• 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 			trace_android_vh_killed_process(current, p, &reap);
1423 			if (reap)
1424 				add_to_oom_reaper(p);
1425 		}
1426 	}
1427 
1428 	return ret;
1429 }
1430 
1431 /*
1432  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1433  * control characters do (^C, ^Z etc)
1434  * - the caller must hold at least a readlock on tasklist_lock
1435  */
__kill_pgrp_info(int sig,struct kernel_siginfo * info,struct pid * pgrp)1436 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1437 {
1438 	struct task_struct *p = NULL;
1439 	int retval, success;
1440 
1441 	success = 0;
1442 	retval = -ESRCH;
1443 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1444 		int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1445 		success |= !err;
1446 		retval = err;
1447 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1448 	return success ? 0 : retval;
1449 }
1450 
kill_pid_info(int sig,struct kernel_siginfo * info,struct pid * pid)1451 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1452 {
1453 	int error = -ESRCH;
1454 	struct task_struct *p;
1455 
1456 	for (;;) {
1457 		rcu_read_lock();
1458 		p = pid_task(pid, PIDTYPE_PID);
1459 		if (p)
1460 			error = group_send_sig_info(sig, info, p, PIDTYPE_TGID);
1461 		rcu_read_unlock();
1462 		if (likely(!p || error != -ESRCH))
1463 			return error;
1464 
1465 		/*
1466 		 * The task was unhashed in between, try again.  If it
1467 		 * is dead, pid_task() will return NULL, if we race with
1468 		 * de_thread() it will find the new leader.
1469 		 */
1470 	}
1471 }
1472 
kill_proc_info(int sig,struct kernel_siginfo * info,pid_t pid)1473 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1474 {
1475 	int error;
1476 	rcu_read_lock();
1477 	error = kill_pid_info(sig, info, find_vpid(pid));
1478 	rcu_read_unlock();
1479 	return error;
1480 }
1481 
kill_as_cred_perm(const struct cred * cred,struct task_struct * target)1482 static inline bool kill_as_cred_perm(const struct cred *cred,
1483 				     struct task_struct *target)
1484 {
1485 	const struct cred *pcred = __task_cred(target);
1486 
1487 	return uid_eq(cred->euid, pcred->suid) ||
1488 	       uid_eq(cred->euid, pcred->uid) ||
1489 	       uid_eq(cred->uid, pcred->suid) ||
1490 	       uid_eq(cred->uid, pcred->uid);
1491 }
1492 
1493 /*
1494  * The usb asyncio usage of siginfo is wrong.  The glibc support
1495  * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1496  * AKA after the generic fields:
1497  *	kernel_pid_t	si_pid;
1498  *	kernel_uid32_t	si_uid;
1499  *	sigval_t	si_value;
1500  *
1501  * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1502  * after the generic fields is:
1503  *	void __user 	*si_addr;
1504  *
1505  * This is a practical problem when there is a 64bit big endian kernel
1506  * and a 32bit userspace.  As the 32bit address will encoded in the low
1507  * 32bits of the pointer.  Those low 32bits will be stored at higher
1508  * address than appear in a 32 bit pointer.  So userspace will not
1509  * see the address it was expecting for it's completions.
1510  *
1511  * There is nothing in the encoding that can allow
1512  * copy_siginfo_to_user32 to detect this confusion of formats, so
1513  * handle this by requiring the caller of kill_pid_usb_asyncio to
1514  * notice when this situration takes place and to store the 32bit
1515  * pointer in sival_int, instead of sival_addr of the sigval_t addr
1516  * parameter.
1517  */
kill_pid_usb_asyncio(int sig,int errno,sigval_t addr,struct pid * pid,const struct cred * cred)1518 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1519 			 struct pid *pid, const struct cred *cred)
1520 {
1521 	struct kernel_siginfo info;
1522 	struct task_struct *p;
1523 	unsigned long flags;
1524 	int ret = -EINVAL;
1525 
1526 	if (!valid_signal(sig))
1527 		return ret;
1528 
1529 	clear_siginfo(&info);
1530 	info.si_signo = sig;
1531 	info.si_errno = errno;
1532 	info.si_code = SI_ASYNCIO;
1533 	*((sigval_t *)&info.si_pid) = addr;
1534 
1535 	rcu_read_lock();
1536 	p = pid_task(pid, PIDTYPE_PID);
1537 	if (!p) {
1538 		ret = -ESRCH;
1539 		goto out_unlock;
1540 	}
1541 	if (!kill_as_cred_perm(cred, p)) {
1542 		ret = -EPERM;
1543 		goto out_unlock;
1544 	}
1545 	ret = security_task_kill(p, &info, sig, cred);
1546 	if (ret)
1547 		goto out_unlock;
1548 
1549 	if (sig) {
1550 		if (lock_task_sighand(p, &flags)) {
1551 			ret = __send_signal(sig, &info, p, PIDTYPE_TGID, false);
1552 			unlock_task_sighand(p, &flags);
1553 		} else
1554 			ret = -ESRCH;
1555 	}
1556 out_unlock:
1557 	rcu_read_unlock();
1558 	return ret;
1559 }
1560 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1561 
1562 /*
1563  * kill_something_info() interprets pid in interesting ways just like kill(2).
1564  *
1565  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1566  * is probably wrong.  Should make it like BSD or SYSV.
1567  */
1568 
kill_something_info(int sig,struct kernel_siginfo * info,pid_t pid)1569 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1570 {
1571 	int ret;
1572 
1573 	if (pid > 0)
1574 		return kill_proc_info(sig, info, pid);
1575 
1576 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1577 	if (pid == INT_MIN)
1578 		return -ESRCH;
1579 
1580 	read_lock(&tasklist_lock);
1581 	if (pid != -1) {
1582 		ret = __kill_pgrp_info(sig, info,
1583 				pid ? find_vpid(-pid) : task_pgrp(current));
1584 	} else {
1585 		int retval = 0, count = 0;
1586 		struct task_struct * p;
1587 
1588 		for_each_process(p) {
1589 			if (task_pid_vnr(p) > 1 &&
1590 					!same_thread_group(p, current)) {
1591 				int err = group_send_sig_info(sig, info, p,
1592 							      PIDTYPE_MAX);
1593 				++count;
1594 				if (err != -EPERM)
1595 					retval = err;
1596 			}
1597 		}
1598 		ret = count ? retval : -ESRCH;
1599 	}
1600 	read_unlock(&tasklist_lock);
1601 
1602 	return ret;
1603 }
1604 
1605 /*
1606  * These are for backward compatibility with the rest of the kernel source.
1607  */
1608 
send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p)1609 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1610 {
1611 	/*
1612 	 * Make sure legacy kernel users don't send in bad values
1613 	 * (normal paths check this in check_kill_permission).
1614 	 */
1615 	if (!valid_signal(sig))
1616 		return -EINVAL;
1617 
1618 	return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1619 }
1620 EXPORT_SYMBOL(send_sig_info);
1621 
1622 #define __si_special(priv) \
1623 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1624 
1625 int
send_sig(int sig,struct task_struct * p,int priv)1626 send_sig(int sig, struct task_struct *p, int priv)
1627 {
1628 	return send_sig_info(sig, __si_special(priv), p);
1629 }
1630 EXPORT_SYMBOL(send_sig);
1631 
force_sig(int sig)1632 void force_sig(int sig)
1633 {
1634 	struct kernel_siginfo info;
1635 
1636 	clear_siginfo(&info);
1637 	info.si_signo = sig;
1638 	info.si_errno = 0;
1639 	info.si_code = SI_KERNEL;
1640 	info.si_pid = 0;
1641 	info.si_uid = 0;
1642 	force_sig_info(&info);
1643 }
1644 EXPORT_SYMBOL(force_sig);
1645 
1646 /*
1647  * When things go south during signal handling, we
1648  * will force a SIGSEGV. And if the signal that caused
1649  * the problem was already a SIGSEGV, we'll want to
1650  * make sure we don't even try to deliver the signal..
1651  */
force_sigsegv(int sig)1652 void force_sigsegv(int sig)
1653 {
1654 	struct task_struct *p = current;
1655 
1656 	if (sig == SIGSEGV) {
1657 		unsigned long flags;
1658 		spin_lock_irqsave(&p->sighand->siglock, flags);
1659 		p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1660 		spin_unlock_irqrestore(&p->sighand->siglock, flags);
1661 	}
1662 	force_sig(SIGSEGV);
1663 }
1664 
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)1665 int force_sig_fault_to_task(int sig, int code, void __user *addr
1666 	___ARCH_SI_TRAPNO(int trapno)
1667 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1668 	, struct task_struct *t)
1669 {
1670 	struct kernel_siginfo info;
1671 
1672 	clear_siginfo(&info);
1673 	info.si_signo = sig;
1674 	info.si_errno = 0;
1675 	info.si_code  = code;
1676 	info.si_addr  = addr;
1677 #ifdef __ARCH_SI_TRAPNO
1678 	info.si_trapno = trapno;
1679 #endif
1680 #ifdef __ia64__
1681 	info.si_imm = imm;
1682 	info.si_flags = flags;
1683 	info.si_isr = isr;
1684 #endif
1685 	return force_sig_info_to_task(&info, t);
1686 }
1687 
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))1688 int force_sig_fault(int sig, int code, void __user *addr
1689 	___ARCH_SI_TRAPNO(int trapno)
1690 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr))
1691 {
1692 	return force_sig_fault_to_task(sig, code, addr
1693 				       ___ARCH_SI_TRAPNO(trapno)
1694 				       ___ARCH_SI_IA64(imm, flags, isr), current);
1695 }
1696 
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)1697 int send_sig_fault(int sig, int code, void __user *addr
1698 	___ARCH_SI_TRAPNO(int trapno)
1699 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1700 	, struct task_struct *t)
1701 {
1702 	struct kernel_siginfo info;
1703 
1704 	clear_siginfo(&info);
1705 	info.si_signo = sig;
1706 	info.si_errno = 0;
1707 	info.si_code  = code;
1708 	info.si_addr  = addr;
1709 #ifdef __ARCH_SI_TRAPNO
1710 	info.si_trapno = trapno;
1711 #endif
1712 #ifdef __ia64__
1713 	info.si_imm = imm;
1714 	info.si_flags = flags;
1715 	info.si_isr = isr;
1716 #endif
1717 	return send_sig_info(info.si_signo, &info, t);
1718 }
1719 
force_sig_mceerr(int code,void __user * addr,short lsb)1720 int force_sig_mceerr(int code, void __user *addr, short lsb)
1721 {
1722 	struct kernel_siginfo info;
1723 
1724 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1725 	clear_siginfo(&info);
1726 	info.si_signo = SIGBUS;
1727 	info.si_errno = 0;
1728 	info.si_code = code;
1729 	info.si_addr = addr;
1730 	info.si_addr_lsb = lsb;
1731 	return force_sig_info(&info);
1732 }
1733 
send_sig_mceerr(int code,void __user * addr,short lsb,struct task_struct * t)1734 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1735 {
1736 	struct kernel_siginfo info;
1737 
1738 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1739 	clear_siginfo(&info);
1740 	info.si_signo = SIGBUS;
1741 	info.si_errno = 0;
1742 	info.si_code = code;
1743 	info.si_addr = addr;
1744 	info.si_addr_lsb = lsb;
1745 	return send_sig_info(info.si_signo, &info, t);
1746 }
1747 EXPORT_SYMBOL(send_sig_mceerr);
1748 
force_sig_bnderr(void __user * addr,void __user * lower,void __user * upper)1749 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1750 {
1751 	struct kernel_siginfo info;
1752 
1753 	clear_siginfo(&info);
1754 	info.si_signo = SIGSEGV;
1755 	info.si_errno = 0;
1756 	info.si_code  = SEGV_BNDERR;
1757 	info.si_addr  = addr;
1758 	info.si_lower = lower;
1759 	info.si_upper = upper;
1760 	return force_sig_info(&info);
1761 }
1762 
1763 #ifdef SEGV_PKUERR
force_sig_pkuerr(void __user * addr,u32 pkey)1764 int force_sig_pkuerr(void __user *addr, u32 pkey)
1765 {
1766 	struct kernel_siginfo info;
1767 
1768 	clear_siginfo(&info);
1769 	info.si_signo = SIGSEGV;
1770 	info.si_errno = 0;
1771 	info.si_code  = SEGV_PKUERR;
1772 	info.si_addr  = addr;
1773 	info.si_pkey  = pkey;
1774 	return force_sig_info(&info);
1775 }
1776 #endif
1777 
1778 /* For the crazy architectures that include trap information in
1779  * the errno field, instead of an actual errno value.
1780  */
force_sig_ptrace_errno_trap(int errno,void __user * addr)1781 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1782 {
1783 	struct kernel_siginfo info;
1784 
1785 	clear_siginfo(&info);
1786 	info.si_signo = SIGTRAP;
1787 	info.si_errno = errno;
1788 	info.si_code  = TRAP_HWBKPT;
1789 	info.si_addr  = addr;
1790 	return force_sig_info(&info);
1791 }
1792 
kill_pgrp(struct pid * pid,int sig,int priv)1793 int kill_pgrp(struct pid *pid, int sig, int priv)
1794 {
1795 	int ret;
1796 
1797 	read_lock(&tasklist_lock);
1798 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1799 	read_unlock(&tasklist_lock);
1800 
1801 	return ret;
1802 }
1803 EXPORT_SYMBOL(kill_pgrp);
1804 
kill_pid(struct pid * pid,int sig,int priv)1805 int kill_pid(struct pid *pid, int sig, int priv)
1806 {
1807 	return kill_pid_info(sig, __si_special(priv), pid);
1808 }
1809 EXPORT_SYMBOL(kill_pid);
1810 
1811 /*
1812  * These functions support sending signals using preallocated sigqueue
1813  * structures.  This is needed "because realtime applications cannot
1814  * afford to lose notifications of asynchronous events, like timer
1815  * expirations or I/O completions".  In the case of POSIX Timers
1816  * we allocate the sigqueue structure from the timer_create.  If this
1817  * allocation fails we are able to report the failure to the application
1818  * with an EAGAIN error.
