• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/kernel/signal.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  1997-11-02  Modified for POSIX.1b signals by Richard Henderson
8  *
9  *  2003-06-02  Jim Houston - Concurrent Computer Corp.
10  *		Changes to use preallocated sigqueue structures
11  *		to allow signals to be sent reliably.
12  */
13 
14 #include <linux/slab.h>
15 #include <linux/export.h>
16 #include <linux/init.h>
17 #include <linux/sched/mm.h>
18 #include <linux/sched/user.h>
19 #include <linux/sched/debug.h>
20 #include <linux/sched/task.h>
21 #include <linux/sched/task_stack.h>
22 #include <linux/sched/cputime.h>
23 #include <linux/file.h>
24 #include <linux/fs.h>
25 #include <linux/proc_fs.h>
26 #include <linux/tty.h>
27 #include <linux/binfmts.h>
28 #include <linux/coredump.h>
29 #include <linux/security.h>
30 #include <linux/syscalls.h>
31 #include <linux/ptrace.h>
32 #include <linux/signal.h>
33 #include <linux/signalfd.h>
34 #include <linux/ratelimit.h>
35 #include <linux/tracehook.h>
36 #include <linux/capability.h>
37 #include <linux/freezer.h>
38 #include <linux/pid_namespace.h>
39 #include <linux/nsproxy.h>
40 #include <linux/user_namespace.h>
41 #include <linux/uprobes.h>
42 #include <linux/compat.h>
43 #include <linux/cn_proc.h>
44 #include <linux/compiler.h>
45 #include <linux/posix-timers.h>
46 #include <linux/livepatch.h>
47 #include <linux/cgroup.h>
48 #include <linux/audit.h>
49 #include <linux/oom.h>
50 
51 #define CREATE_TRACE_POINTS
52 #include <trace/events/signal.h>
53 
54 #include <asm/param.h>
55 #include <linux/uaccess.h>
56 #include <asm/unistd.h>
57 #include <asm/siginfo.h>
58 #include <asm/cacheflush.h>
59 
60 #undef CREATE_TRACE_POINTS
61 #include <trace/hooks/signal.h>
62 /*
63  * SLAB caches for signal bits.
64  */
65 
66 static struct kmem_cache *sigqueue_cachep;
67 
68 int print_fatal_signals __read_mostly;
69 
sig_handler(struct task_struct * t,int sig)70 static void __user *sig_handler(struct task_struct *t, int sig)
71 {
72 	return t->sighand->action[sig - 1].sa.sa_handler;
73 }
74 
sig_handler_ignored(void __user * handler,int sig)75 static inline bool sig_handler_ignored(void __user *handler, int sig)
76 {
77 	/* Is it explicitly or implicitly ignored? */
78 	return handler == SIG_IGN ||
79 	       (handler == SIG_DFL && sig_kernel_ignore(sig));
80 }
81 
sig_task_ignored(struct task_struct * t,int sig,bool force)82 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
83 {
84 	void __user *handler;
85 
86 	handler = sig_handler(t, sig);
87 
88 	/* SIGKILL and SIGSTOP may not be sent to the global init */
89 	if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
90 		return true;
91 
92 	if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
93 	    handler == SIG_DFL && !(force && sig_kernel_only(sig)))
94 		return true;
95 
96 	/* Only allow kernel generated signals to this kthread */
97 	if (unlikely((t->flags & PF_KTHREAD) &&
98 		     (handler == SIG_KTHREAD_KERNEL) && !force))
99 		return true;
100 
101 	return sig_handler_ignored(handler, sig);
102 }
103 
sig_ignored(struct task_struct * t,int sig,bool force)104 static bool sig_ignored(struct task_struct *t, int sig, bool force)
105 {
106 	/*
107 	 * Blocked signals are never ignored, since the
108 	 * signal handler may change by the time it is
109 	 * unblocked.
110 	 */
111 	if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
112 		return false;
113 
114 	/*
115 	 * Tracers may want to know about even ignored signal unless it
116 	 * is SIGKILL which can't be reported anyway but can be ignored
117 	 * by SIGNAL_UNKILLABLE task.
118 	 */
119 	if (t->ptrace && sig != SIGKILL)
120 		return false;
121 
122 	return sig_task_ignored(t, sig, force);
123 }
124 
125 /*
126  * Re-calculate pending state from the set of locally pending
127  * signals, globally pending signals, and blocked signals.
128  */
has_pending_signals(sigset_t * signal,sigset_t * blocked)129 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
130 {
131 	unsigned long ready;
132 	long i;
133 
134 	switch (_NSIG_WORDS) {
135 	default:
136 		for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
137 			ready |= signal->sig[i] &~ blocked->sig[i];
138 		break;
139 
140 	case 4: ready  = signal->sig[3] &~ blocked->sig[3];
141 		ready |= signal->sig[2] &~ blocked->sig[2];
142 		ready |= signal->sig[1] &~ blocked->sig[1];
143 		ready |= signal->sig[0] &~ blocked->sig[0];
144 		break;
145 
146 	case 2: ready  = signal->sig[1] &~ blocked->sig[1];
147 		ready |= signal->sig[0] &~ blocked->sig[0];
148 		break;
149 
150 	case 1: ready  = signal->sig[0] &~ blocked->sig[0];
151 	}
152 	return ready !=	0;
153 }
154 
155 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
156 
recalc_sigpending_tsk(struct task_struct * t)157 static bool recalc_sigpending_tsk(struct task_struct *t)
158 {
159 	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
160 	    PENDING(&t->pending, &t->blocked) ||
161 	    PENDING(&t->signal->shared_pending, &t->blocked) ||
162 	    cgroup_task_frozen(t)) {
163 		set_tsk_thread_flag(t, TIF_SIGPENDING);
164 		return true;
165 	}
166 
167 	/*
168 	 * We must never clear the flag in another thread, or in current
169 	 * when it's possible the current syscall is returning -ERESTART*.
170 	 * So we don't clear it here, and only callers who know they should do.
171 	 */
172 	return false;
173 }
174 
175 /*
176  * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
177  * This is superfluous when called on current, the wakeup is a harmless no-op.
178  */
recalc_sigpending_and_wake(struct task_struct * t)179 void recalc_sigpending_and_wake(struct task_struct *t)
180 {
181 	if (recalc_sigpending_tsk(t))
182 		signal_wake_up(t, 0);
183 }
184 
recalc_sigpending(void)185 void recalc_sigpending(void)
186 {
187 	if (!recalc_sigpending_tsk(current) && !freezing(current) &&
188 	    !klp_patch_pending(current))
189 		clear_thread_flag(TIF_SIGPENDING);
190 
191 }
192 EXPORT_SYMBOL(recalc_sigpending);
193 
calculate_sigpending(void)194 void calculate_sigpending(void)
195 {
196 	/* Have any signals or users of TIF_SIGPENDING been delayed
197 	 * until after fork?
198 	 */
199 	spin_lock_irq(&current->sighand->siglock);
200 	set_tsk_thread_flag(current, TIF_SIGPENDING);
201 	recalc_sigpending();
202 	spin_unlock_irq(&current->sighand->siglock);
203 }
204 
205 /* Given the mask, find the first available signal that should be serviced. */
206 
207 #define SYNCHRONOUS_MASK \
208 	(sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
209 	 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
210 
next_signal(struct sigpending * pending,sigset_t * mask)211 int next_signal(struct sigpending *pending, sigset_t *mask)
212 {
213 	unsigned long i, *s, *m, x;
214 	int sig = 0;
215 
216 	s = pending->signal.sig;
217 	m = mask->sig;
218 
219 	/*
220 	 * Handle the first word specially: it contains the
221 	 * synchronous signals that need to be dequeued first.
222 	 */
223 	x = *s &~ *m;
224 	if (x) {
225 		if (x & SYNCHRONOUS_MASK)
226 			x &= SYNCHRONOUS_MASK;
227 		sig = ffz(~x) + 1;
228 		return sig;
229 	}
230 
231 	switch (_NSIG_WORDS) {
232 	default:
233 		for (i = 1; i < _NSIG_WORDS; ++i) {
234 			x = *++s &~ *++m;
235 			if (!x)
236 				continue;
237 			sig = ffz(~x) + i*_NSIG_BPW + 1;
238 			break;
239 		}
240 		break;
241 
242 	case 2:
243 		x = s[1] &~ m[1];
244 		if (!x)
245 			break;
246 		sig = ffz(~x) + _NSIG_BPW + 1;
247 		break;
248 
249 	case 1:
250 		/* Nothing to do */
251 		break;
252 	}
253 
254 	return sig;
255 }
256 
print_dropped_signal(int sig)257 static inline void print_dropped_signal(int sig)
258 {
259 	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
260 
261 	if (!print_fatal_signals)
262 		return;
263 
264 	if (!__ratelimit(&ratelimit_state))
265 		return;
266 
267 	pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
268 				current->comm, current->pid, sig);
269 }
270 
271 /**
272  * task_set_jobctl_pending - set jobctl pending bits
273  * @task: target task
274  * @mask: pending bits to set
275  *
276  * Clear @mask from @task->jobctl.  @mask must be subset of
277  * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
278  * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
279  * cleared.  If @task is already being killed or exiting, this function
280  * becomes noop.
281  *
282  * CONTEXT:
283  * Must be called with @task->sighand->siglock held.
284  *
285  * RETURNS:
286  * %true if @mask is set, %false if made noop because @task was dying.
287  */
task_set_jobctl_pending(struct task_struct * task,unsigned long mask)288 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
289 {
290 	BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
291 			JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
292 	BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
293 
294 	if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
295 		return false;
296 
297 	if (mask & JOBCTL_STOP_SIGMASK)
298 		task->jobctl &= ~JOBCTL_STOP_SIGMASK;
299 
300 	task->jobctl |= mask;
301 	return true;
302 }
303 
304 /**
305  * task_clear_jobctl_trapping - clear jobctl trapping bit
306  * @task: target task
307  *
308  * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
309  * Clear it and wake up the ptracer.  Note that we don't need any further
310  * locking.  @task->siglock guarantees that @task->parent points to the
311  * ptracer.
312  *
313  * CONTEXT:
314  * Must be called with @task->sighand->siglock held.
315  */
task_clear_jobctl_trapping(struct task_struct * task)316 void task_clear_jobctl_trapping(struct task_struct *task)
317 {
318 	if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
319 		task->jobctl &= ~JOBCTL_TRAPPING;
320 		smp_mb();	/* advised by wake_up_bit() */
321 		wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
322 	}
323 }
324 
325 /**
326  * task_clear_jobctl_pending - clear jobctl pending bits
327  * @task: target task
328  * @mask: pending bits to clear
329  *
330  * Clear @mask from @task->jobctl.  @mask must be subset of
331  * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
332  * STOP bits are cleared together.
333  *
334  * If clearing of @mask leaves no stop or trap pending, this function calls
335  * task_clear_jobctl_trapping().
336  *
337  * CONTEXT:
338  * Must be called with @task->sighand->siglock held.
339  */
task_clear_jobctl_pending(struct task_struct * task,unsigned long mask)340 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
341 {
342 	BUG_ON(mask & ~JOBCTL_PENDING_MASK);
343 
344 	if (mask & JOBCTL_STOP_PENDING)
345 		mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
346 
347 	task->jobctl &= ~mask;
348 
349 	if (!(task->jobctl & JOBCTL_PENDING_MASK))
350 		task_clear_jobctl_trapping(task);
351 }
352 
353 /**
354  * task_participate_group_stop - participate in a group stop
355  * @task: task participating in a group stop
356  *
357  * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
358  * Group stop states are cleared and the group stop count is consumed if
359  * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
360  * stop, the appropriate `SIGNAL_*` flags are set.
361  *
362  * CONTEXT:
363  * Must be called with @task->sighand->siglock held.
364  *
365  * RETURNS:
366  * %true if group stop completion should be notified to the parent, %false
367  * otherwise.
368  */
task_participate_group_stop(struct task_struct * task)369 static bool task_participate_group_stop(struct task_struct *task)
370 {
371 	struct signal_struct *sig = task->signal;
372 	bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
373 
374 	WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
375 
376 	task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
377 
378 	if (!consume)
379 		return false;
380 
381 	if (!WARN_ON_ONCE(sig->group_stop_count == 0))
382 		sig->group_stop_count--;
383 
384 	/*
385 	 * Tell the caller to notify completion iff we are entering into a
386 	 * fresh group stop.  Read comment in do_signal_stop() for details.
387 	 */
388 	if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
389 		signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
390 		return true;
391 	}
392 	return false;
393 }
394 
task_join_group_stop(struct task_struct * task)395 void task_join_group_stop(struct task_struct *task)
396 {
397 	unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
398 	struct signal_struct *sig = current->signal;
399 
400 	if (sig->group_stop_count) {
401 		sig->group_stop_count++;
402 		mask |= JOBCTL_STOP_CONSUME;
403 	} else if (!(sig->flags & SIGNAL_STOP_STOPPED))
404 		return;
405 
406 	/* Have the new thread join an on-going signal group stop */
407 	task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
408 }
409 
410 /*
411  * allocate a new signal queue record
412  * - this may be called without locks if and only if t == current, otherwise an
413  *   appropriate lock must be held to stop the target task from exiting
414  */
415 static struct sigqueue *
__sigqueue_alloc(int sig,struct task_struct * t,gfp_t flags,int override_rlimit)416 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t flags, int override_rlimit)
417 {
418 	struct sigqueue *q = NULL;
419 	struct user_struct *user;
420 	int sigpending;
421 
422 	/*
423 	 * Protect access to @t credentials. This can go away when all
424 	 * callers hold rcu read lock.
425 	 *
426 	 * NOTE! A pending signal will hold on to the user refcount,
427 	 * and we get/put the refcount only when the sigpending count
428 	 * changes from/to zero.
429 	 */
430 	rcu_read_lock();
431 	user = __task_cred(t)->user;
432 	sigpending = atomic_inc_return(&user->sigpending);
433 	if (sigpending == 1)
434 		get_uid(user);
435 	rcu_read_unlock();
436 
437 	if (override_rlimit || likely(sigpending <= task_rlimit(t, RLIMIT_SIGPENDING))) {
438 		q = kmem_cache_alloc(sigqueue_cachep, flags);
439 	} else {
440 		print_dropped_signal(sig);
441 	}
442 
443 	if (unlikely(q == NULL)) {
444 		if (atomic_dec_and_test(&user->sigpending))
445 			free_uid(user);
446 	} else {
447 		INIT_LIST_HEAD(&q->list);
448 		q->flags = 0;
449 		q->user = user;
450 	}
451 
452 	return q;
453 }
454 
__sigqueue_free(struct sigqueue * q)455 static void __sigqueue_free(struct sigqueue *q)
456 {
457 	if (q->flags & SIGQUEUE_PREALLOC)
458 		return;
459 	if (atomic_dec_and_test(&q->user->sigpending))
460 		free_uid(q->user);
461 	kmem_cache_free(sigqueue_cachep, q);
462 }
463 
flush_sigqueue(struct sigpending * queue)464 void flush_sigqueue(struct sigpending *queue)
465 {
466 	struct sigqueue *q;
467 
468 	sigemptyset(&queue->signal);
469 	while (!list_empty(&queue->list)) {
470 		q = list_entry(queue->list.next, struct sigqueue , list);
471 		list_del_init(&q->list);
472 		__sigqueue_free(q);
473 	}
474 }
475 
476 /*
477  * Flush all pending signals for this kthread.
478  */
flush_signals(struct task_struct * t)479 void flush_signals(struct task_struct *t)
480 {
481 	unsigned long flags;
482 
483 	spin_lock_irqsave(&t->sighand->siglock, flags);
484 	clear_tsk_thread_flag(t, TIF_SIGPENDING);
485 	flush_sigqueue(&t->pending);
486 	flush_sigqueue(&t->signal->shared_pending);
487 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
488 }
489 EXPORT_SYMBOL(flush_signals);
490 
491 #ifdef CONFIG_POSIX_TIMERS
__flush_itimer_signals(struct sigpending * pending)492 static void __flush_itimer_signals(struct sigpending *pending)
493 {
494 	sigset_t signal, retain;
495 	struct sigqueue *q, *n;
496 
497 	signal = pending->signal;
498 	sigemptyset(&retain);
499 
500 	list_for_each_entry_safe(q, n, &pending->list, list) {
501 		int sig = q->info.si_signo;
502 
503 		if (likely(q->info.si_code != SI_TIMER)) {
504 			sigaddset(&retain, sig);
505 		} else {
506 			sigdelset(&signal, sig);
507 			list_del_init(&q->list);
508 			__sigqueue_free(q);
509 		}
510 	}
511 
512 	sigorsets(&pending->signal, &signal, &retain);
513 }
514 
flush_itimer_signals(void)515 void flush_itimer_signals(void)
516 {
517 	struct task_struct *tsk = current;
518 	unsigned long flags;
519 
520 	spin_lock_irqsave(&tsk->sighand->siglock, flags);
521 	__flush_itimer_signals(&tsk->pending);
522 	__flush_itimer_signals(&tsk->signal->shared_pending);
523 	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
524 }
525 #endif
526 
ignore_signals(struct task_struct * t)527 void ignore_signals(struct task_struct *t)
528 {
529 	int i;
530 
531 	for (i = 0; i < _NSIG; ++i)
532 		t->sighand->action[i].sa.sa_handler = SIG_IGN;
533 
534 	flush_signals(t);
535 }
536 
537 /*
538  * Flush all handlers for a task.
539  */
540 
541 void
flush_signal_handlers(struct task_struct * t,int force_default)542 flush_signal_handlers(struct task_struct *t, int force_default)
543 {
544 	int i;
545 	struct k_sigaction *ka = &t->sighand->action[0];
546 	for (i = _NSIG ; i != 0 ; i--) {
547 		if (force_default || ka->sa.sa_handler != SIG_IGN)
548 			ka->sa.sa_handler = SIG_DFL;
549 		ka->sa.sa_flags = 0;
550 #ifdef __ARCH_HAS_SA_RESTORER
551 		ka->sa.sa_restorer = NULL;
552 #endif
553 		sigemptyset(&ka->sa.sa_mask);
554 		ka++;
555 	}
556 }
557 
unhandled_signal(struct task_struct * tsk,int sig)558 bool unhandled_signal(struct task_struct *tsk, int sig)
559 {
560 	void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
561 	if (is_global_init(tsk))
562 		return true;
563 
564 	if (handler != SIG_IGN && handler != SIG_DFL)
565 		return false;
566 
567 	/* if ptraced, let the tracer determine */
568 	return !tsk->ptrace;
569 }
570 
collect_signal(int sig,struct sigpending * list,kernel_siginfo_t * info,bool * resched_timer)571 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
572 			   bool *resched_timer)
573 {
574 	struct sigqueue *q, *first = NULL;
575 
576 	/*
577 	 * Collect the siginfo appropriate to this signal.  Check if
578 	 * there is another siginfo for the same signal.
579 	*/
580 	list_for_each_entry(q, &list->list, list) {
581 		if (q->info.si_signo == sig) {
582 			if (first)
583 				goto still_pending;
584 			first = q;
585 		}
586 	}
587 
588 	sigdelset(&list->signal, sig);
589 
590 	if (first) {
591 still_pending:
592 		list_del_init(&first->list);
593 		copy_siginfo(info, &first->info);
594 
595 		*resched_timer =
596 			(first->flags & SIGQUEUE_PREALLOC) &&
597 			(info->si_code == SI_TIMER) &&
598 			(info->si_sys_private);
599 
600 		__sigqueue_free(first);
601 	} else {
602 		/*
603 		 * Ok, it wasn't in the queue.  This must be
604 		 * a fast-pathed signal or we must have been
605 		 * out of queue space.  So zero out the info.
606 		 */
607 		clear_siginfo(info);
608 		info->si_signo = sig;
609 		info->si_errno = 0;
610 		info->si_code = SI_USER;
611 		info->si_pid = 0;
612 		info->si_uid = 0;
613 	}
614 }
615 
__dequeue_signal(struct sigpending * pending,sigset_t * mask,kernel_siginfo_t * info,bool * resched_timer)616 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
617 			kernel_siginfo_t *info, bool *resched_timer)
618 {
619 	int sig = next_signal(pending, mask);
620 
621 	if (sig)
622 		collect_signal(sig, pending, info, resched_timer);
623 	return sig;
624 }
625 
626 /*
627  * Dequeue a signal and return the element to the caller, which is
628  * expected to free it.
