 		 */
604 		clear_siginfo(info);
605 		info->si_signo = sig;
606 		info->si_errno = 0;
607 		info->si_code = SI_USER;
608 		info->si_pid = 0;
609 		info->si_uid = 0;
610 	}
611 }
612 
613 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
614 			kernel_siginfo_t *info, bool *resched_timer)
615 {
616 	int sig = next_signal(pending, mask);
617 
618 	if (sig)
619 		collect_signal(sig, pending, info, resched_timer);
620 	return sig;
621 }
622 
623 /*
624  * Dequeue a signal and return the element to the caller, which is
625  * expected to free it.
626  *
627  * All callers have to hold the siglock.
628  */
629 int dequeue_signal(struct task_struct *tsk, sigset_t *mask,
630 		   kernel_siginfo_t *info, enum pid_type *type)
631 {
632 	bool resched_timer = false;
633 	int signr;
634 
635 	/* We only dequeue private signals from ourselves, we don't let
636 	 * signalfd steal them
637 	 */
638 	*type = PIDTYPE_PID;
639 	signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
640 	if (!signr) {
641 		*type = PIDTYPE_TGID;
642 		signr = __dequeue_signal(&tsk->signal->shared_pending,
643 					 mask, info, &resched_timer);
644 #ifdef CONFIG_POSIX_TIMERS
645 		/*
646 		 * itimer signal ?
647 		 *
648 		 * itimers are process shared and we restart periodic
649 		 * itimers in the signal delivery path to prevent DoS
650 		 * attacks in the high resolution timer case. This is
651 		 * compliant with the old way of self-restarting
652 		 * itimers, as the SIGALRM is a legacy signal and only
653 		 * queued once. Changing the restart behaviour to
654 		 * restart the timer in the signal dequeue path is
655 		 * reducing the timer noise on heavy loaded !highres
656 		 * systems too.
657 		 */
658 		if (unlikely(signr == SIGALRM)) {
659 			struct hrtimer *tmr = &tsk->signal->real_timer;
660 
661 			if (!hrtimer_is_queued(tmr) &&
662 			    tsk->signal->it_real_incr != 0) {
663 				hrtimer_forward(tmr, tmr->base->get_time(),
664 						tsk->signal->it_real_incr);
665 				hrtimer_restart(tmr);
666 			}
667 		}
668 #endif
669 	}
670 
671 	recalc_sigpending();
672 	if (!signr)
673 		return 0;
674 
675 	if (unlikely(sig_kernel_stop(signr))) {
676 		/*
677 		 * Set a marker that we have dequeued a stop signal.  Our
678 		 * caller might release the siglock and then the pending
679 		 * stop signal it is about to process is no longer in the
680 		 * pending bitmasks, but must still be cleared by a SIGCONT
681 		 * (and overruled by a SIGKILL).  So those cases clear this
682 		 * shared flag after we've set it.  Note that this flag may
683 		 * remain set after the signal we return is ignored or
684 		 * handled.  That doesn't matter because its only purpose
685 		 * is to alert stop-signal processing code when another
686 		 * processor has come along and cleared the flag.
687 		 */
688 		current->jobctl |= JOBCTL_STOP_DEQUEUED;
689 	}
690 #ifdef CONFIG_POSIX_TIMERS
691 	if (resched_timer) {
692 		/*
693 		 * Release the siglock to ensure proper locking order
694 		 * of timer locks outside of siglocks.  Note, we leave
695 		 * irqs disabled here, since the posix-timers code is
696 		 * about to disable them again anyway.
697 		 */
698 		spin_unlock(&tsk->sighand->siglock);
699 		posixtimer_rearm(info);
700 		spin_lock(&tsk->sighand->siglock);
701 
702 		/* Don't expose the si_sys_private value to userspace */
703 		info->si_sys_private = 0;
704 	}
705 #endif
706 	return signr;
707 }
708 EXPORT_SYMBOL_GPL(dequeue_signal);
709 
710 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
711 {
712 	struct task_struct *tsk = current;
713 	struct sigpending *pending = &tsk->pending;
714 	struct sigqueue *q, *sync = NULL;
715 
716 	/*
717 	 * Might a synchronous signal be in the queue?
718 	 */
719 	if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
720 		return 0;
721 
722 	/*
723 	 * Return the first synchronous signal in the queue.
724 	 */
725 	list_for_each_entry(q, &pending->list, list) {
726 		/* Synchronous signals have a positive si_code */
727 		if ((q->info.si_code > SI_USER) &&
728 		    (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
729 			sync = q;
730 			goto next;
731 		}
732 	}
733 	return 0;
734 next:
735 	/*
736 	 * Check if there is another siginfo for the same signal.
737 	 */
738 	list_for_each_entry_continue(q, &pending->list, list) {
739 		if (q->info.si_signo == sync->info.si_signo)
740 			goto still_pending;
741 	}
742 
743 	sigdelset(&pending->signal, sync->info.si_signo);
744 	recalc_sigpending();
745 still_pending:
746 	list_del_init(&sync->list);
747 	copy_siginfo(info, &sync->info);
748 	__sigqueue_free(sync);
749 	return info->si_signo;
750 }
751 
752 /*
753  * Tell a process that it has a new active signal..
754  *
755  * NOTE! we rely on the previous spin_lock to
756  * lock interrupts for us! We can only be called with
757  * "siglock" held, and the local interrupt must
758  * have been disabled when that got acquired!
759  *
760  * No need to set need_resched since signal event passing
761  * goes through ->blocked
762  */
763 void signal_wake_up_state(struct task_struct *t, unsigned int state)
764 {
765 	set_tsk_thread_flag(t, TIF_SIGPENDING);
766 	/*
767 	 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
768 	 * case. We don't check t->state here because there is a race with it
769 	 * executing another processor and just now entering stopped state.
770 	 * By using wake_up_state, we ensure the process will wake up and
771 	 * handle its death signal.
772 	 */
773 	if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
774 		kick_process(t);
775 }
776 
777 /*
778  * Remove signals in mask from the pending set and queue.
779  * Returns 1 if any signals were found.
780  *
781  * All callers must be holding the siglock.
782  */
783 static void flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
784 {
785 	struct sigqueue *q, *n;
786 	sigset_t m;
787 
788 	sigandsets(&m, mask, &s->signal);
789 	if (sigisemptyset(&m))
790 		return;
791 
792 	sigandnsets(&s->signal, &s->signal, mask);
793 	list_for_each_entry_safe(q, n, &s->list, list) {
794 		if (sigismember(mask, q->info.si_signo)) {
795 			list_del_init(&q->list);
796 			__sigqueue_free(q);
797 		}
798 	}
799 }
800 
801 static inline int is_si_special(const struct kernel_siginfo *info)
802 {
803 	return info <= SEND_SIG_PRIV;
804 }
805 
806 static inline bool si_fromuser(const struct kernel_siginfo *info)
807 {
808 	return info == SEND_SIG_NOINFO ||
809 		(!is_si_special(info) && SI_FROMUSER(info));
810 }
811 
812 /*
813  * called with RCU read lock from check_kill_permission()
814  */
815 static bool kill_ok_by_cred(struct task_struct *t)
816 {
817 	const struct cred *cred = current_cred();
818 	const struct cred *tcred = __task_cred(t);
819 
820 	return uid_eq(cred->euid, tcred->suid) ||
821 	       uid_eq(cred->euid, tcred->uid) ||
822 	       uid_eq(cred->uid, tcred->suid) ||
823 	       uid_eq(cred->uid, tcred->uid) ||
824 	       ns_capable(tcred->user_ns, CAP_KILL);
825 }
826 
827 /*
828  * Bad permissions for sending the signal
829  * - the caller must hold the RCU read lock
830  */
831 static int check_kill_permission(int sig, struct kernel_siginfo *info,
832 				 struct task_struct *t)
833 {
834 	struct pid *sid;
835 	int error;
836 
837 	if (!valid_signal(sig))
838 		return -EINVAL;
839 
840 	if (!si_fromuser(info))
841 		return 0;
842 
843 	error = audit_signal_info(sig, t); /* Let audit system see the signal */
844 	if (error)
845 		return error;
846 
847 	if (!same_thread_group(current, t) &&
848 	    !kill_ok_by_cred(t)) {
849 		switch (sig) {
850 		case SIGCONT:
851 			sid = task_session(t);
852 			/*
853 			 * We don't return the error if sid == NULL. The
854 			 * task was unhashed, the caller must notice this.
855 			 */
856 			if (!sid || sid == task_session(current))
857 				break;
858 			fallthrough;
859 		default:
860 			return -EPERM;
861 		}
862 	}
863 
864 	return security_task_kill(t, info, sig, NULL);
865 }
866 
867 /**
868  * ptrace_trap_notify - schedule trap to notify ptracer
869  * @t: tracee wanting to notify tracer
870  *
871  * This function schedules sticky ptrace trap which is cleared on the next
872  * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
873  * ptracer.
874  *
875  * If @t is running, STOP trap will be taken.  If trapped for STOP and
876  * ptracer is listening for events, tracee is woken up so that it can
877  * re-trap for the new event.  If trapped otherwise, STOP trap will be
878  * eventually taken without returning to userland after the existing traps
879  * are finished by PTRACE_CONT.
880  *
881  * CONTEXT:
882  * Must be called with @task->sighand->siglock held.
883  */
884 static void ptrace_trap_notify(struct task_struct *t)
885 {
886 	WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
887 	assert_spin_locked(&t->sighand->siglock);
888 
889 	task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
890 	ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
891 }
892 
893 /*
894  * Handle magic process-wide effects of stop/continue signals. Unlike
895  * the signal actions, these happen immediately at signal-generation
896  * time regardless of blocking, ignoring, or handling.  This does the
897  * actual continuing for SIGCONT, but not the actual stopping for stop
898  * signals. The process stop is done as a signal action for SIG_DFL.
899  *
900  * Returns true if the signal should be actually delivered, otherwise
901  * it should be dropped.
902  */
903 static bool prepare_signal(int sig, struct task_struct *p, bool force)
904 {
905 	struct signal_struct *signal = p->signal;
906 	struct task_struct *t;
907 	sigset_t flush;
908 
909 	if (signal->flags & SIGNAL_GROUP_EXIT) {
910 		if (signal->core_state)
911 			return sig == SIGKILL;
912 		/*
913 		 * The process is in the middle of dying, nothing to do.
914 		 */
915 	} else if (sig_kernel_stop(sig)) {
916 		/*
917 		 * This is a stop signal.  Remove SIGCONT from all queues.
918 		 */
919 		siginitset(&flush, sigmask(SIGCONT));
920 		flush_sigqueue_mask(&flush, &signal->shared_pending);
921 		for_each_thread(p, t)
922 			flush_sigqueue_mask(&flush, &t->pending);
923 	} else if (sig == SIGCONT) {
924 		unsigned int why;
925 		/*
926 		 * Remove all stop signals from all queues, wake all threads.
927 		 */
928 		siginitset(&flush, SIG_KERNEL_STOP_MASK);
929 		flush_sigqueue_mask(&flush, &signal->shared_pending);
930 		for_each_thread(p, t) {
931 			flush_sigqueue_mask(&flush, &t->pending);
932 			task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
933 			if (likely(!(t->ptrace & PT_SEIZED)))
934 				wake_up_state(t, __TASK_STOPPED);
935 			else
936 				ptrace_trap_notify(t);
937 		}
938 
939 		/*
940 		 * Notify the parent with CLD_CONTINUED if we were stopped.
941 		 *
942 		 * If we were in the middle of a group stop, we pretend it
943 		 * was already finished, and then continued. Since SIGCHLD
944 		 * doesn't queue we report only CLD_STOPPED, as if the next
945 		 * CLD_CONTINUED was dropped.
946 		 */
947 		why = 0;
948 		if (signal->flags & SIGNAL_STOP_STOPPED)
949 			why |= SIGNAL_CLD_CONTINUED;
950 		else if (signal->group_stop_count)
951 			why |= SIGNAL_CLD_STOPPED;
952 
953 		if (why) {
954 			/*
955 			 * The first thread which returns from do_signal_stop()
956 			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
957 			 * notify its parent. See get_signal().
958 			 */
959 			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
960 			signal->group_stop_count = 0;
961 			signal->group_exit_code = 0;
962 		}
963 	}
964 
965 	return !sig_ignored(p, sig, force);
966 }
967 
968 /*
969  * Test if P wants to take SIG.  After we've checked all threads with this,
970  * it's equivalent to finding no threads not blocking SIG.  Any threads not
971  * blocking SIG were ruled out because they are not running and already
972  * have pending signals.  Such threads will dequeue from the shared queue
973  * as soon as they're available, so putting the signal on the shared queue
974  * will be equivalent to sending it to one such thread.
975  */
976 static inline bool wants_signal(int sig, struct task_struct *p)
977 {
978 	if (sigismember(&p->blocked, sig))
979 		return false;
980 
981 	if (p->flags & PF_EXITING)
982 		return false;
983 
984 	if (sig == SIGKILL)
985 		return true;
986 
987 	if (task_is_stopped_or_traced(p))
988 		return false;
989 
990 	return task_curr(p) || !task_sigpending(p);
991 }
992 
993 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
994 {
995 	struct signal_struct *signal = p->signal;
996 	struct task_struct *t;
997 
998 	/*
999 	 * Now find a thread we can wake up to take the signal off the queue.
1000 	 *
1001 	 * If the main thread wants the signal, it gets first crack.
1002 	 * Probably the least surprising to the average bear.
1003 	 */
1004 	if (wants_signal(sig, p))
1005 		t = p;
1006 	else if ((type == PIDTYPE_PID) || thread_group_empty(p))
1007 		/*
1008 		 * There is just one thread and it does not need to be woken.
1009 		 * It will dequeue unblocked signals before it runs again.
1010 		 */
1011 		return;
1012 	else {
1013 		/*
1014 		 * Otherwise try to find a suitable thread.
1015 		 */
1016 		t = signal->curr_target;
1017 		while (!wants_signal(sig, t)) {
1018 			t = next_thread(t);
1019 			if (t == signal->curr_target)
1020 				/*
1021 				 * No thread needs to be woken.
1022 				 * Any eligible threads will see
1023 				 * the signal in the queue soon.
1024 				 */
1025 				return;
1026 		}
1027 		signal->curr_target = t;
1028 	}
1029 
1030 	/*
1031 	 * Found a killable thread.  If the signal will be fatal,
1032 	 * then start taking the whole group down immediately.
1033 	 */
1034 	if (sig_fatal(p, sig) &&
1035 	    (signal->core_state || !(signal->flags & SIGNAL_GROUP_EXIT)) &&
1036 	    !sigismember(&t->real_blocked, sig) &&
1037 	    (sig == SIGKILL || !p->ptrace)) {
1038 		/*
1039 		 * This signal will be fatal to the whole group.
1040 		 */
1041 		if (!sig_kernel_coredump(sig)) {
1042 			/*
1043 			 * Start a group exit and wake everybody up.
1044 			 * This way we don't have other threads
1045 			 * running and doing things after a slower
1046 			 * thread has the fatal signal pending.
1047 			 */
1048 			signal->flags = SIGNAL_GROUP_EXIT;
1049 			signal->group_exit_code = sig;
1050 			signal->group_stop_count = 0;
1051 			t = p;
1052 			do {
1053 				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1054 				sigaddset(&t->pending.signal, SIGKILL);
1055 				signal_wake_up(t, 1);
1056 			} while_each_thread(p, t);
1057 			return;
1058 		}
1059 	}
1060 
1061 	/*
1062 	 * The signal is already in the shared-pending queue.
1063 	 * Tell the chosen thread to wake up and dequeue it.
1064 	 */
1065 	signal_wake_up(t, sig == SIGKILL);
1066 	return;
1067 }
1068 
1069 static inline bool legacy_queue(struct sigpending *signals, int sig)
1070 {
1071 	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1072 }
1073 
1074 static int __send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1075 			enum pid_type type, bool force)
1076 {
1077 	struct sigpending *pending;
1078 	struct sigqueue *q;
1079 	int override_rlimit;
1080 	int ret = 0, result;
1081 
1082 	assert_spin_locked(&t->sighand->siglock);
1083 
1084 	result = TRACE_SIGNAL_IGNORED;
1085 	if (!prepare_signal(sig, t, force))
1086 		goto ret;
1087 
1088 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1089 	/*
1090 	 * Short-circuit ignored signals and support queuing
1091 	 * exactly one non-rt signal, so that we can get more
1092 	 * detailed information about the cause of the signal.
1093 	 */
1094 	result = TRACE_SIGNAL_ALREADY_PENDING;
1095 	if (legacy_queue(pending, sig))
1096 		goto ret;
1097 
1098 	result = TRACE_SIGNAL_DELIVERED;
1099 	/*
1100 	 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1101 	 */
1102 	if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1103 		goto out_set;
1104 
1105 	/*
1106 	 * Real-time signals must be queued if sent by sigqueue, or
1107 	 * some other real-time mechanism.  It is implementation
1108 	 * defined whether kill() does so.  We attempt to do so, on
1109 	 * the principle of least surprise, but since kill is not
1110 	 * allowed to fail with EAGAIN when low on memory we just
1111 	 * make sure at least one signal gets delivered and don't
1112 	 * pass on the info struct.