1819  */
sigqueue_alloc(void)1820 struct sigqueue *sigqueue_alloc(void)
1821 {
1822 	struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1823 
1824 	if (q)
1825 		q->flags |= SIGQUEUE_PREALLOC;
1826 
1827 	return q;
1828 }
1829 
sigqueue_free(struct sigqueue * q)1830 void sigqueue_free(struct sigqueue *q)
1831 {
1832 	unsigned long flags;
1833 	spinlock_t *lock = &current->sighand->siglock;
1834 
1835 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1836 	/*
1837 	 * We must hold ->siglock while testing q->list
1838 	 * to serialize with collect_signal() or with
1839 	 * __exit_signal()->flush_sigqueue().
1840 	 */
1841 	spin_lock_irqsave(lock, flags);
1842 	q->flags &= ~SIGQUEUE_PREALLOC;
1843 	/*
1844 	 * If it is queued it will be freed when dequeued,
1845 	 * like the "regular" sigqueue.
1846 	 */
1847 	if (!list_empty(&q->list))
1848 		q = NULL;
1849 	spin_unlock_irqrestore(lock, flags);
1850 
1851 	if (q)
1852 		__sigqueue_free(q);
1853 }
1854 
send_sigqueue(struct sigqueue * q,struct pid * pid,enum pid_type type)1855 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1856 {
1857 	int sig = q->info.si_signo;
1858 	struct sigpending *pending;
1859 	struct task_struct *t;
1860 	unsigned long flags;
1861 	int ret, result;
1862 
1863 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1864 
1865 	ret = -1;
1866 	rcu_read_lock();
1867 	t = pid_task(pid, type);
1868 	if (!t || !likely(lock_task_sighand(t, &flags)))
1869 		goto ret;
1870 
1871 	ret = 1; /* the signal is ignored */
1872 	result = TRACE_SIGNAL_IGNORED;
1873 	if (!prepare_signal(sig, t, false))
1874 		goto out;
1875 
1876 	ret = 0;
1877 	if (unlikely(!list_empty(&q->list))) {
1878 		/*
1879 		 * If an SI_TIMER entry is already queue just increment
1880 		 * the overrun count.
1881 		 */
1882 		BUG_ON(q->info.si_code != SI_TIMER);
1883 		q->info.si_overrun++;
1884 		result = TRACE_SIGNAL_ALREADY_PENDING;
1885 		goto out;
1886 	}
1887 	q->info.si_overrun = 0;
1888 
1889 	signalfd_notify(t, sig);
1890 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1891 	list_add_tail(&q->list, &pending->list);
1892 	sigaddset(&pending->signal, sig);
1893 	complete_signal(sig, t, type);
1894 	result = TRACE_SIGNAL_DELIVERED;
1895 out:
1896 	trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
1897 	unlock_task_sighand(t, &flags);
1898 ret:
1899 	rcu_read_unlock();
1900 	return ret;
1901 }
1902 
do_notify_pidfd(struct task_struct * task)1903 static void do_notify_pidfd(struct task_struct *task)
1904 {
1905 	struct pid *pid;
1906 
1907 	WARN_ON(task->exit_state == 0);
1908 	pid = task_pid(task);
1909 	wake_up_all(&pid->wait_pidfd);
1910 }
1911 
1912 /*
1913  * Let a parent know about the death of a child.
1914  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1915  *
1916  * Returns true if our parent ignored us and so we've switched to
1917  * self-reaping.
1918  */
do_notify_parent(struct task_struct * tsk,int sig)1919 bool do_notify_parent(struct task_struct *tsk, int sig)
1920 {
1921 	struct kernel_siginfo info;
1922 	unsigned long flags;
1923 	struct sighand_struct *psig;
1924 	bool autoreap = false;
1925 	u64 utime, stime;
1926 
1927 	WARN_ON_ONCE(sig == -1);
1928 
1929 	/* do_notify_parent_cldstop should have been called instead.  */
1930 	WARN_ON_ONCE(task_is_stopped_or_traced(tsk));
1931 
1932 	WARN_ON_ONCE(!tsk->ptrace &&
1933 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1934 
1935 	/* Wake up all pidfd waiters */
1936 	do_notify_pidfd(tsk);
1937 
1938 	if (sig != SIGCHLD) {
1939 		/*
1940 		 * This is only possible if parent == real_parent.
1941 		 * Check if it has changed security domain.
1942 		 */
1943 		if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
1944 			sig = SIGCHLD;
1945 	}
1946 
1947 	clear_siginfo(&info);
1948 	info.si_signo = sig;
1949 	info.si_errno = 0;
1950 	/*
1951 	 * We are under tasklist_lock here so our parent is tied to
1952 	 * us and cannot change.
1953 	 *
1954 	 * task_active_pid_ns will always return the same pid namespace
1955 	 * until a task passes through release_task.
1956 	 *
1957 	 * write_lock() currently calls preempt_disable() which is the
1958 	 * same as rcu_read_lock(), but according to Oleg, this is not
1959 	 * correct to rely on this
1960 	 */
1961 	rcu_read_lock();
1962 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1963 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1964 				       task_uid(tsk));
1965 	rcu_read_unlock();
1966 
1967 	task_cputime(tsk, &utime, &stime);
1968 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
1969 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
1970 
1971 	info.si_status = tsk->exit_code & 0x7f;
1972 	if (tsk->exit_code & 0x80)
1973 		info.si_code = CLD_DUMPED;
1974 	else if (tsk->exit_code & 0x7f)
1975 		info.si_code = CLD_KILLED;
1976 	else {
1977 		info.si_code = CLD_EXITED;
1978 		info.si_status = tsk->exit_code >> 8;
1979 	}
1980 
1981 	psig = tsk->parent->sighand;
1982 	spin_lock_irqsave(&psig->siglock, flags);
1983 	if (!tsk->ptrace && sig == SIGCHLD &&
1984 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1985 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1986 		/*
1987 		 * We are exiting and our parent doesn't care.  POSIX.1
1988 		 * defines special semantics for setting SIGCHLD to SIG_IGN
1989 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
1990 		 * automatically and not left for our parent's wait4 call.
1991 		 * Rather than having the parent do it as a magic kind of
1992 		 * signal handler, we just set this to tell do_exit that we
1993 		 * can be cleaned up without becoming a zombie.  Note that
1994 		 * we still call __wake_up_parent in this case, because a
1995 		 * blocked sys_wait4 might now return -ECHILD.
1996 		 *
1997 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1998 		 * is implementation-defined: we do (if you don't want
1999 		 * it, just use SIG_IGN instead).
2000 		 */
2001 		autoreap = true;
2002 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2003 			sig = 0;
2004 	}
2005 	/*
2006 	 * Send with __send_signal as si_pid and si_uid are in the
2007 	 * parent's namespaces.
2008 	 */
2009 	if (valid_signal(sig) && sig)
2010 		__send_signal(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2011 	__wake_up_parent(tsk, tsk->parent);
2012 	spin_unlock_irqrestore(&psig->siglock, flags);
2013 
2014 	return autoreap;
2015 }
2016 
2017 /**
2018  * do_notify_parent_cldstop - notify parent of stopped/continued state change
2019  * @tsk: task reporting the state change
2020  * @for_ptracer: the notification is for ptracer
2021  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2022  *
2023  * Notify @tsk's parent that the stopped/continued state has changed.  If
2024  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2025  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2026  *
2027  * CONTEXT:
2028  * Must be called with tasklist_lock at least read locked.
2029  */
do_notify_parent_cldstop(struct task_struct * tsk,bool for_ptracer,int why)2030 static void do_notify_parent_cldstop(struct task_struct *tsk,
2031 				     bool for_ptracer, int why)
2032 {
2033 	struct kernel_siginfo info;
2034 	unsigned long flags;
2035 	struct task_struct *parent;
2036 	struct sighand_struct *sighand;
2037 	u64 utime, stime;
2038 
2039 	if (for_ptracer) {
2040 		parent = tsk->parent;
2041 	} else {
2042 		tsk = tsk->group_leader;
2043 		parent = tsk->real_parent;
2044 	}
2045 
2046 	clear_siginfo(&info);
2047 	info.si_signo = SIGCHLD;
2048 	info.si_errno = 0;
2049 	/*
2050 	 * see comment in do_notify_parent() about the following 4 lines
2051 	 */
2052 	rcu_read_lock();
2053 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2054 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2055 	rcu_read_unlock();
2056 
2057 	task_cputime(tsk, &utime, &stime);
2058 	info.si_utime = nsec_to_clock_t(utime);
2059 	info.si_stime = nsec_to_clock_t(stime);
2060 
2061  	info.si_code = why;
2062  	switch (why) {
2063  	case CLD_CONTINUED:
2064  		info.si_status = SIGCONT;
2065  		break;
2066  	case CLD_STOPPED:
2067  		info.si_status = tsk->signal->group_exit_code & 0x7f;
2068  		break;
2069  	case CLD_TRAPPED:
2070  		info.si_status = tsk->exit_code & 0x7f;
2071  		break;
2072  	default:
2073  		BUG();
2074  	}
2075 
2076 	sighand = parent->sighand;
2077 	spin_lock_irqsave(&sighand->siglock, flags);
2078 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2079 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2080 		__group_send_sig_info(SIGCHLD, &info, parent);
2081 	/*
2082 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2083 	 */
2084 	__wake_up_parent(tsk, parent);
2085 	spin_unlock_irqrestore(&sighand->siglock, flags);
2086 }
2087 
may_ptrace_stop(void)2088 static inline bool may_ptrace_stop(void)
2089 {
2090 	if (!likely(current->ptrace))
2091 		return false;
2092 	/*
2093 	 * Are we in the middle of do_coredump?
2094 	 * If so and our tracer is also part of the coredump stopping
2095 	 * is a deadlock situation, and pointless because our tracer
2096 	 * is dead so don't allow us to stop.
2097 	 * If SIGKILL was already sent before the caller unlocked
2098 	 * ->siglock we must see ->core_state != NULL. Otherwise it
2099 	 * is safe to enter schedule().
2100 	 *
2101 	 * This is almost outdated, a task with the pending SIGKILL can't
2102 	 * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
2103 	 * after SIGKILL was already dequeued.
2104 	 */
2105 	if (unlikely(current->mm->core_state) &&
2106 	    unlikely(current->mm == current->parent->mm))
2107 		return false;
2108 
2109 	return true;
2110 }
2111 
2112 
2113 /*
2114  * This must be called with current->sighand->siglock held.
2115  *
2116  * This should be the path for all ptrace stops.
2117  * We always set current->last_siginfo while stopped here.
2118  * That makes it a way to test a stopped process for
2119  * being ptrace-stopped vs being job-control-stopped.
2120  *
2121  * If we actually decide not to stop at all because the tracer
2122  * is gone, we keep current->exit_code unless clear_code.
2123  */
ptrace_stop(int exit_code,int why,int clear_code,kernel_siginfo_t * info)2124 static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t *info)
2125 	__releases(&current->sighand->siglock)
2126 	__acquires(&current->sighand->siglock)
2127 {
2128 	bool gstop_done = false;
2129 
2130 	if (arch_ptrace_stop_needed(exit_code, info)) {
2131 		/*
2132 		 * The arch code has something special to do before a
2133 		 * ptrace stop.  This is allowed to block, e.g. for faults
2134 		 * on user stack pages.  We can't keep the siglock while
2135 		 * calling arch_ptrace_stop, so we must release it now.
2136 		 * To preserve proper semantics, we must do this before
2137 		 * any signal bookkeeping like checking group_stop_count.
2138 		 */
2139 		spin_unlock_irq(&current->sighand->siglock);
2140 		arch_ptrace_stop(exit_code, info);
2141 		spin_lock_irq(&current->sighand->siglock);
2142 	}
2143 
2144 	/*
2145 	 * schedule() will not sleep if there is a pending signal that
2146 	 * can awaken the task.
2147 	 */
2148 	set_special_state(TASK_TRACED);
2149 
2150 	/*
2151 	 * We're committing to trapping.  TRACED should be visible before
2152 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2153 	 * Also, transition to TRACED and updates to ->jobctl should be
2154 	 * atomic with respect to siglock and should be done after the arch
2155 	 * hook as siglock is released and regrabbed across it.
2156 	 *
2157 	 *     TRACER				    TRACEE
2158 	 *
2159 	 *     ptrace_attach()
2160 	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2161 	 *     do_wait()
2162 	 *       set_current_state()                smp_wmb();
2163 	 *       ptrace_do_wait()
2164 	 *         wait_task_stopped()
2165 	 *           task_stopped_code()
2166 	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2167 	 */
2168 	smp_wmb();
2169 
2170 	current->last_siginfo = info;
2171 	current->exit_code = exit_code;
2172 
2173 	/*
2174 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2175 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2176 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2177 	 * could be clear now.  We act as if SIGCONT is received after
2178 	 * TASK_TRACED is entered - ignore it.
2179 	 */
2180 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2181 		gstop_done = task_participate_group_stop(current);
2182 
2183 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2184 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2185 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2186 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2187 
2188 	/* entering a trap, clear TRAPPING */
2189 	task_clear_jobctl_trapping(current);
2190 
2191 	spin_unlock_irq(&current->sighand->siglock);
2192 	read_lock(&tasklist_lock);
2193 	if (may_ptrace_stop()) {
2194 		/*
2195 		 * Notify parents of the stop.
2196 		 *
2197 		 * While ptraced, there are two parents - the ptracer and
2198 		 * the real_parent of the group_leader.  The ptracer should
2199 		 * know about every stop while the real parent is only
2200 		 * interested in the completion of group stop.  The states
2201 		 * for the two don't interact with each other.  Notify
2202 		 * separately unless they're gonna be duplicates.
2203 		 */
2204 		do_notify_parent_cldstop(current, true, why);
2205 		if (gstop_done && ptrace_reparented(current))
2206 			do_notify_parent_cldstop(current, false, why);
2207 
2208 		/*
2209 		 * Don't want to allow preemption here, because
2210 		 * sys_ptrace() needs this task to be inactive.
2211 		 *
2212 		 * XXX: implement read_unlock_no_resched().