629  *
630  * All callers have to hold the siglock.
631  */
dequeue_signal(struct task_struct * tsk,sigset_t * mask,kernel_siginfo_t * info)632 int dequeue_signal(struct task_struct *tsk, sigset_t *mask, kernel_siginfo_t *info)
633 {
634 	bool resched_timer = false;
635 	int signr;
636 
637 	/* We only dequeue private signals from ourselves, we don't let
638 	 * signalfd steal them
639 	 */
640 	signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
641 	if (!signr) {
642 		signr = __dequeue_signal(&tsk->signal->shared_pending,
643 					 mask, info, &resched_timer);
644 #ifdef CONFIG_POSIX_TIMERS
645 		/*
646 		 * itimer signal ?
647 		 *
648 		 * itimers are process shared and we restart periodic
649 		 * itimers in the signal delivery path to prevent DoS
650 		 * attacks in the high resolution timer case. This is
651 		 * compliant with the old way of self-restarting
652 		 * itimers, as the SIGALRM is a legacy signal and only
653 		 * queued once. Changing the restart behaviour to
654 		 * restart the timer in the signal dequeue path is
655 		 * reducing the timer noise on heavy loaded !highres
656 		 * systems too.
657 		 */
658 		if (unlikely(signr == SIGALRM)) {
659 			struct hrtimer *tmr = &tsk->signal->real_timer;
660 
661 			if (!hrtimer_is_queued(tmr) &&
662 			    tsk->signal->it_real_incr != 0) {
663 				hrtimer_forward(tmr, tmr->base->get_time(),
664 						tsk->signal->it_real_incr);
665 				hrtimer_restart(tmr);
666 			}
667 		}
668 #endif
669 	}
670 
671 	recalc_sigpending();
672 	if (!signr)
673 		return 0;
674 
675 	if (unlikely(sig_kernel_stop(signr))) {
676 		/*
677 		 * Set a marker that we have dequeued a stop signal.  Our
678 		 * caller might release the siglock and then the pending
679 		 * stop signal it is about to process is no longer in the
680 		 * pending bitmasks, but must still be cleared by a SIGCONT
681 		 * (and overruled by a SIGKILL).  So those cases clear this
682 		 * shared flag after we've set it.  Note that this flag may
683 		 * remain set after the signal we return is ignored or
684 		 * handled.  That doesn't matter because its only purpose
685 		 * is to alert stop-signal processing code when another
686 		 * processor has come along and cleared the flag.
687 		 */
688 		current->jobctl |= JOBCTL_STOP_DEQUEUED;
689 	}
690 #ifdef CONFIG_POSIX_TIMERS
691 	if (resched_timer) {
692 		/*
693 		 * Release the siglock to ensure proper locking order
694 		 * of timer locks outside of siglocks.  Note, we leave
695 		 * irqs disabled here, since the posix-timers code is
696 		 * about to disable them again anyway.
697 		 */
698 		spin_unlock(&tsk->sighand->siglock);
699 		posixtimer_rearm(info);
700 		spin_lock(&tsk->sighand->siglock);
701 
702 		/* Don't expose the si_sys_private value to userspace */
703 		info->si_sys_private = 0;
704 	}
705 #endif
706 	return signr;
707 }
708 EXPORT_SYMBOL_GPL(dequeue_signal);
709 
dequeue_synchronous_signal(kernel_siginfo_t * info)710 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
711 {
712 	struct task_struct *tsk = current;
713 	struct sigpending *pending = &tsk->pending;
714 	struct sigqueue *q, *sync = NULL;
715 
716 	/*
717 	 * Might a synchronous signal be in the queue?
718 	 */
719 	if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
720 		return 0;
721 
722 	/*
723 	 * Return the first synchronous signal in the queue.
724 	 */
725 	list_for_each_entry(q, &pending->list, list) {
726 		/* Synchronous signals have a positive si_code */
727 		if ((q->info.si_code > SI_USER) &&
728 		    (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
729 			sync = q;
730 			goto next;
731 		}
732 	}
733 	return 0;
734 next:
735 	/*
736 	 * Check if there is another siginfo for the same signal.
737 	 */
738 	list_for_each_entry_continue(q, &pending->list, list) {
739 		if (q->info.si_signo == sync->info.si_signo)
740 			goto still_pending;
741 	}
742 
743 	sigdelset(&pending->signal, sync->info.si_signo);
744 	recalc_sigpending();
745 still_pending:
746 	list_del_init(&sync->list);
747 	copy_siginfo(info, &sync->info);
748 	__sigqueue_free(sync);
749 	return info->si_signo;
750 }
751 
752 /*
753  * Tell a process that it has a new active signal..
754  *
755  * NOTE! we rely on the previous spin_lock to
756  * lock interrupts for us! We can only be called with
757  * "siglock" held, and the local interrupt must
758  * have been disabled when that got acquired!
759  *
760  * No need to set need_resched since signal event passing
761  * goes through ->blocked
762  */
signal_wake_up_state(struct task_struct * t,unsigned int state)763 void signal_wake_up_state(struct task_struct *t, unsigned int state)
764 {
765 	set_tsk_thread_flag(t, TIF_SIGPENDING);
766 	/*
767 	 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
768 	 * case. We don't check t->state here because there is a race with it
769 	 * executing another processor and just now entering stopped state.
770 	 * By using wake_up_state, we ensure the process will wake up and
771 	 * handle its death signal.
772 	 */
773 	if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
774 		kick_process(t);
775 }
776 
777 /*
778  * Remove signals in mask from the pending set and queue.
779  * Returns 1 if any signals were found.
780  *
781  * All callers must be holding the siglock.
782  */
flush_sigqueue_mask(sigset_t * mask,struct sigpending * s)783 static void flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
784 {
785 	struct sigqueue *q, *n;
786 	sigset_t m;
787 
788 	sigandsets(&m, mask, &s->signal);
789 	if (sigisemptyset(&m))
790 		return;
791 
792 	sigandnsets(&s->signal, &s->signal, mask);
793 	list_for_each_entry_safe(q, n, &s->list, list) {
794 		if (sigismember(mask, q->info.si_signo)) {
795 			list_del_init(&q->list);
796 			__sigqueue_free(q);
797 		}
798 	}
799 }
800 
is_si_special(const struct kernel_siginfo * info)801 static inline int is_si_special(const struct kernel_siginfo *info)
802 {
803 	return info <= SEND_SIG_PRIV;
804 }
805 
si_fromuser(const struct kernel_siginfo * info)806 static inline bool si_fromuser(const struct kernel_siginfo *info)
807 {
808 	return info == SEND_SIG_NOINFO ||
809 		(!is_si_special(info) && SI_FROMUSER(info));
810 }
811 
812 /*
813  * called with RCU read lock from check_kill_permission()
814  */
kill_ok_by_cred(struct task_struct * t)815 static bool kill_ok_by_cred(struct task_struct *t)
816 {
817 	const struct cred *cred = current_cred();
818 	const struct cred *tcred = __task_cred(t);
819 
820 	return uid_eq(cred->euid, tcred->suid) ||
821 	       uid_eq(cred->euid, tcred->uid) ||
822 	       uid_eq(cred->uid, tcred->suid) ||
823 	       uid_eq(cred->uid, tcred->uid) ||
824 	       ns_capable(tcred->user_ns, CAP_KILL);
825 }
826 
827 /*
828  * Bad permissions for sending the signal
829  * - the caller must hold the RCU read lock
830  */
check_kill_permission(int sig,struct kernel_siginfo * info,struct task_struct * t)831 static int check_kill_permission(int sig, struct kernel_siginfo *info,
832 				 struct task_struct *t)
833 {
834 	struct pid *sid;
835 	int error;
836 
837 	if (!valid_signal(sig))
838 		return -EINVAL;
839 
840 	if (!si_fromuser(info))
841 		return 0;
842 
843 	error = audit_signal_info(sig, t); /* Let audit system see the signal */
844 	if (error)
845 		return error;
846 
847 	if (!same_thread_group(current, t) &&
848 	    !kill_ok_by_cred(t)) {
849 		switch (sig) {
850 		case SIGCONT:
851 			sid = task_session(t);
852 			/*
853 			 * We don't return the error if sid == NULL. The
854 			 * task was unhashed, the caller must notice this.
855 			 */
856 			if (!sid || sid == task_session(current))
857 				break;
858 			fallthrough;
859 		default:
860 			return -EPERM;
861 		}
862 	}
863 
864 	return security_task_kill(t, info, sig, NULL);
865 }
866 
867 /**
868  * ptrace_trap_notify - schedule trap to notify ptracer
869  * @t: tracee wanting to notify tracer
870  *
871  * This function schedules sticky ptrace trap which is cleared on the next
872  * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
873  * ptracer.
874  *
875  * If @t is running, STOP trap will be taken.  If trapped for STOP and
876  * ptracer is listening for events, tracee is woken up so that it can
877  * re-trap for the new event.  If trapped otherwise, STOP trap will be
878  * eventually taken without returning to userland after the existing traps
879  * are finished by PTRACE_CONT.
880  *
881  * CONTEXT:
882  * Must be called with @task->sighand->siglock held.
883  */
ptrace_trap_notify(struct task_struct * t)884 static void ptrace_trap_notify(struct task_struct *t)
885 {
886 	WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
887 	assert_spin_locked(&t->sighand->siglock);
888 
889 	task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
890 	ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
891 }
892 
893 /*
894  * Handle magic process-wide effects of stop/continue signals. Unlike
895  * the signal actions, these happen immediately at signal-generation
896  * time regardless of blocking, ignoring, or handling.  This does the
897  * actual continuing for SIGCONT, but not the actual stopping for stop
898  * signals. The process stop is done as a signal action for SIG_DFL.
899  *
900  * Returns true if the signal should be actually delivered, otherwise
901  * it should be dropped.
902  */
prepare_signal(int sig,struct task_struct * p,bool force)903 static bool prepare_signal(int sig, struct task_struct *p, bool force)
904 {
905 	struct signal_struct *signal = p->signal;
906 	struct task_struct *t;
907 	sigset_t flush;
908 
909 	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
910 		if (!(signal->flags & SIGNAL_GROUP_EXIT))
911 			return sig == SIGKILL;
912 		/*
913 		 * The process is in the middle of dying, nothing to do.
914 		 */
915 	} else if (sig_kernel_stop(sig)) {
916 		/*
917 		 * This is a stop signal.  Remove SIGCONT from all queues.
918 		 */
919 		siginitset(&flush, sigmask(SIGCONT));
920 		flush_sigqueue_mask(&flush, &signal->shared_pending);
921 		for_each_thread(p, t)
922 			flush_sigqueue_mask(&flush, &t->pending);
923 	} else if (sig == SIGCONT) {
924 		unsigned int why;
925 		/*
926 		 * Remove all stop signals from all queues, wake all threads.
927 		 */
928 		siginitset(&flush, SIG_KERNEL_STOP_MASK);
929 		flush_sigqueue_mask(&flush, &signal->shared_pending);
930 		for_each_thread(p, t) {
931 			flush_sigqueue_mask(&flush, &t->pending);
932 			task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
933 			if (likely(!(t->ptrace & PT_SEIZED)))
934 				wake_up_state(t, __TASK_STOPPED);
935 			else
936 				ptrace_trap_notify(t);
937 		}
938 
939 		/*
940 		 * Notify the parent with CLD_CONTINUED if we were stopped.
941 		 *
942 		 * If we were in the middle of a group stop, we pretend it
943 		 * was already finished, and then continued. Since SIGCHLD
944 		 * doesn't queue we report only CLD_STOPPED, as if the next
945 		 * CLD_CONTINUED was dropped.
946 		 */
947 		why = 0;
948 		if (signal->flags & SIGNAL_STOP_STOPPED)
949 			why |= SIGNAL_CLD_CONTINUED;
950 		else if (signal->group_stop_count)
951 			why |= SIGNAL_CLD_STOPPED;
952 
953 		if (why) {
954 			/*
955 			 * The first thread which returns from do_signal_stop()
956 			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
957 			 * notify its parent. See get_signal().
958 			 */
959 			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
960 			signal->group_stop_count = 0;
961 			signal->group_exit_code = 0;
962 		}
963 	}
964 
965 	return !sig_ignored(p, sig, force);
966 }
967 
968 /*
969  * Test if P wants to take SIG.  After we've checked all threads with this,
970  * it's equivalent to finding no threads not blocking SIG.  Any threads not
971  * blocking SIG were ruled out because they are not running and already
972  * have pending signals.  Such threads will dequeue from the shared queue
973  * as soon as they're available, so putting the signal on the shared queue
974  * will be equivalent to sending it to one such thread.
975  */
wants_signal(int sig,struct task_struct * p)976 static inline bool wants_signal(int sig, struct task_struct *p)
977 {
978 	if (sigismember(&p->blocked, sig))
979 		return false;
980 
981 	if (p->flags & PF_EXITING)
982 		return false;
983 
984 	if (sig == SIGKILL)
985 		return true;
986 
987 	if (task_is_stopped_or_traced(p))
988 		return false;
989 
990 	return task_curr(p) || !task_sigpending(p);
991 }
992 
complete_signal(int sig,struct task_struct * p,enum pid_type type)993 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
994 {
995 	struct signal_struct *signal = p->signal;
996 	struct task_struct *t;
997 
998 	/*
999 	 * Now find a thread we can wake up to take the signal off the queue.
1000 	 *
1001 	 * If the main thread wants the signal, it gets first crack.
1002 	 * Probably the least surprising to the average bear.
1003 	 */
1004 	if (wants_signal(sig, p))
1005 		t = p;
1006 	else if ((type == PIDTYPE_PID) || thread_group_empty(p))
1007 		/*
1008 		 * There is just one thread and it does not need to be woken.
1009 		 * It will dequeue unblocked signals before it runs again.
1010 		 */
1011 		return;
1012 	else {
1013 		/*
1014 		 * Otherwise try to find a suitable thread.
1015 		 */
1016 		t = signal->curr_target;
1017 		while (!wants_signal(sig, t)) {
1018 			t = next_thread(t);
1019 			if (t == signal->curr_target)
1020 				/*
1021 				 * No thread needs to be woken.
1022 				 * Any eligible threads will see
1023 				 * the signal in the queue soon.
1024 				 */
1025 				return;
1026 		}
1027 		signal->curr_target = t;
1028 	}
1029 
1030 	/*
1031 	 * Found a killable thread.  If the signal will be fatal,
1032 	 * then start taking the whole group down immediately.
1033 	 */
1034 	if (sig_fatal(p, sig) &&
1035 	    !(signal->flags & SIGNAL_GROUP_EXIT) &&
1036 	    !sigismember(&t->real_blocked, sig) &&
1037 	    (sig == SIGKILL || !p->ptrace)) {
1038 		/*
1039 		 * This signal will be fatal to the whole group.
1040 		 */
1041 		if (!sig_kernel_coredump(sig)) {
1042 			/*
1043 			 * Start a group exit and wake everybody up.
1044 			 * This way we don't have other threads
1045 			 * running and doing things after a slower
1046 			 * thread has the fatal signal pending.
1047 			 */
1048 			signal->flags = SIGNAL_GROUP_EXIT;
1049 			signal->group_exit_code = sig;
1050 			signal->group_stop_count = 0;
1051 			t = p;
1052 			do {
1053 				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1054 				sigaddset(&t->pending.signal, SIGKILL);
1055 				signal_wake_up(t, 1);
1056 			} while_each_thread(p, t);
1057 			return;
1058 		}
1059 	}
1060 
1061 	/*
1062 	 * The signal is already in the shared-pending queue.
1063 	 * Tell the chosen thread to wake up and dequeue it.
1064 	 */
1065 	signal_wake_up(t, sig == SIGKILL);
1066 	return;
1067 }
1068 
legacy_queue(struct sigpending * signals,int sig)1069 static inline bool legacy_queue(struct sigpending *signals, int sig)
1070 {
1071 	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1072 }
1073 
__send_signal(int sig,struct kernel_siginfo * info,struct task_struct * t,enum pid_type type,bool force)1074 static int __send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1075 			enum pid_type type, bool force)
1076 {
1077 	struct sigpending *pending;
1078 	struct sigqueue *q;
1079 	int override_rlimit;
1080 	int ret = 0, result;
1081 
1082 	assert_spin_locked(&t->sighand->siglock);
1083 
1084 	result = TRACE_SIGNAL_IGNORED;
1085 	if (!prepare_signal(sig, t, force))
1086 		goto ret;
1087 
1088 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1089 	/*
1090 	 * Short-circuit ignored signals and support queuing
1091 	 * exactly one non-rt signal, so that we can get more
1092 	 * detailed information about the cause of the signal.
1093 	 */
1094 	result = TRACE_SIGNAL_ALREADY_PENDING;
1095 	if (legacy_queue(pending, sig))
1096 		goto ret;
1097 
1098 	result = TRACE_SIGNAL_DELIVERED;
1099 	/*
1100 	 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1101 	 */
1102 	if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1103 		goto out_set;
1104 
1105 	/*
1106 	 * Real-time signals must be queued if sent by sigqueue, or
1107 	 * some other real-time mechanism.  It is implementation
1108 	 * defined whether kill() does so.  We attempt to do so, on
1109 	 * the principle of least surprise, but since kill is not
1110 	 * allowed to fail with EAGAIN when low on memory we just
1111 	 * make sure at least one signal gets delivered and don't
1112 	 * pass on the info struct.
1113 	 */
1114 	if (sig < SIGRTMIN)
1115 		override_rlimit = (is_si_special(info) || info->si_code >= 0);
1116 	else
1117 		override_rlimit = 0;
1118 
1119 	q = __sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit);
1120 	if (q) {
1121 		list_add_tail(&q->list, &pending->list);
1122 		switch ((unsigned long) info) {
1123 		case (unsigned long) SEND_SIG_NOINFO:
1124 			clear_siginfo(&q->info);
1125 			q->info.si_signo = sig;
1126 			q->info.si_errno = 0;
1127 			q->info.si_code = SI_USER;
1128 			q->info.si_pid = task_tgid_nr_ns(current,
1129 							task_active_pid_ns(t));
1130 			rcu_read_lock();
1131 			q->info.si_uid =
1132 				from_kuid_munged(task_cred_xxx(t, user_ns),
1133 						 current_uid());
1134 			rcu_read_unlock();
1135 			break;
1136 		case (unsigned long) SEND_SIG_PRIV:
1137 			clear_siginfo(&q->info);
1138 			q->info.si_signo = sig;
1139 			q->info.si_errno = 0;
1140 			q->info.si_code = SI_KERNEL;
1141 			q->info.si_pid = 0;
1142 			q->info.si_uid = 0;
1143 			break;
1144 		default:
1145 			copy_siginfo(&q->info, info);
1146 			break;
1147 		}
1148 	} else if (!is_si_special(info) &&
1149 		   sig >= SIGRTMIN && info->si_code != SI_USER) {
1150 		/*
1151 		 * Queue overflow, abort.  We may abort if the
1152 		 * signal was rt and sent by user using something
1153 		 * other than kill().
1154 		 */
1155 		result = TRACE_SIGNAL_OVERFLOW_FAIL;
1156 		ret = -EAGAIN;
1157 		goto ret;
1158 	} else {
1159 		/*
1160 		 * This is a silent loss of information.  We still
1161 		 * send the signal, but the *info bits are lost.
