xref: /linux/kernel/signal.c (revision 69fb09f6ccdb2f070557fd1f4c56c4d646694c8e)
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 	if (action->sa.sa_handler == SIG_DFL)
1198 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1199 	ret = specific_send_sig_info(sig, info, t);
1200 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1201 
1202 	return ret;
1203 }
1204 
1205 /*
1206  * Nuke all other threads in the group.
1207  */
1208 int zap_other_threads(struct task_struct *p)
1209 {
1210 	struct task_struct *t = p;
1211 	int count = 0;
1212 
1213 	p->signal->group_stop_count = 0;
1214 
1215 	while_each_thread(p, t) {
1216 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1217 		count++;
1218 
1219 		/* Don't bother with already dead threads */
1220 		if (t->exit_state)
1221 			continue;
1222 		sigaddset(&t->pending.signal, SIGKILL);
1223 		signal_wake_up(t, 1);
1224 	}
1225 
1226 	return count;
1227 }
1228 
1229 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1230 					   unsigned long *flags)
1231 {
1232 	struct sighand_struct *sighand;
1233 
1234 	for (;;) {
1235 		/*
1236 		 * Disable interrupts early to avoid deadlocks.
1237 		 * See rcu_read_unlock() comment header for details.
1238 		 */
1239 		local_irq_save(*flags);
1240 		rcu_read_lock();
1241 		sighand = rcu_dereference(tsk->sighand);
1242 		if (unlikely(sighand == NULL)) {
1243 			rcu_read_unlock();
1244 			local_irq_restore(*flags);
1245 			break;
1246 		}
1247 		/*
1248 		 * This sighand can be already freed and even reused, but
1249 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1250 		 * initializes ->siglock: this slab can't go away, it has
1251 		 * the same object type, ->siglock can't be reinitialized.
1252 		 *
1253 		 * We need to ensure that tsk->sighand is still the same
1254 		 * after we take the lock, we can race with de_thread() or
1255 		 * __exit_signal(). In the latter case the next iteration
1256 		 * must see ->sighand == NULL.
1257 		 */
1258 		spin_lock(&sighand->siglock);
1259 		if (likely(sighand == tsk->sighand)) {
1260 			rcu_read_unlock();
1261 			break;
1262 		}
1263 		spin_unlock(&sighand->siglock);
1264 		rcu_read_unlock();
1265 		local_irq_restore(*flags);
1266 	}
1267 
1268 	return sighand;
1269 }
1270 
1271 /*
1272  * send signal info to all the members of a group
1273  */
1274 int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1275 {
1276 	int ret;
1277 
1278 	rcu_read_lock();
1279 	ret = check_kill_permission(sig, info, p);
1280 	rcu_read_unlock();
1281 
1282 	if (!ret && sig)
1283 		ret = do_send_sig_info(sig, info, p, true);
1284 
1285 	return ret;
1286 }
1287 
1288 /*
1289  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1290  * control characters do (^C, ^Z etc)
1291  * - the caller must hold at least a readlock on tasklist_lock
1292  */
1293 int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp)
1294 {
1295 	struct task_struct *p = NULL;
1296 	int retval, success;
1297 
1298 	success = 0;
1299 	retval = -ESRCH;
1300 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1301 		int err = group_send_sig_info(sig, info, p);
1302 		success |= !err;
1303 		retval = err;
1304 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1305 	return success ? 0 : retval;
1306 }
1307 
1308 int kill_pid_info(int sig, struct siginfo *info, struct pid *pid)
1309 {
1310 	int error = -ESRCH;
1311 	struct task_struct *p;
1312 
1313 	for (;;) {
1314 		rcu_read_lock();
1315 		p = pid_task(pid, PIDTYPE_PID);
1316 		if (p)
1317 			error = group_send_sig_info(sig, info, p);
1318 		rcu_read_unlock();
1319 		if (likely(!p || error != -ESRCH))
1320 			return error;
1321 
1322 		/*
1323 		 * The task was unhashed in between, try again.  If it
1324 		 * is dead, pid_task() will return NULL, if we race with
1325 		 * de_thread() it will find the new leader.
1326 		 */
1327 	}
1328 }
1329 
1330 static int kill_proc_info(int sig, struct siginfo *info, pid_t pid)
1331 {
1332 	int error;
1333 	rcu_read_lock();
1334 	error = kill_pid_info(sig, info, find_vpid(pid));
1335 	rcu_read_unlock();
1336 	return error;
1337 }
1338 
1339 static int kill_as_cred_perm(const struct cred *cred,
1340 			     struct task_struct *target)
1341 {
1342 	const struct cred *pcred = __task_cred(target);
1343 	if (!uid_eq(cred->euid, pcred->suid) && !uid_eq(cred->euid, pcred->uid) &&
1344 	    !uid_eq(cred->uid,  pcred->suid) && !uid_eq(cred->uid,  pcred->uid))
1345 		return 0;
1346 	return 1;
1347 }
1348 
1349 /* like kill_pid_info(), but doesn't use uid/euid of "current" */
1350 int kill_pid_info_as_cred(int sig, struct siginfo *info, struct pid *pid,
1351 			 const struct cred *cred, u32 secid)
1352 {
1353 	int ret = -EINVAL;
1354 	struct task_struct *p;
1355 	unsigned long flags;
1356 
1357 	if (!valid_signal(sig))
1358 		return ret;
1359 
1360 	rcu_read_lock();
1361 	p = pid_task(pid, PIDTYPE_PID);
1362 	if (!p) {
1363 		ret = -ESRCH;
1364 		goto out_unlock;
1365 	}
1366 	if (si_fromuser(info) && !kill_as_cred_perm(cred, p)) {
1367 		ret = -EPERM;
1368 		goto out_unlock;
1369 	}
1370 	ret = security_task_kill(p, info, sig, secid);
1371 	if (ret)
1372 		goto out_unlock;
1373 
1374 	if (sig) {
1375 		if (lock_task_sighand(p, &flags)) {
1376 			ret = __send_signal(sig, info, p, 1, 0);
1377 			unlock_task_sighand(p, &flags);
1378 		} else
1379 			ret = -ESRCH;
1380 	}
1381 out_unlock:
1382 	rcu_read_unlock();
1383 	return ret;
1384 }
1385 EXPORT_SYMBOL_GPL(kill_pid_info_as_cred);
1386 
1387 /*
1388  * kill_something_info() interprets pid in interesting ways just like kill(2).
1389  *
1390  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1391  * is probably wrong.  Should make it like BSD or SYSV.
1392  */
1393 
1394 static int kill_something_info(int sig, struct siginfo *info, pid_t pid)
1395 {
1396 	int ret;
1397 
1398 	if (pid > 0) {
1399 		rcu_read_lock();
1400 		ret = kill_pid_info(sig, info, find_vpid(pid));
1401 		rcu_read_unlock();
1402 		return ret;
1403 	}
1404 
1405 	read_lock(&tasklist_lock);
1406 	if (pid != -1) {
1407 		ret = __kill_pgrp_info(sig, info,
1408 				pid ? find_vpid(-pid) : task_pgrp(current));
1409 	} else {
1410 		int retval = 0, count = 0;
1411 		struct task_struct * p;
1412 
1413 		for_each_process(p) {
1414 			if (task_pid_vnr(p) > 1 &&
1415 					!same_thread_group(p, current)) {
1416 				int err = group_send_sig_info(sig, info, p);
1417 				++count;
1418 				if (err != -EPERM)
1419 					retval = err;
1420 			}
1421 		}
1422 		ret = count ? retval : -ESRCH;
1423 	}
1424 	read_unlock(&tasklist_lock);
1425 
1426 	return ret;
1427 }
1428 
1429 /*
1430  * These are for backward compatibility with the rest of the kernel source.
1431  */
1432 
1433 int send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1434 {
1435 	/*
1436 	 * Make sure legacy kernel users don't send in bad values
1437 	 * (normal paths check this in check_kill_permission).
1438 	 */
1439 	if (!valid_signal(sig))
1440 		return -EINVAL;
1441 
1442 	return do_send_sig_info(sig, info, p, false);
1443 }
1444 
1445 #define __si_special(priv) \
1446 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1447 
1448 int
1449 send_sig(int sig, struct task_struct *p, int priv)
1450 {
1451 	return send_sig_info(sig, __si_special(priv), p);
1452 }
1453 
1454 void
1455 force_sig(int sig, struct task_struct *p)
1456 {
1457 	force_sig_info(sig, SEND_SIG_PRIV, p);
1458 }
1459 
1460 /*
1461  * When things go south during signal handling, we
1462  * will force a SIGSEGV. And if the signal that caused
1463  * the problem was already a SIGSEGV, we'll want to
1464  * make sure we don't even try to deliver the signal..
1465  */
1466 int
1467 force_sigsegv(int sig, struct task_struct *p)
1468 {
1469 	if (sig == SIGSEGV) {
1470 		unsigned long flags;
1471 		spin_lock_irqsave(&p->sighand->siglock, flags);
1472 		p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1473 		spin_unlock_irqrestore(&p->sighand->siglock, flags);
1474 	}
1475 	force_sig(SIGSEGV, p);
1476 	return 0;
1477 }
1478 
1479 int kill_pgrp(struct pid *pid, int sig, int priv)
1480 {
1481 	int ret;
1482 
1483 	read_lock(&tasklist_lock);
1484 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1485 	read_unlock(&tasklist_lock);
1486 
1487 	return ret;
1488 }
1489 EXPORT_SYMBOL(kill_pgrp);
1490 
1491 int kill_pid(struct pid *pid, int sig, int priv)
1492 {
1493 	return kill_pid_info(sig, __si_special(priv), pid);
1494 }
1495 EXPORT_SYMBOL(kill_pid);
1496 
1497 /*
1498  * These functions support sending signals using preallocated sigqueue
1499  * structures.  This is needed "because realtime applications cannot
1500  * afford to lose notifications of asynchronous events, like timer
1501  * expirations or I/O completions".  In the case of POSIX Timers
1502  * we allocate the sigqueue structure from the timer_create.  If this
1503  * allocation fails we are able to report the failure to the application
1504  * with an EAGAIN error.
1505  */
1506 struct sigqueue *sigqueue_alloc(void)
1507 {
1508 	struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1509 
1510 	if (q)
1511 		q->flags |= SIGQUEUE_PREALLOC;
1512 
1513 	return q;
1514 }
1515 
1516 void sigqueue_free(struct sigqueue *q)
1517 {
1518 	unsigned long flags;
1519 	spinlock_t *lock = &current->sighand->siglock;
1520 
1521 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1522 	/*
1523 	 * We must hold ->siglock while testing q->list
1524 	 * to serialize with collect_signal() or with
1525 	 * __exit_signal()->flush_sigqueue().
