xref: /openbmc/linux/kernel/signal.c (revision 9d64fc08)
1 /*
2  *  linux/kernel/signal.c
3  *
4  *  Copyright (C) 1991, 1992  Linus Torvalds
5  *
6  *  1997-11-02  Modified for POSIX.1b signals by Richard Henderson
7  *
8  *  2003-06-02  Jim Houston - Concurrent Computer Corp.
9  *		Changes to use preallocated sigqueue structures
10  *		to allow signals to be sent reliably.
11  */
12 
13 #include <linux/slab.h>
14 #include <linux/export.h>
15 #include <linux/init.h>
16 #include <linux/sched/mm.h>
17 #include <linux/sched/user.h>
18 #include <linux/sched/debug.h>
19 #include <linux/sched/task.h>
20 #include <linux/sched/task_stack.h>
21 #include <linux/sched/cputime.h>
22 #include <linux/fs.h>
23 #include <linux/tty.h>
24 #include <linux/binfmts.h>
25 #include <linux/coredump.h>
26 #include <linux/security.h>
27 #include <linux/syscalls.h>
28 #include <linux/ptrace.h>
29 #include <linux/signal.h>
30 #include <linux/signalfd.h>
31 #include <linux/ratelimit.h>
32 #include <linux/tracehook.h>
33 #include <linux/capability.h>
34 #include <linux/freezer.h>
35 #include <linux/pid_namespace.h>
36 #include <linux/nsproxy.h>
37 #include <linux/user_namespace.h>
38 #include <linux/uprobes.h>
39 #include <linux/compat.h>
40 #include <linux/cn_proc.h>
41 #include <linux/compiler.h>
42 #include <linux/posix-timers.h>
43 
44 #define CREATE_TRACE_POINTS
45 #include <trace/events/signal.h>
46 
47 #include <asm/param.h>
48 #include <linux/uaccess.h>
49 #include <asm/unistd.h>
50 #include <asm/siginfo.h>
51 #include <asm/cacheflush.h>
52 #include "audit.h"	/* audit_signal_info() */
53 
54 /*
55  * SLAB caches for signal bits.
56  */
57 
58 static struct kmem_cache *sigqueue_cachep;
59 
60 int print_fatal_signals __read_mostly;
61 
62 static void __user *sig_handler(struct task_struct *t, int sig)
63 {
64 	return t->sighand->action[sig - 1].sa.sa_handler;
65 }
66 
67 static int sig_handler_ignored(void __user *handler, int sig)
68 {
69 	/* Is it explicitly or implicitly ignored? */
70 	return handler == SIG_IGN ||
71 		(handler == SIG_DFL && sig_kernel_ignore(sig));
72 }
73 
74 static int sig_task_ignored(struct task_struct *t, int sig, bool force)
75 {
76 	void __user *handler;
77 
78 	handler = sig_handler(t, sig);
79 
80 	if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
81 			handler == SIG_DFL && !force)
82 		return 1;
83 
84 	return sig_handler_ignored(handler, sig);
85 }
86 
87 static int sig_ignored(struct task_struct *t, int sig, bool force)
88 {
89 	/*
90 	 * Blocked signals are never ignored, since the
91 	 * signal handler may change by the time it is
92 	 * unblocked.
93 	 */
94 	if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
95 		return 0;
96 
97 	if (!sig_task_ignored(t, sig, force))
98 		return 0;
99 
100 	/*
101 	 * Tracers may want to know about even ignored signals.
102 	 */
103 	return !t->ptrace;
104 }
105 
106 /*
107  * Re-calculate pending state from the set of locally pending
108  * signals, globally pending signals, and blocked signals.
109  */
110 static inline int has_pending_signals(sigset_t *signal, sigset_t *blocked)
111 {
112 	unsigned long ready;
113 	long i;
114 
115 	switch (_NSIG_WORDS) {
116 	default:
117 		for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
118 			ready |= signal->sig[i] &~ blocked->sig[i];
119 		break;
120 
121 	case 4: ready  = signal->sig[3] &~ blocked->sig[3];
122 		ready |= signal->sig[2] &~ blocked->sig[2];
123 		ready |= signal->sig[1] &~ blocked->sig[1];
124 		ready |= signal->sig[0] &~ blocked->sig[0];
125 		break;
126 
127 	case 2: ready  = signal->sig[1] &~ blocked->sig[1];
128 		ready |= signal->sig[0] &~ blocked->sig[0];
129 		break;
130 
131 	case 1: ready  = signal->sig[0] &~ blocked->sig[0];
132 	}
133 	return ready !=	0;
134 }
135 
136 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
137 
138 static int recalc_sigpending_tsk(struct task_struct *t)
139 {
140 	if ((t->jobctl & JOBCTL_PENDING_MASK) ||
141 	    PENDING(&t->pending, &t->blocked) ||
142 	    PENDING(&t->signal->shared_pending, &t->blocked)) {
143 		set_tsk_thread_flag(t, TIF_SIGPENDING);
144 		return 1;
145 	}
146 	/*
147 	 * We must never clear the flag in another thread, or in current
148 	 * when it's possible the current syscall is returning -ERESTART*.
149 	 * So we don't clear it here, and only callers who know they should do.
150 	 */
151 	return 0;
152 }
153 
154 /*
155  * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
156  * This is superfluous when called on current, the wakeup is a harmless no-op.
157  */
158 void recalc_sigpending_and_wake(struct task_struct *t)
159 {
160 	if (recalc_sigpending_tsk(t))
161 		signal_wake_up(t, 0);
162 }
163 
164 void recalc_sigpending(void)
165 {
166 	if (!recalc_sigpending_tsk(current) && !freezing(current))
167 		clear_thread_flag(TIF_SIGPENDING);
168 
169 }
170 
171 /* Given the mask, find the first available signal that should be serviced. */
172 
173 #define SYNCHRONOUS_MASK \
174 	(sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
175 	 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
176 
177 int next_signal(struct sigpending *pending, sigset_t *mask)
178 {
179 	unsigned long i, *s, *m, x;
180 	int sig = 0;
181 
182 	s = pending->signal.sig;
183 	m = mask->sig;
184 
185 	/*
186 	 * Handle the first word specially: it contains the
187 	 * synchronous signals that need to be dequeued first.
188 	 */
189 	x = *s &~ *m;
190 	if (x) {
191 		if (x & SYNCHRONOUS_MASK)
192 			x &= SYNCHRONOUS_MASK;
193 		sig = ffz(~x) + 1;
194 		return sig;
195 	}
196 
197 	switch (_NSIG_WORDS) {
198 	default:
199 		for (i = 1; i < _NSIG_WORDS; ++i) {
200 			x = *++s &~ *++m;
201 			if (!x)
202 				continue;
203 			sig = ffz(~x) + i*_NSIG_BPW + 1;
204 			break;
205 		}
206 		break;
207 
208 	case 2:
209 		x = s[1] &~ m[1];
210 		if (!x)
211 			break;
212 		sig = ffz(~x) + _NSIG_BPW + 1;
213 		break;
214 
215 	case 1:
216 		/* Nothing to do */
217 		break;
218 	}
219 
220 	return sig;
221 }
222 
223 static inline void print_dropped_signal(int sig)
224 {
225 	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
226 
227 	if (!print_fatal_signals)
228 		return;
229 
230 	if (!__ratelimit(&ratelimit_state))
231 		return;
232 
233 	pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
234 				current->comm, current->pid, sig);
235 }
236 
237 /**
238  * task_set_jobctl_pending - set jobctl pending bits
239  * @task: target task
240  * @mask: pending bits to set
241  *
242  * Clear @mask from @task->jobctl.  @mask must be subset of
243  * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
244  * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
245  * cleared.  If @task is already being killed or exiting, this function
246  * becomes noop.
247  *
248  * CONTEXT:
249  * Must be called with @task->sighand->siglock held.
250  *
251  * RETURNS:
252  * %true if @mask is set, %false if made noop because @task was dying.
253  */
254 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
255 {
256 	BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
257 			JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
258 	BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
259 
260 	if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
261 		return false;
262 
263 	if (mask & JOBCTL_STOP_SIGMASK)
264 		task->jobctl &= ~JOBCTL_STOP_SIGMASK;
265 
266 	task->jobctl |= mask;
267 	return true;
268 }
269 
270 /**
271  * task_clear_jobctl_trapping - clear jobctl trapping bit
272  * @task: target task
273  *
274  * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
275  * Clear it and wake up the ptracer.  Note that we don't need any further
276  * locking.  @task->siglock guarantees that @task->parent points to the
277  * ptracer.
278  *
279  * CONTEXT:
280  * Must be called with @task->sighand->siglock held.
281  */
282 void task_clear_jobctl_trapping(struct task_struct *task)
283 {
284 	if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
285 		task->jobctl &= ~JOBCTL_TRAPPING;
286 		smp_mb();	/* advised by wake_up_bit() */
287 		wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
288 	}
289 }
290 
291 /**
292  * task_clear_jobctl_pending - clear jobctl pending bits
293  * @task: target task
294  * @mask: pending bits to clear
295  *
296  * Clear @mask from @task->jobctl.  @mask must be subset of
297  * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
298  * STOP bits are cleared together.
299  *
300  * If clearing of @mask leaves no stop or trap pending, this function calls
301  * task_clear_jobctl_trapping().
302  *
303  * CONTEXT:
304  * Must be called with @task->sighand->siglock held.
305  */
306 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
307 {
308 	BUG_ON(mask & ~JOBCTL_PENDING_MASK);
309 
310 	if (mask & JOBCTL_STOP_PENDING)
311 		mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
312 
313 	task->jobctl &= ~mask;
314 
315 	if (!(task->jobctl & JOBCTL_PENDING_MASK))
316 		task_clear_jobctl_trapping(task);
317 }
318 
319 /**
320  * task_participate_group_stop - participate in a group stop
321  * @task: task participating in a group stop
322  *
323  * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
324  * Group stop states are cleared and the group stop count is consumed if
325  * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
326  * stop, the appropriate %SIGNAL_* flags are set.
327  *
328  * CONTEXT:
329  * Must be called with @task->sighand->siglock held.
330  *
331  * RETURNS:
332  * %true if group stop completion should be notified to the parent, %false
333  * otherwise.
334  */
335 static bool task_participate_group_stop(struct task_struct *task)
336 {
337 	struct signal_struct *sig = task->signal;
338 	bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
339 
340 	WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
341 
342 	task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
343 
344 	if (!consume)
345 		return false;
346 
347 	if (!WARN_ON_ONCE(sig->group_stop_count == 0))
348 		sig->group_stop_count--;
349 
350 	/*
351 	 * Tell the caller to notify completion iff we are entering into a
352 	 * fresh group stop.  Read comment in do_signal_stop() for details.
353 	 */
354 	if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
355 		signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
356 		return true;
357 	}
358 	return false;
359 }
360 
361 /*
362  * allocate a new signal queue record
363  * - this may be called without locks if and only if t == current, otherwise an
364  *   appropriate lock must be held to stop the target task from exiting
365  */
366 static struct sigqueue *
367 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t flags, int override_rlimit)
368 {
369 	struct sigqueue *q = NULL;
370 	struct user_struct *user;
371 
372 	/*
373 	 * Protect access to @t credentials. This can go away when all
374 	 * callers hold rcu read lock.
375 	 */
376 	rcu_read_lock();
377 	user = get_uid(__task_cred(t)->user);
378 	atomic_inc(&user->sigpending);
379 	rcu_read_unlock();
380 
381 	if (override_rlimit ||
382 	    atomic_read(&user->sigpending) <=
383 			task_rlimit(t, RLIMIT_SIGPENDING)) {
384 		q = kmem_cache_alloc(sigqueue_cachep, flags);
385 	} else {
386 		print_dropped_signal(sig);
387 	}
388 
389 	if (unlikely(q == NULL)) {
390 		atomic_dec(&user->sigpending);
391 		free_uid(user);
392 	} else {
393 		INIT_LIST_HEAD(&q->list);
394 		q->flags = 0;
395 		q->user = user;
396 	}
397 
398 	return q;
399 }
400 
401 static void __sigqueue_free(struct sigqueue *q)
402 {
403 	if (q->flags & SIGQUEUE_PREALLOC)
404 		return;
405 	atomic_dec(&q->user->sigpending);
406 	free_uid(q->user);
407 	kmem_cache_free(sigqueue_cachep, q);
408 }
409 
410 void flush_sigqueue(struct sigpending *queue)
411 {
412 	struct sigqueue *q;
413 
414 	sigemptyset(&queue->signal);
415 	while (!list_empty(&queue->list)) {
416 		q = list_entry(queue->list.next, struct sigqueue , list);
417 		list_del_init(&q->list);
418 		__sigqueue_free(q);
419 	}
420 }
421 
422 /*
423  * Flush all pending signals for this kthread.
424  */
425 void flush_signals(struct task_struct *t)
426 {
427 	unsigned long flags;
428 
429 	spin_lock_irqsave(&t->sighand->siglock, flags);
430 	clear_tsk_thread_flag(t, TIF_SIGPENDING);
431 	flush_sigqueue(&t->pending);
432 	flush_sigqueue(&t->signal->shared_pending);
433 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
434 }
435 
436 #ifdef CONFIG_POSIX_TIMERS
437 static void __flush_itimer_signals(struct sigpending *pending)
438 {
439 	sigset_t signal, retain;
440 	struct sigqueue *q, *n;
441 
442 	signal = pending->signal;
443 	sigemptyset(&retain);
444 
445 	list_for_each_entry_safe(q, n, &pending->list, list) {
446 		int sig = q->info.si_signo;
447 
448 		if (likely(q->info.si_code != SI_TIMER)) {
449 			sigaddset(&retain, sig);
450 		} else {
451 			sigdelset(&signal, sig);
452 			list_del_init(&q->list);
453 			__sigqueue_free(q);
454 		}
455 	}
456 
457 	sigorsets(&pending->signal, &signal, &retain);
458 }
459 
460 void flush_itimer_signals(void)
461 {
462 	struct task_struct *tsk = current;
463 	unsigned long flags;
464 
465 	spin_lock_irqsave(&tsk->sighand->siglock, flags);
466 	__flush_itimer_signals(&tsk->pending);
467 	__flush_itimer_signals(&tsk->signal->shared_pending);
468 	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
469 }
470 #endif
471 
472 void ignore_signals(struct task_struct *t)
473 {
474 	int i;
475 
476 	for (i = 0; i < _NSIG; ++i)
477 		t->sighand->action[i].sa.sa_handler = SIG_IGN;
478 
479 	flush_signals(t);
480 }
481 
482 /*
483  * Flush all handlers for a task.
484  */
485 
486 void
487 flush_signal_handlers(struct task_struct *t, int force_default)
488 {
489 	int i;
490 	struct k_sigaction *ka = &t->sighand->action[0];
491 	for (i = _NSIG ; i != 0 ; i--) {
492 		if (force_default || ka->sa.sa_handler != SIG_IGN)
493 			ka->sa.sa_handler = SIG_DFL;
494 		ka->sa.sa_flags = 0;
495 #ifdef __ARCH_HAS_SA_RESTORER
496 		ka->sa.sa_restorer = NULL;
497 #endif
498 		sigemptyset(&ka->sa.sa_mask);
499 		ka++;
500 	}
501 }
502 
503 int unhandled_signal(struct task_struct *tsk, int sig)
504 {
505 	void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
506 	if (is_global_init(tsk))
507 		return 1;
508 	if (handler != SIG_IGN && handler != SIG_DFL)
509 		return 0;
510 	/* if ptraced, let the tracer determine */
511 	return !tsk->ptrace;
512 }
513 
514 static void collect_signal(int sig, struct sigpending *list, siginfo_t *info,
515 			   bool *resched_timer)
516 {
517 	struct sigqueue *q, *first = NULL;
518 
519 	/*
520 	 * Collect the siginfo appropriate to this signal.  Check if
521 	 * there is another siginfo for the same signal.