1162 		 */
1163 		result = TRACE_SIGNAL_LOSE_INFO;
1164 	}
1165 
1166 out_set:
1167 	signalfd_notify(t, sig);
1168 	sigaddset(&pending->signal, sig);
1169 
1170 	/* Let multiprocess signals appear after on-going forks */
1171 	if (type > PIDTYPE_TGID) {
1172 		struct multiprocess_signals *delayed;
1173 		hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1174 			sigset_t *signal = &delayed->signal;
1175 			/* Can't queue both a stop and a continue signal */
1176 			if (sig == SIGCONT)
1177 				sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1178 			else if (sig_kernel_stop(sig))
1179 				sigdelset(signal, SIGCONT);
1180 			sigaddset(signal, sig);
1181 		}
1182 	}
1183 
1184 	complete_signal(sig, t, type);
1185 ret:
1186 	trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1187 	return ret;
1188 }
1189 
has_si_pid_and_uid(struct kernel_siginfo * info)1190 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1191 {
1192 	bool ret = false;
1193 	switch (siginfo_layout(info->si_signo, info->si_code)) {
1194 	case SIL_KILL:
1195 	case SIL_CHLD:
1196 	case SIL_RT:
1197 		ret = true;
1198 		break;
1199 	case SIL_TIMER:
1200 	case SIL_POLL:
1201 	case SIL_FAULT:
1202 	case SIL_FAULT_MCEERR:
1203 	case SIL_FAULT_BNDERR:
1204 	case SIL_FAULT_PKUERR:
1205 	case SIL_SYS:
1206 		ret = false;
1207 		break;
1208 	}
1209 	return ret;
1210 }
1211 
send_signal(int sig,struct kernel_siginfo * info,struct task_struct * t,enum pid_type type)1212 static int send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1213 			enum pid_type type)
1214 {
1215 	/* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1216 	bool force = false;
1217 
1218 	if (info == SEND_SIG_NOINFO) {
1219 		/* Force if sent from an ancestor pid namespace */
1220 		force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1221 	} else if (info == SEND_SIG_PRIV) {
1222 		/* Don't ignore kernel generated signals */
1223 		force = true;
1224 	} else if (has_si_pid_and_uid(info)) {
1225 		/* SIGKILL and SIGSTOP is special or has ids */
1226 		struct user_namespace *t_user_ns;
1227 
1228 		rcu_read_lock();
1229 		t_user_ns = task_cred_xxx(t, user_ns);
1230 		if (current_user_ns() != t_user_ns) {
1231 			kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1232 			info->si_uid = from_kuid_munged(t_user_ns, uid);
1233 		}
1234 		rcu_read_unlock();
1235 
1236 		/* A kernel generated signal? */
1237 		force = (info->si_code == SI_KERNEL);
1238 
1239 		/* From an ancestor pid namespace? */
1240 		if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1241 			info->si_pid = 0;
1242 			force = true;
1243 		}
1244 	}
1245 	return __send_signal(sig, info, t, type, force);
1246 }
1247 
print_fatal_signal(int signr)1248 static void print_fatal_signal(int signr)
1249 {
1250 	struct pt_regs *regs = signal_pt_regs();
1251 	pr_info("potentially unexpected fatal signal %d.\n", signr);
1252 
1253 #if defined(__i386__) && !defined(__arch_um__)
1254 	pr_info("code at %08lx: ", regs->ip);
1255 	{
1256 		int i;
1257 		for (i = 0; i < 16; i++) {
1258 			unsigned char insn;
1259 
1260 			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1261 				break;
1262 			pr_cont("%02x ", insn);
1263 		}
1264 	}
1265 	pr_cont("\n");
1266 #endif
1267 	preempt_disable();
1268 	show_regs(regs);
1269 	preempt_enable();
1270 }
1271 
setup_print_fatal_signals(char * str)1272 static int __init setup_print_fatal_signals(char *str)
1273 {
1274 	get_option (&str, &print_fatal_signals);
1275 
1276 	return 1;
1277 }
1278 
1279 __setup("print-fatal-signals=", setup_print_fatal_signals);
1280 
1281 int
__group_send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p)1282 __group_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1283 {
1284 	return send_signal(sig, info, p, PIDTYPE_TGID);
1285 }
1286 
do_send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p,enum pid_type type)1287 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1288 			enum pid_type type)
1289 {
1290 	unsigned long flags;
1291 	int ret = -ESRCH;
1292 	trace_android_vh_do_send_sig_info(sig, current, p);
1293 	if (lock_task_sighand(p, &flags)) {
1294 		ret = send_signal(sig, info, p, type);
1295 		unlock_task_sighand(p, &flags);
1296 	}
1297 
1298 	return ret;
1299 }
1300 
1301 /*
1302  * Force a signal that the process can't ignore: if necessary
1303  * we unblock the signal and change any SIG_IGN to SIG_DFL.
1304  *
1305  * Note: If we unblock the signal, we always reset it to SIG_DFL,
1306  * since we do not want to have a signal handler that was blocked
1307  * be invoked when user space had explicitly blocked it.
1308  *
1309  * We don't want to have recursive SIGSEGV's etc, for example,
1310  * that is why we also clear SIGNAL_UNKILLABLE.
1311  */
1312 static int
force_sig_info_to_task(struct kernel_siginfo * info,struct task_struct * t)1313 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t)
1314 {
1315 	unsigned long int flags;
1316 	int ret, blocked, ignored;
1317 	struct k_sigaction *action;
1318 	int sig = info->si_signo;
1319 
1320 	spin_lock_irqsave(&t->sighand->siglock, flags);
1321 	action = &t->sighand->action[sig-1];
1322 	ignored = action->sa.sa_handler == SIG_IGN;
1323 	blocked = sigismember(&t->blocked, sig);
1324 	if (blocked || ignored) {
1325 		action->sa.sa_handler = SIG_DFL;
1326 		if (blocked) {
1327 			sigdelset(&t->blocked, sig);
1328 			recalc_sigpending_and_wake(t);
1329 		}
1330 	}
1331 	/*
1332 	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1333 	 * debugging to leave init killable.
1334 	 */
1335 	if (action->sa.sa_handler == SIG_DFL && !t->ptrace)
1336 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1337 	ret = send_signal(sig, info, t, PIDTYPE_PID);
1338 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1339 
1340 	return ret;
1341 }
1342 
force_sig_info(struct kernel_siginfo * info)1343 int force_sig_info(struct kernel_siginfo *info)
1344 {
1345 	return force_sig_info_to_task(info, current);
1346 }
1347 
1348 /*
1349  * Nuke all other threads in the group.
1350  */
zap_other_threads(struct task_struct * p)1351 int zap_other_threads(struct task_struct *p)
1352 {
1353 	struct task_struct *t = p;
1354 	int count = 0;
1355 
1356 	p->signal->group_stop_count = 0;
1357 
1358 	while_each_thread(p, t) {
1359 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1360 		count++;
1361 
1362 		/* Don't bother with already dead threads */
1363 		if (t->exit_state)
1364 			continue;
1365 		sigaddset(&t->pending.signal, SIGKILL);
1366 		signal_wake_up(t, 1);
1367 	}
1368 
1369 	return count;
1370 }
1371 
__lock_task_sighand(struct task_struct * tsk,unsigned long * flags)1372 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1373 					   unsigned long *flags)
1374 {
1375 	struct sighand_struct *sighand;
1376 
1377 	rcu_read_lock();
1378 	for (;;) {
1379 		sighand = rcu_dereference(tsk->sighand);
1380 		if (unlikely(sighand == NULL))
1381 			break;
1382 
1383 		/*
1384 		 * This sighand can be already freed and even reused, but
1385 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1386 		 * initializes ->siglock: this slab can't go away, it has
1387 		 * the same object type, ->siglock can't be reinitialized.
1388 		 *
1389 		 * We need to ensure that tsk->sighand is still the same
1390 		 * after we take the lock, we can race with de_thread() or
1391 		 * __exit_signal(). In the latter case the next iteration
1392 		 * must see ->sighand == NULL.
1393 		 */
1394 		spin_lock_irqsave(&sighand->siglock, *flags);
1395 		if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1396 			break;
1397 		spin_unlock_irqrestore(&sighand->siglock, *flags);
1398 	}
1399 	rcu_read_unlock();
1400 
1401 	return sighand;
1402 }
1403 
1404 /*
1405  * send signal info to all the members of a group
1406  */
group_send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p,enum pid_type type)1407 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1408 			struct task_struct *p, enum pid_type type)
1409 {
1410 	int ret;
1411 
1412 	rcu_read_lock();
1413 	ret = check_kill_permission(sig, info, p);
1414 	rcu_read_unlock();
1415 
1416 	if (!ret && sig) {
1417 		ret = do_send_sig_info(sig, info, p, type);
1418 		if (!ret && sig == SIGKILL) {
1419 			bool reap = false;
1420 
1421 			trace_android_vh_process_killed(current, &reap);
1422 			if (reap)
1423 				add_to_oom_reaper(p);
1424 		}
1425 	}
1426 
1427 	return ret;
1428 }
1429 
1430 /*
1431  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1432  * control characters do (^C, ^Z etc)
1433  * - the caller must hold at least a readlock on tasklist_lock
1434  */
__kill_pgrp_info(int sig,struct kernel_siginfo * info,struct pid * pgrp)1435 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1436 {
1437 	struct task_struct *p = NULL;
1438 	int retval, success;
1439 
1440 	success = 0;
1441 	retval = -ESRCH;
1442 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1443 		int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1444 		success |= !err;
1445 		retval = err;
1446 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1447 	return success ? 0 : retval;
1448 }
1449 
kill_pid_info(int sig,struct kernel_siginfo * info,struct pid * pid)1450 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1451 {
1452 	int error = -ESRCH;
1453 	struct task_struct *p;
1454 
1455 	for (;;) {
1456 		rcu_read_lock();
1457 		p = pid_task(pid, PIDTYPE_PID);
1458 		if (p)
1459 			error = group_send_sig_info(sig, info, p, PIDTYPE_TGID);
1460 		rcu_read_unlock();
1461 		if (likely(!p || error != -ESRCH))
1462 			return error;
1463 
1464 		/*
1465 		 * The task was unhashed in between, try again.  If it
1466 		 * is dead, pid_task() will return NULL, if we race with
1467 		 * de_thread() it will find the new leader.
1468 		 */
1469 	}
1470 }
1471 
kill_proc_info(int sig,struct kernel_siginfo * info,pid_t pid)1472 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1473 {
1474 	int error;
1475 	rcu_read_lock();
1476 	error = kill_pid_info(sig, info, find_vpid(pid));
1477 	rcu_read_unlock();
1478 	return error;
1479 }
1480 
kill_as_cred_perm(const struct cred * cred,struct task_struct * target)1481 static inline bool kill_as_cred_perm(const struct cred *cred,
1482 				     struct task_struct *target)
1483 {
1484 	const struct cred *pcred = __task_cred(target);
1485 
1486 	return uid_eq(cred->euid, pcred->suid) ||
1487 	       uid_eq(cred->euid, pcred->uid) ||
1488 	       uid_eq(cred->uid, pcred->suid) ||
1489 	       uid_eq(cred->uid, pcred->uid);
1490 }
1491 
1492 /*
1493  * The usb asyncio usage of siginfo is wrong.  The glibc support
1494  * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1495  * AKA after the generic fields:
1496  *	kernel_pid_t	si_pid;
1497  *	kernel_uid32_t	si_uid;
1498  *	sigval_t	si_value;
1499  *
1500  * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1501  * after the generic fields is:
1502  *	void __user 	*si_addr;
1503  *
1504  * This is a practical problem when there is a 64bit big endian kernel
1505  * and a 32bit userspace.  As the 32bit address will encoded in the low
1506  * 32bits of the pointer.  Those low 32bits will be stored at higher
1507  * address than appear in a 32 bit pointer.  So userspace will not
1508  * see the address it was expecting for it's completions.
1509  *
1510  * There is nothing in the encoding that can allow
1511  * copy_siginfo_to_user32 to detect this confusion of formats, so
1512  * handle this by requiring the caller of kill_pid_usb_asyncio to
1513  * notice when this situration takes place and to store the 32bit
1514  * pointer in sival_int, instead of sival_addr of the sigval_t addr
1515  * parameter.
1516  */
kill_pid_usb_asyncio(int sig,int errno,sigval_t addr,struct pid * pid,const struct cred * cred)1517 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1518 			 struct pid *pid, const struct cred *cred)
1519 {
1520 	struct kernel_siginfo info;
1521 	struct task_struct *p;
1522 	unsigned long flags;
1523 	int ret = -EINVAL;
1524 
1525 	if (!valid_signal(sig))
1526 		return ret;
1527 
1528 	clear_siginfo(&info);
1529 	info.si_signo = sig;
1530 	info.si_errno = errno;
1531 	info.si_code = SI_ASYNCIO;
1532 	*((sigval_t *)&info.si_pid) = addr;
1533 
1534 	rcu_read_lock();
1535 	p = pid_task(pid, PIDTYPE_PID);
1536 	if (!p) {
1537 		ret = -ESRCH;
1538 		goto out_unlock;
1539 	}
1540 	if (!kill_as_cred_perm(cred, p)) {
1541 		ret = -EPERM;
1542 		goto out_unlock;
1543 	}
1544 	ret = security_task_kill(p, &info, sig, cred);
1545 	if (ret)
1546 		goto out_unlock;
1547 
1548 	if (sig) {
1549 		if (lock_task_sighand(p, &flags)) {
1550 			ret = __send_signal(sig, &info, p, PIDTYPE_TGID, false);
1551 			unlock_task_sighand(p, &flags);
1552 		} else
1553 			ret = -ESRCH;
1554 	}
1555 out_unlock:
1556 	rcu_read_unlock();
1557 	return ret;
1558 }
1559 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1560 
1561 /*
1562  * kill_something_info() interprets pid in interesting ways just like kill(2).
1563  *
1564  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1565  * is probably wrong.  Should make it like BSD or SYSV.
1566  */
1567 
kill_something_info(int sig,struct kernel_siginfo * info,pid_t pid)1568 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1569 {
1570 	int ret;
1571 
1572 	if (pid > 0)
1573 		return kill_proc_info(sig, info, pid);
1574 
1575 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1576 	if (pid == INT_MIN)
1577 		return -ESRCH;
1578 
1579 	read_lock(&tasklist_lock);
1580 	if (pid != -1) {
1581 		ret = __kill_pgrp_info(sig, info,
1582 				pid ? find_vpid(-pid) : task_pgrp(current));
1583 	} else {
1584 		int retval = 0, count = 0;
1585 		struct task_struct * p;
1586 
1587 		for_each_process(p) {
1588 			if (task_pid_vnr(p) > 1 &&
1589 					!same_thread_group(p, current)) {
1590 				int err = group_send_sig_info(sig, info, p,
1591 							      PIDTYPE_MAX);
1592 				++count;
1593 				if (err != -EPERM)
1594 					retval = err;
1595 			}
1596 		}
1597 		ret = count ? retval : -ESRCH;
1598 	}
1599 	read_unlock(&tasklist_lock);
1600 
1601 	return ret;
1602 }
1603 
1604 /*
1605  * These are for backward compatibility with the rest of the kernel source.
1606  */
1607 
send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p)1608 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1609 {
1610 	/*
1611 	 * Make sure legacy kernel users don't send in bad values
1612 	 * (normal paths check this in check_kill_permission).
1613 	 */
1614 	if (!valid_signal(sig))
1615 		return -EINVAL;
1616 
1617 	return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1618 }
1619 EXPORT_SYMBOL(send_sig_info);
1620 
1621 #define __si_special(priv) \
1622 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1623 
1624 int
send_sig(int sig,struct task_struct * p,int priv)1625 send_sig(int sig, struct task_struct *p, int priv)
1626 {
1627 	return send_sig_info(sig, __si_special(priv), p);
1628 }
1629 EXPORT_SYMBOL(send_sig);
1630 
force_sig(int sig)1631 void force_sig(int sig)
1632 {
1633 	struct kernel_siginfo info;
1634 
1635 	clear_siginfo(&info);
1636 	info.si_signo = sig;
1637 	info.si_errno = 0;
1638 	info.si_code = SI_KERNEL;
1639 	info.si_pid = 0;
1640 	info.si_uid = 0;
1641 	force_sig_info(&info);
1642 }
1643 EXPORT_SYMBOL(force_sig);
1644 
1645 /*
1646  * When things go south during signal handling, we
1647  * will force a SIGSEGV. And if the signal that caused
1648  * the problem was already a SIGSEGV, we'll want to
1649  * make sure we don't even try to deliver the signal..
1650  */
force_sigsegv(int sig)1651 void force_sigsegv(int sig)
1652 {
1653 	struct task_struct *p = current;
1654 
1655 	if (sig == SIGSEGV) {
1656 		unsigned long flags;
1657 		spin_lock_irqsave(&p->sighand->siglock, flags);
1658 		p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1659 		spin_unlock_irqrestore(&p->sighand->siglock, flags);
1660 	}
1661 	force_sig(SIGSEGV);
1662 }
1663 
force_sig_fault_to_task(int sig,int code,void __user * addr ___ARCH_SI_TRAPNO (int trapno)___ARCH_SI_IA64 (int imm,unsigned int flags,unsigned long isr),struct task_struct * t)1664 int force_sig_fault_to_task(int sig, int code, void __user *addr
1665 	___ARCH_SI_TRAPNO(int trapno)
1666 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1667 	, struct task_struct *t)
1668 {
1669 	struct kernel_siginfo info;
1670 
1671 	clear_siginfo(&info);
1672 	info.si_signo = sig;
1673 	info.si_errno = 0;
1674 	info.si_code  = code;
1675 	info.si_addr  = addr;
1676 #ifdef __ARCH_SI_TRAPNO
1677 	info.si_trapno = trapno;
1678 #endif
1679 #ifdef __ia64__
1680 	info.si_imm = imm;
1681 	info.si_flags = flags;
1682 	info.si_isr = isr;
1683 #endif
1684 	return force_sig_info_to_task(&info, t);
1685 }
1686 
force_sig_fault(int sig,int code,void __user * addr ___ARCH_SI_TRAPNO (int trapno)___ARCH_SI_IA64 (int imm,unsigned int flags,unsigned long isr))1687 int force_sig_fault(int sig, int code, void __user *addr
1688 	___ARCH_SI_TRAPNO(int trapno)
1689 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr))
1690 {
1691 	return force_sig_fault_to_task(sig, code, addr
1692 				       ___ARCH_SI_TRAPNO(trapno)
1693 				       ___ARCH_SI_IA64(imm, flags, isr), current);
1694 }
1695 
send_sig_fault(int sig,int code,void __user * addr ___ARCH_SI_TRAPNO (int trapno)___ARCH_SI_IA64 (int imm,unsigned int flags,unsigned long isr),struct task_struct * t)1696 int send_sig_fault(int sig, int code, void __user *addr
1697 	___ARCH_SI_TRAPNO(int trapno)
1698 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1699 	, struct task_struct *t)
1700 {
1701 	struct kernel_siginfo info;
1702 
1703 	clear_siginfo(&info);
1704 	info.si_signo = sig;
1705 	info.si_errno = 0;
1706 	info.si_code  = code;
1707 	info.si_addr  = addr;
1708 #ifdef __ARCH_SI_TRAPNO
1709 	info.si_trapno = trapno;
1710 #endif
1711 #ifdef __ia64__
1712 	info.si_imm = imm;
1713 	info.si_flags = flags;
1714 	info.si_isr = isr;
1715 #endif
1716 	return send_sig_info(info.si_signo, &info, t);
1717 }
1718 
force_sig_mceerr(int code,void __user * addr,short lsb)1719 int force_sig_mceerr(int code, void __user *addr, short lsb)
1720 {
1721 	struct kernel_siginfo info;
1722 
1723 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1724 	clear_siginfo(&info);
1725 	info.si_signo = SIGBUS;
1726 	info.si_errno = 0;
1727 	info.si_code = code;
1728 	info.si_addr = addr;
1729 	info.si_addr_lsb = lsb;
1730 	return force_sig_info(&info);
1731 }
1732 
send_sig_mceerr(int code,void __user * addr,short lsb,struct task_struct * t)1733 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1734 {
1735 	struct kernel_siginfo info;
1736 
1737 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1738 	clear_siginfo(&info);
1739 	info.si_signo = SIGBUS;
1740 	info.si_errno = 0;
1741 	info.si_code = code;
1742 	info.si_addr = addr;
1743 	info.si_addr_lsb = lsb;
1744 	return send_sig_info(info.si_signo, &info, t);
1745 }
1746 EXPORT_SYMBOL(send_sig_mceerr);
1747 
force_sig_bnderr(void __user * addr,void __user * lower,void __user * upper)1748 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1749 {
1750 	struct kernel_siginfo info;
1751 
1752 	clear_siginfo(&info);
1753 	info.si_signo = SIGSEGV;
1754 	info.si_errno = 0;
1755 	info.si_code  = SEGV_BNDERR;
1756 	info.si_addr  = addr;
1757 	info.si_lower = lower;
1758 	info.si_upper = upper;
1759 	return force_sig_info(&info);
1760 }
1761 
1762 #ifdef SEGV_PKUERR
force_sig_pkuerr(void __user * addr,u32 pkey)1763 int force_sig_pkuerr(void __user *addr, u32 pkey)
1764 {
1765 	struct kernel_siginfo info;
1766 
1767 	clear_siginfo(&info);
1768 	info.si_signo = SIGSEGV;
1769 	info.si_errno = 0;
1770 	info.si_code  = SEGV_PKUERR;
1771 	info.si_addr  = addr;
1772 	info.si_pkey  = pkey;
1773 	return force_sig_info(&info);
1774 }
1775 #endif
1776 
1777 /* For the crazy architectures that include trap information in
1778  * the errno field, instead of an actual errno value.