1113 	 */
1114 	if (sig < SIGRTMIN)
1115 		override_rlimit = (is_si_special(info) || info->si_code >= 0);
1116 	else
1117 		override_rlimit = 0;
1118 
1119 	q = __sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit, 0);
1120 
1121 	if (q) {
1122 		list_add_tail(&q->list, &pending->list);
1123 		switch ((unsigned long) info) {
1124 		case (unsigned long) SEND_SIG_NOINFO:
1125 			clear_siginfo(&q->info);
1126 			q->info.si_signo = sig;
1127 			q->info.si_errno = 0;
1128 			q->info.si_code = SI_USER;
1129 			q->info.si_pid = task_tgid_nr_ns(current,
1130 							task_active_pid_ns(t));
1131 			rcu_read_lock();
1132 			q->info.si_uid =
1133 				from_kuid_munged(task_cred_xxx(t, user_ns),
1134 						 current_uid());
1135 			rcu_read_unlock();
1136 			break;
1137 		case (unsigned long) SEND_SIG_PRIV:
1138 			clear_siginfo(&q->info);
1139 			q->info.si_signo = sig;
1140 			q->info.si_errno = 0;
1141 			q->info.si_code = SI_KERNEL;
1142 			q->info.si_pid = 0;
1143 			q->info.si_uid = 0;
1144 			break;
1145 		default:
1146 			copy_siginfo(&q->info, info);
1147 			break;
1148 		}
1149 	} else if (!is_si_special(info) &&
1150 		   sig >= SIGRTMIN && info->si_code != SI_USER) {
1151 		/*
1152 		 * Queue overflow, abort.  We may abort if the
1153 		 * signal was rt and sent by user using something
1154 		 * other than kill().
1155 		 */
1156 		result = TRACE_SIGNAL_OVERFLOW_FAIL;
1157 		ret = -EAGAIN;
1158 		goto ret;
1159 	} else {
1160 		/*
1161 		 * This is a silent loss of information.  We still
1162 		 * send the signal, but the *info bits are lost.
1163 		 */
1164 		result = TRACE_SIGNAL_LOSE_INFO;
1165 	}
1166 
1167 out_set:
1168 	signalfd_notify(t, sig);
1169 	sigaddset(&pending->signal, sig);
1170 
1171 	/* Let multiprocess signals appear after on-going forks */
1172 	if (type > PIDTYPE_TGID) {
1173 		struct multiprocess_signals *delayed;
1174 		hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1175 			sigset_t *signal = &delayed->signal;
1176 			/* Can't queue both a stop and a continue signal */
1177 			if (sig == SIGCONT)
1178 				sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1179 			else if (sig_kernel_stop(sig))
1180 				sigdelset(signal, SIGCONT);
1181 			sigaddset(signal, sig);
1182 		}
1183 	}
1184 
1185 	complete_signal(sig, t, type);
1186 ret:
1187 	trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1188 	return ret;
1189 }
1190 
1191 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1192 {
1193 	bool ret = false;
1194 	switch (siginfo_layout(info->si_signo, info->si_code)) {
1195 	case SIL_KILL:
1196 	case SIL_CHLD:
1197 	case SIL_RT:
1198 		ret = true;
1199 		break;
1200 	case SIL_TIMER:
1201 	case SIL_POLL:
1202 	case SIL_FAULT:
1203 	case SIL_FAULT_TRAPNO:
1204 	case SIL_FAULT_MCEERR:
1205 	case SIL_FAULT_BNDERR:
1206 	case SIL_FAULT_PKUERR:
1207 	case SIL_FAULT_PERF_EVENT:
1208 	case SIL_SYS:
1209 		ret = false;
1210 		break;
1211 	}
1212 	return ret;
1213 }
1214 
1215 static int send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1216 			enum pid_type type)
1217 {
1218 	/* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1219 	bool force = false;
1220 
1221 	if (info == SEND_SIG_NOINFO) {
1222 		/* Force if sent from an ancestor pid namespace */
1223 		force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1224 	} else if (info == SEND_SIG_PRIV) {
1225 		/* Don't ignore kernel generated signals */
1226 		force = true;
1227 	} else if (has_si_pid_and_uid(info)) {
1228 		/* SIGKILL and SIGSTOP is special or has ids */
1229 		struct user_namespace *t_user_ns;
1230 
1231 		rcu_read_lock();
1232 		t_user_ns = task_cred_xxx(t, user_ns);
1233 		if (current_user_ns() != t_user_ns) {
1234 			kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1235 			info->si_uid = from_kuid_munged(t_user_ns, uid);
1236 		}
1237 		rcu_read_unlock();
1238 
1239 		/* A kernel generated signal? */
1240 		force = (info->si_code == SI_KERNEL);
1241 
1242 		/* From an ancestor pid namespace? */
1243 		if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1244 			info->si_pid = 0;
1245 			force = true;
1246 		}
1247 	}
1248 	return __send_signal(sig, info, t, type, force);
1249 }
1250 
1251 static void print_fatal_signal(int signr)
1252 {
1253 	struct pt_regs *regs = signal_pt_regs();
1254 	pr_info("potentially unexpected fatal signal %d.\n", signr);
1255 
1256 #if defined(__i386__) && !defined(__arch_um__)
1257 	pr_info("code at %08lx: ", regs->ip);
1258 	{
1259 		int i;
1260 		for (i = 0; i < 16; i++) {
1261 			unsigned char insn;
1262 
1263 			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1264 				break;
1265 			pr_cont("%02x ", insn);
1266 		}
1267 	}
1268 	pr_cont("\n");
1269 #endif
1270 	preempt_disable();
1271 	show_regs(regs);
1272 	preempt_enable();
1273 }
1274 
1275 static int __init setup_print_fatal_signals(char *str)
1276 {
1277 	get_option (&str, &print_fatal_signals);
1278 
1279 	return 1;
1280 }
1281 
1282 __setup("print-fatal-signals=", setup_print_fatal_signals);
1283 
1284 int
1285 __group_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1286 {
1287 	return send_signal(sig, info, p, PIDTYPE_TGID);
1288 }
1289 
1290 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1291 			enum pid_type type)
1292 {
1293 	unsigned long flags;
1294 	int ret = -ESRCH;
1295 
1296 	if (lock_task_sighand(p, &flags)) {
1297 		ret = send_signal(sig, info, p, type);
1298 		unlock_task_sighand(p, &flags);
1299 	}
1300 
1301 	return ret;
1302 }
1303 
1304 enum sig_handler {
1305 	HANDLER_CURRENT, /* If reachable use the current handler */
1306 	HANDLER_SIG_DFL, /* Always use SIG_DFL handler semantics */
1307 	HANDLER_EXIT,	 /* Only visible as the process exit code */
1308 };
1309 
1310 /*
1311  * Force a signal that the process can't ignore: if necessary
1312  * we unblock the signal and change any SIG_IGN to SIG_DFL.
1313  *
1314  * Note: If we unblock the signal, we always reset it to SIG_DFL,
1315  * since we do not want to have a signal handler that was blocked
1316  * be invoked when user space had explicitly blocked it.
1317  *
1318  * We don't want to have recursive SIGSEGV's etc, for example,
1319  * that is why we also clear SIGNAL_UNKILLABLE.
1320  */
1321 static int
1322 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t,
1323 	enum sig_handler handler)
1324 {
1325 	unsigned long int flags;
1326 	int ret, blocked, ignored;
1327 	struct k_sigaction *action;
1328 	int sig = info->si_signo;
1329 
1330 	spin_lock_irqsave(&t->sighand->siglock, flags);
1331 	action = &t->sighand->action[sig-1];
1332 	ignored = action->sa.sa_handler == SIG_IGN;
1333 	blocked = sigismember(&t->blocked, sig);
1334 	if (blocked || ignored || (handler != HANDLER_CURRENT)) {
1335 		action->sa.sa_handler = SIG_DFL;
1336 		if (handler == HANDLER_EXIT)
1337 			action->sa.sa_flags |= SA_IMMUTABLE;
1338 		if (blocked) {
1339 			sigdelset(&t->blocked, sig);
1340 			recalc_sigpending_and_wake(t);
1341 		}
1342 	}
1343 	/*
1344 	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1345 	 * debugging to leave init killable. But HANDLER_EXIT is always fatal.
1346 	 */
1347 	if (action->sa.sa_handler == SIG_DFL &&
1348 	    (!t->ptrace || (handler == HANDLER_EXIT)))
1349 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1350 	ret = send_signal(sig, info, t, PIDTYPE_PID);
1351 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1352 
1353 	return ret;
1354 }
1355 
1356 int force_sig_info(struct kernel_siginfo *info)
1357 {
1358 	return force_sig_info_to_task(info, current, HANDLER_CURRENT);
1359 }
1360 
1361 /*
1362  * Nuke all other threads in the group.
1363  */
1364 int zap_other_threads(struct task_struct *p)
1365 {
1366 	struct task_struct *t = p;
1367 	int count = 0;
1368 
1369 	p->signal->group_stop_count = 0;
1370 
1371 	while_each_thread(p, t) {
1372 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1373 		count++;
1374 
1375 		/* Don't bother with already dead threads */
1376 		if (t->exit_state)
1377 			continue;
1378 		sigaddset(&t->pending.signal, SIGKILL);
1379 		signal_wake_up(t, 1);
1380 	}
1381 
1382 	return count;
1383 }
1384 
1385 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1386 					   unsigned long *flags)
1387 {
1388 	struct sighand_struct *sighand;
1389 
1390 	rcu_read_lock();
1391 	for (;;) {
1392 		sighand = rcu_dereference(tsk->sighand);
1393 		if (unlikely(sighand == NULL))
1394 			break;
1395 
1396 		/*
1397 		 * This sighand can be already freed and even reused, but
1398 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1399 		 * initializes ->siglock: this slab can't go away, it has
1400 		 * the same object type, ->siglock can't be reinitialized.
1401 		 *
1402 		 * We need to ensure that tsk->sighand is still the same
1403 		 * after we take the lock, we can race with de_thread() or
1404 		 * __exit_signal(). In the latter case the next iteration
1405 		 * must see ->sighand == NULL.
1406 		 */
1407 		spin_lock_irqsave(&sighand->siglock, *flags);
1408 		if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1409 			break;
1410 		spin_unlock_irqrestore(&sighand->siglock, *flags);
1411 	}
1412 	rcu_read_unlock();
1413 
1414 	return sighand;
1415 }
1416 
1417 #ifdef CONFIG_LOCKDEP
1418 void lockdep_assert_task_sighand_held(struct task_struct *task)
1419 {
1420 	struct sighand_struct *sighand;
1421 
1422 	rcu_read_lock();
1423 	sighand = rcu_dereference(task->sighand);
1424 	if (sighand)
1425 		lockdep_assert_held(&sighand->siglock);
1426 	else
1427 		WARN_ON_ONCE(1);
1428 	rcu_read_unlock();
1429 }
1430 #endif
1431 
1432 /*
1433  * send signal info to all the members of a group
1434  */
1435 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1436 			struct task_struct *p, enum pid_type type)
1437 {
1438 	int ret;
1439 
1440 	rcu_read_lock();
1441 	ret = check_kill_permission(sig, info, p);
1442 	rcu_read_unlock();
1443 
1444 	if (!ret && sig)
1445 		ret = do_send_sig_info(sig, info, p, type);
1446 
1447 	return ret;
1448 }
1449 
1450 /*
1451  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1452  * control characters do (^C, ^Z etc)
1453  * - the caller must hold at least a readlock on tasklist_lock
1454  */
1455 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1456 {
1457 	struct task_struct *p = NULL;
1458 	int retval, success;
1459 
1460 	success = 0;
1461 	retval = -ESRCH;
1462 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1463 		int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1464 		success |= !err;
1465 		retval = err;
1466 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1467 	return success ? 0 : retval;
1468 }
1469 
1470 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1471 {
1472 	int error = -ESRCH;
1473 	struct task_struct *p;
1474 
1475 	for (;;) {
1476 		rcu_read_lock();
1477 		p = pid_task(pid, PIDTYPE_PID);
1478 		if (p)
1479 			error = group_send_sig_info(sig, info, p, PIDTYPE_TGID);
1480 		rcu_read_unlock();
1481 		if (likely(!p || error != -ESRCH))
1482 			return error;
1483 
1484 		/*
1485 		 * The task was unhashed in between, try again.  If it
1486 		 * is dead, pid_task() will return NULL, if we race with
1487 		 * de_thread() it will find the new leader.
1488 		 */
1489 	}
1490 }
1491 
1492 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1493 {
1494 	int error;
1495 	rcu_read_lock();
1496 	error = kill_pid_info(sig, info, find_vpid(pid));
1497 	rcu_read_unlock();
1498 	return error;
1499 }
1500 
1501 static inline bool kill_as_cred_perm(const struct cred *cred,
1502 				     struct task_struct *target)
1503 {
1504 	const struct cred *pcred = __task_cred(target);
1505 
1506 	return uid_eq(cred->euid, pcred->suid) ||
1507 	       uid_eq(cred->euid, pcred->uid) ||
1508 	       uid_eq(cred->uid, pcred->suid) ||
1509 	       uid_eq(cred->uid, pcred->uid);
1510 }
1511 
1512 /*
1513  * The usb asyncio usage of siginfo is wrong.  The glibc support
1514  * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1515  * AKA after the generic fields:
1516  *	kernel_pid_t	si_pid;
1517  *	kernel_uid32_t	si_uid;
1518  *	sigval_t	si_value;
1519  *
1520  * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1521  * after the generic fields is:
1522  *	void __user 	*si_addr;
1523  *
1524  * This is a practical problem when there is a 64bit big endian kernel
1525  * and a 32bit userspace.  As the 32bit address will encoded in the low
1526  * 32bits of the pointer.  Those low 32bits will be stored at higher
1527  * address than appear in a 32 bit pointer.  So userspace will not
1528  * see the address it was expecting for it's completions.
1529  *
1530  * There is nothing in the encoding that can allow
1531  * copy_siginfo_to_user32 to detect this confusion of formats, so
1532  * handle this by requiring the caller of kill_pid_usb_asyncio to
1533  * notice when this situration takes place and to store the 32bit
1534  * pointer in sival_int, instead of sival_addr of the sigval_t addr
1535  * parameter.
1536  */
1537 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1538 			 struct pid *pid, const struct cred *cred)
1539 {
1540 	struct kernel_siginfo info;
1541 	struct task_struct *p;
1542 	unsigned long flags;
1543 	int ret = -EINVAL;
1544 
1545 	if (!valid_signal(sig))
1546 		return ret;
1547 
1548 	clear_siginfo(&info);
1549 	info.si_signo = sig;
1550 	info.si_errno = errno;
1551 	info.si_code = SI_ASYNCIO;
1552 	*((sigval_t *)&info.si_pid) = addr;
1553 
1554 	rcu_read_lock();
1555 	p = pid_task(pid, PIDTYPE_PID);
1556 	if (!p) {
1557 		ret = -ESRCH;
1558 		goto out_unlock;
1559 	}
1560 	if (!kill_as_cred_perm(cred, p)) {
1561 		ret = -EPERM;
1562 		goto out_unlock;
1563 	}
1564 	ret = security_task_kill(p, &info, sig, cred);
1565 	if (ret)
1566 		goto out_unlock;
1567 
1568 	if (sig) {
1569 		if (lock_task_sighand(p, &flags)) {
1570 			ret = __send_signal(sig, &info, p, PIDTYPE_TGID, false);
1571 			unlock_task_sighand(p, &flags);
1572 		} else
1573 			ret = -ESRCH;
1574 	}
1575 out_unlock:
1576 	rcu_read_unlock();
1577 	return ret;
1578 }
1579 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1580 
1581 /*
1582  * kill_something_info() interprets pid in interesting ways just like kill(2).
1583  *
1584  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1585  * is probably wrong.  Should make it like BSD or SYSV.
1586  */
1587 
1588 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1589 {
1590 	int ret;
1591 
1592 	if (pid > 0)
1593 		return kill_proc_info(sig, info, pid);
1594 
1595 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1596 	if (pid == INT_MIN)
1597 		return -ESRCH;
1598 
1599 	read_lock(&tasklist_lock);
1600 	if (pid != -1) {
1601 		ret = __kill_pgrp_info(sig, info,
1602 				pid ? find_vpid(-pid) : task_pgrp(current));
1603 	} else {
1604 		int retval = 0, count = 0;
1605 		struct task_struct * p;
1606 
1607 		for_each_process(p) {
1608 			if (task_pid_vnr(p) > 1 &&
1609 					!same_thread_group(p, current)) {
1610 				int err = group_send_sig_info(sig, info, p,
1611 							      PIDTYPE_MAX);
1612 				++count;
1613 				if (err != -EPERM)
1614 					retval = err;
1615 			}
1616 		}
1617 		ret = count ? retval : -ESRCH;
1618 	}
1619 	read_unlock(&tasklist_lock);
1620 
1621 	return ret;
1622 }
1623 
1624 /*
1625  * These are for backward compatibility with the rest of the kernel source.