522 	*/
523 	list_for_each_entry(q, &list->list, list) {
524 		if (q->info.si_signo == sig) {
525 			if (first)
526 				goto still_pending;
527 			first = q;
528 		}
529 	}
530 
531 	sigdelset(&list->signal, sig);
532 
533 	if (first) {
534 still_pending:
535 		list_del_init(&first->list);
536 		copy_siginfo(info, &first->info);
537 
538 		*resched_timer =
539 			(first->flags & SIGQUEUE_PREALLOC) &&
540 			(info->si_code == SI_TIMER) &&
541 			(info->si_sys_private);
542 
543 		__sigqueue_free(first);
544 	} else {
545 		/*
546 		 * Ok, it wasn't in the queue.  This must be
547 		 * a fast-pathed signal or we must have been
548 		 * out of queue space.  So zero out the info.
549 		 */
550 		info->si_signo = sig;
551 		info->si_errno = 0;
552 		info->si_code = SI_USER;
553 		info->si_pid = 0;
554 		info->si_uid = 0;
555 	}
556 }
557 
558 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
559 			siginfo_t *info, bool *resched_timer)
560 {
561 	int sig = next_signal(pending, mask);
562 
563 	if (sig)
564 		collect_signal(sig, pending, info, resched_timer);
565 	return sig;
566 }
567 
568 /*
569  * Dequeue a signal and return the element to the caller, which is
570  * expected to free it.
571  *
572  * All callers have to hold the siglock.
573  */
574 int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info)
575 {
576 	bool resched_timer = false;
577 	int signr;
578 
579 	/* We only dequeue private signals from ourselves, we don't let
580 	 * signalfd steal them
581 	 */
582 	signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
583 	if (!signr) {
584 		signr = __dequeue_signal(&tsk->signal->shared_pending,
585 					 mask, info, &resched_timer);
586 #ifdef CONFIG_POSIX_TIMERS
587 		/*
588 		 * itimer signal ?
589 		 *
590 		 * itimers are process shared and we restart periodic
591 		 * itimers in the signal delivery path to prevent DoS
592 		 * attacks in the high resolution timer case. This is
593 		 * compliant with the old way of self-restarting
594 		 * itimers, as the SIGALRM is a legacy signal and only
595 		 * queued once. Changing the restart behaviour to
596 		 * restart the timer in the signal dequeue path is
597 		 * reducing the timer noise on heavy loaded !highres
598 		 * systems too.
599 		 */
600 		if (unlikely(signr == SIGALRM)) {
601 			struct hrtimer *tmr = &tsk->signal->real_timer;
602 
603 			if (!hrtimer_is_queued(tmr) &&
604 			    tsk->signal->it_real_incr != 0) {
605 				hrtimer_forward(tmr, tmr->base->get_time(),
606 						tsk->signal->it_real_incr);
607 				hrtimer_restart(tmr);
608 			}
609 		}
610 #endif
611 	}
612 
613 	recalc_sigpending();
614 	if (!signr)
615 		return 0;
616 
617 	if (unlikely(sig_kernel_stop(signr))) {
618 		/*
619 		 * Set a marker that we have dequeued a stop signal.  Our
620 		 * caller might release the siglock and then the pending
621 		 * stop signal it is about to process is no longer in the
622 		 * pending bitmasks, but must still be cleared by a SIGCONT
623 		 * (and overruled by a SIGKILL).  So those cases clear this
624 		 * shared flag after we've set it.  Note that this flag may
625 		 * remain set after the signal we return is ignored or
626 		 * handled.  That doesn't matter because its only purpose
627 		 * is to alert stop-signal processing code when another
628 		 * processor has come along and cleared the flag.
629 		 */
630 		current->jobctl |= JOBCTL_STOP_DEQUEUED;
631 	}
632 #ifdef CONFIG_POSIX_TIMERS
633 	if (resched_timer) {
634 		/*
635 		 * Release the siglock to ensure proper locking order
636 		 * of timer locks outside of siglocks.  Note, we leave
637 		 * irqs disabled here, since the posix-timers code is
638 		 * about to disable them again anyway.
639 		 */
640 		spin_unlock(&tsk->sighand->siglock);
641 		posixtimer_rearm(info);
642 		spin_lock(&tsk->sighand->siglock);
643 	}
644 #endif
645 	return signr;
646 }
647 
648 /*
649  * Tell a process that it has a new active signal..
650  *
651  * NOTE! we rely on the previous spin_lock to
652  * lock interrupts for us! We can only be called with
653  * "siglock" held, and the local interrupt must
654  * have been disabled when that got acquired!
655  *
656  * No need to set need_resched since signal event passing
657  * goes through ->blocked
658  */
659 void signal_wake_up_state(struct task_struct *t, unsigned int state)
660 {
661 	set_tsk_thread_flag(t, TIF_SIGPENDING);
662 	/*
663 	 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
664 	 * case. We don't check t->state here because there is a race with it
665 	 * executing another processor and just now entering stopped state.
666 	 * By using wake_up_state, we ensure the process will wake up and
667 	 * handle its death signal.
668 	 */
669 	if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
670 		kick_process(t);
671 }
672 
673 /*
674  * Remove signals in mask from the pending set and queue.
675  * Returns 1 if any signals were found.
676  *
677  * All callers must be holding the siglock.
678  */
679 static int flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
680 {
681 	struct sigqueue *q, *n;
682 	sigset_t m;
683 
684 	sigandsets(&m, mask, &s->signal);
685 	if (sigisemptyset(&m))
686 		return 0;
687 
688 	sigandnsets(&s->signal, &s->signal, mask);
689 	list_for_each_entry_safe(q, n, &s->list, list) {
690 		if (sigismember(mask, q->info.si_signo)) {
691 			list_del_init(&q->list);
692 			__sigqueue_free(q);
693 		}
694 	}
695 	return 1;
696 }
697 
698 static inline int is_si_special(const struct siginfo *info)
699 {
700 	return info <= SEND_SIG_FORCED;
701 }
702 
703 static inline bool si_fromuser(const struct siginfo *info)
704 {
705 	return info == SEND_SIG_NOINFO ||
706 		(!is_si_special(info) && SI_FROMUSER(info));
707 }
708 
709 /*
710  * called with RCU read lock from check_kill_permission()
711  */
712 static int kill_ok_by_cred(struct task_struct *t)
713 {
714 	const struct cred *cred = current_cred();
715 	const struct cred *tcred = __task_cred(t);
716 
717 	if (uid_eq(cred->euid, tcred->suid) ||
718 	    uid_eq(cred->euid, tcred->uid)  ||
719 	    uid_eq(cred->uid,  tcred->suid) ||
720 	    uid_eq(cred->uid,  tcred->uid))
721 		return 1;
722 
723 	if (ns_capable(tcred->user_ns, CAP_KILL))
724 		return 1;
725 
726 	return 0;
727 }
728 
729 /*
730  * Bad permissions for sending the signal
731  * - the caller must hold the RCU read lock
732  */
733 static int check_kill_permission(int sig, struct siginfo *info,
734 				 struct task_struct *t)
735 {
736 	struct pid *sid;
737 	int error;
738 
739 	if (!valid_signal(sig))
740 		return -EINVAL;
741 
742 	if (!si_fromuser(info))
743 		return 0;
744 
745 	error = audit_signal_info(sig, t); /* Let audit system see the signal */
746 	if (error)
747 		return error;
748 
749 	if (!same_thread_group(current, t) &&
750 	    !kill_ok_by_cred(t)) {
751 		switch (sig) {
752 		case SIGCONT:
753 			sid = task_session(t);
754 			/*
755 			 * We don't return the error if sid == NULL. The
756 			 * task was unhashed, the caller must notice this.
757 			 */
758 			if (!sid || sid == task_session(current))
759 				break;
760 		default:
761 			return -EPERM;
762 		}
763 	}
764 
765 	return security_task_kill(t, info, sig, 0);
766 }
767 
768 /**
769  * ptrace_trap_notify - schedule trap to notify ptracer
770  * @t: tracee wanting to notify tracer
771  *
772  * This function schedules sticky ptrace trap which is cleared on the next
773  * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
774  * ptracer.
775  *
776  * If @t is running, STOP trap will be taken.  If trapped for STOP and
777  * ptracer is listening for events, tracee is woken up so that it can
778  * re-trap for the new event.  If trapped otherwise, STOP trap will be
779  * eventually taken without returning to userland after the existing traps
780  * are finished by PTRACE_CONT.
781  *
782  * CONTEXT:
783  * Must be called with @task->sighand->siglock held.
784  */
785 static void ptrace_trap_notify(struct task_struct *t)
786 {
787 	WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
788 	assert_spin_locked(&t->sighand->siglock);
789 
790 	task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
791 	ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
792 }
793 
794 /*
795  * Handle magic process-wide effects of stop/continue signals. Unlike
796  * the signal actions, these happen immediately at signal-generation
797  * time regardless of blocking, ignoring, or handling.  This does the
798  * actual continuing for SIGCONT, but not the actual stopping for stop
799  * signals. The process stop is done as a signal action for SIG_DFL.
800  *
801  * Returns true if the signal should be actually delivered, otherwise
802  * it should be dropped.
803  */
804 static bool prepare_signal(int sig, struct task_struct *p, bool force)
805 {
806 	struct signal_struct *signal = p->signal;
807 	struct task_struct *t;
808 	sigset_t flush;
809 
810 	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
811 		if (!(signal->flags & SIGNAL_GROUP_EXIT))
812 			return sig == SIGKILL;
813 		/*
814 		 * The process is in the middle of dying, nothing to do.
815 		 */
816 	} else if (sig_kernel_stop(sig)) {
817 		/*
818 		 * This is a stop signal.  Remove SIGCONT from all queues.
819 		 */
820 		siginitset(&flush, sigmask(SIGCONT));
821 		flush_sigqueue_mask(&flush, &signal->shared_pending);
822 		for_each_thread(p, t)
823 			flush_sigqueue_mask(&flush, &t->pending);
824 	} else if (sig == SIGCONT) {
825 		unsigned int why;
826 		/*
827 		 * Remove all stop signals from all queues, wake all threads.
828 		 */
829 		siginitset(&flush, SIG_KERNEL_STOP_MASK);
830 		flush_sigqueue_mask(&flush, &signal->shared_pending);
831 		for_each_thread(p, t) {
832 			flush_sigqueue_mask(&flush, &t->pending);
833 			task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
834 			if (likely(!(t->ptrace & PT_SEIZED)))
835 				wake_up_state(t, __TASK_STOPPED);
836 			else
837 				ptrace_trap_notify(t);
838 		}
839 
840 		/*
841 		 * Notify the parent with CLD_CONTINUED if we were stopped.
842 		 *
843 		 * If we were in the middle of a group stop, we pretend it
844 		 * was already finished, and then continued. Since SIGCHLD
845 		 * doesn't queue we report only CLD_STOPPED, as if the next
846 		 * CLD_CONTINUED was dropped.
847 		 */
848 		why = 0;
849 		if (signal->flags & SIGNAL_STOP_STOPPED)
850 			why |= SIGNAL_CLD_CONTINUED;
851 		else if (signal->group_stop_count)
852 			why |= SIGNAL_CLD_STOPPED;
853 
854 		if (why) {
855 			/*
856 			 * The first thread which returns from do_signal_stop()
857 			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
858 			 * notify its parent. See get_signal_to_deliver().
859 			 */
860 			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
861 			signal->group_stop_count = 0;
862 			signal->group_exit_code = 0;
863 		}
864 	}
865 
866 	return !sig_ignored(p, sig, force);
867 }
868 
869 /*
870  * Test if P wants to take SIG.  After we've checked all threads with this,
871  * it's equivalent to finding no threads not blocking SIG.  Any threads not
872  * blocking SIG were ruled out because they are not running and already
873  * have pending signals.  Such threads will dequeue from the shared queue
874  * as soon as they're available, so putting the signal on the shared queue
875  * will be equivalent to sending it to one such thread.
876  */
877 static inline int wants_signal(int sig, struct task_struct *p)
878 {
879 	if (sigismember(&p->blocked, sig))
880 		return 0;
881 	if (p->flags & PF_EXITING)
882 		return 0;
883 	if (sig == SIGKILL)
884 		return 1;
885 	if (task_is_stopped_or_traced(p))
886 		return 0;
887 	return task_curr(p) || !signal_pending(p);
888 }
889 
890 static void complete_signal(int sig, struct task_struct *p, int group)
891 {
892 	struct signal_struct *signal = p->signal;
893 	struct task_struct *t;
894 
895 	/*
896 	 * Now find a thread we can wake up to take the signal off the queue.
897 	 *
898 	 * If the main thread wants the signal, it gets first crack.
899 	 * Probably the least surprising to the average bear.
900 	 */
901 	if (wants_signal(sig, p))
902 		t = p;
903 	else if (!group || thread_group_empty(p))
904 		/*
905 		 * There is just one thread and it does not need to be woken.
906 		 * It will dequeue unblocked signals before it runs again.
907 		 */
908 		return;
909 	else {
910 		/*
911 		 * Otherwise try to find a suitable thread.
912 		 */
913 		t = signal->curr_target;
914 		while (!wants_signal(sig, t)) {
915 			t = next_thread(t);
916 			if (t == signal->curr_target)
917 				/*
918 				 * No thread needs to be woken.
919 				 * Any eligible threads will see
920 				 * the signal in the queue soon.
921 				 */
922 				return;
923 		}
924 		signal->curr_target = t;
925 	}
926 
927 	/*
928 	 * Found a killable thread.  If the signal will be fatal,
929 	 * then start taking the whole group down immediately.
930 	 */
931 	if (sig_fatal(p, sig) &&
932 	    !(signal->flags & (SIGNAL_UNKILLABLE | SIGNAL_GROUP_EXIT)) &&
933 	    !sigismember(&t->real_blocked, sig) &&
934 	    (sig == SIGKILL || !t->ptrace)) {
935 		/*
936 		 * This signal will be fatal to the whole group.
937 		 */
938 		if (!sig_kernel_coredump(sig)) {
939 			/*
940 			 * Start a group exit and wake everybody up.
941 			 * This way we don't have other threads
942 			 * running and doing things after a slower
943 			 * thread has the fatal signal pending.
944 			 */
945 			signal->flags = SIGNAL_GROUP_EXIT;
946 			signal->group_exit_code = sig;
947 			signal->group_stop_count = 0;
948 			t = p;
949 			do {
950 				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
951 				sigaddset(&t->pending.signal, SIGKILL);
952 				signal_wake_up(t, 1);
953 			} while_each_thread(p, t);
954 			return;
955 		}
956 	}
957 
958 	/*
959 	 * The signal is already in the shared-pending queue.
960 	 * Tell the chosen thread to wake up and dequeue it.