1779  */
force_sig_ptrace_errno_trap(int errno,void __user * addr)1780 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1781 {
1782 	struct kernel_siginfo info;
1783 
1784 	clear_siginfo(&info);
1785 	info.si_signo = SIGTRAP;
1786 	info.si_errno = errno;
1787 	info.si_code  = TRAP_HWBKPT;
1788 	info.si_addr  = addr;
1789 	return force_sig_info(&info);
1790 }
1791 
kill_pgrp(struct pid * pid,int sig,int priv)1792 int kill_pgrp(struct pid *pid, int sig, int priv)
1793 {
1794 	int ret;
1795 
1796 	read_lock(&tasklist_lock);
1797 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1798 	read_unlock(&tasklist_lock);
1799 
1800 	return ret;
1801 }
1802 EXPORT_SYMBOL(kill_pgrp);
1803 
kill_pid(struct pid * pid,int sig,int priv)1804 int kill_pid(struct pid *pid, int sig, int priv)
1805 {
1806 	return kill_pid_info(sig, __si_special(priv), pid);
1807 }
1808 EXPORT_SYMBOL(kill_pid);
1809 
1810 /*
1811  * These functions support sending signals using preallocated sigqueue
1812  * structures.  This is needed "because realtime applications cannot
1813  * afford to lose notifications of asynchronous events, like timer
1814  * expirations or I/O completions".  In the case of POSIX Timers
1815  * we allocate the sigqueue structure from the timer_create.  If this
1816  * allocation fails we are able to report the failure to the application
1817  * with an EAGAIN error.
1818  */
sigqueue_alloc(void)1819 struct sigqueue *sigqueue_alloc(void)
1820 {
1821 	struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1822 
1823 	if (q)
1824 		q->flags |= SIGQUEUE_PREALLOC;
1825 
1826 	return q;
1827 }
1828 
sigqueue_free(struct sigqueue * q)1829 void sigqueue_free(struct sigqueue *q)
1830 {
1831 	unsigned long flags;
1832 	spinlock_t *lock = &current->sighand->siglock;
1833 
1834 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1835 	/*
1836 	 * We must hold ->siglock while testing q->list
1837 	 * to serialize with collect_signal() or with
1838 	 * __exit_signal()->flush_sigqueue().
1839 	 */
1840 	spin_lock_irqsave(lock, flags);
1841 	q->flags &= ~SIGQUEUE_PREALLOC;
1842 	/*
1843 	 * If it is queued it will be freed when dequeued,
1844 	 * like the "regular" sigqueue.
1845 	 */
1846 	if (!list_empty(&q->list))
1847 		q = NULL;
1848 	spin_unlock_irqrestore(lock, flags);
1849 
1850 	if (q)
1851 		__sigqueue_free(q);
1852 }
1853 
send_sigqueue(struct sigqueue * q,struct pid * pid,enum pid_type type)1854 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1855 {
1856 	int sig = q->info.si_signo;
1857 	struct sigpending *pending;
1858 	struct task_struct *t;
1859 	unsigned long flags;
1860 	int ret, result;
1861 
1862 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1863 
1864 	ret = -1;
1865 	rcu_read_lock();
1866 	t = pid_task(pid, type);
1867 	if (!t || !likely(lock_task_sighand(t, &flags)))
1868 		goto ret;
1869 
1870 	ret = 1; /* the signal is ignored */
1871 	result = TRACE_SIGNAL_IGNORED;
1872 	if (!prepare_signal(sig, t, false))
1873 		goto out;
1874 
1875 	ret = 0;
1876 	if (unlikely(!list_empty(&q->list))) {
1877 		/*
1878 		 * If an SI_TIMER entry is already queue just increment
1879 		 * the overrun count.
1880 		 */
1881 		BUG_ON(q->info.si_code != SI_TIMER);
1882 		q->info.si_overrun++;
1883 		result = TRACE_SIGNAL_ALREADY_PENDING;
1884 		goto out;
1885 	}
1886 	q->info.si_overrun = 0;
1887 
1888 	signalfd_notify(t, sig);
1889 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1890 	list_add_tail(&q->list, &pending->list);
1891 	sigaddset(&pending->signal, sig);
1892 	complete_signal(sig, t, type);
1893 	result = TRACE_SIGNAL_DELIVERED;
1894 out:
1895 	trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
1896 	unlock_task_sighand(t, &flags);
1897 ret:
1898 	rcu_read_unlock();
1899 	return ret;
1900 }
1901 
do_notify_pidfd(struct task_struct * task)1902 static void do_notify_pidfd(struct task_struct *task)
1903 {
1904 	struct pid *pid;
1905 
1906 	WARN_ON(task->exit_state == 0);
1907 	pid = task_pid(task);
1908 	wake_up_all(&pid->wait_pidfd);
1909 }
1910 
1911 /*
1912  * Let a parent know about the death of a child.
1913  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1914  *
1915  * Returns true if our parent ignored us and so we've switched to
1916  * self-reaping.
1917  */
do_notify_parent(struct task_struct * tsk,int sig)1918 bool do_notify_parent(struct task_struct *tsk, int sig)
1919 {
1920 	struct kernel_siginfo info;
1921 	unsigned long flags;
1922 	struct sighand_struct *psig;
1923 	bool autoreap = false;
1924 	u64 utime, stime;
1925 
1926 	WARN_ON_ONCE(sig == -1);
1927 
1928 	/* do_notify_parent_cldstop should have been called instead.  */
1929 	WARN_ON_ONCE(task_is_stopped_or_traced(tsk));
1930 
1931 	WARN_ON_ONCE(!tsk->ptrace &&
1932 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1933 
1934 	/* Wake up all pidfd waiters */
1935 	do_notify_pidfd(tsk);
1936 
1937 	if (sig != SIGCHLD) {
1938 		/*
1939 		 * This is only possible if parent == real_parent.
1940 		 * Check if it has changed security domain.
1941 		 */
1942 		if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
1943 			sig = SIGCHLD;
1944 	}
1945 
1946 	clear_siginfo(&info);
1947 	info.si_signo = sig;
1948 	info.si_errno = 0;
1949 	/*
1950 	 * We are under tasklist_lock here so our parent is tied to
1951 	 * us and cannot change.
1952 	 *
1953 	 * task_active_pid_ns will always return the same pid namespace
1954 	 * until a task passes through release_task.
1955 	 *
1956 	 * write_lock() currently calls preempt_disable() which is the
1957 	 * same as rcu_read_lock(), but according to Oleg, this is not
1958 	 * correct to rely on this
1959 	 */
1960 	rcu_read_lock();
1961 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1962 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1963 				       task_uid(tsk));
1964 	rcu_read_unlock();
1965 
1966 	task_cputime(tsk, &utime, &stime);
1967 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
1968 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
1969 
1970 	info.si_status = tsk->exit_code & 0x7f;
1971 	if (tsk->exit_code & 0x80)
1972 		info.si_code = CLD_DUMPED;
1973 	else if (tsk->exit_code & 0x7f)
1974 		info.si_code = CLD_KILLED;
1975 	else {
1976 		info.si_code = CLD_EXITED;
1977 		info.si_status = tsk->exit_code >> 8;
1978 	}
1979 
1980 	psig = tsk->parent->sighand;
1981 	spin_lock_irqsave(&psig->siglock, flags);
1982 	if (!tsk->ptrace && sig == SIGCHLD &&
1983 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1984 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1985 		/*
1986 		 * We are exiting and our parent doesn't care.  POSIX.1
1987 		 * defines special semantics for setting SIGCHLD to SIG_IGN
1988 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
1989 		 * automatically and not left for our parent's wait4 call.
1990 		 * Rather than having the parent do it as a magic kind of
1991 		 * signal handler, we just set this to tell do_exit that we
1992 		 * can be cleaned up without becoming a zombie.  Note that
1993 		 * we still call __wake_up_parent in this case, because a
1994 		 * blocked sys_wait4 might now return -ECHILD.
1995 		 *
1996 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1997 		 * is implementation-defined: we do (if you don't want
1998 		 * it, just use SIG_IGN instead).
1999 		 */
2000 		autoreap = true;
2001 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2002 			sig = 0;
2003 	}
2004 	/*
2005 	 * Send with __send_signal as si_pid and si_uid are in the
2006 	 * parent's namespaces.
2007 	 */
2008 	if (valid_signal(sig) && sig)
2009 		__send_signal(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2010 	__wake_up_parent(tsk, tsk->parent);
2011 	spin_unlock_irqrestore(&psig->siglock, flags);
2012 
2013 	return autoreap;
2014 }
2015 
2016 /**
2017  * do_notify_parent_cldstop - notify parent of stopped/continued state change
2018  * @tsk: task reporting the state change
2019  * @for_ptracer: the notification is for ptracer
2020  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2021  *
2022  * Notify @tsk's parent that the stopped/continued state has changed.  If
2023  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2024  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2025  *
2026  * CONTEXT:
2027  * Must be called with tasklist_lock at least read locked.
2028  */
do_notify_parent_cldstop(struct task_struct * tsk,bool for_ptracer,int why)2029 static void do_notify_parent_cldstop(struct task_struct *tsk,
2030 				     bool for_ptracer, int why)
2031 {
2032 	struct kernel_siginfo info;
2033 	unsigned long flags;
2034 	struct task_struct *parent;
2035 	struct sighand_struct *sighand;
2036 	u64 utime, stime;
2037 
2038 	if (for_ptracer) {
2039 		parent = tsk->parent;
2040 	} else {
2041 		tsk = tsk->group_leader;
2042 		parent = tsk->real_parent;
2043 	}
2044 
2045 	clear_siginfo(&info);
2046 	info.si_signo = SIGCHLD;
2047 	info.si_errno = 0;
2048 	/*
2049 	 * see comment in do_notify_parent() about the following 4 lines
2050 	 */
2051 	rcu_read_lock();
2052 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2053 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2054 	rcu_read_unlock();
2055 
2056 	task_cputime(tsk, &utime, &stime);
2057 	info.si_utime = nsec_to_clock_t(utime);
2058 	info.si_stime = nsec_to_clock_t(stime);
2059 
2060  	info.si_code = why;
2061  	switch (why) {
2062  	case CLD_CONTINUED:
2063  		info.si_status = SIGCONT;
2064  		break;
2065  	case CLD_STOPPED:
2066  		info.si_status = tsk->signal->group_exit_code & 0x7f;
2067  		break;
2068  	case CLD_TRAPPED:
2069  		info.si_status = tsk->exit_code & 0x7f;
2070  		break;
2071  	default:
2072  		BUG();
2073  	}
2074 
2075 	sighand = parent->sighand;
2076 	spin_lock_irqsave(&sighand->siglock, flags);
2077 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2078 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2079 		__group_send_sig_info(SIGCHLD, &info, parent);
2080 	/*
2081 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2082 	 */
2083 	__wake_up_parent(tsk, parent);
2084 	spin_unlock_irqrestore(&sighand->siglock, flags);
2085 }
2086 
may_ptrace_stop(void)2087 static inline bool may_ptrace_stop(void)
2088 {
2089 	if (!likely(current->ptrace))
2090 		return false;
2091 	/*
2092 	 * Are we in the middle of do_coredump?
2093 	 * If so and our tracer is also part of the coredump stopping
2094 	 * is a deadlock situation, and pointless because our tracer
2095 	 * is dead so don't allow us to stop.
2096 	 * If SIGKILL was already sent before the caller unlocked
2097 	 * ->siglock we must see ->core_state != NULL. Otherwise it
2098 	 * is safe to enter schedule().
2099 	 *
2100 	 * This is almost outdated, a task with the pending SIGKILL can't
2101 	 * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
2102 	 * after SIGKILL was already dequeued.
2103 	 */
2104 	if (unlikely(current->mm->core_state) &&
2105 	    unlikely(current->mm == current->parent->mm))
2106 		return false;
2107 
2108 	return true;
2109 }
2110 
2111 
2112 /*
2113  * This must be called with current->sighand->siglock held.
2114  *
2115  * This should be the path for all ptrace stops.
2116  * We always set current->last_siginfo while stopped here.
2117  * That makes it a way to test a stopped process for
2118  * being ptrace-stopped vs being job-control-stopped.
2119  *
2120  * If we actually decide not to stop at all because the tracer
2121  * is gone, we keep current->exit_code unless clear_code.
2122  */
ptrace_stop(int exit_code,int why,int clear_code,kernel_siginfo_t * info)2123 static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t *info)
2124 	__releases(&current->sighand->siglock)
2125 	__acquires(&current->sighand->siglock)
2126 {
2127 	bool gstop_done = false;
2128 
2129 	if (arch_ptrace_stop_needed(exit_code, info)) {
2130 		/*
2131 		 * The arch code has something special to do before a
2132 		 * ptrace stop.  This is allowed to block, e.g. for faults
2133 		 * on user stack pages.  We can't keep the siglock while
2134 		 * calling arch_ptrace_stop, so we must release it now.
2135 		 * To preserve proper semantics, we must do this before
2136 		 * any signal bookkeeping like checking group_stop_count.
2137 		 */
2138 		spin_unlock_irq(&current->sighand->siglock);
2139 		arch_ptrace_stop(exit_code, info);
2140 		spin_lock_irq(&current->sighand->siglock);
2141 	}
2142 
2143 	/*
2144 	 * schedule() will not sleep if there is a pending signal that
2145 	 * can awaken the task.
2146 	 */
2147 	set_special_state(TASK_TRACED);
2148 
2149 	/*
2150 	 * We're committing to trapping.  TRACED should be visible before
2151 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2152 	 * Also, transition to TRACED and updates to ->jobctl should be
2153 	 * atomic with respect to siglock and should be done after the arch
2154 	 * hook as siglock is released and regrabbed across it.
2155 	 *
2156 	 *     TRACER				    TRACEE
2157 	 *
2158 	 *     ptrace_attach()
2159 	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2160 	 *     do_wait()
2161 	 *       set_current_state()                smp_wmb();
2162 	 *       ptrace_do_wait()
2163 	 *         wait_task_stopped()
2164 	 *           task_stopped_code()
2165 	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2166 	 */
2167 	smp_wmb();
2168 
2169 	current->last_siginfo = info;
2170 	current->exit_code = exit_code;
2171 
2172 	/*
2173 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2174 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2175 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2176 	 * could be clear now.  We act as if SIGCONT is received after
2177 	 * TASK_TRACED is entered - ignore it.
2178 	 */
2179 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2180 		gstop_done = task_participate_group_stop(current);
2181 
2182 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2183 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2184 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2185 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2186 
2187 	/* entering a trap, clear TRAPPING */
2188 	task_clear_jobctl_trapping(current);
2189 
2190 	spin_unlock_irq(&current->sighand->siglock);
2191 	read_lock(&tasklist_lock);
2192 	if (may_ptrace_stop()) {
2193 		/*
2194 		 * Notify parents of the stop.
2195 		 *
2196 		 * While ptraced, there are two parents - the ptracer and
2197 		 * the real_parent of the group_leader.  The ptracer should
2198 		 * know about every stop while the real parent is only
2199 		 * interested in the completion of group stop.  The states
2200 		 * for the two don't interact with each other.  Notify
2201 		 * separately unless they're gonna be duplicates.
2202 		 */
2203 		do_notify_parent_cldstop(current, true, why);
2204 		if (gstop_done && ptrace_reparented(current))
2205 			do_notify_parent_cldstop(current, false, why);
2206 
2207 		/*
2208 		 * Don't want to allow preemption here, because
2209 		 * sys_ptrace() needs this task to be inactive.
2210 		 *
2211 		 * XXX: implement read_unlock_no_resched().
2212 		 */
2213 		preempt_disable();
2214 		read_unlock(&tasklist_lock);
2215 		cgroup_enter_frozen();
2216 		preempt_enable_no_resched();
2217 		freezable_schedule();
2218 		cgroup_leave_frozen(true);
2219 	} else {
2220 		/*
2221 		 * By the time we got the lock, our tracer went away.
2222 		 * Don't drop the lock yet, another tracer may come.
2223 		 *
2224 		 * If @gstop_done, the ptracer went away between group stop
2225 		 * completion and here.  During detach, it would have set
2226 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
2227 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
2228 		 * the real parent of the group stop completion is enough.
2229 		 */
2230 		if (gstop_done)
2231 			do_notify_parent_cldstop(current, false, why);
2232 
2233 		/* tasklist protects us from ptrace_freeze_traced() */
2234 		__set_current_state(TASK_RUNNING);
2235 		if (clear_code)
2236 			current->exit_code = 0;
2237 		read_unlock(&tasklist_lock);
2238 	}
2239 
2240 	/*
2241 	 * We are back.  Now reacquire the siglock before touching
2242 	 * last_siginfo, so that we are sure to have synchronized with
2243 	 * any signal-sending on another CPU that wants to examine it.
2244 	 */
2245 	spin_lock_irq(&current->sighand->siglock);
2246 	current->last_siginfo = NULL;
2247 
2248 	/* LISTENING can be set only during STOP traps, clear it */
2249 	current->jobctl &= ~JOBCTL_LISTENING;
2250 
2251 	/*
2252 	 * Queued signals ignored us while we were stopped for tracing.
2253 	 * So check for any that we should take before resuming user mode.
2254 	 * This sets TIF_SIGPENDING, but never clears it.
2255 	 */
2256 	recalc_sigpending_tsk(current);
2257 }
2258 
ptrace_do_notify(int signr,int exit_code,int why)2259 static void ptrace_do_notify(int signr, int exit_code, int why)
2260 {
2261 	kernel_siginfo_t info;
2262 
2263 	clear_siginfo(&info);
2264 	info.si_signo = signr;
2265 	info.si_code = exit_code;
2266 	info.si_pid = task_pid_vnr(current);
2267 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2268 
2269 	/* Let the debugger run.  */
2270 	ptrace_stop(exit_code, why, 1, &info);
2271 }
2272 
ptrace_notify(int exit_code)2273 void ptrace_notify(int exit_code)
2274 {
2275 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2276 	if (unlikely(current->task_works))
2277 		task_work_run();
2278 
2279 	spin_lock_irq(&current->sighand->siglock);
2280 	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
2281 	spin_unlock_irq(&current->sighand->siglock);
2282 }
2283 
2284 /**
2285  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2286  * @signr: signr causing group stop if initiating
2287  *
2288  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2289  * and participate in it.  If already set, participate in the existing
2290  * group stop.  If participated in a group stop (and thus slept), %true is
2291  * returned with siglock released.
2292  *
2293  * If ptraced, this function doesn't handle stop itself.  Instead,
2294  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2295  * untouched.  The caller must ensure that INTERRUPT trap handling takes
2296  * places afterwards.
2297  *
2298  * CONTEXT:
2299  * Must be called with @current->sighand->siglock held, which is released
2300  * on %true return.
2301  *
2302  * RETURNS:
2303  * %false if group stop is already cancelled or ptrace trap is scheduled.
2304  * %true if participated in group stop.
2305  */
do_signal_stop(int signr)2306 static bool do_signal_stop(int signr)
2307 	__releases(&current->sighand->siglock)
2308 {
2309 	struct signal_struct *sig = current->signal;
2310 
2311 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2312 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2313 		struct task_struct *t;
2314 
2315 		/* signr will be recorded in task->jobctl for retries */
2316 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2317 
2318 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2319 		    unlikely(signal_group_exit(sig)))
2320 			return false;
2321 		/*
2322 		 * There is no group stop already in progress.  We must
2323 		 * initiate one now.
