xref: /linux/kernel/signal.c (revision 527d1470744d338c912f94bc1f4dba08ffdff349)
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, override_rlimit);
1040 	if (q) {
1041 		list_add_tail(&q->list, &pending->list);
1042 		switch ((unsigned long) info) {
1043 		case (unsigned long) SEND_SIG_NOINFO:
1044 			q->info.si_signo = sig;
1045 			q->info.si_errno = 0;
1046 			q->info.si_code = SI_USER;
1047 			q->info.si_pid = task_tgid_nr_ns(current,
1048 							task_active_pid_ns(t));
1049 			q->info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1050 			break;
1051 		case (unsigned long) SEND_SIG_PRIV:
1052 			q->info.si_signo = sig;
1053 			q->info.si_errno = 0;
1054 			q->info.si_code = SI_KERNEL;
1055 			q->info.si_pid = 0;
1056 			q->info.si_uid = 0;
1057 			break;
1058 		default:
1059 			copy_siginfo(&q->info, info);
1060 			if (from_ancestor_ns)
1061 				q->info.si_pid = 0;
1062 			break;
1063 		}
1064 
1065 		userns_fixup_signal_uid(&q->info, t);
1066 
1067 	} else if (!is_si_special(info)) {
1068 		if (sig >= SIGRTMIN && info->si_code != SI_USER) {
1069 			/*
1070 			 * Queue overflow, abort.  We may abort if the
1071 			 * signal was rt and sent by user using something
1072 			 * other than kill().
1073 			 */
1074 			result = TRACE_SIGNAL_OVERFLOW_FAIL;
1075 			ret = -EAGAIN;
1076 			goto ret;
1077 		} else {
1078 			/*
1079 			 * This is a silent loss of information.  We still
1080 			 * send the signal, but the *info bits are lost.
1081 			 */
1082 			result = TRACE_SIGNAL_LOSE_INFO;
1083 		}
1084 	}
1085 
1086 out_set:
1087 	signalfd_notify(t, sig);
1088 	sigaddset(&pending->signal, sig);
1089 	complete_signal(sig, t, group);
1090 ret:
1091 	trace_signal_generate(sig, info, t, group, result);
1092 	return ret;
1093 }
1094 
1095 static int send_signal(int sig, struct siginfo *info, struct task_struct *t,
1096 			int group)
1097 {
1098 	int from_ancestor_ns = 0;
1099 
1100 #ifdef CONFIG_PID_NS
1101 	from_ancestor_ns = si_fromuser(info) &&
1102 			   !task_pid_nr_ns(current, task_active_pid_ns(t));
1103 #endif
1104 
1105 	return __send_signal(sig, info, t, group, from_ancestor_ns);
1106 }
1107 
1108 static void print_fatal_signal(int signr)
1109 {
1110 	struct pt_regs *regs = signal_pt_regs();
1111 	pr_info("potentially unexpected fatal signal %d.\n", signr);
1112 
1113 #if defined(__i386__) && !defined(__arch_um__)
1114 	pr_info("code at %08lx: ", regs->ip);
1115 	{
1116 		int i;
1117 		for (i = 0; i < 16; i++) {
1118 			unsigned char insn;
1119 
1120 			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1121 				break;
1122 			pr_cont("%02x ", insn);
1123 		}
1124 	}
1125 	pr_cont("\n");
1126 #endif
1127 	preempt_disable();
1128 	show_regs(regs);
1129 	preempt_enable();
1130 }
1131 
1132 static int __init setup_print_fatal_signals(char *str)
1133 {
1134 	get_option (&str, &print_fatal_signals);
1135 
1136 	return 1;
1137 }
1138 
1139 __setup("print-fatal-signals=", setup_print_fatal_signals);
1140 
1141 int
1142 __group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1143 {
1144 	return send_signal(sig, info, p, 1);
1145 }
1146 
1147 static int
1148 specific_send_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1149 {
1150 	return send_signal(sig, info, t, 0);
1151 }
1152 
1153 int do_send_sig_info(int sig, struct siginfo *info, struct task_struct *p,
1154 			bool group)
1155 {
1156 	unsigned long flags;
1157 	int ret = -ESRCH;
1158 
1159 	if (lock_task_sighand(p, &flags)) {
1160 		ret = send_signal(sig, info, p, group);
1161 		unlock_task_sighand(p, &flags);
1162 	}
1163 
1164 	return ret;
1165 }
1166 
1167 /*
1168  * Force a signal that the process can't ignore: if necessary
1169  * we unblock the signal and change any SIG_IGN to SIG_DFL.
1170  *
1171  * Note: If we unblock the signal, we always reset it to SIG_DFL,
1172  * since we do not want to have a signal handler that was blocked
1173  * be invoked when user space had explicitly blocked it.
1174  *
1175  * We don't want to have recursive SIGSEGV's etc, for example,
1176  * that is why we also clear SIGNAL_UNKILLABLE.
1177  */
1178 int
1179 force_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1180 {
1181 	unsigned long int flags;
1182 	int ret, blocked, ignored;
1183 	struct k_sigaction *action;
1184 
1185 	spin_lock_irqsave(&t->sighand->siglock, flags);
1186 	action = &t->sighand->action[sig-1];
1187 	ignored = action->sa.sa_handler == SIG_IGN;
1188 	blocked = sigismember(&t->blocked, sig);
1189 	if (blocked || ignored) {
1190 		action->sa.sa_handler = SIG_DFL;
1191 		if (blocked) {
1192 			sigdelset(&t->blocked, sig);
1193 			recalc_sigpending_and_wake(t);
1194 		}
1195 	}
1196 	/*
1197 	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1198 	 * debugging to leave init killable.
1199 	 */
1200 	if (action->sa.sa_handler == SIG_DFL && !t->ptrace)
1201 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1202 	ret = specific_send_sig_info(sig, info, t);
1203 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1204 
1205 	return ret;
1206 }
1207 
1208 /*
1209  * Nuke all other threads in the group.
1210  */
1211 int zap_other_threads(struct task_struct *p)
1212 {
1213 	struct task_struct *t = p;
1214 	int count = 0;
1215 
1216 	p->signal->group_stop_count = 0;
1217 
1218 	while_each_thread(p, t) {
1219 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1220 		count++;
1221 
1222 		/* Don't bother with already dead threads */
1223 		if (t->exit_state)
1224 			continue;
1225 		sigaddset(&t->pending.signal, SIGKILL);
1226 		signal_wake_up(t, 1);
1227 	}
1228 
1229 	return count;
1230 }
1231 
1232 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1233 					   unsigned long *flags)
1234 {
1235 	struct sighand_struct *sighand;
1236 
1237 	for (;;) {
1238 		/*
1239 		 * Disable interrupts early to avoid deadlocks.
1240 		 * See rcu_read_unlock() comment header for details.
1241 		 */
1242 		local_irq_save(*flags);
1243 		rcu_read_lock();
1244 		sighand = rcu_dereference(tsk->sighand);
1245 		if (unlikely(sighand == NULL)) {
1246 			rcu_read_unlock();
1247 			local_irq_restore(*flags);
1248 			break;
1249 		}
1250 		/*
1251 		 * This sighand can be already freed and even reused, but
1252 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1253 		 * initializes ->siglock: this slab can't go away, it has
1254 		 * the same object type, ->siglock can't be reinitialized.
1255 		 *
1256 		 * We need to ensure that tsk->sighand is still the same
1257 		 * after we take the lock, we can race with de_thread() or
1258 		 * __exit_signal(). In the latter case the next iteration
1259 		 * must see ->sighand == NULL.
1260 		 */
1261 		spin_lock(&sighand->siglock);
1262 		if (likely(sighand == tsk->sighand)) {
1263 			rcu_read_unlock();
1264 			break;
1265 		}
1266 		spin_unlock(&sighand->siglock);
1267 		rcu_read_unlock();
1268 		local_irq_restore(*flags);
1269 	}
1270 
1271 	return sighand;
1272 }
1273 
1274 /*
1275  * send signal info to all the members of a group
1276  */
1277 int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1278 {
1279 	int ret;
1280 
1281 	rcu_read_lock();
1282 	ret = check_kill_permission(sig, info, p);
1283 	rcu_read_unlock();
1284 
1285 	if (!ret && sig)
1286 		ret = do_send_sig_info(sig, info, p, true);
1287 
1288 	return ret;
1289 }
1290 
1291 /*
1292  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1293  * control characters do (^C, ^Z etc)
1294  * - the caller must hold at least a readlock on tasklist_lock
1295  */
1296 int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp)
1297 {
1298 	struct task_struct *p = NULL;
1299 	int retval, success;
1300 
1301 	success = 0;
1302 	retval = -ESRCH;
1303 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1304 		int err = group_send_sig_info(sig, info, p);
1305 		success |= !err;
1306 		retval = err;
1307 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1308 	return success ? 0 : retval;
1309 }
1310 
1311 int kill_pid_info(int sig, struct siginfo *info, struct pid *pid)
1312 {
1313 	int error = -ESRCH;
1314 	struct task_struct *p;
1315 
1316 	for (;;) {
1317 		rcu_read_lock();
1318 		p = pid_task(pid, PIDTYPE_PID);
1319 		if (p)
1320 			error = group_send_sig_info(sig, info, p);
1321 		rcu_read_unlock();
1322 		if (likely(!p || error != -ESRCH))
1323 			return error;
1324 
1325 		/*
1326 		 * The task was unhashed in between, try again.  If it
1327 		 * is dead, pid_task() will return NULL, if we race with
1328 		 * de_thread() it will find the new leader.
1329 		 */
1330 	}
1331 }
1332 
1333 static int kill_proc_info(int sig, struct siginfo *info, pid_t pid)
1334 {
1335 	int error;
1336 	rcu_read_lock();
1337 	error = kill_pid_info(sig, info, find_vpid(pid));
1338 	rcu_read_unlock();
1339 	return error;
1340 }
1341 
1342 static int kill_as_cred_perm(const struct cred *cred,
1343 			     struct task_struct *target)
1344 {
1345 	const struct cred *pcred = __task_cred(target);
1346 	if (!uid_eq(cred->euid, pcred->suid) && !uid_eq(cred->euid, pcred->uid) &&
1347 	    !uid_eq(cred->uid,  pcred->suid) && !uid_eq(cred->uid,  pcred->uid))
1348 		return 0;
1349 	return 1;
1350 }
1351 
1352 /* like kill_pid_info(), but doesn't use uid/euid of "current" */
1353 int kill_pid_info_as_cred(int sig, struct siginfo *info, struct pid *pid,
1354 			 const struct cred *cred, u32 secid)
1355 {
1356 	int ret = -EINVAL;
1357 	struct task_struct *p;
1358 	unsigned long flags;
1359 
1360 	if (!valid_signal(sig))
1361 		return ret;
1362 
1363 	rcu_read_lock();
1364 	p = pid_task(pid, PIDTYPE_PID);
1365 	if (!p) {
1366 		ret = -ESRCH;
1367 		goto out_unlock;
1368 	}
1369 	if (si_fromuser(info) && !kill_as_cred_perm(cred, p)) {
1370 		ret = -EPERM;
1371 		goto out_unlock;
1372 	}
1373 	ret = security_task_kill(p, info, sig, secid);
1374 	if (ret)
1375 		goto out_unlock;
1376 
1377 	if (sig) {
1378 		if (lock_task_sighand(p, &flags)) {
1379 			ret = __send_signal(sig, info, p, 1, 0);
1380 			unlock_task_sighand(p, &flags);
1381 		} else
1382 			ret = -ESRCH;
1383 	}
1384 out_unlock:
1385 	rcu_read_unlock();
1386 	return ret;
1387 }
1388 EXPORT_SYMBOL_GPL(kill_pid_info_as_cred);
1389 
1390 /*
1391  * kill_something_info() interprets pid in interesting ways just like kill(2).