961 	 */
962 	signal_wake_up(t, sig == SIGKILL);
963 	return;
964 }
965 
966 static inline int legacy_queue(struct sigpending *signals, int sig)
967 {
968 	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
969 }
970 
971 #ifdef CONFIG_USER_NS
972 static inline void userns_fixup_signal_uid(struct siginfo *info, struct task_struct *t)
973 {
974 	if (current_user_ns() == task_cred_xxx(t, user_ns))
975 		return;
976 
977 	if (SI_FROMKERNEL(info))
978 		return;
979 
980 	rcu_read_lock();
981 	info->si_uid = from_kuid_munged(task_cred_xxx(t, user_ns),
982 					make_kuid(current_user_ns(), info->si_uid));
983 	rcu_read_unlock();
984 }
985 #else
986 static inline void userns_fixup_signal_uid(struct siginfo *info, struct task_struct *t)
987 {
988 	return;
989 }
990 #endif
991 
992 static int __send_signal(int sig, struct siginfo *info, struct task_struct *t,
993 			int group, int from_ancestor_ns)
994 {
995 	struct sigpending *pending;
996 	struct sigqueue *q;
997 	int override_rlimit;
998 	int ret = 0, result;
999 
1000 	assert_spin_locked(&t->sighand->siglock);
1001 
1002 	result = TRACE_SIGNAL_IGNORED;
1003 	if (!prepare_signal(sig, t,
1004 			from_ancestor_ns || (info == SEND_SIG_FORCED)))
1005 		goto ret;
1006 
1007 	pending = group ? &t->signal->shared_pending : &t->pending;
1008 	/*
1009 	 * Short-circuit ignored signals and support queuing
1010 	 * exactly one non-rt signal, so that we can get more
1011 	 * detailed information about the cause of the signal.
1012 	 */
1013 	result = TRACE_SIGNAL_ALREADY_PENDING;
1014 	if (legacy_queue(pending, sig))
1015 		goto ret;
1016 
1017 	result = TRACE_SIGNAL_DELIVERED;
1018 	/*
1019 	 * fast-pathed signals for kernel-internal things like SIGSTOP
1020 	 * or SIGKILL.
1021 	 */
1022 	if (info == SEND_SIG_FORCED)
1023 		goto out_set;
1024 
1025 	/*
1026 	 * Real-time signals must be queued if sent by sigqueue, or
1027 	 * some other real-time mechanism.  It is implementation
1028 	 * defined whether kill() does so.  We attempt to do so, on
1029 	 * the principle of least surprise, but since kill is not
1030 	 * allowed to fail with EAGAIN when low on memory we just
1031 	 * make sure at least one signal gets delivered and don't
1032 	 * pass on the info struct.
1033 	 */
1034 	if (sig < SIGRTMIN)
1035 		override_rlimit = (is_si_special(info) || info->si_code >= 0);
1036 	else
1037 		override_rlimit = 0;
1038 
1039 	q = __sigqueue_alloc(sig, t, GFP_ATOMIC | __GFP_NOTRACK_FALSE_POSITIVE,
1040 		override_rlimit);
1041 	if (q) {
1042 		list_add_tail(&q->list, &pending->list);
1043 		switch ((unsigned long) info) {
1044 		case (unsigned long) SEND_SIG_NOINFO:
1045 			q->info.si_signo = sig;
1046 			q->info.si_errno = 0;
1047 			q->info.si_code = SI_USER;
1048 			q->info.si_pid = task_tgid_nr_ns(current,
1049 							task_active_pid_ns(t));
1050 			q->info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1051 			break;
1052 		case (unsigned long) SEND_SIG_PRIV:
1053 			q->info.si_signo = sig;
1054 			q->info.si_errno = 0;
1055 			q->info.si_code = SI_KERNEL;
1056 			q->info.si_pid = 0;
1057 			q->info.si_uid = 0;
1058 			break;
1059 		default:
1060 			copy_siginfo(&q->info, info);
1061 			if (from_ancestor_ns)
1062 				q->info.si_pid = 0;
1063 			break;
1064 		}
1065 
1066 		userns_fixup_signal_uid(&q->info, t);
1067 
1068 	} else if (!is_si_special(info)) {
1069 		if (sig >= SIGRTMIN && info->si_code != SI_USER) {
1070 			/*
1071 			 * Queue overflow, abort.  We may abort if the
1072 			 * signal was rt and sent by user using something
1073 			 * other than kill().
1074 			 */
1075 			result = TRACE_SIGNAL_OVERFLOW_FAIL;
1076 			ret = -EAGAIN;
1077 			goto ret;
1078 		} else {
1079 			/*
1080 			 * This is a silent loss of information.  We still
1081 			 * send the signal, but the *info bits are lost.
1082 			 */
1083 			result = TRACE_SIGNAL_LOSE_INFO;
1084 		}
1085 	}
1086 
1087 out_set:
1088 	signalfd_notify(t, sig);
1089 	sigaddset(&pending->signal, sig);
1090 	complete_signal(sig, t, group);
1091 ret:
1092 	trace_signal_generate(sig, info, t, group, result);
1093 	return ret;
1094 }
1095 
1096 static int send_signal(int sig, struct siginfo *info, struct task_struct *t,
1097 			int group)
1098 {
1099 	int from_ancestor_ns = 0;
1100 
1101 #ifdef CONFIG_PID_NS
1102 	from_ancestor_ns = si_fromuser(info) &&
1103 			   !task_pid_nr_ns(current, task_active_pid_ns(t));
1104 #endif
1105 
1106 	return __send_signal(sig, info, t, group, from_ancestor_ns);
1107 }
1108 
1109 static void print_fatal_signal(int signr)
1110 {
1111 	struct pt_regs *regs = signal_pt_regs();
1112 	pr_info("potentially unexpected fatal signal %d.\n", signr);
1113 
1114 #if defined(__i386__) && !defined(__arch_um__)
1115 	pr_info("code at %08lx: ", regs->ip);
1116 	{
1117 		int i;
1118 		for (i = 0; i < 16; i++) {
1119 			unsigned char insn;
1120 
1121 			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1122 				break;
1123 			pr_cont("%02x ", insn);
1124 		}
1125 	}
1126 	pr_cont("\n");
1127 #endif
1128 	preempt_disable();
1129 	show_regs(regs);
1130 	preempt_enable();
1131 }
1132 
1133 static int __init setup_print_fatal_signals(char *str)
1134 {
1135 	get_option (&str, &print_fatal_signals);
1136 
1137 	return 1;
1138 }
1139 
1140 __setup("print-fatal-signals=", setup_print_fatal_signals);
1141 
1142 int
1143 __group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1144 {
1145 	return send_signal(sig, info, p, 1);
1146 }
1147 
1148 static int
1149 specific_send_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1150 {
1151 	return send_signal(sig, info, t, 0);
1152 }
1153 
1154 int do_send_sig_info(int sig, struct siginfo *info, struct task_struct *p,
1155 			bool group)
1156 {
1157 	unsigned long flags;
1158 	int ret = -ESRCH;
1159 
1160 	if (lock_task_sighand(p, &flags)) {
1161 		ret = send_signal(sig, info, p, group);
1162 		unlock_task_sighand(p, &flags);
1163 	}
1164 
1165 	return ret;
1166 }
1167 
1168 /*
1169  * Force a signal that the process can't ignore: if necessary
1170  * we unblock the signal and change any SIG_IGN to SIG_DFL.
1171  *
1172  * Note: If we unblock the signal, we always reset it to SIG_DFL,
1173  * since we do not want to have a signal handler that was blocked
1174  * be invoked when user space had explicitly blocked it.
1175  *
1176  * We don't want to have recursive SIGSEGV's etc, for example,
1177  * that is why we also clear SIGNAL_UNKILLABLE.
1178  */
1179 int
1180 force_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1181 {
1182 	unsigned long int flags;
1183 	int ret, blocked, ignored;
1184 	struct k_sigaction *action;
1185 
1186 	spin_lock_irqsave(&t->sighand->siglock, flags);
1187 	action = &t->sighand->action[sig-1];
1188 	ignored = action->sa.sa_handler == SIG_IGN;
1189 	blocked = sigismember(&t->blocked, sig);
1190 	if (blocked || ignored) {
1191 		action->sa.sa_handler = SIG_DFL;
1192 		if (blocked) {
1193 			sigdelset(&t->blocked, sig);
1194 			recalc_sigpending_and_wake(t);
1195 		}
1196 	}
1197 	/*
1198 	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1199 	 * debugging to leave init killable.
1200 	 */
1201 	if (action->sa.sa_handler == SIG_DFL && !t->ptrace)
1202 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1203 	ret = specific_send_sig_info(sig, info, t);
1204 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1205 
1206 	return ret;
1207 }
1208 
1209 /*
1210  * Nuke all other threads in the group.
1211  */
1212 int zap_other_threads(struct task_struct *p)
1213 {
1214 	struct task_struct *t = p;
1215 	int count = 0;
1216 
1217 	p->signal->group_stop_count = 0;
1218 
1219 	while_each_thread(p, t) {
1220 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1221 		count++;
1222 
1223 		/* Don't bother with already dead threads */
1224 		if (t->exit_state)
1225 			continue;
1226 		sigaddset(&t->pending.signal, SIGKILL);
1227 		signal_wake_up(t, 1);
1228 	}
1229 
1230 	return count;
1231 }
1232 
1233 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1234 					   unsigned long *flags)
1235 {
1236 	struct sighand_struct *sighand;
1237 
1238 	for (;;) {
1239 		/*
1240 		 * Disable interrupts early to avoid deadlocks.
1241 		 * See rcu_read_unlock() comment header for details.
1242 		 */
1243 		local_irq_save(*flags);
1244 		rcu_read_lock();
1245 		sighand = rcu_dereference(tsk->sighand);
1246 		if (unlikely(sighand == NULL)) {
1247 			rcu_read_unlock();
1248 			local_irq_restore(*flags);
1249 			break;
1250 		}
1251 		/*
1252 		 * This sighand can be already freed and even reused, but
1253 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1254 		 * initializes ->siglock: this slab can't go away, it has
1255 		 * the same object type, ->siglock can't be reinitialized.
1256 		 *
1257 		 * We need to ensure that tsk->sighand is still the same
1258 		 * after we take the lock, we can race with de_thread() or
1259 		 * __exit_signal(). In the latter case the next iteration
1260 		 * must see ->sighand == NULL.
1261 		 */
1262 		spin_lock(&sighand->siglock);
1263 		if (likely(sighand == tsk->sighand)) {
1264 			rcu_read_unlock();
1265 			break;
1266 		}
1267 		spin_unlock(&sighand->siglock);
1268 		rcu_read_unlock();
1269 		local_irq_restore(*flags);
1270 	}
1271 
1272 	return sighand;
1273 }
1274 
1275 /*
1276  * send signal info to all the members of a group
1277  */
1278 int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1279 {
1280 	int ret;
1281 
1282 	rcu_read_lock();
1283 	ret = check_kill_permission(sig, info, p);
1284 	rcu_read_unlock();
1285 
1286 	if (!ret && sig)
1287 		ret = do_send_sig_info(sig, info, p, true);
1288 
1289 	return ret;
1290 }
1291 
1292 /*
1293  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1294  * control characters do (^C, ^Z etc)
1295  * - the caller must hold at least a readlock on tasklist_lock
1296  */
1297 int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp)
1298 {
1299 	struct task_struct *p = NULL;
1300 	int retval, success;
1301 
1302 	success = 0;
1303 	retval = -ESRCH;
1304 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1305 		int err = group_send_sig_info(sig, info, p);
1306 		success |= !err;
1307 		retval = err;
1308 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1309 	return success ? 0 : retval;
1310 }
1311 
1312 int kill_pid_info(int sig, struct siginfo *info, struct pid *pid)
1313 {
1314 	int error = -ESRCH;
1315 	struct task_struct *p;
1316 
1317 	for (;;) {
1318 		rcu_read_lock();
1319 		p = pid_task(pid, PIDTYPE_PID);
1320 		if (p)
1321 			error = group_send_sig_info(sig, info, p);
1322 		rcu_read_unlock();
1323 		if (likely(!p || error != -ESRCH))
1324 			return error;
1325 
1326 		/*
1327 		 * The task was unhashed in between, try again.  If it
1328 		 * is dead, pid_task() will return NULL, if we race with
1329 		 * de_thread() it will find the new leader.
1330 		 */
1331 	}
1332 }
1333 
1334 static int kill_proc_info(int sig, struct siginfo *info, pid_t pid)
1335 {
1336 	int error;
1337 	rcu_read_lock();
1338 	error = kill_pid_info(sig, info, find_vpid(pid));
1339 	rcu_read_unlock();
1340 	return error;
1341 }
1342 
1343 static int kill_as_cred_perm(const struct cred *cred,
1344 			     struct task_struct *target)
1345 {
1346 	const struct cred *pcred = __task_cred(target);
1347 	if (!uid_eq(cred->euid, pcred->suid) && !uid_eq(cred->euid, pcred->uid) &&
1348 	    !uid_eq(cred->uid,  pcred->suid) && !uid_eq(cred->uid,  pcred->uid))
1349 		return 0;
1350 	return 1;
1351 }
1352 
1353 /* like kill_pid_info(), but doesn't use uid/euid of "current" */
1354 int kill_pid_info_as_cred(int sig, struct siginfo *info, struct pid *pid,
1355 			 const struct cred *cred, u32 secid)
1356 {
1357 	int ret = -EINVAL;
1358 	struct task_struct *p;
1359 	unsigned long flags;
1360 
1361 	if (!valid_signal(sig))
1362 		return ret;
1363 
1364 	rcu_read_lock();
1365 	p = pid_task(pid, PIDTYPE_PID);
1366 	if (!p) {
1367 		ret = -ESRCH;
1368 		goto out_unlock;
1369 	}
1370 	if (si_fromuser(info) && !kill_as_cred_perm(cred, p)) {
1371 		ret = -EPERM;
1372 		goto out_unlock;
1373 	}
1374 	ret = security_task_kill(p, info, sig, secid);
1375 	if (ret)
1376 		goto out_unlock;
1377 
1378 	if (sig) {
1379 		if (lock_task_sighand(p, &flags)) {
1380 			ret = __send_signal(sig, info, p, 1, 0);
1381 			unlock_task_sighand(p, &flags);
1382 		} else
1383 			ret = -ESRCH;
1384 	}
1385 out_unlock:
1386 	rcu_read_unlock();
1387 	return ret;
1388 }
1389 EXPORT_SYMBOL_GPL(kill_pid_info_as_cred);
1390 
1391 /*
1392  * kill_something_info() interprets pid in interesting ways just like kill(2).
1393  *
1394  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1395  * is probably wrong.  Should make it like BSD or SYSV.
1396  */
1397 
1398 static int kill_something_info(int sig, struct siginfo *info, pid_t pid)
1399 {
1400 	int ret;
1401 
1402 	if (pid > 0) {
1403 		rcu_read_lock();
1404 		ret = kill_pid_info(sig, info, find_vpid(pid));
1405 		rcu_read_unlock();
1406 		return ret;
1407 	}
1408 
1409 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1410 	if (pid == INT_MIN)
1411 		return -ESRCH;
1412 
1413 	read_lock(&tasklist_lock);
1414 	if (pid != -1) {
1415 		ret = __kill_pgrp_info(sig, info,
1416 				pid ? find_vpid(-pid) : task_pgrp(current));
1417 	} else {
1418 		int retval = 0, count = 0;
1419 		struct task_struct * p;
1420 
1421 		for_each_process(p) {
1422 			if (task_pid_vnr(p) > 1 &&
1423 					!same_thread_group(p, current)) {
1424 				int err = group_send_sig_info(sig, info, p);
1425 				++count;
1426 				if (err != -EPERM)
1427 					retval = err;
1428 			}
1429 		}
1430 		ret = count ? retval : -ESRCH;
1431 	}
1432 	read_unlock(&tasklist_lock);
1433 
1434 	return ret;
1435 }
1436 
1437 /*
1438  * These are for backward compatibility with the rest of the kernel source.
1439  */
1440 
1441 int send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1442 {
1443 	/*
1444 	 * Make sure legacy kernel users don't send in bad values
1445 	 * (normal paths check this in check_kill_permission).