2324 		 *
2325 		 * While ptraced, a task may be resumed while group stop is
2326 		 * still in effect and then receive a stop signal and
2327 		 * initiate another group stop.  This deviates from the
2328 		 * usual behavior as two consecutive stop signals can't
2329 		 * cause two group stops when !ptraced.  That is why we
2330 		 * also check !task_is_stopped(t) below.
2331 		 *
2332 		 * The condition can be distinguished by testing whether
2333 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2334 		 * group_exit_code in such case.
2335 		 *
2336 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2337 		 * an intervening stop signal is required to cause two
2338 		 * continued events regardless of ptrace.
2339 		 */
2340 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2341 			sig->group_exit_code = signr;
2342 
2343 		sig->group_stop_count = 0;
2344 
2345 		if (task_set_jobctl_pending(current, signr | gstop))
2346 			sig->group_stop_count++;
2347 
2348 		t = current;
2349 		while_each_thread(current, t) {
2350 			/*
2351 			 * Setting state to TASK_STOPPED for a group
2352 			 * stop is always done with the siglock held,
2353 			 * so this check has no races.
2354 			 */
2355 			if (!task_is_stopped(t) &&
2356 			    task_set_jobctl_pending(t, signr | gstop)) {
2357 				sig->group_stop_count++;
2358 				if (likely(!(t->ptrace & PT_SEIZED)))
2359 					signal_wake_up(t, 0);
2360 				else
2361 					ptrace_trap_notify(t);
2362 			}
2363 		}
2364 	}
2365 
2366 	if (likely(!current->ptrace)) {
2367 		int notify = 0;
2368 
2369 		/*
2370 		 * If there are no other threads in the group, or if there
2371 		 * is a group stop in progress and we are the last to stop,
2372 		 * report to the parent.
2373 		 */
2374 		if (task_participate_group_stop(current))
2375 			notify = CLD_STOPPED;
2376 
2377 		set_special_state(TASK_STOPPED);
2378 		spin_unlock_irq(&current->sighand->siglock);
2379 
2380 		/*
2381 		 * Notify the parent of the group stop completion.  Because
2382 		 * we're not holding either the siglock or tasklist_lock
2383 		 * here, ptracer may attach inbetween; however, this is for
2384 		 * group stop and should always be delivered to the real
2385 		 * parent of the group leader.  The new ptracer will get
2386 		 * its notification when this task transitions into
2387 		 * TASK_TRACED.
2388 		 */
2389 		if (notify) {
2390 			read_lock(&tasklist_lock);
2391 			do_notify_parent_cldstop(current, false, notify);
2392 			read_unlock(&tasklist_lock);
2393 		}
2394 
2395 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2396 		cgroup_enter_frozen();
2397 		freezable_schedule();
2398 		return true;
2399 	} else {
2400 		/*
2401 		 * While ptraced, group stop is handled by STOP trap.
2402 		 * Schedule it and let the caller deal with it.
2403 		 */
2404 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2405 		return false;
2406 	}
2407 }
2408 
2409 /**
2410  * do_jobctl_trap - take care of ptrace jobctl traps
2411  *
2412  * When PT_SEIZED, it's used for both group stop and explicit
2413  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2414  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2415  * the stop signal; otherwise, %SIGTRAP.
2416  *
2417  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2418  * number as exit_code and no siginfo.
2419  *
2420  * CONTEXT:
2421  * Must be called with @current->sighand->siglock held, which may be
2422  * released and re-acquired before returning with intervening sleep.
2423  */
do_jobctl_trap(void)2424 static void do_jobctl_trap(void)
2425 {
2426 	struct signal_struct *signal = current->signal;
2427 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2428 
2429 	if (current->ptrace & PT_SEIZED) {
2430 		if (!signal->group_stop_count &&
2431 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2432 			signr = SIGTRAP;
2433 		WARN_ON_ONCE(!signr);
2434 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2435 				 CLD_STOPPED);
2436 	} else {
2437 		WARN_ON_ONCE(!signr);
2438 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2439 		current->exit_code = 0;
2440 	}
2441 }
2442 
2443 /**
2444  * do_freezer_trap - handle the freezer jobctl trap
2445  *
2446  * Puts the task into frozen state, if only the task is not about to quit.
2447  * In this case it drops JOBCTL_TRAP_FREEZE.
2448  *
2449  * CONTEXT:
2450  * Must be called with @current->sighand->siglock held,
2451  * which is always released before returning.
2452  */
do_freezer_trap(void)2453 static void do_freezer_trap(void)
2454 	__releases(&current->sighand->siglock)
2455 {
2456 	/*
2457 	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2458 	 * let's make another loop to give it a chance to be handled.
2459 	 * In any case, we'll return back.
2460 	 */
2461 	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2462 	     JOBCTL_TRAP_FREEZE) {
2463 		spin_unlock_irq(&current->sighand->siglock);
2464 		return;
2465 	}
2466 
2467 	/*
2468 	 * Now we're sure that there is no pending fatal signal and no
2469 	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2470 	 * immediately (if there is a non-fatal signal pending), and
2471 	 * put the task into sleep.
2472 	 */
2473 	__set_current_state(TASK_INTERRUPTIBLE);
2474 	clear_thread_flag(TIF_SIGPENDING);
2475 	spin_unlock_irq(&current->sighand->siglock);
2476 	cgroup_enter_frozen();
2477 	freezable_schedule();
2478 }
2479 
ptrace_signal(int signr,kernel_siginfo_t * info)2480 static int ptrace_signal(int signr, kernel_siginfo_t *info)
2481 {
2482 	/*
2483 	 * We do not check sig_kernel_stop(signr) but set this marker
2484 	 * unconditionally because we do not know whether debugger will
2485 	 * change signr. This flag has no meaning unless we are going
2486 	 * to stop after return from ptrace_stop(). In this case it will
2487 	 * be checked in do_signal_stop(), we should only stop if it was
2488 	 * not cleared by SIGCONT while we were sleeping. See also the
2489 	 * comment in dequeue_signal().
2490 	 */
2491 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2492 	ptrace_stop(signr, CLD_TRAPPED, 0, info);
2493 
2494 	/* We're back.  Did the debugger cancel the sig?  */
2495 	signr = current->exit_code;
2496 	if (signr == 0)
2497 		return signr;
2498 
2499 	current->exit_code = 0;
2500 
2501 	/*
2502 	 * Update the siginfo structure if the signal has
2503 	 * changed.  If the debugger wanted something
2504 	 * specific in the siginfo structure then it should
2505 	 * have updated *info via PTRACE_SETSIGINFO.
2506 	 */
2507 	if (signr != info->si_signo) {
2508 		clear_siginfo(info);
2509 		info->si_signo = signr;
2510 		info->si_errno = 0;
2511 		info->si_code = SI_USER;
2512 		rcu_read_lock();
2513 		info->si_pid = task_pid_vnr(current->parent);
2514 		info->si_uid = from_kuid_munged(current_user_ns(),
2515 						task_uid(current->parent));
2516 		rcu_read_unlock();
2517 	}
2518 
2519 	/* If the (new) signal is now blocked, requeue it.  */
2520 	if (sigismember(&current->blocked, signr)) {
2521 		send_signal(signr, info, current, PIDTYPE_PID);
2522 		signr = 0;
2523 	}
2524 
2525 	return signr;
2526 }
2527 
hide_si_addr_tag_bits(struct ksignal * ksig)2528 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2529 {
2530 	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2531 	case SIL_FAULT:
2532 	case SIL_FAULT_MCEERR:
2533 	case SIL_FAULT_BNDERR:
2534 	case SIL_FAULT_PKUERR:
2535 		ksig->info.si_addr = arch_untagged_si_addr(
2536 			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2537 		break;
2538 	case SIL_KILL:
2539 	case SIL_TIMER:
2540 	case SIL_POLL:
2541 	case SIL_CHLD:
2542 	case SIL_RT:
2543 	case SIL_SYS:
2544 		break;
2545 	}
2546 }
2547 
get_signal(struct ksignal * ksig)2548 bool get_signal(struct ksignal *ksig)
2549 {
2550 	struct sighand_struct *sighand = current->sighand;
2551 	struct signal_struct *signal = current->signal;
2552 	int signr;
2553 
2554 	if (unlikely(current->task_works))
2555 		task_work_run();
2556 
2557 	/*
2558 	 * For non-generic architectures, check for TIF_NOTIFY_SIGNAL so
2559 	 * that the arch handlers don't all have to do it. If we get here
2560 	 * without TIF_SIGPENDING, just exit after running signal work.
2561 	 */
2562 	if (!IS_ENABLED(CONFIG_GENERIC_ENTRY)) {
2563 		if (test_thread_flag(TIF_NOTIFY_SIGNAL))
2564 			tracehook_notify_signal();
2565 		if (!task_sigpending(current))
2566 			return false;
2567 	}
2568 
2569 	if (unlikely(uprobe_deny_signal()))
2570 		return false;
2571 
2572 	/*
2573 	 * Do this once, we can't return to user-mode if freezing() == T.
2574 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2575 	 * thus do not need another check after return.
2576 	 */
2577 	try_to_freeze();
2578 
2579 relock:
2580 	spin_lock_irq(&sighand->siglock);
2581 
2582 	/*
2583 	 * Every stopped thread goes here after wakeup. Check to see if
2584 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2585 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2586 	 */
2587 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2588 		int why;
2589 
2590 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2591 			why = CLD_CONTINUED;
2592 		else
2593 			why = CLD_STOPPED;
2594 
2595 		signal->flags &= ~SIGNAL_CLD_MASK;
2596 
2597 		spin_unlock_irq(&sighand->siglock);
2598 
2599 		/*
2600 		 * Notify the parent that we're continuing.  This event is
2601 		 * always per-process and doesn't make whole lot of sense
2602 		 * for ptracers, who shouldn't consume the state via
2603 		 * wait(2) either, but, for backward compatibility, notify
2604 		 * the ptracer of the group leader too unless it's gonna be
2605 		 * a duplicate.
2606 		 */
2607 		read_lock(&tasklist_lock);
2608 		do_notify_parent_cldstop(current, false, why);
2609 
2610 		if (ptrace_reparented(current->group_leader))
2611 			do_notify_parent_cldstop(current->group_leader,
2612 						true, why);
2613 		read_unlock(&tasklist_lock);
2614 
2615 		goto relock;
2616 	}
2617 
2618 	/* Has this task already been marked for death? */
2619 	if (signal_group_exit(signal)) {
2620 		ksig->info.si_signo = signr = SIGKILL;
2621 		sigdelset(&current->pending.signal, SIGKILL);
2622 		trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2623 				&sighand->action[SIGKILL - 1]);
2624 		recalc_sigpending();
2625 		goto fatal;
2626 	}
2627 
2628 	for (;;) {
2629 		struct k_sigaction *ka;
2630 
2631 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2632 		    do_signal_stop(0))
2633 			goto relock;
2634 
2635 		if (unlikely(current->jobctl &
2636 			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2637 			if (current->jobctl & JOBCTL_TRAP_MASK) {
2638 				do_jobctl_trap();
2639 				spin_unlock_irq(&sighand->siglock);
2640 			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2641 				do_freezer_trap();
2642 
2643 			goto relock;
2644 		}
2645 
2646 		/*
2647 		 * If the task is leaving the frozen state, let's update
2648 		 * cgroup counters and reset the frozen bit.
2649 		 */
2650 		if (unlikely(cgroup_task_frozen(current))) {
2651 			spin_unlock_irq(&sighand->siglock);
2652 			cgroup_leave_frozen(false);
2653 			goto relock;
2654 		}
2655 
2656 		/*
2657 		 * Signals generated by the execution of an instruction
2658 		 * need to be delivered before any other pending signals
2659 		 * so that the instruction pointer in the signal stack
2660 		 * frame points to the faulting instruction.
2661 		 */
2662 		signr = dequeue_synchronous_signal(&ksig->info);
2663 		if (!signr)
2664 			signr = dequeue_signal(current, &current->blocked, &ksig->info);
2665 
2666 		if (!signr)
2667 			break; /* will return 0 */
2668 
2669 		if (unlikely(current->ptrace) && signr != SIGKILL) {
2670 			signr = ptrace_signal(signr, &ksig->info);
2671 			if (!signr)
2672 				continue;
2673 		}
2674 
2675 		ka = &sighand->action[signr-1];
2676 
2677 		/* Trace actually delivered signals. */
2678 		trace_signal_deliver(signr, &ksig->info, ka);
2679 
2680 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2681 			continue;
2682 		if (ka->sa.sa_handler != SIG_DFL) {
2683 			/* Run the handler.  */
2684 			ksig->ka = *ka;
2685 
2686 			if (ka->sa.sa_flags & SA_ONESHOT)
2687 				ka->sa.sa_handler = SIG_DFL;
2688 
2689 			break; /* will return non-zero "signr" value */
2690 		}
2691 
2692 		/*
2693 		 * Now we are doing the default action for this signal.
2694 		 */
2695 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2696 			continue;
2697 
2698 		/*
2699 		 * Global init gets no signals it doesn't want.
2700 		 * Container-init gets no signals it doesn't want from same
2701 		 * container.
2702 		 *
2703 		 * Note that if global/container-init sees a sig_kernel_only()
2704 		 * signal here, the signal must have been generated internally
2705 		 * or must have come from an ancestor namespace. In either
2706 		 * case, the signal cannot be dropped.
2707 		 */
2708 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2709 				!sig_kernel_only(signr))
2710 			continue;
2711 
2712 		if (sig_kernel_stop(signr)) {
2713 			/*
2714 			 * The default action is to stop all threads in
2715 			 * the thread group.  The job control signals
2716 			 * do nothing in an orphaned pgrp, but SIGSTOP
2717 			 * always works.  Note that siglock needs to be
2718 			 * dropped during the call to is_orphaned_pgrp()
2719 			 * because of lock ordering with tasklist_lock.
2720 			 * This allows an intervening SIGCONT to be posted.
2721 			 * We need to check for that and bail out if necessary.
2722 			 */
2723 			if (signr != SIGSTOP) {
2724 				spin_unlock_irq(&sighand->siglock);
2725 
2726 				/* signals can be posted during this window */
2727 
2728 				if (is_current_pgrp_orphaned())
2729 					goto relock;
2730 
2731 				spin_lock_irq(&sighand->siglock);
2732 			}
2733 
2734 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2735 				/* It released the siglock.  */
2736 				goto relock;
2737 			}
2738 
2739 			/*
2740 			 * We didn't actually stop, due to a race
2741 			 * with SIGCONT or something like that.
2742 			 */
2743 			continue;
2744 		}
2745 
2746 	fatal:
2747 		spin_unlock_irq(&sighand->siglock);
2748 		if (unlikely(cgroup_task_frozen(current)))
2749 			cgroup_leave_frozen(true);
2750 
2751 		/*
2752 		 * Anything else is fatal, maybe with a core dump.
2753 		 */
2754 		current->flags |= PF_SIGNALED;
2755 
2756 		if (sig_kernel_coredump(signr)) {
2757 			if (print_fatal_signals)
2758 				print_fatal_signal(ksig->info.si_signo);
2759 			proc_coredump_connector(current);
2760 			/*
2761 			 * If it was able to dump core, this kills all
2762 			 * other threads in the group and synchronizes with
2763 			 * their demise.  If we lost the race with another
2764 			 * thread getting here, it set group_exit_code
2765 			 * first and our do_group_exit call below will use
2766 			 * that value and ignore the one we pass it.
2767 			 */
2768 			do_coredump(&ksig->info);
2769 		}
2770 
2771 		/*
2772 		 * PF_IO_WORKER threads will catch and exit on fatal signals
2773 		 * themselves. They have cleanup that must be performed, so
2774 		 * we cannot call do_exit() on their behalf.
2775 		 */
2776 		if (current->flags & PF_IO_WORKER)
2777 			goto out;
2778 
2779 		/*
2780 		 * Death signals, no core dump.
2781 		 */
2782 		do_group_exit(ksig->info.si_signo);
2783 		/* NOTREACHED */
2784 	}
2785 	spin_unlock_irq(&sighand->siglock);
2786 out:
2787 	ksig->sig = signr;
2788 
2789 	if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2790 		hide_si_addr_tag_bits(ksig);
2791 
2792 	return ksig->sig > 0;
2793 }
2794 
2795 /**
2796  * signal_delivered -
2797  * @ksig:		kernel signal struct
2798  * @stepping:		nonzero if debugger single-step or block-step in use
2799  *
2800  * This function should be called when a signal has successfully been
2801  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2802  * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2803  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2804  */
signal_delivered(struct ksignal * ksig,int stepping)2805 static void signal_delivered(struct ksignal *ksig, int stepping)
2806 {
2807 	sigset_t blocked;
2808 
2809 	/* A signal was successfully delivered, and the
2810 	   saved sigmask was stored on the signal frame,
2811 	   and will be restored by sigreturn.  So we can
2812 	   simply clear the restore sigmask flag.  */
2813 	clear_restore_sigmask();
2814 
2815 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2816 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2817 		sigaddset(&blocked, ksig->sig);
2818 	set_current_blocked(&blocked);
2819 	tracehook_signal_handler(stepping);
2820 }
2821 
signal_setup_done(int failed,struct ksignal * ksig,int stepping)2822 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2823 {
2824 	if (failed)
2825 		force_sigsegv(ksig->sig);
2826 	else
2827 		signal_delivered(ksig, stepping);
2828 }
2829 
2830 /*
2831  * It could be that complete_signal() picked us to notify about the
2832  * group-wide signal. Other threads should be notified now to take
2833  * the shared signals in @which since we will not.
2834  */
retarget_shared_pending(struct task_struct * tsk,sigset_t * which)2835 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2836 {
2837 	sigset_t retarget;
2838 	struct task_struct *t;
2839 
2840 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2841 	if (sigisemptyset(&retarget))
2842 		return;
2843 
2844 	t = tsk;
2845 	while_each_thread(tsk, t) {
2846 		if (t->flags & PF_EXITING)
2847 			continue;
2848 
2849 		if (!has_pending_signals(&retarget, &t->blocked))
2850 			continue;
2851 		/* Remove the signals this thread can handle. */
2852 		sigandsets(&retarget, &retarget, &t->blocked);
2853 
2854 		if (!task_sigpending(t))
2855 			signal_wake_up(t, 0);
2856 
2857 		if (sigisemptyset(&retarget))
2858 			break;
2859 	}
2860 }
2861 
exit_signals(struct task_struct * tsk)2862 void exit_signals(struct task_struct *tsk)
2863 {
2864 	int group_stop = 0;
2865 	sigset_t unblocked;
2866 
2867 	/*
2868 	 * @tsk is about to have PF_EXITING set - lock out users which
2869 	 * expect stable threadgroup.
2870 	 */
2871 	cgroup_threadgroup_change_begin(tsk);
2872 
2873 	if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2874 		tsk->flags |= PF_EXITING;
2875 		cgroup_threadgroup_change_end(tsk);
2876 		return;
2877 	}
2878 
2879 	spin_lock_irq(&tsk->sighand->siglock);
2880 	/*
2881 	 * From now this task is not visible for group-wide signals,
2882 	 * see wants_signal(), do_signal_stop().
2883 	 */
2884 	tsk->flags |= PF_EXITING;
2885 
2886 	cgroup_threadgroup_change_end(tsk);
2887 
2888 	if (!task_sigpending(tsk))
2889 		goto out;
2890 
2891 	unblocked = tsk->blocked;
2892 	signotset(&unblocked);
2893 	retarget_shared_pending(tsk, &unblocked);
2894 
2895 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2896 	    task_participate_group_stop(tsk))
2897 		group_stop = CLD_STOPPED;
2898 out:
2899 	spin_unlock_irq(&tsk->sighand->siglock);
2900 
2901 	/*
2902 	 * If group stop has completed, deliver the notification.  This
2903 	 * should always go to the real parent of the group leader.
2904 	 */
2905 	if (unlikely(group_stop)) {
2906 		read_lock(&tasklist_lock);
2907 		do_notify_parent_cldstop(tsk, false, group_stop);
2908 		read_unlock(&tasklist_lock);
2909 	}
2910 }
2911 
2912 /*
2913  * System call entry points.