1392  *
1393  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1394  * is probably wrong.  Should make it like BSD or SYSV.
1395  */
1396 
1397 static int kill_something_info(int sig, struct siginfo *info, pid_t pid)
1398 {
1399 	int ret;
1400 
1401 	if (pid > 0) {
1402 		rcu_read_lock();
1403 		ret = kill_pid_info(sig, info, find_vpid(pid));
1404 		rcu_read_unlock();
1405 		return ret;
1406 	}
1407 
1408 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1409 	if (pid == INT_MIN)
1410 		return -ESRCH;
1411 
1412 	read_lock(&tasklist_lock);
1413 	if (pid != -1) {
1414 		ret = __kill_pgrp_info(sig, info,
1415 				pid ? find_vpid(-pid) : task_pgrp(current));
1416 	} else {
1417 		int retval = 0, count = 0;
1418 		struct task_struct * p;
1419 
1420 		for_each_process(p) {
1421 			if (task_pid_vnr(p) > 1 &&
1422 					!same_thread_group(p, current)) {
1423 				int err = group_send_sig_info(sig, info, p);
1424 				++count;
1425 				if (err != -EPERM)
1426 					retval = err;
1427 			}
1428 		}
1429 		ret = count ? retval : -ESRCH;
1430 	}
1431 	read_unlock(&tasklist_lock);
1432 
1433 	return ret;
1434 }
1435 
1436 /*
1437  * These are for backward compatibility with the rest of the kernel source.
1438  */
1439 
1440 int send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1441 {
1442 	/*
1443 	 * Make sure legacy kernel users don't send in bad values
1444 	 * (normal paths check this in check_kill_permission).
1445 	 */
1446 	if (!valid_signal(sig))
1447 		return -EINVAL;
1448 
1449 	return do_send_sig_info(sig, info, p, false);
1450 }
1451 
1452 #define __si_special(priv) \
1453 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1454 
1455 int
1456 send_sig(int sig, struct task_struct *p, int priv)
1457 {
1458 	return send_sig_info(sig, __si_special(priv), p);
1459 }
1460 
1461 void
1462 force_sig(int sig, struct task_struct *p)
1463 {
1464 	force_sig_info(sig, SEND_SIG_PRIV, p);
1465 }
1466 
1467 /*
1468  * When things go south during signal handling, we
1469  * will force a SIGSEGV. And if the signal that caused
1470  * the problem was already a SIGSEGV, we'll want to
1471  * make sure we don't even try to deliver the signal..
1472  */
1473 int
1474 force_sigsegv(int sig, struct task_struct *p)
1475 {
1476 	if (sig == SIGSEGV) {
1477 		unsigned long flags;
1478 		spin_lock_irqsave(&p->sighand->siglock, flags);
1479 		p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1480 		spin_unlock_irqrestore(&p->sighand->siglock, flags);
1481 	}
1482 	force_sig(SIGSEGV, p);
1483 	return 0;
1484 }
1485 
1486 int kill_pgrp(struct pid *pid, int sig, int priv)
1487 {
1488 	int ret;
1489 
1490 	read_lock(&tasklist_lock);
1491 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1492 	read_unlock(&tasklist_lock);
1493 
1494 	return ret;
1495 }
1496 EXPORT_SYMBOL(kill_pgrp);
1497 
1498 int kill_pid(struct pid *pid, int sig, int priv)
1499 {
1500 	return kill_pid_info(sig, __si_special(priv), pid);
1501 }
1502 EXPORT_SYMBOL(kill_pid);
1503 
1504 /*
1505  * These functions support sending signals using preallocated sigqueue
1506  * structures.  This is needed "because realtime applications cannot
1507  * afford to lose notifications of asynchronous events, like timer
1508  * expirations or I/O completions".  In the case of POSIX Timers
1509  * we allocate the sigqueue structure from the timer_create.  If this
1510  * allocation fails we are able to report the failure to the application
1511  * with an EAGAIN error.
1512  */
1513 struct sigqueue *sigqueue_alloc(void)
1514 {
1515 	struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1516 
1517 	if (q)
1518 		q->flags |= SIGQUEUE_PREALLOC;
1519 
1520 	return q;
1521 }
1522 
1523 void sigqueue_free(struct sigqueue *q)
1524 {
1525 	unsigned long flags;
1526 	spinlock_t *lock = &current->sighand->siglock;
1527 
1528 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1529 	/*
1530 	 * We must hold ->siglock while testing q->list
1531 	 * to serialize with collect_signal() or with
1532 	 * __exit_signal()->flush_sigqueue().
1533 	 */
1534 	spin_lock_irqsave(lock, flags);
1535 	q->flags &= ~SIGQUEUE_PREALLOC;
1536 	/*
1537 	 * If it is queued it will be freed when dequeued,
1538 	 * like the "regular" sigqueue.
1539 	 */
1540 	if (!list_empty(&q->list))
1541 		q = NULL;
1542 	spin_unlock_irqrestore(lock, flags);
1543 
1544 	if (q)
1545 		__sigqueue_free(q);
1546 }
1547 
1548 int send_sigqueue(struct sigqueue *q, struct task_struct *t, int group)
1549 {
1550 	int sig = q->info.si_signo;
1551 	struct sigpending *pending;
1552 	unsigned long flags;
1553 	int ret, result;
1554 
1555 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1556 
1557 	ret = -1;
1558 	if (!likely(lock_task_sighand(t, &flags)))
1559 		goto ret;
1560 
1561 	ret = 1; /* the signal is ignored */
1562 	result = TRACE_SIGNAL_IGNORED;
1563 	if (!prepare_signal(sig, t, false))
1564 		goto out;
1565 
1566 	ret = 0;
1567 	if (unlikely(!list_empty(&q->list))) {
1568 		/*
1569 		 * If an SI_TIMER entry is already queue just increment
1570 		 * the overrun count.
1571 		 */
1572 		BUG_ON(q->info.si_code != SI_TIMER);
1573 		q->info.si_overrun++;
1574 		result = TRACE_SIGNAL_ALREADY_PENDING;
1575 		goto out;
1576 	}
1577 	q->info.si_overrun = 0;
1578 
1579 	signalfd_notify(t, sig);
1580 	pending = group ? &t->signal->shared_pending : &t->pending;
1581 	list_add_tail(&q->list, &pending->list);
1582 	sigaddset(&pending->signal, sig);
1583 	complete_signal(sig, t, group);
1584 	result = TRACE_SIGNAL_DELIVERED;
1585 out:
1586 	trace_signal_generate(sig, &q->info, t, group, result);
1587 	unlock_task_sighand(t, &flags);
1588 ret:
1589 	return ret;
1590 }
1591 
1592 /*
1593  * Let a parent know about the death of a child.
1594  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1595  *
1596  * Returns true if our parent ignored us and so we've switched to
1597  * self-reaping.
1598  */
1599 bool do_notify_parent(struct task_struct *tsk, int sig)
1600 {
1601 	struct siginfo info;
1602 	unsigned long flags;
1603 	struct sighand_struct *psig;
1604 	bool autoreap = false;
1605 	u64 utime, stime;
1606 
1607 	BUG_ON(sig == -1);
1608 
1609  	/* do_notify_parent_cldstop should have been called instead.  */
1610  	BUG_ON(task_is_stopped_or_traced(tsk));
1611 
1612 	BUG_ON(!tsk->ptrace &&
1613 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1614 
1615 	if (sig != SIGCHLD) {
1616 		/*
1617 		 * This is only possible if parent == real_parent.
1618 		 * Check if it has changed security domain.
1619 		 */
1620 		if (tsk->parent_exec_id != tsk->parent->self_exec_id)
1621 			sig = SIGCHLD;
1622 	}
1623 
1624 	info.si_signo = sig;
1625 	info.si_errno = 0;
1626 	/*
1627 	 * We are under tasklist_lock here so our parent is tied to
1628 	 * us and cannot change.
1629 	 *
1630 	 * task_active_pid_ns will always return the same pid namespace
1631 	 * until a task passes through release_task.
1632 	 *
1633 	 * write_lock() currently calls preempt_disable() which is the
1634 	 * same as rcu_read_lock(), but according to Oleg, this is not
1635 	 * correct to rely on this
1636 	 */
1637 	rcu_read_lock();
1638 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1639 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1640 				       task_uid(tsk));
1641 	rcu_read_unlock();
1642 
1643 	task_cputime(tsk, &utime, &stime);
1644 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
1645 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
1646 
1647 	info.si_status = tsk->exit_code & 0x7f;
1648 	if (tsk->exit_code & 0x80)
1649 		info.si_code = CLD_DUMPED;
1650 	else if (tsk->exit_code & 0x7f)
1651 		info.si_code = CLD_KILLED;
1652 	else {
1653 		info.si_code = CLD_EXITED;
1654 		info.si_status = tsk->exit_code >> 8;
1655 	}
1656 
1657 	psig = tsk->parent->sighand;
1658 	spin_lock_irqsave(&psig->siglock, flags);
1659 	if (!tsk->ptrace && sig == SIGCHLD &&
1660 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1661 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1662 		/*
1663 		 * We are exiting and our parent doesn't care.  POSIX.1
1664 		 * defines special semantics for setting SIGCHLD to SIG_IGN
1665 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
1666 		 * automatically and not left for our parent's wait4 call.
1667 		 * Rather than having the parent do it as a magic kind of
1668 		 * signal handler, we just set this to tell do_exit that we
1669 		 * can be cleaned up without becoming a zombie.  Note that
1670 		 * we still call __wake_up_parent in this case, because a
1671 		 * blocked sys_wait4 might now return -ECHILD.
1672 		 *
1673 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1674 		 * is implementation-defined: we do (if you don't want
1675 		 * it, just use SIG_IGN instead).
1676 		 */
1677 		autoreap = true;
1678 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
1679 			sig = 0;
1680 	}
1681 	if (valid_signal(sig) && sig)
1682 		__group_send_sig_info(sig, &info, tsk->parent);
1683 	__wake_up_parent(tsk, tsk->parent);
1684 	spin_unlock_irqrestore(&psig->siglock, flags);
1685 
1686 	return autoreap;
1687 }
1688 
1689 /**
1690  * do_notify_parent_cldstop - notify parent of stopped/continued state change
1691  * @tsk: task reporting the state change
1692  * @for_ptracer: the notification is for ptracer
1693  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
1694  *
1695  * Notify @tsk's parent that the stopped/continued state has changed.  If
1696  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
1697  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
1698  *
1699  * CONTEXT:
1700  * Must be called with tasklist_lock at least read locked.