1446 	 */
1447 	if (!valid_signal(sig))
1448 		return -EINVAL;
1449 
1450 	return do_send_sig_info(sig, info, p, false);
1451 }
1452 
1453 #define __si_special(priv) \
1454 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1455 
1456 int
1457 send_sig(int sig, struct task_struct *p, int priv)
1458 {
1459 	return send_sig_info(sig, __si_special(priv), p);
1460 }
1461 
1462 void
1463 force_sig(int sig, struct task_struct *p)
1464 {
1465 	force_sig_info(sig, SEND_SIG_PRIV, p);
1466 }
1467 
1468 /*
1469  * When things go south during signal handling, we
1470  * will force a SIGSEGV. And if the signal that caused
1471  * the problem was already a SIGSEGV, we'll want to
1472  * make sure we don't even try to deliver the signal..
1473  */
1474 int
1475 force_sigsegv(int sig, struct task_struct *p)
1476 {
1477 	if (sig == SIGSEGV) {
1478 		unsigned long flags;
1479 		spin_lock_irqsave(&p->sighand->siglock, flags);
1480 		p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1481 		spin_unlock_irqrestore(&p->sighand->siglock, flags);
1482 	}
1483 	force_sig(SIGSEGV, p);
1484 	return 0;
1485 }
1486 
1487 int kill_pgrp(struct pid *pid, int sig, int priv)
1488 {
1489 	int ret;
1490 
1491 	read_lock(&tasklist_lock);
1492 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1493 	read_unlock(&tasklist_lock);
1494 
1495 	return ret;
1496 }
1497 EXPORT_SYMBOL(kill_pgrp);
1498 
1499 int kill_pid(struct pid *pid, int sig, int priv)
1500 {
1501 	return kill_pid_info(sig, __si_special(priv), pid);
1502 }
1503 EXPORT_SYMBOL(kill_pid);
1504 
1505 /*
1506  * These functions support sending signals using preallocated sigqueue
1507  * structures.  This is needed "because realtime applications cannot
1508  * afford to lose notifications of asynchronous events, like timer
1509  * expirations or I/O completions".  In the case of POSIX Timers
1510  * we allocate the sigqueue structure from the timer_create.  If this
1511  * allocation fails we are able to report the failure to the application
1512  * with an EAGAIN error.
1513  */
1514 struct sigqueue *sigqueue_alloc(void)
1515 {
1516 	struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1517 
1518 	if (q)
1519 		q->flags |= SIGQUEUE_PREALLOC;
1520 
1521 	return q;
1522 }
1523 
1524 void sigqueue_free(struct sigqueue *q)
1525 {
1526 	unsigned long flags;
1527 	spinlock_t *lock = &current->sighand->siglock;
1528 
1529 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1530 	/*
1531 	 * We must hold ->siglock while testing q->list
1532 	 * to serialize with collect_signal() or with
1533 	 * __exit_signal()->flush_sigqueue().
1534 	 */
1535 	spin_lock_irqsave(lock, flags);
1536 	q->flags &= ~SIGQUEUE_PREALLOC;
1537 	/*
1538 	 * If it is queued it will be freed when dequeued,
1539 	 * like the "regular" sigqueue.
1540 	 */
1541 	if (!list_empty(&q->list))
1542 		q = NULL;
1543 	spin_unlock_irqrestore(lock, flags);
1544 
1545 	if (q)
1546 		__sigqueue_free(q);
1547 }
1548 
1549 int send_sigqueue(struct sigqueue *q, struct task_struct *t, int group)
1550 {
1551 	int sig = q->info.si_signo;
1552 	struct sigpending *pending;
1553 	unsigned long flags;
1554 	int ret, result;
1555 
1556 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1557 
1558 	ret = -1;
1559 	if (!likely(lock_task_sighand(t, &flags)))
1560 		goto ret;
1561 
1562 	ret = 1; /* the signal is ignored */
1563 	result = TRACE_SIGNAL_IGNORED;
1564 	if (!prepare_signal(sig, t, false))
1565 		goto out;
1566 
1567 	ret = 0;
1568 	if (unlikely(!list_empty(&q->list))) {
1569 		/*
1570 		 * If an SI_TIMER entry is already queue just increment
1571 		 * the overrun count.
1572 		 */
1573 		BUG_ON(q->info.si_code != SI_TIMER);
1574 		q->info.si_overrun++;
1575 		result = TRACE_SIGNAL_ALREADY_PENDING;
1576 		goto out;
1577 	}
1578 	q->info.si_overrun = 0;
1579 
1580 	signalfd_notify(t, sig);
1581 	pending = group ? &t->signal->shared_pending : &t->pending;
1582 	list_add_tail(&q->list, &pending->list);
1583 	sigaddset(&pending->signal, sig);
1584 	complete_signal(sig, t, group);
1585 	result = TRACE_SIGNAL_DELIVERED;
1586 out:
1587 	trace_signal_generate(sig, &q->info, t, group, result);
1588 	unlock_task_sighand(t, &flags);
1589 ret:
1590 	return ret;
1591 }
1592 
1593 /*
1594  * Let a parent know about the death of a child.
1595  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1596  *
1597  * Returns true if our parent ignored us and so we've switched to
1598  * self-reaping.
1599  */
1600 bool do_notify_parent(struct task_struct *tsk, int sig)
1601 {
1602 	struct siginfo info;
1603 	unsigned long flags;
1604 	struct sighand_struct *psig;
1605 	bool autoreap = false;
1606 	u64 utime, stime;
1607 
1608 	BUG_ON(sig == -1);
1609 
1610  	/* do_notify_parent_cldstop should have been called instead.  */
1611  	BUG_ON(task_is_stopped_or_traced(tsk));
1612 
1613 	BUG_ON(!tsk->ptrace &&
1614 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1615 
1616 	if (sig != SIGCHLD) {
1617 		/*
1618 		 * This is only possible if parent == real_parent.
1619 		 * Check if it has changed security domain.
1620 		 */
1621 		if (tsk->parent_exec_id != tsk->parent->self_exec_id)
1622 			sig = SIGCHLD;
1623 	}
1624 
1625 	info.si_signo = sig;
1626 	info.si_errno = 0;
1627 	/*
1628 	 * We are under tasklist_lock here so our parent is tied to
1629 	 * us and cannot change.
1630 	 *
1631 	 * task_active_pid_ns will always return the same pid namespace
1632 	 * until a task passes through release_task.
1633 	 *
1634 	 * write_lock() currently calls preempt_disable() which is the
1635 	 * same as rcu_read_lock(), but according to Oleg, this is not
1636 	 * correct to rely on this
1637 	 */
1638 	rcu_read_lock();
1639 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1640 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1641 				       task_uid(tsk));
1642 	rcu_read_unlock();
1643 
1644 	task_cputime(tsk, &utime, &stime);
1645 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
1646 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
1647 
1648 	info.si_status = tsk->exit_code & 0x7f;
1649 	if (tsk->exit_code & 0x80)
1650 		info.si_code = CLD_DUMPED;
1651 	else if (tsk->exit_code & 0x7f)
1652 		info.si_code = CLD_KILLED;
1653 	else {
1654 		info.si_code = CLD_EXITED;
1655 		info.si_status = tsk->exit_code >> 8;
1656 	}
1657 
1658 	psig = tsk->parent->sighand;
1659 	spin_lock_irqsave(&psig->siglock, flags);
1660 	if (!tsk->ptrace && sig == SIGCHLD &&
1661 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1662 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1663 		/*
1664 		 * We are exiting and our parent doesn't care.  POSIX.1
1665 		 * defines special semantics for setting SIGCHLD to SIG_IGN
1666 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
1667 		 * automatically and not left for our parent's wait4 call.
1668 		 * Rather than having the parent do it as a magic kind of
1669 		 * signal handler, we just set this to tell do_exit that we
1670 		 * can be cleaned up without becoming a zombie.  Note that
1671 		 * we still call __wake_up_parent in this case, because a
1672 		 * blocked sys_wait4 might now return -ECHILD.
1673 		 *
1674 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1675 		 * is implementation-defined: we do (if you don't want
1676 		 * it, just use SIG_IGN instead).
1677 		 */
1678 		autoreap = true;
1679 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
1680 			sig = 0;
1681 	}
1682 	if (valid_signal(sig) && sig)
1683 		__group_send_sig_info(sig, &info, tsk->parent);
1684 	__wake_up_parent(tsk, tsk->parent);
1685 	spin_unlock_irqrestore(&psig->siglock, flags);
1686 
1687 	return autoreap;
1688 }
1689 
1690 /**
1691  * do_notify_parent_cldstop - notify parent of stopped/continued state change
1692  * @tsk: task reporting the state change
1693  * @for_ptracer: the notification is for ptracer
1694  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
1695  *
1696  * Notify @tsk's parent that the stopped/continued state has changed.  If
1697  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
1698  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
1699  *
1700  * CONTEXT:
1701  * Must be called with tasklist_lock at least read locked.
1702  */
1703 static void do_notify_parent_cldstop(struct task_struct *tsk,
1704 				     bool for_ptracer, int why)
1705 {
1706 	struct siginfo info;
1707 	unsigned long flags;
1708 	struct task_struct *parent;
1709 	struct sighand_struct *sighand;
1710 	u64 utime, stime;
1711 
1712 	if (for_ptracer) {
1713 		parent = tsk->parent;
1714 	} else {
1715 		tsk = tsk->group_leader;
1716 		parent = tsk->real_parent;
1717 	}
1718 
1719 	info.si_signo = SIGCHLD;
1720 	info.si_errno = 0;
1721 	/*
1722 	 * see comment in do_notify_parent() about the following 4 lines
1723 	 */
1724 	rcu_read_lock();
1725 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
1726 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
1727 	rcu_read_unlock();
1728 
1729 	task_cputime(tsk, &utime, &stime);
1730 	info.si_utime = nsec_to_clock_t(utime);
1731 	info.si_stime = nsec_to_clock_t(stime);
1732 
1733  	info.si_code = why;
1734  	switch (why) {
1735  	case CLD_CONTINUED:
1736  		info.si_status = SIGCONT;
1737  		break;
1738  	case CLD_STOPPED:
1739  		info.si_status = tsk->signal->group_exit_code & 0x7f;
1740  		break;
1741  	case CLD_TRAPPED:
1742  		info.si_status = tsk->exit_code & 0x7f;
1743  		break;
1744  	default:
1745  		BUG();
1746  	}
1747 
1748 	sighand = parent->sighand;
1749 	spin_lock_irqsave(&sighand->siglock, flags);
1750 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
1751 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
1752 		__group_send_sig_info(SIGCHLD, &info, parent);
1753 	/*
1754 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
1755 	 */
1756 	__wake_up_parent(tsk, parent);
1757 	spin_unlock_irqrestore(&sighand->siglock, flags);
1758 }
1759 
1760 static inline int may_ptrace_stop(void)
1761 {
1762 	if (!likely(current->ptrace))
1763 		return 0;
1764 	/*
1765 	 * Are we in the middle of do_coredump?
1766 	 * If so and our tracer is also part of the coredump stopping
1767 	 * is a deadlock situation, and pointless because our tracer
1768 	 * is dead so don't allow us to stop.
1769 	 * If SIGKILL was already sent before the caller unlocked
1770 	 * ->siglock we must see ->core_state != NULL. Otherwise it
1771 	 * is safe to enter schedule().
1772 	 *
1773 	 * This is almost outdated, a task with the pending SIGKILL can't
1774 	 * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
1775 	 * after SIGKILL was already dequeued.
1776 	 */
1777 	if (unlikely(current->mm->core_state) &&
1778 	    unlikely(current->mm == current->parent->mm))
1779 		return 0;
1780 
1781 	return 1;
1782 }
1783 
1784 /*
1785  * Return non-zero if there is a SIGKILL that should be waking us up.
1786  * Called with the siglock held.
1787  */
1788 static int sigkill_pending(struct task_struct *tsk)
1789 {
1790 	return	sigismember(&tsk->pending.signal, SIGKILL) ||
1791 		sigismember(&tsk->signal->shared_pending.signal, SIGKILL);
1792 }
1793 
1794 /*
1795  * This must be called with current->sighand->siglock held.
1796  *
1797  * This should be the path for all ptrace stops.
1798  * We always set current->last_siginfo while stopped here.
1799  * That makes it a way to test a stopped process for
1800  * being ptrace-stopped vs being job-control-stopped.
1801  *
1802  * If we actually decide not to stop at all because the tracer
1803  * is gone, we keep current->exit_code unless clear_code.
1804  */
1805 static void ptrace_stop(int exit_code, int why, int clear_code, siginfo_t *info)
1806 	__releases(&current->sighand->siglock)
1807 	__acquires(&current->sighand->siglock)
1808 {
1809 	bool gstop_done = false;
1810 
1811 	if (arch_ptrace_stop_needed(exit_code, info)) {
1812 		/*
1813 		 * The arch code has something special to do before a
1814 		 * ptrace stop.  This is allowed to block, e.g. for faults
1815 		 * on user stack pages.  We can't keep the siglock while
1816 		 * calling arch_ptrace_stop, so we must release it now.
1817 		 * To preserve proper semantics, we must do this before
1818 		 * any signal bookkeeping like checking group_stop_count.
1819 		 * Meanwhile, a SIGKILL could come in before we retake the
1820 		 * siglock.  That must prevent us from sleeping in TASK_TRACED.
1821 		 * So after regaining the lock, we must check for SIGKILL.
1822 		 */
1823 		spin_unlock_irq(&current->sighand->siglock);
1824 		arch_ptrace_stop(exit_code, info);
1825 		spin_lock_irq(&current->sighand->siglock);
1826 		if (sigkill_pending(current))
1827 			return;
1828 	}
1829 
1830 	/*
1831 	 * We're committing to trapping.  TRACED should be visible before
1832 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
1833 	 * Also, transition to TRACED and updates to ->jobctl should be
1834 	 * atomic with respect to siglock and should be done after the arch
1835 	 * hook as siglock is released and regrabbed across it.
1836 	 */
1837 	set_current_state(TASK_TRACED);
1838 
1839 	current->last_siginfo = info;
1840 	current->exit_code = exit_code;
1841 
1842 	/*
1843 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
1844 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
1845 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
1846 	 * could be clear now.  We act as if SIGCONT is received after
1847 	 * TASK_TRACED is entered - ignore it.
1848 	 */
1849 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
1850 		gstop_done = task_participate_group_stop(current);
1851 
1852 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
1853 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
1854 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
1855 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
1856 
1857 	/* entering a trap, clear TRAPPING */
1858 	task_clear_jobctl_trapping(current);
1859 
1860 	spin_unlock_irq(&current->sighand->siglock);
1861 	read_lock(&tasklist_lock);
1862 	if (may_ptrace_stop()) {
1863 		/*
1864 		 * Notify parents of the stop.
1865 		 *
1866 		 * While ptraced, there are two parents - the ptracer and
1867 		 * the real_parent of the group_leader.  The ptracer should
1868 		 * know about every stop while the real parent is only
1869 		 * interested in the completion of group stop.  The states
1870 		 * for the two don't interact with each other.  Notify
1871 		 * separately unless they're gonna be duplicates.
1872 		 */
1873 		do_notify_parent_cldstop(current, true, why);
1874 		if (gstop_done && ptrace_reparented(current))
1875 			do_notify_parent_cldstop(current, false, why);
1876 
1877 		/*
1878 		 * Don't want to allow preemption here, because
1879 		 * sys_ptrace() needs this task to be inactive.
1880 		 *
1881 		 * XXX: implement read_unlock_no_resched().