1526 	 */
1527 	spin_lock_irqsave(lock, flags);
1528 	q->flags &= ~SIGQUEUE_PREALLOC;
1529 	/*
1530 	 * If it is queued it will be freed when dequeued,
1531 	 * like the "regular" sigqueue.
1532 	 */
1533 	if (!list_empty(&q->list))
1534 		q = NULL;
1535 	spin_unlock_irqrestore(lock, flags);
1536 
1537 	if (q)
1538 		__sigqueue_free(q);
1539 }
1540 
1541 int send_sigqueue(struct sigqueue *q, struct task_struct *t, int group)
1542 {
1543 	int sig = q->info.si_signo;
1544 	struct sigpending *pending;
1545 	unsigned long flags;
1546 	int ret, result;
1547 
1548 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1549 
1550 	ret = -1;
1551 	if (!likely(lock_task_sighand(t, &flags)))
1552 		goto ret;
1553 
1554 	ret = 1; /* the signal is ignored */
1555 	result = TRACE_SIGNAL_IGNORED;
1556 	if (!prepare_signal(sig, t, false))
1557 		goto out;
1558 
1559 	ret = 0;
1560 	if (unlikely(!list_empty(&q->list))) {
1561 		/*
1562 		 * If an SI_TIMER entry is already queue just increment
1563 		 * the overrun count.
1564 		 */
1565 		BUG_ON(q->info.si_code != SI_TIMER);
1566 		q->info.si_overrun++;
1567 		result = TRACE_SIGNAL_ALREADY_PENDING;
1568 		goto out;
1569 	}
1570 	q->info.si_overrun = 0;
1571 
1572 	signalfd_notify(t, sig);
1573 	pending = group ? &t->signal->shared_pending : &t->pending;
1574 	list_add_tail(&q->list, &pending->list);
1575 	sigaddset(&pending->signal, sig);
1576 	complete_signal(sig, t, group);
1577 	result = TRACE_SIGNAL_DELIVERED;
1578 out:
1579 	trace_signal_generate(sig, &q->info, t, group, result);
1580 	unlock_task_sighand(t, &flags);
1581 ret:
1582 	return ret;
1583 }
1584 
1585 /*
1586  * Let a parent know about the death of a child.
1587  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1588  *
1589  * Returns true if our parent ignored us and so we've switched to
1590  * self-reaping.
1591  */
1592 bool do_notify_parent(struct task_struct *tsk, int sig)
1593 {
1594 	struct siginfo info;
1595 	unsigned long flags;
1596 	struct sighand_struct *psig;
1597 	bool autoreap = false;
1598 	u64 utime, stime;
1599 
1600 	BUG_ON(sig == -1);
1601 
1602  	/* do_notify_parent_cldstop should have been called instead.  */
1603  	BUG_ON(task_is_stopped_or_traced(tsk));
1604 
1605 	BUG_ON(!tsk->ptrace &&
1606 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1607 
1608 	if (sig != SIGCHLD) {
1609 		/*
1610 		 * This is only possible if parent == real_parent.
1611 		 * Check if it has changed security domain.
1612 		 */
1613 		if (tsk->parent_exec_id != tsk->parent->self_exec_id)
1614 			sig = SIGCHLD;
1615 	}
1616 
1617 	info.si_signo = sig;
1618 	info.si_errno = 0;
1619 	/*
1620 	 * We are under tasklist_lock here so our parent is tied to
1621 	 * us and cannot change.
1622 	 *
1623 	 * task_active_pid_ns will always return the same pid namespace
1624 	 * until a task passes through release_task.
1625 	 *
1626 	 * write_lock() currently calls preempt_disable() which is the
1627 	 * same as rcu_read_lock(), but according to Oleg, this is not
1628 	 * correct to rely on this
1629 	 */
1630 	rcu_read_lock();
1631 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1632 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1633 				       task_uid(tsk));
1634 	rcu_read_unlock();
1635 
1636 	task_cputime(tsk, &utime, &stime);
1637 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
1638 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
1639 
1640 	info.si_status = tsk->exit_code & 0x7f;
1641 	if (tsk->exit_code & 0x80)
1642 		info.si_code = CLD_DUMPED;
1643 	else if (tsk->exit_code & 0x7f)
1644 		info.si_code = CLD_KILLED;
1645 	else {
1646 		info.si_code = CLD_EXITED;
1647 		info.si_status = tsk->exit_code >> 8;
1648 	}
1649 
1650 	psig = tsk->parent->sighand;
1651 	spin_lock_irqsave(&psig->siglock, flags);
1652 	if (!tsk->ptrace && sig == SIGCHLD &&
1653 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1654 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1655 		/*
1656 		 * We are exiting and our parent doesn't care.  POSIX.1
1657 		 * defines special semantics for setting SIGCHLD to SIG_IGN
1658 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
1659 		 * automatically and not left for our parent's wait4 call.
1660 		 * Rather than having the parent do it as a magic kind of
1661 		 * signal handler, we just set this to tell do_exit that we
1662 		 * can be cleaned up without becoming a zombie.  Note that
1663 		 * we still call __wake_up_parent in this case, because a
1664 		 * blocked sys_wait4 might now return -ECHILD.
1665 		 *
1666 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1667 		 * is implementation-defined: we do (if you don't want
1668 		 * it, just use SIG_IGN instead).
1669 		 */
1670 		autoreap = true;
1671 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
1672 			sig = 0;
1673 	}
1674 	if (valid_signal(sig) && sig)
1675 		__group_send_sig_info(sig, &info, tsk->parent);
1676 	__wake_up_parent(tsk, tsk->parent);
1677 	spin_unlock_irqrestore(&psig->siglock, flags);
1678 
1679 	return autoreap;
1680 }
1681 
1682 /**
1683  * do_notify_parent_cldstop - notify parent of stopped/continued state change
1684  * @tsk: task reporting the state change
1685  * @for_ptracer: the notification is for ptracer
1686  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
1687  *
1688  * Notify @tsk's parent that the stopped/continued state has changed.  If
1689  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
1690  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
1691  *
1692  * CONTEXT:
1693  * Must be called with tasklist_lock at least read locked.
1694  */
1695 static void do_notify_parent_cldstop(struct task_struct *tsk,
1696 				     bool for_ptracer, int why)
1697 {
1698 	struct siginfo info;
1699 	unsigned long flags;
1700 	struct task_struct *parent;
1701 	struct sighand_struct *sighand;
1702 	u64 utime, stime;
1703 
1704 	if (for_ptracer) {
1705 		parent = tsk->parent;
1706 	} else {
1707 		tsk = tsk->group_leader;
1708 		parent = tsk->real_parent;
1709 	}
1710 
1711 	info.si_signo = SIGCHLD;
1712 	info.si_errno = 0;
1713 	/*
1714 	 * see comment in do_notify_parent() about the following 4 lines
1715 	 */
1716 	rcu_read_lock();
1717 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
1718 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
1719 	rcu_read_unlock();
1720 
1721 	task_cputime(tsk, &utime, &stime);
1722 	info.si_utime = nsec_to_clock_t(utime);
1723 	info.si_stime = nsec_to_clock_t(stime);
1724 
1725  	info.si_code = why;
1726  	switch (why) {
1727  	case CLD_CONTINUED:
1728  		info.si_status = SIGCONT;
1729  		break;
1730  	case CLD_STOPPED:
1731  		info.si_status = tsk->signal->group_exit_code & 0x7f;
1732  		break;
1733  	case CLD_TRAPPED:
1734  		info.si_status = tsk->exit_code & 0x7f;
1735  		break;
1736  	default:
1737  		BUG();
1738  	}
1739 
1740 	sighand = parent->sighand;
1741 	spin_lock_irqsave(&sighand->siglock, flags);
1742 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
1743 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
1744 		__group_send_sig_info(SIGCHLD, &info, parent);
1745 	/*
1746 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
1747 	 */
1748 	__wake_up_parent(tsk, parent);
1749 	spin_unlock_irqrestore(&sighand->siglock, flags);
1750 }
1751 
1752 static inline int may_ptrace_stop(void)
1753 {
1754 	if (!likely(current->ptrace))
1755 		return 0;
1756 	/*
1757 	 * Are we in the middle of do_coredump?
1758 	 * If so and our tracer is also part of the coredump stopping
1759 	 * is a deadlock situation, and pointless because our tracer
1760 	 * is dead so don't allow us to stop.
1761 	 * If SIGKILL was already sent before the caller unlocked
1762 	 * ->siglock we must see ->core_state != NULL. Otherwise it
1763 	 * is safe to enter schedule().
1764 	 *
1765 	 * This is almost outdated, a task with the pending SIGKILL can't
1766 	 * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
1767 	 * after SIGKILL was already dequeued.
1768 	 */
1769 	if (unlikely(current->mm->core_state) &&
1770 	    unlikely(current->mm == current->parent->mm))
1771 		return 0;
1772 
1773 	return 1;
1774 }
1775 
1776 /*
1777  * Return non-zero if there is a SIGKILL that should be waking us up.
1778  * Called with the siglock held.
1779  */
1780 static int sigkill_pending(struct task_struct *tsk)
1781 {
1782 	return	sigismember(&tsk->pending.signal, SIGKILL) ||
1783 		sigismember(&tsk->signal->shared_pending.signal, SIGKILL);
1784 }
1785 
1786 /*
1787  * This must be called with current->sighand->siglock held.
1788  *
1789  * This should be the path for all ptrace stops.
1790  * We always set current->last_siginfo while stopped here.
1791  * That makes it a way to test a stopped process for
1792  * being ptrace-stopped vs being job-control-stopped.
1793  *
1794  * If we actually decide not to stop at all because the tracer
1795  * is gone, we keep current->exit_code unless clear_code.
1796  */
1797 static void ptrace_stop(int exit_code, int why, int clear_code, siginfo_t *info)
1798 	__releases(&current->sighand->siglock)
1799 	__acquires(&current->sighand->siglock)
1800 {
1801 	bool gstop_done = false;
1802 
1803 	if (arch_ptrace_stop_needed(exit_code, info)) {
1804 		/*
1805 		 * The arch code has something special to do before a
1806 		 * ptrace stop.  This is allowed to block, e.g. for faults
1807 		 * on user stack pages.  We can't keep the siglock while
1808 		 * calling arch_ptrace_stop, so we must release it now.
1809 		 * To preserve proper semantics, we must do this before
1810 		 * any signal bookkeeping like checking group_stop_count.