1701  */
1702 static void do_notify_parent_cldstop(struct task_struct *tsk,
1703 				     bool for_ptracer, int why)
1704 {
1705 	struct siginfo info;
1706 	unsigned long flags;
1707 	struct task_struct *parent;
1708 	struct sighand_struct *sighand;
1709 	u64 utime, stime;
1710 
1711 	if (for_ptracer) {
1712 		parent = tsk->parent;
1713 	} else {
1714 		tsk = tsk->group_leader;
1715 		parent = tsk->real_parent;
1716 	}
1717 
1718 	info.si_signo = SIGCHLD;
1719 	info.si_errno = 0;
1720 	/*
1721 	 * see comment in do_notify_parent() about the following 4 lines
1722 	 */
1723 	rcu_read_lock();
1724 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
1725 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
1726 	rcu_read_unlock();
1727 
1728 	task_cputime(tsk, &utime, &stime);
1729 	info.si_utime = nsec_to_clock_t(utime);
1730 	info.si_stime = nsec_to_clock_t(stime);
1731 
1732  	info.si_code = why;
1733  	switch (why) {
1734  	case CLD_CONTINUED:
1735  		info.si_status = SIGCONT;
1736  		break;
1737  	case CLD_STOPPED:
1738  		info.si_status = tsk->signal->group_exit_code & 0x7f;
1739  		break;
1740  	case CLD_TRAPPED:
1741  		info.si_status = tsk->exit_code & 0x7f;
1742  		break;
1743  	default:
1744  		BUG();
1745  	}
1746 
1747 	sighand = parent->sighand;
1748 	spin_lock_irqsave(&sighand->siglock, flags);
1749 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
1750 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
1751 		__group_send_sig_info(SIGCHLD, &info, parent);
1752 	/*
1753 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
1754 	 */
1755 	__wake_up_parent(tsk, parent);
1756 	spin_unlock_irqrestore(&sighand->siglock, flags);
1757 }
1758 
1759 static inline int may_ptrace_stop(void)
1760 {
1761 	if (!likely(current->ptrace))
1762 		return 0;
1763 	/*
1764 	 * Are we in the middle of do_coredump?
1765 	 * If so and our tracer is also part of the coredump stopping
1766 	 * is a deadlock situation, and pointless because our tracer
1767 	 * is dead so don't allow us to stop.
1768 	 * If SIGKILL was already sent before the caller unlocked
1769 	 * ->siglock we must see ->core_state != NULL. Otherwise it
1770 	 * is safe to enter schedule().
1771 	 *
1772 	 * This is almost outdated, a task with the pending SIGKILL can't
1773 	 * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
1774 	 * after SIGKILL was already dequeued.
1775 	 */
1776 	if (unlikely(current->mm->core_state) &&
1777 	    unlikely(current->mm == current->parent->mm))
1778 		return 0;
1779 
1780 	return 1;
1781 }
1782 
1783 /*
1784  * Return non-zero if there is a SIGKILL that should be waking us up.
1785  * Called with the siglock held.
1786  */
1787 static int sigkill_pending(struct task_struct *tsk)
1788 {
1789 	return	sigismember(&tsk->pending.signal, SIGKILL) ||
1790 		sigismember(&tsk->signal->shared_pending.signal, SIGKILL);
1791 }
1792 
1793 /*
1794  * This must be called with current->sighand->siglock held.
1795  *
1796  * This should be the path for all ptrace stops.
1797  * We always set current->last_siginfo while stopped here.
1798  * That makes it a way to test a stopped process for
1799  * being ptrace-stopped vs being job-control-stopped.
1800  *
1801  * If we actually decide not to stop at all because the tracer
1802  * is gone, we keep current->exit_code unless clear_code.
1803  */
1804 static void ptrace_stop(int exit_code, int why, int clear_code, siginfo_t *info)
1805 	__releases(&current->sighand->siglock)
1806 	__acquires(&current->sighand->siglock)
1807 {
1808 	bool gstop_done = false;
1809 
1810 	if (arch_ptrace_stop_needed(exit_code, info)) {
1811 		/*
1812 		 * The arch code has something special to do before a
1813 		 * ptrace stop.  This is allowed to block, e.g. for faults
1814 		 * on user stack pages.  We can't keep the siglock while
1815 		 * calling arch_ptrace_stop, so we must release it now.
1816 		 * To preserve proper semantics, we must do this before
1817 		 * any signal bookkeeping like checking group_stop_count.
1818 		 * Meanwhile, a SIGKILL could come in before we retake the
1819 		 * siglock.  That must prevent us from sleeping in TASK_TRACED.
1820 		 * So after regaining the lock, we must check for SIGKILL.
1821 		 */
1822 		spin_unlock_irq(&current->sighand->siglock);
1823 		arch_ptrace_stop(exit_code, info);
1824 		spin_lock_irq(&current->sighand->siglock);
1825 		if (sigkill_pending(current))
1826 			return;
1827 	}
1828 
1829 	/*
1830 	 * We're committing to trapping.  TRACED should be visible before
1831 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
1832 	 * Also, transition to TRACED and updates to ->jobctl should be
1833 	 * atomic with respect to siglock and should be done after the arch
1834 	 * hook as siglock is released and regrabbed across it.
1835 	 */
1836 	set_current_state(TASK_TRACED);
1837 
1838 	current->last_siginfo = info;
1839 	current->exit_code = exit_code;
1840 
1841 	/*
1842 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
1843 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
1844 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
1845 	 * could be clear now.  We act as if SIGCONT is received after
1846 	 * TASK_TRACED is entered - ignore it.
1847 	 */
1848 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
1849 		gstop_done = task_participate_group_stop(current);
1850 
1851 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
1852 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
1853 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
1854 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
1855 
1856 	/* entering a trap, clear TRAPPING */
1857 	task_clear_jobctl_trapping(current);
1858 
1859 	spin_unlock_irq(&current->sighand->siglock);
1860 	read_lock(&tasklist_lock);
1861 	if (may_ptrace_stop()) {
1862 		/*
1863 		 * Notify parents of the stop.
1864 		 *
1865 		 * While ptraced, there are two parents - the ptracer and
1866 		 * the real_parent of the group_leader.  The ptracer should
1867 		 * know about every stop while the real parent is only
1868 		 * interested in the completion of group stop.  The states
1869 		 * for the two don't interact with each other.  Notify
1870 		 * separately unless they're gonna be duplicates.
1871 		 */
1872 		do_notify_parent_cldstop(current, true, why);
1873 		if (gstop_done && ptrace_reparented(current))
1874 			do_notify_parent_cldstop(current, false, why);
1875 
1876 		/*
1877 		 * Don't want to allow preemption here, because
1878 		 * sys_ptrace() needs this task to be inactive.
1879 		 *
1880 		 * XXX: implement read_unlock_no_resched().
1881 		 */
1882 		preempt_disable();
1883 		read_unlock(&tasklist_lock);
1884 		preempt_enable_no_resched();
1885 		freezable_schedule();
1886 	} else {
1887 		/*
1888 		 * By the time we got the lock, our tracer went away.
1889 		 * Don't drop the lock yet, another tracer may come.
1890 		 *
1891 		 * If @gstop_done, the ptracer went away between group stop
1892 		 * completion and here.  During detach, it would have set
1893 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
1894 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
1895 		 * the real parent of the group stop completion is enough.
1896 		 */
1897 		if (gstop_done)
1898 			do_notify_parent_cldstop(current, false, why);
1899 
1900 		/* tasklist protects us from ptrace_freeze_traced() */
1901 		__set_current_state(TASK_RUNNING);
1902 		if (clear_code)
1903 			current->exit_code = 0;
1904 		read_unlock(&tasklist_lock);
1905 	}
1906 
1907 	/*
1908 	 * We are back.  Now reacquire the siglock before touching
1909 	 * last_siginfo, so that we are sure to have synchronized with
1910 	 * any signal-sending on another CPU that wants to examine it.
1911 	 */
1912 	spin_lock_irq(&current->sighand->siglock);
1913 	current->last_siginfo = NULL;
1914 
1915 	/* LISTENING can be set only during STOP traps, clear it */
1916 	current->jobctl &= ~JOBCTL_LISTENING;
1917 
1918 	/*
1919 	 * Queued signals ignored us while we were stopped for tracing.
1920 	 * So check for any that we should take before resuming user mode.
1921 	 * This sets TIF_SIGPENDING, but never clears it.
1922 	 */
1923 	recalc_sigpending_tsk(current);
1924 }
1925 
1926 static void ptrace_do_notify(int signr, int exit_code, int why)
1927 {
1928 	siginfo_t info;
1929 
1930 	memset(&info, 0, sizeof info);
1931 	info.si_signo = signr;
1932 	info.si_code = exit_code;
1933 	info.si_pid = task_pid_vnr(current);
1934 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1935 
1936 	/* Let the debugger run.  */
1937 	ptrace_stop(exit_code, why, 1, &info);
1938 }
1939 
1940 void ptrace_notify(int exit_code)
1941 {
1942 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
1943 	if (unlikely(current->task_works))
1944 		task_work_run();
1945 
1946 	spin_lock_irq(&current->sighand->siglock);
1947 	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
1948 	spin_unlock_irq(&current->sighand->siglock);
1949 }
1950 
1951 /**
1952  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
1953  * @signr: signr causing group stop if initiating
1954  *
1955  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
1956  * and participate in it.  If already set, participate in the existing
1957  * group stop.  If participated in a group stop (and thus slept), %true is
1958  * returned with siglock released.
1959  *
1960  * If ptraced, this function doesn't handle stop itself.  Instead,
1961  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
1962  * untouched.  The caller must ensure that INTERRUPT trap handling takes
1963  * places afterwards.
1964  *
1965  * CONTEXT:
1966  * Must be called with @current->sighand->siglock held, which is released
1967  * on %true return.
1968  *
1969  * RETURNS:
1970  * %false if group stop is already cancelled or ptrace trap is scheduled.
1971  * %true if participated in group stop.
1972  */
1973 static bool do_signal_stop(int signr)
1974 	__releases(&current->sighand->siglock)
1975 {
1976 	struct signal_struct *sig = current->signal;
1977 
1978 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
1979 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
1980 		struct task_struct *t;
1981 
1982 		/* signr will be recorded in task->jobctl for retries */
1983 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
1984 
1985 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
1986 		    unlikely(signal_group_exit(sig)))
1987 			return false;
1988 		/*
1989 		 * There is no group stop already in progress.  We must
1990 		 * initiate one now.
1991 		 *
1992 		 * While ptraced, a task may be resumed while group stop is
1993 		 * still in effect and then receive a stop signal and
1994 		 * initiate another group stop.  This deviates from the
1995 		 * usual behavior as two consecutive stop signals can't
1996 		 * cause two group stops when !ptraced.  That is why we
1997 		 * also check !task_is_stopped(t) below.
1998 		 *
1999 		 * The condition can be distinguished by testing whether
2000 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2001 		 * group_exit_code in such case.
2002 		 *
2003 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2004 		 * an intervening stop signal is required to cause two
2005 		 * continued events regardless of ptrace.
2006 		 */
2007 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2008 			sig->group_exit_code = signr;
2009 
2010 		sig->group_stop_count = 0;
2011 
2012 		if (task_set_jobctl_pending(current, signr | gstop))
2013 			sig->group_stop_count++;
2014 
2015 		t = current;
2016 		while_each_thread(current, t) {
2017 			/*
2018 			 * Setting state to TASK_STOPPED for a group
2019 			 * stop is always done with the siglock held,
2020 			 * so this check has no races.
2021 			 */
2022 			if (!task_is_stopped(t) &&
2023 			    task_set_jobctl_pending(t, signr | gstop)) {
2024 				sig->group_stop_count++;
2025 				if (likely(!(t->ptrace & PT_SEIZED)))
2026 					signal_wake_up(t, 0);
2027 				else
2028 					ptrace_trap_notify(t);
2029 			}
2030 		}
2031 	}
2032 
2033 	if (likely(!current->ptrace)) {
2034 		int notify = 0;
2035 
2036 		/*
2037 		 * If there are no other threads in the group, or if there
2038 		 * is a group stop in progress and we are the last to stop,
2039 		 * report to the parent.