2213 		 */
2214 		preempt_disable();
2215 		read_unlock(&tasklist_lock);
2216 		cgroup_enter_frozen();
2217 		preempt_enable_no_resched();
2218 		freezable_schedule();
2219 		cgroup_leave_frozen(true);
2220 	} else {
2221 		/*
2222 		 * By the time we got the lock, our tracer went away.
2223 		 * Don't drop the lock yet, another tracer may come.
2224 		 *
2225 		 * If @gstop_done, the ptracer went away between group stop
2226 		 * completion and here.  During detach, it would have set
2227 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
2228 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
2229 		 * the real parent of the group stop completion is enough.
2230 		 */
2231 		if (gstop_done)
2232 			do_notify_parent_cldstop(current, false, why);
2233 
2234 		/* tasklist protects us from ptrace_freeze_traced() */
2235 		__set_current_state(TASK_RUNNING);
2236 		if (clear_code)
2237 			current->exit_code = 0;
2238 		read_unlock(&tasklist_lock);
2239 	}
2240 
2241 	/*
2242 	 * We are back.  Now reacquire the siglock before touching
2243 	 * last_siginfo, so that we are sure to have synchronized with
2244 	 * any signal-sending on another CPU that wants to examine it.
2245 	 */
2246 	spin_lock_irq(&current->sighand->siglock);
2247 	current->last_siginfo = NULL;
2248 
2249 	/* LISTENING can be set only during STOP traps, clear it */
2250 	current->jobctl &= ~JOBCTL_LISTENING;
2251 
2252 	/*
2253 	 * Queued signals ignored us while we were stopped for tracing.
2254 	 * So check for any that we should take before resuming user mode.
2255 	 * This sets TIF_SIGPENDING, but never clears it.
2256 	 */
2257 	recalc_sigpending_tsk(current);
2258 }
2259 
ptrace_do_notify(int signr,int exit_code,int why)2260 static void ptrace_do_notify(int signr, int exit_code, int why)
2261 {
2262 	kernel_siginfo_t info;
2263 
2264 	clear_siginfo(&info);
2265 	info.si_signo = signr;
2266 	info.si_code = exit_code;
2267 	info.si_pid = task_pid_vnr(current);
2268 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2269 
2270 	/* Let the debugger run.  */
2271 	ptrace_stop(exit_code, why, 1, &info);
2272 }
2273 
ptrace_notify(int exit_code)2274 void ptrace_notify(int exit_code)
2275 {
2276 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2277 	if (unlikely(current->task_works))
2278 		task_work_run();
2279 
2280 	spin_lock_irq(&current->sighand->siglock);
2281 	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
2282 	spin_unlock_irq(&current->sighand->siglock);
2283 }
2284 
2285 /**
2286  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2287  * @signr: signr causing group stop if initiating
2288  *
2289  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2290  * and participate in it.  If already set, participate in the existing
2291  * group stop.  If participated in a group stop (and thus slept), %true is
2292  * returned with siglock released.
2293  *
2294  * If ptraced, this function doesn't handle stop itself.  Instead,
2295  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2296  * untouched.  The caller must ensure that INTERRUPT trap handling takes
2297  * places afterwards.
2298  *
2299  * CONTEXT:
2300  * Must be called with @current->sighand->siglock held, which is released
2301  * on %true return.
2302  *
2303  * RETURNS:
2304  * %false if group stop is already cancelled or ptrace trap is scheduled.
2305  * %true if participated in group stop.
2306  */
do_signal_stop(int signr)2307 static bool do_signal_stop(int signr)
2308 	__releases(&current->sighand->siglock)
2309 {
2310 	struct signal_struct *sig = current->signal;
2311 
2312 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2313 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2314 		struct task_struct *t;
2315 
2316 		/* signr will be recorded in task->jobctl for retries */
2317 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2318 
2319 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2320 		    unlikely(signal_group_exit(sig)))
2321 			return false;
2322 		/*
2323 		 * There is no group stop already in progress.  We must
2324 		 * initiate one now.
2325 		 *
2326 		 * While ptraced, a task may be resumed while group stop is
2327 		 * still in effect and then receive a stop signal and
2328 		 * initiate another group stop.  This deviates from the
2329 		 * usual behavior as two consecutive stop signals can't
2330 		 * cause two group stops when !ptraced.  That is why we
2331 		 * also check !task_is_stopped(t) below.
2332 		 *
2333 		 * The condition can be distinguished by testing whether
2334 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2335 		 * group_exit_code in such case.
2336 		 *
2337 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2338 		 * an intervening stop signal is required to cause two
2339 		 * continued events regardless of ptrace.
2340 		 */
2341 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2342 			sig->group_exit_code = signr;
2343 
2344 		sig->group_stop_count = 0;
2345 
2346 		if (task_set_jobctl_pending(current, signr | gstop))
2347 			sig->group_stop_count++;
2348 
2349 		t = current;
2350 		while_each_thread(current, t) {
2351 			/*
2352 			 * Setting state to TASK_STOPPED for a group
2353 			 * stop is always done with the siglock held,
2354 			 * so this check has no races.
2355 			 */
2356 			if (!task_is_stopped(t) &&
2357 			    task_set_jobctl_pending(t, signr | gstop)) {
2358 				sig->group_stop_count++;
2359 				if (likely(!(t->ptrace & PT_SEIZED)))
2360 					signal_wake_up(t, 0);
2361 				else
2362 					ptrace_trap_notify(t);
2363 			}
2364 		}
2365 	}
2366 
2367 	if (likely(!current->ptrace)) {
2368 		int notify = 0;
2369 
2370 		/*
2371 		 * If there are no other threads in the group, or if there
2372 		 * is a group stop in progress and we are the last to stop,
2373 		 * report to the parent.
2374 		 */
2375 		if (task_participate_group_stop(current))
2376 			notify = CLD_STOPPED;
2377 
2378 		set_special_state(TASK_STOPPED);
2379 		spin_unlock_irq(&current->sighand->siglock);
2380 
2381 		/*
2382 		 * Notify the parent of the group stop completion.  Because
2383 		 * we're not holding either the siglock or tasklist_lock
2384 		 * here, ptracer may attach inbetween; however, this is for
2385 		 * group stop and should always be delivered to the real
2386 		 * parent of the group leader.  The new ptracer will get
2387 		 * its notification when this task transitions into
2388 		 * TASK_TRACED.
2389 		 */
2390 		if (notify) {
2391 			read_lock(&tasklist_lock);
2392 			do_notify_parent_cldstop(current, false, notify);
2393 			read_unlock(&tasklist_lock);
2394 		}
2395 
2396 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2397 		cgroup_enter_frozen();
2398 		freezable_schedule();
2399 		return true;
2400 	} else {
2401 		/*
2402 		 * While ptraced, group stop is handled by STOP trap.
2403 		 * Schedule it and let the caller deal with it.
2404 		 */
2405 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2406 		return false;
2407 	}
2408 }
2409 
2410 /**
2411  * do_jobctl_trap - take care of ptrace jobctl traps
2412  *
2413  * When PT_SEIZED, it's used for both group stop and explicit
2414  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2415  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2416  * the stop signal; otherwise, %SIGTRAP.
2417  *
2418  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2419  * number as exit_code and no siginfo.
2420  *
2421  * CONTEXT:
2422  * Must be called with @current->sighand->siglock held, which may be
2423  * released and re-acquired before returning with intervening sleep.
2424  */
do_jobctl_trap(void)2425 static void do_jobctl_trap(void)
2426 {
2427 	struct signal_struct *signal = current->signal;
2428 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2429 
2430 	if (current->ptrace & PT_SEIZED) {
2431 		if (!signal->group_stop_count &&
2432 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2433 			signr = SIGTRAP;
2434 		WARN_ON_ONCE(!signr);
2435 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2436 				 CLD_STOPPED);
2437 	} else {
2438 		WARN_ON_ONCE(!signr);
2439 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2440 		current->exit_code = 0;
2441 	}
2442 }
2443 
2444 /**
2445  * do_freezer_trap - handle the freezer jobctl trap
2446  *
2447  * Puts the task into frozen state, if only the task is not about to quit.
2448  * In this case it drops JOBCTL_TRAP_FREEZE.
2449  *
2450  * CONTEXT:
2451  * Must be called with @current->sighand->siglock held,
2452  * which is always released before returning.
2453  */
do_freezer_trap(void)2454 static void do_freezer_trap(void)
2455 	__releases(&current->sighand->siglock)
2456 {
2457 	/*
2458 	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2459 	 * let's make another loop to give it a chance to be handled.
2460 	 * In any case, we'll return back.
2461 	 */
2462 	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2463 	     JOBCTL_TRAP_FREEZE) {
2464 		spin_unlock_irq(&current->sighand->siglock);
2465 		return;
2466 	}
2467 
2468 	/*
2469 	 * Now we're sure that there is no pending fatal signal and no
2470 	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2471 	 * immediately (if there is a non-fatal signal pending), and
2472 	 * put the task into sleep.
2473 	 */
2474 	__set_current_state(TASK_INTERRUPTIBLE);
2475 	clear_thread_flag(TIF_SIGPENDING);
2476 	spin_unlock_irq(&current->sighand->siglock);
2477 	cgroup_enter_frozen();
2478 	freezable_schedule();
2479 }
2480 
ptrace_signal(int signr,kernel_siginfo_t * info)2481 static int ptrace_signal(int signr, kernel_siginfo_t *info)
2482 {
2483 	/*
2484 	 * We do not check sig_kernel_stop(signr) but set this marker
2485 	 * unconditionally because we do not know whether debugger will
2486 	 * change signr. This flag has no meaning unless we are going
2487 	 * to stop after return from ptrace_stop(). In this case it will
2488 	 * be checked in do_signal_stop(), we should only stop if it was
2489 	 * not cleared by SIGCONT while we were sleeping. See also the
2490 	 * comment in dequeue_signal().
2491 	 */
2492 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2493 	ptrace_stop(signr, CLD_TRAPPED, 0, info);
2494 
2495 	/* We're back.  Did the debugger cancel the sig?  */
2496 	signr = current->exit_code;
2497 	if (signr == 0)
2498 		return signr;
2499 
2500 	current->exit_code = 0;
2501 
2502 	/*
2503 	 * Update the siginfo structure if the signal has
2504 	 * changed.  If the debugger wanted something
2505 	 * specific in the siginfo structure then it should
2506 	 * have updated *info via PTRACE_SETSIGINFO.
2507 	 */
2508 	if (signr != info->si_signo) {
2509 		clear_siginfo(info);
2510 		info->si_signo = signr;
2511 		info->si_errno = 0;
2512 		info->si_code = SI_USER;
2513 		rcu_read_lock();
2514 		info->si_pid = task_pid_vnr(current->parent);
2515 		info->si_uid = from_kuid_munged(current_user_ns(),
2516 						task_uid(current->parent));
2517 		rcu_read_unlock();
2518 	}
2519 
2520 	/* If the (new) signal is now blocked, requeue it.  */
2521 	if (sigismember(&current->blocked, signr)) {
2522 		send_signal(signr, info, current, PIDTYPE_PID);
2523 		signr = 0;
2524 	}
2525 
2526 	return signr;
2527 }
2528 
hide_si_addr_tag_bits(struct ksignal * ksig)2529 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2530 {
2531 	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2532 	case SIL_FAULT:
2533 	case SIL_FAULT_MCEERR:
2534 	case SIL_FAULT_BNDERR:
2535 	case SIL_FAULT_PKUERR:
2536 		ksig->info.si_addr = arch_untagged_si_addr(
2537 			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2538 		break;
2539 	case SIL_KILL:
2540 	case SIL_TIMER:
2541 	case SIL_POLL:
2542 	case SIL_CHLD:
2543 	case SIL_RT:
2544 	case SIL_SYS:
2545 		break;
2546 	}
2547 }
2548 
get_signal(struct ksignal * ksig)2549 bool get_signal(struct ksignal *ksig)
2550 {
2551 	struct sighand_struct *sighand = current->sighand;
2552 	struct signal_struct *signal = current->signal;
2553 	int signr;
2554 
2555 	if (unlikely(current->task_works))
2556 		task_work_run();
2557 
2558 	/*
2559 	 * For non-generic architectures, check for TIF_NOTIFY_SIGNAL so
2560 	 * that the arch handlers don't all have to do it. If we get here
2561 	 * without TIF_SIGPENDING, just exit after running signal work.
2562 	 */
2563 	if (!IS_ENABLED(CONFIG_GENERIC_ENTRY)) {
2564 		if (test_thread_flag(TIF_NOTIFY_SIGNAL))
2565 			tracehook_notify_signal();
2566 		if (!task_sigpending(current))
2567 			return false;
2568 	}
2569 
2570 	if (unlikely(uprobe_deny_signal()))
2571 		return false;
2572 
2573 	/*
2574 	 * Do this once, we can't return to user-mode if freezing() == T.
2575 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2576 	 * thus do not need another check after return.
2577 	 */
2578 	try_to_freeze();
2579 
2580 relock:
2581 	spin_lock_irq(&sighand->siglock);
2582 
2583 	/*
2584 	 * Every stopped thread goes here after wakeup. Check to see if
2585 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2586 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2587 	 */
2588 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2589 		int why;
2590 
2591 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2592 			why = CLD_CONTINUED;
2593 		else
2594 			why = CLD_STOPPED;
2595 
2596 		signal->flags &= ~SIGNAL_CLD_MASK;
2597 
2598 		spin_unlock_irq(&sighand->siglock);
2599 
2600 		/*
2601 		 * Notify the parent that we're continuing.  This event is
2602 		 * always per-process and doesn't make whole lot of sense
2603 		 * for ptracers, who shouldn't consume the state via
2604 		 * wait(2) either, but, for backward compatibility, notify
2605 		 * the ptracer of the group leader too unless it's gonna be
2606 		 * a duplicate.