1626  */
1627 
1628 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1629 {
1630 	/*
1631 	 * Make sure legacy kernel users don't send in bad values
1632 	 * (normal paths check this in check_kill_permission).
1633 	 */
1634 	if (!valid_signal(sig))
1635 		return -EINVAL;
1636 
1637 	return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1638 }
1639 EXPORT_SYMBOL(send_sig_info);
1640 
1641 #define __si_special(priv) \
1642 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1643 
1644 int
1645 send_sig(int sig, struct task_struct *p, int priv)
1646 {
1647 	return send_sig_info(sig, __si_special(priv), p);
1648 }
1649 EXPORT_SYMBOL(send_sig);
1650 
1651 void force_sig(int sig)
1652 {
1653 	struct kernel_siginfo info;
1654 
1655 	clear_siginfo(&info);
1656 	info.si_signo = sig;
1657 	info.si_errno = 0;
1658 	info.si_code = SI_KERNEL;
1659 	info.si_pid = 0;
1660 	info.si_uid = 0;
1661 	force_sig_info(&info);
1662 }
1663 EXPORT_SYMBOL(force_sig);
1664 
1665 void force_fatal_sig(int sig)
1666 {
1667 	struct kernel_siginfo info;
1668 
1669 	clear_siginfo(&info);
1670 	info.si_signo = sig;
1671 	info.si_errno = 0;
1672 	info.si_code = SI_KERNEL;
1673 	info.si_pid = 0;
1674 	info.si_uid = 0;
1675 	force_sig_info_to_task(&info, current, HANDLER_SIG_DFL);
1676 }
1677 
1678 void force_exit_sig(int sig)
1679 {
1680 	struct kernel_siginfo info;
1681 
1682 	clear_siginfo(&info);
1683 	info.si_signo = sig;
1684 	info.si_errno = 0;
1685 	info.si_code = SI_KERNEL;
1686 	info.si_pid = 0;
1687 	info.si_uid = 0;
1688 	force_sig_info_to_task(&info, current, HANDLER_EXIT);
1689 }
1690 
1691 /*
1692  * When things go south during signal handling, we
1693  * will force a SIGSEGV. And if the signal that caused
1694  * the problem was already a SIGSEGV, we'll want to
1695  * make sure we don't even try to deliver the signal..
1696  */
1697 void force_sigsegv(int sig)
1698 {
1699 	if (sig == SIGSEGV)
1700 		force_fatal_sig(SIGSEGV);
1701 	else
1702 		force_sig(SIGSEGV);
1703 }
1704 
1705 int force_sig_fault_to_task(int sig, int code, void __user *addr
1706 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1707 	, struct task_struct *t)
1708 {
1709 	struct kernel_siginfo info;
1710 
1711 	clear_siginfo(&info);
1712 	info.si_signo = sig;
1713 	info.si_errno = 0;
1714 	info.si_code  = code;
1715 	info.si_addr  = addr;
1716 #ifdef __ia64__
1717 	info.si_imm = imm;
1718 	info.si_flags = flags;
1719 	info.si_isr = isr;
1720 #endif
1721 	return force_sig_info_to_task(&info, t, HANDLER_CURRENT);
1722 }
1723 
1724 int force_sig_fault(int sig, int code, void __user *addr
1725 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr))
1726 {
1727 	return force_sig_fault_to_task(sig, code, addr
1728 				       ___ARCH_SI_IA64(imm, flags, isr), current);
1729 }
1730 
1731 int send_sig_fault(int sig, int code, void __user *addr
1732 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1733 	, struct task_struct *t)
1734 {
1735 	struct kernel_siginfo info;
1736 
1737 	clear_siginfo(&info);
1738 	info.si_signo = sig;
1739 	info.si_errno = 0;
1740 	info.si_code  = code;
1741 	info.si_addr  = addr;
1742 #ifdef __ia64__
1743 	info.si_imm = imm;
1744 	info.si_flags = flags;
1745 	info.si_isr = isr;
1746 #endif
1747 	return send_sig_info(info.si_signo, &info, t);
1748 }
1749 
1750 int force_sig_mceerr(int code, void __user *addr, short lsb)
1751 {
1752 	struct kernel_siginfo info;
1753 
1754 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1755 	clear_siginfo(&info);
1756 	info.si_signo = SIGBUS;
1757 	info.si_errno = 0;
1758 	info.si_code = code;
1759 	info.si_addr = addr;
1760 	info.si_addr_lsb = lsb;
1761 	return force_sig_info(&info);
1762 }
1763 
1764 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1765 {
1766 	struct kernel_siginfo info;
1767 
1768 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1769 	clear_siginfo(&info);
1770 	info.si_signo = SIGBUS;
1771 	info.si_errno = 0;
1772 	info.si_code = code;
1773 	info.si_addr = addr;
1774 	info.si_addr_lsb = lsb;
1775 	return send_sig_info(info.si_signo, &info, t);
1776 }
1777 EXPORT_SYMBOL(send_sig_mceerr);
1778 
1779 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1780 {
1781 	struct kernel_siginfo info;
1782 
1783 	clear_siginfo(&info);
1784 	info.si_signo = SIGSEGV;
1785 	info.si_errno = 0;
1786 	info.si_code  = SEGV_BNDERR;
1787 	info.si_addr  = addr;
1788 	info.si_lower = lower;
1789 	info.si_upper = upper;
1790 	return force_sig_info(&info);
1791 }
1792 
1793 #ifdef SEGV_PKUERR
1794 int force_sig_pkuerr(void __user *addr, u32 pkey)
1795 {
1796 	struct kernel_siginfo info;
1797 
1798 	clear_siginfo(&info);
1799 	info.si_signo = SIGSEGV;
1800 	info.si_errno = 0;
1801 	info.si_code  = SEGV_PKUERR;
1802 	info.si_addr  = addr;
1803 	info.si_pkey  = pkey;
1804 	return force_sig_info(&info);
1805 }
1806 #endif
1807 
1808 int send_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 	/*
1821 	 * Signals generated by perf events should not terminate the whole
1822 	 * process if SIGTRAP is blocked, however, delivering the signal
1823 	 * asynchronously is better than not delivering at all. But tell user
1824 	 * space if the signal was asynchronous, so it can clearly be
1825 	 * distinguished from normal synchronous ones.
1826 	 */
1827 	info.si_perf_flags = sigismember(&current->blocked, info.si_signo) ?
1828 				     TRAP_PERF_FLAG_ASYNC :
1829 				     0;
1830 
1831 	return send_sig_info(info.si_signo, &info, current);
1832 }
1833 
1834 /**
1835  * force_sig_seccomp - signals the task to allow in-process syscall emulation
1836  * @syscall: syscall number to send to userland
1837  * @reason: filter-supplied reason code to send to userland (via si_errno)
1838  * @force_coredump: true to trigger a coredump
1839  *
1840  * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
1841  */
1842 int force_sig_seccomp(int syscall, int reason, bool force_coredump)
1843 {
1844 	struct kernel_siginfo info;
1845 
1846 	clear_siginfo(&info);
1847 	info.si_signo = SIGSYS;
1848 	info.si_code = SYS_SECCOMP;
1849 	info.si_call_addr = (void __user *)KSTK_EIP(current);
1850 	info.si_errno = reason;
1851 	info.si_arch = syscall_get_arch(current);
1852 	info.si_syscall = syscall;
1853 	return force_sig_info_to_task(&info, current,
1854 		force_coredump ? HANDLER_EXIT : HANDLER_CURRENT);
1855 }
1856 
1857 /* For the crazy architectures that include trap information in
1858  * the errno field, instead of an actual errno value.
1859  */
1860 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1861 {
1862 	struct kernel_siginfo info;
1863 
1864 	clear_siginfo(&info);
1865 	info.si_signo = SIGTRAP;
1866 	info.si_errno = errno;
1867 	info.si_code  = TRAP_HWBKPT;
1868 	info.si_addr  = addr;
1869 	return force_sig_info(&info);
1870 }
1871 
1872 /* For the rare architectures that include trap information using
1873  * si_trapno.
1874  */
1875 int force_sig_fault_trapno(int sig, int code, void __user *addr, int trapno)
1876 {
1877 	struct kernel_siginfo info;
1878 
1879 	clear_siginfo(&info);
1880 	info.si_signo = sig;
1881 	info.si_errno = 0;
1882 	info.si_code  = code;
1883 	info.si_addr  = addr;
1884 	info.si_trapno = trapno;
1885 	return force_sig_info(&info);
1886 }
1887 
1888 /* For the rare architectures that include trap information using
1889  * si_trapno.
1890  */
1891 int send_sig_fault_trapno(int sig, int code, void __user *addr, int trapno,
1892 			  struct task_struct *t)
1893 {
1894 	struct kernel_siginfo info;
1895 
1896 	clear_siginfo(&info);
1897 	info.si_signo = sig;
1898 	info.si_errno = 0;
1899 	info.si_code  = code;
1900 	info.si_addr  = addr;
1901 	info.si_trapno = trapno;
1902 	return send_sig_info(info.si_signo, &info, t);
1903 }
1904 
1905 int kill_pgrp(struct pid *pid, int sig, int priv)
1906 {
1907 	int ret;
1908 
1909 	read_lock(&tasklist_lock);
1910 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1911 	read_unlock(&tasklist_lock);
1912 
1913 	return ret;
1914 }
1915 EXPORT_SYMBOL(kill_pgrp);
1916 
1917 int kill_pid(struct pid *pid, int sig, int priv)
1918 {
1919 	return kill_pid_info(sig, __si_special(priv), pid);
1920 }
1921 EXPORT_SYMBOL(kill_pid);
1922 
1923 /*
1924  * These functions support sending signals using preallocated sigqueue
1925  * structures.  This is needed "because realtime applications cannot
1926  * afford to lose notifications of asynchronous events, like timer
1927  * expirations or I/O completions".  In the case of POSIX Timers
1928  * we allocate the sigqueue structure from the timer_create.  If this
1929  * allocation fails we are able to report the failure to the application
1930  * with an EAGAIN error.
1931  */
1932 struct sigqueue *sigqueue_alloc(void)
1933 {
1934 	return __sigqueue_alloc(-1, current, GFP_KERNEL, 0, SIGQUEUE_PREALLOC);
1935 }
1936 
1937 void sigqueue_free(struct sigqueue *q)
1938 {
1939 	unsigned long flags;
1940 	spinlock_t *lock = &current->sighand->siglock;
1941 
1942 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1943 	/*
1944 	 * We must hold ->siglock while testing q->list
1945 	 * to serialize with collect_signal() or with
1946 	 * __exit_signal()->flush_sigqueue().
1947 	 */
1948 	spin_lock_irqsave(lock, flags);
1949 	q->flags &= ~SIGQUEUE_PREALLOC;
1950 	/*
1951 	 * If it is queued it will be freed when dequeued,
1952 	 * like the "regular" sigqueue.
1953 	 */
1954 	if (!list_empty(&q->list))
1955 		q = NULL;
1956 	spin_unlock_irqrestore(lock, flags);
1957 
1958 	if (q)
1959 		__sigqueue_free(q);
1960 }
1961 
1962 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1963 {
1964 	int sig = q->info.si_signo;
1965 	struct sigpending *pending;
1966 	struct task_struct *t;
1967 	unsigned long flags;
1968 	int ret, result;
1969 
1970 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1971 
1972 	ret = -1;
1973 	rcu_read_lock();
1974 	t = pid_task(pid, type);
1975 	if (!t || !likely(lock_task_sighand(t, &flags)))
1976 		goto ret;
1977 
1978 	ret = 1; /* the signal is ignored */
1979 	result = TRACE_SIGNAL_IGNORED;
1980 	if (!prepare_signal(sig, t, false))
1981 		goto out;
1982 
1983 	ret = 0;
1984 	if (unlikely(!list_empty(&q->list))) {
1985 		/*
1986 		 * If an SI_TIMER entry is already queue just increment
1987 		 * the overrun count.
1988 		 */
1989 		BUG_ON(q->info.si_code != SI_TIMER);
1990 		q->info.si_overrun++;
1991 		result = TRACE_SIGNAL_ALREADY_PENDING;
1992 		goto out;
1993 	}
1994 	q->info.si_overrun = 0;
1995 
1996 	signalfd_notify(t, sig);
1997 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1998 	list_add_tail(&q->list, &pending->list);
1999 	sigaddset(&pending->signal, sig);
2000 	complete_signal(sig, t, type);
2001 	result = TRACE_SIGNAL_DELIVERED;
2002 out:
2003 	trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
2004 	unlock_task_sighand(t, &flags);
2005 ret:
2006 	rcu_read_unlock();
2007 	return ret;
2008 }
2009 
2010 static void do_notify_pidfd(struct task_struct *task)
2011 {
2012 	struct pid *pid;
2013 
2014 	WARN_ON(task->exit_state == 0);
2015 	pid = task_pid(task);
2016 	wake_up_all(&pid->wait_pidfd);
2017 }
2018 
2019 /*
2020  * Let a parent know about the death of a child.
2021  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
2022  *
2023  * Returns true if our parent ignored us and so we've switched to
2024  * self-reaping.
2025  */
2026 bool do_notify_parent(struct task_struct *tsk, int sig)
2027 {
2028 	struct kernel_siginfo info;
2029 	unsigned long flags;
2030 	struct sighand_struct *psig;
2031 	bool autoreap = false;
2032 	u64 utime, stime;
2033 
2034 	BUG_ON(sig == -1);
2035 
2036  	/* do_notify_parent_cldstop should have been called instead.  */
2037  	BUG_ON(task_is_stopped_or_traced(tsk));
2038 
2039 	BUG_ON(!tsk->ptrace &&
2040 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
2041 
2042 	/* Wake up all pidfd waiters */
2043 	do_notify_pidfd(tsk);
2044 
2045 	if (sig != SIGCHLD) {
2046 		/*
2047 		 * This is only possible if parent == real_parent.
2048 		 * Check if it has changed security domain.
2049 		 */
2050 		if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
2051 			sig = SIGCHLD;
2052 	}
2053 
2054 	clear_siginfo(&info);
2055 	info.si_signo = sig;
2056 	info.si_errno = 0;
2057 	/*
2058 	 * We are under tasklist_lock here so our parent is tied to
2059 	 * us and cannot change.
2060 	 *
2061 	 * task_active_pid_ns will always return the same pid namespace
2062 	 * until a task passes through release_task.
2063 	 *
2064 	 * write_lock() currently calls preempt_disable() which is the
2065 	 * same as rcu_read_lock(), but according to Oleg, this is not
2066 	 * correct to rely on this
2067 	 */
2068 	rcu_read_lock();
2069 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
2070 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
2071 				       task_uid(tsk));
2072 	rcu_read_unlock();
2073 
2074 	task_cputime(tsk, &utime, &stime);
2075 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
2076 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
2077 
2078 	info.si_status = tsk->exit_code & 0x7f;
2079 	if (tsk->exit_code & 0x80)
2080 		info.si_code = CLD_DUMPED;
2081 	else if (tsk->exit_code & 0x7f)
2082 		info.si_code = CLD_KILLED;
2083 	else {
2084 		info.si_code = CLD_EXITED;
2085 		info.si_status = tsk->exit_code >> 8;
2086 	}
2087 
2088 	psig = tsk->parent->sighand;
2089 	spin_lock_irqsave(&psig->siglock, flags);
2090 	if (!tsk->ptrace && sig == SIGCHLD &&
2091 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
2092 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
2093 		/*
2094 		 * We are exiting and our parent doesn't care.  POSIX.1
2095 		 * defines special semantics for setting SIGCHLD to SIG_IGN
2096 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
2097 		 * automatically and not left for our parent's wait4 call.
2098 		 * Rather than having the parent do it as a magic kind of
2099 		 * signal handler, we just set this to tell do_exit that we
2100 		 * can be cleaned up without becoming a zombie.  Note that
2101 		 * we still call __wake_up_parent in this case, because a
2102 		 * blocked sys_wait4 might now return -ECHILD.
2103 		 *
2104 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
2105 		 * is implementation-defined: we do (if you don't want
2106 		 * it, just use SIG_IGN instead).
2107 		 */
2108 		autoreap = true;
2109 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2110 			sig = 0;
2111 	}
2112 	/*
2113 	 * Send with __send_signal as si_pid and si_uid are in the
2114 	 * parent's namespaces.
2115 	 */
2116 	if (valid_signal(sig) && sig)
2117 		__send_signal(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2118 	__wake_up_parent(tsk, tsk->parent);
2119 	spin_unlock_irqrestore(&psig->siglock, flags);
2120 
2121 	return autoreap;
2122 }
2123 
2124 /**
2125  * do_notify_parent_cldstop - notify parent of stopped/continued state change
2126  * @tsk: task reporting the state change
2127  * @for_ptracer: the notification is for ptracer
2128  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2129  *
2130  * Notify @tsk's parent that the stopped/continued state has changed.  If
2131  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2132  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2133  *
2134  * CONTEXT:
2135  * Must be called with tasklist_lock at least read locked.