2040 		 */
2041 		if (task_participate_group_stop(current))
2042 			notify = CLD_STOPPED;
2043 
2044 		__set_current_state(TASK_STOPPED);
2045 		spin_unlock_irq(&current->sighand->siglock);
2046 
2047 		/*
2048 		 * Notify the parent of the group stop completion.  Because
2049 		 * we're not holding either the siglock or tasklist_lock
2050 		 * here, ptracer may attach inbetween; however, this is for
2051 		 * group stop and should always be delivered to the real
2052 		 * parent of the group leader.  The new ptracer will get
2053 		 * its notification when this task transitions into
2054 		 * TASK_TRACED.
2055 		 */
2056 		if (notify) {
2057 			read_lock(&tasklist_lock);
2058 			do_notify_parent_cldstop(current, false, notify);
2059 			read_unlock(&tasklist_lock);
2060 		}
2061 
2062 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2063 		freezable_schedule();
2064 		return true;
2065 	} else {
2066 		/*
2067 		 * While ptraced, group stop is handled by STOP trap.
2068 		 * Schedule it and let the caller deal with it.
2069 		 */
2070 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2071 		return false;
2072 	}
2073 }
2074 
2075 /**
2076  * do_jobctl_trap - take care of ptrace jobctl traps
2077  *
2078  * When PT_SEIZED, it's used for both group stop and explicit
2079  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2080  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2081  * the stop signal; otherwise, %SIGTRAP.
2082  *
2083  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2084  * number as exit_code and no siginfo.
2085  *
2086  * CONTEXT:
2087  * Must be called with @current->sighand->siglock held, which may be
2088  * released and re-acquired before returning with intervening sleep.
2089  */
2090 static void do_jobctl_trap(void)
2091 {
2092 	struct signal_struct *signal = current->signal;
2093 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2094 
2095 	if (current->ptrace & PT_SEIZED) {
2096 		if (!signal->group_stop_count &&
2097 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2098 			signr = SIGTRAP;
2099 		WARN_ON_ONCE(!signr);
2100 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2101 				 CLD_STOPPED);
2102 	} else {
2103 		WARN_ON_ONCE(!signr);
2104 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2105 		current->exit_code = 0;
2106 	}
2107 }
2108 
2109 static int ptrace_signal(int signr, siginfo_t *info)
2110 {
2111 	/*
2112 	 * We do not check sig_kernel_stop(signr) but set this marker
2113 	 * unconditionally because we do not know whether debugger will
2114 	 * change signr. This flag has no meaning unless we are going
2115 	 * to stop after return from ptrace_stop(). In this case it will
2116 	 * be checked in do_signal_stop(), we should only stop if it was
2117 	 * not cleared by SIGCONT while we were sleeping. See also the
2118 	 * comment in dequeue_signal().
2119 	 */
2120 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2121 	ptrace_stop(signr, CLD_TRAPPED, 0, info);
2122 
2123 	/* We're back.  Did the debugger cancel the sig?  */
2124 	signr = current->exit_code;
2125 	if (signr == 0)
2126 		return signr;
2127 
2128 	current->exit_code = 0;
2129 
2130 	/*
2131 	 * Update the siginfo structure if the signal has
2132 	 * changed.  If the debugger wanted something
2133 	 * specific in the siginfo structure then it should
2134 	 * have updated *info via PTRACE_SETSIGINFO.
2135 	 */
2136 	if (signr != info->si_signo) {
2137 		info->si_signo = signr;
2138 		info->si_errno = 0;
2139 		info->si_code = SI_USER;
2140 		rcu_read_lock();
2141 		info->si_pid = task_pid_vnr(current->parent);
2142 		info->si_uid = from_kuid_munged(current_user_ns(),
2143 						task_uid(current->parent));
2144 		rcu_read_unlock();
2145 	}
2146 
2147 	/* If the (new) signal is now blocked, requeue it.  */
2148 	if (sigismember(&current->blocked, signr)) {
2149 		specific_send_sig_info(signr, info, current);
2150 		signr = 0;
2151 	}
2152 
2153 	return signr;
2154 }
2155 
2156 int get_signal(struct ksignal *ksig)
2157 {
2158 	struct sighand_struct *sighand = current->sighand;
2159 	struct signal_struct *signal = current->signal;
2160 	int signr;
2161 
2162 	if (unlikely(current->task_works))
2163 		task_work_run();
2164 
2165 	if (unlikely(uprobe_deny_signal()))
2166 		return 0;
2167 
2168 	/*
2169 	 * Do this once, we can't return to user-mode if freezing() == T.
2170 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2171 	 * thus do not need another check after return.
2172 	 */
2173 	try_to_freeze();
2174 
2175 relock:
2176 	spin_lock_irq(&sighand->siglock);
2177 	/*
2178 	 * Every stopped thread goes here after wakeup. Check to see if
2179 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2180 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2181 	 */
2182 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2183 		int why;
2184 
2185 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2186 			why = CLD_CONTINUED;
2187 		else
2188 			why = CLD_STOPPED;
2189 
2190 		signal->flags &= ~SIGNAL_CLD_MASK;
2191 
2192 		spin_unlock_irq(&sighand->siglock);
2193 
2194 		/*
2195 		 * Notify the parent that we're continuing.  This event is
2196 		 * always per-process and doesn't make whole lot of sense
2197 		 * for ptracers, who shouldn't consume the state via
2198 		 * wait(2) either, but, for backward compatibility, notify
2199 		 * the ptracer of the group leader too unless it's gonna be
2200 		 * a duplicate.
2201 		 */
2202 		read_lock(&tasklist_lock);
2203 		do_notify_parent_cldstop(current, false, why);
2204 
2205 		if (ptrace_reparented(current->group_leader))
2206 			do_notify_parent_cldstop(current->group_leader,
2207 						true, why);
2208 		read_unlock(&tasklist_lock);
2209 
2210 		goto relock;
2211 	}
2212 
2213 	for (;;) {
2214 		struct k_sigaction *ka;
2215 
2216 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2217 		    do_signal_stop(0))
2218 			goto relock;
2219 
2220 		if (unlikely(current->jobctl & JOBCTL_TRAP_MASK)) {
2221 			do_jobctl_trap();
2222 			spin_unlock_irq(&sighand->siglock);
2223 			goto relock;
2224 		}
2225 
2226 		signr = dequeue_signal(current, &current->blocked, &ksig->info);
2227 
2228 		if (!signr)
2229 			break; /* will return 0 */
2230 
2231 		if (unlikely(current->ptrace) && signr != SIGKILL) {
2232 			signr = ptrace_signal(signr, &ksig->info);
2233 			if (!signr)
2234 				continue;
2235 		}
2236 
2237 		ka = &sighand->action[signr-1];
2238 
2239 		/* Trace actually delivered signals. */
2240 		trace_signal_deliver(signr, &ksig->info, ka);
2241 
2242 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2243 			continue;
2244 		if (ka->sa.sa_handler != SIG_DFL) {
2245 			/* Run the handler.  */
2246 			ksig->ka = *ka;
2247 
2248 			if (ka->sa.sa_flags & SA_ONESHOT)
2249 				ka->sa.sa_handler = SIG_DFL;
2250 
2251 			break; /* will return non-zero "signr" value */
2252 		}
2253 
2254 		/*
2255 		 * Now we are doing the default action for this signal.
2256 		 */
2257 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2258 			continue;
2259 
2260 		/*
2261 		 * Global init gets no signals it doesn't want.
2262 		 * Container-init gets no signals it doesn't want from same
2263 		 * container.
2264 		 *
2265 		 * Note that if global/container-init sees a sig_kernel_only()
2266 		 * signal here, the signal must have been generated internally
2267 		 * or must have come from an ancestor namespace. In either
2268 		 * case, the signal cannot be dropped.
2269 		 */
2270 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2271 				!sig_kernel_only(signr))
2272 			continue;
2273 
2274 		if (sig_kernel_stop(signr)) {
2275 			/*
2276 			 * The default action is to stop all threads in
2277 			 * the thread group.  The job control signals
2278 			 * do nothing in an orphaned pgrp, but SIGSTOP
2279 			 * always works.  Note that siglock needs to be
2280 			 * dropped during the call to is_orphaned_pgrp()
2281 			 * because of lock ordering with tasklist_lock.
2282 			 * This allows an intervening SIGCONT to be posted.
2283 			 * We need to check for that and bail out if necessary.
2284 			 */
2285 			if (signr != SIGSTOP) {
2286 				spin_unlock_irq(&sighand->siglock);
2287 
2288 				/* signals can be posted during this window */
2289 
2290 				if (is_current_pgrp_orphaned())
2291 					goto relock;
2292 
2293 				spin_lock_irq(&sighand->siglock);
2294 			}
2295 
2296 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2297 				/* It released the siglock.  */
2298 				goto relock;
2299 			}
2300 
2301 			/*
2302 			 * We didn't actually stop, due to a race
2303 			 * with SIGCONT or something like that.
2304 			 */
2305 			continue;
2306 		}
2307 
2308 		spin_unlock_irq(&sighand->siglock);
2309 
2310 		/*
2311 		 * Anything else is fatal, maybe with a core dump.
2312 		 */
2313 		current->flags |= PF_SIGNALED;
2314 
2315 		if (sig_kernel_coredump(signr)) {
2316 			if (print_fatal_signals)
2317 				print_fatal_signal(ksig->info.si_signo);
2318 			proc_coredump_connector(current);
2319 			/*
2320 			 * If it was able to dump core, this kills all
2321 			 * other threads in the group and synchronizes with
2322 			 * their demise.  If we lost the race with another
2323 			 * thread getting here, it set group_exit_code
2324 			 * first and our do_group_exit call below will use
2325 			 * that value and ignore the one we pass it.
2326 			 */
2327 			do_coredump(&ksig->info);
2328 		}
2329 
2330 		/*
2331 		 * Death signals, no core dump.
2332 		 */
2333 		do_group_exit(ksig->info.si_signo);
2334 		/* NOTREACHED */
2335 	}
2336 	spin_unlock_irq(&sighand->siglock);
2337 
2338 	ksig->sig = signr;
2339 	return ksig->sig > 0;
2340 }
2341 
2342 /**
2343  * signal_delivered -
2344  * @ksig:		kernel signal struct
2345  * @stepping:		nonzero if debugger single-step or block-step in use
2346  *
2347  * This function should be called when a signal has successfully been
2348  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2349  * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2350  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2351  */
2352 static void signal_delivered(struct ksignal *ksig, int stepping)
2353 {
2354 	sigset_t blocked;
2355 
2356 	/* A signal was successfully delivered, and the
2357 	   saved sigmask was stored on the signal frame,
2358 	   and will be restored by sigreturn.  So we can
2359 	   simply clear the restore sigmask flag.  */
2360 	clear_restore_sigmask();
2361 
2362 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2363 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2364 		sigaddset(&blocked, ksig->sig);
2365 	set_current_blocked(&blocked);
2366 	tracehook_signal_handler(stepping);
2367 }
2368 
2369 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2370 {
2371 	if (failed)
2372 		force_sigsegv(ksig->sig, current);
2373 	else
2374 		signal_delivered(ksig, stepping);
2375 }
2376 
2377 /*
2378  * It could be that complete_signal() picked us to notify about the
2379  * group-wide signal. Other threads should be notified now to take
2380  * the shared signals in @which since we will not.
