signal.c

   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/mm.h>
  26#include <linux/proc_fs.h>
  27#include <linux/tty.h>
  28#include <linux/binfmts.h>
  29#include <linux/coredump.h>
  30#include <linux/security.h>
  31#include <linux/syscalls.h>
  32#include <linux/ptrace.h>
  33#include <linux/signal.h>
  34#include <linux/signalfd.h>
  35#include <linux/ratelimit.h>
  36#include <linux/task_work.h>
  37#include <linux/capability.h>
  38#include <linux/freezer.h>
  39#include <linux/pid_namespace.h>
  40#include <linux/nsproxy.h>
  41#include <linux/user_namespace.h>
  42#include <linux/uprobes.h>
  43#include <linux/compat.h>
  44#include <linux/cn_proc.h>
  45#include <linux/compiler.h>
  46#include <linux/posix-timers.h>
  47#include <linux/cgroup.h>
  48#include <linux/audit.h>
  49#include <linux/sysctl.h>
  50#include <uapi/linux/pidfd.h>
  51
  52#define CREATE_TRACE_POINTS
  53#include <trace/events/signal.h>
  54
  55#include <asm/param.h>
  56#include <linux/uaccess.h>
  57#include <asm/unistd.h>
  58#include <asm/siginfo.h>
  59#include <asm/cacheflush.h>
  60#include <asm/syscall.h>	/* for syscall_get_* */
  61
  62#include "time/posix-timers.h"
  63
  64/*
  65 * SLAB caches for signal bits.
  66 */
  67
  68static struct kmem_cache *sigqueue_cachep;
  69
  70int print_fatal_signals __read_mostly;
  71
  72static void __user *sig_handler(struct task_struct *t, int sig)
  73{
  74	return t->sighand->action[sig - 1].sa.sa_handler;
  75}
  76
  77static inline bool sig_handler_ignored(void __user *handler, int sig)
  78{
  79	/* Is it explicitly or implicitly ignored? */
  80	return handler == SIG_IGN ||
  81	       (handler == SIG_DFL && sig_kernel_ignore(sig));
  82}
  83
  84static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
  85{
  86	void __user *handler;
  87
  88	handler = sig_handler(t, sig);
  89
  90	/* SIGKILL and SIGSTOP may not be sent to the global init */
  91	if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
  92		return true;
  93
  94	if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
  95	    handler == SIG_DFL && !(force && sig_kernel_only(sig)))
  96		return true;
  97
  98	/* Only allow kernel generated signals to this kthread */
  99	if (unlikely((t->flags & PF_KTHREAD) &&
 100		     (handler == SIG_KTHREAD_KERNEL) && !force))
 101		return true;
 102
 103	return sig_handler_ignored(handler, sig);
 104}
 105
 106static bool sig_ignored(struct task_struct *t, int sig, bool force)
 107{
 108	/*
 109	 * Blocked signals are never ignored, since the
 110	 * signal handler may change by the time it is
 111	 * unblocked.
 112	 */
 113	if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
 114		return false;
 115
 116	/*
 117	 * Tracers may want to know about even ignored signal unless it
 118	 * is SIGKILL which can't be reported anyway but can be ignored
 119	 * by SIGNAL_UNKILLABLE task.
 120	 */
 121	if (t->ptrace && sig != SIGKILL)
 122		return false;
 123
 124	return sig_task_ignored(t, sig, force);
 125}
 126
 127/*
 128 * Re-calculate pending state from the set of locally pending
 129 * signals, globally pending signals, and blocked signals.
 130 */
 131static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
 132{
 133	unsigned long ready;
 134	long i;
 135
 136	switch (_NSIG_WORDS) {
 137	default:
 138		for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
 139			ready |= signal->sig[i] &~ blocked->sig[i];
 140		break;
 141
 142	case 4: ready  = signal->sig[3] &~ blocked->sig[3];
 143		ready |= signal->sig[2] &~ blocked->sig[2];
 144		ready |= signal->sig[1] &~ blocked->sig[1];
 145		ready |= signal->sig[0] &~ blocked->sig[0];
 146		break;
 147
 148	case 2: ready  = signal->sig[1] &~ blocked->sig[1];
 149		ready |= signal->sig[0] &~ blocked->sig[0];
 150		break;
 151
 152	case 1: ready  = signal->sig[0] &~ blocked->sig[0];
 153	}
 154	return ready !=	0;
 155}
 156
 157#define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
 158
 159static bool recalc_sigpending_tsk(struct task_struct *t)
 160{
 161	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
 162	    PENDING(&t->pending, &t->blocked) ||
 163	    PENDING(&t->signal->shared_pending, &t->blocked) ||
 164	    cgroup_task_frozen(t)) {
 165		set_tsk_thread_flag(t, TIF_SIGPENDING);
 166		return true;
 167	}
 168
 169	/*
 170	 * We must never clear the flag in another thread, or in current
 171	 * when it's possible the current syscall is returning -ERESTART*.
 172	 * So we don't clear it here, and only callers who know they should do.
 173	 */
 174	return false;
 175}
 176
 177void recalc_sigpending(void)
 178{
 179	if (!recalc_sigpending_tsk(current) && !freezing(current)) {
 180		if (unlikely(test_thread_flag(TIF_SIGPENDING)))
 181			clear_thread_flag(TIF_SIGPENDING);
 182	}
 183}
 184EXPORT_SYMBOL(recalc_sigpending);
 185
 186void calculate_sigpending(void)
 187{
 188	/* Have any signals or users of TIF_SIGPENDING been delayed
 189	 * until after fork?
 190	 */
 191	spin_lock_irq(&current->sighand->siglock);
 192	set_tsk_thread_flag(current, TIF_SIGPENDING);
 193	recalc_sigpending();
 194	spin_unlock_irq(&current->sighand->siglock);
 195}
 196
 197/* Given the mask, find the first available signal that should be serviced. */
 198
 199#define SYNCHRONOUS_MASK \
 200	(sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
 201	 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
 202
 203int next_signal(struct sigpending *pending, sigset_t *mask)
 204{
 205	unsigned long i, *s, *m, x;
 206	int sig = 0;
 207
 208	s = pending->signal.sig;
 209	m = mask->sig;
 210
 211	/*
 212	 * Handle the first word specially: it contains the
 213	 * synchronous signals that need to be dequeued first.
 214	 */
 215	x = *s &~ *m;
 216	if (x) {
 217		if (x & SYNCHRONOUS_MASK)
 218			x &= SYNCHRONOUS_MASK;
 219		sig = ffz(~x) + 1;
 220		return sig;
 221	}
 222
 223	switch (_NSIG_WORDS) {
 224	default:
 225		for (i = 1; i < _NSIG_WORDS; ++i) {
 226			x = *++s &~ *++m;
 227			if (!x)
 228				continue;
 229			sig = ffz(~x) + i*_NSIG_BPW + 1;
 230			break;
 231		}
 232		break;
 233
 234	case 2:
 235		x = s[1] &~ m[1];
 236		if (!x)
 237			break;
 238		sig = ffz(~x) + _NSIG_BPW + 1;
 239		break;
 240
 241	case 1:
 242		/* Nothing to do */
 243		break;
 244	}
 245
 246	return sig;
 247}
 248
 249static inline void print_dropped_signal(int sig)
 250{
 251	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
 252
 253	if (!print_fatal_signals)
 254		return;
 255
 256	if (!__ratelimit(&ratelimit_state))
 257		return;
 258
 259	pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
 260				current->comm, current->pid, sig);
 261}
 262
 263/**
 264 * task_set_jobctl_pending - set jobctl pending bits
 265 * @task: target task
 266 * @mask: pending bits to set
 267 *
 268 * Clear @mask from @task->jobctl.  @mask must be subset of
 269 * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
 270 * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
 271 * cleared.  If @task is already being killed or exiting, this function
 272 * becomes noop.
 273 *
 274 * CONTEXT:
 275 * Must be called with @task->sighand->siglock held.
 276 *
 277 * RETURNS:
 278 * %true if @mask is set, %false if made noop because @task was dying.
 279 */
 280bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
 281{
 282	BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
 283			JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
 284	BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
 285
 286	if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
 287		return false;
 288
 289	if (mask & JOBCTL_STOP_SIGMASK)
 290		task->jobctl &= ~JOBCTL_STOP_SIGMASK;
 291
 292	task->jobctl |= mask;
 293	return true;
 294}
 295
 296/**
 297 * task_clear_jobctl_trapping - clear jobctl trapping bit
 298 * @task: target task
 299 *
 300 * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
 301 * Clear it and wake up the ptracer.  Note that we don't need any further
 302 * locking.  @task->siglock guarantees that @task->parent points to the
 303 * ptracer.
 304 *
 305 * CONTEXT:
 306 * Must be called with @task->sighand->siglock held.
 307 */
 308void task_clear_jobctl_trapping(struct task_struct *task)
 309{
 310	if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
 311		task->jobctl &= ~JOBCTL_TRAPPING;
 312		smp_mb();	/* advised by wake_up_bit() */
 313		wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
 314	}
 315}
 316
 317/**
 318 * task_clear_jobctl_pending - clear jobctl pending bits
 319 * @task: target task
 320 * @mask: pending bits to clear
 321 *
 322 * Clear @mask from @task->jobctl.  @mask must be subset of
 323 * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
 324 * STOP bits are cleared together.
 325 *
 326 * If clearing of @mask leaves no stop or trap pending, this function calls
 327 * task_clear_jobctl_trapping().
 328 *
 329 * CONTEXT:
 330 * Must be called with @task->sighand->siglock held.
 331 */
 332void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
 333{
 334	BUG_ON(mask & ~JOBCTL_PENDING_MASK);
 335
 336	if (mask & JOBCTL_STOP_PENDING)
 337		mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
 338
 339	task->jobctl &= ~mask;
 340
 341	if (!(task->jobctl & JOBCTL_PENDING_MASK))
 342		task_clear_jobctl_trapping(task);
 343}
 344
 345/**
 346 * task_participate_group_stop - participate in a group stop
 347 * @task: task participating in a group stop
 348 *
 349 * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
 350 * Group stop states are cleared and the group stop count is consumed if
 351 * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
 352 * stop, the appropriate `SIGNAL_*` flags are set.
 353 *
 354 * CONTEXT:
 355 * Must be called with @task->sighand->siglock held.
 356 *
 357 * RETURNS:
 358 * %true if group stop completion should be notified to the parent, %false
 359 * otherwise.
 360 */
 361static bool task_participate_group_stop(struct task_struct *task)
 362{
 363	struct signal_struct *sig = task->signal;
 364	bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
 365
 366	WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
 367
 368	task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
 369
 370	if (!consume)
 371		return false;
 372
 373	if (!WARN_ON_ONCE(sig->group_stop_count == 0))
 374		sig->group_stop_count--;
 375
 376	/*
 377	 * Tell the caller to notify completion iff we are entering into a
 378	 * fresh group stop.  Read comment in do_signal_stop() for details.
 379	 */
 380	if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
 381		signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
 382		return true;
 383	}
 384	return false;
 385}
 386
 387void task_join_group_stop(struct task_struct *task)
 388{
 389	unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
 390	struct signal_struct *sig = current->signal;
 391
 392	if (sig->group_stop_count) {
 393		sig->group_stop_count++;
 394		mask |= JOBCTL_STOP_CONSUME;
 395	} else if (!(sig->flags & SIGNAL_STOP_STOPPED))
 396		return;
 397
 398	/* Have the new thread join an on-going signal group stop */
 399	task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
 400}
 401
 402static struct ucounts *sig_get_ucounts(struct task_struct *t, int sig,
 403				       int override_rlimit)
 404{
 405	struct ucounts *ucounts;
 406	long sigpending;
 407
 408	/*
 409	 * Protect access to @t credentials. This can go away when all
 410	 * callers hold rcu read lock.
 411	 *
 412	 * NOTE! A pending signal will hold on to the user refcount,
 413	 * and we get/put the refcount only when the sigpending count
 414	 * changes from/to zero.
 415	 */
 416	rcu_read_lock();
 417	ucounts = task_ucounts(t);
 418	sigpending = inc_rlimit_get_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING,
 419					    override_rlimit);
 420	rcu_read_unlock();
 421	if (!sigpending)
 422		return NULL;
 423
 424	if (unlikely(!override_rlimit && sigpending > task_rlimit(t, RLIMIT_SIGPENDING))) {
 425		dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
 426		print_dropped_signal(sig);
 427		return NULL;
 428	}
 429
 430	return ucounts;
 431}
 432
 433static void __sigqueue_init(struct sigqueue *q, struct ucounts *ucounts,
 434			    const unsigned int sigqueue_flags)
 435{
 436	INIT_LIST_HEAD(&q->list);
 437	q->flags = sigqueue_flags;
 438	q->ucounts = ucounts;
 439}
 440
 441/*
 442 * allocate a new signal queue record
 443 * - this may be called without locks if and only if t == current, otherwise an
 444 *   appropriate lock must be held to stop the target task from exiting
 445 */
 446static struct sigqueue *sigqueue_alloc(int sig, struct task_struct *t, gfp_t gfp_flags,
 447				       int override_rlimit)
 448{
 449	struct ucounts *ucounts = sig_get_ucounts(t, sig, override_rlimit);
 450	struct sigqueue *q;
 451
 452	if (!ucounts)
 453		return NULL;
 454
 455	q = kmem_cache_alloc(sigqueue_cachep, gfp_flags);
 456	if (!q) {
 457		dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
 458		return NULL;
 459	}
 460
 461	__sigqueue_init(q, ucounts, 0);
 462	return q;
 463}
 464
 465static void __sigqueue_free(struct sigqueue *q)
 466{
 467	if (q->flags & SIGQUEUE_PREALLOC) {
 468		posixtimer_sigqueue_putref(q);
 469		return;
 470	}
 471	if (q->ucounts) {
 472		dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING);
 473		q->ucounts = NULL;
 474	}
 475	kmem_cache_free(sigqueue_cachep, q);
 476}
 477
 478void flush_sigqueue(struct sigpending *queue)
 479{
 480	struct sigqueue *q;
 481
 482	sigemptyset(&queue->signal);
 483	while (!list_empty(&queue->list)) {
 484		q = list_entry(queue->list.next, struct sigqueue , list);
 485		list_del_init(&q->list);
 486		__sigqueue_free(q);
 487	}
 488}
 489
 490/*
 491 * Flush all pending signals for this kthread.
 492 */
 493void flush_signals(struct task_struct *t)
 494{
 495	unsigned long flags;
 496
 497	spin_lock_irqsave(&t->sighand->siglock, flags);
 498	clear_tsk_thread_flag(t, TIF_SIGPENDING);
 499	flush_sigqueue(&t->pending);
 500	flush_sigqueue(&t->signal->shared_pending);
 501	spin_unlock_irqrestore(&t->sighand->siglock, flags);
 502}
 503EXPORT_SYMBOL(flush_signals);
 504
 505void ignore_signals(struct task_struct *t)
 506{
 507	int i;
 508
 509	for (i = 0; i < _NSIG; ++i)
 510		t->sighand->action[i].sa.sa_handler = SIG_IGN;
 511
 512	flush_signals(t);
 513}
 514
 515/*
 516 * Flush all handlers for a task.
 517 */
 518
 519void
 520flush_signal_handlers(struct task_struct *t, int force_default)
 521{
 522	int i;
 523	struct k_sigaction *ka = &t->sighand->action[0];
 524	for (i = _NSIG ; i != 0 ; i--) {
 525		if (force_default || ka->sa.sa_handler != SIG_IGN)
 526			ka->sa.sa_handler = SIG_DFL;
 527		ka->sa.sa_flags = 0;
 528#ifdef __ARCH_HAS_SA_RESTORER
 529		ka->sa.sa_restorer = NULL;
 530#endif
 531		sigemptyset(&ka->sa.sa_mask);
 532		ka++;
 533	}
 534}
 535
 536bool unhandled_signal(struct task_struct *tsk, int sig)
 537{
 538	void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
 539	if (is_global_init(tsk))
 540		return true;
 541
 542	if (handler != SIG_IGN && handler != SIG_DFL)
 543		return false;
 544
 545	/* If dying, we handle all new signals by ignoring them */
 546	if (fatal_signal_pending(tsk))
 547		return false;
 548
 549	/* if ptraced, let the tracer determine */
 550	return !tsk->ptrace;
 551}
 552
 553static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
 554			   struct sigqueue **timer_sigq)
 555{
 556	struct sigqueue *q, *first = NULL;
 557
 558	/*
 559	 * Collect the siginfo appropriate to this signal.  Check if
 560	 * there is another siginfo for the same signal.