2136  */
2137 static void do_notify_parent_cldstop(struct task_struct *tsk,
2138 				     bool for_ptracer, int why)
2139 {
2140 	struct kernel_siginfo info;
2141 	unsigned long flags;
2142 	struct task_struct *parent;
2143 	struct sighand_struct *sighand;
2144 	u64 utime, stime;
2145 
2146 	if (for_ptracer) {
2147 		parent = tsk->parent;
2148 	} else {
2149 		tsk = tsk->group_leader;
2150 		parent = tsk->real_parent;
2151 	}
2152 
2153 	clear_siginfo(&info);
2154 	info.si_signo = SIGCHLD;
2155 	info.si_errno = 0;
2156 	/*
2157 	 * see comment in do_notify_parent() about the following 4 lines
2158 	 */
2159 	rcu_read_lock();
2160 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2161 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2162 	rcu_read_unlock();
2163 
2164 	task_cputime(tsk, &utime, &stime);
2165 	info.si_utime = nsec_to_clock_t(utime);
2166 	info.si_stime = nsec_to_clock_t(stime);
2167 
2168  	info.si_code = why;
2169  	switch (why) {
2170  	case CLD_CONTINUED:
2171  		info.si_status = SIGCONT;
2172  		break;
2173  	case CLD_STOPPED:
2174  		info.si_status = tsk->signal->group_exit_code & 0x7f;
2175  		break;
2176  	case CLD_TRAPPED:
2177  		info.si_status = tsk->exit_code & 0x7f;
2178  		break;
2179  	default:
2180  		BUG();
2181  	}
2182 
2183 	sighand = parent->sighand;
2184 	spin_lock_irqsave(&sighand->siglock, flags);
2185 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2186 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2187 		__group_send_sig_info(SIGCHLD, &info, parent);
2188 	/*
2189 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2190 	 */
2191 	__wake_up_parent(tsk, parent);
2192 	spin_unlock_irqrestore(&sighand->siglock, flags);
2193 }
2194 
2195 /*
2196  * This must be called with current->sighand->siglock held.
2197  *
2198  * This should be the path for all ptrace stops.
2199  * We always set current->last_siginfo while stopped here.
2200  * That makes it a way to test a stopped process for
2201  * being ptrace-stopped vs being job-control-stopped.
2202  *
2203  * Returns the signal the ptracer requested the code resume
2204  * with.  If the code did not stop because the tracer is gone,
2205  * the stop signal remains unchanged unless clear_code.
2206  */
2207 static int ptrace_stop(int exit_code, int why, int clear_code,
2208 			unsigned long message, kernel_siginfo_t *info)
2209 	__releases(&current->sighand->siglock)
2210 	__acquires(&current->sighand->siglock)
2211 {
2212 	bool gstop_done = false;
2213 	bool read_code = true;
2214 
2215 	if (arch_ptrace_stop_needed()) {
2216 		/*
2217 		 * The arch code has something special to do before a
2218 		 * ptrace stop.  This is allowed to block, e.g. for faults
2219 		 * on user stack pages.  We can't keep the siglock while
2220 		 * calling arch_ptrace_stop, so we must release it now.
2221 		 * To preserve proper semantics, we must do this before
2222 		 * any signal bookkeeping like checking group_stop_count.
2223 		 */
2224 		spin_unlock_irq(&current->sighand->siglock);
2225 		arch_ptrace_stop();
2226 		spin_lock_irq(&current->sighand->siglock);
2227 	}
2228 
2229 	/*
2230 	 * schedule() will not sleep if there is a pending signal that
2231 	 * can awaken the task.
2232 	 */
2233 	set_special_state(TASK_TRACED);
2234 
2235 	/*
2236 	 * We're committing to trapping.  TRACED should be visible before
2237 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2238 	 * Also, transition to TRACED and updates to ->jobctl should be
2239 	 * atomic with respect to siglock and should be done after the arch
2240 	 * hook as siglock is released and regrabbed across it.
2241 	 *
2242 	 *     TRACER				    TRACEE
2243 	 *
2244 	 *     ptrace_attach()
2245 	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2246 	 *     do_wait()
2247 	 *       set_current_state()                smp_wmb();
2248 	 *       ptrace_do_wait()
2249 	 *         wait_task_stopped()
2250 	 *           task_stopped_code()
2251 	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2252 	 */
2253 	smp_wmb();
2254 
2255 	current->ptrace_message = message;
2256 	current->last_siginfo = info;
2257 	current->exit_code = exit_code;
2258 
2259 	/*
2260 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2261 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2262 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2263 	 * could be clear now.  We act as if SIGCONT is received after
2264 	 * TASK_TRACED is entered - ignore it.
2265 	 */
2266 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2267 		gstop_done = task_participate_group_stop(current);
2268 
2269 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2270 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2271 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2272 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2273 
2274 	/* entering a trap, clear TRAPPING */
2275 	task_clear_jobctl_trapping(current);
2276 
2277 	spin_unlock_irq(&current->sighand->siglock);
2278 	read_lock(&tasklist_lock);
2279 	if (likely(current->ptrace)) {
2280 		/*
2281 		 * Notify parents of the stop.
2282 		 *
2283 		 * While ptraced, there are two parents - the ptracer and
2284 		 * the real_parent of the group_leader.  The ptracer should
2285 		 * know about every stop while the real parent is only
2286 		 * interested in the completion of group stop.  The states
2287 		 * for the two don't interact with each other.  Notify
2288 		 * separately unless they're gonna be duplicates.
2289 		 */
2290 		do_notify_parent_cldstop(current, true, why);
2291 		if (gstop_done && ptrace_reparented(current))
2292 			do_notify_parent_cldstop(current, false, why);
2293 
2294 		/*
2295 		 * Don't want to allow preemption here, because
2296 		 * sys_ptrace() needs this task to be inactive.
2297 		 *
2298 		 * XXX: implement read_unlock_no_resched().
2299 		 */
2300 		preempt_disable();
2301 		read_unlock(&tasklist_lock);
2302 		cgroup_enter_frozen();
2303 		preempt_enable_no_resched();
2304 		freezable_schedule();
2305 		cgroup_leave_frozen(true);
2306 	} else {
2307 		/*
2308 		 * By the time we got the lock, our tracer went away.
2309 		 * Don't drop the lock yet, another tracer may come.
2310 		 *
2311 		 * If @gstop_done, the ptracer went away between group stop
2312 		 * completion and here.  During detach, it would have set
2313 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
2314 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
2315 		 * the real parent of the group stop completion is enough.
2316 		 */
2317 		if (gstop_done)
2318 			do_notify_parent_cldstop(current, false, why);
2319 
2320 		/* tasklist protects us from ptrace_freeze_traced() */
2321 		__set_current_state(TASK_RUNNING);
2322 		read_code = false;
2323 		if (clear_code)
2324 			exit_code = 0;
2325 		read_unlock(&tasklist_lock);
2326 	}
2327 
2328 	/*
2329 	 * We are back.  Now reacquire the siglock before touching
2330 	 * last_siginfo, so that we are sure to have synchronized with
2331 	 * any signal-sending on another CPU that wants to examine it.
2332 	 */
2333 	spin_lock_irq(&current->sighand->siglock);
2334 	if (read_code)
2335 		exit_code = current->exit_code;
2336 	current->last_siginfo = NULL;
2337 	current->ptrace_message = 0;
2338 	current->exit_code = 0;
2339 
2340 	/* LISTENING can be set only during STOP traps, clear it */
2341 	current->jobctl &= ~JOBCTL_LISTENING;
2342 
2343 	/*
2344 	 * Queued signals ignored us while we were stopped for tracing.
2345 	 * So check for any that we should take before resuming user mode.
2346 	 * This sets TIF_SIGPENDING, but never clears it.
2347 	 */
2348 	recalc_sigpending_tsk(current);
2349 	return exit_code;
2350 }
2351 
2352 static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
2353 {
2354 	kernel_siginfo_t info;
2355 
2356 	clear_siginfo(&info);
2357 	info.si_signo = signr;
2358 	info.si_code = exit_code;
2359 	info.si_pid = task_pid_vnr(current);
2360 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2361 
2362 	/* Let the debugger run.  */
2363 	return ptrace_stop(exit_code, why, 1, message, &info);
2364 }
2365 
2366 int ptrace_notify(int exit_code, unsigned long message)
2367 {
2368 	int signr;
2369 
2370 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2371 	if (unlikely(task_work_pending(current)))
2372 		task_work_run();
2373 
2374 	spin_lock_irq(&current->sighand->siglock);
2375 	signr = ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
2376 	spin_unlock_irq(&current->sighand->siglock);
2377 	return signr;
2378 }
2379 
2380 /**
2381  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2382  * @signr: signr causing group stop if initiating
2383  *
2384  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2385  * and participate in it.  If already set, participate in the existing
2386  * group stop.  If participated in a group stop (and thus slept), %true is
2387  * returned with siglock released.
2388  *
2389  * If ptraced, this function doesn't handle stop itself.  Instead,
2390  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2391  * untouched.  The caller must ensure that INTERRUPT trap handling takes
2392  * places afterwards.
2393  *
2394  * CONTEXT:
2395  * Must be called with @current->sighand->siglock held, which is released
2396  * on %true return.
2397  *
2398  * RETURNS:
2399  * %false if group stop is already cancelled or ptrace trap is scheduled.
2400  * %true if participated in group stop.
2401  */
2402 static bool do_signal_stop(int signr)
2403 	__releases(&current->sighand->siglock)
2404 {
2405 	struct signal_struct *sig = current->signal;
2406 
2407 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2408 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2409 		struct task_struct *t;
2410 
2411 		/* signr will be recorded in task->jobctl for retries */
2412 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2413 
2414 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2415 		    unlikely(sig->flags & SIGNAL_GROUP_EXIT) ||
2416 		    unlikely(sig->group_exec_task))
2417 			return false;
2418 		/*
2419 		 * There is no group stop already in progress.  We must
2420 		 * initiate one now.
2421 		 *
2422 		 * While ptraced, a task may be resumed while group stop is
2423 		 * still in effect and then receive a stop signal and
2424 		 * initiate another group stop.  This deviates from the
2425 		 * usual behavior as two consecutive stop signals can't
2426 		 * cause two group stops when !ptraced.  That is why we
2427 		 * also check !task_is_stopped(t) below.
2428 		 *
2429 		 * The condition can be distinguished by testing whether
2430 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2431 		 * group_exit_code in such case.
2432 		 *
2433 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2434 		 * an intervening stop signal is required to cause two
2435 		 * continued events regardless of ptrace.
2436 		 */
2437 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2438 			sig->group_exit_code = signr;
2439 
2440 		sig->group_stop_count = 0;
2441 
2442 		if (task_set_jobctl_pending(current, signr | gstop))
2443 			sig->group_stop_count++;
2444 
2445 		t = current;
2446 		while_each_thread(current, t) {
2447 			/*
2448 			 * Setting state to TASK_STOPPED for a group
2449 			 * stop is always done with the siglock held,
2450 			 * so this check has no races.
2451 			 */
2452 			if (!task_is_stopped(t) &&
2453 			    task_set_jobctl_pending(t, signr | gstop)) {
2454 				sig->group_stop_count++;
2455 				if (likely(!(t->ptrace & PT_SEIZED)))
2456 					signal_wake_up(t, 0);
2457 				else
2458 					ptrace_trap_notify(t);
2459 			}
2460 		}
2461 	}
2462 
2463 	if (likely(!current->ptrace)) {
2464 		int notify = 0;
2465 
2466 		/*
2467 		 * If there are no other threads in the group, or if there
2468 		 * is a group stop in progress and we are the last to stop,
2469 		 * report to the parent.
2470 		 */
2471 		if (task_participate_group_stop(current))
2472 			notify = CLD_STOPPED;
2473 
2474 		set_special_state(TASK_STOPPED);
2475 		spin_unlock_irq(&current->sighand->siglock);
2476 
2477 		/*
2478 		 * Notify the parent of the group stop completion.  Because
2479 		 * we're not holding either the siglock or tasklist_lock
2480 		 * here, ptracer may attach inbetween; however, this is for
2481 		 * group stop and should always be delivered to the real
2482 		 * parent of the group leader.  The new ptracer will get
2483 		 * its notification when this task transitions into
2484 		 * TASK_TRACED.
2485 		 */
2486 		if (notify) {
2487 			read_lock(&tasklist_lock);
2488 			do_notify_parent_cldstop(current, false, notify);
2489 			read_unlock(&tasklist_lock);
2490 		}
2491 
2492 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2493 		cgroup_enter_frozen();
2494 		freezable_schedule();
2495 		return true;
2496 	} else {
2497 		/*
2498 		 * While ptraced, group stop is handled by STOP trap.
2499 		 * Schedule it and let the caller deal with it.
2500 		 */
2501 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2502 		return false;
2503 	}
2504 }
2505 
2506 /**
2507  * do_jobctl_trap - take care of ptrace jobctl traps
2508  *
2509  * When PT_SEIZED, it's used for both group stop and explicit
2510  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2511  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2512  * the stop signal; otherwise, %SIGTRAP.
2513  *
2514  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2515  * number as exit_code and no siginfo.
2516  *
2517  * CONTEXT:
2518  * Must be called with @current->sighand->siglock held, which may be
2519  * released and re-acquired before returning with intervening sleep.
2520  */
2521 static void do_jobctl_trap(void)
2522 {
2523 	struct signal_struct *signal = current->signal;
2524 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2525 
2526 	if (current->ptrace & PT_SEIZED) {
2527 		if (!signal->group_stop_count &&
2528 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2529 			signr = SIGTRAP;
2530 		WARN_ON_ONCE(!signr);
2531 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2532 				 CLD_STOPPED, 0);
2533 	} else {
2534 		WARN_ON_ONCE(!signr);
2535 		ptrace_stop(signr, CLD_STOPPED, 0, 0, NULL);
2536 	}
2537 }
2538 
2539 /**
2540  * do_freezer_trap - handle the freezer jobctl trap
2541  *
2542  * Puts the task into frozen state, if only the task is not about to quit.
2543  * In this case it drops JOBCTL_TRAP_FREEZE.
2544  *
2545  * CONTEXT:
2546  * Must be called with @current->sighand->siglock held,
2547  * which is always released before returning.
2548  */
2549 static void do_freezer_trap(void)
2550 	__releases(&current->sighand->siglock)
2551 {
2552 	/*
2553 	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2554 	 * let's make another loop to give it a chance to be handled.
2555 	 * In any case, we'll return back.
2556 	 */
2557 	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2558 	     JOBCTL_TRAP_FREEZE) {
2559 		spin_unlock_irq(&current->sighand->siglock);
2560 		return;
2561 	}
2562 
2563 	/*
2564 	 * Now we're sure that there is no pending fatal signal and no
2565 	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2566 	 * immediately (if there is a non-fatal signal pending), and
2567 	 * put the task into sleep.
2568 	 */
2569 	__set_current_state(TASK_INTERRUPTIBLE);
2570 	clear_thread_flag(TIF_SIGPENDING);
2571 	spin_unlock_irq(&current->sighand->siglock);
2572 	cgroup_enter_frozen();
2573 	freezable_schedule();
2574 }
2575 
2576 static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
2577 {
2578 	/*
2579 	 * We do not check sig_kernel_stop(signr) but set this marker
2580 	 * unconditionally because we do not know whether debugger will
2581 	 * change signr. This flag has no meaning unless we are going
2582 	 * to stop after return from ptrace_stop(). In this case it will
2583 	 * be checked in do_signal_stop(), we should only stop if it was
2584 	 * not cleared by SIGCONT while we were sleeping. See also the
2585 	 * comment in dequeue_signal().
2586 	 */
2587 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2588 	signr = ptrace_stop(signr, CLD_TRAPPED, 0, 0, info);
2589 
2590 	/* We're back.  Did the debugger cancel the sig?  */
2591 	if (signr == 0)
2592 		return signr;
2593 
2594 	/*
2595 	 * Update the siginfo structure if the signal has
2596 	 * changed.  If the debugger wanted something
2597 	 * specific in the siginfo structure then it should
2598 	 * have updated *info via PTRACE_SETSIGINFO.
2599 	 */
2600 	if (signr != info->si_signo) {
2601 		clear_siginfo(info);
2602 		info->si_signo = signr;
2603 		info->si_errno = 0;
2604 		info->si_code = SI_USER;
2605 		rcu_read_lock();
2606 		info->si_pid = task_pid_vnr(current->parent);
2607 		info->si_uid = from_kuid_munged(current_user_ns(),
2608 						task_uid(current->parent));
2609 		rcu_read_unlock();
2610 	}
2611 
2612 	/* If the (new) signal is now blocked, requeue it.  */
2613 	if (sigismember(&current->blocked, signr) ||
2614 	    fatal_signal_pending(current)) {
2615 		send_signal(signr, info, current, type);
2616 		signr = 0;
2617 	}
2618 
2619 	return signr;
2620 }
2621 
2622 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2623 {
2624 	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2625 	case SIL_FAULT:
2626 	case SIL_FAULT_TRAPNO:
2627 	case SIL_FAULT_MCEERR:
2628 	case SIL_FAULT_BNDERR:
2629 	case SIL_FAULT_PKUERR:
2630 	case SIL_FAULT_PERF_EVENT:
2631 		ksig->info.si_addr = arch_untagged_si_addr(
2632 			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2633 		break;
2634 	case SIL_KILL:
2635 	case SIL_TIMER:
2636 	case SIL_POLL:
2637 	case SIL_CHLD:
2638 	case SIL_RT:
2639 	case SIL_SYS:
2640 		break;
2641 	}
2642 }
2643 
2644 bool get_signal(struct ksignal *ksig)
2645 {
2646 	struct sighand_struct *sighand = current->sighand;
2647 	struct signal_struct *signal = current->signal;
2648 	int signr;
2649 
2650 	clear_notify_signal();
2651 	if (unlikely(task_work_pending(current)))
2652 		task_work_run();
2653 
2654 	if (!task_sigpending(current))
2655 		return false;
2656 
2657 	if (unlikely(uprobe_deny_signal()))
2658 		return false;
2659 
2660 	/*
2661 	 * Do this once, we can't return to user-mode if freezing() == T.