2607 		 */
2608 		read_lock(&tasklist_lock);
2609 		do_notify_parent_cldstop(current, false, why);
2610 
2611 		if (ptrace_reparented(current->group_leader))
2612 			do_notify_parent_cldstop(current->group_leader,
2613 						true, why);
2614 		read_unlock(&tasklist_lock);
2615 
2616 		goto relock;
2617 	}
2618 
2619 	/* Has this task already been marked for death? */
2620 	if (signal_group_exit(signal)) {
2621 		ksig->info.si_signo = signr = SIGKILL;
2622 		sigdelset(&current->pending.signal, SIGKILL);
2623 		trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2624 				&sighand->action[SIGKILL - 1]);
2625 		recalc_sigpending();
2626 		goto fatal;
2627 	}
2628 
2629 	for (;;) {
2630 		struct k_sigaction *ka;
2631 
2632 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2633 		    do_signal_stop(0))
2634 			goto relock;
2635 
2636 		if (unlikely(current->jobctl &
2637 			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2638 			if (current->jobctl & JOBCTL_TRAP_MASK) {
2639 				do_jobctl_trap();
2640 				spin_unlock_irq(&sighand->siglock);
2641 			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2642 				do_freezer_trap();
2643 
2644 			goto relock;
2645 		}
2646 
2647 		/*
2648 		 * If the task is leaving the frozen state, let's update
2649 		 * cgroup counters and reset the frozen bit.
2650 		 */
2651 		if (unlikely(cgroup_task_frozen(current))) {
2652 			spin_unlock_irq(&sighand->siglock);
2653 			cgroup_leave_frozen(false);
2654 			goto relock;
2655 		}
2656 
2657 		/*
2658 		 * Signals generated by the execution of an instruction
2659 		 * need to be delivered before any other pending signals
2660 		 * so that the instruction pointer in the signal stack
2661 		 * frame points to the faulting instruction.
2662 		 */
2663 		signr = dequeue_synchronous_signal(&ksig->info);
2664 		if (!signr)
2665 			signr = dequeue_signal(current, &current->blocked, &ksig->info);
2666 
2667 		if (!signr)
2668 			break; /* will return 0 */
2669 
2670 		if (unlikely(current->ptrace) && signr != SIGKILL) {
2671 			signr = ptrace_signal(signr, &ksig->info);
2672 			if (!signr)
2673 				continue;
2674 		}
2675 
2676 		ka = &sighand->action[signr-1];
2677 
2678 		/* Trace actually delivered signals. */
2679 		trace_signal_deliver(signr, &ksig->info, ka);
2680 
2681 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2682 			continue;
2683 		if (ka->sa.sa_handler != SIG_DFL) {
2684 			/* Run the handler.  */
2685 			ksig->ka = *ka;
2686 
2687 			if (ka->sa.sa_flags & SA_ONESHOT)
2688 				ka->sa.sa_handler = SIG_DFL;
2689 
2690 			break; /* will return non-zero "signr" value */
2691 		}
2692 
2693 		/*
2694 		 * Now we are doing the default action for this signal.
2695 		 */
2696 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2697 			continue;
2698 
2699 		/*
2700 		 * Global init gets no signals it doesn't want.
2701 		 * Container-init gets no signals it doesn't want from same
2702 		 * container.
2703 		 *
2704 		 * Note that if global/container-init sees a sig_kernel_only()
2705 		 * signal here, the signal must have been generated internally
2706 		 * or must have come from an ancestor namespace. In either
2707 		 * case, the signal cannot be dropped.
2708 		 */
2709 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2710 				!sig_kernel_only(signr))
2711 			continue;
2712 
2713 		if (sig_kernel_stop(signr)) {
2714 			/*
2715 			 * The default action is to stop all threads in
2716 			 * the thread group.  The job control signals
2717 			 * do nothing in an orphaned pgrp, but SIGSTOP
2718 			 * always works.  Note that siglock needs to be
2719 			 * dropped during the call to is_orphaned_pgrp()
2720 			 * because of lock ordering with tasklist_lock.
2721 			 * This allows an intervening SIGCONT to be posted.
2722 			 * We need to check for that and bail out if necessary.
2723 			 */
2724 			if (signr != SIGSTOP) {
2725 				spin_unlock_irq(&sighand->siglock);
2726 
2727 				/* signals can be posted during this window */
2728 
2729 				if (is_current_pgrp_orphaned())
2730 					goto relock;
2731 
2732 				spin_lock_irq(&sighand->siglock);
2733 			}
2734 
2735 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2736 				/* It released the siglock.  */
2737 				goto relock;
2738 			}
2739 
2740 			/*
2741 			 * We didn't actually stop, due to a race
2742 			 * with SIGCONT or something like that.
2743 			 */
2744 			continue;
2745 		}
2746 
2747 	fatal:
2748 		spin_unlock_irq(&sighand->siglock);
2749 		if (unlikely(cgroup_task_frozen(current)))
2750 			cgroup_leave_frozen(true);
2751 
2752 		/*
2753 		 * Anything else is fatal, maybe with a core dump.
2754 		 */
2755 		current->flags |= PF_SIGNALED;
2756 
2757 		if (sig_kernel_coredump(signr)) {
2758 			if (print_fatal_signals)
2759 				print_fatal_signal(ksig->info.si_signo);
2760 			proc_coredump_connector(current);
2761 			/*
2762 			 * If it was able to dump core, this kills all
2763 			 * other threads in the group and synchronizes with
2764 			 * their demise.  If we lost the race with another
2765 			 * thread getting here, it set group_exit_code
2766 			 * first and our do_group_exit call below will use
2767 			 * that value and ignore the one we pass it.
2768 			 */
2769 			do_coredump(&ksig->info);
2770 		}
2771 
2772 		/*
2773 		 * PF_IO_WORKER threads will catch and exit on fatal signals
2774 		 * themselves. They have cleanup that must be performed, so
2775 		 * we cannot call do_exit() on their behalf.
2776 		 */
2777 		if (current->flags & PF_IO_WORKER)
2778 			goto out;
2779 
2780 		/*
2781 		 * Death signals, no core dump.
2782 		 */
2783 		do_group_exit(ksig->info.si_signo);
2784 		/* NOTREACHED */
2785 	}
2786 	spin_unlock_irq(&sighand->siglock);
2787 out:
2788 	ksig->sig = signr;
2789 
2790 	if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2791 		hide_si_addr_tag_bits(ksig);
2792 
2793 	return ksig->sig > 0;
2794 }
2795 
2796 /**
2797  * signal_delivered -
2798  * @ksig:		kernel signal struct
2799  * @stepping:		nonzero if debugger single-step or block-step in use
2800  *
2801  * This function should be called when a signal has successfully been
2802  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2803  * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2804  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2805  */
signal_delivered(struct ksignal * ksig,int stepping)2806 static void signal_delivered(struct ksignal *ksig, int stepping)
2807 {
2808 	sigset_t blocked;
2809 
2810 	/* A signal was successfully delivered, and the
2811 	   saved sigmask was stored on the signal frame,
2812 	   and will be restored by sigreturn.  So we can
2813 	   simply clear the restore sigmask flag.  */
2814 	clear_restore_sigmask();
2815 
2816 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2817 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2818 		sigaddset(&blocked, ksig->sig);
2819 	set_current_blocked(&blocked);
2820 	tracehook_signal_handler(stepping);
2821 }
2822 
signal_setup_done(int failed,struct ksignal * ksig,int stepping)2823 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2824 {
2825 	if (failed)
2826 		force_sigsegv(ksig->sig);
2827 	else
2828 		signal_delivered(ksig, stepping);
2829 }
2830 
2831 /*
2832  * It could be that complete_signal() picked us to notify about the
2833  * group-wide signal. Other threads should be notified now to take
2834  * the shared signals in @which since we will not.
2835  */
retarget_shared_pending(struct task_struct * tsk,sigset_t * which)2836 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2837 {
2838 	sigset_t retarget;
2839 	struct task_struct *t;
2840 
2841 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2842 	if (sigisemptyset(&retarget))
2843 		return;
2844 
2845 	t = tsk;
2846 	while_each_thread(tsk, t) {
2847 		if (t->flags & PF_EXITING)
2848 			continue;
2849 
2850 		if (!has_pending_signals(&retarget, &t->blocked))
2851 			continue;
2852 		/* Remove the signals this thread can handle. */
2853 		sigandsets(&retarget, &retarget, &t->blocked);
2854 
2855 		if (!task_sigpending(t))
2856 			signal_wake_up(t, 0);
2857 
2858 		if (sigisemptyset(&retarget))
2859 			break;
2860 	}
2861 }
2862 
exit_signals(struct task_struct * tsk)2863 void exit_signals(struct task_struct *tsk)
2864 {
2865 	int group_stop = 0;
2866 	sigset_t unblocked;
2867 
2868 	/*
2869 	 * @tsk is about to have PF_EXITING set - lock out users which
2870 	 * expect stable threadgroup.
2871 	 */
2872 	cgroup_threadgroup_change_begin(tsk);
2873 
2874 	if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2875 		tsk->flags |= PF_EXITING;
2876 		cgroup_threadgroup_change_end(tsk);
2877 		return;
2878 	}
2879 
2880 	spin_lock_irq(&tsk->sighand->siglock);
2881 	/*
2882 	 * From now this task is not visible for group-wide signals,
2883 	 * see wants_signal(), do_signal_stop().
2884 	 */
2885 	tsk->flags |= PF_EXITING;
2886 
2887 	cgroup_threadgroup_change_end(tsk);
2888 
2889 	if (!task_sigpending(tsk))
2890 		goto out;
2891 
2892 	unblocked = tsk->blocked;
2893 	signotset(&unblocked);
2894 	retarget_shared_pending(tsk, &unblocked);
2895 
2896 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2897 	    task_participate_group_stop(tsk))
2898 		group_stop = CLD_STOPPED;
2899 out:
2900 	spin_unlock_irq(&tsk->sighand->siglock);
2901 
2902 	/*
2903 	 * If group stop has completed, deliver the notification.  This
2904 	 * should always go to the real parent of the group leader.
2905 	 */
2906 	if (unlikely(group_stop)) {
2907 		read_lock(&tasklist_lock);
2908 		do_notify_parent_cldstop(tsk, false, group_stop);
2909 		read_unlock(&tasklist_lock);
2910 	}
2911 }
2912 
2913 /*
2914  * System call entry points.
2915  */
2916 
2917 /**
2918  *  sys_restart_syscall - restart a system call
2919  */
SYSCALL_DEFINE0(restart_syscall)2920 SYSCALL_DEFINE0(restart_syscall)
2921 {
2922 	struct restart_block *restart = &current->restart_block;
2923 	return restart->fn(restart);
2924 }
2925 
do_no_restart_syscall(struct restart_block * param)2926 long do_no_restart_syscall(struct restart_block *param)
2927 {
2928 	return -EINTR;
2929 }
2930 
__set_task_blocked(struct task_struct * tsk,const sigset_t * newset)2931 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2932 {
2933 	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
2934 		sigset_t newblocked;
2935 		/* A set of now blocked but previously unblocked signals. */
2936 		sigandnsets(&newblocked, newset, &current->blocked);
2937 		retarget_shared_pending(tsk, &newblocked);
2938 	}
2939 	tsk->blocked = *newset;
2940 	recalc_sigpending();
2941 }
2942 
2943 /**
2944  * set_current_blocked - change current->blocked mask
2945  * @newset: new mask
2946  *
2947  * It is wrong to change ->blocked directly, this helper should be used
2948  * to ensure the process can't miss a shared signal we are going to block.
2949  */
set_current_blocked(sigset_t * newset)2950 void set_current_blocked(sigset_t *newset)
2951 {
2952 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2953 	__set_current_blocked(newset);
2954 }
2955 
__set_current_blocked(const sigset_t * newset)2956 void __set_current_blocked(const sigset_t *newset)
2957 {
2958 	struct task_struct *tsk = current;
2959 
2960 	/*
2961 	 * In case the signal mask hasn't changed, there is nothing we need
2962 	 * to do. The current->blocked shouldn't be modified by other task.
2963 	 */
2964 	if (sigequalsets(&tsk->blocked, newset))
2965 		return;
2966 
2967 	spin_lock_irq(&tsk->sighand->siglock);
2968 	__set_task_blocked(tsk, newset);
2969 	spin_unlock_irq(&tsk->sighand->siglock);
2970 }
2971 
2972 /*
2973  * This is also useful for kernel threads that want to temporarily
2974  * (or permanently) block certain signals.
2975  *
2976  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2977  * interface happily blocks "unblockable" signals like SIGKILL
2978  * and friends.
2979  */
sigprocmask(int how,sigset_t * set,sigset_t * oldset)2980 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2981 {
2982 	struct task_struct *tsk = current;
2983 	sigset_t newset;
2984 
2985 	/* Lockless, only current can change ->blocked, never from irq */
2986 	if (oldset)
2987 		*oldset = tsk->blocked;
2988 
2989 	switch (how) {
2990 	case SIG_BLOCK:
2991 		sigorsets(&newset, &tsk->blocked, set);
2992 		break;
2993 	case SIG_UNBLOCK:
2994 		sigandnsets(&newset, &tsk->blocked, set);
2995 		break;
2996 	case SIG_SETMASK:
2997 		newset = *set;
2998 		break;
2999 	default:
3000 		return -EINVAL;
3001 	}
3002 
3003 	__set_current_blocked(&newset);
3004 	return 0;
3005 }
3006 EXPORT_SYMBOL(sigprocmask);
3007 
3008 /*
3009  * The api helps set app-provided sigmasks.
3010  *
3011  * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3012  * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3013  *
3014  * Note that it does set_restore_sigmask() in advance, so it must be always
3015  * paired with restore_saved_sigmask_unless() before return from syscall.