1882 		 */
1883 		preempt_disable();
1884 		read_unlock(&tasklist_lock);
1885 		preempt_enable_no_resched();
1886 		freezable_schedule();
1887 	} else {
1888 		/*
1889 		 * By the time we got the lock, our tracer went away.
1890 		 * Don't drop the lock yet, another tracer may come.
1891 		 *
1892 		 * If @gstop_done, the ptracer went away between group stop
1893 		 * completion and here.  During detach, it would have set
1894 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
1895 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
1896 		 * the real parent of the group stop completion is enough.
1897 		 */
1898 		if (gstop_done)
1899 			do_notify_parent_cldstop(current, false, why);
1900 
1901 		/* tasklist protects us from ptrace_freeze_traced() */
1902 		__set_current_state(TASK_RUNNING);
1903 		if (clear_code)
1904 			current->exit_code = 0;
1905 		read_unlock(&tasklist_lock);
1906 	}
1907 
1908 	/*
1909 	 * We are back.  Now reacquire the siglock before touching
1910 	 * last_siginfo, so that we are sure to have synchronized with
1911 	 * any signal-sending on another CPU that wants to examine it.
1912 	 */
1913 	spin_lock_irq(&current->sighand->siglock);
1914 	current->last_siginfo = NULL;
1915 
1916 	/* LISTENING can be set only during STOP traps, clear it */
1917 	current->jobctl &= ~JOBCTL_LISTENING;
1918 
1919 	/*
1920 	 * Queued signals ignored us while we were stopped for tracing.
1921 	 * So check for any that we should take before resuming user mode.
1922 	 * This sets TIF_SIGPENDING, but never clears it.
1923 	 */
1924 	recalc_sigpending_tsk(current);
1925 }
1926 
1927 static void ptrace_do_notify(int signr, int exit_code, int why)
1928 {
1929 	siginfo_t info;
1930 
1931 	memset(&info, 0, sizeof info);
1932 	info.si_signo = signr;
1933 	info.si_code = exit_code;
1934 	info.si_pid = task_pid_vnr(current);
1935 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1936 
1937 	/* Let the debugger run.  */
1938 	ptrace_stop(exit_code, why, 1, &info);
1939 }
1940 
1941 void ptrace_notify(int exit_code)
1942 {
1943 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
1944 	if (unlikely(current->task_works))
1945 		task_work_run();
1946 
1947 	spin_lock_irq(&current->sighand->siglock);
1948 	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
1949 	spin_unlock_irq(&current->sighand->siglock);
1950 }
1951 
1952 /**
1953  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
1954  * @signr: signr causing group stop if initiating
1955  *
1956  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
1957  * and participate in it.  If already set, participate in the existing
1958  * group stop.  If participated in a group stop (and thus slept), %true is
1959  * returned with siglock released.
1960  *
1961  * If ptraced, this function doesn't handle stop itself.  Instead,
1962  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
1963  * untouched.  The caller must ensure that INTERRUPT trap handling takes
1964  * places afterwards.
1965  *
1966  * CONTEXT:
1967  * Must be called with @current->sighand->siglock held, which is released
1968  * on %true return.
1969  *
1970  * RETURNS:
1971  * %false if group stop is already cancelled or ptrace trap is scheduled.
1972  * %true if participated in group stop.
1973  */
1974 static bool do_signal_stop(int signr)
1975 	__releases(&current->sighand->siglock)
1976 {
1977 	struct signal_struct *sig = current->signal;
1978 
1979 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
1980 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
1981 		struct task_struct *t;
1982 
1983 		/* signr will be recorded in task->jobctl for retries */
1984 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
1985 
1986 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
1987 		    unlikely(signal_group_exit(sig)))
1988 			return false;
1989 		/*
1990 		 * There is no group stop already in progress.  We must
1991 		 * initiate one now.
1992 		 *
1993 		 * While ptraced, a task may be resumed while group stop is
1994 		 * still in effect and then receive a stop signal and
1995 		 * initiate another group stop.  This deviates from the
1996 		 * usual behavior as two consecutive stop signals can't
1997 		 * cause two group stops when !ptraced.  That is why we
1998 		 * also check !task_is_stopped(t) below.
1999 		 *
2000 		 * The condition can be distinguished by testing whether
2001 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2002 		 * group_exit_code in such case.
2003 		 *
2004 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2005 		 * an intervening stop signal is required to cause two
2006 		 * continued events regardless of ptrace.
2007 		 */
2008 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2009 			sig->group_exit_code = signr;
2010 
2011 		sig->group_stop_count = 0;
2012 
2013 		if (task_set_jobctl_pending(current, signr | gstop))
2014 			sig->group_stop_count++;
2015 
2016 		t = current;
2017 		while_each_thread(current, t) {
2018 			/*
2019 			 * Setting state to TASK_STOPPED for a group
2020 			 * stop is always done with the siglock held,
2021 			 * so this check has no races.
2022 			 */
2023 			if (!task_is_stopped(t) &&
2024 			    task_set_jobctl_pending(t, signr | gstop)) {
2025 				sig->group_stop_count++;
2026 				if (likely(!(t->ptrace & PT_SEIZED)))
2027 					signal_wake_up(t, 0);
2028 				else
2029 					ptrace_trap_notify(t);
2030 			}
2031 		}
2032 	}
2033 
2034 	if (likely(!current->ptrace)) {
2035 		int notify = 0;
2036 
2037 		/*
2038 		 * If there are no other threads in the group, or if there
2039 		 * is a group stop in progress and we are the last to stop,
2040 		 * report to the parent.
2041 		 */
2042 		if (task_participate_group_stop(current))
2043 			notify = CLD_STOPPED;
2044 
2045 		__set_current_state(TASK_STOPPED);
2046 		spin_unlock_irq(&current->sighand->siglock);
2047 
2048 		/*
2049 		 * Notify the parent of the group stop completion.  Because
2050 		 * we're not holding either the siglock or tasklist_lock
2051 		 * here, ptracer may attach inbetween; however, this is for
2052 		 * group stop and should always be delivered to the real
2053 		 * parent of the group leader.  The new ptracer will get
2054 		 * its notification when this task transitions into
2055 		 * TASK_TRACED.
2056 		 */
2057 		if (notify) {
2058 			read_lock(&tasklist_lock);
2059 			do_notify_parent_cldstop(current, false, notify);
2060 			read_unlock(&tasklist_lock);
2061 		}
2062 
2063 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2064 		freezable_schedule();
2065 		return true;
2066 	} else {
2067 		/*
2068 		 * While ptraced, group stop is handled by STOP trap.
2069 		 * Schedule it and let the caller deal with it.
2070 		 */
2071 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2072 		return false;
2073 	}
2074 }
2075 
2076 /**
2077  * do_jobctl_trap - take care of ptrace jobctl traps
2078  *
2079  * When PT_SEIZED, it's used for both group stop and explicit
2080  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2081  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2082  * the stop signal; otherwise, %SIGTRAP.
2083  *
2084  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2085  * number as exit_code and no siginfo.
2086  *
2087  * CONTEXT:
2088  * Must be called with @current->sighand->siglock held, which may be
2089  * released and re-acquired before returning with intervening sleep.
2090  */
2091 static void do_jobctl_trap(void)
2092 {
2093 	struct signal_struct *signal = current->signal;
2094 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2095 
2096 	if (current->ptrace & PT_SEIZED) {
2097 		if (!signal->group_stop_count &&
2098 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2099 			signr = SIGTRAP;
2100 		WARN_ON_ONCE(!signr);
2101 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2102 				 CLD_STOPPED);
2103 	} else {
2104 		WARN_ON_ONCE(!signr);
2105 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2106 		current->exit_code = 0;
2107 	}
2108 }
2109 
2110 static int ptrace_signal(int signr, siginfo_t *info)
2111 {
2112 	/*
2113 	 * We do not check sig_kernel_stop(signr) but set this marker
2114 	 * unconditionally because we do not know whether debugger will
2115 	 * change signr. This flag has no meaning unless we are going
2116 	 * to stop after return from ptrace_stop(). In this case it will
2117 	 * be checked in do_signal_stop(), we should only stop if it was
2118 	 * not cleared by SIGCONT while we were sleeping. See also the
2119 	 * comment in dequeue_signal().
2120 	 */
2121 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2122 	ptrace_stop(signr, CLD_TRAPPED, 0, info);
2123 
2124 	/* We're back.  Did the debugger cancel the sig?  */
2125 	signr = current->exit_code;
2126 	if (signr == 0)
2127 		return signr;
2128 
2129 	current->exit_code = 0;
2130 
2131 	/*
2132 	 * Update the siginfo structure if the signal has
2133 	 * changed.  If the debugger wanted something
2134 	 * specific in the siginfo structure then it should
2135 	 * have updated *info via PTRACE_SETSIGINFO.
2136 	 */
2137 	if (signr != info->si_signo) {
2138 		info->si_signo = signr;
2139 		info->si_errno = 0;
2140 		info->si_code = SI_USER;
2141 		rcu_read_lock();
2142 		info->si_pid = task_pid_vnr(current->parent);
2143 		info->si_uid = from_kuid_munged(current_user_ns(),
2144 						task_uid(current->parent));
2145 		rcu_read_unlock();
2146 	}
2147 
2148 	/* If the (new) signal is now blocked, requeue it.  */
2149 	if (sigismember(&current->blocked, signr)) {
2150 		specific_send_sig_info(signr, info, current);
2151 		signr = 0;
2152 	}
2153 
2154 	return signr;
2155 }
2156 
2157 int get_signal(struct ksignal *ksig)
2158 {
2159 	struct sighand_struct *sighand = current->sighand;
2160 	struct signal_struct *signal = current->signal;
2161 	int signr;
2162 
2163 	if (unlikely(current->task_works))
2164 		task_work_run();
2165 
2166 	if (unlikely(uprobe_deny_signal()))
2167 		return 0;
2168 
2169 	/*
2170 	 * Do this once, we can't return to user-mode if freezing() == T.
2171 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2172 	 * thus do not need another check after return.
2173 	 */
2174 	try_to_freeze();
2175 
2176 relock:
2177 	spin_lock_irq(&sighand->siglock);
2178 	/*
2179 	 * Every stopped thread goes here after wakeup. Check to see if
2180 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2181 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2182 	 */
2183 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2184 		int why;
2185 
2186 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2187 			why = CLD_CONTINUED;
2188 		else
2189 			why = CLD_STOPPED;
2190 
2191 		signal->flags &= ~SIGNAL_CLD_MASK;
2192 
2193 		spin_unlock_irq(&sighand->siglock);
2194 
2195 		/*
2196 		 * Notify the parent that we're continuing.  This event is
2197 		 * always per-process and doesn't make whole lot of sense
2198 		 * for ptracers, who shouldn't consume the state via
2199 		 * wait(2) either, but, for backward compatibility, notify
2200 		 * the ptracer of the group leader too unless it's gonna be
2201 		 * a duplicate.
2202 		 */
2203 		read_lock(&tasklist_lock);
2204 		do_notify_parent_cldstop(current, false, why);
2205 
2206 		if (ptrace_reparented(current->group_leader))
2207 			do_notify_parent_cldstop(current->group_leader,
2208 						true, why);
2209 		read_unlock(&tasklist_lock);
2210 
2211 		goto relock;
2212 	}
2213 
2214 	for (;;) {
2215 		struct k_sigaction *ka;
2216 
2217 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2218 		    do_signal_stop(0))
2219 			goto relock;
2220 
2221 		if (unlikely(current->jobctl & JOBCTL_TRAP_MASK)) {
2222 			do_jobctl_trap();
2223 			spin_unlock_irq(&sighand->siglock);
2224 			goto relock;
2225 		}
2226 
2227 		signr = dequeue_signal(current, &current->blocked, &ksig->info);
2228 
2229 		if (!signr)
2230 			break; /* will return 0 */
2231 
2232 		if (unlikely(current->ptrace) && signr != SIGKILL) {
2233 			signr = ptrace_signal(signr, &ksig->info);
2234 			if (!signr)
2235 				continue;
2236 		}
2237 
2238 		ka = &sighand->action[signr-1];
2239 
2240 		/* Trace actually delivered signals. */
2241 		trace_signal_deliver(signr, &ksig->info, ka);
2242 
2243 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2244 			continue;
2245 		if (ka->sa.sa_handler != SIG_DFL) {
2246 			/* Run the handler.  */
2247 			ksig->ka = *ka;
2248 
2249 			if (ka->sa.sa_flags & SA_ONESHOT)
2250 				ka->sa.sa_handler = SIG_DFL;
2251 
2252 			break; /* will return non-zero "signr" value */
2253 		}
2254 
2255 		/*
2256 		 * Now we are doing the default action for this signal.
2257 		 */
2258 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2259 			continue;
2260 
2261 		/*
2262 		 * Global init gets no signals it doesn't want.
2263 		 * Container-init gets no signals it doesn't want from same
2264 		 * container.
2265 		 *
2266 		 * Note that if global/container-init sees a sig_kernel_only()
2267 		 * signal here, the signal must have been generated internally
2268 		 * or must have come from an ancestor namespace. In either
2269 		 * case, the signal cannot be dropped.
2270 		 */
2271 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2272 				!sig_kernel_only(signr))
2273 			continue;
2274 
2275 		if (sig_kernel_stop(signr)) {
2276 			/*
2277 			 * The default action is to stop all threads in
2278 			 * the thread group.  The job control signals
2279 			 * do nothing in an orphaned pgrp, but SIGSTOP
2280 			 * always works.  Note that siglock needs to be
2281 			 * dropped during the call to is_orphaned_pgrp()
2282 			 * because of lock ordering with tasklist_lock.
2283 			 * This allows an intervening SIGCONT to be posted.
2284 			 * We need to check for that and bail out if necessary.
2285 			 */
2286 			if (signr != SIGSTOP) {
2287 				spin_unlock_irq(&sighand->siglock);
2288 
2289 				/* signals can be posted during this window */
2290 
2291 				if (is_current_pgrp_orphaned())
2292 					goto relock;
2293 
2294 				spin_lock_irq(&sighand->siglock);
2295 			}
2296 
2297 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2298 				/* It released the siglock.  */
2299 				goto relock;
2300 			}
2301 
2302 			/*
2303 			 * We didn't actually stop, due to a race
2304 			 * with SIGCONT or something like that.
2305 			 */
2306 			continue;
2307 		}
2308 
2309 		spin_unlock_irq(&sighand->siglock);
2310 
2311 		/*
2312 		 * Anything else is fatal, maybe with a core dump.
2313 		 */
2314 		current->flags |= PF_SIGNALED;
2315 
2316 		if (sig_kernel_coredump(signr)) {
2317 			if (print_fatal_signals)
2318 				print_fatal_signal(ksig->info.si_signo);
2319 			proc_coredump_connector(current);
2320 			/*
2321 			 * If it was able to dump core, this kills all
2322 			 * other threads in the group and synchronizes with
2323 			 * their demise.  If we lost the race with another
2324 			 * thread getting here, it set group_exit_code
2325 			 * first and our do_group_exit call below will use
2326 			 * that value and ignore the one we pass it.
2327 			 */
2328 			do_coredump(&ksig->info);
2329 		}
2330 
2331 		/*
2332 		 * Death signals, no core dump.