2662 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2663 	 * thus do not need another check after return.
2664 	 */
2665 	try_to_freeze();
2666 
2667 relock:
2668 	spin_lock_irq(&sighand->siglock);
2669 
2670 	/*
2671 	 * Every stopped thread goes here after wakeup. Check to see if
2672 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2673 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2674 	 */
2675 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2676 		int why;
2677 
2678 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2679 			why = CLD_CONTINUED;
2680 		else
2681 			why = CLD_STOPPED;
2682 
2683 		signal->flags &= ~SIGNAL_CLD_MASK;
2684 
2685 		spin_unlock_irq(&sighand->siglock);
2686 
2687 		/*
2688 		 * Notify the parent that we're continuing.  This event is
2689 		 * always per-process and doesn't make whole lot of sense
2690 		 * for ptracers, who shouldn't consume the state via
2691 		 * wait(2) either, but, for backward compatibility, notify
2692 		 * the ptracer of the group leader too unless it's gonna be
2693 		 * a duplicate.
2694 		 */
2695 		read_lock(&tasklist_lock);
2696 		do_notify_parent_cldstop(current, false, why);
2697 
2698 		if (ptrace_reparented(current->group_leader))
2699 			do_notify_parent_cldstop(current->group_leader,
2700 						true, why);
2701 		read_unlock(&tasklist_lock);
2702 
2703 		goto relock;
2704 	}
2705 
2706 	for (;;) {
2707 		struct k_sigaction *ka;
2708 		enum pid_type type;
2709 
2710 		/* Has this task already been marked for death? */
2711 		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
2712 		     signal->group_exec_task) {
2713 			ksig->info.si_signo = signr = SIGKILL;
2714 			sigdelset(&current->pending.signal, SIGKILL);
2715 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2716 				&sighand->action[SIGKILL - 1]);
2717 			recalc_sigpending();
2718 			goto fatal;
2719 		}
2720 
2721 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2722 		    do_signal_stop(0))
2723 			goto relock;
2724 
2725 		if (unlikely(current->jobctl &
2726 			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2727 			if (current->jobctl & JOBCTL_TRAP_MASK) {
2728 				do_jobctl_trap();
2729 				spin_unlock_irq(&sighand->siglock);
2730 			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2731 				do_freezer_trap();
2732 
2733 			goto relock;
2734 		}
2735 
2736 		/*
2737 		 * If the task is leaving the frozen state, let's update
2738 		 * cgroup counters and reset the frozen bit.
2739 		 */
2740 		if (unlikely(cgroup_task_frozen(current))) {
2741 			spin_unlock_irq(&sighand->siglock);
2742 			cgroup_leave_frozen(false);
2743 			goto relock;
2744 		}
2745 
2746 		/*
2747 		 * Signals generated by the execution of an instruction
2748 		 * need to be delivered before any other pending signals
2749 		 * so that the instruction pointer in the signal stack
2750 		 * frame points to the faulting instruction.
2751 		 */
2752 		type = PIDTYPE_PID;
2753 		signr = dequeue_synchronous_signal(&ksig->info);
2754 		if (!signr)
2755 			signr = dequeue_signal(current, &current->blocked,
2756 					       &ksig->info, &type);
2757 
2758 		if (!signr)
2759 			break; /* will return 0 */
2760 
2761 		if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2762 		    !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2763 			signr = ptrace_signal(signr, &ksig->info, type);
2764 			if (!signr)
2765 				continue;
2766 		}
2767 
2768 		ka = &sighand->action[signr-1];
2769 
2770 		/* Trace actually delivered signals. */
2771 		trace_signal_deliver(signr, &ksig->info, ka);
2772 
2773 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2774 			continue;
2775 		if (ka->sa.sa_handler != SIG_DFL) {
2776 			/* Run the handler.  */
2777 			ksig->ka = *ka;
2778 
2779 			if (ka->sa.sa_flags & SA_ONESHOT)
2780 				ka->sa.sa_handler = SIG_DFL;
2781 
2782 			break; /* will return non-zero "signr" value */
2783 		}
2784 
2785 		/*
2786 		 * Now we are doing the default action for this signal.
2787 		 */
2788 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2789 			continue;
2790 
2791 		/*
2792 		 * Global init gets no signals it doesn't want.
2793 		 * Container-init gets no signals it doesn't want from same
2794 		 * container.
2795 		 *
2796 		 * Note that if global/container-init sees a sig_kernel_only()
2797 		 * signal here, the signal must have been generated internally
2798 		 * or must have come from an ancestor namespace. In either
2799 		 * case, the signal cannot be dropped.
2800 		 */
2801 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2802 				!sig_kernel_only(signr))
2803 			continue;
2804 
2805 		if (sig_kernel_stop(signr)) {
2806 			/*
2807 			 * The default action is to stop all threads in
2808 			 * the thread group.  The job control signals
2809 			 * do nothing in an orphaned pgrp, but SIGSTOP
2810 			 * always works.  Note that siglock needs to be
2811 			 * dropped during the call to is_orphaned_pgrp()
2812 			 * because of lock ordering with tasklist_lock.
2813 			 * This allows an intervening SIGCONT to be posted.
2814 			 * We need to check for that and bail out if necessary.
2815 			 */
2816 			if (signr != SIGSTOP) {
2817 				spin_unlock_irq(&sighand->siglock);
2818 
2819 				/* signals can be posted during this window */
2820 
2821 				if (is_current_pgrp_orphaned())
2822 					goto relock;
2823 
2824 				spin_lock_irq(&sighand->siglock);
2825 			}
2826 
2827 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2828 				/* It released the siglock.  */
2829 				goto relock;
2830 			}
2831 
2832 			/*
2833 			 * We didn't actually stop, due to a race
2834 			 * with SIGCONT or something like that.
2835 			 */
2836 			continue;
2837 		}
2838 
2839 	fatal:
2840 		spin_unlock_irq(&sighand->siglock);
2841 		if (unlikely(cgroup_task_frozen(current)))
2842 			cgroup_leave_frozen(true);
2843 
2844 		/*
2845 		 * Anything else is fatal, maybe with a core dump.
2846 		 */
2847 		current->flags |= PF_SIGNALED;
2848 
2849 		if (sig_kernel_coredump(signr)) {
2850 			if (print_fatal_signals)
2851 				print_fatal_signal(ksig->info.si_signo);
2852 			proc_coredump_connector(current);
2853 			/*
2854 			 * If it was able to dump core, this kills all
2855 			 * other threads in the group and synchronizes with
2856 			 * their demise.  If we lost the race with another
2857 			 * thread getting here, it set group_exit_code
2858 			 * first and our do_group_exit call below will use
2859 			 * that value and ignore the one we pass it.
2860 			 */
2861 			do_coredump(&ksig->info);
2862 		}
2863 
2864 		/*
2865 		 * PF_IO_WORKER threads will catch and exit on fatal signals
2866 		 * themselves. They have cleanup that must be performed, so
2867 		 * we cannot call do_exit() on their behalf.
2868 		 */
2869 		if (current->flags & PF_IO_WORKER)
2870 			goto out;
2871 
2872 		/*
2873 		 * Death signals, no core dump.
2874 		 */
2875 		do_group_exit(ksig->info.si_signo);
2876 		/* NOTREACHED */
2877 	}
2878 	spin_unlock_irq(&sighand->siglock);
2879 out:
2880 	ksig->sig = signr;
2881 
2882 	if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2883 		hide_si_addr_tag_bits(ksig);
2884 
2885 	return ksig->sig > 0;
2886 }
2887 
2888 /**
2889  * signal_delivered - called after signal delivery to update blocked signals
2890  * @ksig:		kernel signal struct
2891  * @stepping:		nonzero if debugger single-step or block-step in use
2892  *
2893  * This function should be called when a signal has successfully been
2894  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2895  * is always blocked), and the signal itself is blocked unless %SA_NODEFER
2896  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2897  */
2898 static void signal_delivered(struct ksignal *ksig, int stepping)
2899 {
2900 	sigset_t blocked;
2901 
2902 	/* A signal was successfully delivered, and the
2903 	   saved sigmask was stored on the signal frame,
2904 	   and will be restored by sigreturn.  So we can
2905 	   simply clear the restore sigmask flag.  */
2906 	clear_restore_sigmask();
2907 
2908 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2909 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2910 		sigaddset(&blocked, ksig->sig);
2911 	set_current_blocked(&blocked);
2912 	if (current->sas_ss_flags & SS_AUTODISARM)
2913 		sas_ss_reset(current);
2914 	if (stepping)
2915 		ptrace_notify(SIGTRAP, 0);
2916 }
2917 
2918 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2919 {
2920 	if (failed)
2921 		force_sigsegv(ksig->sig);
2922 	else
2923 		signal_delivered(ksig, stepping);
2924 }
2925 
2926 /*
2927  * It could be that complete_signal() picked us to notify about the
2928  * group-wide signal. Other threads should be notified now to take
2929  * the shared signals in @which since we will not.
2930  */
2931 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2932 {
2933 	sigset_t retarget;
2934 	struct task_struct *t;
2935 
2936 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2937 	if (sigisemptyset(&retarget))
2938 		return;
2939 
2940 	t = tsk;
2941 	while_each_thread(tsk, t) {
2942 		if (t->flags & PF_EXITING)
2943 			continue;
2944 
2945 		if (!has_pending_signals(&retarget, &t->blocked))
2946 			continue;
2947 		/* Remove the signals this thread can handle. */
2948 		sigandsets(&retarget, &retarget, &t->blocked);
2949 
2950 		if (!task_sigpending(t))
2951 			signal_wake_up(t, 0);
2952 
2953 		if (sigisemptyset(&retarget))
2954 			break;
2955 	}
2956 }
2957 
2958 void exit_signals(struct task_struct *tsk)
2959 {
2960 	int group_stop = 0;
2961 	sigset_t unblocked;
2962 
2963 	/*
2964 	 * @tsk is about to have PF_EXITING set - lock out users which
2965 	 * expect stable threadgroup.
2966 	 */
2967 	cgroup_threadgroup_change_begin(tsk);
2968 
2969 	if (thread_group_empty(tsk) || (tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
2970 		tsk->flags |= PF_EXITING;
2971 		cgroup_threadgroup_change_end(tsk);
2972 		return;
2973 	}
2974 
2975 	spin_lock_irq(&tsk->sighand->siglock);
2976 	/*
2977 	 * From now this task is not visible for group-wide signals,
2978 	 * see wants_signal(), do_signal_stop().
2979 	 */
2980 	tsk->flags |= PF_EXITING;
2981 
2982 	cgroup_threadgroup_change_end(tsk);
2983 
2984 	if (!task_sigpending(tsk))
2985 		goto out;
2986 
2987 	unblocked = tsk->blocked;
2988 	signotset(&unblocked);
2989 	retarget_shared_pending(tsk, &unblocked);
2990 
2991 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2992 	    task_participate_group_stop(tsk))
2993 		group_stop = CLD_STOPPED;
2994 out:
2995 	spin_unlock_irq(&tsk->sighand->siglock);
2996 
2997 	/*
2998 	 * If group stop has completed, deliver the notification.  This
2999 	 * should always go to the real parent of the group leader.
3000 	 */
3001 	if (unlikely(group_stop)) {
3002 		read_lock(&tasklist_lock);
3003 		do_notify_parent_cldstop(tsk, false, group_stop);
3004 		read_unlock(&tasklist_lock);
3005 	}
3006 }
3007 
3008 /*
3009  * System call entry points.
3010  */
3011 
3012 /**
3013  *  sys_restart_syscall - restart a system call
3014  */
3015 SYSCALL_DEFINE0(restart_syscall)
3016 {
3017 	struct restart_block *restart = &current->restart_block;
3018 	return restart->fn(restart);
3019 }
3020 
3021 long do_no_restart_syscall(struct restart_block *param)
3022 {
3023 	return -EINTR;
3024 }
3025 
3026 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
3027 {
3028 	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3029 		sigset_t newblocked;
3030 		/* A set of now blocked but previously unblocked signals. */
3031 		sigandnsets(&newblocked, newset, &current->blocked);
3032 		retarget_shared_pending(tsk, &newblocked);
3033 	}
3034 	tsk->blocked = *newset;
3035 	recalc_sigpending();
3036 }
3037 
3038 /**
3039  * set_current_blocked - change current->blocked mask
3040  * @newset: new mask
3041  *
3042  * It is wrong to change ->blocked directly, this helper should be used
3043  * to ensure the process can't miss a shared signal we are going to block.
3044  */
3045 void set_current_blocked(sigset_t *newset)
3046 {
3047 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3048 	__set_current_blocked(newset);
3049 }
3050 
3051 void __set_current_blocked(const sigset_t *newset)
3052 {
3053 	struct task_struct *tsk = current;
3054 
3055 	/*
3056 	 * In case the signal mask hasn't changed, there is nothing we need
3057 	 * to do. The current->blocked shouldn't be modified by other task.
3058 	 */
3059 	if (sigequalsets(&tsk->blocked, newset))
3060 		return;
3061 
3062 	spin_lock_irq(&tsk->sighand->siglock);
3063 	__set_task_blocked(tsk, newset);
3064 	spin_unlock_irq(&tsk->sighand->siglock);
3065 }
3066 
3067 /*
3068  * This is also useful for kernel threads that want to temporarily
3069  * (or permanently) block certain signals.
3070  *
3071  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3072  * interface happily blocks "unblockable" signals like SIGKILL
3073  * and friends.
3074  */
3075 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3076 {
3077 	struct task_struct *tsk = current;
3078 	sigset_t newset;
3079 
3080 	/* Lockless, only current can change ->blocked, never from irq */
3081 	if (oldset)
3082 		*oldset = tsk->blocked;
3083 
3084 	switch (how) {
3085 	case SIG_BLOCK:
3086 		sigorsets(&newset, &tsk->blocked, set);
3087 		break;
3088 	case SIG_UNBLOCK:
3089 		sigandnsets(&newset, &tsk->blocked, set);
3090 		break;
3091 	case SIG_SETMASK:
3092 		newset = *set;
3093 		break;
3094 	default:
3095 		return -EINVAL;
3096 	}
3097 
3098 	__set_current_blocked(&newset);
3099 	return 0;
3100 }
3101 EXPORT_SYMBOL(sigprocmask);
3102 
3103 /*
3104  * The api helps set app-provided sigmasks.
3105  *
3106  * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3107  * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3108  *
3109  * Note that it does set_restore_sigmask() in advance, so it must be always
3110  * paired with restore_saved_sigmask_unless() before return from syscall.