 561	*/
 562	list_for_each_entry(q, &list->list, list) {
 563		if (q->info.si_signo == sig) {
 564			if (first)
 565				goto still_pending;
 566			first = q;
 567		}
 568	}
 569
 570	sigdelset(&list->signal, sig);
 571
 572	if (first) {
 573still_pending:
 574		list_del_init(&first->list);
 575		copy_siginfo(info, &first->info);
 576
 577		/*
 578		 * posix-timer signals are preallocated and freed when the last
 579		 * reference count is dropped in posixtimer_deliver_signal() or
 580		 * immediately on timer deletion when the signal is not pending.
 581		 * Spare the extra round through __sigqueue_free() which is
 582		 * ignoring preallocated signals.
 583		 */
 584		if (unlikely((first->flags & SIGQUEUE_PREALLOC) && (info->si_code == SI_TIMER)))
 585			*timer_sigq = first;
 586		else
 587			__sigqueue_free(first);
 588	} else {
 589		/*
 590		 * Ok, it wasn't in the queue.  This must be
 591		 * a fast-pathed signal or we must have been
 592		 * out of queue space.  So zero out the info.
 593		 */
 594		clear_siginfo(info);
 595		info->si_signo = sig;
 596		info->si_errno = 0;
 597		info->si_code = SI_USER;
 598		info->si_pid = 0;
 599		info->si_uid = 0;
 600	}
 601}
 602
 603static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
 604			    kernel_siginfo_t *info, struct sigqueue **timer_sigq)
 605{
 606	int sig = next_signal(pending, mask);
 607
 608	if (sig)
 609		collect_signal(sig, pending, info, timer_sigq);
 610	return sig;
 611}
 612
 613/*
 614 * Try to dequeue a signal. If a deliverable signal is found fill in the
 615 * caller provided siginfo and return the signal number. Otherwise return
 616 * 0.
 617 */
 618int dequeue_signal(sigset_t *mask, kernel_siginfo_t *info, enum pid_type *type)
 619{
 620	struct task_struct *tsk = current;
 621	struct sigqueue *timer_sigq;
 622	int signr;
 623
 624	lockdep_assert_held(&tsk->sighand->siglock);
 625
 626again:
 627	*type = PIDTYPE_PID;
 628	timer_sigq = NULL;
 629	signr = __dequeue_signal(&tsk->pending, mask, info, &timer_sigq);
 630	if (!signr) {
 631		*type = PIDTYPE_TGID;
 632		signr = __dequeue_signal(&tsk->signal->shared_pending,
 633					 mask, info, &timer_sigq);
 634
 635		if (unlikely(signr == SIGALRM))
 636			posixtimer_rearm_itimer(tsk);
 637	}
 638
 639	recalc_sigpending();
 640	if (!signr)
 641		return 0;
 642
 643	if (unlikely(sig_kernel_stop(signr))) {
 644		/*
 645		 * Set a marker that we have dequeued a stop signal.  Our
 646		 * caller might release the siglock and then the pending
 647		 * stop signal it is about to process is no longer in the
 648		 * pending bitmasks, but must still be cleared by a SIGCONT
 649		 * (and overruled by a SIGKILL).  So those cases clear this
 650		 * shared flag after we've set it.  Note that this flag may
 651		 * remain set after the signal we return is ignored or
 652		 * handled.  That doesn't matter because its only purpose
 653		 * is to alert stop-signal processing code when another
 654		 * processor has come along and cleared the flag.
 655		 */
 656		current->jobctl |= JOBCTL_STOP_DEQUEUED;
 657	}
 658
 659	if (IS_ENABLED(CONFIG_POSIX_TIMERS) && unlikely(timer_sigq)) {
 660		if (!posixtimer_deliver_signal(info, timer_sigq))
 661			goto again;
 662	}
 663
 664	return signr;
 665}
 666EXPORT_SYMBOL_GPL(dequeue_signal);
 667
 668static int dequeue_synchronous_signal(kernel_siginfo_t *info)
 669{
 670	struct task_struct *tsk = current;
 671	struct sigpending *pending = &tsk->pending;
 672	struct sigqueue *q, *sync = NULL;
 673
 674	/*
 675	 * Might a synchronous signal be in the queue?
 676	 */
 677	if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
 678		return 0;
 679
 680	/*
 681	 * Return the first synchronous signal in the queue.
 682	 */
 683	list_for_each_entry(q, &pending->list, list) {
 684		/* Synchronous signals have a positive si_code */
 685		if ((q->info.si_code > SI_USER) &&
 686		    (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
 687			sync = q;
 688			goto next;
 689		}
 690	}
 691	return 0;
 692next:
 693	/*
 694	 * Check if there is another siginfo for the same signal.
 695	 */
 696	list_for_each_entry_continue(q, &pending->list, list) {
 697		if (q->info.si_signo == sync->info.si_signo)
 698			goto still_pending;
 699	}
 700
 701	sigdelset(&pending->signal, sync->info.si_signo);
 702	recalc_sigpending();
 703still_pending:
 704	list_del_init(&sync->list);
 705	copy_siginfo(info, &sync->info);
 706	__sigqueue_free(sync);
 707	return info->si_signo;
 708}
 709
 710/*
 711 * Tell a process that it has a new active signal..
 712 *
 713 * NOTE! we rely on the previous spin_lock to
 714 * lock interrupts for us! We can only be called with
 715 * "siglock" held, and the local interrupt must
 716 * have been disabled when that got acquired!
 717 *
 718 * No need to set need_resched since signal event passing
 719 * goes through ->blocked
 720 */
 721void signal_wake_up_state(struct task_struct *t, unsigned int state)
 722{
 723	lockdep_assert_held(&t->sighand->siglock);
 724
 725	set_tsk_thread_flag(t, TIF_SIGPENDING);
 726
 727	/*
 728	 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
 729	 * case. We don't check t->state here because there is a race with it
 730	 * executing another processor and just now entering stopped state.
 731	 * By using wake_up_state, we ensure the process will wake up and
 732	 * handle its death signal.
 733	 */
 734	if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
 735		kick_process(t);
 736}
 737
 738static inline void posixtimer_sig_ignore(struct task_struct *tsk, struct sigqueue *q);
 739
 740static void sigqueue_free_ignored(struct task_struct *tsk, struct sigqueue *q)
 741{
 742	if (likely(!(q->flags & SIGQUEUE_PREALLOC) || q->info.si_code != SI_TIMER))
 743		__sigqueue_free(q);
 744	else
 745		posixtimer_sig_ignore(tsk, q);
 746}
 747
 748/* Remove signals in mask from the pending set and queue. */
 749static void flush_sigqueue_mask(struct task_struct *p, sigset_t *mask, struct sigpending *s)
 750{
 751	struct sigqueue *q, *n;
 752	sigset_t m;
 753
 754	lockdep_assert_held(&p->sighand->siglock);
 755
 756	sigandsets(&m, mask, &s->signal);
 757	if (sigisemptyset(&m))
 758		return;
 759
 760	sigandnsets(&s->signal, &s->signal, mask);
 761	list_for_each_entry_safe(q, n, &s->list, list) {
 762		if (sigismember(mask, q->info.si_signo)) {
 763			list_del_init(&q->list);
 764			sigqueue_free_ignored(p, q);
 765		}
 766	}
 767}
 768
 769static inline int is_si_special(const struct kernel_siginfo *info)
 770{
 771	return info <= SEND_SIG_PRIV;
 772}
 773
 774static inline bool si_fromuser(const struct kernel_siginfo *info)
 775{
 776	return info == SEND_SIG_NOINFO ||
 777		(!is_si_special(info) && SI_FROMUSER(info));
 778}
 779
 780/*
 781 * called with RCU read lock from check_kill_permission()
 782 */
 783static bool kill_ok_by_cred(struct task_struct *t)
 784{
 785	const struct cred *cred = current_cred();
 786	const struct cred *tcred = __task_cred(t);
 787
 788	return uid_eq(cred->euid, tcred->suid) ||
 789	       uid_eq(cred->euid, tcred->uid) ||
 790	       uid_eq(cred->uid, tcred->suid) ||
 791	       uid_eq(cred->uid, tcred->uid) ||
 792	       ns_capable(tcred->user_ns, CAP_KILL);
 793}
 794
 795/*
 796 * Bad permissions for sending the signal
 797 * - the caller must hold the RCU read lock
 798 */
 799static int check_kill_permission(int sig, struct kernel_siginfo *info,
 800				 struct task_struct *t)
 801{
 802	struct pid *sid;
 803	int error;
 804
 805	if (!valid_signal(sig))
 806		return -EINVAL;
 807
 808	if (!si_fromuser(info))
 809		return 0;
 810
 811	error = audit_signal_info(sig, t); /* Let audit system see the signal */
 812	if (error)
 813		return error;
 814
 815	if (!same_thread_group(current, t) &&
 816	    !kill_ok_by_cred(t)) {
 817		switch (sig) {
 818		case SIGCONT:
 819			sid = task_session(t);
 820			/*
 821			 * We don't return the error if sid == NULL. The
 822			 * task was unhashed, the caller must notice this.
 823			 */
 824			if (!sid || sid == task_session(current))
 825				break;
 826			fallthrough;
 827		default:
 828			return -EPERM;
 829		}
 830	}
 831
 832	return security_task_kill(t, info, sig, NULL);
 833}
 834
 835/**
 836 * ptrace_trap_notify - schedule trap to notify ptracer
 837 * @t: tracee wanting to notify tracer
 838 *
 839 * This function schedules sticky ptrace trap which is cleared on the next
 840 * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
 841 * ptracer.
 842 *
 843 * If @t is running, STOP trap will be taken.  If trapped for STOP and
 844 * ptracer is listening for events, tracee is woken up so that it can
 845 * re-trap for the new event.  If trapped otherwise, STOP trap will be
 846 * eventually taken without returning to userland after the existing traps
 847 * are finished by PTRACE_CONT.
 848 *
 849 * CONTEXT:
 850 * Must be called with @task->sighand->siglock held.
 851 */
 852static void ptrace_trap_notify(struct task_struct *t)
 853{
 854	WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
 855	lockdep_assert_held(&t->sighand->siglock);
 856
 857	task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
 858	ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
 859}
 860
 861/*
 862 * Handle magic process-wide effects of stop/continue signals. Unlike
 863 * the signal actions, these happen immediately at signal-generation
 864 * time regardless of blocking, ignoring, or handling.  This does the
 865 * actual continuing for SIGCONT, but not the actual stopping for stop
 866 * signals. The process stop is done as a signal action for SIG_DFL.
 867 *
 868 * Returns true if the signal should be actually delivered, otherwise
 869 * it should be dropped.
 870 */
 871static bool prepare_signal(int sig, struct task_struct *p, bool force)
 872{
 873	struct signal_struct *signal = p->signal;
 874	struct task_struct *t;
 875	sigset_t flush;
 876
 877	if (signal->flags & SIGNAL_GROUP_EXIT) {
 878		if (signal->core_state)
 879			return sig == SIGKILL;
 880		/*
 881		 * The process is in the middle of dying, drop the signal.
 882		 */
 883		return false;
 884	} else if (sig_kernel_stop(sig)) {
 885		/*
 886		 * This is a stop signal.  Remove SIGCONT from all queues.
 887		 */
 888		siginitset(&flush, sigmask(SIGCONT));
 889		flush_sigqueue_mask(p, &flush, &signal->shared_pending);
 890		for_each_thread(p, t)
 891			flush_sigqueue_mask(p, &flush, &t->pending);
 892	} else if (sig == SIGCONT) {
 893		unsigned int why;
 894		/*
 895		 * Remove all stop signals from all queues, wake all threads.
 896		 */
 897		siginitset(&flush, SIG_KERNEL_STOP_MASK);
 898		flush_sigqueue_mask(p, &flush, &signal->shared_pending);
 899		for_each_thread(p, t) {
 900			flush_sigqueue_mask(p, &flush, &t->pending);
 901			task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
 902			if (likely(!(t->ptrace & PT_SEIZED))) {
 903				t->jobctl &= ~JOBCTL_STOPPED;
 904				wake_up_state(t, __TASK_STOPPED);
 905			} else
 906				ptrace_trap_notify(t);
 907		}
 908
 909		/*
 910		 * Notify the parent with CLD_CONTINUED if we were stopped.
 911		 *
 912		 * If we were in the middle of a group stop, we pretend it
 913		 * was already finished, and then continued. Since SIGCHLD
 914		 * doesn't queue we report only CLD_STOPPED, as if the next
 915		 * CLD_CONTINUED was dropped.
 916		 */
 917		why = 0;
 918		if (signal->flags & SIGNAL_STOP_STOPPED)
 919			why |= SIGNAL_CLD_CONTINUED;
 920		else if (signal->group_stop_count)
 921			why |= SIGNAL_CLD_STOPPED;
 922
 923		if (why) {
 924			/*
 925			 * The first thread which returns from do_signal_stop()
 926			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
 927			 * notify its parent. See get_signal().
 928			 */
 929			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
 930			signal->group_stop_count = 0;
 931			signal->group_exit_code = 0;
 932		}
 933	}
 934
 935	return !sig_ignored(p, sig, force);
 936}
 937
 938/*
 939 * Test if P wants to take SIG.  After we've checked all threads with this,
 940 * it's equivalent to finding no threads not blocking SIG.  Any threads not
 941 * blocking SIG were ruled out because they are not running and already
 942 * have pending signals.  Such threads will dequeue from the shared queue
 943 * as soon as they're available, so putting the signal on the shared queue
 944 * will be equivalent to sending it to one such thread.
 945 */
 946static inline bool wants_signal(int sig, struct task_struct *p)
 947{
 948	if (sigismember(&p->blocked, sig))
 949		return false;
 950
 951	if (p->flags & PF_EXITING)
 952		return false;
 953
 954	if (sig == SIGKILL)
 955		return true;
 956
 957	if (task_is_stopped_or_traced(p))
 958		return false;
 959
 960	return task_curr(p) || !task_sigpending(p);
 961}
 962
 963static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
 964{
 965	struct signal_struct *signal = p->signal;
 966	struct task_struct *t;
 967
 968	/*
 969	 * Now find a thread we can wake up to take the signal off the queue.
 970	 *
 971	 * Try the suggested task first (may or may not be the main thread).
 972	 */
 973	if (wants_signal(sig, p))
 974		t = p;
 975	else if ((type == PIDTYPE_PID) || thread_group_empty(p))
 976		/*
 977		 * There is just one thread and it does not need to be woken.
 978		 * It will dequeue unblocked signals before it runs again.
 979		 */
 980		return;
 981	else {
 982		/*
 983		 * Otherwise try to find a suitable thread.
 984		 */
 985		t = signal->curr_target;
 986		while (!wants_signal(sig, t)) {
 987			t = next_thread(t);
 988			if (t == signal->curr_target)
 989				/*
 990				 * No thread needs to be woken.
 991				 * Any eligible threads will see
 992				 * the signal in the queue soon.
 993				 */
 994				return;
 995		}
 996		signal->curr_target = t;
 997	}
 998
 999	/*
1000	 * Found a killable thread.  If the signal will be fatal,
1001	 * then start taking the whole group down immediately.
1002	 */
1003	if (sig_fatal(p, sig) &&
1004	    (signal->core_state || !(signal->flags & SIGNAL_GROUP_EXIT)) &&
1005	    !sigismember(&t->real_blocked, sig) &&
1006	    (sig == SIGKILL || !p->ptrace)) {
1007		/*
1008		 * This signal will be fatal to the whole group.
1009		 */
1010		if (!sig_kernel_coredump(sig)) {
1011			/*
1012			 * Start a group exit and wake everybody up.
1013			 * This way we don't have other threads
1014			 * running and doing things after a slower
1015			 * thread has the fatal signal pending.
1016			 */
1017			signal->flags = SIGNAL_GROUP_EXIT;
1018			signal->group_exit_code = sig;
1019			signal->group_stop_count = 0;
1020			__for_each_thread(signal, t) {
1021				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1022				sigaddset(&t->pending.signal, SIGKILL);
1023				signal_wake_up(t, 1);
1024			}
1025			return;
1026		}
1027	}
1028
1029	/*
1030	 * The signal is already in the shared-pending queue.