1811 		 * Meanwhile, a SIGKILL could come in before we retake the
1812 		 * siglock.  That must prevent us from sleeping in TASK_TRACED.
1813 		 * So after regaining the lock, we must check for SIGKILL.
1814 		 */
1815 		spin_unlock_irq(&current->sighand->siglock);
1816 		arch_ptrace_stop(exit_code, info);
1817 		spin_lock_irq(&current->sighand->siglock);
1818 		if (sigkill_pending(current))
1819 			return;
1820 	}
1821 
1822 	/*
1823 	 * We're committing to trapping.  TRACED should be visible before
1824 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
1825 	 * Also, transition to TRACED and updates to ->jobctl should be
1826 	 * atomic with respect to siglock and should be done after the arch
1827 	 * hook as siglock is released and regrabbed across it.
1828 	 */
1829 	set_current_state(TASK_TRACED);
1830 
1831 	current->last_siginfo = info;
1832 	current->exit_code = exit_code;
1833 
1834 	/*
1835 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
1836 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
1837 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
1838 	 * could be clear now.  We act as if SIGCONT is received after
1839 	 * TASK_TRACED is entered - ignore it.
1840 	 */
1841 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
1842 		gstop_done = task_participate_group_stop(current);
1843 
1844 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
1845 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
1846 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
1847 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
1848 
1849 	/* entering a trap, clear TRAPPING */
1850 	task_clear_jobctl_trapping(current);
1851 
1852 	spin_unlock_irq(&current->sighand->siglock);
1853 	read_lock(&tasklist_lock);
1854 	if (may_ptrace_stop()) {
1855 		/*
1856 		 * Notify parents of the stop.
1857 		 *
1858 		 * While ptraced, there are two parents - the ptracer and
1859 		 * the real_parent of the group_leader.  The ptracer should
1860 		 * know about every stop while the real parent is only
1861 		 * interested in the completion of group stop.  The states
1862 		 * for the two don't interact with each other.  Notify
1863 		 * separately unless they're gonna be duplicates.
1864 		 */
1865 		do_notify_parent_cldstop(current, true, why);
1866 		if (gstop_done && ptrace_reparented(current))
1867 			do_notify_parent_cldstop(current, false, why);
1868 
1869 		/*
1870 		 * Don't want to allow preemption here, because
1871 		 * sys_ptrace() needs this task to be inactive.
1872 		 *
1873 		 * XXX: implement read_unlock_no_resched().
1874 		 */
1875 		preempt_disable();
1876 		read_unlock(&tasklist_lock);
1877 		preempt_enable_no_resched();
1878 		freezable_schedule();
1879 	} else {
1880 		/*
1881 		 * By the time we got the lock, our tracer went away.
1882 		 * Don't drop the lock yet, another tracer may come.
1883 		 *
1884 		 * If @gstop_done, the ptracer went away between group stop
1885 		 * completion and here.  During detach, it would have set
1886 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
1887 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
1888 		 * the real parent of the group stop completion is enough.
1889 		 */
1890 		if (gstop_done)
1891 			do_notify_parent_cldstop(current, false, why);
1892 
1893 		/* tasklist protects us from ptrace_freeze_traced() */
1894 		__set_current_state(TASK_RUNNING);
1895 		if (clear_code)
1896 			current->exit_code = 0;
1897 		read_unlock(&tasklist_lock);
1898 	}
1899 
1900 	/*
1901 	 * We are back.  Now reacquire the siglock before touching
1902 	 * last_siginfo, so that we are sure to have synchronized with
1903 	 * any signal-sending on another CPU that wants to examine it.
1904 	 */
1905 	spin_lock_irq(&current->sighand->siglock);
1906 	current->last_siginfo = NULL;
1907 
1908 	/* LISTENING can be set only during STOP traps, clear it */
1909 	current->jobctl &= ~JOBCTL_LISTENING;
1910 
1911 	/*
1912 	 * Queued signals ignored us while we were stopped for tracing.
1913 	 * So check for any that we should take before resuming user mode.
1914 	 * This sets TIF_SIGPENDING, but never clears it.
1915 	 */
1916 	recalc_sigpending_tsk(current);
1917 }
1918 
1919 static void ptrace_do_notify(int signr, int exit_code, int why)
1920 {
1921 	siginfo_t info;
1922 
1923 	memset(&info, 0, sizeof info);
1924 	info.si_signo = signr;
1925 	info.si_code = exit_code;
1926 	info.si_pid = task_pid_vnr(current);
1927 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1928 
1929 	/* Let the debugger run.  */
1930 	ptrace_stop(exit_code, why, 1, &info);
1931 }
1932 
1933 void ptrace_notify(int exit_code)
1934 {
1935 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
1936 	if (unlikely(current->task_works))
1937 		task_work_run();
1938 
1939 	spin_lock_irq(&current->sighand->siglock);
1940 	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
1941 	spin_unlock_irq(&current->sighand->siglock);
1942 }
1943 
1944 /**
1945  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
1946  * @signr: signr causing group stop if initiating
1947  *
1948  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
1949  * and participate in it.  If already set, participate in the existing
1950  * group stop.  If participated in a group stop (and thus slept), %true is
1951  * returned with siglock released.
1952  *
1953  * If ptraced, this function doesn't handle stop itself.  Instead,
1954  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
1955  * untouched.  The caller must ensure that INTERRUPT trap handling takes
1956  * places afterwards.
1957  *
1958  * CONTEXT:
1959  * Must be called with @current->sighand->siglock held, which is released
1960  * on %true return.
1961  *
1962  * RETURNS:
1963  * %false if group stop is already cancelled or ptrace trap is scheduled.
1964  * %true if participated in group stop.
1965  */
1966 static bool do_signal_stop(int signr)
1967 	__releases(&current->sighand->siglock)
1968 {
1969 	struct signal_struct *sig = current->signal;
1970 
1971 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
1972 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
1973 		struct task_struct *t;
1974 
1975 		/* signr will be recorded in task->jobctl for retries */
1976 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
1977 
1978 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
1979 		    unlikely(signal_group_exit(sig)))
1980 			return false;
1981 		/*
1982 		 * There is no group stop already in progress.  We must
1983 		 * initiate one now.
1984 		 *
1985 		 * While ptraced, a task may be resumed while group stop is
1986 		 * still in effect and then receive a stop signal and
1987 		 * initiate another group stop.  This deviates from the
1988 		 * usual behavior as two consecutive stop signals can't
1989 		 * cause two group stops when !ptraced.  That is why we
1990 		 * also check !task_is_stopped(t) below.
1991 		 *
1992 		 * The condition can be distinguished by testing whether
1993 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
1994 		 * group_exit_code in such case.
1995 		 *
1996 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
1997 		 * an intervening stop signal is required to cause two
1998 		 * continued events regardless of ptrace.
1999 		 */
2000 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2001 			sig->group_exit_code = signr;
2002 
2003 		sig->group_stop_count = 0;
2004 
2005 		if (task_set_jobctl_pending(current, signr | gstop))
2006 			sig->group_stop_count++;
2007 
2008 		t = current;
2009 		while_each_thread(current, t) {
2010 			/*
2011 			 * Setting state to TASK_STOPPED for a group
2012 			 * stop is always done with the siglock held,
2013 			 * so this check has no races.
2014 			 */
2015 			if (!task_is_stopped(t) &&
2016 			    task_set_jobctl_pending(t, signr | gstop)) {
2017 				sig->group_stop_count++;
2018 				if (likely(!(t->ptrace & PT_SEIZED)))
2019 					signal_wake_up(t, 0);
2020 				else
2021 					ptrace_trap_notify(t);
2022 			}
2023 		}
2024 	}
2025 
2026 	if (likely(!current->ptrace)) {
2027 		int notify = 0;
2028 
2029 		/*
2030 		 * If there are no other threads in the group, or if there
2031 		 * is a group stop in progress and we are the last to stop,
2032 		 * report to the parent.
2033 		 */
2034 		if (task_participate_group_stop(current))
2035 			notify = CLD_STOPPED;
2036 
2037 		__set_current_state(TASK_STOPPED);
2038 		spin_unlock_irq(&current->sighand->siglock);
2039 
2040 		/*
2041 		 * Notify the parent of the group stop completion.  Because
2042 		 * we're not holding either the siglock or tasklist_lock
2043 		 * here, ptracer may attach inbetween; however, this is for
2044 		 * group stop and should always be delivered to the real
2045 		 * parent of the group leader.  The new ptracer will get
2046 		 * its notification when this task transitions into
2047 		 * TASK_TRACED.
2048 		 */
2049 		if (notify) {
2050 			read_lock(&tasklist_lock);
2051 			do_notify_parent_cldstop(current, false, notify);
2052 			read_unlock(&tasklist_lock);
2053 		}
2054 
2055 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2056 		freezable_schedule();
2057 		return true;
2058 	} else {
2059 		/*
2060 		 * While ptraced, group stop is handled by STOP trap.
2061 		 * Schedule it and let the caller deal with it.
2062 		 */
2063 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2064 		return false;
2065 	}
2066 }
2067 
2068 /**
2069  * do_jobctl_trap - take care of ptrace jobctl traps
2070  *
2071  * When PT_SEIZED, it's used for both group stop and explicit
2072  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2073  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2074  * the stop signal; otherwise, %SIGTRAP.
2075  *
2076  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2077  * number as exit_code and no siginfo.
2078  *
2079  * CONTEXT:
2080  * Must be called with @current->sighand->siglock held, which may be
2081  * released and re-acquired before returning with intervening sleep.
2082  */
2083 static void do_jobctl_trap(void)
2084 {
2085 	struct signal_struct *signal = current->signal;
2086 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2087 
2088 	if (current->ptrace & PT_SEIZED) {
2089 		if (!signal->group_stop_count &&
2090 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2091 			signr = SIGTRAP;
2092 		WARN_ON_ONCE(!signr);
2093 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2094 				 CLD_STOPPED);
2095 	} else {
2096 		WARN_ON_ONCE(!signr);
2097 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2098 		current->exit_code = 0;
2099 	}
2100 }
2101 
2102 static int ptrace_signal(int signr, siginfo_t *info)
2103 {
2104 	/*
2105 	 * We do not check sig_kernel_stop(signr) but set this marker
2106 	 * unconditionally because we do not know whether debugger will
2107 	 * change signr. This flag has no meaning unless we are going
2108 	 * to stop after return from ptrace_stop(). In this case it will
2109 	 * be checked in do_signal_stop(), we should only stop if it was
2110 	 * not cleared by SIGCONT while we were sleeping. See also the
2111 	 * comment in dequeue_signal().