2914  */
2915 
2916 /**
2917  *  sys_restart_syscall - restart a system call
2918  */
SYSCALL_DEFINE0(restart_syscall)2919 SYSCALL_DEFINE0(restart_syscall)
2920 {
2921 	struct restart_block *restart = &current->restart_block;
2922 	return restart->fn(restart);
2923 }
2924 
do_no_restart_syscall(struct restart_block * param)2925 long do_no_restart_syscall(struct restart_block *param)
2926 {
2927 	return -EINTR;
2928 }
2929 
__set_task_blocked(struct task_struct * tsk,const sigset_t * newset)2930 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2931 {
2932 	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
2933 		sigset_t newblocked;
2934 		/* A set of now blocked but previously unblocked signals. */
2935 		sigandnsets(&newblocked, newset, &current->blocked);
2936 		retarget_shared_pending(tsk, &newblocked);
2937 	}
2938 	tsk->blocked = *newset;
2939 	recalc_sigpending();
2940 }
2941 
2942 /**
2943  * set_current_blocked - change current->blocked mask
2944  * @newset: new mask
2945  *
2946  * It is wrong to change ->blocked directly, this helper should be used
2947  * to ensure the process can't miss a shared signal we are going to block.
2948  */
set_current_blocked(sigset_t * newset)2949 void set_current_blocked(sigset_t *newset)
2950 {
2951 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2952 	__set_current_blocked(newset);
2953 }
2954 
__set_current_blocked(const sigset_t * newset)2955 void __set_current_blocked(const sigset_t *newset)
2956 {
2957 	struct task_struct *tsk = current;
2958 
2959 	/*
2960 	 * In case the signal mask hasn't changed, there is nothing we need
2961 	 * to do. The current->blocked shouldn't be modified by other task.
2962 	 */
2963 	if (sigequalsets(&tsk->blocked, newset))
2964 		return;
2965 
2966 	spin_lock_irq(&tsk->sighand->siglock);
2967 	__set_task_blocked(tsk, newset);
2968 	spin_unlock_irq(&tsk->sighand->siglock);
2969 }
2970 
2971 /*
2972  * This is also useful for kernel threads that want to temporarily
2973  * (or permanently) block certain signals.
2974  *
2975  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2976  * interface happily blocks "unblockable" signals like SIGKILL
2977  * and friends.
2978  */
sigprocmask(int how,sigset_t * set,sigset_t * oldset)2979 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2980 {
2981 	struct task_struct *tsk = current;
2982 	sigset_t newset;
2983 
2984 	/* Lockless, only current can change ->blocked, never from irq */
2985 	if (oldset)
2986 		*oldset = tsk->blocked;
2987 
2988 	switch (how) {
2989 	case SIG_BLOCK:
2990 		sigorsets(&newset, &tsk->blocked, set);
2991 		break;
2992 	case SIG_UNBLOCK:
2993 		sigandnsets(&newset, &tsk->blocked, set);
2994 		break;
2995 	case SIG_SETMASK:
2996 		newset = *set;
2997 		break;
2998 	default:
2999 		return -EINVAL;
3000 	}
3001 
3002 	__set_current_blocked(&newset);
3003 	return 0;
3004 }
3005 EXPORT_SYMBOL(sigprocmask);
3006 
3007 /*
3008  * The api helps set app-provided sigmasks.
3009  *
3010  * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3011  * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3012  *
3013  * Note that it does set_restore_sigmask() in advance, so it must be always
3014  * paired with restore_saved_sigmask_unless() before return from syscall.
3015  */
set_user_sigmask(const sigset_t __user * umask,size_t sigsetsize)3016 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3017 {
3018 	sigset_t kmask;
3019 
3020 	if (!umask)
3021 		return 0;
3022 	if (sigsetsize != sizeof(sigset_t))
3023 		return -EINVAL;
3024 	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3025 		return -EFAULT;
3026 
3027 	set_restore_sigmask();
3028 	current->saved_sigmask = current->blocked;
3029 	set_current_blocked(&kmask);
3030 
3031 	return 0;
3032 }
3033 
3034 #ifdef CONFIG_COMPAT
set_compat_user_sigmask(const compat_sigset_t __user * umask,size_t sigsetsize)3035 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3036 			    size_t sigsetsize)
3037 {
3038 	sigset_t kmask;
3039 
3040 	if (!umask)
3041 		return 0;
3042 	if (sigsetsize != sizeof(compat_sigset_t))
3043 		return -EINVAL;
3044 	if (get_compat_sigset(&kmask, umask))
3045 		return -EFAULT;
3046 
3047 	set_restore_sigmask();
3048 	current->saved_sigmask = current->blocked;
3049 	set_current_blocked(&kmask);
3050 
3051 	return 0;
3052 }
3053 #endif
3054 
3055 /**
3056  *  sys_rt_sigprocmask - change the list of currently blocked signals
3057  *  @how: whether to add, remove, or set signals
3058  *  @nset: stores pending signals
3059  *  @oset: previous value of signal mask if non-null
3060  *  @sigsetsize: size of sigset_t type
3061  */
SYSCALL_DEFINE4(rt_sigprocmask,int,how,sigset_t __user *,nset,sigset_t __user *,oset,size_t,sigsetsize)3062 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3063 		sigset_t __user *, oset, size_t, sigsetsize)
3064 {
3065 	sigset_t old_set, new_set;
3066 	int error;
3067 
3068 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3069 	if (sigsetsize != sizeof(sigset_t))
3070 		return -EINVAL;
3071 
3072 	old_set = current->blocked;
3073 
3074 	if (nset) {
3075 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3076 			return -EFAULT;
3077 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3078 
3079 		error = sigprocmask(how, &new_set, NULL);
3080 		if (error)
3081 			return error;
3082 	}
3083 
3084 	if (oset) {
3085 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3086 			return -EFAULT;
3087 	}
3088 
3089 	return 0;
3090 }
3091 
3092 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_sigprocmask,int,how,compat_sigset_t __user *,nset,compat_sigset_t __user *,oset,compat_size_t,sigsetsize)3093 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3094 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3095 {
3096 	sigset_t old_set = current->blocked;
3097 
3098 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3099 	if (sigsetsize != sizeof(sigset_t))
3100 		return -EINVAL;
3101 
3102 	if (nset) {
3103 		sigset_t new_set;
3104 		int error;
3105 		if (get_compat_sigset(&new_set, nset))
3106 			return -EFAULT;
3107 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3108 
3109 		error = sigprocmask(how, &new_set, NULL);
3110 		if (error)
3111 			return error;
3112 	}
3113 	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3114 }
3115 #endif
3116 
do_sigpending(sigset_t * set)3117 static void do_sigpending(sigset_t *set)
3118 {
3119 	spin_lock_irq(&current->sighand->siglock);
3120 	sigorsets(set, &current->pending.signal,
3121 		  &current->signal->shared_pending.signal);
3122 	spin_unlock_irq(&current->sighand->siglock);
3123 
3124 	/* Outside the lock because only this thread touches it.  */
3125 	sigandsets(set, &current->blocked, set);
3126 }
3127 
3128 /**
3129  *  sys_rt_sigpending - examine a pending signal that has been raised
3130  *			while blocked
3131  *  @uset: stores pending signals
3132  *  @sigsetsize: size of sigset_t type or larger
3133  */
SYSCALL_DEFINE2(rt_sigpending,sigset_t __user *,uset,size_t,sigsetsize)3134 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3135 {
3136 	sigset_t set;
3137 
3138 	if (sigsetsize > sizeof(*uset))
3139 		return -EINVAL;
3140 
3141 	do_sigpending(&set);
3142 
3143 	if (copy_to_user(uset, &set, sigsetsize))
3144 		return -EFAULT;
3145 
3146 	return 0;
3147 }
3148 
3149 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(rt_sigpending,compat_sigset_t __user *,uset,compat_size_t,sigsetsize)3150 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3151 		compat_size_t, sigsetsize)
3152 {
3153 	sigset_t set;
3154 
3155 	if (sigsetsize > sizeof(*uset))
3156 		return -EINVAL;
3157 
3158 	do_sigpending(&set);
3159 
3160 	return put_compat_sigset(uset, &set, sigsetsize);
3161 }
3162 #endif
3163 
3164 static const struct {
3165 	unsigned char limit, layout;
3166 } sig_sicodes[] = {
3167 	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3168 	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3169 	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3170 	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3171 	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3172 #if defined(SIGEMT)
3173 	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3174 #endif
3175 	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3176 	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3177 	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3178 };
3179 
known_siginfo_layout(unsigned sig,int si_code)3180 static bool known_siginfo_layout(unsigned sig, int si_code)
3181 {
3182 	if (si_code == SI_KERNEL)
3183 		return true;
3184 	else if ((si_code > SI_USER)) {
3185 		if (sig_specific_sicodes(sig)) {
3186 			if (si_code <= sig_sicodes[sig].limit)
3187 				return true;
3188 		}
3189 		else if (si_code <= NSIGPOLL)
3190 			return true;
3191 	}
3192 	else if (si_code >= SI_DETHREAD)
3193 		return true;
3194 	else if (si_code == SI_ASYNCNL)
3195 		return true;
3196 	return false;
3197 }
3198 
siginfo_layout(unsigned sig,int si_code)3199 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3200 {
3201 	enum siginfo_layout layout = SIL_KILL;
3202 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3203 		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3204 		    (si_code <= sig_sicodes[sig].limit)) {
3205 			layout = sig_sicodes[sig].layout;
3206 			/* Handle the exceptions */
3207 			if ((sig == SIGBUS) &&
3208 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3209 				layout = SIL_FAULT_MCEERR;
3210 			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3211 				layout = SIL_FAULT_BNDERR;
3212 #ifdef SEGV_PKUERR
3213 			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3214 				layout = SIL_FAULT_PKUERR;
3215 #endif
3216 		}
3217 		else if (si_code <= NSIGPOLL)
3218 			layout = SIL_POLL;
3219 	} else {
3220 		if (si_code == SI_TIMER)
3221 			layout = SIL_TIMER;
3222 		else if (si_code == SI_SIGIO)
3223 			layout = SIL_POLL;
3224 		else if (si_code < 0)
3225 			layout = SIL_RT;
3226 	}
3227 	return layout;
3228 }
3229 
si_expansion(const siginfo_t __user * info)3230 static inline char __user *si_expansion(const siginfo_t __user *info)
3231 {
3232 	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3233 }
3234 
copy_siginfo_to_user(siginfo_t __user * to,const kernel_siginfo_t * from)3235 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3236 {
3237 	char __user *expansion = si_expansion(to);
3238 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3239 		return -EFAULT;
3240 	if (clear_user(expansion, SI_EXPANSION_SIZE))
3241 		return -EFAULT;
3242 	return 0;
3243 }
3244 
post_copy_siginfo_from_user(kernel_siginfo_t * info,const siginfo_t __user * from)3245 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3246 				       const siginfo_t __user *from)
3247 {
3248 	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3249 		char __user *expansion = si_expansion(from);
3250 		char buf[SI_EXPANSION_SIZE];
3251 		int i;
3252 		/*
3253 		 * An unknown si_code might need more than
3254 		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3255 		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3256 		 * will return this data to userspace exactly.
3257 		 */
3258 		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3259 			return -EFAULT;
3260 		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3261 			if (buf[i] != 0)
3262 				return -E2BIG;
3263 		}
3264 	}
3265 	return 0;
3266 }
3267 
__copy_siginfo_from_user(int signo,kernel_siginfo_t * to,const siginfo_t __user * from)3268 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3269 				    const siginfo_t __user *from)
3270 {
3271 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3272 		return -EFAULT;
3273 	to->si_signo = signo;
3274 	return post_copy_siginfo_from_user(to, from);
3275 }
3276 
copy_siginfo_from_user(kernel_siginfo_t * to,const siginfo_t __user * from)3277 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3278 {
3279 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3280 		return -EFAULT;
3281 	return post_copy_siginfo_from_user(to, from);
3282 }
3283 
3284 #ifdef CONFIG_COMPAT
3285 /**
3286  * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3287  * @to: compat siginfo destination
3288  * @from: kernel siginfo source
3289  *
3290  * Note: This function does not work properly for the SIGCHLD on x32, but
3291  * fortunately it doesn't have to.  The only valid callers for this function are
3292  * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3293  * The latter does not care because SIGCHLD will never cause a coredump.
3294  */
copy_siginfo_to_external32(struct compat_siginfo * to,const struct kernel_siginfo * from)3295 void copy_siginfo_to_external32(struct compat_siginfo *to,
3296 		const struct kernel_siginfo *from)
3297 {
3298 	memset(to, 0, sizeof(*to));
3299 
3300 	to->si_signo = from->si_signo;
3301 	to->si_errno = from->si_errno;
3302 	to->si_code  = from->si_code;
3303 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3304 	case SIL_KILL:
3305 		to->si_pid = from->si_pid;
3306 		to->si_uid = from->si_uid;
3307 		break;
3308 	case SIL_TIMER:
3309 		to->si_tid     = from->si_tid;
3310 		to->si_overrun = from->si_overrun;
3311 		to->si_int     = from->si_int;
3312 		break;
3313 	case SIL_POLL:
3314 		to->si_band = from->si_band;
3315 		to->si_fd   = from->si_fd;
3316 		break;
3317 	case SIL_FAULT:
3318 		to->si_addr = ptr_to_compat(from->si_addr);
3319 #ifdef __ARCH_SI_TRAPNO
3320 		to->si_trapno = from->si_trapno;
3321 #endif
3322 		break;
3323 	case SIL_FAULT_MCEERR:
3324 		to->si_addr = ptr_to_compat(from->si_addr);
3325 #ifdef __ARCH_SI_TRAPNO
3326 		to->si_trapno = from->si_trapno;
3327 #endif
3328 		to->si_addr_lsb = from->si_addr_lsb;
3329 		break;
3330 	case SIL_FAULT_BNDERR:
3331 		to->si_addr = ptr_to_compat(from->si_addr);
3332 #ifdef __ARCH_SI_TRAPNO
3333 		to->si_trapno = from->si_trapno;
3334 #endif
3335 		to->si_lower = ptr_to_compat(from->si_lower);
3336 		to->si_upper = ptr_to_compat(from->si_upper);
3337 		break;
3338 	case SIL_FAULT_PKUERR:
3339 		to->si_addr = ptr_to_compat(from->si_addr);
3340 #ifdef __ARCH_SI_TRAPNO
3341 		to->si_trapno = from->si_trapno;
3342 #endif
3343 		to->si_pkey = from->si_pkey;
3344 		break;
3345 	case SIL_CHLD:
3346 		to->si_pid = from->si_pid;
3347 		to->si_uid = from->si_uid;
3348 		to->si_status = from->si_status;
3349 		to->si_utime = from->si_utime;
3350 		to->si_stime = from->si_stime;
3351 		break;
3352 	case SIL_RT:
3353 		to->si_pid = from->si_pid;
3354 		to->si_uid = from->si_uid;
3355 		to->si_int = from->si_int;
3356 		break;
3357 	case SIL_SYS:
3358 		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3359 		to->si_syscall   = from->si_syscall;
3360 		to->si_arch      = from->si_arch;
3361 		break;
3362 	}
3363 }
3364 
__copy_siginfo_to_user32(struct compat_siginfo __user * to,const struct kernel_siginfo * from)3365 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3366 			   const struct kernel_siginfo *from)
3367 {
3368 	struct compat_siginfo new;
3369 
3370 	copy_siginfo_to_external32(&new, from);
3371 	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3372 		return -EFAULT;
3373 	return 0;
3374 }
3375 
post_copy_siginfo_from_user32(kernel_siginfo_t * to,const struct compat_siginfo * from)3376 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3377 					 const struct compat_siginfo *from)
3378 {
3379 	clear_siginfo(to);
3380 	to->si_signo = from->si_signo;
3381 	to->si_errno = from->si_errno;
3382 	to->si_code  = from->si_code;
3383 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3384 	case SIL_KILL:
3385 		to->si_pid = from->si_pid;
3386 		to->si_uid = from->si_uid;
3387 		break;
3388 	case SIL_TIMER:
3389 		to->si_tid     = from->si_tid;
3390 		to->si_overrun = from->si_overrun;
3391 		to->si_int     = from->si_int;
3392 		break;
3393 	case SIL_POLL:
3394 		to->si_band = from->si_band;
3395 		to->si_fd   = from->si_fd;
3396 		break;
3397 	case SIL_FAULT:
3398 		to->si_addr = compat_ptr(from->si_addr);
3399 #ifdef __ARCH_SI_TRAPNO
3400 		to->si_trapno = from->si_trapno;
3401 #endif
3402 		break;
3403 	case SIL_FAULT_MCEERR:
3404 		to->si_addr = compat_ptr(from->si_addr);
3405 #ifdef __ARCH_SI_TRAPNO
3406 		to->si_trapno = from->si_trapno;
3407 #endif
3408 		to->si_addr_lsb = from->si_addr_lsb;
3409 		break;
3410 	case SIL_FAULT_BNDERR:
3411 		to->si_addr = compat_ptr(from->si_addr);
3412 #ifdef __ARCH_SI_TRAPNO
3413 		to->si_trapno = from->si_trapno;
3414 #endif
3415 		to->si_lower = compat_ptr(from->si_lower);
3416 		to->si_upper = compat_ptr(from->si_upper);
3417 		break;
3418 	case SIL_FAULT_PKUERR:
3419 		to->si_addr = compat_ptr(from->si_addr);
3420 #ifdef __ARCH_SI_TRAPNO
3421 		to->si_trapno = from->si_trapno;
3422 #endif
3423 		to->si_pkey = from->si_pkey;
3424 		break;
3425 	case SIL_CHLD:
3426 		to->si_pid    = from->si_pid;
3427 		to->si_uid    = from->si_uid;
3428 		to->si_status = from->si_status;
3429 #ifdef CONFIG_X86_X32_ABI
3430 		if (in_x32_syscall()) {
3431 			to->si_utime = from->_sifields._sigchld_x32._utime;
3432 			to->si_stime = from->_sifields._sigchld_x32._stime;
3433 		} else
3434 #endif
3435 		{
3436 			to->si_utime = from->si_utime;
3437 			to->si_stime = from->si_stime;
3438 		}
3439 		break;
3440 	case SIL_RT:
3441 		to->si_pid = from->si_pid;
3442 		to->si_uid = from->si_uid;
3443 		to->si_int = from->si_int;
3444 		break;
3445 	case SIL_SYS:
3446 		to->si_call_addr = compat_ptr(from->si_call_addr);
3447 		to->si_syscall   = from->si_syscall;
3448 		to->si_arch      = from->si_arch;
3449 		break;
3450 	}
3451 	return 0;
3452 }
3453 
__copy_siginfo_from_user32(int signo,struct kernel_siginfo * to,const struct compat_siginfo __user * ufrom)3454 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3455 				      const struct compat_siginfo __user *ufrom)
3456 {
3457 	struct compat_siginfo from;
3458 
3459 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3460 		return -EFAULT;
3461 
3462 	from.si_signo = signo;
3463 	return post_copy_siginfo_from_user32(to, &from);
3464 }
3465 
copy_siginfo_from_user32(struct kernel_siginfo * to,const struct compat_siginfo __user * ufrom)3466 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3467 			     const struct compat_siginfo __user *ufrom)
3468 {
3469 	struct compat_siginfo from;
3470 
3471 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3472 		return -EFAULT;
3473 
3474 	return post_copy_siginfo_from_user32(to, &from);
3475 }
3476 #endif /* CONFIG_COMPAT */
3477 
3478 /**
3479  *  do_sigtimedwait - wait for queued signals specified in @which
3480  *  @which: queued signals to wait for
3481  *  @info: if non-null, the signal's siginfo is returned here
3482  *  @ts: upper bound on process time suspension
3483  */
do_sigtimedwait(const sigset_t * which,kernel_siginfo_t * info,const struct timespec64 * ts)3484 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3485 		    const struct timespec64 *ts)
3486 {
3487 	ktime_t *to = NULL, timeout = KTIME_MAX;
3488 	struct task_struct *tsk = current;
3489 	sigset_t mask = *which;
3490 	int sig, ret = 0;
3491 
3492 	if (ts) {
3493 		if (!timespec64_valid(ts))
3494 			return -EINVAL;
3495 		timeout = timespec64_to_ktime(*ts);
3496 		to = &timeout;
3497 	}
3498 
3499 	/*
3500 	 * Invert the set of allowed signals to get those we want to block.