1031	 * Tell the chosen thread to wake up and dequeue it.
1032	 */
1033	signal_wake_up(t, sig == SIGKILL);
1034	return;
1035}
1036
1037static inline bool legacy_queue(struct sigpending *signals, int sig)
1038{
1039	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1040}
1041
1042static int __send_signal_locked(int sig, struct kernel_siginfo *info,
1043				struct task_struct *t, enum pid_type type, bool force)
1044{
1045	struct sigpending *pending;
1046	struct sigqueue *q;
1047	int override_rlimit;
1048	int ret = 0, result;
1049
1050	lockdep_assert_held(&t->sighand->siglock);
1051
1052	result = TRACE_SIGNAL_IGNORED;
1053	if (!prepare_signal(sig, t, force))
1054		goto ret;
1055
1056	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1057	/*
1058	 * Short-circuit ignored signals and support queuing
1059	 * exactly one non-rt signal, so that we can get more
1060	 * detailed information about the cause of the signal.
1061	 */
1062	result = TRACE_SIGNAL_ALREADY_PENDING;
1063	if (legacy_queue(pending, sig))
1064		goto ret;
1065
1066	result = TRACE_SIGNAL_DELIVERED;
1067	/*
1068	 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1069	 */
1070	if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1071		goto out_set;
1072
1073	/*
1074	 * Real-time signals must be queued if sent by sigqueue, or
1075	 * some other real-time mechanism.  It is implementation
1076	 * defined whether kill() does so.  We attempt to do so, on
1077	 * the principle of least surprise, but since kill is not
1078	 * allowed to fail with EAGAIN when low on memory we just
1079	 * make sure at least one signal gets delivered and don't
1080	 * pass on the info struct.
1081	 */
1082	if (sig < SIGRTMIN)
1083		override_rlimit = (is_si_special(info) || info->si_code >= 0);
1084	else
1085		override_rlimit = 0;
1086
1087	q = sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit);
1088
1089	if (q) {
1090		list_add_tail(&q->list, &pending->list);
1091		switch ((unsigned long) info) {
1092		case (unsigned long) SEND_SIG_NOINFO:
1093			clear_siginfo(&q->info);
1094			q->info.si_signo = sig;
1095			q->info.si_errno = 0;
1096			q->info.si_code = SI_USER;
1097			q->info.si_pid = task_tgid_nr_ns(current,
1098							task_active_pid_ns(t));
1099			rcu_read_lock();
1100			q->info.si_uid =
1101				from_kuid_munged(task_cred_xxx(t, user_ns),
1102						 current_uid());
1103			rcu_read_unlock();
1104			break;
1105		case (unsigned long) SEND_SIG_PRIV:
1106			clear_siginfo(&q->info);
1107			q->info.si_signo = sig;
1108			q->info.si_errno = 0;
1109			q->info.si_code = SI_KERNEL;
1110			q->info.si_pid = 0;
1111			q->info.si_uid = 0;
1112			break;
1113		default:
1114			copy_siginfo(&q->info, info);
1115			break;
1116		}
1117	} else if (!is_si_special(info) &&
1118		   sig >= SIGRTMIN && info->si_code != SI_USER) {
1119		/*
1120		 * Queue overflow, abort.  We may abort if the
1121		 * signal was rt and sent by user using something
1122		 * other than kill().
1123		 */
1124		result = TRACE_SIGNAL_OVERFLOW_FAIL;
1125		ret = -EAGAIN;
1126		goto ret;
1127	} else {
1128		/*
1129		 * This is a silent loss of information.  We still
1130		 * send the signal, but the *info bits are lost.
1131		 */
1132		result = TRACE_SIGNAL_LOSE_INFO;
1133	}
1134
1135out_set:
1136	signalfd_notify(t, sig);
1137	sigaddset(&pending->signal, sig);
1138
1139	/* Let multiprocess signals appear after on-going forks */
1140	if (type > PIDTYPE_TGID) {
1141		struct multiprocess_signals *delayed;
1142		hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1143			sigset_t *signal = &delayed->signal;
1144			/* Can't queue both a stop and a continue signal */
1145			if (sig == SIGCONT)
1146				sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1147			else if (sig_kernel_stop(sig))
1148				sigdelset(signal, SIGCONT);
1149			sigaddset(signal, sig);
1150		}
1151	}
1152
1153	complete_signal(sig, t, type);
1154ret:
1155	trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1156	return ret;
1157}
1158
1159static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1160{
1161	bool ret = false;
1162	switch (siginfo_layout(info->si_signo, info->si_code)) {
1163	case SIL_KILL:
1164	case SIL_CHLD:
1165	case SIL_RT:
1166		ret = true;
1167		break;
1168	case SIL_TIMER:
1169	case SIL_POLL:
1170	case SIL_FAULT:
1171	case SIL_FAULT_TRAPNO:
1172	case SIL_FAULT_MCEERR:
1173	case SIL_FAULT_BNDERR:
1174	case SIL_FAULT_PKUERR:
1175	case SIL_FAULT_PERF_EVENT:
1176	case SIL_SYS:
1177		ret = false;
1178		break;
1179	}
1180	return ret;
1181}
1182
1183int send_signal_locked(int sig, struct kernel_siginfo *info,
1184		       struct task_struct *t, enum pid_type type)
1185{
1186	/* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1187	bool force = false;
1188
1189	if (info == SEND_SIG_NOINFO) {
1190		/* Force if sent from an ancestor pid namespace */
1191		force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1192	} else if (info == SEND_SIG_PRIV) {
1193		/* Don't ignore kernel generated signals */
1194		force = true;
1195	} else if (has_si_pid_and_uid(info)) {
1196		/* SIGKILL and SIGSTOP is special or has ids */
1197		struct user_namespace *t_user_ns;
1198
1199		rcu_read_lock();
1200		t_user_ns = task_cred_xxx(t, user_ns);
1201		if (current_user_ns() != t_user_ns) {
1202			kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1203			info->si_uid = from_kuid_munged(t_user_ns, uid);
1204		}
1205		rcu_read_unlock();
1206
1207		/* A kernel generated signal? */
1208		force = (info->si_code == SI_KERNEL);
1209
1210		/* From an ancestor pid namespace? */
1211		if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1212			info->si_pid = 0;
1213			force = true;
1214		}
1215	}
1216	return __send_signal_locked(sig, info, t, type, force);
1217}
1218
1219static void print_fatal_signal(int signr)
1220{
1221	struct pt_regs *regs = task_pt_regs(current);
1222	struct file *exe_file;
1223
1224	exe_file = get_task_exe_file(current);
1225	if (exe_file) {
1226		pr_info("%pD: %s: potentially unexpected fatal signal %d.\n",
1227			exe_file, current->comm, signr);
1228		fput(exe_file);
1229	} else {
1230		pr_info("%s: potentially unexpected fatal signal %d.\n",
1231			current->comm, signr);
1232	}
1233
1234#if defined(__i386__) && !defined(__arch_um__)
1235	pr_info("code at %08lx: ", regs->ip);
1236	{
1237		int i;
1238		for (i = 0; i < 16; i++) {
1239			unsigned char insn;
1240
1241			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1242				break;
1243			pr_cont("%02x ", insn);
1244		}
1245	}
1246	pr_cont("\n");
1247#endif
1248	preempt_disable();
1249	show_regs(regs);
1250	preempt_enable();
1251}
1252
1253static int __init setup_print_fatal_signals(char *str)
1254{
1255	get_option (&str, &print_fatal_signals);
1256
1257	return 1;
1258}
1259
1260__setup("print-fatal-signals=", setup_print_fatal_signals);
1261
1262int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1263			enum pid_type type)
1264{
1265	unsigned long flags;
1266	int ret = -ESRCH;
1267
1268	if (lock_task_sighand(p, &flags)) {
1269		ret = send_signal_locked(sig, info, p, type);
1270		unlock_task_sighand(p, &flags);
1271	}
1272
1273	return ret;
1274}
1275
1276enum sig_handler {
1277	HANDLER_CURRENT, /* If reachable use the current handler */
1278	HANDLER_SIG_DFL, /* Always use SIG_DFL handler semantics */
1279	HANDLER_EXIT,	 /* Only visible as the process exit code */
1280};
1281
1282/*
1283 * Force a signal that the process can't ignore: if necessary
1284 * we unblock the signal and change any SIG_IGN to SIG_DFL.
1285 *
1286 * Note: If we unblock the signal, we always reset it to SIG_DFL,
1287 * since we do not want to have a signal handler that was blocked
1288 * be invoked when user space had explicitly blocked it.
1289 *
1290 * We don't want to have recursive SIGSEGV's etc, for example,
1291 * that is why we also clear SIGNAL_UNKILLABLE.
1292 */
1293static int
1294force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t,
1295	enum sig_handler handler)
1296{
1297	unsigned long int flags;
1298	int ret, blocked, ignored;
1299	struct k_sigaction *action;
1300	int sig = info->si_signo;
1301
1302	spin_lock_irqsave(&t->sighand->siglock, flags);
1303	action = &t->sighand->action[sig-1];
1304	ignored = action->sa.sa_handler == SIG_IGN;
1305	blocked = sigismember(&t->blocked, sig);
1306	if (blocked || ignored || (handler != HANDLER_CURRENT)) {
1307		action->sa.sa_handler = SIG_DFL;
1308		if (handler == HANDLER_EXIT)
1309			action->sa.sa_flags |= SA_IMMUTABLE;
1310		if (blocked)
1311			sigdelset(&t->blocked, sig);
1312	}
1313	/*
1314	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1315	 * debugging to leave init killable. But HANDLER_EXIT is always fatal.
1316	 */
1317	if (action->sa.sa_handler == SIG_DFL &&
1318	    (!t->ptrace || (handler == HANDLER_EXIT)))
1319		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1320	ret = send_signal_locked(sig, info, t, PIDTYPE_PID);
1321	/* This can happen if the signal was already pending and blocked */
1322	if (!task_sigpending(t))
1323		signal_wake_up(t, 0);
1324	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1325
1326	return ret;
1327}
1328
1329int force_sig_info(struct kernel_siginfo *info)
1330{
1331	return force_sig_info_to_task(info, current, HANDLER_CURRENT);
1332}
1333
1334/*
1335 * Nuke all other threads in the group.
1336 */
1337int zap_other_threads(struct task_struct *p)
1338{
1339	struct task_struct *t;
1340	int count = 0;
1341
1342	p->signal->group_stop_count = 0;
1343
1344	for_other_threads(p, t) {
1345		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1346		count++;
1347
1348		/* Don't bother with already dead threads */
1349		if (t->exit_state)
1350			continue;
1351		sigaddset(&t->pending.signal, SIGKILL);
1352		signal_wake_up(t, 1);
1353	}
1354
1355	return count;
1356}
1357
1358struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1359					   unsigned long *flags)
1360{
1361	struct sighand_struct *sighand;
1362
1363	rcu_read_lock();
1364	for (;;) {
1365		sighand = rcu_dereference(tsk->sighand);
1366		if (unlikely(sighand == NULL))
1367			break;
1368
1369		/*
1370		 * This sighand can be already freed and even reused, but
1371		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1372		 * initializes ->siglock: this slab can't go away, it has
1373		 * the same object type, ->siglock can't be reinitialized.
1374		 *
1375		 * We need to ensure that tsk->sighand is still the same
1376		 * after we take the lock, we can race with de_thread() or
1377		 * __exit_signal(). In the latter case the next iteration
1378		 * must see ->sighand == NULL.
1379		 */
1380		spin_lock_irqsave(&sighand->siglock, *flags);
1381		if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1382			break;
1383		spin_unlock_irqrestore(&sighand->siglock, *flags);
1384	}
1385	rcu_read_unlock();
1386
1387	return sighand;
1388}
1389
1390#ifdef CONFIG_LOCKDEP
1391void lockdep_assert_task_sighand_held(struct task_struct *task)
1392{
1393	struct sighand_struct *sighand;
1394
1395	rcu_read_lock();
1396	sighand = rcu_dereference(task->sighand);
1397	if (sighand)
1398		lockdep_assert_held(&sighand->siglock);
1399	else
1400		WARN_ON_ONCE(1);
1401	rcu_read_unlock();
1402}
1403#endif
1404
1405/*
1406 * send signal info to all the members of a thread group or to the
1407 * individual thread if type == PIDTYPE_PID.
1408 */
1409int group_send_sig_info(int sig, struct kernel_siginfo *info,
1410			struct task_struct *p, enum pid_type type)
1411{
1412	int ret;
1413
1414	rcu_read_lock();
1415	ret = check_kill_permission(sig, info, p);
1416	rcu_read_unlock();
1417
1418	if (!ret && sig)
1419		ret = do_send_sig_info(sig, info, p, type);
1420
1421	return ret;
1422}
1423
1424/*
1425 * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1426 * control characters do (^C, ^Z etc)
1427 * - the caller must hold at least a readlock on tasklist_lock
1428 */
1429int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1430{
1431	struct task_struct *p = NULL;
1432	int ret = -ESRCH;
1433
1434	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1435		int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1436		/*
1437		 * If group_send_sig_info() succeeds at least once ret
1438		 * becomes 0 and after that the code below has no effect.
1439		 * Otherwise we return the last err or -ESRCH if this
1440		 * process group is empty.
1441		 */
1442		if (ret)
1443			ret = err;
1444	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1445
1446	return ret;
1447}
1448
1449static int kill_pid_info_type(int sig, struct kernel_siginfo *info,
1450				struct pid *pid, enum pid_type type)
1451{
1452	int error = -ESRCH;
1453	struct task_struct *p;
1454
1455	for (;;) {
1456		rcu_read_lock();
1457		p = pid_task(pid, PIDTYPE_PID);
1458		if (p)
1459			error = group_send_sig_info(sig, info, p, type);
1460		rcu_read_unlock();
1461		if (likely(!p || error != -ESRCH))
1462			return error;
1463		/*
1464		 * The task was unhashed in between, try again.  If it
1465		 * is dead, pid_task() will return NULL, if we race with
1466		 * de_thread() it will find the new leader.
1467		 */
1468	}
1469}
1470
1471int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1472{
1473	return kill_pid_info_type(sig, info, pid, PIDTYPE_TGID);
1474}
1475
1476static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1477{
1478	int error;
1479	rcu_read_lock();
1480	error = kill_pid_info(sig, info, find_vpid(pid));
1481	rcu_read_unlock();
1482	return error;
1483}
1484
1485static inline bool kill_as_cred_perm(const struct cred *cred,
1486				     struct task_struct *target)
1487{
1488	const struct cred *pcred = __task_cred(target);
1489
1490	return uid_eq(cred->euid, pcred->suid) ||
1491	       uid_eq(cred->euid, pcred->uid) ||
1492	       uid_eq(cred->uid, pcred->suid) ||
1493	       uid_eq(cred->uid, pcred->uid);
1494}
1495
1496/*
1497 * The usb asyncio usage of siginfo is wrong.  The glibc support
1498 * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1499 * AKA after the generic fields:
1500 *	kernel_pid_t	si_pid;
1501 *	kernel_uid32_t	si_uid;
1502 *	sigval_t	si_value;
1503 *
1504 * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1505 * after the generic fields is:
1506 *	void __user 	*si_addr;
1507 *
1508 * This is a practical problem when there is a 64bit big endian kernel
1509 * and a 32bit userspace.  As the 32bit address will encoded in the low
1510 * 32bits of the pointer.  Those low 32bits will be stored at higher
1511 * address than appear in a 32 bit pointer.  So userspace will not
1512 * see the address it was expecting for it's completions.
1513 *
1514 * There is nothing in the encoding that can allow
1515 * copy_siginfo_to_user32 to detect this confusion of formats, so
1516 * handle this by requiring the caller of kill_pid_usb_asyncio to
1517 * notice when this situration takes place and to store the 32bit
1518 * pointer in sival_int, instead of sival_addr of the sigval_t addr
1519 * parameter.