2112 	 */
2113 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2114 	ptrace_stop(signr, CLD_TRAPPED, 0, info);
2115 
2116 	/* We're back.  Did the debugger cancel the sig?  */
2117 	signr = current->exit_code;
2118 	if (signr == 0)
2119 		return signr;
2120 
2121 	current->exit_code = 0;
2122 
2123 	/*
2124 	 * Update the siginfo structure if the signal has
2125 	 * changed.  If the debugger wanted something
2126 	 * specific in the siginfo structure then it should
2127 	 * have updated *info via PTRACE_SETSIGINFO.
2128 	 */
2129 	if (signr != info->si_signo) {
2130 		info->si_signo = signr;
2131 		info->si_errno = 0;
2132 		info->si_code = SI_USER;
2133 		rcu_read_lock();
2134 		info->si_pid = task_pid_vnr(current->parent);
2135 		info->si_uid = from_kuid_munged(current_user_ns(),
2136 						task_uid(current->parent));
2137 		rcu_read_unlock();
2138 	}
2139 
2140 	/* If the (new) signal is now blocked, requeue it.  */
2141 	if (sigismember(&current->blocked, signr)) {
2142 		specific_send_sig_info(signr, info, current);
2143 		signr = 0;
2144 	}
2145 
2146 	return signr;
2147 }
2148 
2149 int get_signal(struct ksignal *ksig)
2150 {
2151 	struct sighand_struct *sighand = current->sighand;
2152 	struct signal_struct *signal = current->signal;
2153 	int signr;
2154 
2155 	if (unlikely(current->task_works))
2156 		task_work_run();
2157 
2158 	if (unlikely(uprobe_deny_signal()))
2159 		return 0;
2160 
2161 	/*
2162 	 * Do this once, we can't return to user-mode if freezing() == T.
2163 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2164 	 * thus do not need another check after return.
2165 	 */
2166 	try_to_freeze();
2167 
2168 relock:
2169 	spin_lock_irq(&sighand->siglock);
2170 	/*
2171 	 * Every stopped thread goes here after wakeup. Check to see if
2172 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2173 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2174 	 */
2175 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2176 		int why;
2177 
2178 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2179 			why = CLD_CONTINUED;
2180 		else
2181 			why = CLD_STOPPED;
2182 
2183 		signal->flags &= ~SIGNAL_CLD_MASK;
2184 
2185 		spin_unlock_irq(&sighand->siglock);
2186 
2187 		/*
2188 		 * Notify the parent that we're continuing.  This event is
2189 		 * always per-process and doesn't make whole lot of sense
2190 		 * for ptracers, who shouldn't consume the state via
2191 		 * wait(2) either, but, for backward compatibility, notify
2192 		 * the ptracer of the group leader too unless it's gonna be
2193 		 * a duplicate.
2194 		 */
2195 		read_lock(&tasklist_lock);
2196 		do_notify_parent_cldstop(current, false, why);
2197 
2198 		if (ptrace_reparented(current->group_leader))
2199 			do_notify_parent_cldstop(current->group_leader,
2200 						true, why);
2201 		read_unlock(&tasklist_lock);
2202 
2203 		goto relock;
2204 	}
2205 
2206 	for (;;) {
2207 		struct k_sigaction *ka;
2208 
2209 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2210 		    do_signal_stop(0))
2211 			goto relock;
2212 
2213 		if (unlikely(current->jobctl & JOBCTL_TRAP_MASK)) {
2214 			do_jobctl_trap();
2215 			spin_unlock_irq(&sighand->siglock);
2216 			goto relock;
2217 		}
2218 
2219 		signr = dequeue_signal(current, &current->blocked, &ksig->info);
2220 
2221 		if (!signr)
2222 			break; /* will return 0 */
2223 
2224 		if (unlikely(current->ptrace) && signr != SIGKILL) {
2225 			signr = ptrace_signal(signr, &ksig->info);
2226 			if (!signr)
2227 				continue;
2228 		}
2229 
2230 		ka = &sighand->action[signr-1];
2231 
2232 		/* Trace actually delivered signals. */
2233 		trace_signal_deliver(signr, &ksig->info, ka);
2234 
2235 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2236 			continue;
2237 		if (ka->sa.sa_handler != SIG_DFL) {
2238 			/* Run the handler.  */
2239 			ksig->ka = *ka;
2240 
2241 			if (ka->sa.sa_flags & SA_ONESHOT)
2242 				ka->sa.sa_handler = SIG_DFL;
2243 
2244 			break; /* will return non-zero "signr" value */
2245 		}
2246 
2247 		/*
2248 		 * Now we are doing the default action for this signal.
2249 		 */
2250 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2251 			continue;
2252 
2253 		/*
2254 		 * Global init gets no signals it doesn't want.
2255 		 * Container-init gets no signals it doesn't want from same
2256 		 * container.
2257 		 *
2258 		 * Note that if global/container-init sees a sig_kernel_only()
2259 		 * signal here, the signal must have been generated internally
2260 		 * or must have come from an ancestor namespace. In either
2261 		 * case, the signal cannot be dropped.
2262 		 */
2263 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2264 				!sig_kernel_only(signr))
2265 			continue;
2266 
2267 		if (sig_kernel_stop(signr)) {
2268 			/*
2269 			 * The default action is to stop all threads in
2270 			 * the thread group.  The job control signals
2271 			 * do nothing in an orphaned pgrp, but SIGSTOP
2272 			 * always works.  Note that siglock needs to be
2273 			 * dropped during the call to is_orphaned_pgrp()
2274 			 * because of lock ordering with tasklist_lock.
2275 			 * This allows an intervening SIGCONT to be posted.
2276 			 * We need to check for that and bail out if necessary.
2277 			 */
2278 			if (signr != SIGSTOP) {
2279 				spin_unlock_irq(&sighand->siglock);
2280 
2281 				/* signals can be posted during this window */
2282 
2283 				if (is_current_pgrp_orphaned())
2284 					goto relock;
2285 
2286 				spin_lock_irq(&sighand->siglock);
2287 			}
2288 
2289 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2290 				/* It released the siglock.  */
2291 				goto relock;
2292 			}
2293 
2294 			/*
2295 			 * We didn't actually stop, due to a race
2296 			 * with SIGCONT or something like that.
2297 			 */
2298 			continue;
2299 		}
2300 
2301 		spin_unlock_irq(&sighand->siglock);
2302 
2303 		/*
2304 		 * Anything else is fatal, maybe with a core dump.
2305 		 */
2306 		current->flags |= PF_SIGNALED;
2307 
2308 		if (sig_kernel_coredump(signr)) {
2309 			if (print_fatal_signals)
2310 				print_fatal_signal(ksig->info.si_signo);
2311 			proc_coredump_connector(current);
2312 			/*
2313 			 * If it was able to dump core, this kills all
2314 			 * other threads in the group and synchronizes with
2315 			 * their demise.  If we lost the race with another
2316 			 * thread getting here, it set group_exit_code
2317 			 * first and our do_group_exit call below will use
2318 			 * that value and ignore the one we pass it.
2319 			 */
2320 			do_coredump(&ksig->info);
2321 		}
2322 
2323 		/*
2324 		 * Death signals, no core dump.
2325 		 */
2326 		do_group_exit(ksig->info.si_signo);
2327 		/* NOTREACHED */
2328 	}
2329 	spin_unlock_irq(&sighand->siglock);
2330 
2331 	ksig->sig = signr;
2332 	return ksig->sig > 0;
2333 }
2334 
2335 /**
2336  * signal_delivered -
2337  * @ksig:		kernel signal struct
2338  * @stepping:		nonzero if debugger single-step or block-step in use
2339  *
2340  * This function should be called when a signal has successfully been
2341  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2342  * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2343  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2344  */
2345 static void signal_delivered(struct ksignal *ksig, int stepping)
2346 {
2347 	sigset_t blocked;
2348 
2349 	/* A signal was successfully delivered, and the
2350 	   saved sigmask was stored on the signal frame,
2351 	   and will be restored by sigreturn.  So we can
2352 	   simply clear the restore sigmask flag.  */
2353 	clear_restore_sigmask();
2354 
2355 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2356 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2357 		sigaddset(&blocked, ksig->sig);
2358 	set_current_blocked(&blocked);
2359 	tracehook_signal_handler(stepping);
2360 }
2361 
2362 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2363 {
2364 	if (failed)
2365 		force_sigsegv(ksig->sig, current);
2366 	else
2367 		signal_delivered(ksig, stepping);
2368 }
2369 
2370 /*
2371  * It could be that complete_signal() picked us to notify about the
2372  * group-wide signal. Other threads should be notified now to take
2373  * the shared signals in @which since we will not.
2374  */
2375 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2376 {
2377 	sigset_t retarget;
2378 	struct task_struct *t;
2379 
2380 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2381 	if (sigisemptyset(&retarget))
2382 		return;
2383 
2384 	t = tsk;
2385 	while_each_thread(tsk, t) {
2386 		if (t->flags & PF_EXITING)
2387 			continue;
2388 
2389 		if (!has_pending_signals(&retarget, &t->blocked))
2390 			continue;
2391 		/* Remove the signals this thread can handle. */
2392 		sigandsets(&retarget, &retarget, &t->blocked);
2393 
2394 		if (!signal_pending(t))
2395 			signal_wake_up(t, 0);
2396 
2397 		if (sigisemptyset(&retarget))
2398 			break;
2399 	}
2400 }
2401 
2402 void exit_signals(struct task_struct *tsk)
2403 {
2404 	int group_stop = 0;
2405 	sigset_t unblocked;
2406 
2407 	/*
2408 	 * @tsk is about to have PF_EXITING set - lock out users which
2409 	 * expect stable threadgroup.
2410 	 */
2411 	cgroup_threadgroup_change_begin(tsk);
2412 
2413 	if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2414 		tsk->flags |= PF_EXITING;
2415 		cgroup_threadgroup_change_end(tsk);
2416 		return;
2417 	}
2418 
2419 	spin_lock_irq(&tsk->sighand->siglock);
2420 	/*
2421 	 * From now this task is not visible for group-wide signals,
2422 	 * see wants_signal(), do_signal_stop().
2423 	 */
2424 	tsk->flags |= PF_EXITING;
2425 
2426 	cgroup_threadgroup_change_end(tsk);
2427 
2428 	if (!signal_pending(tsk))
2429 		goto out;
2430 
2431 	unblocked = tsk->blocked;
2432 	signotset(&unblocked);
2433 	retarget_shared_pending(tsk, &unblocked);
2434 
2435 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2436 	    task_participate_group_stop(tsk))
2437 		group_stop = CLD_STOPPED;
2438 out:
2439 	spin_unlock_irq(&tsk->sighand->siglock);
2440 
2441 	/*
2442 	 * If group stop has completed, deliver the notification.  This
2443 	 * should always go to the real parent of the group leader.