3016  */
set_user_sigmask(const sigset_t __user * umask,size_t sigsetsize)3017 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3018 {
3019 	sigset_t kmask;
3020 
3021 	if (!umask)
3022 		return 0;
3023 	if (sigsetsize != sizeof(sigset_t))
3024 		return -EINVAL;
3025 	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3026 		return -EFAULT;
3027 
3028 	set_restore_sigmask();
3029 	current->saved_sigmask = current->blocked;
3030 	set_current_blocked(&kmask);
3031 
3032 	return 0;
3033 }
3034 
3035 #ifdef CONFIG_COMPAT
set_compat_user_sigmask(const compat_sigset_t __user * umask,size_t sigsetsize)3036 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3037 			    size_t sigsetsize)
3038 {
3039 	sigset_t kmask;
3040 
3041 	if (!umask)
3042 		return 0;
3043 	if (sigsetsize != sizeof(compat_sigset_t))
3044 		return -EINVAL;
3045 	if (get_compat_sigset(&kmask, umask))
3046 		return -EFAULT;
3047 
3048 	set_restore_sigmask();
3049 	current->saved_sigmask = current->blocked;
3050 	set_current_blocked(&kmask);
3051 
3052 	return 0;
3053 }
3054 #endif
3055 
3056 /**
3057  *  sys_rt_sigprocmask - change the list of currently blocked signals
3058  *  @how: whether to add, remove, or set signals
3059  *  @nset: stores pending signals
3060  *  @oset: previous value of signal mask if non-null
3061  *  @sigsetsize: size of sigset_t type
3062  */
SYSCALL_DEFINE4(rt_sigprocmask,int,how,sigset_t __user *,nset,sigset_t __user *,oset,size_t,sigsetsize)3063 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3064 		sigset_t __user *, oset, size_t, sigsetsize)
3065 {
3066 	sigset_t old_set, new_set;
3067 	int error;
3068 
3069 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3070 	if (sigsetsize != sizeof(sigset_t))
3071 		return -EINVAL;
3072 
3073 	old_set = current->blocked;
3074 
3075 	if (nset) {
3076 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3077 			return -EFAULT;
3078 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3079 
3080 		error = sigprocmask(how, &new_set, NULL);
3081 		if (error)
3082 			return error;
3083 	}
3084 
3085 	if (oset) {
3086 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3087 			return -EFAULT;
3088 	}
3089 
3090 	return 0;
3091 }
3092 
3093 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_sigprocmask,int,how,compat_sigset_t __user *,nset,compat_sigset_t __user *,oset,compat_size_t,sigsetsize)3094 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3095 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3096 {
3097 	sigset_t old_set = current->blocked;
3098 
3099 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3100 	if (sigsetsize != sizeof(sigset_t))
3101 		return -EINVAL;
3102 
3103 	if (nset) {
3104 		sigset_t new_set;
3105 		int error;
3106 		if (get_compat_sigset(&new_set, nset))
3107 			return -EFAULT;
3108 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3109 
3110 		error = sigprocmask(how, &new_set, NULL);
3111 		if (error)
3112 			return error;
3113 	}
3114 	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3115 }
3116 #endif
3117 
do_sigpending(sigset_t * set)3118 static void do_sigpending(sigset_t *set)
3119 {
3120 	spin_lock_irq(&current->sighand->siglock);
3121 	sigorsets(set, &current->pending.signal,
3122 		  &current->signal->shared_pending.signal);
3123 	spin_unlock_irq(&current->sighand->siglock);
3124 
3125 	/* Outside the lock because only this thread touches it.  */
3126 	sigandsets(set, &current->blocked, set);
3127 }
3128 
3129 /**
3130  *  sys_rt_sigpending - examine a pending signal that has been raised
3131  *			while blocked
3132  *  @uset: stores pending signals
3133  *  @sigsetsize: size of sigset_t type or larger
3134  */
SYSCALL_DEFINE2(rt_sigpending,sigset_t __user *,uset,size_t,sigsetsize)3135 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3136 {
3137 	sigset_t set;
3138 
3139 	if (sigsetsize > sizeof(*uset))
3140 		return -EINVAL;
3141 
3142 	do_sigpending(&set);
3143 
3144 	if (copy_to_user(uset, &set, sigsetsize))
3145 		return -EFAULT;
3146 
3147 	return 0;
3148 }
3149 
3150 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(rt_sigpending,compat_sigset_t __user *,uset,compat_size_t,sigsetsize)3151 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3152 		compat_size_t, sigsetsize)
3153 {
3154 	sigset_t set;
3155 
3156 	if (sigsetsize > sizeof(*uset))
3157 		return -EINVAL;
3158 
3159 	do_sigpending(&set);
3160 
3161 	return put_compat_sigset(uset, &set, sigsetsize);
3162 }
3163 #endif
3164 
3165 static const struct {
3166 	unsigned char limit, layout;
3167 } sig_sicodes[] = {
3168 	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3169 	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3170 	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3171 	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3172 	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3173 #if defined(SIGEMT)
3174 	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3175 #endif
3176 	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3177 	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3178 	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3179 };
3180 
known_siginfo_layout(unsigned sig,int si_code)3181 static bool known_siginfo_layout(unsigned sig, int si_code)
3182 {
3183 	if (si_code == SI_KERNEL)
3184 		return true;
3185 	else if ((si_code > SI_USER)) {
3186 		if (sig_specific_sicodes(sig)) {
3187 			if (si_code <= sig_sicodes[sig].limit)
3188 				return true;
3189 		}
3190 		else if (si_code <= NSIGPOLL)
3191 			return true;
3192 	}
3193 	else if (si_code >= SI_DETHREAD)
3194 		return true;
3195 	else if (si_code == SI_ASYNCNL)
3196 		return true;
3197 	return false;
3198 }
3199 
siginfo_layout(unsigned sig,int si_code)3200 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3201 {
3202 	enum siginfo_layout layout = SIL_KILL;
3203 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3204 		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3205 		    (si_code <= sig_sicodes[sig].limit)) {
3206 			layout = sig_sicodes[sig].layout;
3207 			/* Handle the exceptions */
3208 			if ((sig == SIGBUS) &&
3209 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3210 				layout = SIL_FAULT_MCEERR;
3211 			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3212 				layout = SIL_FAULT_BNDERR;
3213 #ifdef SEGV_PKUERR
3214 			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3215 				layout = SIL_FAULT_PKUERR;
3216 #endif
3217 		}
3218 		else if (si_code <= NSIGPOLL)
3219 			layout = SIL_POLL;
3220 	} else {
3221 		if (si_code == SI_TIMER)
3222 			layout = SIL_TIMER;
3223 		else if (si_code == SI_SIGIO)
3224 			layout = SIL_POLL;
3225 		else if (si_code < 0)
3226 			layout = SIL_RT;
3227 	}
3228 	return layout;
3229 }
3230 
si_expansion(const siginfo_t __user * info)3231 static inline char __user *si_expansion(const siginfo_t __user *info)
3232 {
3233 	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3234 }
3235 
copy_siginfo_to_user(siginfo_t __user * to,const kernel_siginfo_t * from)3236 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3237 {
3238 	char __user *expansion = si_expansion(to);
3239 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3240 		return -EFAULT;
3241 	if (clear_user(expansion, SI_EXPANSION_SIZE))
3242 		return -EFAULT;
3243 	return 0;
3244 }
3245 
post_copy_siginfo_from_user(kernel_siginfo_t * info,const siginfo_t __user * from)3246 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3247 				       const siginfo_t __user *from)
3248 {
3249 	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3250 		char __user *expansion = si_expansion(from);
3251 		char buf[SI_EXPANSION_SIZE];
3252 		int i;
3253 		/*
3254 		 * An unknown si_code might need more than
3255 		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3256 		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3257 		 * will return this data to userspace exactly.
3258 		 */
3259 		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3260 			return -EFAULT;
3261 		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3262 			if (buf[i] != 0)
3263 				return -E2BIG;
3264 		}
3265 	}
3266 	return 0;
3267 }
3268 
__copy_siginfo_from_user(int signo,kernel_siginfo_t * to,const siginfo_t __user * from)3269 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3270 				    const siginfo_t __user *from)
3271 {
3272 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3273 		return -EFAULT;
3274 	to->si_signo = signo;
3275 	return post_copy_siginfo_from_user(to, from);
3276 }
3277 
copy_siginfo_from_user(kernel_siginfo_t * to,const siginfo_t __user * from)3278 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3279 {
3280 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3281 		return -EFAULT;
3282 	return post_copy_siginfo_from_user(to, from);
3283 }
3284 
3285 #ifdef CONFIG_COMPAT
3286 /**
3287  * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3288  * @to: compat siginfo destination
3289  * @from: kernel siginfo source
3290  *
3291  * Note: This function does not work properly for the SIGCHLD on x32, but
3292  * fortunately it doesn't have to.  The only valid callers for this function are
3293  * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3294  * The latter does not care because SIGCHLD will never cause a coredump.
3295  */
copy_siginfo_to_external32(struct compat_siginfo * to,const struct kernel_siginfo * from)3296 void copy_siginfo_to_external32(struct compat_siginfo *to,
3297 		const struct kernel_siginfo *from)
3298 {
3299 	memset(to, 0, sizeof(*to));
3300 
3301 	to->si_signo = from->si_signo;
3302 	to->si_errno = from->si_errno;
3303 	to->si_code  = from->si_code;
3304 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3305 	case SIL_KILL:
3306 		to->si_pid = from->si_pid;
3307 		to->si_uid = from->si_uid;
3308 		break;
3309 	case SIL_TIMER:
3310 		to->si_tid     = from->si_tid;
3311 		to->si_overrun = from->si_overrun;
3312 		to->si_int     = from->si_int;
3313 		break;
3314 	case SIL_POLL:
3315 		to->si_band = from->si_band;
3316 		to->si_fd   = from->si_fd;
3317 		break;
3318 	case SIL_FAULT:
3319 		to->si_addr = ptr_to_compat(from->si_addr);
3320 #ifdef __ARCH_SI_TRAPNO
3321 		to->si_trapno = from->si_trapno;
3322 #endif
3323 		break;
3324 	case SIL_FAULT_MCEERR:
3325 		to->si_addr = ptr_to_compat(from->si_addr);
3326 #ifdef __ARCH_SI_TRAPNO
3327 		to->si_trapno = from->si_trapno;
3328 #endif
3329 		to->si_addr_lsb = from->si_addr_lsb;
3330 		break;
3331 	case SIL_FAULT_BNDERR:
3332 		to->si_addr = ptr_to_compat(from->si_addr);
3333 #ifdef __ARCH_SI_TRAPNO
3334 		to->si_trapno = from->si_trapno;
3335 #endif
3336 		to->si_lower = ptr_to_compat(from->si_lower);
3337 		to->si_upper = ptr_to_compat(from->si_upper);
3338 		break;
3339 	case SIL_FAULT_PKUERR:
3340 		to->si_addr = ptr_to_compat(from->si_addr);
3341 #ifdef __ARCH_SI_TRAPNO
3342 		to->si_trapno = from->si_trapno;
3343 #endif
3344 		to->si_pkey = from->si_pkey;
3345 		break;
3346 	case SIL_CHLD:
3347 		to->si_pid = from->si_pid;
3348 		to->si_uid = from->si_uid;
3349 		to->si_status = from->si_status;
3350 		to->si_utime = from->si_utime;
3351 		to->si_stime = from->si_stime;
3352 		break;
3353 	case SIL_RT:
3354 		to->si_pid = from->si_pid;
3355 		to->si_uid = from->si_uid;
3356 		to->si_int = from->si_int;
3357 		break;
3358 	case SIL_SYS:
3359 		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3360 		to->si_syscall   = from->si_syscall;
3361 		to->si_arch      = from->si_arch;
3362 		break;
3363 	}
3364 }
3365 
__copy_siginfo_to_user32(struct compat_siginfo __user * to,const struct kernel_siginfo * from)3366 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3367 			   const struct kernel_siginfo *from)
3368 {
3369 	struct compat_siginfo new;
3370 
3371 	copy_siginfo_to_external32(&new, from);
3372 	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3373 		return -EFAULT;
3374 	return 0;
3375 }
3376 
post_copy_siginfo_from_user32(kernel_siginfo_t * to,const struct compat_siginfo * from)3377 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3378 					 const struct compat_siginfo *from)
3379 {
3380 	clear_siginfo(to);
3381 	to->si_signo = from->si_signo;
3382 	to->si_errno = from->si_errno;
3383 	to->si_code  = from->si_code;
3384 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3385 	case SIL_KILL:
3386 		to->si_pid = from->si_pid;
3387 		to->si_uid = from->si_uid;
3388 		break;
3389 	case SIL_TIMER:
3390 		to->si_tid     = from->si_tid;
3391 		to->si_overrun = from->si_overrun;
3392 		to->si_int     = from->si_int;
3393 		break;
3394 	case SIL_POLL:
3395 		to->si_band = from->si_band;
3396 		to->si_fd   = from->si_fd;
3397 		break;
3398 	case SIL_FAULT:
3399 		to->si_addr = compat_ptr(from->si_addr);
3400 #ifdef __ARCH_SI_TRAPNO
3401 		to->si_trapno = from->si_trapno;
3402 #endif
3403 		break;
3404 	case SIL_FAULT_MCEERR:
3405 		to->si_addr = compat_ptr(from->si_addr);
3406 #ifdef __ARCH_SI_TRAPNO
3407 		to->si_trapno = from->si_trapno;
3408 #endif
3409 		to->si_addr_lsb = from->si_addr_lsb;
3410 		break;
3411 	case SIL_FAULT_BNDERR:
3412 		to->si_addr = compat_ptr(from->si_addr);
3413 #ifdef __ARCH_SI_TRAPNO
3414 		to->si_trapno = from->si_trapno;
3415 #endif
3416 		to->si_lower = compat_ptr(from->si_lower);
3417 		to->si_upper = compat_ptr(from->si_upper);
3418 		break;
3419 	case SIL_FAULT_PKUERR:
3420 		to->si_addr = compat_ptr(from->si_addr);
3421 #ifdef __ARCH_SI_TRAPNO
3422 		to->si_trapno = from->si_trapno;
3423 #endif
3424 		to->si_pkey = from->si_pkey;
3425 		break;
3426 	case SIL_CHLD:
3427 		to->si_pid    = from->si_pid;
3428 		to->si_uid    = from->si_uid;
3429 		to->si_status = from->si_status;
3430 #ifdef CONFIG_X86_X32_ABI
3431 		if (in_x32_syscall()) {
3432 			to->si_utime = from->_sifields._sigchld_x32._utime;
3433 			to->si_stime = from->_sifields._sigchld_x32._stime;
3434 		} else
3435 #endif
3436 		{
3437 			to->si_utime = from->si_utime;
3438 			to->si_stime = from->si_stime;
3439 		}
3440 		break;
3441 	case SIL_RT:
3442 		to->si_pid = from->si_pid;
3443 		to->si_uid = from->si_uid;
3444 		to->si_int = from->si_int;
3445 		break;
3446 	case SIL_SYS:
3447 		to->si_call_addr = compat_ptr(from->si_call_addr);
3448 		to->si_syscall   = from->si_syscall;
3449 		to->si_arch      = from->si_arch;
3450 		break;
3451 	}
3452 	return 0;
3453 }
3454 
__copy_siginfo_from_user32(int signo,struct kernel_siginfo * to,const struct compat_siginfo __user * ufrom)3455 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3456 				      const struct compat_siginfo __user *ufrom)
3457 {
3458 	struct compat_siginfo from;
3459 
3460 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3461 		return -EFAULT;
3462 
3463 	from.si_signo = signo;
3464 	return post_copy_siginfo_from_user32(to, &from);
3465 }
3466 
copy_siginfo_from_user32(struct kernel_siginfo * to,const struct compat_siginfo __user * ufrom)3467 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3468 			     const struct compat_siginfo __user *ufrom)
3469 {
3470 	struct compat_siginfo from;
3471 
3472 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3473 		return -EFAULT;
3474 
3475 	return post_copy_siginfo_from_user32(to, &from);
3476 }
3477 #endif /* CONFIG_COMPAT */
3478 
3479 /**
3480  *  do_sigtimedwait - wait for queued signals specified in @which
3481  *  @which: queued signals to wait for
3482  *  @info: if non-null, the signal's siginfo is returned here
3483  *  @ts: upper bound on process time suspension
3484  */
do_sigtimedwait(const sigset_t * which,kernel_siginfo_t * info,const struct timespec64 * ts)3485 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3486 		    const struct timespec64 *ts)
3487 {
3488 	ktime_t *to = NULL, timeout = KTIME_MAX;
3489 	struct task_struct *tsk = current;
3490 	sigset_t mask = *which;
3491 	int sig, ret = 0;
3492 
3493 	if (ts) {
3494 		if (!timespec64_valid(ts))
3495 			return -EINVAL;
3496 		timeout = timespec64_to_ktime(*ts);
3497 		to = &timeout;
3498 	}
3499 
3500 	/*
3501 	 * Invert the set of allowed signals to get those we want to block.