3111  */
3112 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3113 {
3114 	sigset_t kmask;
3115 
3116 	if (!umask)
3117 		return 0;
3118 	if (sigsetsize != sizeof(sigset_t))
3119 		return -EINVAL;
3120 	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3121 		return -EFAULT;
3122 
3123 	set_restore_sigmask();
3124 	current->saved_sigmask = current->blocked;
3125 	set_current_blocked(&kmask);
3126 
3127 	return 0;
3128 }
3129 
3130 #ifdef CONFIG_COMPAT
3131 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3132 			    size_t sigsetsize)
3133 {
3134 	sigset_t kmask;
3135 
3136 	if (!umask)
3137 		return 0;
3138 	if (sigsetsize != sizeof(compat_sigset_t))
3139 		return -EINVAL;
3140 	if (get_compat_sigset(&kmask, umask))
3141 		return -EFAULT;
3142 
3143 	set_restore_sigmask();
3144 	current->saved_sigmask = current->blocked;
3145 	set_current_blocked(&kmask);
3146 
3147 	return 0;
3148 }
3149 #endif
3150 
3151 /**
3152  *  sys_rt_sigprocmask - change the list of currently blocked signals
3153  *  @how: whether to add, remove, or set signals
3154  *  @nset: stores pending signals
3155  *  @oset: previous value of signal mask if non-null
3156  *  @sigsetsize: size of sigset_t type
3157  */
3158 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3159 		sigset_t __user *, oset, size_t, sigsetsize)
3160 {
3161 	sigset_t old_set, new_set;
3162 	int error;
3163 
3164 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3165 	if (sigsetsize != sizeof(sigset_t))
3166 		return -EINVAL;
3167 
3168 	old_set = current->blocked;
3169 
3170 	if (nset) {
3171 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3172 			return -EFAULT;
3173 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3174 
3175 		error = sigprocmask(how, &new_set, NULL);
3176 		if (error)
3177 			return error;
3178 	}
3179 
3180 	if (oset) {
3181 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3182 			return -EFAULT;
3183 	}
3184 
3185 	return 0;
3186 }
3187 
3188 #ifdef CONFIG_COMPAT
3189 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3190 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3191 {
3192 	sigset_t old_set = current->blocked;
3193 
3194 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3195 	if (sigsetsize != sizeof(sigset_t))
3196 		return -EINVAL;
3197 
3198 	if (nset) {
3199 		sigset_t new_set;
3200 		int error;
3201 		if (get_compat_sigset(&new_set, nset))
3202 			return -EFAULT;
3203 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3204 
3205 		error = sigprocmask(how, &new_set, NULL);
3206 		if (error)
3207 			return error;
3208 	}
3209 	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3210 }
3211 #endif
3212 
3213 static void do_sigpending(sigset_t *set)
3214 {
3215 	spin_lock_irq(&current->sighand->siglock);
3216 	sigorsets(set, &current->pending.signal,
3217 		  &current->signal->shared_pending.signal);
3218 	spin_unlock_irq(&current->sighand->siglock);
3219 
3220 	/* Outside the lock because only this thread touches it.  */
3221 	sigandsets(set, &current->blocked, set);
3222 }
3223 
3224 /**
3225  *  sys_rt_sigpending - examine a pending signal that has been raised
3226  *			while blocked
3227  *  @uset: stores pending signals
3228  *  @sigsetsize: size of sigset_t type or larger
3229  */
3230 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3231 {
3232 	sigset_t set;
3233 
3234 	if (sigsetsize > sizeof(*uset))
3235 		return -EINVAL;
3236 
3237 	do_sigpending(&set);
3238 
3239 	if (copy_to_user(uset, &set, sigsetsize))
3240 		return -EFAULT;
3241 
3242 	return 0;
3243 }
3244 
3245 #ifdef CONFIG_COMPAT
3246 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3247 		compat_size_t, sigsetsize)
3248 {
3249 	sigset_t set;
3250 
3251 	if (sigsetsize > sizeof(*uset))
3252 		return -EINVAL;
3253 
3254 	do_sigpending(&set);
3255 
3256 	return put_compat_sigset(uset, &set, sigsetsize);
3257 }
3258 #endif
3259 
3260 static const struct {
3261 	unsigned char limit, layout;
3262 } sig_sicodes[] = {
3263 	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3264 	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3265 	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3266 	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3267 	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3268 #if defined(SIGEMT)
3269 	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3270 #endif
3271 	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3272 	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3273 	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3274 };
3275 
3276 static bool known_siginfo_layout(unsigned sig, int si_code)
3277 {
3278 	if (si_code == SI_KERNEL)
3279 		return true;
3280 	else if ((si_code > SI_USER)) {
3281 		if (sig_specific_sicodes(sig)) {
3282 			if (si_code <= sig_sicodes[sig].limit)
3283 				return true;
3284 		}
3285 		else if (si_code <= NSIGPOLL)
3286 			return true;
3287 	}
3288 	else if (si_code >= SI_DETHREAD)
3289 		return true;
3290 	else if (si_code == SI_ASYNCNL)
3291 		return true;
3292 	return false;
3293 }
3294 
3295 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3296 {
3297 	enum siginfo_layout layout = SIL_KILL;
3298 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3299 		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3300 		    (si_code <= sig_sicodes[sig].limit)) {
3301 			layout = sig_sicodes[sig].layout;
3302 			/* Handle the exceptions */
3303 			if ((sig == SIGBUS) &&
3304 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3305 				layout = SIL_FAULT_MCEERR;
3306 			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3307 				layout = SIL_FAULT_BNDERR;
3308 #ifdef SEGV_PKUERR
3309 			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3310 				layout = SIL_FAULT_PKUERR;
3311 #endif
3312 			else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3313 				layout = SIL_FAULT_PERF_EVENT;
3314 			else if (IS_ENABLED(CONFIG_SPARC) &&
3315 				 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3316 				layout = SIL_FAULT_TRAPNO;
3317 			else if (IS_ENABLED(CONFIG_ALPHA) &&
3318 				 ((sig == SIGFPE) ||
3319 				  ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3320 				layout = SIL_FAULT_TRAPNO;
3321 		}
3322 		else if (si_code <= NSIGPOLL)
3323 			layout = SIL_POLL;
3324 	} else {
3325 		if (si_code == SI_TIMER)
3326 			layout = SIL_TIMER;
3327 		else if (si_code == SI_SIGIO)
3328 			layout = SIL_POLL;
3329 		else if (si_code < 0)
3330 			layout = SIL_RT;
3331 	}
3332 	return layout;
3333 }
3334 
3335 static inline char __user *si_expansion(const siginfo_t __user *info)
3336 {
3337 	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3338 }
3339 
3340 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3341 {
3342 	char __user *expansion = si_expansion(to);
3343 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3344 		return -EFAULT;
3345 	if (clear_user(expansion, SI_EXPANSION_SIZE))
3346 		return -EFAULT;
3347 	return 0;
3348 }
3349 
3350 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3351 				       const siginfo_t __user *from)
3352 {
3353 	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3354 		char __user *expansion = si_expansion(from);
3355 		char buf[SI_EXPANSION_SIZE];
3356 		int i;
3357 		/*
3358 		 * An unknown si_code might need more than
3359 		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3360 		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3361 		 * will return this data to userspace exactly.
3362 		 */
3363 		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3364 			return -EFAULT;
3365 		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3366 			if (buf[i] != 0)
3367 				return -E2BIG;
3368 		}
3369 	}
3370 	return 0;
3371 }
3372 
3373 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3374 				    const siginfo_t __user *from)
3375 {
3376 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3377 		return -EFAULT;
3378 	to->si_signo = signo;
3379 	return post_copy_siginfo_from_user(to, from);
3380 }
3381 
3382 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3383 {
3384 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3385 		return -EFAULT;
3386 	return post_copy_siginfo_from_user(to, from);
3387 }
3388 
3389 #ifdef CONFIG_COMPAT
3390 /**
3391  * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3392  * @to: compat siginfo destination
3393  * @from: kernel siginfo source
3394  *
3395  * Note: This function does not work properly for the SIGCHLD on x32, but
3396  * fortunately it doesn't have to.  The only valid callers for this function are
3397  * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3398  * The latter does not care because SIGCHLD will never cause a coredump.
3399  */
3400 void copy_siginfo_to_external32(struct compat_siginfo *to,
3401 		const struct kernel_siginfo *from)
3402 {
3403 	memset(to, 0, sizeof(*to));
3404 
3405 	to->si_signo = from->si_signo;
3406 	to->si_errno = from->si_errno;
3407 	to->si_code  = from->si_code;
3408 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3409 	case SIL_KILL:
3410 		to->si_pid = from->si_pid;
3411 		to->si_uid = from->si_uid;
3412 		break;
3413 	case SIL_TIMER:
3414 		to->si_tid     = from->si_tid;
3415 		to->si_overrun = from->si_overrun;
3416 		to->si_int     = from->si_int;
3417 		break;
3418 	case SIL_POLL:
3419 		to->si_band = from->si_band;
3420 		to->si_fd   = from->si_fd;
3421 		break;
3422 	case SIL_FAULT:
3423 		to->si_addr = ptr_to_compat(from->si_addr);
3424 		break;
3425 	case SIL_FAULT_TRAPNO:
3426 		to->si_addr = ptr_to_compat(from->si_addr);
3427 		to->si_trapno = from->si_trapno;
3428 		break;
3429 	case SIL_FAULT_MCEERR:
3430 		to->si_addr = ptr_to_compat(from->si_addr);
3431 		to->si_addr_lsb = from->si_addr_lsb;
3432 		break;
3433 	case SIL_FAULT_BNDERR:
3434 		to->si_addr = ptr_to_compat(from->si_addr);
3435 		to->si_lower = ptr_to_compat(from->si_lower);
3436 		to->si_upper = ptr_to_compat(from->si_upper);
3437 		break;
3438 	case SIL_FAULT_PKUERR:
3439 		to->si_addr = ptr_to_compat(from->si_addr);
3440 		to->si_pkey = from->si_pkey;
3441 		break;
3442 	case SIL_FAULT_PERF_EVENT:
3443 		to->si_addr = ptr_to_compat(from->si_addr);
3444 		to->si_perf_data = from->si_perf_data;
3445 		to->si_perf_type = from->si_perf_type;
3446 		to->si_perf_flags = from->si_perf_flags;
3447 		break;
3448 	case SIL_CHLD:
3449 		to->si_pid = from->si_pid;
3450 		to->si_uid = from->si_uid;
3451 		to->si_status = from->si_status;
3452 		to->si_utime = from->si_utime;
3453 		to->si_stime = from->si_stime;
3454 		break;
3455 	case SIL_RT:
3456 		to->si_pid = from->si_pid;
3457 		to->si_uid = from->si_uid;
3458 		to->si_int = from->si_int;
3459 		break;
3460 	case SIL_SYS:
3461 		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3462 		to->si_syscall   = from->si_syscall;
3463 		to->si_arch      = from->si_arch;
3464 		break;
3465 	}
3466 }
3467 
3468 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3469 			   const struct kernel_siginfo *from)
3470 {
3471 	struct compat_siginfo new;
3472 
3473 	copy_siginfo_to_external32(&new, from);
3474 	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3475 		return -EFAULT;
3476 	return 0;
3477 }
3478 
3479 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3480 					 const struct compat_siginfo *from)
3481 {
3482 	clear_siginfo(to);
3483 	to->si_signo = from->si_signo;
3484 	to->si_errno = from->si_errno;
3485 	to->si_code  = from->si_code;
3486 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3487 	case SIL_KILL:
3488 		to->si_pid = from->si_pid;
3489 		to->si_uid = from->si_uid;
3490 		break;
3491 	case SIL_TIMER:
3492 		to->si_tid     = from->si_tid;
3493 		to->si_overrun = from->si_overrun;
3494 		to->si_int     = from->si_int;
3495 		break;
3496 	case SIL_POLL:
3497 		to->si_band = from->si_band;
3498 		to->si_fd   = from->si_fd;
3499 		break;
3500 	case SIL_FAULT:
3501 		to->si_addr = compat_ptr(from->si_addr);
3502 		break;
3503 	case SIL_FAULT_TRAPNO:
3504 		to->si_addr = compat_ptr(from->si_addr);
3505 		to->si_trapno = from->si_trapno;
3506 		break;
3507 	case SIL_FAULT_MCEERR:
3508 		to->si_addr = compat_ptr(from->si_addr);
3509 		to->si_addr_lsb = from->si_addr_lsb;
3510 		break;
3511 	case SIL_FAULT_BNDERR:
3512 		to->si_addr = compat_ptr(from->si_addr);
3513 		to->si_lower = compat_ptr(from->si_lower);
3514 		to->si_upper = compat_ptr(from->si_upper);
3515 		break;
3516 	case SIL_FAULT_PKUERR:
3517 		to->si_addr = compat_ptr(from->si_addr);
3518 		to->si_pkey = from->si_pkey;
3519 		break;
3520 	case SIL_FAULT_PERF_EVENT:
3521 		to->si_addr = compat_ptr(from->si_addr);
3522 		to->si_perf_data = from->si_perf_data;
3523 		to->si_perf_type = from->si_perf_type;
3524 		to->si_perf_flags = from->si_perf_flags;
3525 		break;
3526 	case SIL_CHLD:
3527 		to->si_pid    = from->si_pid;
3528 		to->si_uid    = from->si_uid;
3529 		to->si_status = from->si_status;
3530 #ifdef CONFIG_X86_X32_ABI
3531 		if (in_x32_syscall()) {
3532 			to->si_utime = from->_sifields._sigchld_x32._utime;
3533 			to->si_stime = from->_sifields._sigchld_x32._stime;
3534 		} else
3535 #endif
3536 		{
3537 			to->si_utime = from->si_utime;
3538 			to->si_stime = from->si_stime;
3539 		}
3540 		break;
3541 	case SIL_RT:
3542 		to->si_pid = from->si_pid;
3543 		to->si_uid = from->si_uid;
3544 		to->si_int = from->si_int;
3545 		break;
3546 	case SIL_SYS:
3547 		to->si_call_addr = compat_ptr(from->si_call_addr);
3548 		to->si_syscall   = from->si_syscall;
3549 		to->si_arch      = from->si_arch;
3550 		break;
3551 	}
3552 	return 0;
3553 }
3554 
3555 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3556 				      const struct compat_siginfo __user *ufrom)
3557 {
3558 	struct compat_siginfo from;
3559 
3560 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3561 		return -EFAULT;
3562 
3563 	from.si_signo = signo;
3564 	return post_copy_siginfo_from_user32(to, &from);
3565 }
3566 
3567 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3568 			     const struct compat_siginfo __user *ufrom)
3569 {
3570 	struct compat_siginfo from;
3571 
3572 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3573 		return -EFAULT;
3574 
3575 	return post_copy_siginfo_from_user32(to, &from);
3576 }
3577 #endif /* CONFIG_COMPAT */
3578 
3579 /**
3580  *  do_sigtimedwait - wait for queued signals specified in @which
3581  *  @which: queued signals to wait for
3582  *  @info: if non-null, the signal's siginfo is returned here
3583  *  @ts: upper bound on process time suspension
3584  */
3585 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3586 		    const struct timespec64 *ts)
3587 {
3588 	ktime_t *to = NULL, timeout = KTIME_MAX;
3589 	struct task_struct *tsk = current;
3590 	sigset_t mask = *which;
3591 	enum pid_type type;
3592 	int sig, ret = 0;
3593 
3594 	if (ts) {
3595 		if (!timespec64_valid(ts))
3596 			return -EINVAL;
3597 		timeout = timespec64_to_ktime(*ts);
3598 		to = &timeout;
3599 	}
3600 
3601 	/*
3602 	 * Invert the set of allowed signals to get those we want to block.
3603 	 */
3604 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3605 	signotset(&mask);
3606 
3607 	spin_lock_irq(&tsk->sighand->siglock);
3608 	sig = dequeue_signal(tsk, &mask, info, &type);
3609 	if (!sig && timeout) {
3610 		/*
3611 		 * None ready, temporarily unblock those we're interested
3612 		 * while we are sleeping in so that we'll be awakened when
3613 		 * they arrive. Unblocking is always fine, we can avoid
3614 		 * set_current_blocked().