2444 	 */
2445 	if (unlikely(group_stop)) {
2446 		read_lock(&tasklist_lock);
2447 		do_notify_parent_cldstop(tsk, false, group_stop);
2448 		read_unlock(&tasklist_lock);
2449 	}
2450 }
2451 
2452 EXPORT_SYMBOL(recalc_sigpending);
2453 EXPORT_SYMBOL_GPL(dequeue_signal);
2454 EXPORT_SYMBOL(flush_signals);
2455 EXPORT_SYMBOL(force_sig);
2456 EXPORT_SYMBOL(send_sig);
2457 EXPORT_SYMBOL(send_sig_info);
2458 EXPORT_SYMBOL(sigprocmask);
2459 
2460 /*
2461  * System call entry points.
2462  */
2463 
2464 /**
2465  *  sys_restart_syscall - restart a system call
2466  */
2467 SYSCALL_DEFINE0(restart_syscall)
2468 {
2469 	struct restart_block *restart = &current->restart_block;
2470 	return restart->fn(restart);
2471 }
2472 
2473 long do_no_restart_syscall(struct restart_block *param)
2474 {
2475 	return -EINTR;
2476 }
2477 
2478 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2479 {
2480 	if (signal_pending(tsk) && !thread_group_empty(tsk)) {
2481 		sigset_t newblocked;
2482 		/* A set of now blocked but previously unblocked signals. */
2483 		sigandnsets(&newblocked, newset, &current->blocked);
2484 		retarget_shared_pending(tsk, &newblocked);
2485 	}
2486 	tsk->blocked = *newset;
2487 	recalc_sigpending();
2488 }
2489 
2490 /**
2491  * set_current_blocked - change current->blocked mask
2492  * @newset: new mask
2493  *
2494  * It is wrong to change ->blocked directly, this helper should be used
2495  * to ensure the process can't miss a shared signal we are going to block.
2496  */
2497 void set_current_blocked(sigset_t *newset)
2498 {
2499 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2500 	__set_current_blocked(newset);
2501 }
2502 
2503 void __set_current_blocked(const sigset_t *newset)
2504 {
2505 	struct task_struct *tsk = current;
2506 
2507 	/*
2508 	 * In case the signal mask hasn't changed, there is nothing we need
2509 	 * to do. The current->blocked shouldn't be modified by other task.
2510 	 */
2511 	if (sigequalsets(&tsk->blocked, newset))
2512 		return;
2513 
2514 	spin_lock_irq(&tsk->sighand->siglock);
2515 	__set_task_blocked(tsk, newset);
2516 	spin_unlock_irq(&tsk->sighand->siglock);
2517 }
2518 
2519 /*
2520  * This is also useful for kernel threads that want to temporarily
2521  * (or permanently) block certain signals.
2522  *
2523  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2524  * interface happily blocks "unblockable" signals like SIGKILL
2525  * and friends.
2526  */
2527 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2528 {
2529 	struct task_struct *tsk = current;
2530 	sigset_t newset;
2531 
2532 	/* Lockless, only current can change ->blocked, never from irq */
2533 	if (oldset)
2534 		*oldset = tsk->blocked;
2535 
2536 	switch (how) {
2537 	case SIG_BLOCK:
2538 		sigorsets(&newset, &tsk->blocked, set);
2539 		break;
2540 	case SIG_UNBLOCK:
2541 		sigandnsets(&newset, &tsk->blocked, set);
2542 		break;
2543 	case SIG_SETMASK:
2544 		newset = *set;
2545 		break;
2546 	default:
2547 		return -EINVAL;
2548 	}
2549 
2550 	__set_current_blocked(&newset);
2551 	return 0;
2552 }
2553 
2554 /**
2555  *  sys_rt_sigprocmask - change the list of currently blocked signals
2556  *  @how: whether to add, remove, or set signals
2557  *  @nset: stores pending signals
2558  *  @oset: previous value of signal mask if non-null
2559  *  @sigsetsize: size of sigset_t type
2560  */
2561 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
2562 		sigset_t __user *, oset, size_t, sigsetsize)
2563 {
2564 	sigset_t old_set, new_set;
2565 	int error;
2566 
2567 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2568 	if (sigsetsize != sizeof(sigset_t))
2569 		return -EINVAL;
2570 
2571 	old_set = current->blocked;
2572 
2573 	if (nset) {
2574 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
2575 			return -EFAULT;
2576 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2577 
2578 		error = sigprocmask(how, &new_set, NULL);
2579 		if (error)
2580 			return error;
2581 	}
2582 
2583 	if (oset) {
2584 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
2585 			return -EFAULT;
2586 	}
2587 
2588 	return 0;
2589 }
2590 
2591 #ifdef CONFIG_COMPAT
2592 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
2593 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
2594 {
2595 #ifdef __BIG_ENDIAN
2596 	sigset_t old_set = current->blocked;
2597 
2598 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2599 	if (sigsetsize != sizeof(sigset_t))
2600 		return -EINVAL;
2601 
2602 	if (nset) {
2603 		compat_sigset_t new32;
2604 		sigset_t new_set;
2605 		int error;
2606 		if (copy_from_user(&new32, nset, sizeof(compat_sigset_t)))
2607 			return -EFAULT;
2608 
2609 		sigset_from_compat(&new_set, &new32);
2610 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2611 
2612 		error = sigprocmask(how, &new_set, NULL);
2613 		if (error)
2614 			return error;
2615 	}
2616 	if (oset) {
2617 		compat_sigset_t old32;
2618 		sigset_to_compat(&old32, &old_set);
2619 		if (copy_to_user(oset, &old32, sizeof(compat_sigset_t)))
2620 			return -EFAULT;
2621 	}
2622 	return 0;
2623 #else
2624 	return sys_rt_sigprocmask(how, (sigset_t __user *)nset,
2625 				  (sigset_t __user *)oset, sigsetsize);
2626 #endif
2627 }
2628 #endif
2629 
2630 static int do_sigpending(void *set, unsigned long sigsetsize)
2631 {
2632 	if (sigsetsize > sizeof(sigset_t))
2633 		return -EINVAL;
2634 
2635 	spin_lock_irq(&current->sighand->siglock);
2636 	sigorsets(set, &current->pending.signal,
2637 		  &current->signal->shared_pending.signal);
2638 	spin_unlock_irq(&current->sighand->siglock);
2639 
2640 	/* Outside the lock because only this thread touches it.  */
2641 	sigandsets(set, &current->blocked, set);
2642 	return 0;
2643 }
2644 
2645 /**
2646  *  sys_rt_sigpending - examine a pending signal that has been raised
2647  *			while blocked
2648  *  @uset: stores pending signals
2649  *  @sigsetsize: size of sigset_t type or larger
2650  */
2651 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
2652 {
2653 	sigset_t set;
2654 	int err = do_sigpending(&set, sigsetsize);
2655 	if (!err && copy_to_user(uset, &set, sigsetsize))
2656 		err = -EFAULT;
2657 	return err;
2658 }
2659 
2660 #ifdef CONFIG_COMPAT
2661 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
2662 		compat_size_t, sigsetsize)
2663 {
2664 #ifdef __BIG_ENDIAN
2665 	sigset_t set;
2666 	int err = do_sigpending(&set, sigsetsize);
2667 	if (!err) {
2668 		compat_sigset_t set32;
2669 		sigset_to_compat(&set32, &set);
2670 		/* we can get here only if sigsetsize <= sizeof(set) */
2671 		if (copy_to_user(uset, &set32, sigsetsize))
2672 			err = -EFAULT;
2673 	}
2674 	return err;
2675 #else
2676 	return sys_rt_sigpending((sigset_t __user *)uset, sigsetsize);
2677 #endif
2678 }
2679 #endif
2680 
2681 #ifndef HAVE_ARCH_COPY_SIGINFO_TO_USER
2682 
2683 int copy_siginfo_to_user(siginfo_t __user *to, const siginfo_t *from)
2684 {
2685 	int err;
2686 
2687 	if (!access_ok (VERIFY_WRITE, to, sizeof(siginfo_t)))
2688 		return -EFAULT;
2689 	if (from->si_code < 0)
2690 		return __copy_to_user(to, from, sizeof(siginfo_t))
2691 			? -EFAULT : 0;
2692 	/*
2693 	 * If you change siginfo_t structure, please be sure
2694 	 * this code is fixed accordingly.
2695 	 * Please remember to update the signalfd_copyinfo() function
2696 	 * inside fs/signalfd.c too, in case siginfo_t changes.
2697 	 * It should never copy any pad contained in the structure
2698 	 * to avoid security leaks, but must copy the generic
2699 	 * 3 ints plus the relevant union member.
2700 	 */
2701 	err = __put_user(from->si_signo, &to->si_signo);
2702 	err |= __put_user(from->si_errno, &to->si_errno);
2703 	err |= __put_user((short)from->si_code, &to->si_code);
2704 	switch (from->si_code & __SI_MASK) {
2705 	case __SI_KILL:
2706 		err |= __put_user(from->si_pid, &to->si_pid);
2707 		err |= __put_user(from->si_uid, &to->si_uid);
2708 		break;
2709 	case __SI_TIMER:
2710 		 err |= __put_user(from->si_tid, &to->si_tid);
2711 		 err |= __put_user(from->si_overrun, &to->si_overrun);
2712 		 err |= __put_user(from->si_ptr, &to->si_ptr);
2713 		break;
2714 	case __SI_POLL:
2715 		err |= __put_user(from->si_band, &to->si_band);
2716 		err |= __put_user(from->si_fd, &to->si_fd);
2717 		break;
2718 	case __SI_FAULT:
2719 		err |= __put_user(from->si_addr, &to->si_addr);
2720 #ifdef __ARCH_SI_TRAPNO
2721 		err |= __put_user(from->si_trapno, &to->si_trapno);
2722 #endif
2723 #ifdef BUS_MCEERR_AO
2724 		/*
2725 		 * Other callers might not initialize the si_lsb field,
2726 		 * so check explicitly for the right codes here.