2381  */
2382 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2383 {
2384 	sigset_t retarget;
2385 	struct task_struct *t;
2386 
2387 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2388 	if (sigisemptyset(&retarget))
2389 		return;
2390 
2391 	t = tsk;
2392 	while_each_thread(tsk, t) {
2393 		if (t->flags & PF_EXITING)
2394 			continue;
2395 
2396 		if (!has_pending_signals(&retarget, &t->blocked))
2397 			continue;
2398 		/* Remove the signals this thread can handle. */
2399 		sigandsets(&retarget, &retarget, &t->blocked);
2400 
2401 		if (!signal_pending(t))
2402 			signal_wake_up(t, 0);
2403 
2404 		if (sigisemptyset(&retarget))
2405 			break;
2406 	}
2407 }
2408 
2409 void exit_signals(struct task_struct *tsk)
2410 {
2411 	int group_stop = 0;
2412 	sigset_t unblocked;
2413 
2414 	/*
2415 	 * @tsk is about to have PF_EXITING set - lock out users which
2416 	 * expect stable threadgroup.
2417 	 */
2418 	cgroup_threadgroup_change_begin(tsk);
2419 
2420 	if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2421 		tsk->flags |= PF_EXITING;
2422 		cgroup_threadgroup_change_end(tsk);
2423 		return;
2424 	}
2425 
2426 	spin_lock_irq(&tsk->sighand->siglock);
2427 	/*
2428 	 * From now this task is not visible for group-wide signals,
2429 	 * see wants_signal(), do_signal_stop().
2430 	 */
2431 	tsk->flags |= PF_EXITING;
2432 
2433 	cgroup_threadgroup_change_end(tsk);
2434 
2435 	if (!signal_pending(tsk))
2436 		goto out;
2437 
2438 	unblocked = tsk->blocked;
2439 	signotset(&unblocked);
2440 	retarget_shared_pending(tsk, &unblocked);
2441 
2442 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2443 	    task_participate_group_stop(tsk))
2444 		group_stop = CLD_STOPPED;
2445 out:
2446 	spin_unlock_irq(&tsk->sighand->siglock);
2447 
2448 	/*
2449 	 * If group stop has completed, deliver the notification.  This
2450 	 * should always go to the real parent of the group leader.
2451 	 */
2452 	if (unlikely(group_stop)) {
2453 		read_lock(&tasklist_lock);
2454 		do_notify_parent_cldstop(tsk, false, group_stop);
2455 		read_unlock(&tasklist_lock);
2456 	}
2457 }
2458 
2459 EXPORT_SYMBOL(recalc_sigpending);
2460 EXPORT_SYMBOL_GPL(dequeue_signal);
2461 EXPORT_SYMBOL(flush_signals);
2462 EXPORT_SYMBOL(force_sig);
2463 EXPORT_SYMBOL(send_sig);
2464 EXPORT_SYMBOL(send_sig_info);
2465 EXPORT_SYMBOL(sigprocmask);
2466 
2467 /*
2468  * System call entry points.
2469  */
2470 
2471 /**
2472  *  sys_restart_syscall - restart a system call
2473  */
2474 SYSCALL_DEFINE0(restart_syscall)
2475 {
2476 	struct restart_block *restart = &current->restart_block;
2477 	return restart->fn(restart);
2478 }
2479 
2480 long do_no_restart_syscall(struct restart_block *param)
2481 {
2482 	return -EINTR;
2483 }
2484 
2485 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2486 {
2487 	if (signal_pending(tsk) && !thread_group_empty(tsk)) {
2488 		sigset_t newblocked;
2489 		/* A set of now blocked but previously unblocked signals. */
2490 		sigandnsets(&newblocked, newset, &current->blocked);
2491 		retarget_shared_pending(tsk, &newblocked);
2492 	}
2493 	tsk->blocked = *newset;
2494 	recalc_sigpending();
2495 }
2496 
2497 /**
2498  * set_current_blocked - change current->blocked mask
2499  * @newset: new mask
2500  *
2501  * It is wrong to change ->blocked directly, this helper should be used
2502  * to ensure the process can't miss a shared signal we are going to block.
2503  */
2504 void set_current_blocked(sigset_t *newset)
2505 {
2506 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2507 	__set_current_blocked(newset);
2508 }
2509 
2510 void __set_current_blocked(const sigset_t *newset)
2511 {
2512 	struct task_struct *tsk = current;
2513 
2514 	/*
2515 	 * In case the signal mask hasn't changed, there is nothing we need
2516 	 * to do. The current->blocked shouldn't be modified by other task.
2517 	 */
2518 	if (sigequalsets(&tsk->blocked, newset))
2519 		return;
2520 
2521 	spin_lock_irq(&tsk->sighand->siglock);
2522 	__set_task_blocked(tsk, newset);
2523 	spin_unlock_irq(&tsk->sighand->siglock);
2524 }
2525 
2526 /*
2527  * This is also useful for kernel threads that want to temporarily
2528  * (or permanently) block certain signals.
2529  *
2530  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2531  * interface happily blocks "unblockable" signals like SIGKILL
2532  * and friends.
2533  */
2534 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2535 {
2536 	struct task_struct *tsk = current;
2537 	sigset_t newset;
2538 
2539 	/* Lockless, only current can change ->blocked, never from irq */
2540 	if (oldset)
2541 		*oldset = tsk->blocked;
2542 
2543 	switch (how) {
2544 	case SIG_BLOCK:
2545 		sigorsets(&newset, &tsk->blocked, set);
2546 		break;
2547 	case SIG_UNBLOCK:
2548 		sigandnsets(&newset, &tsk->blocked, set);
2549 		break;
2550 	case SIG_SETMASK:
2551 		newset = *set;
2552 		break;
2553 	default:
2554 		return -EINVAL;
2555 	}
2556 
2557 	__set_current_blocked(&newset);
2558 	return 0;
2559 }
2560 
2561 /**
2562  *  sys_rt_sigprocmask - change the list of currently blocked signals
2563  *  @how: whether to add, remove, or set signals
2564  *  @nset: stores pending signals
2565  *  @oset: previous value of signal mask if non-null
2566  *  @sigsetsize: size of sigset_t type
2567  */
2568 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
2569 		sigset_t __user *, oset, size_t, sigsetsize)
2570 {
2571 	sigset_t old_set, new_set;
2572 	int error;
2573 
2574 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2575 	if (sigsetsize != sizeof(sigset_t))
2576 		return -EINVAL;
2577 
2578 	old_set = current->blocked;
2579 
2580 	if (nset) {
2581 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
2582 			return -EFAULT;
2583 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2584 
2585 		error = sigprocmask(how, &new_set, NULL);
2586 		if (error)
2587 			return error;
2588 	}
2589 
2590 	if (oset) {
2591 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
2592 			return -EFAULT;
2593 	}
2594 
2595 	return 0;
2596 }
2597 
2598 #ifdef CONFIG_COMPAT
2599 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
2600 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
2601 {
2602 #ifdef __BIG_ENDIAN
2603 	sigset_t old_set = current->blocked;
2604 
2605 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2606 	if (sigsetsize != sizeof(sigset_t))
2607 		return -EINVAL;
2608 
2609 	if (nset) {
2610 		compat_sigset_t new32;
2611 		sigset_t new_set;
2612 		int error;
2613 		if (copy_from_user(&new32, nset, sizeof(compat_sigset_t)))
2614 			return -EFAULT;
2615 
2616 		sigset_from_compat(&new_set, &new32);
2617 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2618 
2619 		error = sigprocmask(how, &new_set, NULL);
2620 		if (error)
2621 			return error;
2622 	}
2623 	if (oset) {
2624 		compat_sigset_t old32;
2625 		sigset_to_compat(&old32, &old_set);
2626 		if (copy_to_user(oset, &old32, sizeof(compat_sigset_t)))
2627 			return -EFAULT;
2628 	}
2629 	return 0;
2630 #else
2631 	return sys_rt_sigprocmask(how, (sigset_t __user *)nset,
2632 				  (sigset_t __user *)oset, sigsetsize);
2633 #endif
2634 }
2635 #endif
2636 
2637 static int do_sigpending(void *set, unsigned long sigsetsize)
2638 {
2639 	if (sigsetsize > sizeof(sigset_t))
2640 		return -EINVAL;
2641 
2642 	spin_lock_irq(&current->sighand->siglock);
2643 	sigorsets(set, &current->pending.signal,
2644 		  &current->signal->shared_pending.signal);
2645 	spin_unlock_irq(&current->sighand->siglock);
2646 
2647 	/* Outside the lock because only this thread touches it.  */
2648 	sigandsets(set, &current->blocked, set);
2649 	return 0;
2650 }
2651 
2652 /**
2653  *  sys_rt_sigpending - examine a pending signal that has been raised
2654  *			while blocked
2655  *  @uset: stores pending signals
2656  *  @sigsetsize: size of sigset_t type or larger
2657  */
2658 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
2659 {
2660 	sigset_t set;
2661 	int err = do_sigpending(&set, sigsetsize);
2662 	if (!err && copy_to_user(uset, &set, sigsetsize))
2663 		err = -EFAULT;
2664 	return err;
2665 }
2666 
2667 #ifdef CONFIG_COMPAT
2668 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
2669 		compat_size_t, sigsetsize)
2670 {
2671 #ifdef __BIG_ENDIAN
2672 	sigset_t set;
2673 	int err = do_sigpending(&set, sigsetsize);
2674 	if (!err) {
2675 		compat_sigset_t set32;
2676 		sigset_to_compat(&set32, &set);
2677 		/* we can get here only if sigsetsize <= sizeof(set) */
2678 		if (copy_to_user(uset, &set32, sigsetsize))
2679 			err = -EFAULT;
2680 	}
2681 	return err;
2682 #else
2683 	return sys_rt_sigpending((sigset_t __user *)uset, sigsetsize);
2684 #endif
2685 }
2686 #endif
2687 
2688 enum siginfo_layout siginfo_layout(int sig, int si_code)
2689 {
2690 	enum siginfo_layout layout = SIL_KILL;
2691 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
2692 		static const struct {
2693 			unsigned char limit, layout;
2694 		} filter[] = {
2695 			[SIGILL]  = { NSIGILL,  SIL_FAULT },
2696 			[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
2697 			[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
2698 			[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
2699 			[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
2700 #if defined(SIGEMT) && defined(NSIGEMT)
2701 			[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
2702 #endif
2703 			[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
2704 			[SIGPOLL] = { NSIGPOLL, SIL_POLL },
2705 #ifdef __ARCH_SIGSYS
2706 			[SIGSYS]  = { NSIGSYS,  SIL_SYS },
2707 #endif
2708 		};
2709 		if ((sig < ARRAY_SIZE(filter)) && (si_code <= filter[sig].limit))
2710 			layout = filter[sig].layout;
2711 		else if (si_code <= NSIGPOLL)
2712 			layout = SIL_POLL;
2713 	} else {
2714 		if (si_code == SI_TIMER)
2715 			layout = SIL_TIMER;
2716 		else if (si_code == SI_SIGIO)
2717 			layout = SIL_POLL;
2718 		else if (si_code < 0)
2719 			layout = SIL_RT;
2720 		/* Tests to support buggy kernel ABIs */
2721 #ifdef TRAP_FIXME
2722 		if ((sig == SIGTRAP) && (si_code == TRAP_FIXME))
2723 			layout = SIL_FAULT;
2724 #endif
2725 #ifdef FPE_FIXME
2726 		if ((sig == SIGFPE) && (si_code == FPE_FIXME))
2727 			layout = SIL_FAULT;
2728 #endif
2729 	}
2730 	return layout;
2731 }
2732 
2733 #ifndef HAVE_ARCH_COPY_SIGINFO_TO_USER
2734 
2735 int copy_siginfo_to_user(siginfo_t __user *to, const siginfo_t *from)
2736 {
2737 	int err;
2738 
2739 	if (!access_ok (VERIFY_WRITE, to, sizeof(siginfo_t)))
2740 		return -EFAULT;
2741 	if (from->si_code < 0)
2742 		return __copy_to_user(to, from, sizeof(siginfo_t))
2743 			? -EFAULT : 0;
2744 	/*
2745 	 * If you change siginfo_t structure, please be sure
2746 	 * this code is fixed accordingly.