1520 */
1521int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1522			 struct pid *pid, const struct cred *cred)
1523{
1524	struct kernel_siginfo info;
1525	struct task_struct *p;
1526	unsigned long flags;
1527	int ret = -EINVAL;
1528
1529	if (!valid_signal(sig))
1530		return ret;
1531
1532	clear_siginfo(&info);
1533	info.si_signo = sig;
1534	info.si_errno = errno;
1535	info.si_code = SI_ASYNCIO;
1536	*((sigval_t *)&info.si_pid) = addr;
1537
1538	rcu_read_lock();
1539	p = pid_task(pid, PIDTYPE_PID);
1540	if (!p) {
1541		ret = -ESRCH;
1542		goto out_unlock;
1543	}
1544	if (!kill_as_cred_perm(cred, p)) {
1545		ret = -EPERM;
1546		goto out_unlock;
1547	}
1548	ret = security_task_kill(p, &info, sig, cred);
1549	if (ret)
1550		goto out_unlock;
1551
1552	if (sig) {
1553		if (lock_task_sighand(p, &flags)) {
1554			ret = __send_signal_locked(sig, &info, p, PIDTYPE_TGID, false);
1555			unlock_task_sighand(p, &flags);
1556		} else
1557			ret = -ESRCH;
1558	}
1559out_unlock:
1560	rcu_read_unlock();
1561	return ret;
1562}
1563EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1564
1565/*
1566 * kill_something_info() interprets pid in interesting ways just like kill(2).
1567 *
1568 * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1569 * is probably wrong.  Should make it like BSD or SYSV.
1570 */
1571
1572static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1573{
1574	int ret;
1575
1576	if (pid > 0)
1577		return kill_proc_info(sig, info, pid);
1578
1579	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1580	if (pid == INT_MIN)
1581		return -ESRCH;
1582
1583	read_lock(&tasklist_lock);
1584	if (pid != -1) {
1585		ret = __kill_pgrp_info(sig, info,
1586				pid ? find_vpid(-pid) : task_pgrp(current));
1587	} else {
1588		int retval = 0, count = 0;
1589		struct task_struct * p;
1590
1591		for_each_process(p) {
1592			if (task_pid_vnr(p) > 1 &&
1593					!same_thread_group(p, current)) {
1594				int err = group_send_sig_info(sig, info, p,
1595							      PIDTYPE_MAX);
1596				++count;
1597				if (err != -EPERM)
1598					retval = err;
1599			}
1600		}
1601		ret = count ? retval : -ESRCH;
1602	}
1603	read_unlock(&tasklist_lock);
1604
1605	return ret;
1606}
1607
1608/*
1609 * These are for backward compatibility with the rest of the kernel source.
1610 */
1611
1612int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1613{
1614	/*
1615	 * Make sure legacy kernel users don't send in bad values
1616	 * (normal paths check this in check_kill_permission).
1617	 */
1618	if (!valid_signal(sig))
1619		return -EINVAL;
1620
1621	return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1622}
1623EXPORT_SYMBOL(send_sig_info);
1624
1625#define __si_special(priv) \
1626	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1627
1628int
1629send_sig(int sig, struct task_struct *p, int priv)
1630{
1631	return send_sig_info(sig, __si_special(priv), p);
1632}
1633EXPORT_SYMBOL(send_sig);
1634
1635void force_sig(int sig)
1636{
1637	struct kernel_siginfo info;
1638
1639	clear_siginfo(&info);
1640	info.si_signo = sig;
1641	info.si_errno = 0;
1642	info.si_code = SI_KERNEL;
1643	info.si_pid = 0;
1644	info.si_uid = 0;
1645	force_sig_info(&info);
1646}
1647EXPORT_SYMBOL(force_sig);
1648
1649void force_fatal_sig(int sig)
1650{
1651	struct kernel_siginfo info;
1652
1653	clear_siginfo(&info);
1654	info.si_signo = sig;
1655	info.si_errno = 0;
1656	info.si_code = SI_KERNEL;
1657	info.si_pid = 0;
1658	info.si_uid = 0;
1659	force_sig_info_to_task(&info, current, HANDLER_SIG_DFL);
1660}
1661
1662void force_exit_sig(int sig)
1663{
1664	struct kernel_siginfo info;
1665
1666	clear_siginfo(&info);
1667	info.si_signo = sig;
1668	info.si_errno = 0;
1669	info.si_code = SI_KERNEL;
1670	info.si_pid = 0;
1671	info.si_uid = 0;
1672	force_sig_info_to_task(&info, current, HANDLER_EXIT);
1673}
1674
1675/*
1676 * When things go south during signal handling, we
1677 * will force a SIGSEGV. And if the signal that caused
1678 * the problem was already a SIGSEGV, we'll want to
1679 * make sure we don't even try to deliver the signal..
1680 */
1681void force_sigsegv(int sig)
1682{
1683	if (sig == SIGSEGV)
1684		force_fatal_sig(SIGSEGV);
1685	else
1686		force_sig(SIGSEGV);
1687}
1688
1689int force_sig_fault_to_task(int sig, int code, void __user *addr,
1690			    struct task_struct *t)
1691{
1692	struct kernel_siginfo info;
1693
1694	clear_siginfo(&info);
1695	info.si_signo = sig;
1696	info.si_errno = 0;
1697	info.si_code  = code;
1698	info.si_addr  = addr;
1699	return force_sig_info_to_task(&info, t, HANDLER_CURRENT);
1700}
1701
1702int force_sig_fault(int sig, int code, void __user *addr)
1703{
1704	return force_sig_fault_to_task(sig, code, addr, current);
1705}
1706
1707int send_sig_fault(int sig, int code, void __user *addr, struct task_struct *t)
1708{
1709	struct kernel_siginfo info;
1710
1711	clear_siginfo(&info);
1712	info.si_signo = sig;
1713	info.si_errno = 0;
1714	info.si_code  = code;
1715	info.si_addr  = addr;
1716	return send_sig_info(info.si_signo, &info, t);
1717}
1718
1719int force_sig_mceerr(int code, void __user *addr, short lsb)
1720{
1721	struct kernel_siginfo info;
1722
1723	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1724	clear_siginfo(&info);
1725	info.si_signo = SIGBUS;
1726	info.si_errno = 0;
1727	info.si_code = code;
1728	info.si_addr = addr;
1729	info.si_addr_lsb = lsb;
1730	return force_sig_info(&info);
1731}
1732
1733int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1734{
1735	struct kernel_siginfo info;
1736
1737	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1738	clear_siginfo(&info);
1739	info.si_signo = SIGBUS;
1740	info.si_errno = 0;
1741	info.si_code = code;
1742	info.si_addr = addr;
1743	info.si_addr_lsb = lsb;
1744	return send_sig_info(info.si_signo, &info, t);
1745}
1746EXPORT_SYMBOL(send_sig_mceerr);
1747
1748int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1749{
1750	struct kernel_siginfo info;
1751
1752	clear_siginfo(&info);
1753	info.si_signo = SIGSEGV;
1754	info.si_errno = 0;
1755	info.si_code  = SEGV_BNDERR;
1756	info.si_addr  = addr;
1757	info.si_lower = lower;
1758	info.si_upper = upper;
1759	return force_sig_info(&info);
1760}
1761
1762#ifdef SEGV_PKUERR
1763int force_sig_pkuerr(void __user *addr, u32 pkey)
1764{
1765	struct kernel_siginfo info;
1766
1767	clear_siginfo(&info);
1768	info.si_signo = SIGSEGV;
1769	info.si_errno = 0;
1770	info.si_code  = SEGV_PKUERR;
1771	info.si_addr  = addr;
1772	info.si_pkey  = pkey;
1773	return force_sig_info(&info);
1774}
1775#endif
1776
1777int send_sig_perf(void __user *addr, u32 type, u64 sig_data)
1778{
1779	struct kernel_siginfo info;
1780
1781	clear_siginfo(&info);
1782	info.si_signo     = SIGTRAP;
1783	info.si_errno     = 0;
1784	info.si_code      = TRAP_PERF;
1785	info.si_addr      = addr;
1786	info.si_perf_data = sig_data;
1787	info.si_perf_type = type;
1788
1789	/*
1790	 * Signals generated by perf events should not terminate the whole
1791	 * process if SIGTRAP is blocked, however, delivering the signal
1792	 * asynchronously is better than not delivering at all. But tell user
1793	 * space if the signal was asynchronous, so it can clearly be
1794	 * distinguished from normal synchronous ones.
1795	 */
1796	info.si_perf_flags = sigismember(&current->blocked, info.si_signo) ?
1797				     TRAP_PERF_FLAG_ASYNC :
1798				     0;
1799
1800	return send_sig_info(info.si_signo, &info, current);
1801}
1802
1803/**
1804 * force_sig_seccomp - signals the task to allow in-process syscall emulation
1805 * @syscall: syscall number to send to userland
1806 * @reason: filter-supplied reason code to send to userland (via si_errno)
1807 * @force_coredump: true to trigger a coredump
1808 *
1809 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
1810 */
1811int force_sig_seccomp(int syscall, int reason, bool force_coredump)
1812{
1813	struct kernel_siginfo info;
1814
1815	clear_siginfo(&info);
1816	info.si_signo = SIGSYS;
1817	info.si_code = SYS_SECCOMP;
1818	info.si_call_addr = (void __user *)KSTK_EIP(current);
1819	info.si_errno = reason;
1820	info.si_arch = syscall_get_arch(current);
1821	info.si_syscall = syscall;
1822	return force_sig_info_to_task(&info, current,
1823		force_coredump ? HANDLER_EXIT : HANDLER_CURRENT);
1824}
1825
1826/* For the crazy architectures that include trap information in
1827 * the errno field, instead of an actual errno value.
1828 */
1829int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1830{
1831	struct kernel_siginfo info;
1832
1833	clear_siginfo(&info);
1834	info.si_signo = SIGTRAP;
1835	info.si_errno = errno;
1836	info.si_code  = TRAP_HWBKPT;
1837	info.si_addr  = addr;
1838	return force_sig_info(&info);
1839}
1840
1841/* For the rare architectures that include trap information using
1842 * si_trapno.
1843 */
1844int force_sig_fault_trapno(int sig, int code, void __user *addr, int trapno)
1845{
1846	struct kernel_siginfo info;
1847
1848	clear_siginfo(&info);
1849	info.si_signo = sig;
1850	info.si_errno = 0;
1851	info.si_code  = code;
1852	info.si_addr  = addr;
1853	info.si_trapno = trapno;
1854	return force_sig_info(&info);
1855}
1856
1857/* For the rare architectures that include trap information using
1858 * si_trapno.
1859 */
1860int send_sig_fault_trapno(int sig, int code, void __user *addr, int trapno,
1861			  struct task_struct *t)
1862{
1863	struct kernel_siginfo info;
1864
1865	clear_siginfo(&info);
1866	info.si_signo = sig;
1867	info.si_errno = 0;
1868	info.si_code  = code;
1869	info.si_addr  = addr;
1870	info.si_trapno = trapno;
1871	return send_sig_info(info.si_signo, &info, t);
1872}
1873
1874static int kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1875{
1876	int ret;
1877	read_lock(&tasklist_lock);
1878	ret = __kill_pgrp_info(sig, info, pgrp);
1879	read_unlock(&tasklist_lock);
1880	return ret;
1881}
1882
1883int kill_pgrp(struct pid *pid, int sig, int priv)
1884{
1885	return kill_pgrp_info(sig, __si_special(priv), pid);
1886}
1887EXPORT_SYMBOL(kill_pgrp);
1888
1889int kill_pid(struct pid *pid, int sig, int priv)
1890{
1891	return kill_pid_info(sig, __si_special(priv), pid);
1892}
1893EXPORT_SYMBOL(kill_pid);
1894
1895#ifdef CONFIG_POSIX_TIMERS
1896/*
1897 * These functions handle POSIX timer signals. POSIX timers use
1898 * preallocated sigqueue structs for sending signals.
1899 */
1900static void __flush_itimer_signals(struct sigpending *pending)
1901{
1902	sigset_t signal, retain;
1903	struct sigqueue *q, *n;
1904
1905	signal = pending->signal;
1906	sigemptyset(&retain);
1907
1908	list_for_each_entry_safe(q, n, &pending->list, list) {
1909		int sig = q->info.si_signo;
1910
1911		if (likely(q->info.si_code != SI_TIMER)) {
1912			sigaddset(&retain, sig);
1913		} else {
1914			sigdelset(&signal, sig);
1915			list_del_init(&q->list);
1916			__sigqueue_free(q);
1917		}
1918	}
1919
1920	sigorsets(&pending->signal, &signal, &retain);
1921}
1922
1923void flush_itimer_signals(void)
1924{
1925	struct task_struct *tsk = current;
1926
1927	guard(spinlock_irqsave)(&tsk->sighand->siglock);
1928	__flush_itimer_signals(&tsk->pending);
1929	__flush_itimer_signals(&tsk->signal->shared_pending);
1930}
1931
1932bool posixtimer_init_sigqueue(struct sigqueue *q)
1933{
1934	struct ucounts *ucounts = sig_get_ucounts(current, -1, 0);
1935
1936	if (!ucounts)
1937		return false;
1938	clear_siginfo(&q->info);
1939	__sigqueue_init(q, ucounts, SIGQUEUE_PREALLOC);
1940	return true;
1941}
1942
1943static void posixtimer_queue_sigqueue(struct sigqueue *q, struct task_struct *t, enum pid_type type)
1944{
1945	struct sigpending *pending;
1946	int sig = q->info.si_signo;
1947
1948	signalfd_notify(t, sig);
1949	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1950	list_add_tail(&q->list, &pending->list);
1951	sigaddset(&pending->signal, sig);
1952	complete_signal(sig, t, type);
1953}
1954
1955/*
1956 * This function is used by POSIX timers to deliver a timer signal.
1957 * Where type is PIDTYPE_PID (such as for timers with SIGEV_THREAD_ID
1958 * set), the signal must be delivered to the specific thread (queues
1959 * into t->pending).
1960 *
1961 * Where type is not PIDTYPE_PID, signals must be delivered to the
1962 * process. In this case, prefer to deliver to current if it is in
1963 * the same thread group as the target process and its sighand is
1964 * stable, which avoids unnecessarily waking up a potentially idle task.
1965 */
1966static inline struct task_struct *posixtimer_get_target(struct k_itimer *tmr)
1967{
1968	struct task_struct *t = pid_task(tmr->it_pid, tmr->it_pid_type);
1969
1970	if (t && tmr->it_pid_type != PIDTYPE_PID &&
1971	    same_thread_group(t, current) && !current->exit_state)
1972		t = current;
1973	return t;
1974}
1975
1976void posixtimer_send_sigqueue(struct k_itimer *tmr)
1977{
1978	struct sigqueue *q = &tmr->sigq;
1979	int sig = q->info.si_signo;
1980	struct task_struct *t;
1981	unsigned long flags;
1982	int result;
1983
1984	guard(rcu)();
1985
1986	t = posixtimer_get_target(tmr);
1987	if (!t)
1988		return;
1989
1990	if (!likely(lock_task_sighand(t, &flags)))
1991		return;
1992
1993	/*
1994	 * Update @tmr::sigqueue_seq for posix timer signals with sighand
1995	 * locked to prevent a race against dequeue_signal().
1996	 */
1997	tmr->it_sigqueue_seq = tmr->it_signal_seq;
1998
1999	/*
2000	 * Set the signal delivery status under sighand lock, so that the
2001	 * ignored signal handling can distinguish between a periodic and a
2002	 * non-periodic timer.
2003	 */
2004	tmr->it_sig_periodic = tmr->it_status == POSIX_TIMER_REQUEUE_PENDING;
2005
2006	if (!prepare_signal(sig, t, false)) {
2007		result = TRACE_SIGNAL_IGNORED;
2008
2009		if (!list_empty(&q->list)) {
2010			/*
2011			 * The signal was ignored and blocked. The timer
2012			 * expiry queued it because blocked signals are
2013			 * queued independent of the ignored state.
2014			 *
2015			 * The unblocking set SIGPENDING, but the signal
2016			 * was not yet dequeued from the pending list.
2017			 * So prepare_signal() sees unblocked and ignored,
2018			 * which ends up here. Leave it queued like a
2019			 * regular signal.
2020			 *
2021			 * The same happens when the task group is exiting
2022			 * and the signal is already queued.
2023			 * prepare_signal() treats SIGNAL_GROUP_EXIT as
2024			 * ignored independent of its queued state. This
2025			 * gets cleaned up in __exit_signal().
2026			 */
2027			goto out;
2028		}
2029
2030		/* Periodic timers with SIG_IGN are queued on the ignored list */
2031		if (tmr->it_sig_periodic) {
2032			/*
2033			 * Already queued means the timer was rearmed after
2034			 * the previous expiry got it on the ignore list.