2727 		 */
2728 		if (from->si_signo == SIGBUS &&
2729 		    (from->si_code == BUS_MCEERR_AR || from->si_code == BUS_MCEERR_AO))
2730 			err |= __put_user(from->si_addr_lsb, &to->si_addr_lsb);
2731 #endif
2732 #ifdef SEGV_BNDERR
2733 		if (from->si_signo == SIGSEGV && from->si_code == SEGV_BNDERR) {
2734 			err |= __put_user(from->si_lower, &to->si_lower);
2735 			err |= __put_user(from->si_upper, &to->si_upper);
2736 		}
2737 #endif
2738 #ifdef SEGV_PKUERR
2739 		if (from->si_signo == SIGSEGV && from->si_code == SEGV_PKUERR)
2740 			err |= __put_user(from->si_pkey, &to->si_pkey);
2741 #endif
2742 		break;
2743 	case __SI_CHLD:
2744 		err |= __put_user(from->si_pid, &to->si_pid);
2745 		err |= __put_user(from->si_uid, &to->si_uid);
2746 		err |= __put_user(from->si_status, &to->si_status);
2747 		err |= __put_user(from->si_utime, &to->si_utime);
2748 		err |= __put_user(from->si_stime, &to->si_stime);
2749 		break;
2750 	case __SI_RT: /* This is not generated by the kernel as of now. */
2751 	case __SI_MESGQ: /* But this is */
2752 		err |= __put_user(from->si_pid, &to->si_pid);
2753 		err |= __put_user(from->si_uid, &to->si_uid);
2754 		err |= __put_user(from->si_ptr, &to->si_ptr);
2755 		break;
2756 #ifdef __ARCH_SIGSYS
2757 	case __SI_SYS:
2758 		err |= __put_user(from->si_call_addr, &to->si_call_addr);
2759 		err |= __put_user(from->si_syscall, &to->si_syscall);
2760 		err |= __put_user(from->si_arch, &to->si_arch);
2761 		break;
2762 #endif
2763 	default: /* this is just in case for now ... */
2764 		err |= __put_user(from->si_pid, &to->si_pid);
2765 		err |= __put_user(from->si_uid, &to->si_uid);
2766 		break;
2767 	}
2768 	return err;
2769 }
2770 
2771 #endif
2772 
2773 /**
2774  *  do_sigtimedwait - wait for queued signals specified in @which
2775  *  @which: queued signals to wait for
2776  *  @info: if non-null, the signal's siginfo is returned here
2777  *  @ts: upper bound on process time suspension
2778  */
2779 int do_sigtimedwait(const sigset_t *which, siginfo_t *info,
2780 		    const struct timespec *ts)
2781 {
2782 	ktime_t *to = NULL, timeout = KTIME_MAX;
2783 	struct task_struct *tsk = current;
2784 	sigset_t mask = *which;
2785 	int sig, ret = 0;
2786 
2787 	if (ts) {
2788 		if (!timespec_valid(ts))
2789 			return -EINVAL;
2790 		timeout = timespec_to_ktime(*ts);
2791 		to = &timeout;
2792 	}
2793 
2794 	/*
2795 	 * Invert the set of allowed signals to get those we want to block.
2796 	 */
2797 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
2798 	signotset(&mask);
2799 
2800 	spin_lock_irq(&tsk->sighand->siglock);
2801 	sig = dequeue_signal(tsk, &mask, info);
2802 	if (!sig && timeout) {
2803 		/*
2804 		 * None ready, temporarily unblock those we're interested
2805 		 * while we are sleeping in so that we'll be awakened when
2806 		 * they arrive. Unblocking is always fine, we can avoid
2807 		 * set_current_blocked().
2808 		 */
2809 		tsk->real_blocked = tsk->blocked;
2810 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
2811 		recalc_sigpending();
2812 		spin_unlock_irq(&tsk->sighand->siglock);
2813 
2814 		__set_current_state(TASK_INTERRUPTIBLE);
2815 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
2816 							 HRTIMER_MODE_REL);
2817 		spin_lock_irq(&tsk->sighand->siglock);
2818 		__set_task_blocked(tsk, &tsk->real_blocked);
2819 		sigemptyset(&tsk->real_blocked);
2820 		sig = dequeue_signal(tsk, &mask, info);
2821 	}
2822 	spin_unlock_irq(&tsk->sighand->siglock);
2823 
2824 	if (sig)
2825 		return sig;
2826 	return ret ? -EINTR : -EAGAIN;
2827 }
2828 
2829 /**
2830  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
2831  *			in @uthese
2832  *  @uthese: queued signals to wait for
2833  *  @uinfo: if non-null, the signal's siginfo is returned here
2834  *  @uts: upper bound on process time suspension
2835  *  @sigsetsize: size of sigset_t type
2836  */
2837 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
2838 		siginfo_t __user *, uinfo, const struct timespec __user *, uts,
2839 		size_t, sigsetsize)
2840 {
2841 	sigset_t these;
2842 	struct timespec ts;
2843 	siginfo_t info;
2844 	int ret;
2845 
2846 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2847 	if (sigsetsize != sizeof(sigset_t))
2848 		return -EINVAL;
2849 
2850 	if (copy_from_user(&these, uthese, sizeof(these)))
2851 		return -EFAULT;
2852 
2853 	if (uts) {
2854 		if (copy_from_user(&ts, uts, sizeof(ts)))
2855 			return -EFAULT;
2856 	}
2857 
2858 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
2859 
2860 	if (ret > 0 && uinfo) {
2861 		if (copy_siginfo_to_user(uinfo, &info))
2862 			ret = -EFAULT;
2863 	}
2864 
2865 	return ret;
2866 }
2867 
2868 /**
2869  *  sys_kill - send a signal to a process
2870  *  @pid: the PID of the process
2871  *  @sig: signal to be sent
2872  */
2873 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
2874 {
2875 	struct siginfo info;
2876 
2877 	info.si_signo = sig;
2878 	info.si_errno = 0;
2879 	info.si_code = SI_USER;
2880 	info.si_pid = task_tgid_vnr(current);
2881 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2882 
2883 	return kill_something_info(sig, &info, pid);
2884 }
2885 
2886 static int
2887 do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info)
2888 {
2889 	struct task_struct *p;
2890 	int error = -ESRCH;
2891 
2892 	rcu_read_lock();
2893 	p = find_task_by_vpid(pid);
2894 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
2895 		error = check_kill_permission(sig, info, p);
2896 		/*
2897 		 * The null signal is a permissions and process existence
2898 		 * probe.  No signal is actually delivered.
2899 		 */
2900 		if (!error && sig) {
2901 			error = do_send_sig_info(sig, info, p, false);
2902 			/*
2903 			 * If lock_task_sighand() failed we pretend the task
2904 			 * dies after receiving the signal. The window is tiny,
2905 			 * and the signal is private anyway.
2906 			 */
2907 			if (unlikely(error == -ESRCH))
2908 				error = 0;
2909 		}
2910 	}
2911 	rcu_read_unlock();
2912 
2913 	return error;
2914 }
2915 
2916 static int do_tkill(pid_t tgid, pid_t pid, int sig)
2917 {
2918 	struct siginfo info = {};
2919 
2920 	info.si_signo = sig;
2921 	info.si_errno = 0;
2922 	info.si_code = SI_TKILL;
2923 	info.si_pid = task_tgid_vnr(current);
2924 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2925 
2926 	return do_send_specific(tgid, pid, sig, &info);
2927 }
2928 
2929 /**
2930  *  sys_tgkill - send signal to one specific thread
2931  *  @tgid: the thread group ID of the thread
2932  *  @pid: the PID of the thread
2933  *  @sig: signal to be sent
2934  *
2935  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
2936  *  exists but it's not belonging to the target process anymore. This
2937  *  method solves the problem of threads exiting and PIDs getting reused.
2938  */
2939 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
2940 {
2941 	/* This is only valid for single tasks */
2942 	if (pid <= 0 || tgid <= 0)
2943 		return -EINVAL;
2944 
2945 	return do_tkill(tgid, pid, sig);
2946 }
2947 
2948 /**
2949  *  sys_tkill - send signal to one specific task
2950  *  @pid: the PID of the task
2951  *  @sig: signal to be sent
2952  *
2953  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
2954  */
2955 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
2956 {
2957 	/* This is only valid for single tasks */
2958 	if (pid <= 0)
2959 		return -EINVAL;
2960 
2961 	return do_tkill(0, pid, sig);
2962 }
2963 
2964 static int do_rt_sigqueueinfo(pid_t pid, int sig, siginfo_t *info)
2965 {
2966 	/* Not even root can pretend to send signals from the kernel.
2967 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
2968 	 */
2969 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
2970 	    (task_pid_vnr(current) != pid))
2971 		return -EPERM;
2972 
2973 	info->si_signo = sig;
2974 
2975 	/* POSIX.1b doesn't mention process groups.  */
2976 	return kill_proc_info(sig, info, pid);
2977 }
2978 
2979 /**
2980  *  sys_rt_sigqueueinfo - send signal information to a signal
2981  *  @pid: the PID of the thread
2982  *  @sig: signal to be sent
2983  *  @uinfo: signal info to be sent
2984  */
2985 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
2986 		siginfo_t __user *, uinfo)
2987 {
2988 	siginfo_t info;
2989 	if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
2990 		return -EFAULT;
2991 	return do_rt_sigqueueinfo(pid, sig, &info);
2992 }
2993 
2994 #ifdef CONFIG_COMPAT
2995 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
2996 			compat_pid_t, pid,
2997 			int, sig,
2998 			struct compat_siginfo __user *, uinfo)
2999 {
3000 	siginfo_t info = {};
3001 	int ret = copy_siginfo_from_user32(&info, uinfo);
3002 	if (unlikely(ret))
3003 		return ret;
3004 	return do_rt_sigqueueinfo(pid, sig, &info);
3005 }
3006 #endif
3007 
3008 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, siginfo_t *info)
3009 {
3010 	/* This is only valid for single tasks */
3011 	if (pid <= 0 || tgid <= 0)
3012 		return -EINVAL;
3013 
3014 	/* Not even root can pretend to send signals from the kernel.