3501 	 */
3502 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3503 	signotset(&mask);
3504 
3505 	spin_lock_irq(&tsk->sighand->siglock);
3506 	sig = dequeue_signal(tsk, &mask, info);
3507 	if (!sig && timeout) {
3508 		/*
3509 		 * None ready, temporarily unblock those we're interested
3510 		 * while we are sleeping in so that we'll be awakened when
3511 		 * they arrive. Unblocking is always fine, we can avoid
3512 		 * set_current_blocked().
3513 		 */
3514 		tsk->real_blocked = tsk->blocked;
3515 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3516 		recalc_sigpending();
3517 		spin_unlock_irq(&tsk->sighand->siglock);
3518 
3519 		__set_current_state(TASK_INTERRUPTIBLE);
3520 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3521 							 HRTIMER_MODE_REL);
3522 		spin_lock_irq(&tsk->sighand->siglock);
3523 		__set_task_blocked(tsk, &tsk->real_blocked);
3524 		sigemptyset(&tsk->real_blocked);
3525 		sig = dequeue_signal(tsk, &mask, info);
3526 	}
3527 	spin_unlock_irq(&tsk->sighand->siglock);
3528 
3529 	if (sig)
3530 		return sig;
3531 	return ret ? -EINTR : -EAGAIN;
3532 }
3533 
3534 /**
3535  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3536  *			in @uthese
3537  *  @uthese: queued signals to wait for
3538  *  @uinfo: if non-null, the signal's siginfo is returned here
3539  *  @uts: upper bound on process time suspension
3540  *  @sigsetsize: size of sigset_t type
3541  */
SYSCALL_DEFINE4(rt_sigtimedwait,const sigset_t __user *,uthese,siginfo_t __user *,uinfo,const struct __kernel_timespec __user *,uts,size_t,sigsetsize)3542 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3543 		siginfo_t __user *, uinfo,
3544 		const struct __kernel_timespec __user *, uts,
3545 		size_t, sigsetsize)
3546 {
3547 	sigset_t these;
3548 	struct timespec64 ts;
3549 	kernel_siginfo_t info;
3550 	int ret;
3551 
3552 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3553 	if (sigsetsize != sizeof(sigset_t))
3554 		return -EINVAL;
3555 
3556 	if (copy_from_user(&these, uthese, sizeof(these)))
3557 		return -EFAULT;
3558 
3559 	if (uts) {
3560 		if (get_timespec64(&ts, uts))
3561 			return -EFAULT;
3562 	}
3563 
3564 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3565 
3566 	if (ret > 0 && uinfo) {
3567 		if (copy_siginfo_to_user(uinfo, &info))
3568 			ret = -EFAULT;
3569 	}
3570 
3571 	return ret;
3572 }
3573 
3574 #ifdef CONFIG_COMPAT_32BIT_TIME
SYSCALL_DEFINE4(rt_sigtimedwait_time32,const sigset_t __user *,uthese,siginfo_t __user *,uinfo,const struct old_timespec32 __user *,uts,size_t,sigsetsize)3575 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3576 		siginfo_t __user *, uinfo,
3577 		const struct old_timespec32 __user *, uts,
3578 		size_t, sigsetsize)
3579 {
3580 	sigset_t these;
3581 	struct timespec64 ts;
3582 	kernel_siginfo_t info;
3583 	int ret;
3584 
3585 	if (sigsetsize != sizeof(sigset_t))
3586 		return -EINVAL;
3587 
3588 	if (copy_from_user(&these, uthese, sizeof(these)))
3589 		return -EFAULT;
3590 
3591 	if (uts) {
3592 		if (get_old_timespec32(&ts, uts))
3593 			return -EFAULT;
3594 	}
3595 
3596 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3597 
3598 	if (ret > 0 && uinfo) {
3599 		if (copy_siginfo_to_user(uinfo, &info))
3600 			ret = -EFAULT;
3601 	}
3602 
3603 	return ret;
3604 }
3605 #endif
3606 
3607 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64,compat_sigset_t __user *,uthese,struct compat_siginfo __user *,uinfo,struct __kernel_timespec __user *,uts,compat_size_t,sigsetsize)3608 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3609 		struct compat_siginfo __user *, uinfo,
3610 		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3611 {
3612 	sigset_t s;
3613 	struct timespec64 t;
3614 	kernel_siginfo_t info;
3615 	long ret;
3616 
3617 	if (sigsetsize != sizeof(sigset_t))
3618 		return -EINVAL;
3619 
3620 	if (get_compat_sigset(&s, uthese))
3621 		return -EFAULT;
3622 
3623 	if (uts) {
3624 		if (get_timespec64(&t, uts))
3625 			return -EFAULT;
3626 	}
3627 
3628 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3629 
3630 	if (ret > 0 && uinfo) {
3631 		if (copy_siginfo_to_user32(uinfo, &info))
3632 			ret = -EFAULT;
3633 	}
3634 
3635 	return ret;
3636 }
3637 
3638 #ifdef CONFIG_COMPAT_32BIT_TIME
COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32,compat_sigset_t __user *,uthese,struct compat_siginfo __user *,uinfo,struct old_timespec32 __user *,uts,compat_size_t,sigsetsize)3639 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3640 		struct compat_siginfo __user *, uinfo,
3641 		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3642 {
3643 	sigset_t s;
3644 	struct timespec64 t;
3645 	kernel_siginfo_t info;
3646 	long ret;
3647 
3648 	if (sigsetsize != sizeof(sigset_t))
3649 		return -EINVAL;
3650 
3651 	if (get_compat_sigset(&s, uthese))
3652 		return -EFAULT;
3653 
3654 	if (uts) {
3655 		if (get_old_timespec32(&t, uts))
3656 			return -EFAULT;
3657 	}
3658 
3659 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3660 
3661 	if (ret > 0 && uinfo) {
3662 		if (copy_siginfo_to_user32(uinfo, &info))
3663 			ret = -EFAULT;
3664 	}
3665 
3666 	return ret;
3667 }
3668 #endif
3669 #endif
3670 
prepare_kill_siginfo(int sig,struct kernel_siginfo * info)3671 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3672 {
3673 	clear_siginfo(info);
3674 	info->si_signo = sig;
3675 	info->si_errno = 0;
3676 	info->si_code = SI_USER;
3677 	info->si_pid = task_tgid_vnr(current);
3678 	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3679 }
3680 
3681 /**
3682  *  sys_kill - send a signal to a process
3683  *  @pid: the PID of the process
3684  *  @sig: signal to be sent
3685  */
SYSCALL_DEFINE2(kill,pid_t,pid,int,sig)3686 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3687 {
3688 	struct kernel_siginfo info;
3689 
3690 	prepare_kill_siginfo(sig, &info);
3691 
3692 	return kill_something_info(sig, &info, pid);
3693 }
3694 
3695 /*
3696  * Verify that the signaler and signalee either are in the same pid namespace
3697  * or that the signaler's pid namespace is an ancestor of the signalee's pid
3698  * namespace.
3699  */
access_pidfd_pidns(struct pid * pid)3700 static bool access_pidfd_pidns(struct pid *pid)
3701 {
3702 	struct pid_namespace *active = task_active_pid_ns(current);
3703 	struct pid_namespace *p = ns_of_pid(pid);
3704 
3705 	for (;;) {
3706 		if (!p)
3707 			return false;
3708 		if (p == active)
3709 			break;
3710 		p = p->parent;
3711 	}
3712 
3713 	return true;
3714 }
3715 
copy_siginfo_from_user_any(kernel_siginfo_t * kinfo,siginfo_t * info)3716 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo, siginfo_t *info)
3717 {
3718 #ifdef CONFIG_COMPAT
3719 	/*
3720 	 * Avoid hooking up compat syscalls and instead handle necessary
3721 	 * conversions here. Note, this is a stop-gap measure and should not be
3722 	 * considered a generic solution.
3723 	 */
3724 	if (in_compat_syscall())
3725 		return copy_siginfo_from_user32(
3726 			kinfo, (struct compat_siginfo __user *)info);
3727 #endif
3728 	return copy_siginfo_from_user(kinfo, info);
3729 }
3730 
pidfd_to_pid(const struct file * file)3731 static struct pid *pidfd_to_pid(const struct file *file)
3732 {
3733 	struct pid *pid;
3734 
3735 	pid = pidfd_pid(file);
3736 	if (!IS_ERR(pid))
3737 		return pid;
3738 
3739 	return tgid_pidfd_to_pid(file);
3740 }
3741 
3742 /**
3743  * sys_pidfd_send_signal - Signal a process through a pidfd
3744  * @pidfd:  file descriptor of the process
3745  * @sig:    signal to send
3746  * @info:   signal info
3747  * @flags:  future flags
3748  *
3749  * The syscall currently only signals via PIDTYPE_PID which covers
3750  * kill(<positive-pid>, <signal>. It does not signal threads or process
3751  * groups.
3752  * In order to extend the syscall to threads and process groups the @flags
3753  * argument should be used. In essence, the @flags argument will determine
3754  * what is signaled and not the file descriptor itself. Put in other words,
3755  * grouping is a property of the flags argument not a property of the file
3756  * descriptor.
3757  *
3758  * Return: 0 on success, negative errno on failure
3759  */
SYSCALL_DEFINE4(pidfd_send_signal,int,pidfd,int,sig,siginfo_t __user *,info,unsigned int,flags)3760 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3761 		siginfo_t __user *, info, unsigned int, flags)
3762 {
3763 	int ret;
3764 	struct fd f;
3765 	struct pid *pid;
3766 	kernel_siginfo_t kinfo;
3767 
3768 	/* Enforce flags be set to 0 until we add an extension. */
3769 	if (flags)
3770 		return -EINVAL;
3771 
3772 	f = fdget(pidfd);
3773 	if (!f.file)
3774 		return -EBADF;
3775 
3776 	/* Is this a pidfd? */
3777 	pid = pidfd_to_pid(f.file);
3778 	if (IS_ERR(pid)) {
3779 		ret = PTR_ERR(pid);
3780 		goto err;
3781 	}
3782 
3783 	ret = -EINVAL;
3784 	if (!access_pidfd_pidns(pid))
3785 		goto err;
3786 
3787 	if (info) {
3788 		ret = copy_siginfo_from_user_any(&kinfo, info);
3789 		if (unlikely(ret))
3790 			goto err;
3791 
3792 		ret = -EINVAL;
3793 		if (unlikely(sig != kinfo.si_signo))
3794 			goto err;
3795 
3796 		/* Only allow sending arbitrary signals to yourself. */
3797 		ret = -EPERM;
3798 		if ((task_pid(current) != pid) &&
3799 		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3800 			goto err;
3801 	} else {
3802 		prepare_kill_siginfo(sig, &kinfo);
3803 	}
3804 
3805 	ret = kill_pid_info(sig, &kinfo, pid);
3806 
3807 err:
3808 	fdput(f);
3809 	return ret;
3810 }
3811 
3812 static int
do_send_specific(pid_t tgid,pid_t pid,int sig,struct kernel_siginfo * info)3813 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3814 {
3815 	struct task_struct *p;
3816 	int error = -ESRCH;
3817 
3818 	rcu_read_lock();
3819 	p = find_task_by_vpid(pid);
3820 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3821 		error = check_kill_permission(sig, info, p);
3822 		/*
3823 		 * The null signal is a permissions and process existence
3824 		 * probe.  No signal is actually delivered.
3825 		 */
3826 		if (!error && sig) {
3827 			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3828 			/*
3829 			 * If lock_task_sighand() failed we pretend the task
3830 			 * dies after receiving the signal. The window is tiny,
3831 			 * and the signal is private anyway.
3832 			 */
3833 			if (unlikely(error == -ESRCH))
3834 				error = 0;
3835 		}
3836 	}
3837 	rcu_read_unlock();
3838 
3839 	return error;
3840 }
3841 
do_tkill(pid_t tgid,pid_t pid,int sig)3842 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3843 {
3844 	struct kernel_siginfo info;
3845 
3846 	clear_siginfo(&info);
3847 	info.si_signo = sig;
3848 	info.si_errno = 0;
3849 	info.si_code = SI_TKILL;
3850 	info.si_pid = task_tgid_vnr(current);
3851 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3852 
3853 	return do_send_specific(tgid, pid, sig, &info);
3854 }
3855 
3856 /**
3857  *  sys_tgkill - send signal to one specific thread
3858  *  @tgid: the thread group ID of the thread
3859  *  @pid: the PID of the thread
3860  *  @sig: signal to be sent
3861  *
3862  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3863  *  exists but it's not belonging to the target process anymore. This
3864  *  method solves the problem of threads exiting and PIDs getting reused.
3865  */
SYSCALL_DEFINE3(tgkill,pid_t,tgid,pid_t,pid,int,sig)3866 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3867 {
3868 	/* This is only valid for single tasks */
3869 	if (pid <= 0 || tgid <= 0)
3870 		return -EINVAL;
3871 
3872 	return do_tkill(tgid, pid, sig);
3873 }
3874 
3875 /**
3876  *  sys_tkill - send signal to one specific task
3877  *  @pid: the PID of the task
3878  *  @sig: signal to be sent
3879  *
3880  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3881  */
SYSCALL_DEFINE2(tkill,pid_t,pid,int,sig)3882 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3883 {
3884 	/* This is only valid for single tasks */
3885 	if (pid <= 0)
3886 		return -EINVAL;
3887 
3888 	return do_tkill(0, pid, sig);
3889 }
3890 
do_rt_sigqueueinfo(pid_t pid,int sig,kernel_siginfo_t * info)3891 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
3892 {
3893 	/* Not even root can pretend to send signals from the kernel.
3894 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3895 	 */
3896 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3897 	    (task_pid_vnr(current) != pid))
3898 		return -EPERM;
3899 
3900 	/* POSIX.1b doesn't mention process groups.  */
3901 	return kill_proc_info(sig, info, pid);
3902 }
3903 
3904 /**
3905  *  sys_rt_sigqueueinfo - send signal information to a signal
3906  *  @pid: the PID of the thread
3907  *  @sig: signal to be sent
3908  *  @uinfo: signal info to be sent
3909  */
SYSCALL_DEFINE3(rt_sigqueueinfo,pid_t,pid,int,sig,siginfo_t __user *,uinfo)3910 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
3911 		siginfo_t __user *, uinfo)
3912 {
3913 	kernel_siginfo_t info;
3914 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
3915 	if (unlikely(ret))
3916 		return ret;
3917 	return do_rt_sigqueueinfo(pid, sig, &info);
3918 }
3919 
3920 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,compat_pid_t,pid,int,sig,struct compat_siginfo __user *,uinfo)3921 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
3922 			compat_pid_t, pid,
3923 			int, sig,
3924 			struct compat_siginfo __user *, uinfo)
3925 {
3926 	kernel_siginfo_t info;
3927 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
3928 	if (unlikely(ret))
3929 		return ret;
3930 	return do_rt_sigqueueinfo(pid, sig, &info);
3931 }
3932 #endif
3933 
do_rt_tgsigqueueinfo(pid_t tgid,pid_t pid,int sig,kernel_siginfo_t * info)3934 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
3935 {
3936 	/* This is only valid for single tasks */
3937 	if (pid <= 0 || tgid <= 0)
3938 		return -EINVAL;
3939 
3940 	/* Not even root can pretend to send signals from the kernel.
3941 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3942 	 */
3943 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3944 	    (task_pid_vnr(current) != pid))
3945 		return -EPERM;
3946 
3947 	return do_send_specific(tgid, pid, sig, info);
3948 }
3949 
SYSCALL_DEFINE4(rt_tgsigqueueinfo,pid_t,tgid,pid_t,pid,int,sig,siginfo_t __user *,uinfo)3950 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
3951 		siginfo_t __user *, uinfo)
3952 {
3953 	kernel_siginfo_t info;
3954 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
3955 	if (unlikely(ret))
3956 		return ret;
3957 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3958 }
3959 
3960 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,compat_pid_t,tgid,compat_pid_t,pid,int,sig,struct compat_siginfo __user *,uinfo)3961 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
3962 			compat_pid_t, tgid,
3963 			compat_pid_t, pid,
3964 			int, sig,
3965 			struct compat_siginfo __user *, uinfo)
3966 {
3967 	kernel_siginfo_t info;
3968 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
3969 	if (unlikely(ret))
3970 		return ret;
3971 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3972 }
3973 #endif
3974 
3975 /*
3976  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
3977  */
kernel_sigaction(int sig,__sighandler_t action)3978 void kernel_sigaction(int sig, __sighandler_t action)
3979 {
3980 	spin_lock_irq(&current->sighand->siglock);
3981 	current->sighand->action[sig - 1].sa.sa_handler = action;
3982 	if (action == SIG_IGN) {
3983 		sigset_t mask;
3984 
3985 		sigemptyset(&mask);
3986 		sigaddset(&mask, sig);
3987 
3988 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
3989 		flush_sigqueue_mask(&mask, &current->pending);
3990 		recalc_sigpending();
3991 	}
3992 	spin_unlock_irq(&current->sighand->siglock);
3993 }
3994 EXPORT_SYMBOL(kernel_sigaction);
3995 
sigaction_compat_abi(struct k_sigaction * act,struct k_sigaction * oact)3996 void __weak sigaction_compat_abi(struct k_sigaction *act,
3997 		struct k_sigaction *oact)
3998 {
3999 }
4000 
do_sigaction(int sig,struct k_sigaction * act,struct k_sigaction * oact)4001 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4002 {
4003 	struct task_struct *p = current, *t;
4004 	struct k_sigaction *k;
4005 	sigset_t mask;
4006 
4007 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4008 		return -EINVAL;
4009 
4010 	k = &p->sighand->action[sig-1];
4011 
4012 	spin_lock_irq(&p->sighand->siglock);
4013 	if (oact)
4014 		*oact = *k;
4015 
4016 	/*
4017 	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4018 	 * e.g. by having an architecture use the bit in their uapi.
4019 	 */
4020 	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4021 
4022 	/*
4023 	 * Clear unknown flag bits in order to allow userspace to detect missing
4024 	 * support for flag bits and to allow the kernel to use non-uapi bits
4025 	 * internally.
4026 	 */
4027 	if (act)
4028 		act->sa.sa_flags &= UAPI_SA_FLAGS;
4029 	if (oact)
4030 		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4031 
4032 	sigaction_compat_abi(act, oact);
4033 
4034 	if (act) {
4035 		sigdelsetmask(&act->sa.sa_mask,
4036 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4037 		*k = *act;
4038 		/*
4039 		 * POSIX 3.3.1.3:
4040 		 *  "Setting a signal action to SIG_IGN for a signal that is
4041 		 *   pending shall cause the pending signal to be discarded,
4042 		 *   whether or not it is blocked."