2035			 * Nothing to do for that case.
2036			 */
2037			if (hlist_unhashed(&tmr->ignored_list)) {
2038				/*
2039				 * Take a signal reference and queue it on
2040				 * the ignored list.
2041				 */
2042				posixtimer_sigqueue_getref(q);
2043				posixtimer_sig_ignore(t, q);
2044			}
2045		} else if (!hlist_unhashed(&tmr->ignored_list)) {
2046			/*
2047			 * Covers the case where a timer was periodic and
2048			 * then the signal was ignored. Later it was rearmed
2049			 * as oneshot timer. The previous signal is invalid
2050			 * now, and this oneshot signal has to be dropped.
2051			 * Remove it from the ignored list and drop the
2052			 * reference count as the signal is not longer
2053			 * queued.
2054			 */
2055			hlist_del_init(&tmr->ignored_list);
2056			posixtimer_putref(tmr);
2057		}
2058		goto out;
2059	}
2060
2061	if (unlikely(!list_empty(&q->list))) {
2062		/* This holds a reference count already */
2063		result = TRACE_SIGNAL_ALREADY_PENDING;
2064		goto out;
2065	}
2066
2067	/*
2068	 * If the signal is on the ignore list, it got blocked after it was
2069	 * ignored earlier. But nothing lifted the ignore. Move it back to
2070	 * the pending list to be consistent with the regular signal
2071	 * handling. This already holds a reference count.
2072	 *
2073	 * If it's not on the ignore list acquire a reference count.
2074	 */
2075	if (likely(hlist_unhashed(&tmr->ignored_list)))
2076		posixtimer_sigqueue_getref(q);
2077	else
2078		hlist_del_init(&tmr->ignored_list);
2079
2080	posixtimer_queue_sigqueue(q, t, tmr->it_pid_type);
2081	result = TRACE_SIGNAL_DELIVERED;
2082out:
2083	trace_signal_generate(sig, &q->info, t, tmr->it_pid_type != PIDTYPE_PID, result);
2084	unlock_task_sighand(t, &flags);
2085}
2086
2087static inline void posixtimer_sig_ignore(struct task_struct *tsk, struct sigqueue *q)
2088{
2089	struct k_itimer *tmr = container_of(q, struct k_itimer, sigq);
2090
2091	/*
2092	 * If the timer is marked deleted already or the signal originates
2093	 * from a non-periodic timer, then just drop the reference
2094	 * count. Otherwise queue it on the ignored list.
2095	 */
2096	if (posixtimer_valid(tmr) && tmr->it_sig_periodic)
2097		hlist_add_head(&tmr->ignored_list, &tsk->signal->ignored_posix_timers);
2098	else
2099		posixtimer_putref(tmr);
2100}
2101
2102static void posixtimer_sig_unignore(struct task_struct *tsk, int sig)
2103{
2104	struct hlist_head *head = &tsk->signal->ignored_posix_timers;
2105	struct hlist_node *tmp;
2106	struct k_itimer *tmr;
2107
2108	if (likely(hlist_empty(head)))
2109		return;
2110
2111	/*
2112	 * Rearming a timer with sighand lock held is not possible due to
2113	 * lock ordering vs. tmr::it_lock. Just stick the sigqueue back and
2114	 * let the signal delivery path deal with it whether it needs to be
2115	 * rearmed or not. This cannot be decided here w/o dropping sighand
2116	 * lock and creating a loop retry horror show.
2117	 */
2118	hlist_for_each_entry_safe(tmr, tmp , head, ignored_list) {
2119		struct task_struct *target;
2120
2121		/*
2122		 * tmr::sigq.info.si_signo is immutable, so accessing it
2123		 * without holding tmr::it_lock is safe.
2124		 */
2125		if (tmr->sigq.info.si_signo != sig)
2126			continue;
2127
2128		hlist_del_init(&tmr->ignored_list);
2129
2130		/* This should never happen and leaks a reference count */
2131		if (WARN_ON_ONCE(!list_empty(&tmr->sigq.list)))
2132			continue;
2133
2134		/*
2135		 * Get the target for the signal. If target is a thread and
2136		 * has exited by now, drop the reference count.
2137		 */
2138		guard(rcu)();
2139		target = posixtimer_get_target(tmr);
2140		if (target)
2141			posixtimer_queue_sigqueue(&tmr->sigq, target, tmr->it_pid_type);
2142		else
2143			posixtimer_putref(tmr);
2144	}
2145}
2146#else /* CONFIG_POSIX_TIMERS */
2147static inline void posixtimer_sig_ignore(struct task_struct *tsk, struct sigqueue *q) { }
2148static inline void posixtimer_sig_unignore(struct task_struct *tsk, int sig) { }
2149#endif /* !CONFIG_POSIX_TIMERS */
2150
2151void do_notify_pidfd(struct task_struct *task)
2152{
2153	struct pid *pid = task_pid(task);
2154
2155	WARN_ON(task->exit_state == 0);
2156
2157	__wake_up(&pid->wait_pidfd, TASK_NORMAL, 0,
2158			poll_to_key(EPOLLIN | EPOLLRDNORM));
2159}
2160
2161/*
2162 * Let a parent know about the death of a child.
2163 * For a stopped/continued status change, use do_notify_parent_cldstop instead.
2164 *
2165 * Returns true if our parent ignored us and so we've switched to
2166 * self-reaping.
2167 */
2168bool do_notify_parent(struct task_struct *tsk, int sig)
2169{
2170	struct kernel_siginfo info;
2171	unsigned long flags;
2172	struct sighand_struct *psig;
2173	bool autoreap = false;
2174	u64 utime, stime;
2175
2176	WARN_ON_ONCE(sig == -1);
2177
2178	/* do_notify_parent_cldstop should have been called instead.  */
2179	WARN_ON_ONCE(task_is_stopped_or_traced(tsk));
2180
2181	WARN_ON_ONCE(!tsk->ptrace &&
2182	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
2183
2184	/* ptraced, or group-leader without sub-threads */
2185	do_notify_pidfd(tsk);
2186
2187	if (sig != SIGCHLD) {
2188		/*
2189		 * This is only possible if parent == real_parent.
2190		 * Check if it has changed security domain.
2191		 */
2192		if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
2193			sig = SIGCHLD;
2194	}
2195
2196	clear_siginfo(&info);
2197	info.si_signo = sig;
2198	info.si_errno = 0;
2199	/*
2200	 * We are under tasklist_lock here so our parent is tied to
2201	 * us and cannot change.
2202	 *
2203	 * task_active_pid_ns will always return the same pid namespace
2204	 * until a task passes through release_task.
2205	 *
2206	 * write_lock() currently calls preempt_disable() which is the
2207	 * same as rcu_read_lock(), but according to Oleg, this is not
2208	 * correct to rely on this
2209	 */
2210	rcu_read_lock();
2211	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
2212	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
2213				       task_uid(tsk));
2214	rcu_read_unlock();
2215
2216	task_cputime(tsk, &utime, &stime);
2217	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
2218	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
2219
2220	info.si_status = tsk->exit_code & 0x7f;
2221	if (tsk->exit_code & 0x80)
2222		info.si_code = CLD_DUMPED;
2223	else if (tsk->exit_code & 0x7f)
2224		info.si_code = CLD_KILLED;
2225	else {
2226		info.si_code = CLD_EXITED;
2227		info.si_status = tsk->exit_code >> 8;
2228	}
2229
2230	psig = tsk->parent->sighand;
2231	spin_lock_irqsave(&psig->siglock, flags);
2232	if (!tsk->ptrace && sig == SIGCHLD &&
2233	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
2234	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
2235		/*
2236		 * We are exiting and our parent doesn't care.  POSIX.1
2237		 * defines special semantics for setting SIGCHLD to SIG_IGN
2238		 * or setting the SA_NOCLDWAIT flag: we should be reaped
2239		 * automatically and not left for our parent's wait4 call.
2240		 * Rather than having the parent do it as a magic kind of
2241		 * signal handler, we just set this to tell do_exit that we
2242		 * can be cleaned up without becoming a zombie.  Note that
2243		 * we still call __wake_up_parent in this case, because a
2244		 * blocked sys_wait4 might now return -ECHILD.
2245		 *
2246		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
2247		 * is implementation-defined: we do (if you don't want
2248		 * it, just use SIG_IGN instead).
2249		 */
2250		autoreap = true;
2251		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2252			sig = 0;
2253	}
2254	/*
2255	 * Send with __send_signal as si_pid and si_uid are in the
2256	 * parent's namespaces.
2257	 */
2258	if (valid_signal(sig) && sig)
2259		__send_signal_locked(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2260	__wake_up_parent(tsk, tsk->parent);
2261	spin_unlock_irqrestore(&psig->siglock, flags);
2262
2263	return autoreap;
2264}
2265
2266/**
2267 * do_notify_parent_cldstop - notify parent of stopped/continued state change
2268 * @tsk: task reporting the state change
2269 * @for_ptracer: the notification is for ptracer
2270 * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2271 *
2272 * Notify @tsk's parent that the stopped/continued state has changed.  If
2273 * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2274 * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2275 *
2276 * CONTEXT:
2277 * Must be called with tasklist_lock at least read locked.
2278 */
2279static void do_notify_parent_cldstop(struct task_struct *tsk,
2280				     bool for_ptracer, int why)
2281{
2282	struct kernel_siginfo info;
2283	unsigned long flags;
2284	struct task_struct *parent;
2285	struct sighand_struct *sighand;
2286	u64 utime, stime;
2287
2288	if (for_ptracer) {
2289		parent = tsk->parent;
2290	} else {
2291		tsk = tsk->group_leader;
2292		parent = tsk->real_parent;
2293	}
2294
2295	clear_siginfo(&info);
2296	info.si_signo = SIGCHLD;
2297	info.si_errno = 0;
2298	/*
2299	 * see comment in do_notify_parent() about the following 4 lines
2300	 */
2301	rcu_read_lock();
2302	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2303	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2304	rcu_read_unlock();
2305
2306	task_cputime(tsk, &utime, &stime);
2307	info.si_utime = nsec_to_clock_t(utime);
2308	info.si_stime = nsec_to_clock_t(stime);
2309
2310 	info.si_code = why;
2311 	switch (why) {
2312 	case CLD_CONTINUED:
2313 		info.si_status = SIGCONT;
2314 		break;
2315 	case CLD_STOPPED:
2316 		info.si_status = tsk->signal->group_exit_code & 0x7f;
2317 		break;
2318 	case CLD_TRAPPED:
2319 		info.si_status = tsk->exit_code & 0x7f;
2320 		break;
2321 	default:
2322 		BUG();
2323 	}
2324
2325	sighand = parent->sighand;
2326	spin_lock_irqsave(&sighand->siglock, flags);
2327	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2328	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2329		send_signal_locked(SIGCHLD, &info, parent, PIDTYPE_TGID);
2330	/*
2331	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2332	 */
2333	__wake_up_parent(tsk, parent);
2334	spin_unlock_irqrestore(&sighand->siglock, flags);
2335}
2336
2337/*
2338 * This must be called with current->sighand->siglock held.
2339 *
2340 * This should be the path for all ptrace stops.
2341 * We always set current->last_siginfo while stopped here.
2342 * That makes it a way to test a stopped process for
2343 * being ptrace-stopped vs being job-control-stopped.
2344 *
2345 * Returns the signal the ptracer requested the code resume
2346 * with.  If the code did not stop because the tracer is gone,
2347 * the stop signal remains unchanged unless clear_code.
2348 */
2349static int ptrace_stop(int exit_code, int why, unsigned long message,
2350		       kernel_siginfo_t *info)
2351	__releases(&current->sighand->siglock)
2352	__acquires(&current->sighand->siglock)
2353{
2354	bool gstop_done = false;
2355
2356	if (arch_ptrace_stop_needed()) {
2357		/*
2358		 * The arch code has something special to do before a
2359		 * ptrace stop.  This is allowed to block, e.g. for faults
2360		 * on user stack pages.  We can't keep the siglock while
2361		 * calling arch_ptrace_stop, so we must release it now.
2362		 * To preserve proper semantics, we must do this before
2363		 * any signal bookkeeping like checking group_stop_count.
2364		 */
2365		spin_unlock_irq(&current->sighand->siglock);
2366		arch_ptrace_stop();
2367		spin_lock_irq(&current->sighand->siglock);
2368	}
2369
2370	/*
2371	 * After this point ptrace_signal_wake_up or signal_wake_up
2372	 * will clear TASK_TRACED if ptrace_unlink happens or a fatal
2373	 * signal comes in.  Handle previous ptrace_unlinks and fatal
2374	 * signals here to prevent ptrace_stop sleeping in schedule.
2375	 */
2376	if (!current->ptrace || __fatal_signal_pending(current))
2377		return exit_code;
2378
2379	set_special_state(TASK_TRACED);
2380	current->jobctl |= JOBCTL_TRACED;
2381
2382	/*
2383	 * We're committing to trapping.  TRACED should be visible before
2384	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2385	 * Also, transition to TRACED and updates to ->jobctl should be
2386	 * atomic with respect to siglock and should be done after the arch
2387	 * hook as siglock is released and regrabbed across it.
2388	 *
2389	 *     TRACER				    TRACEE
2390	 *
2391	 *     ptrace_attach()
2392	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2393	 *     do_wait()
2394	 *       set_current_state()                smp_wmb();
2395	 *       ptrace_do_wait()
2396	 *         wait_task_stopped()
2397	 *           task_stopped_code()
2398	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2399	 */
2400	smp_wmb();
2401
2402	current->ptrace_message = message;
2403	current->last_siginfo = info;
2404	current->exit_code = exit_code;
2405
2406	/*
2407	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2408	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2409	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2410	 * could be clear now.  We act as if SIGCONT is received after
2411	 * TASK_TRACED is entered - ignore it.
2412	 */
2413	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2414		gstop_done = task_participate_group_stop(current);
2415
2416	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2417	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2418	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2419		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2420
2421	/* entering a trap, clear TRAPPING */
2422	task_clear_jobctl_trapping(current);
2423
2424	spin_unlock_irq(&current->sighand->siglock);
2425	read_lock(&tasklist_lock);
2426	/*
2427	 * Notify parents of the stop.
2428	 *
2429	 * While ptraced, there are two parents - the ptracer and
2430	 * the real_parent of the group_leader.  The ptracer should
2431	 * know about every stop while the real parent is only
2432	 * interested in the completion of group stop.  The states
2433	 * for the two don't interact with each other.  Notify
2434	 * separately unless they're gonna be duplicates.
2435	 */
2436	if (current->ptrace)
2437		do_notify_parent_cldstop(current, true, why);
2438	if (gstop_done && (!current->ptrace || ptrace_reparented(current)))
2439		do_notify_parent_cldstop(current, false, why);
2440
2441	/*
2442	 * The previous do_notify_parent_cldstop() invocation woke ptracer.
2443	 * One a PREEMPTION kernel this can result in preemption requirement
2444	 * which will be fulfilled after read_unlock() and the ptracer will be
2445	 * put on the CPU.
2446	 * The ptracer is in wait_task_inactive(, __TASK_TRACED) waiting for
2447	 * this task wait in schedule(). If this task gets preempted then it
2448	 * remains enqueued on the runqueue. The ptracer will observe this and
2449	 * then sleep for a delay of one HZ tick. In the meantime this task
2450	 * gets scheduled, enters schedule() and will wait for the ptracer.
2451	 *
2452	 * This preemption point is not bad from a correctness point of
2453	 * view but extends the runtime by one HZ tick time due to the
2454	 * ptracer's sleep.  The preempt-disable section ensures that there
2455	 * will be no preemption between unlock and schedule() and so
2456	 * improving the performance since the ptracer will observe that
2457	 * the tracee is scheduled out once it gets on the CPU.
2458	 *
2459	 * On PREEMPT_RT locking tasklist_lock does not disable preemption.
2460	 * Therefore the task can be preempted after do_notify_parent_cldstop()
2461	 * before unlocking tasklist_lock so there is no benefit in doing this.
2462	 *
2463	 * In fact disabling preemption is harmful on PREEMPT_RT because
2464	 * the spinlock_t in cgroup_enter_frozen() must not be acquired
2465	 * with preemption disabled due to the 'sleeping' spinlock
2466	 * substitution of RT.