3015 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3016 	 */
3017 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3018 	    (task_pid_vnr(current) != pid))
3019 		return -EPERM;
3020 
3021 	info->si_signo = sig;
3022 
3023 	return do_send_specific(tgid, pid, sig, info);
3024 }
3025 
3026 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
3027 		siginfo_t __user *, uinfo)
3028 {
3029 	siginfo_t info;
3030 
3031 	if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
3032 		return -EFAULT;
3033 
3034 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3035 }
3036 
3037 #ifdef CONFIG_COMPAT
3038 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
3039 			compat_pid_t, tgid,
3040 			compat_pid_t, pid,
3041 			int, sig,
3042 			struct compat_siginfo __user *, uinfo)
3043 {
3044 	siginfo_t info = {};
3045 
3046 	if (copy_siginfo_from_user32(&info, uinfo))
3047 		return -EFAULT;
3048 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3049 }
3050 #endif
3051 
3052 /*
3053  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
3054  */
3055 void kernel_sigaction(int sig, __sighandler_t action)
3056 {
3057 	spin_lock_irq(&current->sighand->siglock);
3058 	current->sighand->action[sig - 1].sa.sa_handler = action;
3059 	if (action == SIG_IGN) {
3060 		sigset_t mask;
3061 
3062 		sigemptyset(&mask);
3063 		sigaddset(&mask, sig);
3064 
3065 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
3066 		flush_sigqueue_mask(&mask, &current->pending);
3067 		recalc_sigpending();
3068 	}
3069 	spin_unlock_irq(&current->sighand->siglock);
3070 }
3071 EXPORT_SYMBOL(kernel_sigaction);
3072 
3073 void __weak sigaction_compat_abi(struct k_sigaction *act,
3074 		struct k_sigaction *oact)
3075 {
3076 }
3077 
3078 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
3079 {
3080 	struct task_struct *p = current, *t;
3081 	struct k_sigaction *k;
3082 	sigset_t mask;
3083 
3084 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
3085 		return -EINVAL;
3086 
3087 	k = &p->sighand->action[sig-1];
3088 
3089 	spin_lock_irq(&p->sighand->siglock);
3090 	if (oact)
3091 		*oact = *k;
3092 
3093 	sigaction_compat_abi(act, oact);
3094 
3095 	if (act) {
3096 		sigdelsetmask(&act->sa.sa_mask,
3097 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
3098 		*k = *act;
3099 		/*
3100 		 * POSIX 3.3.1.3:
3101 		 *  "Setting a signal action to SIG_IGN for a signal that is
3102 		 *   pending shall cause the pending signal to be discarded,
3103 		 *   whether or not it is blocked."
3104 		 *
3105 		 *  "Setting a signal action to SIG_DFL for a signal that is
3106 		 *   pending and whose default action is to ignore the signal
3107 		 *   (for example, SIGCHLD), shall cause the pending signal to
3108 		 *   be discarded, whether or not it is blocked"
3109 		 */
3110 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
3111 			sigemptyset(&mask);
3112 			sigaddset(&mask, sig);
3113 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
3114 			for_each_thread(p, t)
3115 				flush_sigqueue_mask(&mask, &t->pending);
3116 		}
3117 	}
3118 
3119 	spin_unlock_irq(&p->sighand->siglock);
3120 	return 0;
3121 }
3122 
3123 static int
3124 do_sigaltstack (const stack_t __user *uss, stack_t __user *uoss, unsigned long sp)
3125 {
3126 	stack_t oss;
3127 	int error;
3128 
3129 	oss.ss_sp = (void __user *) current->sas_ss_sp;
3130 	oss.ss_size = current->sas_ss_size;
3131 	oss.ss_flags = sas_ss_flags(sp) |
3132 		(current->sas_ss_flags & SS_FLAG_BITS);
3133 
3134 	if (uss) {
3135 		void __user *ss_sp;
3136 		size_t ss_size;
3137 		unsigned ss_flags;
3138 		int ss_mode;
3139 
3140 		error = -EFAULT;
3141 		if (!access_ok(VERIFY_READ, uss, sizeof(*uss)))
3142 			goto out;
3143 		error = __get_user(ss_sp, &uss->ss_sp) |
3144 			__get_user(ss_flags, &uss->ss_flags) |
3145 			__get_user(ss_size, &uss->ss_size);
3146 		if (error)
3147 			goto out;
3148 
3149 		error = -EPERM;
3150 		if (on_sig_stack(sp))
3151 			goto out;
3152 
3153 		ss_mode = ss_flags & ~SS_FLAG_BITS;
3154 		error = -EINVAL;
3155 		if (ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
3156 				ss_mode != 0)
3157 			goto out;
3158 
3159 		if (ss_mode == SS_DISABLE) {
3160 			ss_size = 0;
3161 			ss_sp = NULL;
3162 		} else {
3163 			error = -ENOMEM;
3164 			if (ss_size < MINSIGSTKSZ)
3165 				goto out;
3166 		}
3167 
3168 		current->sas_ss_sp = (unsigned long) ss_sp;
3169 		current->sas_ss_size = ss_size;
3170 		current->sas_ss_flags = ss_flags;
3171 	}
3172 
3173 	error = 0;
3174 	if (uoss) {
3175 		error = -EFAULT;
3176 		if (!access_ok(VERIFY_WRITE, uoss, sizeof(*uoss)))
3177 			goto out;
3178 		error = __put_user(oss.ss_sp, &uoss->ss_sp) |
3179 			__put_user(oss.ss_size, &uoss->ss_size) |
3180 			__put_user(oss.ss_flags, &uoss->ss_flags);
3181 	}
3182 
3183 out:
3184 	return error;
3185 }
3186 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
3187 {
3188 	return do_sigaltstack(uss, uoss, current_user_stack_pointer());
3189 }
3190 
3191 int restore_altstack(const stack_t __user *uss)
3192 {
3193 	int err = do_sigaltstack(uss, NULL, current_user_stack_pointer());
3194 	/* squash all but EFAULT for now */
3195 	return err == -EFAULT ? err : 0;
3196 }
3197 
3198 int __save_altstack(stack_t __user *uss, unsigned long sp)
3199 {
3200 	struct task_struct *t = current;
3201 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
3202 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
3203 		__put_user(t->sas_ss_size, &uss->ss_size);
3204 	if (err)
3205 		return err;
3206 	if (t->sas_ss_flags & SS_AUTODISARM)
3207 		sas_ss_reset(t);
3208 	return 0;
3209 }
3210 
3211 #ifdef CONFIG_COMPAT
3212 COMPAT_SYSCALL_DEFINE2(sigaltstack,
3213 			const compat_stack_t __user *, uss_ptr,
3214 			compat_stack_t __user *, uoss_ptr)
3215 {
3216 	stack_t uss, uoss;
3217 	int ret;
3218 	mm_segment_t seg;
3219 
3220 	if (uss_ptr) {
3221 		compat_stack_t uss32;
3222 
3223 		memset(&uss, 0, sizeof(stack_t));
3224 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
3225 			return -EFAULT;
3226 		uss.ss_sp = compat_ptr(uss32.ss_sp);
3227 		uss.ss_flags = uss32.ss_flags;
3228 		uss.ss_size = uss32.ss_size;
3229 	}
3230 	seg = get_fs();
3231 	set_fs(KERNEL_DS);
3232 	ret = do_sigaltstack((stack_t __force __user *) (uss_ptr ? &uss : NULL),
3233 			     (stack_t __force __user *) &uoss,
3234 			     compat_user_stack_pointer());
3235 	set_fs(seg);
3236 	if (ret >= 0 && uoss_ptr)  {
3237 		if (!access_ok(VERIFY_WRITE, uoss_ptr, sizeof(compat_stack_t)) ||
3238 		    __put_user(ptr_to_compat(uoss.ss_sp), &uoss_ptr->ss_sp) ||
3239 		    __put_user(uoss.ss_flags, &uoss_ptr->ss_flags) ||
3240 		    __put_user(uoss.ss_size, &uoss_ptr->ss_size))
3241 			ret = -EFAULT;
3242 	}
3243 	return ret;
3244 }
3245 
3246 int compat_restore_altstack(const compat_stack_t __user *uss)
3247 {
3248 	int err = compat_sys_sigaltstack(uss, NULL);
3249 	/* squash all but -EFAULT for now */
3250 	return err == -EFAULT ? err : 0;
3251 }
3252 
3253 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
3254 {
3255 	int err;
3256 	struct task_struct *t = current;
3257 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
3258 			 &uss->ss_sp) |
3259 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
3260 		__put_user(t->sas_ss_size, &uss->ss_size);
3261 	if (err)
3262 		return err;
3263 	if (t->sas_ss_flags & SS_AUTODISARM)
3264 		sas_ss_reset(t);
3265 	return 0;
3266 }
3267 #endif
3268 
3269 #ifdef __ARCH_WANT_SYS_SIGPENDING
3270 
3271 /**
3272  *  sys_sigpending - examine pending signals
3273  *  @set: where mask of pending signal is returned
3274  */
3275 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, set)
3276 {
3277 	return sys_rt_sigpending((sigset_t __user *)set, sizeof(old_sigset_t));
3278 }
3279 
3280 #endif
3281 
3282 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
3283 /**
3284  *  sys_sigprocmask - examine and change blocked signals
3285  *  @how: whether to add, remove, or set signals
3286  *  @nset: signals to add or remove (if non-null)
3287  *  @oset: previous value of signal mask if non-null
3288  *
3289  * Some platforms have their own version with special arguments;
3290  * others support only sys_rt_sigprocmask.