2747 	 * Please remember to update the signalfd_copyinfo() function
2748 	 * inside fs/signalfd.c too, in case siginfo_t changes.
2749 	 * It should never copy any pad contained in the structure
2750 	 * to avoid security leaks, but must copy the generic
2751 	 * 3 ints plus the relevant union member.
2752 	 */
2753 	err = __put_user(from->si_signo, &to->si_signo);
2754 	err |= __put_user(from->si_errno, &to->si_errno);
2755 	err |= __put_user(from->si_code, &to->si_code);
2756 	switch (siginfo_layout(from->si_signo, from->si_code)) {
2757 	case SIL_KILL:
2758 		err |= __put_user(from->si_pid, &to->si_pid);
2759 		err |= __put_user(from->si_uid, &to->si_uid);
2760 		break;
2761 	case SIL_TIMER:
2762 		/* Unreached SI_TIMER is negative */
2763 		break;
2764 	case SIL_POLL:
2765 		err |= __put_user(from->si_band, &to->si_band);
2766 		err |= __put_user(from->si_fd, &to->si_fd);
2767 		break;
2768 	case SIL_FAULT:
2769 		err |= __put_user(from->si_addr, &to->si_addr);
2770 #ifdef __ARCH_SI_TRAPNO
2771 		err |= __put_user(from->si_trapno, &to->si_trapno);
2772 #endif
2773 #ifdef BUS_MCEERR_AO
2774 		/*
2775 		 * Other callers might not initialize the si_lsb field,
2776 		 * so check explicitly for the right codes here.
2777 		 */
2778 		if (from->si_signo == SIGBUS &&
2779 		    (from->si_code == BUS_MCEERR_AR || from->si_code == BUS_MCEERR_AO))
2780 			err |= __put_user(from->si_addr_lsb, &to->si_addr_lsb);
2781 #endif
2782 #ifdef SEGV_BNDERR
2783 		if (from->si_signo == SIGSEGV && from->si_code == SEGV_BNDERR) {
2784 			err |= __put_user(from->si_lower, &to->si_lower);
2785 			err |= __put_user(from->si_upper, &to->si_upper);
2786 		}
2787 #endif
2788 #ifdef SEGV_PKUERR
2789 		if (from->si_signo == SIGSEGV && from->si_code == SEGV_PKUERR)
2790 			err |= __put_user(from->si_pkey, &to->si_pkey);
2791 #endif
2792 		break;
2793 	case SIL_CHLD:
2794 		err |= __put_user(from->si_pid, &to->si_pid);
2795 		err |= __put_user(from->si_uid, &to->si_uid);
2796 		err |= __put_user(from->si_status, &to->si_status);
2797 		err |= __put_user(from->si_utime, &to->si_utime);
2798 		err |= __put_user(from->si_stime, &to->si_stime);
2799 		break;
2800 	case SIL_RT:
2801 		err |= __put_user(from->si_pid, &to->si_pid);
2802 		err |= __put_user(from->si_uid, &to->si_uid);
2803 		err |= __put_user(from->si_ptr, &to->si_ptr);
2804 		break;
2805 #ifdef __ARCH_SIGSYS
2806 	case SIL_SYS:
2807 		err |= __put_user(from->si_call_addr, &to->si_call_addr);
2808 		err |= __put_user(from->si_syscall, &to->si_syscall);
2809 		err |= __put_user(from->si_arch, &to->si_arch);
2810 		break;
2811 #endif
2812 	}
2813 	return err;
2814 }
2815 
2816 #endif
2817 
2818 /**
2819  *  do_sigtimedwait - wait for queued signals specified in @which
2820  *  @which: queued signals to wait for
2821  *  @info: if non-null, the signal's siginfo is returned here
2822  *  @ts: upper bound on process time suspension
2823  */
2824 static int do_sigtimedwait(const sigset_t *which, siginfo_t *info,
2825 		    const struct timespec *ts)
2826 {
2827 	ktime_t *to = NULL, timeout = KTIME_MAX;
2828 	struct task_struct *tsk = current;
2829 	sigset_t mask = *which;
2830 	int sig, ret = 0;
2831 
2832 	if (ts) {
2833 		if (!timespec_valid(ts))
2834 			return -EINVAL;
2835 		timeout = timespec_to_ktime(*ts);
2836 		to = &timeout;
2837 	}
2838 
2839 	/*
2840 	 * Invert the set of allowed signals to get those we want to block.
2841 	 */
2842 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
2843 	signotset(&mask);
2844 
2845 	spin_lock_irq(&tsk->sighand->siglock);
2846 	sig = dequeue_signal(tsk, &mask, info);
2847 	if (!sig && timeout) {
2848 		/*
2849 		 * None ready, temporarily unblock those we're interested
2850 		 * while we are sleeping in so that we'll be awakened when
2851 		 * they arrive. Unblocking is always fine, we can avoid
2852 		 * set_current_blocked().
2853 		 */
2854 		tsk->real_blocked = tsk->blocked;
2855 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
2856 		recalc_sigpending();
2857 		spin_unlock_irq(&tsk->sighand->siglock);
2858 
2859 		__set_current_state(TASK_INTERRUPTIBLE);
2860 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
2861 							 HRTIMER_MODE_REL);
2862 		spin_lock_irq(&tsk->sighand->siglock);
2863 		__set_task_blocked(tsk, &tsk->real_blocked);
2864 		sigemptyset(&tsk->real_blocked);
2865 		sig = dequeue_signal(tsk, &mask, info);
2866 	}
2867 	spin_unlock_irq(&tsk->sighand->siglock);
2868 
2869 	if (sig)
2870 		return sig;
2871 	return ret ? -EINTR : -EAGAIN;
2872 }
2873 
2874 /**
2875  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
2876  *			in @uthese
2877  *  @uthese: queued signals to wait for
2878  *  @uinfo: if non-null, the signal's siginfo is returned here
2879  *  @uts: upper bound on process time suspension
2880  *  @sigsetsize: size of sigset_t type
2881  */
2882 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
2883 		siginfo_t __user *, uinfo, const struct timespec __user *, uts,
2884 		size_t, sigsetsize)
2885 {
2886 	sigset_t these;
2887 	struct timespec ts;
2888 	siginfo_t info;
2889 	int ret;
2890 
2891 	/* XXX: Don't preclude handling different sized sigset_t's.  */
2892 	if (sigsetsize != sizeof(sigset_t))
2893 		return -EINVAL;
2894 
2895 	if (copy_from_user(&these, uthese, sizeof(these)))
2896 		return -EFAULT;
2897 
2898 	if (uts) {
2899 		if (copy_from_user(&ts, uts, sizeof(ts)))
2900 			return -EFAULT;
2901 	}
2902 
2903 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
2904 
2905 	if (ret > 0 && uinfo) {
2906 		if (copy_siginfo_to_user(uinfo, &info))
2907 			ret = -EFAULT;
2908 	}
2909 
2910 	return ret;
2911 }
2912 
2913 #ifdef CONFIG_COMPAT
2914 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait, compat_sigset_t __user *, uthese,
2915 		struct compat_siginfo __user *, uinfo,
2916 		struct compat_timespec __user *, uts, compat_size_t, sigsetsize)
2917 {
2918 	compat_sigset_t s32;
2919 	sigset_t s;
2920 	struct timespec t;
2921 	siginfo_t info;
2922 	long ret;
2923 
2924 	if (sigsetsize != sizeof(sigset_t))
2925 		return -EINVAL;
2926 
2927 	if (copy_from_user(&s32, uthese, sizeof(compat_sigset_t)))
2928 		return -EFAULT;
2929 	sigset_from_compat(&s, &s32);
2930 
2931 	if (uts) {
2932 		if (compat_get_timespec(&t, uts))
2933 			return -EFAULT;
2934 	}
2935 
2936 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
2937 
2938 	if (ret > 0 && uinfo) {
2939 		if (copy_siginfo_to_user32(uinfo, &info))
2940 			ret = -EFAULT;
2941 	}
2942 
2943 	return ret;
2944 }
2945 #endif
2946 
2947 /**
2948  *  sys_kill - send a signal to a process
2949  *  @pid: the PID of the process
2950  *  @sig: signal to be sent
2951  */
2952 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
2953 {
2954 	struct siginfo info;
2955 
2956 	info.si_signo = sig;
2957 	info.si_errno = 0;
2958 	info.si_code = SI_USER;
2959 	info.si_pid = task_tgid_vnr(current);
2960 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2961 
2962 	return kill_something_info(sig, &info, pid);
2963 }
2964 
2965 static int
2966 do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info)
2967 {
2968 	struct task_struct *p;
2969 	int error = -ESRCH;
2970 
2971 	rcu_read_lock();
2972 	p = find_task_by_vpid(pid);
2973 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
2974 		error = check_kill_permission(sig, info, p);
2975 		/*
2976 		 * The null signal is a permissions and process existence
2977 		 * probe.  No signal is actually delivered.
2978 		 */
2979 		if (!error && sig) {
2980 			error = do_send_sig_info(sig, info, p, false);
2981 			/*
2982 			 * If lock_task_sighand() failed we pretend the task
2983 			 * dies after receiving the signal. The window is tiny,
2984 			 * and the signal is private anyway.
2985 			 */
2986 			if (unlikely(error == -ESRCH))
2987 				error = 0;
2988 		}
2989 	}
2990 	rcu_read_unlock();
2991 
2992 	return error;
2993 }
2994 
2995 static int do_tkill(pid_t tgid, pid_t pid, int sig)
2996 {
2997 	struct siginfo info = {};
2998 
2999 	info.si_signo = sig;
3000 	info.si_errno = 0;
3001 	info.si_code = SI_TKILL;
3002 	info.si_pid = task_tgid_vnr(current);
3003 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3004 
3005 	return do_send_specific(tgid, pid, sig, &info);
3006 }
3007 
3008 /**
3009  *  sys_tgkill - send signal to one specific thread
3010  *  @tgid: the thread group ID of the thread
3011  *  @pid: the PID of the thread
3012  *  @sig: signal to be sent
3013  *
3014  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3015  *  exists but it's not belonging to the target process anymore. This
3016  *  method solves the problem of threads exiting and PIDs getting reused.
3017  */
3018 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3019 {
3020 	/* This is only valid for single tasks */
3021 	if (pid <= 0 || tgid <= 0)
3022 		return -EINVAL;
3023 
3024 	return do_tkill(tgid, pid, sig);
3025 }
3026 
3027 /**
3028  *  sys_tkill - send signal to one specific task
3029  *  @pid: the PID of the task
3030  *  @sig: signal to be sent
3031  *
3032  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3033  */
3034 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3035 {
3036 	/* This is only valid for single tasks */
3037 	if (pid <= 0)
3038 		return -EINVAL;
3039 
3040 	return do_tkill(0, pid, sig);
3041 }
3042 
3043 static int do_rt_sigqueueinfo(pid_t pid, int sig, siginfo_t *info)
3044 {
3045 	/* Not even root can pretend to send signals from the kernel.