2467	 */
2468	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
2469		preempt_disable();
2470	read_unlock(&tasklist_lock);
2471	cgroup_enter_frozen();
2472	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
2473		preempt_enable_no_resched();
2474	schedule();
2475	cgroup_leave_frozen(true);
2476
2477	/*
2478	 * We are back.  Now reacquire the siglock before touching
2479	 * last_siginfo, so that we are sure to have synchronized with
2480	 * any signal-sending on another CPU that wants to examine it.
2481	 */
2482	spin_lock_irq(&current->sighand->siglock);
2483	exit_code = current->exit_code;
2484	current->last_siginfo = NULL;
2485	current->ptrace_message = 0;
2486	current->exit_code = 0;
2487
2488	/* LISTENING can be set only during STOP traps, clear it */
2489	current->jobctl &= ~(JOBCTL_LISTENING | JOBCTL_PTRACE_FROZEN);
2490
2491	/*
2492	 * Queued signals ignored us while we were stopped for tracing.
2493	 * So check for any that we should take before resuming user mode.
2494	 * This sets TIF_SIGPENDING, but never clears it.
2495	 */
2496	recalc_sigpending_tsk(current);
2497	return exit_code;
2498}
2499
2500static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
2501{
2502	kernel_siginfo_t info;
2503
2504	clear_siginfo(&info);
2505	info.si_signo = signr;
2506	info.si_code = exit_code;
2507	info.si_pid = task_pid_vnr(current);
2508	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2509
2510	/* Let the debugger run.  */
2511	return ptrace_stop(exit_code, why, message, &info);
2512}
2513
2514int ptrace_notify(int exit_code, unsigned long message)
2515{
2516	int signr;
2517
2518	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2519	if (unlikely(task_work_pending(current)))
2520		task_work_run();
2521
2522	spin_lock_irq(&current->sighand->siglock);
2523	signr = ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
2524	spin_unlock_irq(&current->sighand->siglock);
2525	return signr;
2526}
2527
2528/**
2529 * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2530 * @signr: signr causing group stop if initiating
2531 *
2532 * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2533 * and participate in it.  If already set, participate in the existing
2534 * group stop.  If participated in a group stop (and thus slept), %true is
2535 * returned with siglock released.
2536 *
2537 * If ptraced, this function doesn't handle stop itself.  Instead,
2538 * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2539 * untouched.  The caller must ensure that INTERRUPT trap handling takes
2540 * places afterwards.
2541 *
2542 * CONTEXT:
2543 * Must be called with @current->sighand->siglock held, which is released
2544 * on %true return.
2545 *
2546 * RETURNS:
2547 * %false if group stop is already cancelled or ptrace trap is scheduled.
2548 * %true if participated in group stop.
2549 */
2550static bool do_signal_stop(int signr)
2551	__releases(&current->sighand->siglock)
2552{
2553	struct signal_struct *sig = current->signal;
2554
2555	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2556		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2557		struct task_struct *t;
2558
2559		/* signr will be recorded in task->jobctl for retries */
2560		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2561
2562		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2563		    unlikely(sig->flags & SIGNAL_GROUP_EXIT) ||
2564		    unlikely(sig->group_exec_task))
2565			return false;
2566		/*
2567		 * There is no group stop already in progress.  We must
2568		 * initiate one now.
2569		 *
2570		 * While ptraced, a task may be resumed while group stop is
2571		 * still in effect and then receive a stop signal and
2572		 * initiate another group stop.  This deviates from the
2573		 * usual behavior as two consecutive stop signals can't
2574		 * cause two group stops when !ptraced.  That is why we
2575		 * also check !task_is_stopped(t) below.
2576		 *
2577		 * The condition can be distinguished by testing whether
2578		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2579		 * group_exit_code in such case.
2580		 *
2581		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2582		 * an intervening stop signal is required to cause two
2583		 * continued events regardless of ptrace.
2584		 */
2585		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2586			sig->group_exit_code = signr;
2587
2588		sig->group_stop_count = 0;
2589		if (task_set_jobctl_pending(current, signr | gstop))
2590			sig->group_stop_count++;
2591
2592		for_other_threads(current, t) {
2593			/*
2594			 * Setting state to TASK_STOPPED for a group
2595			 * stop is always done with the siglock held,
2596			 * so this check has no races.
2597			 */
2598			if (!task_is_stopped(t) &&
2599			    task_set_jobctl_pending(t, signr | gstop)) {
2600				sig->group_stop_count++;
2601				if (likely(!(t->ptrace & PT_SEIZED)))
2602					signal_wake_up(t, 0);
2603				else
2604					ptrace_trap_notify(t);
2605			}
2606		}
2607	}
2608
2609	if (likely(!current->ptrace)) {
2610		int notify = 0;
2611
2612		/*
2613		 * If there are no other threads in the group, or if there
2614		 * is a group stop in progress and we are the last to stop,
2615		 * report to the parent.
2616		 */
2617		if (task_participate_group_stop(current))
2618			notify = CLD_STOPPED;
2619
2620		current->jobctl |= JOBCTL_STOPPED;
2621		set_special_state(TASK_STOPPED);
2622		spin_unlock_irq(&current->sighand->siglock);
2623
2624		/*
2625		 * Notify the parent of the group stop completion.  Because
2626		 * we're not holding either the siglock or tasklist_lock
2627		 * here, ptracer may attach inbetween; however, this is for
2628		 * group stop and should always be delivered to the real
2629		 * parent of the group leader.  The new ptracer will get
2630		 * its notification when this task transitions into
2631		 * TASK_TRACED.
2632		 */
2633		if (notify) {
2634			read_lock(&tasklist_lock);
2635			do_notify_parent_cldstop(current, false, notify);
2636			read_unlock(&tasklist_lock);
2637		}
2638
2639		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2640		cgroup_enter_frozen();
2641		schedule();
2642		return true;
2643	} else {
2644		/*
2645		 * While ptraced, group stop is handled by STOP trap.
2646		 * Schedule it and let the caller deal with it.
2647		 */
2648		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2649		return false;
2650	}
2651}
2652
2653/**
2654 * do_jobctl_trap - take care of ptrace jobctl traps
2655 *
2656 * When PT_SEIZED, it's used for both group stop and explicit
2657 * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2658 * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2659 * the stop signal; otherwise, %SIGTRAP.
2660 *
2661 * When !PT_SEIZED, it's used only for group stop trap with stop signal
2662 * number as exit_code and no siginfo.
2663 *
2664 * CONTEXT:
2665 * Must be called with @current->sighand->siglock held, which may be
2666 * released and re-acquired before returning with intervening sleep.
2667 */
2668static void do_jobctl_trap(void)
2669{
2670	struct signal_struct *signal = current->signal;
2671	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2672
2673	if (current->ptrace & PT_SEIZED) {
2674		if (!signal->group_stop_count &&
2675		    !(signal->flags & SIGNAL_STOP_STOPPED))
2676			signr = SIGTRAP;
2677		WARN_ON_ONCE(!signr);
2678		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2679				 CLD_STOPPED, 0);
2680	} else {
2681		WARN_ON_ONCE(!signr);
2682		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2683	}
2684}
2685
2686/**
2687 * do_freezer_trap - handle the freezer jobctl trap
2688 *
2689 * Puts the task into frozen state, if only the task is not about to quit.
2690 * In this case it drops JOBCTL_TRAP_FREEZE.
2691 *
2692 * CONTEXT:
2693 * Must be called with @current->sighand->siglock held,
2694 * which is always released before returning.
2695 */
2696static void do_freezer_trap(void)
2697	__releases(&current->sighand->siglock)
2698{
2699	/*
2700	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2701	 * let's make another loop to give it a chance to be handled.
2702	 * In any case, we'll return back.
2703	 */
2704	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2705	     JOBCTL_TRAP_FREEZE) {
2706		spin_unlock_irq(&current->sighand->siglock);
2707		return;
2708	}
2709
2710	/*
2711	 * Now we're sure that there is no pending fatal signal and no
2712	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2713	 * immediately (if there is a non-fatal signal pending), and
2714	 * put the task into sleep.
2715	 */
2716	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
2717	clear_thread_flag(TIF_SIGPENDING);
2718	spin_unlock_irq(&current->sighand->siglock);
2719	cgroup_enter_frozen();
2720	schedule();
2721
2722	/*
2723	 * We could've been woken by task_work, run it to clear
2724	 * TIF_NOTIFY_SIGNAL. The caller will retry if necessary.
2725	 */
2726	clear_notify_signal();
2727	if (unlikely(task_work_pending(current)))
2728		task_work_run();
2729}
2730
2731static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
2732{
2733	/*
2734	 * We do not check sig_kernel_stop(signr) but set this marker
2735	 * unconditionally because we do not know whether debugger will
2736	 * change signr. This flag has no meaning unless we are going
2737	 * to stop after return from ptrace_stop(). In this case it will
2738	 * be checked in do_signal_stop(), we should only stop if it was
2739	 * not cleared by SIGCONT while we were sleeping. See also the
2740	 * comment in dequeue_signal().
2741	 */
2742	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2743	signr = ptrace_stop(signr, CLD_TRAPPED, 0, info);
2744
2745	/* We're back.  Did the debugger cancel the sig?  */
2746	if (signr == 0)
2747		return signr;
2748
2749	/*
2750	 * Update the siginfo structure if the signal has
2751	 * changed.  If the debugger wanted something
2752	 * specific in the siginfo structure then it should
2753	 * have updated *info via PTRACE_SETSIGINFO.
2754	 */
2755	if (signr != info->si_signo) {
2756		clear_siginfo(info);
2757		info->si_signo = signr;
2758		info->si_errno = 0;
2759		info->si_code = SI_USER;
2760		rcu_read_lock();
2761		info->si_pid = task_pid_vnr(current->parent);
2762		info->si_uid = from_kuid_munged(current_user_ns(),
2763						task_uid(current->parent));
2764		rcu_read_unlock();
2765	}
2766
2767	/* If the (new) signal is now blocked, requeue it.  */
2768	if (sigismember(&current->blocked, signr) ||
2769	    fatal_signal_pending(current)) {
2770		send_signal_locked(signr, info, current, type);
2771		signr = 0;
2772	}
2773
2774	return signr;
2775}
2776
2777static void hide_si_addr_tag_bits(struct ksignal *ksig)
2778{
2779	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2780	case SIL_FAULT:
2781	case SIL_FAULT_TRAPNO:
2782	case SIL_FAULT_MCEERR:
2783	case SIL_FAULT_BNDERR:
2784	case SIL_FAULT_PKUERR:
2785	case SIL_FAULT_PERF_EVENT:
2786		ksig->info.si_addr = arch_untagged_si_addr(
2787			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2788		break;
2789	case SIL_KILL:
2790	case SIL_TIMER:
2791	case SIL_POLL:
2792	case SIL_CHLD:
2793	case SIL_RT:
2794	case SIL_SYS:
2795		break;
2796	}
2797}
2798
2799bool get_signal(struct ksignal *ksig)
2800{
2801	struct sighand_struct *sighand = current->sighand;
2802	struct signal_struct *signal = current->signal;
2803	int signr;
2804
2805	clear_notify_signal();
2806	if (unlikely(task_work_pending(current)))
2807		task_work_run();
2808
2809	if (!task_sigpending(current))
2810		return false;
2811
2812	if (unlikely(uprobe_deny_signal()))
2813		return false;
2814
2815	/*
2816	 * Do this once, we can't return to user-mode if freezing() == T.
2817	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2818	 * thus do not need another check after return.
2819	 */
2820	try_to_freeze();
2821
2822relock:
2823	spin_lock_irq(&sighand->siglock);
2824
2825	/*
2826	 * Every stopped thread goes here after wakeup. Check to see if
2827	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2828	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2829	 */
2830	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2831		int why;
2832
2833		if (signal->flags & SIGNAL_CLD_CONTINUED)
2834			why = CLD_CONTINUED;
2835		else
2836			why = CLD_STOPPED;
2837
2838		signal->flags &= ~SIGNAL_CLD_MASK;
2839
2840		spin_unlock_irq(&sighand->siglock);
2841
2842		/*
2843		 * Notify the parent that we're continuing.  This event is
2844		 * always per-process and doesn't make whole lot of sense
2845		 * for ptracers, who shouldn't consume the state via
2846		 * wait(2) either, but, for backward compatibility, notify
2847		 * the ptracer of the group leader too unless it's gonna be
2848		 * a duplicate.
2849		 */
2850		read_lock(&tasklist_lock);
2851		do_notify_parent_cldstop(current, false, why);
2852
2853		if (ptrace_reparented(current->group_leader))
2854			do_notify_parent_cldstop(current->group_leader,
2855						true, why);
2856		read_unlock(&tasklist_lock);
2857
2858		goto relock;
2859	}
2860
2861	for (;;) {
2862		struct k_sigaction *ka;
2863		enum pid_type type;
2864
2865		/* Has this task already been marked for death? */
2866		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
2867		     signal->group_exec_task) {
2868			signr = SIGKILL;
2869			sigdelset(&current->pending.signal, SIGKILL);
2870			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2871					     &sighand->action[SIGKILL-1]);
2872			recalc_sigpending();
2873			/*
2874			 * implies do_group_exit() or return to PF_USER_WORKER,
2875			 * no need to initialize ksig->info/etc.
2876			 */
2877			goto fatal;
2878		}
2879
2880		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2881		    do_signal_stop(0))
2882			goto relock;
2883
2884		if (unlikely(current->jobctl &
2885			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2886			if (current->jobctl & JOBCTL_TRAP_MASK) {
2887				do_jobctl_trap();
2888				spin_unlock_irq(&sighand->siglock);
2889			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2890				do_freezer_trap();
2891
2892			goto relock;
2893		}
2894
2895		/*
2896		 * If the task is leaving the frozen state, let's update
2897		 * cgroup counters and reset the frozen bit.
2898		 */
2899		if (unlikely(cgroup_task_frozen(current))) {
2900			spin_unlock_irq(&sighand->siglock);
2901			cgroup_leave_frozen(false);
2902			goto relock;
2903		}
2904
2905		/*
2906		 * Signals generated by the execution of an instruction
2907		 * need to be delivered before any other pending signals
2908		 * so that the instruction pointer in the signal stack
2909		 * frame points to the faulting instruction.
2910		 */
2911		type = PIDTYPE_PID;
2912		signr = dequeue_synchronous_signal(&ksig->info);
2913		if (!signr)
2914			signr = dequeue_signal(&current->blocked, &ksig->info, &type);
2915
2916		if (!signr)
2917			break; /* will return 0 */
2918
2919		if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2920		    !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2921			signr = ptrace_signal(signr, &ksig->info, type);
2922			if (!signr)
2923				continue;
2924		}
2925
2926		ka = &sighand->action[signr-1];
2927
2928		/* Trace actually delivered signals. */
2929		trace_signal_deliver(signr, &ksig->info, ka);
2930
2931		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2932			continue;
2933		if (ka->sa.sa_handler != SIG_DFL) {
2934			/* Run the handler.  */
2935			ksig->ka = *ka;
2936
2937			if (ka->sa.sa_flags & SA_ONESHOT)
2938				ka->sa.sa_handler = SIG_DFL;
2939
2940			break; /* will return non-zero "signr" value */
2941		}
2942
2943		/*
2944		 * Now we are doing the default action for this signal.
2945		 */
2946		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2947			continue;
2948
2949		/*
2950		 * Global init gets no signals it doesn't want.
2951		 * Container-init gets no signals it doesn't want from same
2952		 * container.
2953		 *
2954		 * Note that if global/container-init sees a sig_kernel_only()
2955		 * signal here, the signal must have been generated internally
2956		 * or must have come from an ancestor namespace. In either
2957		 * case, the signal cannot be dropped.
2958		 */
2959		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2960				!sig_kernel_only(signr))
2961			continue;
2962
2963		if (sig_kernel_stop(signr)) {
2964			/*
2965			 * The default action is to stop all threads in
2966			 * the thread group.  The job control signals
2967			 * do nothing in an orphaned pgrp, but SIGSTOP
2968			 * always works.  Note that siglock needs to be
2969			 * dropped during the call to is_orphaned_pgrp()
2970			 * because of lock ordering with tasklist_lock.
2971			 * This allows an intervening SIGCONT to be posted.
2972			 * We need to check for that and bail out if necessary.