2333 		 */
2334 		do_group_exit(ksig->info.si_signo);
2335 		/* NOTREACHED */
2336 	}
2337 	spin_unlock_irq(&sighand->siglock);
2338 
2339 	ksig->sig = signr;
2340 	return ksig->sig > 0;
2341 }
2342 
2343 /**
2344  * signal_delivered -
2345  * @ksig:		kernel signal struct
2346  * @stepping:		nonzero if debugger single-step or block-step in use
2347  *
2348  * This function should be called when a signal has successfully been
2349  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2350  * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2351  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2352  */
2353 static void signal_delivered(struct ksignal *ksig, int stepping)
2354 {
2355 	sigset_t blocked;
2356 
2357 	/* A signal was successfully delivered, and the
2358 	   saved sigmask was stored on the signal frame,
2359 	   and will be restored by sigreturn.  So we can
2360 	   simply clear the restore sigmask flag.  */
2361 	clear_restore_sigmask();
2362 
2363 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2364 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2365 		sigaddset(&blocked, ksig->sig);
2366 	set_current_blocked(&blocked);
2367 	tracehook_signal_handler(stepping);
2368 }
2369 
2370 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2371 {
2372 	if (failed)
2373 		force_sigsegv(ksig->sig, current);
2374 	else
2375 		signal_delivered(ksig, stepping);
2376 }
2377 
2378 /*
2379  * It could be that complete_signal() picked us to notify about the
2380  * group-wide signal. Other threads should be notified now to take
2381  * the shared signals in @which since we will not.
2382  */
2383 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2384 {
2385 	sigset_t retarget;
2386 	struct task_struct *t;
2387 
2388 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2389 	if (sigisemptyset(&retarget))
2390 		return;
2391 
2392 	t = tsk;
2393 	while_each_thread(tsk, t) {
2394 		if (t->flags & PF_EXITING)
2395 			continue;
2396 
2397 		if (!has_pending_signals(&retarget, &t->blocked))
2398 			continue;
2399 		/* Remove the signals this thread can handle. */
2400 		sigandsets(&retarget, &retarget, &t->blocked);
2401 
2402 		if (!signal_pending(t))
2403 			signal_wake_up(t, 0);
2404 
2405 		if (sigisemptyset(&retarget))
2406 			break;
2407 	}
2408 }
2409 
2410 void exit_signals(struct task_struct *tsk)
2411 {
2412 	int group_stop = 0;
2413 	sigset_t unblocked;
2414 
2415 	/*
2416 	 * @tsk is about to have PF_EXITING set - lock out users which
2417 	 * expect stable threadgroup.
2418 	 */
2419 	cgroup_threadgroup_change_begin(tsk);
2420 
2421 	if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2422 		tsk->flags |= PF_EXITING;
2423 		cgroup_threadgroup_change_end(tsk);
2424 		return;
2425 	}
2426 
2427 	spin_lock_irq(&tsk->sighand->siglock);
2428 	/*
2429 	 * From now this task is not visible for group-wide signals,
2430 	 * see wants_signal(), do_signal_stop().
2431 	 */
2432 	tsk->flags |= PF_EXITING;
2433 
2434 	cgroup_threadgroup_change_end(tsk);
2435 
2436 	if (!signal_pending(tsk))
2437 		goto out;
2438 
2439 	unblocked = tsk->blocked;
2440 	signotset(&unblocked);
2441 	retarget_shared_pending(tsk, &unblocked);
2442 
2443 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2444 	    task_participate_group_stop(tsk))
2445 		group_stop = CLD_STOPPED;
2446 out:
2447 	spin_unlock_irq(&tsk->sighand->siglock);
2448 
2449 	/*
2450 	 * If group stop has completed, deliver the notification.  This
2451 	 * should always go to the real parent of the group leader.
2452 	 */
2453 	if (unlikely(group_stop)) {
2454 		read_lock(&tasklist_lock);
2455 		do_notify_parent_cldstop(tsk, false, group_stop);
2456 		read_unlock(&tasklist_lock);
2457 	}
2458 }
2459 
2460 EXPORT_SYMBOL(recalc_sigpending);
2461 EXPORT_SYMBOL_GPL(dequeue_signal);
2462 EXPORT_SYMBOL(flush_signals);
2463 EXPORT_SYMBOL(force_sig);
2464 EXPORT_SYMBOL(send_sig);
2465 EXPORT_SYMBOL(send_sig_info);
2466 EXPORT_SYMBOL(sigprocmask);
2467 
2468 /*
2469  * System call entry points.
2470  */
2471 
2472 /**
2473  *  sys_restart_syscall - restart a system call
2474  */
2475 SYSCALL_DEFINE0(restart_syscall)
2476 {
2477 	struct restart_block *restart = &current->restart_block;
2478 	return restart->fn(restart);
2479 }
2480 
2481 long do_no_restart_syscall(struct restart_block *param)
2482 {
2483 	return -EINTR;
2484 }
2485 
2486 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2487 {
2488 	if (signal_pending(tsk) && !thread_group_empty(tsk)) {
2489 		sigset_t newblocked;
2490 		/* A set of now blocked but previously unblocked signals. */
2491 		sigandnsets(&newblocked, newset, &current->blocked);
2492 		retarget_shared_pending(tsk, &newblocked);
2493 	}
2494 	tsk->blocked = *newset;
2495 	recalc_sigpending();
2496 }
2497 
2498 /**
2499  * set_current_blocked - change current->blocked mask
2500  * @newset: new mask
2501  *
2502  * It is wrong to change ->blocked directly, this helper should be used
2503  * to ensure the process can't miss a shared signal we are going to block.
2504  */
2505 void set_current_blocked(sigset_t *newset)
2506 {
2507 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2508 	__set_current_blocked(newset);
2509 }
2510 
2511 void __set_current_blocked(const sigset_t *newset)
2512 {
2513 	struct task_struct *tsk = current;
2514 
2515 	/*
2516 	 * In case the signal mask hasn't changed, there is nothing we need
2517 	 * to do. The current->blocked shouldn't be modified by other task.
2518 	 */
2519 	if (sigequalsets(&tsk->blocked, newset))
2520 		return;
2521 
2522 	spin_lock_irq(&tsk->sighand->siglock);
2523 	__set_task_blocked(tsk, newset);
2524 	spin_unlock_irq(&tsk->sighand->siglock);
2525 }
2526 
2527 /*
2528  * This is also useful for kernel threads that want to temporarily
2529  * (or permanently) block certain signals.
2530  *
2531  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2532  * interface happily blocks "unblockable" signals like SIGKILL
2533  * and friends.
2534  */
2535 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2536 {
2537 	struct task_struct *tsk = current;
2538 	sigset_t newset;
2539 
2540 	/* Lockless, only current can change ->blocked, never from irq */
2541 	if (oldset)
2542 		*oldset = tsk->blocked;
2543 
2544 	switch (how) {
2545 	case SIG_BLOCK:
2546 		sigorsets(&newset, &tsk->blocked, set);
2547 		break;
2548 	case SIG_UNBLOCK:
2549 		sigandnsets(&newset, &tsk->blocked, set);
2550 		break;
2551 	case SIG_SETMASK:
2552 		newset = *set;
2553 		break;
2554 	default:
2555 		return -EINVAL;
2556 	}
2557 
2558 	__set_current_blocked(&newset);
2559 	return 0;
2560 }
2561 
2562 /**
2563  *  sys_rt_sigprocmask - change the list of currently blocked signals
2564  *  @how: whether to add, remove, or set signals
2565  *  @nset: stores pending signals
2566  *  @oset: previous value of signal mask if non-null
2567  *  @sigsetsize: size of sigset_t type
2568  */
2569 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
2570 		sigset_t __user *, oset, size_t, sigsetsize)
2571 {
2572 	sigset_t old_set, new_set;
2573 	int error;
2574 
2575 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2576 	if (sigsetsize != sizeof(sigset_t))
2577 		return -EINVAL;
2578 
2579 	old_set = current->blocked;
2580 
2581 	if (nset) {
2582 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
2583 			return -EFAULT;
2584 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2585 
2586 		error = sigprocmask(how, &new_set, NULL);
2587 		if (error)
2588 			return error;
2589 	}
2590 
2591 	if (oset) {
2592 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
2593 			return -EFAULT;
2594 	}
2595 
2596 	return 0;
2597 }
2598 
2599 #ifdef CONFIG_COMPAT
2600 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
2601 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
2602 {
2603 #ifdef __BIG_ENDIAN
2604 	sigset_t old_set = current->blocked;
2605 
2606 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2607 	if (sigsetsize != sizeof(sigset_t))
2608 		return -EINVAL;
2609 
2610 	if (nset) {
2611 		compat_sigset_t new32;
2612 		sigset_t new_set;
2613 		int error;
2614 		if (copy_from_user(&new32, nset, sizeof(compat_sigset_t)))
2615 			return -EFAULT;
2616 
2617 		sigset_from_compat(&new_set, &new32);
2618 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2619 
2620 		error = sigprocmask(how, &new_set, NULL);
2621 		if (error)
2622 			return error;
2623 	}
2624 	if (oset) {
2625 		compat_sigset_t old32;
2626 		sigset_to_compat(&old32, &old_set);
2627 		if (copy_to_user(oset, &old32, sizeof(compat_sigset_t)))
2628 			return -EFAULT;
2629 	}
2630 	return 0;
2631 #else
2632 	return sys_rt_sigprocmask(how, (sigset_t __user *)nset,
2633 				  (sigset_t __user *)oset, sigsetsize);
2634 #endif
2635 }
2636 #endif
2637 
2638 static int do_sigpending(void *set, unsigned long sigsetsize)
2639 {
2640 	if (sigsetsize > sizeof(sigset_t))
2641 		return -EINVAL;
2642 
2643 	spin_lock_irq(&current->sighand->siglock);
2644 	sigorsets(set, &current->pending.signal,
2645 		  &current->signal->shared_pending.signal);
2646 	spin_unlock_irq(&current->sighand->siglock);
2647 
2648 	/* Outside the lock because only this thread touches it.  */
2649 	sigandsets(set, &current->blocked, set);
2650 	return 0;
2651 }
2652 
2653 /**
2654  *  sys_rt_sigpending - examine a pending signal that has been raised
2655  *			while blocked
2656  *  @uset: stores pending signals
2657  *  @sigsetsize: size of sigset_t type or larger
2658  */
2659 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
2660 {
2661 	sigset_t set;
2662 	int err = do_sigpending(&set, sigsetsize);
2663 	if (!err && copy_to_user(uset, &set, sigsetsize))
2664 		err = -EFAULT;
2665 	return err;
2666 }
2667 
2668 #ifdef CONFIG_COMPAT
2669 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
2670 		compat_size_t, sigsetsize)
2671 {
2672 #ifdef __BIG_ENDIAN
2673 	sigset_t set;
2674 	int err = do_sigpending(&set, sigsetsize);
2675 	if (!err) {
2676 		compat_sigset_t set32;
2677 		sigset_to_compat(&set32, &set);
2678 		/* we can get here only if sigsetsize <= sizeof(set) */
2679 		if (copy_to_user(uset, &set32, sigsetsize))
2680 			err = -EFAULT;
2681 	}
2682 	return err;
2683 #else
2684 	return sys_rt_sigpending((sigset_t __user *)uset, sigsetsize);
2685 #endif
2686 }
2687 #endif
2688 
2689 enum siginfo_layout siginfo_layout(int sig, int si_code)
2690 {
2691 	enum siginfo_layout layout = SIL_KILL;
2692 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
2693 		static const struct {
2694 			unsigned char limit, layout;
2695 		} filter[] = {
2696 			[SIGILL]  = { NSIGILL,  SIL_FAULT },
2697 			[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
2698 			[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
2699 			[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
2700 			[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
2701 #if defined(SIGMET) && defined(NSIGEMT)
2702 			[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
2703 #endif
2704 			[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
2705 			[SIGPOLL] = { NSIGPOLL, SIL_POLL },
2706 #ifdef __ARCH_SIGSYS
2707 			[SIGSYS]  = { NSIGSYS,  SIL_SYS },
2708 #endif
2709 		};
2710 		if ((sig < ARRAY_SIZE(filter)) && (si_code <= filter[sig].limit))
2711 			layout = filter[sig].layout;
2712 		else if (si_code <= NSIGPOLL)
2713 			layout = SIL_POLL;
2714 	} else {
2715 		if (si_code == SI_TIMER)
2716 			layout = SIL_TIMER;
2717 		else if (si_code == SI_SIGIO)
2718 			layout = SIL_POLL;
2719 		else if (si_code < 0)
2720 			layout = SIL_RT;
2721 		/* Tests to support buggy kernel ABIs */
2722 #ifdef TRAP_FIXME
2723 		if ((sig == SIGTRAP) && (si_code == TRAP_FIXME))
2724 			layout = SIL_FAULT;
2725 #endif
2726 #ifdef FPE_FIXME
2727 		if ((sig == SIGFPE) && (si_code == FPE_FIXME))
2728 			layout = SIL_FAULT;
2729 #endif
2730 	}
2731 	return layout;
2732 }
2733 
2734 #ifndef HAVE_ARCH_COPY_SIGINFO_TO_USER
2735 
2736 int copy_siginfo_to_user(siginfo_t __user *to, const siginfo_t *from)
2737 {
2738 	int err;
2739 
2740 	if (!access_ok (VERIFY_WRITE, to, sizeof(siginfo_t)))
2741 		return -EFAULT;
2742 	if (from->si_code < 0)
2743 		return __copy_to_user(to, from, sizeof(siginfo_t))
2744 			? -EFAULT : 0;
2745 	/*
2746 	 * If you change siginfo_t structure, please be sure
2747 	 * this code is fixed accordingly.
2748 	 * Please remember to update the signalfd_copyinfo() function
2749 	 * inside fs/signalfd.c too, in case siginfo_t changes.
2750 	 * It should never copy any pad contained in the structure
2751 	 * to avoid security leaks, but must copy the generic
2752 	 * 3 ints plus the relevant union member.
2753 	 */
2754 	err = __put_user(from->si_signo, &to->si_signo);
2755 	err |= __put_user(from->si_errno, &to->si_errno);
2756 	err |= __put_user(from->si_code, &to->si_code);
2757 	switch (siginfo_layout(from->si_signo, from->si_code)) {
2758 	case SIL_KILL:
2759 		err |= __put_user(from->si_pid, &to->si_pid);
2760 		err |= __put_user(from->si_uid, &to->si_uid);
2761 		break;
2762 	case SIL_TIMER:
2763 		/* Unreached SI_TIMER is negative */
2764 		break;
2765 	case SIL_POLL:
2766 		err |= __put_user(from->si_band, &to->si_band);
2767 		err |= __put_user(from->si_fd, &to->si_fd);
2768 		break;
2769 	case SIL_FAULT:
2770 		err |= __put_user(from->si_addr, &to->si_addr);
2771 #ifdef __ARCH_SI_TRAPNO
2772 		err |= __put_user(from->si_trapno, &to->si_trapno);
2773 #endif
2774 #ifdef BUS_MCEERR_AO
2775 		/*
2776 		 * Other callers might not initialize the si_lsb field,
2777 		 * so check explicitly for the right codes here.