3502 	 */
3503 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3504 	signotset(&mask);
3505 
3506 	spin_lock_irq(&tsk->sighand->siglock);
3507 	sig = dequeue_signal(tsk, &mask, info);
3508 	if (!sig && timeout) {
3509 		/*
3510 		 * None ready, temporarily unblock those we're interested
3511 		 * while we are sleeping in so that we'll be awakened when
3512 		 * they arrive. Unblocking is always fine, we can avoid
3513 		 * set_current_blocked().
3514 		 */
3515 		tsk->real_blocked = tsk->blocked;
3516 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3517 		recalc_sigpending();
3518 		spin_unlock_irq(&tsk->sighand->siglock);
3519 
3520 		__set_current_state(TASK_INTERRUPTIBLE);
3521 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3522 							 HRTIMER_MODE_REL);
3523 		spin_lock_irq(&tsk->sighand->siglock);
3524 		__set_task_blocked(tsk, &tsk->real_blocked);
3525 		sigemptyset(&tsk->real_blocked);
3526 		sig = dequeue_signal(tsk, &mask, info);
3527 	}
3528 	spin_unlock_irq(&tsk->sighand->siglock);
3529 
3530 	if (sig)
3531 		return sig;
3532 	return ret ? -EINTR : -EAGAIN;
3533 }
3534 
3535 /**
3536  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3537  *			in @uthese
3538  *  @uthese: queued signals to wait for
3539  *  @uinfo: if non-null, the signal's siginfo is returned here
3540  *  @uts: upper bound on process time suspension
3541  *  @sigsetsize: size of sigset_t type
3542  */
SYSCALL_DEFINE4(rt_sigtimedwait,const sigset_t __user *,uthese,siginfo_t __user *,uinfo,const struct __kernel_timespec __user *,uts,size_t,sigsetsize)3543 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3544 		siginfo_t __user *, uinfo,
3545 		const struct __kernel_timespec __user *, uts,
3546 		size_t, sigsetsize)
3547 {
3548 	sigset_t these;
3549 	struct timespec64 ts;
3550 	kernel_siginfo_t info;
3551 	int ret;
3552 
3553 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3554 	if (sigsetsize != sizeof(sigset_t))
3555 		return -EINVAL;
3556 
3557 	if (copy_from_user(&these, uthese, sizeof(these)))
3558 		return -EFAULT;
3559 
3560 	if (uts) {
3561 		if (get_timespec64(&ts, uts))
3562 			return -EFAULT;
3563 	}
3564 
3565 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3566 
3567 	if (ret > 0 && uinfo) {
3568 		if (copy_siginfo_to_user(uinfo, &info))
3569 			ret = -EFAULT;
3570 	}
3571 
3572 	return ret;
3573 }
3574 
3575 #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)3576 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3577 		siginfo_t __user *, uinfo,
3578 		const struct old_timespec32 __user *, uts,
3579 		size_t, sigsetsize)
3580 {
3581 	sigset_t these;
3582 	struct timespec64 ts;
3583 	kernel_siginfo_t info;
3584 	int ret;
3585 
3586 	if (sigsetsize != sizeof(sigset_t))
3587 		return -EINVAL;
3588 
3589 	if (copy_from_user(&these, uthese, sizeof(these)))
3590 		return -EFAULT;
3591 
3592 	if (uts) {
3593 		if (get_old_timespec32(&ts, uts))
3594 			return -EFAULT;
3595 	}
3596 
3597 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3598 
3599 	if (ret > 0 && uinfo) {
3600 		if (copy_siginfo_to_user(uinfo, &info))
3601 			ret = -EFAULT;
3602 	}
3603 
3604 	return ret;
3605 }
3606 #endif
3607 
3608 #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)3609 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3610 		struct compat_siginfo __user *, uinfo,
3611 		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3612 {
3613 	sigset_t s;
3614 	struct timespec64 t;
3615 	kernel_siginfo_t info;
3616 	long ret;
3617 
3618 	if (sigsetsize != sizeof(sigset_t))
3619 		return -EINVAL;
3620 
3621 	if (get_compat_sigset(&s, uthese))
3622 		return -EFAULT;
3623 
3624 	if (uts) {
3625 		if (get_timespec64(&t, uts))
3626 			return -EFAULT;
3627 	}
3628 
3629 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3630 
3631 	if (ret > 0 && uinfo) {
3632 		if (copy_siginfo_to_user32(uinfo, &info))
3633 			ret = -EFAULT;
3634 	}
3635 
3636 	return ret;
3637 }
3638 
3639 #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)3640 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3641 		struct compat_siginfo __user *, uinfo,
3642 		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3643 {
3644 	sigset_t s;
3645 	struct timespec64 t;
3646 	kernel_siginfo_t info;
3647 	long ret;
3648 
3649 	if (sigsetsize != sizeof(sigset_t))
3650 		return -EINVAL;
3651 
3652 	if (get_compat_sigset(&s, uthese))
3653 		return -EFAULT;
3654 
3655 	if (uts) {
3656 		if (get_old_timespec32(&t, uts))
3657 			return -EFAULT;
3658 	}
3659 
3660 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3661 
3662 	if (ret > 0 && uinfo) {
3663 		if (copy_siginfo_to_user32(uinfo, &info))
3664 			ret = -EFAULT;
3665 	}
3666 
3667 	return ret;
3668 }
3669 #endif
3670 #endif
3671 
prepare_kill_siginfo(int sig,struct kernel_siginfo * info)3672 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3673 {
3674 	clear_siginfo(info);
3675 	info->si_signo = sig;
3676 	info->si_errno = 0;
3677 	info->si_code = SI_USER;
3678 	info->si_pid = task_tgid_vnr(current);
3679 	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3680 }
3681 
3682 /**
3683  *  sys_kill - send a signal to a process
3684  *  @pid: the PID of the process
3685  *  @sig: signal to be sent
3686  */
SYSCALL_DEFINE2(kill,pid_t,pid,int,sig)3687 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3688 {
3689 	struct kernel_siginfo info;
3690 
3691 	prepare_kill_siginfo(sig, &info);
3692 
3693 	return kill_something_info(sig, &info, pid);
3694 }
3695 
3696 /*
3697  * Verify that the signaler and signalee either are in the same pid namespace
3698  * or that the signaler's pid namespace is an ancestor of the signalee's pid
3699  * namespace.
3700  */
access_pidfd_pidns(struct pid * pid)3701 static bool access_pidfd_pidns(struct pid *pid)
3702 {
3703 	struct pid_namespace *active = task_active_pid_ns(current);
3704 	struct pid_namespace *p = ns_of_pid(pid);
3705 
3706 	for (;;) {
3707 		if (!p)
3708 			return false;
3709 		if (p == active)
3710 			break;
3711 		p = p->parent;
3712 	}
3713 
3714 	return true;
3715 }
3716 
copy_siginfo_from_user_any(kernel_siginfo_t * kinfo,siginfo_t * info)3717 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo, siginfo_t *info)
3718 {
3719 #ifdef CONFIG_COMPAT
3720 	/*
3721 	 * Avoid hooking up compat syscalls and instead handle necessary
3722 	 * conversions here. Note, this is a stop-gap measure and should not be
3723 	 * considered a generic solution.
3724 	 */
3725 	if (in_compat_syscall())
3726 		return copy_siginfo_from_user32(
3727 			kinfo, (struct compat_siginfo __user *)info);
3728 #endif
3729 	return copy_siginfo_from_user(kinfo, info);
3730 }
3731 
pidfd_to_pid(const struct file * file)3732 static struct pid *pidfd_to_pid(const struct file *file)
3733 {
3734 	struct pid *pid;
3735 
3736 	pid = pidfd_pid(file);
3737 	if (!IS_ERR(pid))
3738 		return pid;
3739 
3740 	return tgid_pidfd_to_pid(file);
3741 }
3742 
3743 /**
3744  * sys_pidfd_send_signal - Signal a process through a pidfd
3745  * @pidfd:  file descriptor of the process
3746  * @sig:    signal to send
3747  * @info:   signal info
3748  * @flags:  future flags
3749  *
3750  * The syscall currently only signals via PIDTYPE_PID which covers
3751  * kill(<positive-pid>, <signal>. It does not signal threads or process
3752  * groups.
3753  * In order to extend the syscall to threads and process groups the @flags
3754  * argument should be used. In essence, the @flags argument will determine
3755  * what is signaled and not the file descriptor itself. Put in other words,
3756  * grouping is a property of the flags argument not a property of the file
3757  * descriptor.
3758  *
3759  * Return: 0 on success, negative errno on failure
3760  */
SYSCALL_DEFINE4(pidfd_send_signal,int,pidfd,int,sig,siginfo_t __user *,info,unsigned int,flags)3761 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3762 		siginfo_t __user *, info, unsigned int, flags)
3763 {
3764 	int ret;
3765 	struct fd f;
3766 	struct pid *pid;
3767 	kernel_siginfo_t kinfo;
3768 
3769 	/* Enforce flags be set to 0 until we add an extension. */
3770 	if (flags)
3771 		return -EINVAL;
3772 
3773 	f = fdget(pidfd);
3774 	if (!f.file)
3775 		return -EBADF;
3776 
3777 	/* Is this a pidfd? */
3778 	pid = pidfd_to_pid(f.file);
3779 	if (IS_ERR(pid)) {
3780 		ret = PTR_ERR(pid);
3781 		goto err;
3782 	}
3783 
3784 	ret = -EINVAL;
3785 	if (!access_pidfd_pidns(pid))
3786 		goto err;
3787 
3788 	if (info) {
3789 		ret = copy_siginfo_from_user_any(&kinfo, info);
3790 		if (unlikely(ret))
3791 			goto err;
3792 
3793 		ret = -EINVAL;
3794 		if (unlikely(sig != kinfo.si_signo))
3795 			goto err;
3796 
3797 		/* Only allow sending arbitrary signals to yourself. */
3798 		ret = -EPERM;
3799 		if ((task_pid(current) != pid) &&
3800 		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3801 			goto err;
3802 	} else {
3803 		prepare_kill_siginfo(sig, &kinfo);
3804 	}
3805 
3806 	ret = kill_pid_info(sig, &kinfo, pid);
3807 
3808 err:
3809 	fdput(f);
3810 	return ret;
3811 }
3812 
3813 static int
do_send_specific(pid_t tgid,pid_t pid,int sig,struct kernel_siginfo * info)3814 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3815 {
3816 	struct task_struct *p;
3817 	int error = -ESRCH;
3818 
3819 	rcu_read_lock();
3820 	p = find_task_by_vpid(pid);
3821 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3822 		error = check_kill_permission(sig, info, p);
3823 		/*
3824 		 * The null signal is a permissions and process existence
3825 		 * probe.  No signal is actually delivered.
3826 		 */
3827 		if (!error && sig) {
3828 			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3829 			/*
3830 			 * If lock_task_sighand() failed we pretend the task
3831 			 * dies after receiving the signal. The window is tiny,
3832 			 * and the signal is private anyway.
3833 			 */
3834 			if (unlikely(error == -ESRCH))
3835 				error = 0;
3836 		}
3837 	}
3838 	rcu_read_unlock();
3839 
3840 	return error;
3841 }
3842 
do_tkill(pid_t tgid,pid_t pid,int sig)3843 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3844 {
3845 	struct kernel_siginfo info;
3846 
3847 	clear_siginfo(&info);
3848 	info.si_signo = sig;
3849 	info.si_errno = 0;
3850 	info.si_code = SI_TKILL;
3851 	info.si_pid = task_tgid_vnr(current);
3852 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3853 
3854 	return do_send_specific(tgid, pid, sig, &info);
3855 }
3856 
3857 /**
3858  *  sys_tgkill - send signal to one specific thread
3859  *  @tgid: the thread group ID of the thread
3860  *  @pid: the PID of the thread
3861  *  @sig: signal to be sent
3862  *
3863  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3864  *  exists but it's not belonging to the target process anymore. This
3865  *  method solves the problem of threads exiting and PIDs getting reused.