4043 		 *
4044 		 *  "Setting a signal action to SIG_DFL for a signal that is
4045 		 *   pending and whose default action is to ignore the signal
4046 		 *   (for example, SIGCHLD), shall cause the pending signal to
4047 		 *   be discarded, whether or not it is blocked"
4048 		 */
4049 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4050 			sigemptyset(&mask);
4051 			sigaddset(&mask, sig);
4052 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4053 			for_each_thread(p, t)
4054 				flush_sigqueue_mask(&mask, &t->pending);
4055 		}
4056 	}
4057 
4058 	spin_unlock_irq(&p->sighand->siglock);
4059 	return 0;
4060 }
4061 
4062 static int
do_sigaltstack(const stack_t * ss,stack_t * oss,unsigned long sp,size_t min_ss_size)4063 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4064 		size_t min_ss_size)
4065 {
4066 	struct task_struct *t = current;
4067 
4068 	if (oss) {
4069 		memset(oss, 0, sizeof(stack_t));
4070 		oss->ss_sp = (void __user *) t->sas_ss_sp;
4071 		oss->ss_size = t->sas_ss_size;
4072 		oss->ss_flags = sas_ss_flags(sp) |
4073 			(current->sas_ss_flags & SS_FLAG_BITS);
4074 	}
4075 
4076 	if (ss) {
4077 		void __user *ss_sp = ss->ss_sp;
4078 		size_t ss_size = ss->ss_size;
4079 		unsigned ss_flags = ss->ss_flags;
4080 		int ss_mode;
4081 
4082 		if (unlikely(on_sig_stack(sp)))
4083 			return -EPERM;
4084 
4085 		ss_mode = ss_flags & ~SS_FLAG_BITS;
4086 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4087 				ss_mode != 0))
4088 			return -EINVAL;
4089 
4090 		if (ss_mode == SS_DISABLE) {
4091 			ss_size = 0;
4092 			ss_sp = NULL;
4093 		} else {
4094 			if (unlikely(ss_size < min_ss_size))
4095 				return -ENOMEM;
4096 		}
4097 
4098 		t->sas_ss_sp = (unsigned long) ss_sp;
4099 		t->sas_ss_size = ss_size;
4100 		t->sas_ss_flags = ss_flags;
4101 	}
4102 	return 0;
4103 }
4104 
SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss,stack_t __user *,uoss)4105 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4106 {
4107 	stack_t new, old;
4108 	int err;
4109 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4110 		return -EFAULT;
4111 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4112 			      current_user_stack_pointer(),
4113 			      MINSIGSTKSZ);
4114 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4115 		err = -EFAULT;
4116 	return err;
4117 }
4118 
restore_altstack(const stack_t __user * uss)4119 int restore_altstack(const stack_t __user *uss)
4120 {
4121 	stack_t new;
4122 	if (copy_from_user(&new, uss, sizeof(stack_t)))
4123 		return -EFAULT;
4124 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4125 			     MINSIGSTKSZ);
4126 	/* squash all but EFAULT for now */
4127 	return 0;
4128 }
4129 
__save_altstack(stack_t __user * uss,unsigned long sp)4130 int __save_altstack(stack_t __user *uss, unsigned long sp)
4131 {
4132 	struct task_struct *t = current;
4133 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4134 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4135 		__put_user(t->sas_ss_size, &uss->ss_size);
4136 	if (err)
4137 		return err;
4138 	if (t->sas_ss_flags & SS_AUTODISARM)
4139 		sas_ss_reset(t);
4140 	return 0;
4141 }
4142 
4143 #ifdef CONFIG_COMPAT
do_compat_sigaltstack(const compat_stack_t __user * uss_ptr,compat_stack_t __user * uoss_ptr)4144 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4145 				 compat_stack_t __user *uoss_ptr)
4146 {
4147 	stack_t uss, uoss;
4148 	int ret;
4149 
4150 	if (uss_ptr) {
4151 		compat_stack_t uss32;
4152 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4153 			return -EFAULT;
4154 		uss.ss_sp = compat_ptr(uss32.ss_sp);
4155 		uss.ss_flags = uss32.ss_flags;
4156 		uss.ss_size = uss32.ss_size;
4157 	}
4158 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4159 			     compat_user_stack_pointer(),
4160 			     COMPAT_MINSIGSTKSZ);
4161 	if (ret >= 0 && uoss_ptr)  {
4162 		compat_stack_t old;
4163 		memset(&old, 0, sizeof(old));
4164 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4165 		old.ss_flags = uoss.ss_flags;
4166 		old.ss_size = uoss.ss_size;
4167 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4168 			ret = -EFAULT;
4169 	}
4170 	return ret;
4171 }
4172 
COMPAT_SYSCALL_DEFINE2(sigaltstack,const compat_stack_t __user *,uss_ptr,compat_stack_t __user *,uoss_ptr)4173 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4174 			const compat_stack_t __user *, uss_ptr,
4175 			compat_stack_t __user *, uoss_ptr)
4176 {
4177 	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4178 }
4179 
compat_restore_altstack(const compat_stack_t __user * uss)4180 int compat_restore_altstack(const compat_stack_t __user *uss)
4181 {
4182 	int err = do_compat_sigaltstack(uss, NULL);
4183 	/* squash all but -EFAULT for now */
4184 	return err == -EFAULT ? err : 0;
4185 }
4186 
__compat_save_altstack(compat_stack_t __user * uss,unsigned long sp)4187 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4188 {
4189 	int err;
4190 	struct task_struct *t = current;
4191 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4192 			 &uss->ss_sp) |
4193 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4194 		__put_user(t->sas_ss_size, &uss->ss_size);
4195 	if (err)
4196 		return err;
4197 	if (t->sas_ss_flags & SS_AUTODISARM)
4198 		sas_ss_reset(t);
4199 	return 0;
4200 }
4201 #endif
4202 
4203 #ifdef __ARCH_WANT_SYS_SIGPENDING
4204 
4205 /**
4206  *  sys_sigpending - examine pending signals
4207  *  @uset: where mask of pending signal is returned
4208  */
SYSCALL_DEFINE1(sigpending,old_sigset_t __user *,uset)4209 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4210 {
4211 	sigset_t set;
4212 
4213 	if (sizeof(old_sigset_t) > sizeof(*uset))
4214 		return -EINVAL;
4215 
4216 	do_sigpending(&set);
4217 
4218 	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4219 		return -EFAULT;
4220 
4221 	return 0;
4222 }
4223 
4224 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE1(sigpending,compat_old_sigset_t __user *,set32)4225 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4226 {
4227 	sigset_t set;
4228 
4229 	do_sigpending(&set);
4230 
4231 	return put_user(set.sig[0], set32);
4232 }
4233 #endif
4234 
4235 #endif
4236 
4237 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4238 /**
4239  *  sys_sigprocmask - examine and change blocked signals
4240  *  @how: whether to add, remove, or set signals
4241  *  @nset: signals to add or remove (if non-null)
4242  *  @oset: previous value of signal mask if non-null
4243  *
4244  * Some platforms have their own version with special arguments;
4245  * others support only sys_rt_sigprocmask.
4246  */
4247 
SYSCALL_DEFINE3(sigprocmask,int,how,old_sigset_t __user *,nset,old_sigset_t __user *,oset)4248 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4249 		old_sigset_t __user *, oset)
4250 {
4251 	old_sigset_t old_set, new_set;
4252 	sigset_t new_blocked;
4253 
4254 	old_set = current->blocked.sig[0];
4255 
4256 	if (nset) {
4257 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4258 			return -EFAULT;
4259 
4260 		new_blocked = current->blocked;
4261 
4262 		switch (how) {
4263 		case SIG_BLOCK:
4264 			sigaddsetmask(&new_blocked, new_set);
4265 			break;
4266 		case SIG_UNBLOCK:
4267 			sigdelsetmask(&new_blocked, new_set);
4268 			break;
4269 		case SIG_SETMASK:
4270 			new_blocked.sig[0] = new_set;
4271 			break;
4272 		default:
4273 			return -EINVAL;
4274 		}
4275 
4276 		set_current_blocked(&new_blocked);
4277 	}
4278 
4279 	if (oset) {
4280 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4281 			return -EFAULT;
4282 	}
4283 
4284 	return 0;
4285 }
4286 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4287 
4288 #ifndef CONFIG_ODD_RT_SIGACTION
4289 /**
4290  *  sys_rt_sigaction - alter an action taken by a process
4291  *  @sig: signal to be sent
4292  *  @act: new sigaction
4293  *  @oact: used to save the previous sigaction
4294  *  @sigsetsize: size of sigset_t type
4295  */
SYSCALL_DEFINE4(rt_sigaction,int,sig,const struct sigaction __user *,act,struct sigaction __user *,oact,size_t,sigsetsize)4296 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4297 		const struct sigaction __user *, act,
4298 		struct sigaction __user *, oact,
4299 		size_t, sigsetsize)
4300 {
4301 	struct k_sigaction new_sa, old_sa;
4302 	int ret;
4303 
4304 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4305 	if (sigsetsize != sizeof(sigset_t))
4306 		return -EINVAL;
4307 
4308 	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4309 		return -EFAULT;
4310 
4311 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4312 	if (ret)
4313 		return ret;
4314 
4315 	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4316 		return -EFAULT;
4317 
4318 	return 0;
4319 }
4320 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_sigaction,int,sig,const struct compat_sigaction __user *,act,struct compat_sigaction __user *,oact,compat_size_t,sigsetsize)4321 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4322 		const struct compat_sigaction __user *, act,
4323 		struct compat_sigaction __user *, oact,
4324 		compat_size_t, sigsetsize)
4325 {
4326 	struct k_sigaction new_ka, old_ka;
4327 #ifdef __ARCH_HAS_SA_RESTORER
4328 	compat_uptr_t restorer;
4329 #endif
4330 	int ret;
4331 
4332 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4333 	if (sigsetsize != sizeof(compat_sigset_t))
4334 		return -EINVAL;
4335 
4336 	if (act) {
4337 		compat_uptr_t handler;
4338 		ret = get_user(handler, &act->sa_handler);
4339 		new_ka.sa.sa_handler = compat_ptr(handler);
4340 #ifdef __ARCH_HAS_SA_RESTORER
4341 		ret |= get_user(restorer, &act->sa_restorer);
4342 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4343 #endif
4344 		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4345 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4346 		if (ret)
4347 			return -EFAULT;
4348 	}
4349 
4350 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4351 	if (!ret && oact) {
4352 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4353 			       &oact->sa_handler);
4354 		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4355 					 sizeof(oact->sa_mask));
4356 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4357 #ifdef __ARCH_HAS_SA_RESTORER
4358 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4359 				&oact->sa_restorer);
4360 #endif
4361 	}
4362 	return ret;
4363 }
4364 #endif
4365 #endif /* !CONFIG_ODD_RT_SIGACTION */
4366 
4367 #ifdef CONFIG_OLD_SIGACTION
SYSCALL_DEFINE3(sigaction,int,sig,const struct old_sigaction __user *,act,struct old_sigaction __user *,oact)4368 SYSCALL_DEFINE3(sigaction, int, sig,
4369 		const struct old_sigaction __user *, act,
4370 	        struct old_sigaction __user *, oact)
4371 {
4372 	struct k_sigaction new_ka, old_ka;
4373 	int ret;
4374 
4375 	if (act) {
4376 		old_sigset_t mask;
4377 		if (!access_ok(act, sizeof(*act)) ||
4378 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4379 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4380 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4381 		    __get_user(mask, &act->sa_mask))
4382 			return -EFAULT;
4383 #ifdef __ARCH_HAS_KA_RESTORER
4384 		new_ka.ka_restorer = NULL;
4385 #endif
4386 		siginitset(&new_ka.sa.sa_mask, mask);
4387 	}
4388 
4389 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4390 
4391 	if (!ret && oact) {
4392 		if (!access_ok(oact, sizeof(*oact)) ||
4393 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4394 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4395 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4396 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4397 			return -EFAULT;
4398 	}
4399 
4400 	return ret;
4401 }
4402 #endif
4403 #ifdef CONFIG_COMPAT_OLD_SIGACTION
COMPAT_SYSCALL_DEFINE3(sigaction,int,sig,const struct compat_old_sigaction __user *,act,struct compat_old_sigaction __user *,oact)4404 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4405 		const struct compat_old_sigaction __user *, act,
4406 	        struct compat_old_sigaction __user *, oact)
4407 {
4408 	struct k_sigaction new_ka, old_ka;
4409 	int ret;
4410 	compat_old_sigset_t mask;
4411 	compat_uptr_t handler, restorer;
4412 
4413 	if (act) {
4414 		if (!access_ok(act, sizeof(*act)) ||
4415 		    __get_user(handler, &act->sa_handler) ||
4416 		    __get_user(restorer, &act->sa_restorer) ||
4417 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4418 		    __get_user(mask, &act->sa_mask))
4419 			return -EFAULT;
4420 
4421 #ifdef __ARCH_HAS_KA_RESTORER
4422 		new_ka.ka_restorer = NULL;
4423 #endif
4424 		new_ka.sa.sa_handler = compat_ptr(handler);
4425 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4426 		siginitset(&new_ka.sa.sa_mask, mask);
4427 	}
4428 
4429 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4430 
4431 	if (!ret && oact) {
4432 		if (!access_ok(oact, sizeof(*oact)) ||
4433 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4434 			       &oact->sa_handler) ||
4435 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4436 			       &oact->sa_restorer) ||
4437 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4438 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4439 			return -EFAULT;
4440 	}
4441 	return ret;
4442 }
4443 #endif
4444 
4445 #ifdef CONFIG_SGETMASK_SYSCALL
4446 
4447 /*
4448  * For backwards compatibility.  Functionality superseded by sigprocmask.
4449  */
SYSCALL_DEFINE0(sgetmask)4450 SYSCALL_DEFINE0(sgetmask)
4451 {
4452 	/* SMP safe */
4453 	return current->blocked.sig[0];
4454 }
4455 
SYSCALL_DEFINE1(ssetmask,int,newmask)4456 SYSCALL_DEFINE1(ssetmask, int, newmask)
4457 {
4458 	int old = current->blocked.sig[0];
4459 	sigset_t newset;
4460 
4461 	siginitset(&newset, newmask);
4462 	set_current_blocked(&newset);
4463 
4464 	return old;
4465 }
4466 #endif /* CONFIG_SGETMASK_SYSCALL */
4467 
4468 #ifdef __ARCH_WANT_SYS_SIGNAL
4469 /*
4470  * For backwards compatibility.  Functionality superseded by sigaction.
4471  */
SYSCALL_DEFINE2(signal,int,sig,__sighandler_t,handler)4472 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4473 {
4474 	struct k_sigaction new_sa, old_sa;
4475 	int ret;
4476 
4477 	new_sa.sa.sa_handler = handler;
4478 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4479 	sigemptyset(&new_sa.sa.sa_mask);
4480 
4481 	ret = do_sigaction(sig, &new_sa, &old_sa);
4482 
4483 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4484 }
4485 #endif /* __ARCH_WANT_SYS_SIGNAL */
4486 
4487 #ifdef __ARCH_WANT_SYS_PAUSE
4488 
SYSCALL_DEFINE0(pause)4489 SYSCALL_DEFINE0(pause)
4490 {
4491 	while (!signal_pending(current)) {
4492 		__set_current_state(TASK_INTERRUPTIBLE);
4493 		schedule();
4494 	}
4495 	return -ERESTARTNOHAND;
4496 }
4497 
4498 #endif
4499 
sigsuspend(sigset_t * set)4500 static int sigsuspend(sigset_t *set)
4501 {
4502 	current->saved_sigmask = current->blocked;
4503 	set_current_blocked(set);
4504 
4505 	while (!signal_pending(current)) {
4506 		__set_current_state(TASK_INTERRUPTIBLE);
4507 		schedule();
4508 	}
4509 	set_restore_sigmask();
4510 	return -ERESTARTNOHAND;
4511 }
4512 
4513 /**
4514  *  sys_rt_sigsuspend - replace the signal mask for a value with the
4515  *	@unewset value until a signal is received
4516  *  @unewset: new signal mask value
4517  *  @sigsetsize: size of sigset_t type
4518  */
SYSCALL_DEFINE2(rt_sigsuspend,sigset_t __user *,unewset,size_t,sigsetsize)4519 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4520 {
4521 	sigset_t newset;
4522 
4523 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4524 	if (sigsetsize != sizeof(sigset_t))
4525 		return -EINVAL;
4526 
4527 	if (copy_from_user(&newset, unewset, sizeof(newset)))
4528 		return -EFAULT;
4529 	return sigsuspend(&newset);
4530 }
4531 
4532 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(rt_sigsuspend,compat_sigset_t __user *,unewset,compat_size_t,sigsetsize)4533 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4534 {
4535 	sigset_t newset;
4536 
4537 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4538 	if (sigsetsize != sizeof(sigset_t))
4539 		return -EINVAL;
4540 
4541 	if (get_compat_sigset(&newset, unewset))
4542 		return -EFAULT;
4543 	return sigsuspend(&newset);
4544 }
4545 #endif
4546 
4547 #ifdef CONFIG_OLD_SIGSUSPEND
SYSCALL_DEFINE1(sigsuspend,old_sigset_t,mask)4548 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4549 {
4550 	sigset_t blocked;
4551 	siginitset(&blocked, mask);
4552 	return sigsuspend(&blocked);
4553 }
4554 #endif
4555 #ifdef CONFIG_OLD_SIGSUSPEND3
SYSCALL_DEFINE3(sigsuspend,int,unused1,int,unused2,old_sigset_t,mask)4556 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4557 {
4558 	sigset_t blocked;
4559 	siginitset(&blocked, mask);
4560 	return sigsuspend(&blocked);
4561 }
4562 #endif
4563 
arch_vma_name(struct vm_area_struct * vma)4564 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4565 {
4566 	return NULL;
4567 }
4568 
siginfo_buildtime_checks(void)4569 static inline void siginfo_buildtime_checks(void)
4570 {
4571 	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4572 
4573 	/* Verify the offsets in the two siginfos match */
4574 #define CHECK_OFFSET(field) \
4575 	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4576 
4577 	/* kill */
4578 	CHECK_OFFSET(si_pid);
4579 	CHECK_OFFSET(si_uid);
4580 
4581 	/* timer */
4582 	CHECK_OFFSET(si_tid);
4583 	CHECK_OFFSET(si_overrun);
4584 	CHECK_OFFSET(si_value);
4585 
4586 	/* rt */
4587 	CHECK_OFFSET(si_pid);
4588 	CHECK_OFFSET(si_uid);
4589 	CHECK_OFFSET(si_value);
4590 
4591 	/* sigchld */
4592 	CHECK_OFFSET(si_pid);
4593 	CHECK_OFFSET(si_uid);
4594 	CHECK_OFFSET(si_status);
4595 	CHECK_OFFSET(si_utime);
4596 	CHECK_OFFSET(si_stime);
4597 
4598 	/* sigfault */
4599 	CHECK_OFFSET(si_addr);
4600 	CHECK_OFFSET(si_addr_lsb);
4601 	CHECK_OFFSET(si_lower);
4602 	CHECK_OFFSET(si_upper);
4603 	CHECK_OFFSET(si_pkey);
4604 
4605 	/* sigpoll */
4606 	CHECK_OFFSET(si_band);
4607 	CHECK_OFFSET(si_fd);
4608 
4609 	/* sigsys */
4610 	CHECK_OFFSET(si_call_addr);
4611 	CHECK_OFFSET(si_syscall);
4612 	CHECK_OFFSET(si_arch);
4613 #undef CHECK_OFFSET
4614 
4615 	/* usb asyncio */
4616 	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4617 		     offsetof(struct siginfo, si_addr));
4618 	if (sizeof(int) == sizeof(void __user *)) {
4619 		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4620 			     sizeof(void __user *));
4621 	} else {
4622 		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4623 			      sizeof_field(struct siginfo, si_uid)) !=
4624 			     sizeof(void __user *));
4625 		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4626 			     offsetof(struct siginfo, si_uid));
4627 	}
4628 #ifdef CONFIG_COMPAT
4629 	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4630 		     offsetof(struct compat_siginfo, si_addr));
4631 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4632 		     sizeof(compat_uptr_t));
4633 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4634 		     sizeof_field(struct siginfo, si_pid));
4635 #endif
4636 }
4637 
signals_init(void)4638 void __init signals_init(void)
4639 {
4640 	siginfo_buildtime_checks();
4641 
4642 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
4643 }
4644 
4645 #ifdef CONFIG_KGDB_KDB
4646 #include <linux/kdb.h>
4647 /*
4648  * kdb_send_sig - Allows kdb to send signals without exposing
4649  * signal internals.  This function checks if the required locks are
4650  * available before calling the main signal code, to avoid kdb
4651  * deadlocks.
4652  */
kdb_send_sig(struct task_struct * t,int sig)4653 void kdb_send_sig(struct task_struct *t, int sig)
4654 {
4655 	static struct task_struct *kdb_prev_t;
4656 	int new_t, ret;
4657 	if (!spin_trylock(&t->sighand->siglock)) {
4658 		kdb_printf("Can't do kill command now.\n"
4659 			   "The sigmask lock is held somewhere else in "
4660 			   "kernel, try again later\n");
4661 		return;
4662 	}
4663 	new_t = kdb_prev_t != t;
4664 	kdb_prev_t = t;
4665 	if (t->state != TASK_RUNNING && new_t) {
4666 		spin_unlock(&t->sighand->siglock);
4667 		kdb_printf("Process is not RUNNING, sending a signal from "
4668 			   "kdb risks deadlock\n"
4669 			   "on the run queue locks. "
4670 			   "The signal has _not_ been sent.\n"
4671 			   "Reissue the kill command if you want to risk "
4672 			   "the deadlock.\n");
4673 		return;
4674 	}
4675 	ret = send_signal(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4676 	spin_unlock(&t->sighand->siglock);
4677 	if (ret)
4678 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
4679 			   sig, t->pid);
4680 	else
4681 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4682 }
4683 #endif	/* CONFIG_KGDB_KDB */
4684