2973			 */
2974			if (signr != SIGSTOP) {
2975				spin_unlock_irq(&sighand->siglock);
2976
2977				/* signals can be posted during this window */
2978
2979				if (is_current_pgrp_orphaned())
2980					goto relock;
2981
2982				spin_lock_irq(&sighand->siglock);
2983			}
2984
2985			if (likely(do_signal_stop(signr))) {
2986				/* It released the siglock.  */
2987				goto relock;
2988			}
2989
2990			/*
2991			 * We didn't actually stop, due to a race
2992			 * with SIGCONT or something like that.
2993			 */
2994			continue;
2995		}
2996
2997	fatal:
2998		spin_unlock_irq(&sighand->siglock);
2999		if (unlikely(cgroup_task_frozen(current)))
3000			cgroup_leave_frozen(true);
3001
3002		/*
3003		 * Anything else is fatal, maybe with a core dump.
3004		 */
3005		current->flags |= PF_SIGNALED;
3006
3007		if (sig_kernel_coredump(signr)) {
3008			if (print_fatal_signals)
3009				print_fatal_signal(signr);
3010			proc_coredump_connector(current);
3011			/*
3012			 * If it was able to dump core, this kills all
3013			 * other threads in the group and synchronizes with
3014			 * their demise.  If we lost the race with another
3015			 * thread getting here, it set group_exit_code
3016			 * first and our do_group_exit call below will use
3017			 * that value and ignore the one we pass it.
3018			 */
3019			vfs_coredump(&ksig->info);
3020		}
3021
3022		/*
3023		 * PF_USER_WORKER threads will catch and exit on fatal signals
3024		 * themselves. They have cleanup that must be performed, so we
3025		 * cannot call do_exit() on their behalf. Note that ksig won't
3026		 * be properly initialized, PF_USER_WORKER's shouldn't use it.
3027		 */
3028		if (current->flags & PF_USER_WORKER)
3029			goto out;
3030
3031		/*
3032		 * Death signals, no core dump.
3033		 */
3034		do_group_exit(signr);
3035		/* NOTREACHED */
3036	}
3037	spin_unlock_irq(&sighand->siglock);
3038
3039	ksig->sig = signr;
3040
3041	if (signr && !(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
3042		hide_si_addr_tag_bits(ksig);
3043out:
3044	return signr > 0;
3045}
3046
3047/**
3048 * signal_delivered - called after signal delivery to update blocked signals
3049 * @ksig:		kernel signal struct
3050 * @stepping:		nonzero if debugger single-step or block-step in use
3051 *
3052 * This function should be called when a signal has successfully been
3053 * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
3054 * is always blocked), and the signal itself is blocked unless %SA_NODEFER
3055 * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
3056 */
3057static void signal_delivered(struct ksignal *ksig, int stepping)
3058{
3059	sigset_t blocked;
3060
3061	/* A signal was successfully delivered, and the
3062	   saved sigmask was stored on the signal frame,
3063	   and will be restored by sigreturn.  So we can
3064	   simply clear the restore sigmask flag.  */
3065	clear_restore_sigmask();
3066
3067	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
3068	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
3069		sigaddset(&blocked, ksig->sig);
3070	set_current_blocked(&blocked);
3071	if (current->sas_ss_flags & SS_AUTODISARM)
3072		sas_ss_reset(current);
3073	if (stepping)
3074		ptrace_notify(SIGTRAP, 0);
3075}
3076
3077void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
3078{
3079	if (failed)
3080		force_sigsegv(ksig->sig);
3081	else
3082		signal_delivered(ksig, stepping);
3083}
3084
3085/*
3086 * It could be that complete_signal() picked us to notify about the
3087 * group-wide signal. Other threads should be notified now to take
3088 * the shared signals in @which since we will not.
3089 */
3090static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
3091{
3092	sigset_t retarget;
3093	struct task_struct *t;
3094
3095	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
3096	if (sigisemptyset(&retarget))
3097		return;
3098
3099	for_other_threads(tsk, t) {
3100		if (t->flags & PF_EXITING)
3101			continue;
3102
3103		if (!has_pending_signals(&retarget, &t->blocked))
3104			continue;
3105		/* Remove the signals this thread can handle. */
3106		sigandsets(&retarget, &retarget, &t->blocked);
3107
3108		if (!task_sigpending(t))
3109			signal_wake_up(t, 0);
3110
3111		if (sigisemptyset(&retarget))
3112			break;
3113	}
3114}
3115
3116void exit_signals(struct task_struct *tsk)
3117{
3118	int group_stop = 0;
3119	sigset_t unblocked;
3120
3121	/*
3122	 * @tsk is about to have PF_EXITING set - lock out users which
3123	 * expect stable threadgroup.
3124	 */
3125	cgroup_threadgroup_change_begin(tsk);
3126
3127	if (thread_group_empty(tsk) || (tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
3128		tsk->flags |= PF_EXITING;
3129		cgroup_threadgroup_change_end(tsk);
3130		return;
3131	}
3132
3133	spin_lock_irq(&tsk->sighand->siglock);
3134	/*
3135	 * From now this task is not visible for group-wide signals,
3136	 * see wants_signal(), do_signal_stop().
3137	 */
3138	tsk->flags |= PF_EXITING;
3139
3140	cgroup_threadgroup_change_end(tsk);
3141
3142	if (!task_sigpending(tsk))
3143		goto out;
3144
3145	unblocked = tsk->blocked;
3146	signotset(&unblocked);
3147	retarget_shared_pending(tsk, &unblocked);
3148
3149	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
3150	    task_participate_group_stop(tsk))
3151		group_stop = CLD_STOPPED;
3152out:
3153	spin_unlock_irq(&tsk->sighand->siglock);
3154
3155	/*
3156	 * If group stop has completed, deliver the notification.  This
3157	 * should always go to the real parent of the group leader.
3158	 */
3159	if (unlikely(group_stop)) {
3160		read_lock(&tasklist_lock);
3161		do_notify_parent_cldstop(tsk, false, group_stop);
3162		read_unlock(&tasklist_lock);
3163	}
3164}
3165
3166/*
3167 * System call entry points.
3168 */
3169
3170/**
3171 *  sys_restart_syscall - restart a system call
3172 */
3173SYSCALL_DEFINE0(restart_syscall)
3174{
3175	struct restart_block *restart = &current->restart_block;
3176	return restart->fn(restart);
3177}
3178
3179long do_no_restart_syscall(struct restart_block *param)
3180{
3181	return -EINTR;
3182}
3183
3184static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
3185{
3186	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3187		sigset_t newblocked;
3188		/* A set of now blocked but previously unblocked signals. */
3189		sigandnsets(&newblocked, newset, &current->blocked);
3190		retarget_shared_pending(tsk, &newblocked);
3191	}
3192	tsk->blocked = *newset;
3193	recalc_sigpending();
3194}
3195
3196/**
3197 * set_current_blocked - change current->blocked mask
3198 * @newset: new mask
3199 *
3200 * It is wrong to change ->blocked directly, this helper should be used
3201 * to ensure the process can't miss a shared signal we are going to block.
3202 */
3203void set_current_blocked(sigset_t *newset)
3204{
3205	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3206	__set_current_blocked(newset);
3207}
3208
3209void __set_current_blocked(const sigset_t *newset)
3210{
3211	struct task_struct *tsk = current;
3212
3213	/*
3214	 * In case the signal mask hasn't changed, there is nothing we need
3215	 * to do. The current->blocked shouldn't be modified by other task.
3216	 */
3217	if (sigequalsets(&tsk->blocked, newset))
3218		return;
3219
3220	spin_lock_irq(&tsk->sighand->siglock);
3221	__set_task_blocked(tsk, newset);
3222	spin_unlock_irq(&tsk->sighand->siglock);
3223}
3224
3225/*
3226 * This is also useful for kernel threads that want to temporarily
3227 * (or permanently) block certain signals.
3228 *
3229 * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3230 * interface happily blocks "unblockable" signals like SIGKILL
3231 * and friends.
3232 */
3233int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3234{
3235	struct task_struct *tsk = current;
3236	sigset_t newset;
3237
3238	/* Lockless, only current can change ->blocked, never from irq */
3239	if (oldset)
3240		*oldset = tsk->blocked;
3241
3242	switch (how) {
3243	case SIG_BLOCK:
3244		sigorsets(&newset, &tsk->blocked, set);
3245		break;
3246	case SIG_UNBLOCK:
3247		sigandnsets(&newset, &tsk->blocked, set);
3248		break;
3249	case SIG_SETMASK:
3250		newset = *set;
3251		break;
3252	default:
3253		return -EINVAL;
3254	}
3255
3256	__set_current_blocked(&newset);
3257	return 0;
3258}
3259EXPORT_SYMBOL(sigprocmask);
3260
3261/*
3262 * The api helps set app-provided sigmasks.
3263 *
3264 * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3265 * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3266 *
3267 * Note that it does set_restore_sigmask() in advance, so it must be always
3268 * paired with restore_saved_sigmask_unless() before return from syscall.
3269 */
3270int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3271{
3272	sigset_t kmask;
3273
3274	if (!umask)
3275		return 0;
3276	if (sigsetsize != sizeof(sigset_t))
3277		return -EINVAL;
3278	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3279		return -EFAULT;
3280
3281	set_restore_sigmask();
3282	current->saved_sigmask = current->blocked;
3283	set_current_blocked(&kmask);
3284
3285	return 0;
3286}
3287
3288#ifdef CONFIG_COMPAT
3289int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3290			    size_t sigsetsize)
3291{
3292	sigset_t kmask;
3293
3294	if (!umask)
3295		return 0;
3296	if (sigsetsize != sizeof(compat_sigset_t))
3297		return -EINVAL;
3298	if (get_compat_sigset(&kmask, umask))
3299		return -EFAULT;
3300
3301	set_restore_sigmask();
3302	current->saved_sigmask = current->blocked;
3303	set_current_blocked(&kmask);
3304
3305	return 0;
3306}
3307#endif
3308
3309/**
3310 *  sys_rt_sigprocmask - change the list of currently blocked signals
3311 *  @how: whether to add, remove, or set signals
3312 *  @nset: stores pending signals
3313 *  @oset: previous value of signal mask if non-null
3314 *  @sigsetsize: size of sigset_t type
3315 */
3316SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3317		sigset_t __user *, oset, size_t, sigsetsize)
3318{
3319	sigset_t old_set, new_set;
3320	int error;
3321
3322	/* XXX: Don't preclude handling different sized sigset_t's.  */
3323	if (sigsetsize != sizeof(sigset_t))
3324		return -EINVAL;
3325
3326	old_set = current->blocked;
3327
3328	if (nset) {
3329		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3330			return -EFAULT;
3331		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3332
3333		error = sigprocmask(how, &new_set, NULL);
3334		if (error)
3335			return error;
3336	}
3337
3338	if (oset) {
3339		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3340			return -EFAULT;
3341	}
3342
3343	return 0;
3344}
3345
3346#ifdef CONFIG_COMPAT
3347COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3348		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3349{
3350	sigset_t old_set = current->blocked;
3351
3352	/* XXX: Don't preclude handling different sized sigset_t's.  */
3353	if (sigsetsize != sizeof(sigset_t))
3354		return -EINVAL;
3355
3356	if (nset) {
3357		sigset_t new_set;
3358		int error;
3359		if (get_compat_sigset(&new_set, nset))
3360			return -EFAULT;
3361		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3362
3363		error = sigprocmask(how, &new_set, NULL);
3364		if (error)
3365			return error;
3366	}
3367	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3368}
3369#endif
3370
3371static void do_sigpending(sigset_t *set)
3372{
3373	spin_lock_irq(&current->sighand->siglock);
3374	sigorsets(set, &current->pending.signal,
3375		  &current->signal->shared_pending.signal);
3376	spin_unlock_irq(&current->sighand->siglock);
3377
3378	/* Outside the lock because only this thread touches it.  */
3379	sigandsets(set, &current->blocked, set);
3380}
3381
3382/**
3383 *  sys_rt_sigpending - examine a pending signal that has been raised
3384 *			while blocked
3385 *  @uset: stores pending signals
3386 *  @sigsetsize: size of sigset_t type or larger
3387 */
3388SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3389{
3390	sigset_t set;
3391
3392	if (sigsetsize > sizeof(*uset))
3393		return -EINVAL;
3394
3395	do_sigpending(&set);
3396
3397	if (copy_to_user(uset, &set, sigsetsize))
3398		return -EFAULT;
3399
3400	return 0;
3401}
3402
3403#ifdef CONFIG_COMPAT
3404COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3405		compat_size_t, sigsetsize)
3406{
3407	sigset_t set;
3408
3409	if (sigsetsize > sizeof(*uset))
3410		return -EINVAL;
3411
3412	do_sigpending(&set);
3413
3414	return put_compat_sigset(uset, &set, sigsetsize);
3415}
3416#endif
3417
3418static const struct {
3419	unsigned char limit, layout;
3420} sig_sicodes[] = {
3421	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3422	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3423	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3424	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3425	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3426#if defined(SIGEMT)
3427	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3428#endif
3429	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3430	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3431	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3432};
3433
3434static bool known_siginfo_layout(unsigned sig, int si_code)
3435{
3436	if (si_code == SI_KERNEL)
3437		return true;
3438	else if ((si_code > SI_USER)) {
3439		if (sig_specific_sicodes(sig)) {
3440			if (si_code <= sig_sicodes[sig].limit)
3441				return true;
3442		}
3443		else if (si_code <= NSIGPOLL)
3444			return true;
3445	}
3446	else if (si_code >= SI_DETHREAD)
3447		return true;
3448	else if (si_code == SI_ASYNCNL)
3449		return true;
3450	return false;
3451}
3452
3453enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3454{
3455	enum siginfo_layout layout = SIL_KILL;
3456	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3457		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3458		    (si_code <= sig_sicodes[sig].limit)) {
3459			layout = sig_sicodes[sig].layout;
3460			/* Handle the exceptions */
3461			if ((sig == SIGBUS) &&
3462			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3463				layout = SIL_FAULT_MCEERR;
3464			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3465				layout = SIL_FAULT_BNDERR;
3466#ifdef SEGV_PKUERR
3467			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3468				layout = SIL_FAULT_PKUERR;
3469#endif
3470			else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3471				layout = SIL_FAULT_PERF_EVENT;
3472			else if (IS_ENABLED(CONFIG_SPARC) &&
3473				 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3474				layout = SIL_FAULT_TRAPNO;
3475			else if (IS_ENABLED(CONFIG_ALPHA) &&
3476				 ((sig == SIGFPE) ||
3477				  ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3478				layout = SIL_FAULT_TRAPNO;
3479		}
3480		else if (si_code <= NSIGPOLL)
3481			layout = SIL_POLL;
3482	} else {
3483		if (si_code == SI_TIMER)
3484			layout = SIL_TIMER;
3485		else if (si_code == SI_SIGIO)
3486			layout = SIL_POLL;
3487		else if (si_code < 0)
3488			layout = SIL_RT;
3489	}
3490	return layout;
3491}
3492
3493static inline char __user *si_expansion(const siginfo_t __user *info)
3494{
3495	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3496}
3497
3498int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3499{
3500	char __user *expansion = si_expansion(to);
3501	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3502		return -EFAULT;
3503	if (clear_user(expansion, SI_EXPANSION_SIZE))
3504		return -EFAULT;
3505	return 0;
3506}
3507
3508static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3509				       const siginfo_t __user *from)
3510{
3511	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3512		char __user *expansion = si_expansion(from);
3513		char buf[SI_EXPANSION_SIZE];
3514		int i;
3515		/*
3516		 * An unknown si_code might need more than
3517		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3518		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3519		 * will return this data to userspace exactly.
3520		 */
3521		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3522			return -EFAULT;
3523		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3524			if (buf[i] != 0)
3525				return -E2BIG;
3526		}
3527	}
3528	return 0;
3529}
3530
3531static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3532				    const siginfo_t __user *from)
3533{
3534	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3535		return -EFAULT;
3536	to->si_signo = signo;
3537	return post_copy_siginfo_from_user(to, from);
3538}
3539
3540int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3541{
3542	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3543		return -EFAULT;
3544	return post_copy_siginfo_from_user(to, from);
3545}
3546
3547#ifdef CONFIG_COMPAT
3548/**
3549 * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3550 * @to: compat siginfo destination
3551 * @from: kernel siginfo source
3552 *
3553 * Note: This function does not work properly for the SIGCHLD on x32, but
3554 * fortunately it doesn't have to.  The only valid callers for this function are
3555 * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3556 * The latter does not care because SIGCHLD will never cause a coredump.