3866  */
SYSCALL_DEFINE3(tgkill,pid_t,tgid,pid_t,pid,int,sig)3867 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3868 {
3869 	/* This is only valid for single tasks */
3870 	if (pid <= 0 || tgid <= 0)
3871 		return -EINVAL;
3872 
3873 	return do_tkill(tgid, pid, sig);
3874 }
3875 
3876 /**
3877  *  sys_tkill - send signal to one specific task
3878  *  @pid: the PID of the task
3879  *  @sig: signal to be sent
3880  *
3881  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3882  */
SYSCALL_DEFINE2(tkill,pid_t,pid,int,sig)3883 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3884 {
3885 	/* This is only valid for single tasks */
3886 	if (pid <= 0)
3887 		return -EINVAL;
3888 
3889 	return do_tkill(0, pid, sig);
3890 }
3891 
do_rt_sigqueueinfo(pid_t pid,int sig,kernel_siginfo_t * info)3892 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
3893 {
3894 	/* Not even root can pretend to send signals from the kernel.
3895 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3896 	 */
3897 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3898 	    (task_pid_vnr(current) != pid))
3899 		return -EPERM;
3900 
3901 	/* POSIX.1b doesn't mention process groups.  */
3902 	return kill_proc_info(sig, info, pid);
3903 }
3904 
3905 /**
3906  *  sys_rt_sigqueueinfo - send signal information to a signal
3907  *  @pid: the PID of the thread
3908  *  @sig: signal to be sent
3909  *  @uinfo: signal info to be sent
3910  */
SYSCALL_DEFINE3(rt_sigqueueinfo,pid_t,pid,int,sig,siginfo_t __user *,uinfo)3911 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
3912 		siginfo_t __user *, uinfo)
3913 {
3914 	kernel_siginfo_t info;
3915 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
3916 	if (unlikely(ret))
3917 		return ret;
3918 	return do_rt_sigqueueinfo(pid, sig, &info);
3919 }
3920 
3921 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,compat_pid_t,pid,int,sig,struct compat_siginfo __user *,uinfo)3922 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
3923 			compat_pid_t, pid,
3924 			int, sig,
3925 			struct compat_siginfo __user *, uinfo)
3926 {
3927 	kernel_siginfo_t info;
3928 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
3929 	if (unlikely(ret))
3930 		return ret;
3931 	return do_rt_sigqueueinfo(pid, sig, &info);
3932 }
3933 #endif
3934 
do_rt_tgsigqueueinfo(pid_t tgid,pid_t pid,int sig,kernel_siginfo_t * info)3935 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
3936 {
3937 	/* This is only valid for single tasks */
3938 	if (pid <= 0 || tgid <= 0)
3939 		return -EINVAL;
3940 
3941 	/* Not even root can pretend to send signals from the kernel.
3942 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3943 	 */
3944 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3945 	    (task_pid_vnr(current) != pid))
3946 		return -EPERM;
3947 
3948 	return do_send_specific(tgid, pid, sig, info);
3949 }
3950 
SYSCALL_DEFINE4(rt_tgsigqueueinfo,pid_t,tgid,pid_t,pid,int,sig,siginfo_t __user *,uinfo)3951 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
3952 		siginfo_t __user *, uinfo)
3953 {
3954 	kernel_siginfo_t info;
3955 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
3956 	if (unlikely(ret))
3957 		return ret;
3958 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3959 }
3960 
3961 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,compat_pid_t,tgid,compat_pid_t,pid,int,sig,struct compat_siginfo __user *,uinfo)3962 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
3963 			compat_pid_t, tgid,
3964 			compat_pid_t, pid,
3965 			int, sig,
3966 			struct compat_siginfo __user *, uinfo)
3967 {
3968 	kernel_siginfo_t info;
3969 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
3970 	if (unlikely(ret))
3971 		return ret;
3972 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3973 }
3974 #endif
3975 
3976 /*
3977  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
3978  */
kernel_sigaction(int sig,__sighandler_t action)3979 void kernel_sigaction(int sig, __sighandler_t action)
3980 {
3981 	spin_lock_irq(&current->sighand->siglock);
3982 	current->sighand->action[sig - 1].sa.sa_handler = action;
3983 	if (action == SIG_IGN) {
3984 		sigset_t mask;
3985 
3986 		sigemptyset(&mask);
3987 		sigaddset(&mask, sig);
3988 
3989 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
3990 		flush_sigqueue_mask(&mask, &current->pending);
3991 		recalc_sigpending();
3992 	}
3993 	spin_unlock_irq(&current->sighand->siglock);
3994 }
3995 EXPORT_SYMBOL(kernel_sigaction);
3996 
sigaction_compat_abi(struct k_sigaction * act,struct k_sigaction * oact)3997 void __weak sigaction_compat_abi(struct k_sigaction *act,
3998 		struct k_sigaction *oact)
3999 {
4000 }
4001 
do_sigaction(int sig,struct k_sigaction * act,struct k_sigaction * oact)4002 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4003 {
4004 	struct task_struct *p = current, *t;
4005 	struct k_sigaction *k;
4006 	sigset_t mask;
4007 
4008 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4009 		return -EINVAL;
4010 
4011 	k = &p->sighand->action[sig-1];
4012 
4013 	spin_lock_irq(&p->sighand->siglock);
4014 	if (oact)
4015 		*oact = *k;
4016 
4017 	/*
4018 	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4019 	 * e.g. by having an architecture use the bit in their uapi.
4020 	 */
4021 	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4022 
4023 	/*
4024 	 * Clear unknown flag bits in order to allow userspace to detect missing
4025 	 * support for flag bits and to allow the kernel to use non-uapi bits
4026 	 * internally.
4027 	 */
4028 	if (act)
4029 		act->sa.sa_flags &= UAPI_SA_FLAGS;
4030 	if (oact)
4031 		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4032 
4033 	sigaction_compat_abi(act, oact);
4034 
4035 	if (act) {
4036 		sigdelsetmask(&act->sa.sa_mask,
4037 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4038 		*k = *act;
4039 		/*
4040 		 * POSIX 3.3.1.3:
4041 		 *  "Setting a signal action to SIG_IGN for a signal that is
4042 		 *   pending shall cause the pending signal to be discarded,
4043 		 *   whether or not it is blocked."
4044 		 *
4045 		 *  "Setting a signal action to SIG_DFL for a signal that is
4046 		 *   pending and whose default action is to ignore the signal
4047 		 *   (for example, SIGCHLD), shall cause the pending signal to
4048 		 *   be discarded, whether or not it is blocked"
4049 		 */
4050 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4051 			sigemptyset(&mask);
4052 			sigaddset(&mask, sig);
4053 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4054 			for_each_thread(p, t)
4055 				flush_sigqueue_mask(&mask, &t->pending);
4056 		}
4057 	}
4058 
4059 	spin_unlock_irq(&p->sighand->siglock);
4060 	return 0;
4061 }
4062 
4063 static int
do_sigaltstack(const stack_t * ss,stack_t * oss,unsigned long sp,size_t min_ss_size)4064 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4065 		size_t min_ss_size)
4066 {
4067 	struct task_struct *t = current;
4068 
4069 	if (oss) {
4070 		memset(oss, 0, sizeof(stack_t));
4071 		oss->ss_sp = (void __user *) t->sas_ss_sp;
4072 		oss->ss_size = t->sas_ss_size;
4073 		oss->ss_flags = sas_ss_flags(sp) |
4074 			(current->sas_ss_flags & SS_FLAG_BITS);
4075 	}
4076 
4077 	if (ss) {
4078 		void __user *ss_sp = ss->ss_sp;
4079 		size_t ss_size = ss->ss_size;
4080 		unsigned ss_flags = ss->ss_flags;
4081 		int ss_mode;
4082 
4083 		if (unlikely(on_sig_stack(sp)))
4084 			return -EPERM;
4085 
4086 		ss_mode = ss_flags & ~SS_FLAG_BITS;
4087 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4088 				ss_mode != 0))
4089 			return -EINVAL;
4090 
4091 		if (ss_mode == SS_DISABLE) {
4092 			ss_size = 0;
4093 			ss_sp = NULL;
4094 		} else {
4095 			if (unlikely(ss_size < min_ss_size))
4096 				return -ENOMEM;
4097 		}
4098 
4099 		t->sas_ss_sp = (unsigned long) ss_sp;
4100 		t->sas_ss_size = ss_size;
4101 		t->sas_ss_flags = ss_flags;
4102 	}
4103 	return 0;
4104 }
4105 
SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss,stack_t __user *,uoss)4106 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4107 {
4108 	stack_t new, old;
4109 	int err;
4110 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4111 		return -EFAULT;
4112 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4113 			      current_user_stack_pointer(),
4114 			      MINSIGSTKSZ);
4115 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4116 		err = -EFAULT;
4117 	return err;
4118 }
4119 
restore_altstack(const stack_t __user * uss)4120 int restore_altstack(const stack_t __user *uss)
4121 {
4122 	stack_t new;
4123 	if (copy_from_user(&new, uss, sizeof(stack_t)))
4124 		return -EFAULT;
4125 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4126 			     MINSIGSTKSZ);
4127 	/* squash all but EFAULT for now */
4128 	return 0;
4129 }
4130 
__save_altstack(stack_t __user * uss,unsigned long sp)4131 int __save_altstack(stack_t __user *uss, unsigned long sp)
4132 {
4133 	struct task_struct *t = current;
4134 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4135 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4136 		__put_user(t->sas_ss_size, &uss->ss_size);
4137 	if (err)
4138 		return err;
4139 	if (t->sas_ss_flags & SS_AUTODISARM)
4140 		sas_ss_reset(t);
4141 	return 0;
4142 }
4143 
4144 #ifdef CONFIG_COMPAT
do_compat_sigaltstack(const compat_stack_t __user * uss_ptr,compat_stack_t __user * uoss_ptr)4145 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4146 				 compat_stack_t __user *uoss_ptr)
4147 {
4148 	stack_t uss, uoss;
4149 	int ret;
4150 
4151 	if (uss_ptr) {
4152 		compat_stack_t uss32;
4153 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4154 			return -EFAULT;
4155 		uss.ss_sp = compat_ptr(uss32.ss_sp);
4156 		uss.ss_flags = uss32.ss_flags;
4157 		uss.ss_size = uss32.ss_size;
4158 	}
4159 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4160 			     compat_user_stack_pointer(),
4161 			     COMPAT_MINSIGSTKSZ);
4162 	if (ret >= 0 && uoss_ptr)  {
4163 		compat_stack_t old;
4164 		memset(&old, 0, sizeof(old));
4165 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4166 		old.ss_flags = uoss.ss_flags;
4167 		old.ss_size = uoss.ss_size;
4168 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4169 			ret = -EFAULT;
4170 	}
4171 	return ret;
4172 }
4173 
COMPAT_SYSCALL_DEFINE2(sigaltstack,const compat_stack_t __user *,uss_ptr,compat_stack_t __user *,uoss_ptr)4174 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4175 			const compat_stack_t __user *, uss_ptr,
4176 			compat_stack_t __user *, uoss_ptr)
4177 {
4178 	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4179 }
4180 
compat_restore_altstack(const compat_stack_t __user * uss)4181 int compat_restore_altstack(const compat_stack_t __user *uss)
4182 {
4183 	int err = do_compat_sigaltstack(uss, NULL);
4184 	/* squash all but -EFAULT for now */
4185 	return err == -EFAULT ? err : 0;
4186 }
4187 
__compat_save_altstack(compat_stack_t __user * uss,unsigned long sp)4188 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4189 {
4190 	int err;
4191 	struct task_struct *t = current;
4192 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4193 			 &uss->ss_sp) |
4194 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4195 		__put_user(t->sas_ss_size, &uss->ss_size);
4196 	if (err)
4197 		return err;
4198 	if (t->sas_ss_flags & SS_AUTODISARM)
4199 		sas_ss_reset(t);
4200 	return 0;
4201 }
4202 #endif
4203 
4204 #ifdef __ARCH_WANT_SYS_SIGPENDING
4205 
4206 /**
4207  *  sys_sigpending - examine pending signals
4208  *  @uset: where mask of pending signal is returned
4209  */
SYSCALL_DEFINE1(sigpending,old_sigset_t __user *,uset)4210 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4211 {
4212 	sigset_t set;
4213 
4214 	if (sizeof(old_sigset_t) > sizeof(*uset))
4215 		return -EINVAL;
4216 
4217 	do_sigpending(&set);
4218 
4219 	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4220 		return -EFAULT;
4221 
4222 	return 0;
4223 }
4224 
4225 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE1(sigpending,compat_old_sigset_t __user *,set32)4226 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4227 {
4228 	sigset_t set;
4229 
4230 	do_sigpending(&set);
4231 
4232 	return put_user(set.sig[0], set32);
4233 }
4234 #endif
4235 
4236 #endif
4237 
4238 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4239 /**
4240  *  sys_sigprocmask - examine and change blocked signals
4241  *  @how: whether to add, remove, or set signals
4242  *  @nset: signals to add or remove (if non-null)
4243  *  @oset: previous value of signal mask if non-null
4244  *
4245  * Some platforms have their own version with special arguments;
4246  * others support only sys_rt_sigprocmask.