3046 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3047 	 */
3048 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3049 	    (task_pid_vnr(current) != pid))
3050 		return -EPERM;
3051 
3052 	info->si_signo = sig;
3053 
3054 	/* POSIX.1b doesn't mention process groups.  */
3055 	return kill_proc_info(sig, info, pid);
3056 }
3057 
3058 /**
3059  *  sys_rt_sigqueueinfo - send signal information to a signal
3060  *  @pid: the PID of the thread
3061  *  @sig: signal to be sent
3062  *  @uinfo: signal info to be sent
3063  */
3064 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
3065 		siginfo_t __user *, uinfo)
3066 {
3067 	siginfo_t info;
3068 	if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
3069 		return -EFAULT;
3070 	return do_rt_sigqueueinfo(pid, sig, &info);
3071 }
3072 
3073 #ifdef CONFIG_COMPAT
3074 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
3075 			compat_pid_t, pid,
3076 			int, sig,
3077 			struct compat_siginfo __user *, uinfo)
3078 {
3079 	siginfo_t info = {};
3080 	int ret = copy_siginfo_from_user32(&info, uinfo);
3081 	if (unlikely(ret))
3082 		return ret;
3083 	return do_rt_sigqueueinfo(pid, sig, &info);
3084 }
3085 #endif
3086 
3087 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, siginfo_t *info)
3088 {
3089 	/* This is only valid for single tasks */
3090 	if (pid <= 0 || tgid <= 0)
3091 		return -EINVAL;
3092 
3093 	/* Not even root can pretend to send signals from the kernel.
3094 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3095 	 */
3096 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3097 	    (task_pid_vnr(current) != pid))
3098 		return -EPERM;
3099 
3100 	info->si_signo = sig;
3101 
3102 	return do_send_specific(tgid, pid, sig, info);
3103 }
3104 
3105 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
3106 		siginfo_t __user *, uinfo)
3107 {
3108 	siginfo_t info;
3109 
3110 	if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
3111 		return -EFAULT;
3112 
3113 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3114 }
3115 
3116 #ifdef CONFIG_COMPAT
3117 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
3118 			compat_pid_t, tgid,
3119 			compat_pid_t, pid,
3120 			int, sig,
3121 			struct compat_siginfo __user *, uinfo)
3122 {
3123 	siginfo_t info = {};
3124 
3125 	if (copy_siginfo_from_user32(&info, uinfo))
3126 		return -EFAULT;
3127 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3128 }
3129 #endif
3130 
3131 /*
3132  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
3133  */
3134 void kernel_sigaction(int sig, __sighandler_t action)
3135 {
3136 	spin_lock_irq(&current->sighand->siglock);
3137 	current->sighand->action[sig - 1].sa.sa_handler = action;
3138 	if (action == SIG_IGN) {
3139 		sigset_t mask;
3140 
3141 		sigemptyset(&mask);
3142 		sigaddset(&mask, sig);
3143 
3144 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
3145 		flush_sigqueue_mask(&mask, &current->pending);
3146 		recalc_sigpending();
3147 	}
3148 	spin_unlock_irq(&current->sighand->siglock);
3149 }
3150 EXPORT_SYMBOL(kernel_sigaction);
3151 
3152 void __weak sigaction_compat_abi(struct k_sigaction *act,
3153 		struct k_sigaction *oact)
3154 {
3155 }
3156 
3157 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
3158 {
3159 	struct task_struct *p = current, *t;
3160 	struct k_sigaction *k;
3161 	sigset_t mask;
3162 
3163 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
3164 		return -EINVAL;
3165 
3166 	k = &p->sighand->action[sig-1];
3167 
3168 	spin_lock_irq(&p->sighand->siglock);
3169 	if (oact)
3170 		*oact = *k;
3171 
3172 	sigaction_compat_abi(act, oact);
3173 
3174 	if (act) {
3175 		sigdelsetmask(&act->sa.sa_mask,
3176 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
3177 		*k = *act;
3178 		/*
3179 		 * POSIX 3.3.1.3:
3180 		 *  "Setting a signal action to SIG_IGN for a signal that is
3181 		 *   pending shall cause the pending signal to be discarded,
3182 		 *   whether or not it is blocked."
3183 		 *
3184 		 *  "Setting a signal action to SIG_DFL for a signal that is
3185 		 *   pending and whose default action is to ignore the signal
3186 		 *   (for example, SIGCHLD), shall cause the pending signal to
3187 		 *   be discarded, whether or not it is blocked"
3188 		 */
3189 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
3190 			sigemptyset(&mask);
3191 			sigaddset(&mask, sig);
3192 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
3193 			for_each_thread(p, t)
3194 				flush_sigqueue_mask(&mask, &t->pending);
3195 		}
3196 	}
3197 
3198 	spin_unlock_irq(&p->sighand->siglock);
3199 	return 0;
3200 }
3201 
3202 static int
3203 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp)
3204 {
3205 	struct task_struct *t = current;
3206 
3207 	if (oss) {
3208 		memset(oss, 0, sizeof(stack_t));
3209 		oss->ss_sp = (void __user *) t->sas_ss_sp;
3210 		oss->ss_size = t->sas_ss_size;
3211 		oss->ss_flags = sas_ss_flags(sp) |
3212 			(current->sas_ss_flags & SS_FLAG_BITS);
3213 	}
3214 
3215 	if (ss) {
3216 		void __user *ss_sp = ss->ss_sp;
3217 		size_t ss_size = ss->ss_size;
3218 		unsigned ss_flags = ss->ss_flags;
3219 		int ss_mode;
3220 
3221 		if (unlikely(on_sig_stack(sp)))
3222 			return -EPERM;
3223 
3224 		ss_mode = ss_flags & ~SS_FLAG_BITS;
3225 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
3226 				ss_mode != 0))
3227 			return -EINVAL;
3228 
3229 		if (ss_mode == SS_DISABLE) {
3230 			ss_size = 0;
3231 			ss_sp = NULL;
3232 		} else {
3233 			if (unlikely(ss_size < MINSIGSTKSZ))
3234 				return -ENOMEM;
3235 		}
3236 
3237 		t->sas_ss_sp = (unsigned long) ss_sp;
3238 		t->sas_ss_size = ss_size;
3239 		t->sas_ss_flags = ss_flags;
3240 	}
3241 	return 0;
3242 }
3243 
3244 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
3245 {
3246 	stack_t new, old;
3247 	int err;
3248 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
3249 		return -EFAULT;
3250 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
3251 			      current_user_stack_pointer());
3252 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
3253 		err = -EFAULT;
3254 	return err;
3255 }
3256 
3257 int restore_altstack(const stack_t __user *uss)
3258 {
3259 	stack_t new;
3260 	if (copy_from_user(&new, uss, sizeof(stack_t)))
3261 		return -EFAULT;
3262 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer());
3263 	/* squash all but EFAULT for now */
3264 	return 0;
3265 }
3266 
3267 int __save_altstack(stack_t __user *uss, unsigned long sp)
3268 {
3269 	struct task_struct *t = current;
3270 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
3271 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
3272 		__put_user(t->sas_ss_size, &uss->ss_size);
3273 	if (err)
3274 		return err;
3275 	if (t->sas_ss_flags & SS_AUTODISARM)
3276 		sas_ss_reset(t);
3277 	return 0;
3278 }
3279 
3280 #ifdef CONFIG_COMPAT
3281 COMPAT_SYSCALL_DEFINE2(sigaltstack,
3282 			const compat_stack_t __user *, uss_ptr,
3283 			compat_stack_t __user *, uoss_ptr)
3284 {
3285 	stack_t uss, uoss;
3286 	int ret;
3287 
3288 	if (uss_ptr) {
3289 		compat_stack_t uss32;
3290 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
3291 			return -EFAULT;
3292 		uss.ss_sp = compat_ptr(uss32.ss_sp);
3293 		uss.ss_flags = uss32.ss_flags;
3294 		uss.ss_size = uss32.ss_size;
3295 	}
3296 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
3297 			     compat_user_stack_pointer());
3298 	if (ret >= 0 && uoss_ptr)  {
3299 		compat_stack_t old;
3300 		memset(&old, 0, sizeof(old));
3301 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
3302 		old.ss_flags = uoss.ss_flags;
3303 		old.ss_size = uoss.ss_size;
3304 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
3305 			ret = -EFAULT;
3306 	}
3307 	return ret;
3308 }
3309 
3310 int compat_restore_altstack(const compat_stack_t __user *uss)
3311 {
3312 	int err = compat_sys_sigaltstack(uss, NULL);
3313 	/* squash all but -EFAULT for now */
3314 	return err == -EFAULT ? err : 0;
3315 }
3316 
3317 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
3318 {
3319 	int err;
3320 	struct task_struct *t = current;
3321 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
3322 			 &uss->ss_sp) |
3323 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
3324 		__put_user(t->sas_ss_size, &uss->ss_size);
3325 	if (err)
3326 		return err;
3327 	if (t->sas_ss_flags & SS_AUTODISARM)
3328 		sas_ss_reset(t);
3329 	return 0;
3330 }
3331 #endif
3332 
3333 #ifdef __ARCH_WANT_SYS_SIGPENDING
3334 
3335 /**
3336  *  sys_sigpending - examine pending signals
3337  *  @set: where mask of pending signal is returned
3338  */
3339 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, set)
3340 {
3341 	return sys_rt_sigpending((sigset_t __user *)set, sizeof(old_sigset_t));
3342 }
3343 
3344 #ifdef CONFIG_COMPAT
3345 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
3346 {
3347 #ifdef __BIG_ENDIAN
3348 	sigset_t set;
3349 	int err = do_sigpending(&set, sizeof(set.sig[0]));
3350 	if (!err)
3351 		err = put_user(set.sig[0], set32);
3352 	return err;
3353 #else
3354 	return sys_rt_sigpending((sigset_t __user *)set32, sizeof(*set32));
3355 #endif
3356 }
3357 #endif
3358 
3359 #endif
3360 
3361 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
3362 /**
3363  *  sys_sigprocmask - examine and change blocked signals
3364  *  @how: whether to add, remove, or set signals
3365  *  @nset: signals to add or remove (if non-null)
3366  *  @oset: previous value of signal mask if non-null
3367  *
3368  * Some platforms have their own version with special arguments;
3369  * others support only sys_rt_sigprocmask.