2778 		 */
2779 		if (from->si_signo == SIGBUS &&
2780 		    (from->si_code == BUS_MCEERR_AR || from->si_code == BUS_MCEERR_AO))
2781 			err |= __put_user(from->si_addr_lsb, &to->si_addr_lsb);
2782 #endif
2783 #ifdef SEGV_BNDERR
2784 		if (from->si_signo == SIGSEGV && from->si_code == SEGV_BNDERR) {
2785 			err |= __put_user(from->si_lower, &to->si_lower);
2786 			err |= __put_user(from->si_upper, &to->si_upper);
2787 		}
2788 #endif
2789 #ifdef SEGV_PKUERR
2790 		if (from->si_signo == SIGSEGV && from->si_code == SEGV_PKUERR)
2791 			err |= __put_user(from->si_pkey, &to->si_pkey);
2792 #endif
2793 		break;
2794 	case SIL_CHLD:
2795 		err |= __put_user(from->si_pid, &to->si_pid);
2796 		err |= __put_user(from->si_uid, &to->si_uid);
2797 		err |= __put_user(from->si_status, &to->si_status);
2798 		err |= __put_user(from->si_utime, &to->si_utime);
2799 		err |= __put_user(from->si_stime, &to->si_stime);
2800 		break;
2801 	case SIL_RT:
2802 		err |= __put_user(from->si_pid, &to->si_pid);
2803 		err |= __put_user(from->si_uid, &to->si_uid);
2804 		err |= __put_user(from->si_ptr, &to->si_ptr);
2805 		break;
2806 #ifdef __ARCH_SIGSYS
2807 	case SIL_SYS:
2808 		err |= __put_user(from->si_call_addr, &to->si_call_addr);
2809 		err |= __put_user(from->si_syscall, &to->si_syscall);
2810 		err |= __put_user(from->si_arch, &to->si_arch);
2811 		break;
2812 #endif
2813 	}
2814 	return err;
2815 }
2816 
2817 #endif
2818 
2819 /**
2820  *  do_sigtimedwait - wait for queued signals specified in @which
2821  *  @which: queued signals to wait for
2822  *  @info: if non-null, the signal's siginfo is returned here
2823  *  @ts: upper bound on process time suspension
2824  */
2825 static int do_sigtimedwait(const sigset_t *which, siginfo_t *info,
2826 		    const struct timespec *ts)
2827 {
2828 	ktime_t *to = NULL, timeout = KTIME_MAX;
2829 	struct task_struct *tsk = current;
2830 	sigset_t mask = *which;
2831 	int sig, ret = 0;
2832 
2833 	if (ts) {
2834 		if (!timespec_valid(ts))
2835 			return -EINVAL;
2836 		timeout = timespec_to_ktime(*ts);
2837 		to = &timeout;
2838 	}
2839 
2840 	/*
2841 	 * Invert the set of allowed signals to get those we want to block.
2842 	 */
2843 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
2844 	signotset(&mask);
2845 
2846 	spin_lock_irq(&tsk->sighand->siglock);
2847 	sig = dequeue_signal(tsk, &mask, info);
2848 	if (!sig && timeout) {
2849 		/*
2850 		 * None ready, temporarily unblock those we're interested
2851 		 * while we are sleeping in so that we'll be awakened when
2852 		 * they arrive. Unblocking is always fine, we can avoid
2853 		 * set_current_blocked().
2854 		 */
2855 		tsk->real_blocked = tsk->blocked;
2856 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
2857 		recalc_sigpending();
2858 		spin_unlock_irq(&tsk->sighand->siglock);
2859 
2860 		__set_current_state(TASK_INTERRUPTIBLE);
2861 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
2862 							 HRTIMER_MODE_REL);
2863 		spin_lock_irq(&tsk->sighand->siglock);
2864 		__set_task_blocked(tsk, &tsk->real_blocked);
2865 		sigemptyset(&tsk->real_blocked);
2866 		sig = dequeue_signal(tsk, &mask, info);
2867 	}
2868 	spin_unlock_irq(&tsk->sighand->siglock);
2869 
2870 	if (sig)
2871 		return sig;
2872 	return ret ? -EINTR : -EAGAIN;
2873 }
2874 
2875 /**
2876  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
2877  *			in @uthese
2878  *  @uthese: queued signals to wait for
2879  *  @uinfo: if non-null, the signal's siginfo is returned here
2880  *  @uts: upper bound on process time suspension
2881  *  @sigsetsize: size of sigset_t type
2882  */
2883 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
2884 		siginfo_t __user *, uinfo, const struct timespec __user *, uts,
2885 		size_t, sigsetsize)
2886 {
2887 	sigset_t these;
2888 	struct timespec ts;
2889 	siginfo_t info;
2890 	int ret;
2891 
2892 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2893 	if (sigsetsize != sizeof(sigset_t))
2894 		return -EINVAL;
2895 
2896 	if (copy_from_user(&these, uthese, sizeof(these)))
2897 		return -EFAULT;
2898 
2899 	if (uts) {
2900 		if (copy_from_user(&ts, uts, sizeof(ts)))
2901 			return -EFAULT;
2902 	}
2903 
2904 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
2905 
2906 	if (ret > 0 && uinfo) {
2907 		if (copy_siginfo_to_user(uinfo, &info))
2908 			ret = -EFAULT;
2909 	}
2910 
2911 	return ret;
2912 }
2913 
2914 #ifdef CONFIG_COMPAT
2915 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait, compat_sigset_t __user *, uthese,
2916 		struct compat_siginfo __user *, uinfo,
2917 		struct compat_timespec __user *, uts, compat_size_t, sigsetsize)
2918 {
2919 	compat_sigset_t s32;
2920 	sigset_t s;
2921 	struct timespec t;
2922 	siginfo_t info;
2923 	long ret;
2924 
2925 	if (sigsetsize != sizeof(sigset_t))
2926 		return -EINVAL;
2927 
2928 	if (copy_from_user(&s32, uthese, sizeof(compat_sigset_t)))
2929 		return -EFAULT;
2930 	sigset_from_compat(&s, &s32);
2931 
2932 	if (uts) {
2933 		if (compat_get_timespec(&t, uts))
2934 			return -EFAULT;
2935 	}
2936 
2937 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
2938 
2939 	if (ret > 0 && uinfo) {
2940 		if (copy_siginfo_to_user32(uinfo, &info))
2941 			ret = -EFAULT;
2942 	}
2943 
2944 	return ret;
2945 }
2946 #endif
2947 
2948 /**
2949  *  sys_kill - send a signal to a process
2950  *  @pid: the PID of the process
2951  *  @sig: signal to be sent
2952  */
2953 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
2954 {
2955 	struct siginfo info;
2956 
2957 	info.si_signo = sig;
2958 	info.si_errno = 0;
2959 	info.si_code = SI_USER;
2960 	info.si_pid = task_tgid_vnr(current);
2961 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2962 
2963 	return kill_something_info(sig, &info, pid);
2964 }
2965 
2966 static int
2967 do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info)
2968 {
2969 	struct task_struct *p;
2970 	int error = -ESRCH;
2971 
2972 	rcu_read_lock();
2973 	p = find_task_by_vpid(pid);
2974 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
2975 		error = check_kill_permission(sig, info, p);
2976 		/*
2977 		 * The null signal is a permissions and process existence
2978 		 * probe.  No signal is actually delivered.
2979 		 */
2980 		if (!error && sig) {
2981 			error = do_send_sig_info(sig, info, p, false);
2982 			/*
2983 			 * If lock_task_sighand() failed we pretend the task
2984 			 * dies after receiving the signal. The window is tiny,
2985 			 * and the signal is private anyway.
2986 			 */
2987 			if (unlikely(error == -ESRCH))
2988 				error = 0;
2989 		}
2990 	}
2991 	rcu_read_unlock();
2992 
2993 	return error;
2994 }
2995 
2996 static int do_tkill(pid_t tgid, pid_t pid, int sig)
2997 {
2998 	struct siginfo info = {};
2999 
3000 	info.si_signo = sig;
3001 	info.si_errno = 0;
3002 	info.si_code = SI_TKILL;
3003 	info.si_pid = task_tgid_vnr(current);
3004 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3005 
3006 	return do_send_specific(tgid, pid, sig, &info);
3007 }
3008 
3009 /**
3010  *  sys_tgkill - send signal to one specific thread
3011  *  @tgid: the thread group ID of the thread
3012  *  @pid: the PID of the thread
3013  *  @sig: signal to be sent
3014  *
3015  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3016  *  exists but it's not belonging to the target process anymore. This
3017  *  method solves the problem of threads exiting and PIDs getting reused.
3018  */
3019 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3020 {
3021 	/* This is only valid for single tasks */
3022 	if (pid <= 0 || tgid <= 0)
3023 		return -EINVAL;
3024 
3025 	return do_tkill(tgid, pid, sig);
3026 }
3027 
3028 /**
3029  *  sys_tkill - send signal to one specific task
3030  *  @pid: the PID of the task
3031  *  @sig: signal to be sent
3032  *
3033  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3034  */
3035 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3036 {
3037 	/* This is only valid for single tasks */
3038 	if (pid <= 0)
3039 		return -EINVAL;
3040 
3041 	return do_tkill(0, pid, sig);
3042 }
3043 
3044 static int do_rt_sigqueueinfo(pid_t pid, int sig, siginfo_t *info)
3045 {
3046 	/* Not even root can pretend to send signals from the kernel.
3047 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3048 	 */
3049 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3050 	    (task_pid_vnr(current) != pid))
3051 		return -EPERM;
3052 
3053 	info->si_signo = sig;
3054 
3055 	/* POSIX.1b doesn't mention process groups.  */
3056 	return kill_proc_info(sig, info, pid);
3057 }
3058 
3059 /**
3060  *  sys_rt_sigqueueinfo - send signal information to a signal
3061  *  @pid: the PID of the thread
3062  *  @sig: signal to be sent
3063  *  @uinfo: signal info to be sent
3064  */
3065 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
3066 		siginfo_t __user *, uinfo)
3067 {
3068 	siginfo_t info;
3069 	if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
3070 		return -EFAULT;
3071 	return do_rt_sigqueueinfo(pid, sig, &info);
3072 }
3073 
3074 #ifdef CONFIG_COMPAT
3075 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
3076 			compat_pid_t, pid,
3077 			int, sig,
3078 			struct compat_siginfo __user *, uinfo)
3079 {
3080 	siginfo_t info = {};
3081 	int ret = copy_siginfo_from_user32(&info, uinfo);
3082 	if (unlikely(ret))
3083 		return ret;
3084 	return do_rt_sigqueueinfo(pid, sig, &info);
3085 }
3086 #endif
3087 
3088 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, siginfo_t *info)
3089 {
3090 	/* This is only valid for single tasks */
3091 	if (pid <= 0 || tgid <= 0)
3092 		return -EINVAL;
3093 
3094 	/* Not even root can pretend to send signals from the kernel.
3095 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3096 	 */
3097 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3098 	    (task_pid_vnr(current) != pid))
3099 		return -EPERM;
3100 
3101 	info->si_signo = sig;
3102 
3103 	return do_send_specific(tgid, pid, sig, info);
3104 }
3105 
3106 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
3107 		siginfo_t __user *, uinfo)
3108 {
3109 	siginfo_t info;
3110 
3111 	if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
3112 		return -EFAULT;
3113 
3114 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3115 }
3116 
3117 #ifdef CONFIG_COMPAT
3118 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
3119 			compat_pid_t, tgid,
3120 			compat_pid_t, pid,
3121 			int, sig,
3122 			struct compat_siginfo __user *, uinfo)
3123 {
3124 	siginfo_t info = {};
3125 
3126 	if (copy_siginfo_from_user32(&info, uinfo))
3127 		return -EFAULT;
3128 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3129 }
3130 #endif
3131 
3132 /*
3133  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
3134  */
3135 void kernel_sigaction(int sig, __sighandler_t action)
3136 {
3137 	spin_lock_irq(&current->sighand->siglock);
3138 	current->sighand->action[sig - 1].sa.sa_handler = action;
3139 	if (action == SIG_IGN) {
3140 		sigset_t mask;
3141 
3142 		sigemptyset(&mask);
3143 		sigaddset(&mask, sig);
3144 
3145 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
3146 		flush_sigqueue_mask(&mask, &current->pending);
3147 		recalc_sigpending();
3148 	}
3149 	spin_unlock_irq(&current->sighand->siglock);
3150 }
3151 EXPORT_SYMBOL(kernel_sigaction);
3152 
3153 void __weak sigaction_compat_abi(struct k_sigaction *act,
3154 		struct k_sigaction *oact)
3155 {
3156 }
3157 
3158 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
3159 {
3160 	struct task_struct *p = current, *t;
3161 	struct k_sigaction *k;
3162 	sigset_t mask;
3163 
3164 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
3165 		return -EINVAL;
3166 
3167 	k = &p->sighand->action[sig-1];
3168 
3169 	spin_lock_irq(&p->sighand->siglock);
3170 	if (oact)
3171 		*oact = *k;
3172 
3173 	sigaction_compat_abi(act, oact);
3174 
3175 	if (act) {
3176 		sigdelsetmask(&act->sa.sa_mask,
3177 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
3178 		*k = *act;
3179 		/*
3180 		 * POSIX 3.3.1.3:
3181 		 *  "Setting a signal action to SIG_IGN for a signal that is
3182 		 *   pending shall cause the pending signal to be discarded,
3183 		 *   whether or not it is blocked."