4247  */
4248 
SYSCALL_DEFINE3(sigprocmask,int,how,old_sigset_t __user *,nset,old_sigset_t __user *,oset)4249 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4250 		old_sigset_t __user *, oset)
4251 {
4252 	old_sigset_t old_set, new_set;
4253 	sigset_t new_blocked;
4254 
4255 	old_set = current->blocked.sig[0];
4256 
4257 	if (nset) {
4258 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4259 			return -EFAULT;
4260 
4261 		new_blocked = current->blocked;
4262 
4263 		switch (how) {
4264 		case SIG_BLOCK:
4265 			sigaddsetmask(&new_blocked, new_set);
4266 			break;
4267 		case SIG_UNBLOCK:
4268 			sigdelsetmask(&new_blocked, new_set);
4269 			break;
4270 		case SIG_SETMASK:
4271 			new_blocked.sig[0] = new_set;
4272 			break;
4273 		default:
4274 			return -EINVAL;
4275 		}
4276 
4277 		set_current_blocked(&new_blocked);
4278 	}
4279 
4280 	if (oset) {
4281 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4282 			return -EFAULT;
4283 	}
4284 
4285 	return 0;
4286 }
4287 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4288 
4289 #ifndef CONFIG_ODD_RT_SIGACTION
4290 /**
4291  *  sys_rt_sigaction - alter an action taken by a process
4292  *  @sig: signal to be sent
4293  *  @act: new sigaction
4294  *  @oact: used to save the previous sigaction
4295  *  @sigsetsize: size of sigset_t type
4296  */
SYSCALL_DEFINE4(rt_sigaction,int,sig,const struct sigaction __user *,act,struct sigaction __user *,oact,size_t,sigsetsize)4297 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4298 		const struct sigaction __user *, act,
4299 		struct sigaction __user *, oact,
4300 		size_t, sigsetsize)
4301 {
4302 	struct k_sigaction new_sa, old_sa;
4303 	int ret;
4304 
4305 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4306 	if (sigsetsize != sizeof(sigset_t))
4307 		return -EINVAL;
4308 
4309 	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4310 		return -EFAULT;
4311 
4312 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4313 	if (ret)
4314 		return ret;
4315 
4316 	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4317 		return -EFAULT;
4318 
4319 	return 0;
4320 }
4321 #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)4322 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4323 		const struct compat_sigaction __user *, act,
4324 		struct compat_sigaction __user *, oact,
4325 		compat_size_t, sigsetsize)
4326 {
4327 	struct k_sigaction new_ka, old_ka;
4328 #ifdef __ARCH_HAS_SA_RESTORER
4329 	compat_uptr_t restorer;
4330 #endif
4331 	int ret;
4332 
4333 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4334 	if (sigsetsize != sizeof(compat_sigset_t))
4335 		return -EINVAL;
4336 
4337 	if (act) {
4338 		compat_uptr_t handler;
4339 		ret = get_user(handler, &act->sa_handler);
4340 		new_ka.sa.sa_handler = compat_ptr(handler);
4341 #ifdef __ARCH_HAS_SA_RESTORER
4342 		ret |= get_user(restorer, &act->sa_restorer);
4343 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4344 #endif
4345 		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4346 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4347 		if (ret)
4348 			return -EFAULT;
4349 	}
4350 
4351 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4352 	if (!ret && oact) {
4353 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4354 			       &oact->sa_handler);
4355 		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4356 					 sizeof(oact->sa_mask));
4357 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4358 #ifdef __ARCH_HAS_SA_RESTORER
4359 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4360 				&oact->sa_restorer);
4361 #endif
4362 	}
4363 	return ret;
4364 }
4365 #endif
4366 #endif /* !CONFIG_ODD_RT_SIGACTION */
4367 
4368 #ifdef CONFIG_OLD_SIGACTION
SYSCALL_DEFINE3(sigaction,int,sig,const struct old_sigaction __user *,act,struct old_sigaction __user *,oact)4369 SYSCALL_DEFINE3(sigaction, int, sig,
4370 		const struct old_sigaction __user *, act,
4371 	        struct old_sigaction __user *, oact)
4372 {
4373 	struct k_sigaction new_ka, old_ka;
4374 	int ret;
4375 
4376 	if (act) {
4377 		old_sigset_t mask;
4378 		if (!access_ok(act, sizeof(*act)) ||
4379 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4380 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4381 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4382 		    __get_user(mask, &act->sa_mask))
4383 			return -EFAULT;
4384 #ifdef __ARCH_HAS_KA_RESTORER
4385 		new_ka.ka_restorer = NULL;
4386 #endif
4387 		siginitset(&new_ka.sa.sa_mask, mask);
4388 	}
4389 
4390 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4391 
4392 	if (!ret && oact) {
4393 		if (!access_ok(oact, sizeof(*oact)) ||
4394 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4395 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4396 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4397 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4398 			return -EFAULT;
4399 	}
4400 
4401 	return ret;
4402 }
4403 #endif
4404 #ifdef CONFIG_COMPAT_OLD_SIGACTION
COMPAT_SYSCALL_DEFINE3(sigaction,int,sig,const struct compat_old_sigaction __user *,act,struct compat_old_sigaction __user *,oact)4405 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4406 		const struct compat_old_sigaction __user *, act,
4407 	        struct compat_old_sigaction __user *, oact)
4408 {
4409 	struct k_sigaction new_ka, old_ka;
4410 	int ret;
4411 	compat_old_sigset_t mask;
4412 	compat_uptr_t handler, restorer;
4413 
4414 	if (act) {
4415 		if (!access_ok(act, sizeof(*act)) ||
4416 		    __get_user(handler, &act->sa_handler) ||
4417 		    __get_user(restorer, &act->sa_restorer) ||
4418 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4419 		    __get_user(mask, &act->sa_mask))
4420 			return -EFAULT;
4421 
4422 #ifdef __ARCH_HAS_KA_RESTORER
4423 		new_ka.ka_restorer = NULL;
4424 #endif
4425 		new_ka.sa.sa_handler = compat_ptr(handler);
4426 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4427 		siginitset(&new_ka.sa.sa_mask, mask);
4428 	}
4429 
4430 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4431 
4432 	if (!ret && oact) {
4433 		if (!access_ok(oact, sizeof(*oact)) ||
4434 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4435 			       &oact->sa_handler) ||
4436 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4437 			       &oact->sa_restorer) ||
4438 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4439 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4440 			return -EFAULT;
4441 	}
4442 	return ret;
4443 }
4444 #endif
4445 
4446 #ifdef CONFIG_SGETMASK_SYSCALL
4447 
4448 /*
4449  * For backwards compatibility.  Functionality superseded by sigprocmask.
4450  */
SYSCALL_DEFINE0(sgetmask)4451 SYSCALL_DEFINE0(sgetmask)
4452 {
4453 	/* SMP safe */
4454 	return current->blocked.sig[0];
4455 }
4456 
SYSCALL_DEFINE1(ssetmask,int,newmask)4457 SYSCALL_DEFINE1(ssetmask, int, newmask)
4458 {
4459 	int old = current->blocked.sig[0];
4460 	sigset_t newset;
4461 
4462 	siginitset(&newset, newmask);
4463 	set_current_blocked(&newset);
4464 
4465 	return old;
4466 }
4467 #endif /* CONFIG_SGETMASK_SYSCALL */
4468 
4469 #ifdef __ARCH_WANT_SYS_SIGNAL
4470 /*
4471  * For backwards compatibility.  Functionality superseded by sigaction.
4472  */
SYSCALL_DEFINE2(signal,int,sig,__sighandler_t,handler)4473 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4474 {
4475 	struct k_sigaction new_sa, old_sa;
4476 	int ret;
4477 
4478 	new_sa.sa.sa_handler = handler;
4479 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4480 	sigemptyset(&new_sa.sa.sa_mask);
4481 
4482 	ret = do_sigaction(sig, &new_sa, &old_sa);
4483 
4484 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4485 }
4486 #endif /* __ARCH_WANT_SYS_SIGNAL */
4487 
4488 #ifdef __ARCH_WANT_SYS_PAUSE
4489 
SYSCALL_DEFINE0(pause)4490 SYSCALL_DEFINE0(pause)
4491 {
4492 	while (!signal_pending(current)) {
4493 		__set_current_state(TASK_INTERRUPTIBLE);
4494 		schedule();
4495 	}
4496 	return -ERESTARTNOHAND;
4497 }
4498 
4499 #endif
4500 
sigsuspend(sigset_t * set)4501 static int sigsuspend(sigset_t *set)
4502 {
4503 	current->saved_sigmask = current->blocked;
4504 	set_current_blocked(set);
4505 
4506 	while (!signal_pending(current)) {
4507 		__set_current_state(TASK_INTERRUPTIBLE);
4508 		schedule();
4509 	}
4510 	set_restore_sigmask();
4511 	return -ERESTARTNOHAND;
4512 }
4513 
4514 /**
4515  *  sys_rt_sigsuspend - replace the signal mask for a value with the
4516  *	@unewset value until a signal is received
4517  *  @unewset: new signal mask value
4518  *  @sigsetsize: size of sigset_t type
4519  */
SYSCALL_DEFINE2(rt_sigsuspend,sigset_t __user *,unewset,size_t,sigsetsize)4520 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4521 {
4522 	sigset_t newset;
4523 
4524 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4525 	if (sigsetsize != sizeof(sigset_t))
4526 		return -EINVAL;
4527 
4528 	if (copy_from_user(&newset, unewset, sizeof(newset)))
4529 		return -EFAULT;
4530 	return sigsuspend(&newset);
4531 }
4532 
4533 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(rt_sigsuspend,compat_sigset_t __user *,unewset,compat_size_t,sigsetsize)4534 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4535 {
4536 	sigset_t newset;
4537 
4538 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4539 	if (sigsetsize != sizeof(sigset_t))
4540 		return -EINVAL;
4541 
4542 	if (get_compat_sigset(&newset, unewset))
4543 		return -EFAULT;
4544 	return sigsuspend(&newset);
4545 }
4546 #endif
4547 
4548 #ifdef CONFIG_OLD_SIGSUSPEND
SYSCALL_DEFINE1(sigsuspend,old_sigset_t,mask)4549 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4550 {
4551 	sigset_t blocked;
4552 	siginitset(&blocked, mask);
4553 	return sigsuspend(&blocked);
4554 }
4555 #endif
4556 #ifdef CONFIG_OLD_SIGSUSPEND3
SYSCALL_DEFINE3(sigsuspend,int,unused1,int,unused2,old_sigset_t,mask)4557 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4558 {
4559 	sigset_t blocked;
4560 	siginitset(&blocked, mask);
4561 	return sigsuspend(&blocked);
4562 }
4563 #endif
4564 
arch_vma_name(struct vm_area_struct * vma)4565 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4566 {
4567 	return NULL;
4568 }
4569 
siginfo_buildtime_checks(void)4570 static inline void siginfo_buildtime_checks(void)
4571 {
4572 	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4573 
4574 	/* Verify the offsets in the two siginfos match */
4575 #define CHECK_OFFSET(field) \
4576 	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4577 
4578 	/* kill */
4579 	CHECK_OFFSET(si_pid);
4580 	CHECK_OFFSET(si_uid);
4581 
4582 	/* timer */
4583 	CHECK_OFFSET(si_tid);
4584 	CHECK_OFFSET(si_overrun);
4585 	CHECK_OFFSET(si_value);
4586 
4587 	/* rt */
4588 	CHECK_OFFSET(si_pid);
4589 	CHECK_OFFSET(si_uid);
4590 	CHECK_OFFSET(si_value);
4591 
4592 	/* sigchld */
4593 	CHECK_OFFSET(si_pid);
4594 	CHECK_OFFSET(si_uid);
4595 	CHECK_OFFSET(si_status);
4596 	CHECK_OFFSET(si_utime);
4597 	CHECK_OFFSET(si_stime);
4598 
4599 	/* sigfault */
4600 	CHECK_OFFSET(si_addr);
4601 	CHECK_OFFSET(si_addr_lsb);
4602 	CHECK_OFFSET(si_lower);
4603 	CHECK_OFFSET(si_upper);
4604 	CHECK_OFFSET(si_pkey);
4605 
4606 	/* sigpoll */
4607 	CHECK_OFFSET(si_band);
4608 	CHECK_OFFSET(si_fd);
4609 
4610 	/* sigsys */
4611 	CHECK_OFFSET(si_call_addr);
4612 	CHECK_OFFSET(si_syscall);
4613 	CHECK_OFFSET(si_arch);
4614 #undef CHECK_OFFSET
4615 
4616 	/* usb asyncio */
4617 	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4618 		     offsetof(struct siginfo, si_addr));
4619 	if (sizeof(int) == sizeof(void __user *)) {
4620 		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4621 			     sizeof(void __user *));
4622 	} else {
4623 		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4624 			      sizeof_field(struct siginfo, si_uid)) !=
4625 			     sizeof(void __user *));
4626 		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4627 			     offsetof(struct siginfo, si_uid));
4628 	}
4629 #ifdef CONFIG_COMPAT
4630 	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4631 		     offsetof(struct compat_siginfo, si_addr));
4632 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4633 		     sizeof(compat_uptr_t));
4634 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4635 		     sizeof_field(struct siginfo, si_pid));
4636 #endif
4637 }
4638 
signals_init(void)4639 void __init signals_init(void)
4640 {
4641 	siginfo_buildtime_checks();
4642 
4643 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
4644 }
4645 
4646 #ifdef CONFIG_KGDB_KDB
4647 #include <linux/kdb.h>
4648 /*
4649  * kdb_send_sig - Allows kdb to send signals without exposing
4650  * signal internals.  This function checks if the required locks are
4651  * available before calling the main signal code, to avoid kdb
4652  * deadlocks.
4653  */
kdb_send_sig(struct task_struct * t,int sig)4654 void kdb_send_sig(struct task_struct *t, int sig)
4655 {
4656 	static struct task_struct *kdb_prev_t;
4657 	int new_t, ret;
4658 	if (!spin_trylock(&t->sighand->siglock)) {
4659 		kdb_printf("Can't do kill command now.\n"
4660 			   "The sigmask lock is held somewhere else in "
4661 			   "kernel, try again later\n");
4662 		return;
4663 	}
4664 	new_t = kdb_prev_t != t;
4665 	kdb_prev_t = t;
4666 	if (t->state != TASK_RUNNING && new_t) {
4667 		spin_unlock(&t->sighand->siglock);
4668 		kdb_printf("Process is not RUNNING, sending a signal from "
4669 			   "kdb risks deadlock\n"
4670 			   "on the run queue locks. "
4671 			   "The signal has _not_ been sent.\n"
4672 			   "Reissue the kill command if you want to risk "
4673 			   "the deadlock.\n");
4674 		return;
4675 	}
4676 	ret = send_signal(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4677 	spin_unlock(&t->sighand->siglock);
4678 	if (ret)
4679 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
4680 			   sig, t->pid);
4681 	else
4682 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4683 }
4684 #endif	/* CONFIG_KGDB_KDB */
4685