3370  */
3371 
3372 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
3373 		old_sigset_t __user *, oset)
3374 {
3375 	old_sigset_t old_set, new_set;
3376 	sigset_t new_blocked;
3377 
3378 	old_set = current->blocked.sig[0];
3379 
3380 	if (nset) {
3381 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
3382 			return -EFAULT;
3383 
3384 		new_blocked = current->blocked;
3385 
3386 		switch (how) {
3387 		case SIG_BLOCK:
3388 			sigaddsetmask(&new_blocked, new_set);
3389 			break;
3390 		case SIG_UNBLOCK:
3391 			sigdelsetmask(&new_blocked, new_set);
3392 			break;
3393 		case SIG_SETMASK:
3394 			new_blocked.sig[0] = new_set;
3395 			break;
3396 		default:
3397 			return -EINVAL;
3398 		}
3399 
3400 		set_current_blocked(&new_blocked);
3401 	}
3402 
3403 	if (oset) {
3404 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
3405 			return -EFAULT;
3406 	}
3407 
3408 	return 0;
3409 }
3410 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
3411 
3412 #ifndef CONFIG_ODD_RT_SIGACTION
3413 /**
3414  *  sys_rt_sigaction - alter an action taken by a process
3415  *  @sig: signal to be sent
3416  *  @act: new sigaction
3417  *  @oact: used to save the previous sigaction
3418  *  @sigsetsize: size of sigset_t type
3419  */
3420 SYSCALL_DEFINE4(rt_sigaction, int, sig,
3421 		const struct sigaction __user *, act,
3422 		struct sigaction __user *, oact,
3423 		size_t, sigsetsize)
3424 {
3425 	struct k_sigaction new_sa, old_sa;
3426 	int ret = -EINVAL;
3427 
3428 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3429 	if (sigsetsize != sizeof(sigset_t))
3430 		goto out;
3431 
3432 	if (act) {
3433 		if (copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
3434 			return -EFAULT;
3435 	}
3436 
3437 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
3438 
3439 	if (!ret && oact) {
3440 		if (copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
3441 			return -EFAULT;
3442 	}
3443 out:
3444 	return ret;
3445 }
3446 #ifdef CONFIG_COMPAT
3447 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
3448 		const struct compat_sigaction __user *, act,
3449 		struct compat_sigaction __user *, oact,
3450 		compat_size_t, sigsetsize)
3451 {
3452 	struct k_sigaction new_ka, old_ka;
3453 	compat_sigset_t mask;
3454 #ifdef __ARCH_HAS_SA_RESTORER
3455 	compat_uptr_t restorer;
3456 #endif
3457 	int ret;
3458 
3459 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3460 	if (sigsetsize != sizeof(compat_sigset_t))
3461 		return -EINVAL;
3462 
3463 	if (act) {
3464 		compat_uptr_t handler;
3465 		ret = get_user(handler, &act->sa_handler);
3466 		new_ka.sa.sa_handler = compat_ptr(handler);
3467 #ifdef __ARCH_HAS_SA_RESTORER
3468 		ret |= get_user(restorer, &act->sa_restorer);
3469 		new_ka.sa.sa_restorer = compat_ptr(restorer);
3470 #endif
3471 		ret |= copy_from_user(&mask, &act->sa_mask, sizeof(mask));
3472 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
3473 		if (ret)
3474 			return -EFAULT;
3475 		sigset_from_compat(&new_ka.sa.sa_mask, &mask);
3476 	}
3477 
3478 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3479 	if (!ret && oact) {
3480 		sigset_to_compat(&mask, &old_ka.sa.sa_mask);
3481 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
3482 			       &oact->sa_handler);
3483 		ret |= copy_to_user(&oact->sa_mask, &mask, sizeof(mask));
3484 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
3485 #ifdef __ARCH_HAS_SA_RESTORER
3486 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
3487 				&oact->sa_restorer);
3488 #endif
3489 	}
3490 	return ret;
3491 }
3492 #endif
3493 #endif /* !CONFIG_ODD_RT_SIGACTION */
3494 
3495 #ifdef CONFIG_OLD_SIGACTION
3496 SYSCALL_DEFINE3(sigaction, int, sig,
3497 		const struct old_sigaction __user *, act,
3498 	        struct old_sigaction __user *, oact)
3499 {
3500 	struct k_sigaction new_ka, old_ka;
3501 	int ret;
3502 
3503 	if (act) {
3504 		old_sigset_t mask;
3505 		if (!access_ok(VERIFY_READ, act, sizeof(*act)) ||
3506 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
3507 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
3508 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
3509 		    __get_user(mask, &act->sa_mask))
3510 			return -EFAULT;
3511 #ifdef __ARCH_HAS_KA_RESTORER
3512 		new_ka.ka_restorer = NULL;
3513 #endif
3514 		siginitset(&new_ka.sa.sa_mask, mask);
3515 	}
3516 
3517 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3518 
3519 	if (!ret && oact) {
3520 		if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)) ||
3521 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
3522 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
3523 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
3524 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
3525 			return -EFAULT;
3526 	}
3527 
3528 	return ret;
3529 }
3530 #endif
3531 #ifdef CONFIG_COMPAT_OLD_SIGACTION
3532 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
3533 		const struct compat_old_sigaction __user *, act,
3534 	        struct compat_old_sigaction __user *, oact)
3535 {
3536 	struct k_sigaction new_ka, old_ka;
3537 	int ret;
3538 	compat_old_sigset_t mask;
3539 	compat_uptr_t handler, restorer;
3540 
3541 	if (act) {
3542 		if (!access_ok(VERIFY_READ, act, sizeof(*act)) ||
3543 		    __get_user(handler, &act->sa_handler) ||
3544 		    __get_user(restorer, &act->sa_restorer) ||
3545 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
3546 		    __get_user(mask, &act->sa_mask))
3547 			return -EFAULT;
3548 
3549 #ifdef __ARCH_HAS_KA_RESTORER
3550 		new_ka.ka_restorer = NULL;
3551 #endif
3552 		new_ka.sa.sa_handler = compat_ptr(handler);
3553 		new_ka.sa.sa_restorer = compat_ptr(restorer);
3554 		siginitset(&new_ka.sa.sa_mask, mask);
3555 	}
3556 
3557 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
3558 
3559 	if (!ret && oact) {
3560 		if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)) ||
3561 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
3562 			       &oact->sa_handler) ||
3563 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
3564 			       &oact->sa_restorer) ||
3565 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
3566 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
3567 			return -EFAULT;
3568 	}
3569 	return ret;
3570 }
3571 #endif
3572 
3573 #ifdef CONFIG_SGETMASK_SYSCALL
3574 
3575 /*
3576  * For backwards compatibility.  Functionality superseded by sigprocmask.
3577  */
3578 SYSCALL_DEFINE0(sgetmask)
3579 {
3580 	/* SMP safe */
3581 	return current->blocked.sig[0];
3582 }
3583 
3584 SYSCALL_DEFINE1(ssetmask, int, newmask)
3585 {
3586 	int old = current->blocked.sig[0];
3587 	sigset_t newset;
3588 
3589 	siginitset(&newset, newmask);
3590 	set_current_blocked(&newset);
3591 
3592 	return old;
3593 }
3594 #endif /* CONFIG_SGETMASK_SYSCALL */
3595 
3596 #ifdef __ARCH_WANT_SYS_SIGNAL
3597 /*
3598  * For backwards compatibility.  Functionality superseded by sigaction.
3599  */
3600 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
3601 {
3602 	struct k_sigaction new_sa, old_sa;
3603 	int ret;
3604 
3605 	new_sa.sa.sa_handler = handler;
3606 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
3607 	sigemptyset(&new_sa.sa.sa_mask);
3608 
3609 	ret = do_sigaction(sig, &new_sa, &old_sa);
3610 
3611 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
3612 }
3613 #endif /* __ARCH_WANT_SYS_SIGNAL */
3614 
3615 #ifdef __ARCH_WANT_SYS_PAUSE
3616 
3617 SYSCALL_DEFINE0(pause)
3618 {
3619 	while (!signal_pending(current)) {
3620 		__set_current_state(TASK_INTERRUPTIBLE);
3621 		schedule();
3622 	}
3623 	return -ERESTARTNOHAND;
3624 }
3625 
3626 #endif
3627 
3628 static int sigsuspend(sigset_t *set)
3629 {
3630 	current->saved_sigmask = current->blocked;
3631 	set_current_blocked(set);
3632 
3633 	while (!signal_pending(current)) {
3634 		__set_current_state(TASK_INTERRUPTIBLE);
3635 		schedule();
3636 	}
3637 	set_restore_sigmask();
3638 	return -ERESTARTNOHAND;
3639 }
3640 
3641 /**
3642  *  sys_rt_sigsuspend - replace the signal mask for a value with the
3643  *	@unewset value until a signal is received
3644  *  @unewset: new signal mask value
3645  *  @sigsetsize: size of sigset_t type
3646  */
3647 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
3648 {
3649 	sigset_t newset;
3650 
3651 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3652 	if (sigsetsize != sizeof(sigset_t))
3653 		return -EINVAL;
3654 
3655 	if (copy_from_user(&newset, unewset, sizeof(newset)))
3656 		return -EFAULT;
3657 	return sigsuspend(&newset);
3658 }
3659 
3660 #ifdef CONFIG_COMPAT
3661 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
3662 {
3663 #ifdef __BIG_ENDIAN
3664 	sigset_t newset;
3665 	compat_sigset_t newset32;
3666 
3667 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3668 	if (sigsetsize != sizeof(sigset_t))
3669 		return -EINVAL;
3670 
3671 	if (copy_from_user(&newset32, unewset, sizeof(compat_sigset_t)))
3672 		return -EFAULT;
3673 	sigset_from_compat(&newset, &newset32);
3674 	return sigsuspend(&newset);
3675 #else
3676 	/* on little-endian bitmaps don't care about granularity */
3677 	return sys_rt_sigsuspend((sigset_t __user *)unewset, sigsetsize);
3678 #endif
3679 }
3680 #endif
3681 
3682 #ifdef CONFIG_OLD_SIGSUSPEND
3683 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
3684 {
3685 	sigset_t blocked;
3686 	siginitset(&blocked, mask);
3687 	return sigsuspend(&blocked);
3688 }
3689 #endif
3690 #ifdef CONFIG_OLD_SIGSUSPEND3
3691 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
3692 {
3693 	sigset_t blocked;
3694 	siginitset(&blocked, mask);
3695 	return sigsuspend(&blocked);
3696 }
3697 #endif
3698 
3699 __weak const char *arch_vma_name(struct vm_area_struct *vma)
3700 {
3701 	return NULL;
3702 }
3703 
3704 void __init signals_init(void)
3705 {
3706 	/* If this check fails, the __ARCH_SI_PREAMBLE_SIZE value is wrong! */
3707 	BUILD_BUG_ON(__ARCH_SI_PREAMBLE_SIZE
3708 		!= offsetof(struct siginfo, _sifields._pad));
3709 
3710 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
3711 }
3712 
3713 #ifdef CONFIG_KGDB_KDB
3714 #include <linux/kdb.h>
3715 /*
3716  * kdb_send_sig_info - Allows kdb to send signals without exposing
3717  * signal internals.  This function checks if the required locks are
3718  * available before calling the main signal code, to avoid kdb
3719  * deadlocks.
3720  */
3721 void
3722 kdb_send_sig_info(struct task_struct *t, struct siginfo *info)
3723 {
3724 	static struct task_struct *kdb_prev_t;
3725 	int sig, new_t;
3726 	if (!spin_trylock(&t->sighand->siglock)) {
3727 		kdb_printf("Can't do kill command now.\n"
3728 			   "The sigmask lock is held somewhere else in "
3729 			   "kernel, try again later\n");
3730 		return;
3731 	}
3732 	spin_unlock(&t->sighand->siglock);
3733 	new_t = kdb_prev_t != t;
3734 	kdb_prev_t = t;
3735 	if (t->state != TASK_RUNNING && new_t) {
3736 		kdb_printf("Process is not RUNNING, sending a signal from "
3737 			   "kdb risks deadlock\n"
3738 			   "on the run queue locks. "
3739 			   "The signal has _not_ been sent.\n"
3740 			   "Reissue the kill command if you want to risk "
3741 			   "the deadlock.\n");
3742 		return;
3743 	}
3744 	sig = info->si_signo;
3745 	if (send_sig_info(sig, info, t))
3746 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
3747 			   sig, t->pid);
3748 	else
3749 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
3750 }
3751 #endif	/* CONFIG_KGDB_KDB */
3752