3615 		 */
3616 		tsk->real_blocked = tsk->blocked;
3617 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3618 		recalc_sigpending();
3619 		spin_unlock_irq(&tsk->sighand->siglock);
3620 
3621 		__set_current_state(TASK_INTERRUPTIBLE);
3622 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3623 							 HRTIMER_MODE_REL);
3624 		spin_lock_irq(&tsk->sighand->siglock);
3625 		__set_task_blocked(tsk, &tsk->real_blocked);
3626 		sigemptyset(&tsk->real_blocked);
3627 		sig = dequeue_signal(tsk, &mask, info, &type);
3628 	}
3629 	spin_unlock_irq(&tsk->sighand->siglock);
3630 
3631 	if (sig)
3632 		return sig;
3633 	return ret ? -EINTR : -EAGAIN;
3634 }
3635 
3636 /**
3637  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3638  *			in @uthese
3639  *  @uthese: queued signals to wait for
3640  *  @uinfo: if non-null, the signal's siginfo is returned here
3641  *  @uts: upper bound on process time suspension
3642  *  @sigsetsize: size of sigset_t type
3643  */
3644 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3645 		siginfo_t __user *, uinfo,
3646 		const struct __kernel_timespec __user *, uts,
3647 		size_t, sigsetsize)
3648 {
3649 	sigset_t these;
3650 	struct timespec64 ts;
3651 	kernel_siginfo_t info;
3652 	int ret;
3653 
3654 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3655 	if (sigsetsize != sizeof(sigset_t))
3656 		return -EINVAL;
3657 
3658 	if (copy_from_user(&these, uthese, sizeof(these)))
3659 		return -EFAULT;
3660 
3661 	if (uts) {
3662 		if (get_timespec64(&ts, uts))
3663 			return -EFAULT;
3664 	}
3665 
3666 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3667 
3668 	if (ret > 0 && uinfo) {
3669 		if (copy_siginfo_to_user(uinfo, &info))
3670 			ret = -EFAULT;
3671 	}
3672 
3673 	return ret;
3674 }
3675 
3676 #ifdef CONFIG_COMPAT_32BIT_TIME
3677 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3678 		siginfo_t __user *, uinfo,
3679 		const struct old_timespec32 __user *, uts,
3680 		size_t, sigsetsize)
3681 {
3682 	sigset_t these;
3683 	struct timespec64 ts;
3684 	kernel_siginfo_t info;
3685 	int ret;
3686 
3687 	if (sigsetsize != sizeof(sigset_t))
3688 		return -EINVAL;
3689 
3690 	if (copy_from_user(&these, uthese, sizeof(these)))
3691 		return -EFAULT;
3692 
3693 	if (uts) {
3694 		if (get_old_timespec32(&ts, uts))
3695 			return -EFAULT;
3696 	}
3697 
3698 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3699 
3700 	if (ret > 0 && uinfo) {
3701 		if (copy_siginfo_to_user(uinfo, &info))
3702 			ret = -EFAULT;
3703 	}
3704 
3705 	return ret;
3706 }
3707 #endif
3708 
3709 #ifdef CONFIG_COMPAT
3710 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3711 		struct compat_siginfo __user *, uinfo,
3712 		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3713 {
3714 	sigset_t s;
3715 	struct timespec64 t;
3716 	kernel_siginfo_t info;
3717 	long ret;
3718 
3719 	if (sigsetsize != sizeof(sigset_t))
3720 		return -EINVAL;
3721 
3722 	if (get_compat_sigset(&s, uthese))
3723 		return -EFAULT;
3724 
3725 	if (uts) {
3726 		if (get_timespec64(&t, uts))
3727 			return -EFAULT;
3728 	}
3729 
3730 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3731 
3732 	if (ret > 0 && uinfo) {
3733 		if (copy_siginfo_to_user32(uinfo, &info))
3734 			ret = -EFAULT;
3735 	}
3736 
3737 	return ret;
3738 }
3739 
3740 #ifdef CONFIG_COMPAT_32BIT_TIME
3741 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3742 		struct compat_siginfo __user *, uinfo,
3743 		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3744 {
3745 	sigset_t s;
3746 	struct timespec64 t;
3747 	kernel_siginfo_t info;
3748 	long ret;
3749 
3750 	if (sigsetsize != sizeof(sigset_t))
3751 		return -EINVAL;
3752 
3753 	if (get_compat_sigset(&s, uthese))
3754 		return -EFAULT;
3755 
3756 	if (uts) {
3757 		if (get_old_timespec32(&t, uts))
3758 			return -EFAULT;
3759 	}
3760 
3761 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3762 
3763 	if (ret > 0 && uinfo) {
3764 		if (copy_siginfo_to_user32(uinfo, &info))
3765 			ret = -EFAULT;
3766 	}
3767 
3768 	return ret;
3769 }
3770 #endif
3771 #endif
3772 
3773 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3774 {
3775 	clear_siginfo(info);
3776 	info->si_signo = sig;
3777 	info->si_errno = 0;
3778 	info->si_code = SI_USER;
3779 	info->si_pid = task_tgid_vnr(current);
3780 	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3781 }
3782 
3783 /**
3784  *  sys_kill - send a signal to a process
3785  *  @pid: the PID of the process
3786  *  @sig: signal to be sent
3787  */
3788 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3789 {
3790 	struct kernel_siginfo info;
3791 
3792 	prepare_kill_siginfo(sig, &info);
3793 
3794 	return kill_something_info(sig, &info, pid);
3795 }
3796 
3797 /*
3798  * Verify that the signaler and signalee either are in the same pid namespace
3799  * or that the signaler's pid namespace is an ancestor of the signalee's pid
3800  * namespace.
3801  */
3802 static bool access_pidfd_pidns(struct pid *pid)
3803 {
3804 	struct pid_namespace *active = task_active_pid_ns(current);
3805 	struct pid_namespace *p = ns_of_pid(pid);
3806 
3807 	for (;;) {
3808 		if (!p)
3809 			return false;
3810 		if (p == active)
3811 			break;
3812 		p = p->parent;
3813 	}
3814 
3815 	return true;
3816 }
3817 
3818 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3819 		siginfo_t __user *info)
3820 {
3821 #ifdef CONFIG_COMPAT
3822 	/*
3823 	 * Avoid hooking up compat syscalls and instead handle necessary
3824 	 * conversions here. Note, this is a stop-gap measure and should not be
3825 	 * considered a generic solution.
3826 	 */
3827 	if (in_compat_syscall())
3828 		return copy_siginfo_from_user32(
3829 			kinfo, (struct compat_siginfo __user *)info);
3830 #endif
3831 	return copy_siginfo_from_user(kinfo, info);
3832 }
3833 
3834 static struct pid *pidfd_to_pid(const struct file *file)
3835 {
3836 	struct pid *pid;
3837 
3838 	pid = pidfd_pid(file);
3839 	if (!IS_ERR(pid))
3840 		return pid;
3841 
3842 	return tgid_pidfd_to_pid(file);
3843 }
3844 
3845 /**
3846  * sys_pidfd_send_signal - Signal a process through a pidfd
3847  * @pidfd:  file descriptor of the process
3848  * @sig:    signal to send
3849  * @info:   signal info
3850  * @flags:  future flags
3851  *
3852  * The syscall currently only signals via PIDTYPE_PID which covers
3853  * kill(<positive-pid>, <signal>. It does not signal threads or process
3854  * groups.
3855  * In order to extend the syscall to threads and process groups the @flags
3856  * argument should be used. In essence, the @flags argument will determine
3857  * what is signaled and not the file descriptor itself. Put in other words,
3858  * grouping is a property of the flags argument not a property of the file
3859  * descriptor.
3860  *
3861  * Return: 0 on success, negative errno on failure
3862  */
3863 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3864 		siginfo_t __user *, info, unsigned int, flags)
3865 {
3866 	int ret;
3867 	struct fd f;
3868 	struct pid *pid;
3869 	kernel_siginfo_t kinfo;
3870 
3871 	/* Enforce flags be set to 0 until we add an extension. */
3872 	if (flags)
3873 		return -EINVAL;
3874 
3875 	f = fdget(pidfd);
3876 	if (!f.file)
3877 		return -EBADF;
3878 
3879 	/* Is this a pidfd? */
3880 	pid = pidfd_to_pid(f.file);
3881 	if (IS_ERR(pid)) {
3882 		ret = PTR_ERR(pid);
3883 		goto err;
3884 	}
3885 
3886 	ret = -EINVAL;
3887 	if (!access_pidfd_pidns(pid))
3888 		goto err;
3889 
3890 	if (info) {
3891 		ret = copy_siginfo_from_user_any(&kinfo, info);
3892 		if (unlikely(ret))
3893 			goto err;
3894 
3895 		ret = -EINVAL;
3896 		if (unlikely(sig != kinfo.si_signo))
3897 			goto err;
3898 
3899 		/* Only allow sending arbitrary signals to yourself. */
3900 		ret = -EPERM;
3901 		if ((task_pid(current) != pid) &&
3902 		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3903 			goto err;
3904 	} else {
3905 		prepare_kill_siginfo(sig, &kinfo);
3906 	}
3907 
3908 	ret = kill_pid_info(sig, &kinfo, pid);
3909 
3910 err:
3911 	fdput(f);
3912 	return ret;
3913 }
3914 
3915 static int
3916 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3917 {
3918 	struct task_struct *p;
3919 	int error = -ESRCH;
3920 
3921 	rcu_read_lock();
3922 	p = find_task_by_vpid(pid);
3923 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3924 		error = check_kill_permission(sig, info, p);
3925 		/*
3926 		 * The null signal is a permissions and process existence
3927 		 * probe.  No signal is actually delivered.
3928 		 */
3929 		if (!error && sig) {
3930 			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3931 			/*
3932 			 * If lock_task_sighand() failed we pretend the task
3933 			 * dies after receiving the signal. The window is tiny,
3934 			 * and the signal is private anyway.
3935 			 */
3936 			if (unlikely(error == -ESRCH))
3937 				error = 0;
3938 		}
3939 	}
3940 	rcu_read_unlock();
3941 
3942 	return error;
3943 }
3944 
3945 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3946 {
3947 	struct kernel_siginfo info;
3948 
3949 	clear_siginfo(&info);
3950 	info.si_signo = sig;
3951 	info.si_errno = 0;
3952 	info.si_code = SI_TKILL;
3953 	info.si_pid = task_tgid_vnr(current);
3954 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3955 
3956 	return do_send_specific(tgid, pid, sig, &info);
3957 }
3958 
3959 /**
3960  *  sys_tgkill - send signal to one specific thread
3961  *  @tgid: the thread group ID of the thread
3962  *  @pid: the PID of the thread
3963  *  @sig: signal to be sent
3964  *
3965  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3966  *  exists but it's not belonging to the target process anymore. This
3967  *  method solves the problem of threads exiting and PIDs getting reused.
3968  */
3969 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3970 {
3971 	/* This is only valid for single tasks */
3972 	if (pid <= 0 || tgid <= 0)
3973 		return -EINVAL;
3974 
3975 	return do_tkill(tgid, pid, sig);
3976 }
3977 
3978 /**
3979  *  sys_tkill - send signal to one specific task
3980  *  @pid: the PID of the task
3981  *  @sig: signal to be sent
3982  *
3983  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3984  */
3985 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3986 {
3987 	/* This is only valid for single tasks */
3988 	if (pid <= 0)
3989 		return -EINVAL;
3990 
3991 	return do_tkill(0, pid, sig);
3992 }
3993 
3994 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
3995 {
3996 	/* Not even root can pretend to send signals from the kernel.
3997 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3998 	 */
3999 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4000 	    (task_pid_vnr(current) != pid))
4001 		return -EPERM;
4002 
4003 	/* POSIX.1b doesn't mention process groups.  */
4004 	return kill_proc_info(sig, info, pid);
4005 }
4006 
4007 /**
4008  *  sys_rt_sigqueueinfo - send signal information to a signal
4009  *  @pid: the PID of the thread
4010  *  @sig: signal to be sent
4011  *  @uinfo: signal info to be sent
4012  */
4013 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
4014 		siginfo_t __user *, uinfo)
4015 {
4016 	kernel_siginfo_t info;
4017 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4018 	if (unlikely(ret))
4019 		return ret;
4020 	return do_rt_sigqueueinfo(pid, sig, &info);
4021 }
4022 
4023 #ifdef CONFIG_COMPAT
4024 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
4025 			compat_pid_t, pid,
4026 			int, sig,
4027 			struct compat_siginfo __user *, uinfo)
4028 {
4029 	kernel_siginfo_t info;
4030 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4031 	if (unlikely(ret))
4032 		return ret;
4033 	return do_rt_sigqueueinfo(pid, sig, &info);
4034 }
4035 #endif
4036 
4037 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4038 {
4039 	/* This is only valid for single tasks */
4040 	if (pid <= 0 || tgid <= 0)
4041 		return -EINVAL;
4042 
4043 	/* Not even root can pretend to send signals from the kernel.
4044 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4045 	 */
4046 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4047 	    (task_pid_vnr(current) != pid))
4048 		return -EPERM;
4049 
4050 	return do_send_specific(tgid, pid, sig, info);
4051 }
4052 
4053 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4054 		siginfo_t __user *, uinfo)
4055 {
4056 	kernel_siginfo_t info;
4057 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4058 	if (unlikely(ret))
4059 		return ret;
4060 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4061 }
4062 
4063 #ifdef CONFIG_COMPAT
4064 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4065 			compat_pid_t, tgid,
4066 			compat_pid_t, pid,
4067 			int, sig,
4068 			struct compat_siginfo __user *, uinfo)
4069 {
4070 	kernel_siginfo_t info;
4071 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4072 	if (unlikely(ret))
4073 		return ret;
4074 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4075 }
4076 #endif
4077 
4078 /*
4079  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4080  */
4081 void kernel_sigaction(int sig, __sighandler_t action)
4082 {
4083 	spin_lock_irq(&current->sighand->siglock);
4084 	current->sighand->action[sig - 1].sa.sa_handler = action;
4085 	if (action == SIG_IGN) {
4086 		sigset_t mask;
4087 
4088 		sigemptyset(&mask);
4089 		sigaddset(&mask, sig);
4090 
4091 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
4092 		flush_sigqueue_mask(&mask, &current->pending);
4093 		recalc_sigpending();
4094 	}
4095 	spin_unlock_irq(&current->sighand->siglock);
4096 }
4097 EXPORT_SYMBOL(kernel_sigaction);
4098 
4099 void __weak sigaction_compat_abi(struct k_sigaction *act,
4100 		struct k_sigaction *oact)
4101 {
4102 }
4103 
4104 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4105 {
4106 	struct task_struct *p = current, *t;
4107 	struct k_sigaction *k;
4108 	sigset_t mask;
4109 
4110 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4111 		return -EINVAL;
4112 
4113 	k = &p->sighand->action[sig-1];
4114 
4115 	spin_lock_irq(&p->sighand->siglock);
4116 	if (k->sa.sa_flags & SA_IMMUTABLE) {
4117 		spin_unlock_irq(&p->sighand->siglock);
4118 		return -EINVAL;
4119 	}
4120 	if (oact)
4121 		*oact = *k;
4122 
4123 	/*
4124 	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4125 	 * e.g. by having an architecture use the bit in their uapi.
4126 	 */
4127 	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4128 
4129 	/*
4130 	 * Clear unknown flag bits in order to allow userspace to detect missing
4131 	 * support for flag bits and to allow the kernel to use non-uapi bits
4132 	 * internally.
4133 	 */
4134 	if (act)
4135 		act->sa.sa_flags &= UAPI_SA_FLAGS;
4136 	if (oact)
4137 		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4138 
4139 	sigaction_compat_abi(act, oact);
4140 
4141 	if (act) {
4142 		sigdelsetmask(&act->sa.sa_mask,
4143 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4144 		*k = *act;
4145 		/*
4146 		 * POSIX 3.3.1.3:
4147 		 *  "Setting a signal action to SIG_IGN for a signal that is
4148 		 *   pending shall cause the pending signal to be discarded,
4149 		 *   whether or not it is blocked."
4150 		 *
4151 		 *  "Setting a signal action to SIG_DFL for a signal that is
4152 		 *   pending and whose default action is to ignore the signal
4153 		 *   (for example, SIGCHLD), shall cause the pending signal to
4154 		 *   be discarded, whether or not it is blocked"
4155 		 */
4156 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4157 			sigemptyset(&mask);
4158 			sigaddset(&mask, sig);
4159 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4160 			for_each_thread(p, t)
4161 				flush_sigqueue_mask(&mask, &t->pending);
4162 		}
4163 	}
4164 
4165 	spin_unlock_irq(&p->sighand->siglock);
4166 	return 0;
4167 }
4168 
4169 #ifdef CONFIG_DYNAMIC_SIGFRAME
4170 static inline void sigaltstack_lock(void)
4171 	__acquires(&current->sighand->siglock)
4172 {
4173 	spin_lock_irq(&current->sighand->siglock);
4174 }
4175 
4176 static inline void sigaltstack_unlock(void)
4177 	__releases(&current->sighand->siglock)
4178 {
4179 	spin_unlock_irq(&current->sighand->siglock);
4180 }
4181 #else
4182 static inline void sigaltstack_lock(void) { }
4183 static inline void sigaltstack_unlock(void) { }
4184 #endif
4185 
4186 static int
4187 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4188 		size_t min_ss_size)
4189 {
4190 	struct task_struct *t = current;
4191 	int ret = 0;
4192 
4193 	if (oss) {
4194 		memset(oss, 0, sizeof(stack_t));
4195 		oss->ss_sp = (void __user *) t->sas_ss_sp;
4196 		oss->ss_size = t->sas_ss_size;
4197 		oss->ss_flags = sas_ss_flags(sp) |
4198 			(current->sas_ss_flags & SS_FLAG_BITS);
4199 	}
4200 
4201 	if (ss) {
4202 		void __user *ss_sp = ss->ss_sp;
4203 		size_t ss_size = ss->ss_size;
4204 		unsigned ss_flags = ss->ss_flags;
4205 		int ss_mode;
4206 
4207 		if (unlikely(on_sig_stack(sp)))
4208 			return -EPERM;
4209 
4210 		ss_mode = ss_flags & ~SS_FLAG_BITS;
4211 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4212 				ss_mode != 0))
4213 			return -EINVAL;
4214 
4215 		/*
4216 		 * Return before taking any locks if no actual
4217 		 * sigaltstack changes were requested.