3184 		 *
3185 		 *  "Setting a signal action to SIG_DFL for a signal that is
3186 		 *   pending and whose default action is to ignore the signal
3187 		 *   (for example, SIGCHLD), shall cause the pending signal to
3188 		 *   be discarded, whether or not it is blocked"
3189 		 */
3190 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
3191 			sigemptyset(&mask);
3192 			sigaddset(&mask, sig);
3193 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
3194 			for_each_thread(p, t)
3195 				flush_sigqueue_mask(&mask, &t->pending);
3196 		}
3197 	}
3198 
3199 	spin_unlock_irq(&p->sighand->siglock);
3200 	return 0;
3201 }
3202 
3203 static int
3204 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp)
3205 {
3206 	struct task_struct *t = current;
3207 
3208 	if (oss) {
3209 		memset(oss, 0, sizeof(stack_t));
3210 		oss->ss_sp = (void __user *) t->sas_ss_sp;
3211 		oss->ss_size = t->sas_ss_size;
3212 		oss->ss_flags = sas_ss_flags(sp) |
3213 			(current->sas_ss_flags & SS_FLAG_BITS);
3214 	}
3215 
3216 	if (ss) {
3217 		void __user *ss_sp = ss->ss_sp;
3218 		size_t ss_size = ss->ss_size;
3219 		unsigned ss_flags = ss->ss_flags;
3220 		int ss_mode;
3221 
3222 		if (unlikely(on_sig_stack(sp)))
3223 			return -EPERM;
3224 
3225 		ss_mode = ss_flags & ~SS_FLAG_BITS;
3226 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
3227 				ss_mode != 0))
3228 			return -EINVAL;
3229 
3230 		if (ss_mode == SS_DISABLE) {
3231 			ss_size = 0;
3232 			ss_sp = NULL;
3233 		} else {
3234 			if (unlikely(ss_size < MINSIGSTKSZ))
3235 				return -ENOMEM;
3236 		}
3237 
3238 		t->sas_ss_sp = (unsigned long) ss_sp;
3239 		t->sas_ss_size = ss_size;
3240 		t->sas_ss_flags = ss_flags;
3241 	}
3242 	return 0;
3243 }
3244 
3245 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
3246 {
3247 	stack_t new, old;
3248 	int err;
3249 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
3250 		return -EFAULT;
3251 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
3252 			      current_user_stack_pointer());
3253 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
3254 		err = -EFAULT;
3255 	return err;
3256 }
3257 
3258 int restore_altstack(const stack_t __user *uss)
3259 {
3260 	stack_t new;
3261 	if (copy_from_user(&new, uss, sizeof(stack_t)))
3262 		return -EFAULT;
3263 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer());
3264 	/* squash all but EFAULT for now */
3265 	return 0;
3266 }
3267 
3268 int __save_altstack(stack_t __user *uss, unsigned long sp)
3269 {
3270 	struct task_struct *t = current;
3271 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
3272 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
3273 		__put_user(t->sas_ss_size, &uss->ss_size);
3274 	if (err)
3275 		return err;
3276 	if (t->sas_ss_flags & SS_AUTODISARM)
3277 		sas_ss_reset(t);
3278 	return 0;
3279 }
3280 
3281 #ifdef CONFIG_COMPAT
3282 COMPAT_SYSCALL_DEFINE2(sigaltstack,
3283 			const compat_stack_t __user *, uss_ptr,
3284 			compat_stack_t __user *, uoss_ptr)
3285 {
3286 	stack_t uss, uoss;
3287 	int ret;
3288 
3289 	if (uss_ptr) {
3290 		compat_stack_t uss32;
3291 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
3292 			return -EFAULT;
3293 		uss.ss_sp = compat_ptr(uss32.ss_sp);
3294 		uss.ss_flags = uss32.ss_flags;
3295 		uss.ss_size = uss32.ss_size;
3296 	}
3297 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
3298 			     compat_user_stack_pointer());
3299 	if (ret >= 0 && uoss_ptr)  {
3300 		compat_stack_t old;
3301 		memset(&old, 0, sizeof(old));
3302 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
3303 		old.ss_flags = uoss.ss_flags;
3304 		old.ss_size = uoss.ss_size;
3305 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
3306 			ret = -EFAULT;
3307 	}
3308 	return ret;
3309 }
3310 
3311 int compat_restore_altstack(const compat_stack_t __user *uss)
3312 {
3313 	int err = compat_sys_sigaltstack(uss, NULL);
3314 	/* squash all but -EFAULT for now */
3315 	return err == -EFAULT ? err : 0;
3316 }
3317 
3318 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
3319 {
3320 	int err;
3321 	struct task_struct *t = current;
3322 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
3323 			 &uss->ss_sp) |
3324 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
3325 		__put_user(t->sas_ss_size, &uss->ss_size);
3326 	if (err)
3327 		return err;
3328 	if (t->sas_ss_flags & SS_AUTODISARM)
3329 		sas_ss_reset(t);
3330 	return 0;
3331 }
3332 #endif
3333 
3334 #ifdef __ARCH_WANT_SYS_SIGPENDING
3335 
3336 /**
3337  *  sys_sigpending - examine pending signals
3338  *  @set: where mask of pending signal is returned
3339  */
3340 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, set)
3341 {
3342 	return sys_rt_sigpending((sigset_t __user *)set, sizeof(old_sigset_t));
3343 }
3344 
3345 #ifdef CONFIG_COMPAT
3346 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
3347 {
3348 #ifdef __BIG_ENDIAN
3349 	sigset_t set;
3350 	int err = do_sigpending(&set, sizeof(set.sig[0]));
3351 	if (!err)
3352 		err = put_user(set.sig[0], set32);
3353 	return err;
3354 #else
3355 	return sys_rt_sigpending((sigset_t __user *)set32, sizeof(*set32));
3356 #endif
3357 }
3358 #endif
3359 
3360 #endif
3361 
3362 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
3363 /**
3364  *  sys_sigprocmask - examine and change blocked signals
3365  *  @how: whether to add, remove, or set signals
3366  *  @nset: signals to add or remove (if non-null)
3367  *  @oset: previous value of signal mask if non-null
3368  *
3369  * Some platforms have their own version with special arguments;
3370  * others support only sys_rt_sigprocmask.
3371  */
3372 
3373 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
3374 		old_sigset_t __user *, oset)
3375 {
3376 	old_sigset_t old_set, new_set;
3377 	sigset_t new_blocked;
3378 
3379 	old_set = current->blocked.sig[0];
3380 
3381 	if (nset) {
3382 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
3383 			return -EFAULT;
3384 
3385 		new_blocked = current->blocked;
3386 
3387 		switch (how) {
3388 		case SIG_BLOCK:
3389 			sigaddsetmask(&new_blocked, new_set);
3390 			break;
3391 		case SIG_UNBLOCK:
3392 			sigdelsetmask(&new_blocked, new_set);
3393 			break;
3394 		case SIG_SETMASK:
3395 			new_blocked.sig[0] = new_set;
3396 			break;
3397 		default:
3398 			return -EINVAL;
3399 		}
3400 
3401 		set_current_blocked(&new_blocked);
3402 	}
3403 
3404 	if (oset) {
3405 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
3406 			return -EFAULT;
3407 	}
3408 
3409 	return 0;
3410 }
3411 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
3412 
3413 #ifndef CONFIG_ODD_RT_SIGACTION
3414 /**
3415  *  sys_rt_sigaction - alter an action taken by a process
3416  *  @sig: signal to be sent
3417  *  @act: new sigaction
3418  *  @oact: used to save the previous sigaction
3419  *  @sigsetsize: size of sigset_t type
3420  */
3421 SYSCALL_DEFINE4(rt_sigaction, int, sig,
3422 		const struct sigaction __user *, act,
3423 		struct sigaction __user *, oact,
3424 		size_t, sigsetsize)
3425 {
3426 	struct k_sigaction new_sa, old_sa;
3427 	int ret = -EINVAL;
3428 
3429 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3430 	if (sigsetsize != sizeof(sigset_t))
3431 		goto out;
3432 
3433 	if (act) {
3434 		if (copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
3435 			return -EFAULT;
3436 	}
3437 
3438 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
3439 
3440 	if (!ret && oact) {
3441 		if (copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
3442 			return -EFAULT;
3443 	}
3444 out:
3445 	return ret;
3446 }
3447 #ifdef CONFIG_COMPAT
3448 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
3449 		const struct compat_sigaction __user *, act,
3450 		struct compat_sigaction __user *, oact,
3451 		compat_size_t, sigsetsize)
3452 {
3453 	struct k_sigaction new_ka, old_ka;
3454 	compat_sigset_t mask;
3455 #ifdef __ARCH_HAS_SA_RESTORER
3456 	compat_uptr_t restorer;
3457 #endif
3458 	int ret;
3459 
3460 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3461 	if (sigsetsize != sizeof(compat_sigset_t))
3462 		return -EINVAL;
3463 
3464 	if (act) {
3465 		compat_uptr_t handler;
3466 		ret = get_user(handler, &act->sa_handler);
3467 		new_ka.sa.sa_handler = compat_ptr(handler);
3468 #ifdef __ARCH_HAS_SA_RESTORER
3469 		ret |= get_user(restorer, &act->sa_restorer);
3470 		new_ka.sa.sa_restorer = compat_ptr(restorer);
3471 #endif
3472 		ret |= copy_from_user(&mask, &act->sa_mask, sizeof(mask));
3473 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
3474 		if (ret)
3475 			return -EFAULT;
3476 		sigset_from_compat(&new_ka.sa.sa_mask, &mask);
3477 	}
3478 
3479 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3480 	if (!ret && oact) {
3481 		sigset_to_compat(&mask, &old_ka.sa.sa_mask);
3482 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
3483 			       &oact->sa_handler);
3484 		ret |= copy_to_user(&oact->sa_mask, &mask, sizeof(mask));
3485 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
3486 #ifdef __ARCH_HAS_SA_RESTORER
3487 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
3488 				&oact->sa_restorer);
3489 #endif
3490 	}
3491 	return ret;
3492 }
3493 #endif
3494 #endif /* !CONFIG_ODD_RT_SIGACTION */
3495 
3496 #ifdef CONFIG_OLD_SIGACTION
3497 SYSCALL_DEFINE3(sigaction, int, sig,
3498 		const struct old_sigaction __user *, act,
3499 	        struct old_sigaction __user *, oact)
3500 {
3501 	struct k_sigaction new_ka, old_ka;
3502 	int ret;
3503 
3504 	if (act) {
3505 		old_sigset_t mask;
3506 		if (!access_ok(VERIFY_READ, act, sizeof(*act)) ||
3507 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
3508 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
3509 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
3510 		    __get_user(mask, &act->sa_mask))
3511 			return -EFAULT;
3512 #ifdef __ARCH_HAS_KA_RESTORER
3513 		new_ka.ka_restorer = NULL;
3514 #endif
3515 		siginitset(&new_ka.sa.sa_mask, mask);
3516 	}
3517 
3518 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3519 
3520 	if (!ret && oact) {
3521 		if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)) ||
3522 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
3523 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
3524 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
3525 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
3526 			return -EFAULT;
3527 	}
3528 
3529 	return ret;
3530 }
3531 #endif
3532 #ifdef CONFIG_COMPAT_OLD_SIGACTION
3533 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
3534 		const struct compat_old_sigaction __user *, act,
3535 	        struct compat_old_sigaction __user *, oact)
3536 {
3537 	struct k_sigaction new_ka, old_ka;
3538 	int ret;
3539 	compat_old_sigset_t mask;
3540 	compat_uptr_t handler, restorer;
3541 
3542 	if (act) {
3543 		if (!access_ok(VERIFY_READ, act, sizeof(*act)) ||
3544 		    __get_user(handler, &act->sa_handler) ||
3545 		    __get_user(restorer, &act->sa_restorer) ||
3546 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
3547 		    __get_user(mask, &act->sa_mask))
3548 			return -EFAULT;
3549 
3550 #ifdef __ARCH_HAS_KA_RESTORER
3551 		new_ka.ka_restorer = NULL;
3552 #endif
3553 		new_ka.sa.sa_handler = compat_ptr(handler);
3554 		new_ka.sa.sa_restorer = compat_ptr(restorer);
3555 		siginitset(&new_ka.sa.sa_mask, mask);
3556 	}
3557 
3558 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3559 
3560 	if (!ret && oact) {
3561 		if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)) ||
3562 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
3563 			       &oact->sa_handler) ||
3564 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
3565 			       &oact->sa_restorer) ||
3566 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
3567 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
3568 			return -EFAULT;
3569 	}
3570 	return ret;
3571 }
3572 #endif
3573 
3574 #ifdef CONFIG_SGETMASK_SYSCALL
3575 
3576 /*
3577  * For backwards compatibility.  Functionality superseded by sigprocmask.
3578  */
3579 SYSCALL_DEFINE0(sgetmask)
3580 {
3581 	/* SMP safe */
3582 	return current->blocked.sig[0];
3583 }
3584 
3585 SYSCALL_DEFINE1(ssetmask, int, newmask)
3586 {
3587 	int old = current->blocked.sig[0];
3588 	sigset_t newset;
3589 
3590 	siginitset(&newset, newmask);
3591 	set_current_blocked(&newset);
3592 
3593 	return old;
3594 }
3595 #endif /* CONFIG_SGETMASK_SYSCALL */
3596 
3597 #ifdef __ARCH_WANT_SYS_SIGNAL
3598 /*
3599  * For backwards compatibility.  Functionality superseded by sigaction.
3600  */
3601 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
3602 {
3603 	struct k_sigaction new_sa, old_sa;
3604 	int ret;
3605 
3606 	new_sa.sa.sa_handler = handler;
3607 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
3608 	sigemptyset(&new_sa.sa.sa_mask);
3609 
3610 	ret = do_sigaction(sig, &new_sa, &old_sa);
3611 
3612 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
3613 }
3614 #endif /* __ARCH_WANT_SYS_SIGNAL */
3615 
3616 #ifdef __ARCH_WANT_SYS_PAUSE
3617 
3618 SYSCALL_DEFINE0(pause)
3619 {
3620 	while (!signal_pending(current)) {
3621 		__set_current_state(TASK_INTERRUPTIBLE);
3622 		schedule();
3623 	}
3624 	return -ERESTARTNOHAND;
3625 }
3626 
3627 #endif
3628 
3629 static int sigsuspend(sigset_t *set)
3630 {
3631 	current->saved_sigmask = current->blocked;
3632 	set_current_blocked(set);
3633 
3634 	while (!signal_pending(current)) {
3635 		__set_current_state(TASK_INTERRUPTIBLE);
3636 		schedule();
3637 	}
3638 	set_restore_sigmask();
3639 	return -ERESTARTNOHAND;
3640 }
3641 
3642 /**
3643  *  sys_rt_sigsuspend - replace the signal mask for a value with the
3644  *	@unewset value until a signal is received
3645  *  @unewset: new signal mask value
3646  *  @sigsetsize: size of sigset_t type
3647  */
3648 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
3649 {
3650 	sigset_t newset;
3651 
3652 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3653 	if (sigsetsize != sizeof(sigset_t))
3654 		return -EINVAL;
3655 
3656 	if (copy_from_user(&newset, unewset, sizeof(newset)))
3657 		return -EFAULT;
3658 	return sigsuspend(&newset);
3659 }
3660 
3661 #ifdef CONFIG_COMPAT
3662 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
3663 {
3664 #ifdef __BIG_ENDIAN
3665 	sigset_t newset;
3666 	compat_sigset_t newset32;
3667 
3668 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3669 	if (sigsetsize != sizeof(sigset_t))
3670 		return -EINVAL;
3671 
3672 	if (copy_from_user(&newset32, unewset, sizeof(compat_sigset_t)))
3673 		return -EFAULT;
3674 	sigset_from_compat(&newset, &newset32);
3675 	return sigsuspend(&newset);
3676 #else
3677 	/* on little-endian bitmaps don't care about granularity */
3678 	return sys_rt_sigsuspend((sigset_t __user *)unewset, sigsetsize);
3679 #endif
3680 }
3681 #endif
3682 
3683 #ifdef CONFIG_OLD_SIGSUSPEND
3684 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
3685 {
3686 	sigset_t blocked;
3687 	siginitset(&blocked, mask);
3688 	return sigsuspend(&blocked);
3689 }
3690 #endif
3691 #ifdef CONFIG_OLD_SIGSUSPEND3
3692 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
3693 {
3694 	sigset_t blocked;
3695 	siginitset(&blocked, mask);
3696 	return sigsuspend(&blocked);
3697 }
3698 #endif
3699 
3700 __weak const char *arch_vma_name(struct vm_area_struct *vma)
3701 {
3702 	return NULL;
3703 }
3704 
3705 void __init signals_init(void)
3706 {
3707 	/* If this check fails, the __ARCH_SI_PREAMBLE_SIZE value is wrong! */
3708 	BUILD_BUG_ON(__ARCH_SI_PREAMBLE_SIZE
3709 		!= offsetof(struct siginfo, _sifields._pad));
3710 
3711 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
3712 }
3713 
3714 #ifdef CONFIG_KGDB_KDB
3715 #include <linux/kdb.h>
3716 /*
3717  * kdb_send_sig_info - Allows kdb to send signals without exposing
3718  * signal internals.  This function checks if the required locks are
3719  * available before calling the main signal code, to avoid kdb
3720  * deadlocks.
3721  */
3722 void
3723 kdb_send_sig_info(struct task_struct *t, struct siginfo *info)
3724 {
3725 	static struct task_struct *kdb_prev_t;
3726 	int sig, new_t;
3727 	if (!spin_trylock(&t->sighand->siglock)) {
3728 		kdb_printf("Can't do kill command now.\n"
3729 			   "The sigmask lock is held somewhere else in "
3730 			   "kernel, try again later\n");
3731 		return;
3732 	}
3733 	spin_unlock(&t->sighand->siglock);
3734 	new_t = kdb_prev_t != t;
3735 	kdb_prev_t = t;
3736 	if (t->state != TASK_RUNNING && new_t) {
3737 		kdb_printf("Process is not RUNNING, sending a signal from "
3738 			   "kdb risks deadlock\n"
3739 			   "on the run queue locks. "
3740 			   "The signal has _not_ been sent.\n"
3741 			   "Reissue the kill command if you want to risk "
3742 			   "the deadlock.\n");
3743 		return;
3744 	}
3745 	sig = info->si_signo;
3746 	if (send_sig_info(sig, info, t))
3747 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
3748 			   sig, t->pid);
3749 	else
3750 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
3751 }
3752 #endif	/* CONFIG_KGDB_KDB */
3753