3291  */
3292 
3293 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
3294 		old_sigset_t __user *, oset)
3295 {
3296 	old_sigset_t old_set, new_set;
3297 	sigset_t new_blocked;
3298 
3299 	old_set = current->blocked.sig[0];
3300 
3301 	if (nset) {
3302 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
3303 			return -EFAULT;
3304 
3305 		new_blocked = current->blocked;
3306 
3307 		switch (how) {
3308 		case SIG_BLOCK:
3309 			sigaddsetmask(&new_blocked, new_set);
3310 			break;
3311 		case SIG_UNBLOCK:
3312 			sigdelsetmask(&new_blocked, new_set);
3313 			break;
3314 		case SIG_SETMASK:
3315 			new_blocked.sig[0] = new_set;
3316 			break;
3317 		default:
3318 			return -EINVAL;
3319 		}
3320 
3321 		set_current_blocked(&new_blocked);
3322 	}
3323 
3324 	if (oset) {
3325 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
3326 			return -EFAULT;
3327 	}
3328 
3329 	return 0;
3330 }
3331 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
3332 
3333 #ifndef CONFIG_ODD_RT_SIGACTION
3334 /**
3335  *  sys_rt_sigaction - alter an action taken by a process
3336  *  @sig: signal to be sent
3337  *  @act: new sigaction
3338  *  @oact: used to save the previous sigaction
3339  *  @sigsetsize: size of sigset_t type
3340  */
3341 SYSCALL_DEFINE4(rt_sigaction, int, sig,
3342 		const struct sigaction __user *, act,
3343 		struct sigaction __user *, oact,
3344 		size_t, sigsetsize)
3345 {
3346 	struct k_sigaction new_sa, old_sa;
3347 	int ret = -EINVAL;
3348 
3349 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3350 	if (sigsetsize != sizeof(sigset_t))
3351 		goto out;
3352 
3353 	if (act) {
3354 		if (copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
3355 			return -EFAULT;
3356 	}
3357 
3358 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
3359 
3360 	if (!ret && oact) {
3361 		if (copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
3362 			return -EFAULT;
3363 	}
3364 out:
3365 	return ret;
3366 }
3367 #ifdef CONFIG_COMPAT
3368 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
3369 		const struct compat_sigaction __user *, act,
3370 		struct compat_sigaction __user *, oact,
3371 		compat_size_t, sigsetsize)
3372 {
3373 	struct k_sigaction new_ka, old_ka;
3374 	compat_sigset_t mask;
3375 #ifdef __ARCH_HAS_SA_RESTORER
3376 	compat_uptr_t restorer;
3377 #endif
3378 	int ret;
3379 
3380 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3381 	if (sigsetsize != sizeof(compat_sigset_t))
3382 		return -EINVAL;
3383 
3384 	if (act) {
3385 		compat_uptr_t handler;
3386 		ret = get_user(handler, &act->sa_handler);
3387 		new_ka.sa.sa_handler = compat_ptr(handler);
3388 #ifdef __ARCH_HAS_SA_RESTORER
3389 		ret |= get_user(restorer, &act->sa_restorer);
3390 		new_ka.sa.sa_restorer = compat_ptr(restorer);
3391 #endif
3392 		ret |= copy_from_user(&mask, &act->sa_mask, sizeof(mask));
3393 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
3394 		if (ret)
3395 			return -EFAULT;
3396 		sigset_from_compat(&new_ka.sa.sa_mask, &mask);
3397 	}
3398 
3399 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3400 	if (!ret && oact) {
3401 		sigset_to_compat(&mask, &old_ka.sa.sa_mask);
3402 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
3403 			       &oact->sa_handler);
3404 		ret |= copy_to_user(&oact->sa_mask, &mask, sizeof(mask));
3405 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
3406 #ifdef __ARCH_HAS_SA_RESTORER
3407 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
3408 				&oact->sa_restorer);
3409 #endif
3410 	}
3411 	return ret;
3412 }
3413 #endif
3414 #endif /* !CONFIG_ODD_RT_SIGACTION */
3415 
3416 #ifdef CONFIG_OLD_SIGACTION
3417 SYSCALL_DEFINE3(sigaction, int, sig,
3418 		const struct old_sigaction __user *, act,
3419 	        struct old_sigaction __user *, oact)
3420 {
3421 	struct k_sigaction new_ka, old_ka;
3422 	int ret;
3423 
3424 	if (act) {
3425 		old_sigset_t mask;
3426 		if (!access_ok(VERIFY_READ, act, sizeof(*act)) ||
3427 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
3428 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
3429 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
3430 		    __get_user(mask, &act->sa_mask))
3431 			return -EFAULT;
3432 #ifdef __ARCH_HAS_KA_RESTORER
3433 		new_ka.ka_restorer = NULL;
3434 #endif
3435 		siginitset(&new_ka.sa.sa_mask, mask);
3436 	}
3437 
3438 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3439 
3440 	if (!ret && oact) {
3441 		if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)) ||
3442 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
3443 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
3444 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
3445 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
3446 			return -EFAULT;
3447 	}
3448 
3449 	return ret;
3450 }
3451 #endif
3452 #ifdef CONFIG_COMPAT_OLD_SIGACTION
3453 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
3454 		const struct compat_old_sigaction __user *, act,
3455 	        struct compat_old_sigaction __user *, oact)
3456 {
3457 	struct k_sigaction new_ka, old_ka;
3458 	int ret;
3459 	compat_old_sigset_t mask;
3460 	compat_uptr_t handler, restorer;
3461 
3462 	if (act) {
3463 		if (!access_ok(VERIFY_READ, act, sizeof(*act)) ||
3464 		    __get_user(handler, &act->sa_handler) ||
3465 		    __get_user(restorer, &act->sa_restorer) ||
3466 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
3467 		    __get_user(mask, &act->sa_mask))
3468 			return -EFAULT;
3469 
3470 #ifdef __ARCH_HAS_KA_RESTORER
3471 		new_ka.ka_restorer = NULL;
3472 #endif
3473 		new_ka.sa.sa_handler = compat_ptr(handler);
3474 		new_ka.sa.sa_restorer = compat_ptr(restorer);
3475 		siginitset(&new_ka.sa.sa_mask, mask);
3476 	}
3477 
3478 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3479 
3480 	if (!ret && oact) {
3481 		if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)) ||
3482 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
3483 			       &oact->sa_handler) ||
3484 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
3485 			       &oact->sa_restorer) ||
3486 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
3487 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
3488 			return -EFAULT;
3489 	}
3490 	return ret;
3491 }
3492 #endif
3493 
3494 #ifdef CONFIG_SGETMASK_SYSCALL
3495 
3496 /*
3497  * For backwards compatibility.  Functionality superseded by sigprocmask.
3498  */
3499 SYSCALL_DEFINE0(sgetmask)
3500 {
3501 	/* SMP safe */
3502 	return current->blocked.sig[0];
3503 }
3504 
3505 SYSCALL_DEFINE1(ssetmask, int, newmask)
3506 {
3507 	int old = current->blocked.sig[0];
3508 	sigset_t newset;
3509 
3510 	siginitset(&newset, newmask);
3511 	set_current_blocked(&newset);
3512 
3513 	return old;
3514 }
3515 #endif /* CONFIG_SGETMASK_SYSCALL */
3516 
3517 #ifdef __ARCH_WANT_SYS_SIGNAL
3518 /*
3519  * For backwards compatibility.  Functionality superseded by sigaction.
3520  */
3521 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
3522 {
3523 	struct k_sigaction new_sa, old_sa;
3524 	int ret;
3525 
3526 	new_sa.sa.sa_handler = handler;
3527 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
3528 	sigemptyset(&new_sa.sa.sa_mask);
3529 
3530 	ret = do_sigaction(sig, &new_sa, &old_sa);
3531 
3532 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
3533 }
3534 #endif /* __ARCH_WANT_SYS_SIGNAL */
3535 
3536 #ifdef __ARCH_WANT_SYS_PAUSE
3537 
3538 SYSCALL_DEFINE0(pause)
3539 {
3540 	while (!signal_pending(current)) {
3541 		__set_current_state(TASK_INTERRUPTIBLE);
3542 		schedule();
3543 	}
3544 	return -ERESTARTNOHAND;
3545 }
3546 
3547 #endif
3548 
3549 static int sigsuspend(sigset_t *set)
3550 {
3551 	current->saved_sigmask = current->blocked;
3552 	set_current_blocked(set);
3553 
3554 	while (!signal_pending(current)) {
3555 		__set_current_state(TASK_INTERRUPTIBLE);
3556 		schedule();
3557 	}
3558 	set_restore_sigmask();
3559 	return -ERESTARTNOHAND;
3560 }
3561 
3562 /**
3563  *  sys_rt_sigsuspend - replace the signal mask for a value with the
3564  *	@unewset value until a signal is received
3565  *  @unewset: new signal mask value
3566  *  @sigsetsize: size of sigset_t type
3567  */
3568 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
3569 {
3570 	sigset_t newset;
3571 
3572 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3573 	if (sigsetsize != sizeof(sigset_t))
3574 		return -EINVAL;
3575 
3576 	if (copy_from_user(&newset, unewset, sizeof(newset)))
3577 		return -EFAULT;
3578 	return sigsuspend(&newset);
3579 }
3580 
3581 #ifdef CONFIG_COMPAT
3582 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
3583 {
3584 #ifdef __BIG_ENDIAN
3585 	sigset_t newset;
3586 	compat_sigset_t newset32;
3587 
3588 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3589 	if (sigsetsize != sizeof(sigset_t))
3590 		return -EINVAL;
3591 
3592 	if (copy_from_user(&newset32, unewset, sizeof(compat_sigset_t)))
3593 		return -EFAULT;
3594 	sigset_from_compat(&newset, &newset32);
3595 	return sigsuspend(&newset);
3596 #else
3597 	/* on little-endian bitmaps don't care about granularity */
3598 	return sys_rt_sigsuspend((sigset_t __user *)unewset, sigsetsize);
3599 #endif
3600 }
3601 #endif
3602 
3603 #ifdef CONFIG_OLD_SIGSUSPEND
3604 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
3605 {
3606 	sigset_t blocked;
3607 	siginitset(&blocked, mask);
3608 	return sigsuspend(&blocked);
3609 }
3610 #endif
3611 #ifdef CONFIG_OLD_SIGSUSPEND3
3612 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
3613 {
3614 	sigset_t blocked;
3615 	siginitset(&blocked, mask);
3616 	return sigsuspend(&blocked);
3617 }
3618 #endif
3619 
3620 __weak const char *arch_vma_name(struct vm_area_struct *vma)
3621 {
3622 	return NULL;
3623 }
3624 
3625 void __init signals_init(void)
3626 {
3627 	/* If this check fails, the __ARCH_SI_PREAMBLE_SIZE value is wrong! */
3628 	BUILD_BUG_ON(__ARCH_SI_PREAMBLE_SIZE
3629 		!= offsetof(struct siginfo, _sifields._pad));
3630 
3631 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
3632 }
3633 
3634 #ifdef CONFIG_KGDB_KDB
3635 #include <linux/kdb.h>
3636 /*
3637  * kdb_send_sig_info - Allows kdb to send signals without exposing
3638  * signal internals.  This function checks if the required locks are
3639  * available before calling the main signal code, to avoid kdb
3640  * deadlocks.
3641  */
3642 void
3643 kdb_send_sig_info(struct task_struct *t, struct siginfo *info)
3644 {
3645 	static struct task_struct *kdb_prev_t;
3646 	int sig, new_t;
3647 	if (!spin_trylock(&t->sighand->siglock)) {
3648 		kdb_printf("Can't do kill command now.\n"
3649 			   "The sigmask lock is held somewhere else in "
3650 			   "kernel, try again later\n");
3651 		return;
3652 	}
3653 	spin_unlock(&t->sighand->siglock);
3654 	new_t = kdb_prev_t != t;
3655 	kdb_prev_t = t;
3656 	if (t->state != TASK_RUNNING && new_t) {
3657 		kdb_printf("Process is not RUNNING, sending a signal from "
3658 			   "kdb risks deadlock\n"
3659 			   "on the run queue locks. "
3660 			   "The signal has _not_ been sent.\n"
3661 			   "Reissue the kill command if you want to risk "
3662 			   "the deadlock.\n");
3663 		return;
3664 	}
3665 	sig = info->si_signo;
3666 	if (send_sig_info(sig, info, t))
3667 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
3668 			   sig, t->pid);
3669 	else
3670 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
3671 }
3672 #endif	/* CONFIG_KGDB_KDB */
3673