4218 		 */
4219 		if (t->sas_ss_sp == (unsigned long)ss_sp &&
4220 		    t->sas_ss_size == ss_size &&
4221 		    t->sas_ss_flags == ss_flags)
4222 			return 0;
4223 
4224 		sigaltstack_lock();
4225 		if (ss_mode == SS_DISABLE) {
4226 			ss_size = 0;
4227 			ss_sp = NULL;
4228 		} else {
4229 			if (unlikely(ss_size < min_ss_size))
4230 				ret = -ENOMEM;
4231 			if (!sigaltstack_size_valid(ss_size))
4232 				ret = -ENOMEM;
4233 		}
4234 		if (!ret) {
4235 			t->sas_ss_sp = (unsigned long) ss_sp;
4236 			t->sas_ss_size = ss_size;
4237 			t->sas_ss_flags = ss_flags;
4238 		}
4239 		sigaltstack_unlock();
4240 	}
4241 	return ret;
4242 }
4243 
4244 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4245 {
4246 	stack_t new, old;
4247 	int err;
4248 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4249 		return -EFAULT;
4250 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4251 			      current_user_stack_pointer(),
4252 			      MINSIGSTKSZ);
4253 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4254 		err = -EFAULT;
4255 	return err;
4256 }
4257 
4258 int restore_altstack(const stack_t __user *uss)
4259 {
4260 	stack_t new;
4261 	if (copy_from_user(&new, uss, sizeof(stack_t)))
4262 		return -EFAULT;
4263 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4264 			     MINSIGSTKSZ);
4265 	/* squash all but EFAULT for now */
4266 	return 0;
4267 }
4268 
4269 int __save_altstack(stack_t __user *uss, unsigned long sp)
4270 {
4271 	struct task_struct *t = current;
4272 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4273 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4274 		__put_user(t->sas_ss_size, &uss->ss_size);
4275 	return err;
4276 }
4277 
4278 #ifdef CONFIG_COMPAT
4279 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4280 				 compat_stack_t __user *uoss_ptr)
4281 {
4282 	stack_t uss, uoss;
4283 	int ret;
4284 
4285 	if (uss_ptr) {
4286 		compat_stack_t uss32;
4287 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4288 			return -EFAULT;
4289 		uss.ss_sp = compat_ptr(uss32.ss_sp);
4290 		uss.ss_flags = uss32.ss_flags;
4291 		uss.ss_size = uss32.ss_size;
4292 	}
4293 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4294 			     compat_user_stack_pointer(),
4295 			     COMPAT_MINSIGSTKSZ);
4296 	if (ret >= 0 && uoss_ptr)  {
4297 		compat_stack_t old;
4298 		memset(&old, 0, sizeof(old));
4299 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4300 		old.ss_flags = uoss.ss_flags;
4301 		old.ss_size = uoss.ss_size;
4302 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4303 			ret = -EFAULT;
4304 	}
4305 	return ret;
4306 }
4307 
4308 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4309 			const compat_stack_t __user *, uss_ptr,
4310 			compat_stack_t __user *, uoss_ptr)
4311 {
4312 	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4313 }
4314 
4315 int compat_restore_altstack(const compat_stack_t __user *uss)
4316 {
4317 	int err = do_compat_sigaltstack(uss, NULL);
4318 	/* squash all but -EFAULT for now */
4319 	return err == -EFAULT ? err : 0;
4320 }
4321 
4322 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4323 {
4324 	int err;
4325 	struct task_struct *t = current;
4326 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4327 			 &uss->ss_sp) |
4328 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4329 		__put_user(t->sas_ss_size, &uss->ss_size);
4330 	return err;
4331 }
4332 #endif
4333 
4334 #ifdef __ARCH_WANT_SYS_SIGPENDING
4335 
4336 /**
4337  *  sys_sigpending - examine pending signals
4338  *  @uset: where mask of pending signal is returned
4339  */
4340 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4341 {
4342 	sigset_t set;
4343 
4344 	if (sizeof(old_sigset_t) > sizeof(*uset))
4345 		return -EINVAL;
4346 
4347 	do_sigpending(&set);
4348 
4349 	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4350 		return -EFAULT;
4351 
4352 	return 0;
4353 }
4354 
4355 #ifdef CONFIG_COMPAT
4356 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4357 {
4358 	sigset_t set;
4359 
4360 	do_sigpending(&set);
4361 
4362 	return put_user(set.sig[0], set32);
4363 }
4364 #endif
4365 
4366 #endif
4367 
4368 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4369 /**
4370  *  sys_sigprocmask - examine and change blocked signals
4371  *  @how: whether to add, remove, or set signals
4372  *  @nset: signals to add or remove (if non-null)
4373  *  @oset: previous value of signal mask if non-null
4374  *
4375  * Some platforms have their own version with special arguments;
4376  * others support only sys_rt_sigprocmask.
4377  */
4378 
4379 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4380 		old_sigset_t __user *, oset)
4381 {
4382 	old_sigset_t old_set, new_set;
4383 	sigset_t new_blocked;
4384 
4385 	old_set = current->blocked.sig[0];
4386 
4387 	if (nset) {
4388 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4389 			return -EFAULT;
4390 
4391 		new_blocked = current->blocked;
4392 
4393 		switch (how) {
4394 		case SIG_BLOCK:
4395 			sigaddsetmask(&new_blocked, new_set);
4396 			break;
4397 		case SIG_UNBLOCK:
4398 			sigdelsetmask(&new_blocked, new_set);
4399 			break;
4400 		case SIG_SETMASK:
4401 			new_blocked.sig[0] = new_set;
4402 			break;
4403 		default:
4404 			return -EINVAL;
4405 		}
4406 
4407 		set_current_blocked(&new_blocked);
4408 	}
4409 
4410 	if (oset) {
4411 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4412 			return -EFAULT;
4413 	}
4414 
4415 	return 0;
4416 }
4417 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4418 
4419 #ifndef CONFIG_ODD_RT_SIGACTION
4420 /**
4421  *  sys_rt_sigaction - alter an action taken by a process
4422  *  @sig: signal to be sent
4423  *  @act: new sigaction
4424  *  @oact: used to save the previous sigaction
4425  *  @sigsetsize: size of sigset_t type
4426  */
4427 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4428 		const struct sigaction __user *, act,
4429 		struct sigaction __user *, oact,
4430 		size_t, sigsetsize)
4431 {
4432 	struct k_sigaction new_sa, old_sa;
4433 	int ret;
4434 
4435 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4436 	if (sigsetsize != sizeof(sigset_t))
4437 		return -EINVAL;
4438 
4439 	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4440 		return -EFAULT;
4441 
4442 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4443 	if (ret)
4444 		return ret;
4445 
4446 	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4447 		return -EFAULT;
4448 
4449 	return 0;
4450 }
4451 #ifdef CONFIG_COMPAT
4452 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4453 		const struct compat_sigaction __user *, act,
4454 		struct compat_sigaction __user *, oact,
4455 		compat_size_t, sigsetsize)
4456 {
4457 	struct k_sigaction new_ka, old_ka;
4458 #ifdef __ARCH_HAS_SA_RESTORER
4459 	compat_uptr_t restorer;
4460 #endif
4461 	int ret;
4462 
4463 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4464 	if (sigsetsize != sizeof(compat_sigset_t))
4465 		return -EINVAL;
4466 
4467 	if (act) {
4468 		compat_uptr_t handler;
4469 		ret = get_user(handler, &act->sa_handler);
4470 		new_ka.sa.sa_handler = compat_ptr(handler);
4471 #ifdef __ARCH_HAS_SA_RESTORER
4472 		ret |= get_user(restorer, &act->sa_restorer);
4473 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4474 #endif
4475 		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4476 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4477 		if (ret)
4478 			return -EFAULT;
4479 	}
4480 
4481 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4482 	if (!ret && oact) {
4483 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4484 			       &oact->sa_handler);
4485 		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4486 					 sizeof(oact->sa_mask));
4487 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4488 #ifdef __ARCH_HAS_SA_RESTORER
4489 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4490 				&oact->sa_restorer);
4491 #endif
4492 	}
4493 	return ret;
4494 }
4495 #endif
4496 #endif /* !CONFIG_ODD_RT_SIGACTION */
4497 
4498 #ifdef CONFIG_OLD_SIGACTION
4499 SYSCALL_DEFINE3(sigaction, int, sig,
4500 		const struct old_sigaction __user *, act,
4501 	        struct old_sigaction __user *, oact)
4502 {
4503 	struct k_sigaction new_ka, old_ka;
4504 	int ret;
4505 
4506 	if (act) {
4507 		old_sigset_t mask;
4508 		if (!access_ok(act, sizeof(*act)) ||
4509 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4510 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4511 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4512 		    __get_user(mask, &act->sa_mask))
4513 			return -EFAULT;
4514 #ifdef __ARCH_HAS_KA_RESTORER
4515 		new_ka.ka_restorer = NULL;
4516 #endif
4517 		siginitset(&new_ka.sa.sa_mask, mask);
4518 	}
4519 
4520 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4521 
4522 	if (!ret && oact) {
4523 		if (!access_ok(oact, sizeof(*oact)) ||
4524 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4525 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4526 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4527 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4528 			return -EFAULT;
4529 	}
4530 
4531 	return ret;
4532 }
4533 #endif
4534 #ifdef CONFIG_COMPAT_OLD_SIGACTION
4535 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4536 		const struct compat_old_sigaction __user *, act,
4537 	        struct compat_old_sigaction __user *, oact)
4538 {
4539 	struct k_sigaction new_ka, old_ka;
4540 	int ret;
4541 	compat_old_sigset_t mask;
4542 	compat_uptr_t handler, restorer;
4543 
4544 	if (act) {
4545 		if (!access_ok(act, sizeof(*act)) ||
4546 		    __get_user(handler, &act->sa_handler) ||
4547 		    __get_user(restorer, &act->sa_restorer) ||
4548 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4549 		    __get_user(mask, &act->sa_mask))
4550 			return -EFAULT;
4551 
4552 #ifdef __ARCH_HAS_KA_RESTORER
4553 		new_ka.ka_restorer = NULL;
4554 #endif
4555 		new_ka.sa.sa_handler = compat_ptr(handler);
4556 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4557 		siginitset(&new_ka.sa.sa_mask, mask);
4558 	}
4559 
4560 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4561 
4562 	if (!ret && oact) {
4563 		if (!access_ok(oact, sizeof(*oact)) ||
4564 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4565 			       &oact->sa_handler) ||
4566 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4567 			       &oact->sa_restorer) ||
4568 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4569 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4570 			return -EFAULT;
4571 	}
4572 	return ret;
4573 }
4574 #endif
4575 
4576 #ifdef CONFIG_SGETMASK_SYSCALL
4577 
4578 /*
4579  * For backwards compatibility.  Functionality superseded by sigprocmask.
4580  */
4581 SYSCALL_DEFINE0(sgetmask)
4582 {
4583 	/* SMP safe */
4584 	return current->blocked.sig[0];
4585 }
4586 
4587 SYSCALL_DEFINE1(ssetmask, int, newmask)
4588 {
4589 	int old = current->blocked.sig[0];
4590 	sigset_t newset;
4591 
4592 	siginitset(&newset, newmask);
4593 	set_current_blocked(&newset);
4594 
4595 	return old;
4596 }
4597 #endif /* CONFIG_SGETMASK_SYSCALL */
4598 
4599 #ifdef __ARCH_WANT_SYS_SIGNAL
4600 /*
4601  * For backwards compatibility.  Functionality superseded by sigaction.
4602  */
4603 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4604 {
4605 	struct k_sigaction new_sa, old_sa;
4606 	int ret;
4607 
4608 	new_sa.sa.sa_handler = handler;
4609 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4610 	sigemptyset(&new_sa.sa.sa_mask);
4611 
4612 	ret = do_sigaction(sig, &new_sa, &old_sa);
4613 
4614 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4615 }
4616 #endif /* __ARCH_WANT_SYS_SIGNAL */
4617 
4618 #ifdef __ARCH_WANT_SYS_PAUSE
4619 
4620 SYSCALL_DEFINE0(pause)
4621 {
4622 	while (!signal_pending(current)) {
4623 		__set_current_state(TASK_INTERRUPTIBLE);
4624 		schedule();
4625 	}
4626 	return -ERESTARTNOHAND;
4627 }
4628 
4629 #endif
4630 
4631 static int sigsuspend(sigset_t *set)
4632 {
4633 	current->saved_sigmask = current->blocked;
4634 	set_current_blocked(set);
4635 
4636 	while (!signal_pending(current)) {
4637 		__set_current_state(TASK_INTERRUPTIBLE);
4638 		schedule();
4639 	}
4640 	set_restore_sigmask();
4641 	return -ERESTARTNOHAND;
4642 }
4643 
4644 /**
4645  *  sys_rt_sigsuspend - replace the signal mask for a value with the
4646  *	@unewset value until a signal is received
4647  *  @unewset: new signal mask value
4648  *  @sigsetsize: size of sigset_t type
4649  */
4650 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4651 {
4652 	sigset_t newset;
4653 
4654 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4655 	if (sigsetsize != sizeof(sigset_t))
4656 		return -EINVAL;
4657 
4658 	if (copy_from_user(&newset, unewset, sizeof(newset)))
4659 		return -EFAULT;
4660 	return sigsuspend(&newset);
4661 }
4662 
4663 #ifdef CONFIG_COMPAT
4664 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4665 {
4666 	sigset_t newset;
4667 
4668 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4669 	if (sigsetsize != sizeof(sigset_t))
4670 		return -EINVAL;
4671 
4672 	if (get_compat_sigset(&newset, unewset))
4673 		return -EFAULT;
4674 	return sigsuspend(&newset);
4675 }
4676 #endif
4677 
4678 #ifdef CONFIG_OLD_SIGSUSPEND
4679 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4680 {
4681 	sigset_t blocked;
4682 	siginitset(&blocked, mask);
4683 	return sigsuspend(&blocked);
4684 }
4685 #endif
4686 #ifdef CONFIG_OLD_SIGSUSPEND3
4687 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4688 {
4689 	sigset_t blocked;
4690 	siginitset(&blocked, mask);
4691 	return sigsuspend(&blocked);
4692 }
4693 #endif
4694 
4695 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4696 {
4697 	return NULL;
4698 }
4699 
4700 static inline void siginfo_buildtime_checks(void)
4701 {
4702 	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4703 
4704 	/* Verify the offsets in the two siginfos match */
4705 #define CHECK_OFFSET(field) \
4706 	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4707 
4708 	/* kill */
4709 	CHECK_OFFSET(si_pid);
4710 	CHECK_OFFSET(si_uid);
4711 
4712 	/* timer */
4713 	CHECK_OFFSET(si_tid);
4714 	CHECK_OFFSET(si_overrun);
4715 	CHECK_OFFSET(si_value);
4716 
4717 	/* rt */
4718 	CHECK_OFFSET(si_pid);
4719 	CHECK_OFFSET(si_uid);
4720 	CHECK_OFFSET(si_value);
4721 
4722 	/* sigchld */
4723 	CHECK_OFFSET(si_pid);
4724 	CHECK_OFFSET(si_uid);
4725 	CHECK_OFFSET(si_status);
4726 	CHECK_OFFSET(si_utime);
4727 	CHECK_OFFSET(si_stime);
4728 
4729 	/* sigfault */
4730 	CHECK_OFFSET(si_addr);
4731 	CHECK_OFFSET(si_trapno);
4732 	CHECK_OFFSET(si_addr_lsb);
4733 	CHECK_OFFSET(si_lower);
4734 	CHECK_OFFSET(si_upper);
4735 	CHECK_OFFSET(si_pkey);
4736 	CHECK_OFFSET(si_perf_data);
4737 	CHECK_OFFSET(si_perf_type);
4738 	CHECK_OFFSET(si_perf_flags);
4739 
4740 	/* sigpoll */
4741 	CHECK_OFFSET(si_band);
4742 	CHECK_OFFSET(si_fd);
4743 
4744 	/* sigsys */
4745 	CHECK_OFFSET(si_call_addr);
4746 	CHECK_OFFSET(si_syscall);
4747 	CHECK_OFFSET(si_arch);
4748 #undef CHECK_OFFSET
4749 
4750 	/* usb asyncio */
4751 	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4752 		     offsetof(struct siginfo, si_addr));
4753 	if (sizeof(int) == sizeof(void __user *)) {
4754 		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4755 			     sizeof(void __user *));
4756 	} else {
4757 		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4758 			      sizeof_field(struct siginfo, si_uid)) !=
4759 			     sizeof(void __user *));
4760 		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4761 			     offsetof(struct siginfo, si_uid));
4762 	}
4763 #ifdef CONFIG_COMPAT
4764 	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4765 		     offsetof(struct compat_siginfo, si_addr));
4766 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4767 		     sizeof(compat_uptr_t));
4768 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4769 		     sizeof_field(struct siginfo, si_pid));
4770 #endif
4771 }
4772 
4773 void __init signals_init(void)
4774 {
4775 	siginfo_buildtime_checks();
4776 
4777 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
4778 }
4779 
4780 #ifdef CONFIG_KGDB_KDB
4781 #include <linux/kdb.h>
4782 /*
4783  * kdb_send_sig - Allows kdb to send signals without exposing
4784  * signal internals.  This function checks if the required locks are
4785  * available before calling the main signal code, to avoid kdb
4786  * deadlocks.
4787  */
4788 void kdb_send_sig(struct task_struct *t, int sig)
4789 {
4790 	static struct task_struct *kdb_prev_t;
4791 	int new_t, ret;
4792 	if (!spin_trylock(&t->sighand->siglock)) {
4793 		kdb_printf("Can't do kill command now.\n"
4794 			   "The sigmask lock is held somewhere else in "
4795 			   "kernel, try again later\n");
4796 		return;
4797 	}
4798 	new_t = kdb_prev_t != t;
4799 	kdb_prev_t = t;
4800 	if (!task_is_running(t) && new_t) {
4801 		spin_unlock(&t->sighand->siglock);
4802 		kdb_printf("Process is not RUNNING, sending a signal from "
4803 			   "kdb risks deadlock\n"
4804 			   "on the run queue locks. "
4805 			   "The signal has _not_ been sent.\n"
4806 			   "Reissue the kill command if you want to risk "
4807 			   "the deadlock.\n");
4808 		return;
4809 	}
4810 	ret = send_signal(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4811 	spin_unlock(&t->sighand->siglock);
4812 	if (ret)
4813 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
4814 			   sig, t->pid);
4815 	else
4816 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4817 }
4818 #endif	/* CONFIG_KGDB_KDB */
4819