3557 */
3558void copy_siginfo_to_external32(struct compat_siginfo *to,
3559		const struct kernel_siginfo *from)
3560{
3561	memset(to, 0, sizeof(*to));
3562
3563	to->si_signo = from->si_signo;
3564	to->si_errno = from->si_errno;
3565	to->si_code  = from->si_code;
3566	switch(siginfo_layout(from->si_signo, from->si_code)) {
3567	case SIL_KILL:
3568		to->si_pid = from->si_pid;
3569		to->si_uid = from->si_uid;
3570		break;
3571	case SIL_TIMER:
3572		to->si_tid     = from->si_tid;
3573		to->si_overrun = from->si_overrun;
3574		to->si_int     = from->si_int;
3575		break;
3576	case SIL_POLL:
3577		to->si_band = from->si_band;
3578		to->si_fd   = from->si_fd;
3579		break;
3580	case SIL_FAULT:
3581		to->si_addr = ptr_to_compat(from->si_addr);
3582		break;
3583	case SIL_FAULT_TRAPNO:
3584		to->si_addr = ptr_to_compat(from->si_addr);
3585		to->si_trapno = from->si_trapno;
3586		break;
3587	case SIL_FAULT_MCEERR:
3588		to->si_addr = ptr_to_compat(from->si_addr);
3589		to->si_addr_lsb = from->si_addr_lsb;
3590		break;
3591	case SIL_FAULT_BNDERR:
3592		to->si_addr = ptr_to_compat(from->si_addr);
3593		to->si_lower = ptr_to_compat(from->si_lower);
3594		to->si_upper = ptr_to_compat(from->si_upper);
3595		break;
3596	case SIL_FAULT_PKUERR:
3597		to->si_addr = ptr_to_compat(from->si_addr);
3598		to->si_pkey = from->si_pkey;
3599		break;
3600	case SIL_FAULT_PERF_EVENT:
3601		to->si_addr = ptr_to_compat(from->si_addr);
3602		to->si_perf_data = from->si_perf_data;
3603		to->si_perf_type = from->si_perf_type;
3604		to->si_perf_flags = from->si_perf_flags;
3605		break;
3606	case SIL_CHLD:
3607		to->si_pid = from->si_pid;
3608		to->si_uid = from->si_uid;
3609		to->si_status = from->si_status;
3610		to->si_utime = from->si_utime;
3611		to->si_stime = from->si_stime;
3612		break;
3613	case SIL_RT:
3614		to->si_pid = from->si_pid;
3615		to->si_uid = from->si_uid;
3616		to->si_int = from->si_int;
3617		break;
3618	case SIL_SYS:
3619		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3620		to->si_syscall   = from->si_syscall;
3621		to->si_arch      = from->si_arch;
3622		break;
3623	}
3624}
3625
3626int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3627			   const struct kernel_siginfo *from)
3628{
3629	struct compat_siginfo new;
3630
3631	copy_siginfo_to_external32(&new, from);
3632	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3633		return -EFAULT;
3634	return 0;
3635}
3636
3637static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3638					 const struct compat_siginfo *from)
3639{
3640	clear_siginfo(to);
3641	to->si_signo = from->si_signo;
3642	to->si_errno = from->si_errno;
3643	to->si_code  = from->si_code;
3644	switch(siginfo_layout(from->si_signo, from->si_code)) {
3645	case SIL_KILL:
3646		to->si_pid = from->si_pid;
3647		to->si_uid = from->si_uid;
3648		break;
3649	case SIL_TIMER:
3650		to->si_tid     = from->si_tid;
3651		to->si_overrun = from->si_overrun;
3652		to->si_int     = from->si_int;
3653		break;
3654	case SIL_POLL:
3655		to->si_band = from->si_band;
3656		to->si_fd   = from->si_fd;
3657		break;
3658	case SIL_FAULT:
3659		to->si_addr = compat_ptr(from->si_addr);
3660		break;
3661	case SIL_FAULT_TRAPNO:
3662		to->si_addr = compat_ptr(from->si_addr);
3663		to->si_trapno = from->si_trapno;
3664		break;
3665	case SIL_FAULT_MCEERR:
3666		to->si_addr = compat_ptr(from->si_addr);
3667		to->si_addr_lsb = from->si_addr_lsb;
3668		break;
3669	case SIL_FAULT_BNDERR:
3670		to->si_addr = compat_ptr(from->si_addr);
3671		to->si_lower = compat_ptr(from->si_lower);
3672		to->si_upper = compat_ptr(from->si_upper);
3673		break;
3674	case SIL_FAULT_PKUERR:
3675		to->si_addr = compat_ptr(from->si_addr);
3676		to->si_pkey = from->si_pkey;
3677		break;
3678	case SIL_FAULT_PERF_EVENT:
3679		to->si_addr = compat_ptr(from->si_addr);
3680		to->si_perf_data = from->si_perf_data;
3681		to->si_perf_type = from->si_perf_type;
3682		to->si_perf_flags = from->si_perf_flags;
3683		break;
3684	case SIL_CHLD:
3685		to->si_pid    = from->si_pid;
3686		to->si_uid    = from->si_uid;
3687		to->si_status = from->si_status;
3688#ifdef CONFIG_X86_X32_ABI
3689		if (in_x32_syscall()) {
3690			to->si_utime = from->_sifields._sigchld_x32._utime;
3691			to->si_stime = from->_sifields._sigchld_x32._stime;
3692		} else
3693#endif
3694		{
3695			to->si_utime = from->si_utime;
3696			to->si_stime = from->si_stime;
3697		}
3698		break;
3699	case SIL_RT:
3700		to->si_pid = from->si_pid;
3701		to->si_uid = from->si_uid;
3702		to->si_int = from->si_int;
3703		break;
3704	case SIL_SYS:
3705		to->si_call_addr = compat_ptr(from->si_call_addr);
3706		to->si_syscall   = from->si_syscall;
3707		to->si_arch      = from->si_arch;
3708		break;
3709	}
3710	return 0;
3711}
3712
3713static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3714				      const struct compat_siginfo __user *ufrom)
3715{
3716	struct compat_siginfo from;
3717
3718	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3719		return -EFAULT;
3720
3721	from.si_signo = signo;
3722	return post_copy_siginfo_from_user32(to, &from);
3723}
3724
3725int copy_siginfo_from_user32(struct kernel_siginfo *to,
3726			     const struct compat_siginfo __user *ufrom)
3727{
3728	struct compat_siginfo from;
3729
3730	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3731		return -EFAULT;
3732
3733	return post_copy_siginfo_from_user32(to, &from);
3734}
3735#endif /* CONFIG_COMPAT */
3736
3737/**
3738 *  do_sigtimedwait - wait for queued signals specified in @which
3739 *  @which: queued signals to wait for
3740 *  @info: if non-null, the signal's siginfo is returned here
3741 *  @ts: upper bound on process time suspension
3742 */
3743static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3744		    const struct timespec64 *ts)
3745{
3746	ktime_t *to = NULL, timeout = KTIME_MAX;
3747	struct task_struct *tsk = current;
3748	sigset_t mask = *which;
3749	enum pid_type type;
3750	int sig, ret = 0;
3751
3752	if (ts) {
3753		if (!timespec64_valid(ts))
3754			return -EINVAL;
3755		timeout = timespec64_to_ktime(*ts);
3756		to = &timeout;
3757	}
3758
3759	/*
3760	 * Invert the set of allowed signals to get those we want to block.
3761	 */
3762	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3763	signotset(&mask);
3764
3765	spin_lock_irq(&tsk->sighand->siglock);
3766	sig = dequeue_signal(&mask, info, &type);
3767	if (!sig && timeout) {
3768		/*
3769		 * None ready, temporarily unblock those we're interested
3770		 * while we are sleeping in so that we'll be awakened when
3771		 * they arrive. Unblocking is always fine, we can avoid
3772		 * set_current_blocked().
3773		 */
3774		tsk->real_blocked = tsk->blocked;
3775		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3776		recalc_sigpending();
3777		spin_unlock_irq(&tsk->sighand->siglock);
3778
3779		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
3780		ret = schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3781					       HRTIMER_MODE_REL);
3782		spin_lock_irq(&tsk->sighand->siglock);
3783		__set_task_blocked(tsk, &tsk->real_blocked);
3784		sigemptyset(&tsk->real_blocked);
3785		sig = dequeue_signal(&mask, info, &type);
3786	}
3787	spin_unlock_irq(&tsk->sighand->siglock);
3788
3789	if (sig)
3790		return sig;
3791	return ret ? -EINTR : -EAGAIN;
3792}
3793
3794/**
3795 *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3796 *			in @uthese
3797 *  @uthese: queued signals to wait for
3798 *  @uinfo: if non-null, the signal's siginfo is returned here
3799 *  @uts: upper bound on process time suspension
3800 *  @sigsetsize: size of sigset_t type
3801 */
3802SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3803		siginfo_t __user *, uinfo,
3804		const struct __kernel_timespec __user *, uts,
3805		size_t, sigsetsize)
3806{
3807	sigset_t these;
3808	struct timespec64 ts;
3809	kernel_siginfo_t info;
3810	int ret;
3811
3812	/* XXX: Don't preclude handling different sized sigset_t's.  */
3813	if (sigsetsize != sizeof(sigset_t))
3814		return -EINVAL;
3815
3816	if (copy_from_user(&these, uthese, sizeof(these)))
3817		return -EFAULT;
3818
3819	if (uts) {
3820		if (get_timespec64(&ts, uts))
3821			return -EFAULT;
3822	}
3823
3824	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3825
3826	if (ret > 0 && uinfo) {
3827		if (copy_siginfo_to_user(uinfo, &info))
3828			ret = -EFAULT;
3829	}
3830
3831	return ret;
3832}
3833
3834#ifdef CONFIG_COMPAT_32BIT_TIME
3835SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3836		siginfo_t __user *, uinfo,
3837		const struct old_timespec32 __user *, uts,
3838		size_t, sigsetsize)
3839{
3840	sigset_t these;
3841	struct timespec64 ts;
3842	kernel_siginfo_t info;
3843	int ret;
3844
3845	if (sigsetsize != sizeof(sigset_t))
3846		return -EINVAL;
3847
3848	if (copy_from_user(&these, uthese, sizeof(these)))
3849		return -EFAULT;
3850
3851	if (uts) {
3852		if (get_old_timespec32(&ts, uts))
3853			return -EFAULT;
3854	}
3855
3856	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3857
3858	if (ret > 0 && uinfo) {
3859		if (copy_siginfo_to_user(uinfo, &info))
3860			ret = -EFAULT;
3861	}
3862
3863	return ret;
3864}
3865#endif
3866
3867#ifdef CONFIG_COMPAT
3868COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3869		struct compat_siginfo __user *, uinfo,
3870		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3871{
3872	sigset_t s;
3873	struct timespec64 t;
3874	kernel_siginfo_t info;
3875	long ret;
3876
3877	if (sigsetsize != sizeof(sigset_t))
3878		return -EINVAL;
3879
3880	if (get_compat_sigset(&s, uthese))
3881		return -EFAULT;
3882
3883	if (uts) {
3884		if (get_timespec64(&t, uts))
3885			return -EFAULT;
3886	}
3887
3888	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3889
3890	if (ret > 0 && uinfo) {
3891		if (copy_siginfo_to_user32(uinfo, &info))
3892			ret = -EFAULT;
3893	}
3894
3895	return ret;
3896}
3897
3898#ifdef CONFIG_COMPAT_32BIT_TIME
3899COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3900		struct compat_siginfo __user *, uinfo,
3901		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3902{
3903	sigset_t s;
3904	struct timespec64 t;
3905	kernel_siginfo_t info;
3906	long ret;
3907
3908	if (sigsetsize != sizeof(sigset_t))
3909		return -EINVAL;
3910
3911	if (get_compat_sigset(&s, uthese))
3912		return -EFAULT;
3913
3914	if (uts) {
3915		if (get_old_timespec32(&t, uts))
3916			return -EFAULT;
3917	}
3918
3919	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3920
3921	if (ret > 0 && uinfo) {
3922		if (copy_siginfo_to_user32(uinfo, &info))
3923			ret = -EFAULT;
3924	}
3925
3926	return ret;
3927}
3928#endif
3929#endif
3930
3931static void prepare_kill_siginfo(int sig, struct kernel_siginfo *info,
3932				 enum pid_type type)
3933{
3934	clear_siginfo(info);
3935	info->si_signo = sig;
3936	info->si_errno = 0;
3937	info->si_code = (type == PIDTYPE_PID) ? SI_TKILL : SI_USER;
3938	info->si_pid = task_tgid_vnr(current);
3939	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3940}
3941
3942/**
3943 *  sys_kill - send a signal to a process
3944 *  @pid: the PID of the process
3945 *  @sig: signal to be sent
3946 */
3947SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3948{
3949	struct kernel_siginfo info;
3950
3951	prepare_kill_siginfo(sig, &info, PIDTYPE_TGID);
3952
3953	return kill_something_info(sig, &info, pid);
3954}
3955
3956/*
3957 * Verify that the signaler and signalee either are in the same pid namespace
3958 * or that the signaler's pid namespace is an ancestor of the signalee's pid
3959 * namespace.
3960 */
3961static bool access_pidfd_pidns(struct pid *pid)
3962{
3963	struct pid_namespace *active = task_active_pid_ns(current);
3964	struct pid_namespace *p = ns_of_pid(pid);
3965
3966	for (;;) {
3967		if (!p)
3968			return false;
3969		if (p == active)
3970			break;
3971		p = p->parent;
3972	}
3973
3974	return true;
3975}
3976
3977static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3978		siginfo_t __user *info)
3979{
3980#ifdef CONFIG_COMPAT
3981	/*
3982	 * Avoid hooking up compat syscalls and instead handle necessary
3983	 * conversions here. Note, this is a stop-gap measure and should not be
3984	 * considered a generic solution.
3985	 */
3986	if (in_compat_syscall())
3987		return copy_siginfo_from_user32(
3988			kinfo, (struct compat_siginfo __user *)info);
3989#endif
3990	return copy_siginfo_from_user(kinfo, info);
3991}
3992
3993static struct pid *pidfd_to_pid(const struct file *file)
3994{
3995	struct pid *pid;
3996
3997	pid = pidfd_pid(file);
3998	if (!IS_ERR(pid))
3999		return pid;
4000
4001	return tgid_pidfd_to_pid(file);
4002}
4003
4004#define PIDFD_SEND_SIGNAL_FLAGS                            \
4005	(PIDFD_SIGNAL_THREAD | PIDFD_SIGNAL_THREAD_GROUP | \
4006	 PIDFD_SIGNAL_PROCESS_GROUP)
4007
4008static int do_pidfd_send_signal(struct pid *pid, int sig, enum pid_type type,
4009				siginfo_t __user *info, unsigned int flags)
4010{
4011	kernel_siginfo_t kinfo;
4012
4013	switch (flags) {
4014	case PIDFD_SIGNAL_THREAD:
4015		type = PIDTYPE_PID;
4016		break;
4017	case PIDFD_SIGNAL_THREAD_GROUP:
4018		type = PIDTYPE_TGID;
4019		break;
4020	case PIDFD_SIGNAL_PROCESS_GROUP:
4021		type = PIDTYPE_PGID;
4022		break;
4023	}
4024
4025	if (info) {
4026		int ret;
4027
4028		ret = copy_siginfo_from_user_any(&kinfo, info);
4029		if (unlikely(ret))
4030			return ret;
4031
4032		if (unlikely(sig != kinfo.si_signo))
4033			return -EINVAL;
4034
4035		/* Only allow sending arbitrary signals to yourself. */
4036		if ((task_pid(current) != pid || type > PIDTYPE_TGID) &&
4037		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
4038			return -EPERM;
4039	} else {
4040		prepare_kill_siginfo(sig, &kinfo, type);
4041	}
4042
4043	if (type == PIDTYPE_PGID)
4044		return kill_pgrp_info(sig, &kinfo, pid);
4045
4046	return kill_pid_info_type(sig, &kinfo, pid, type);
4047}
4048
4049/**
4050 * sys_pidfd_send_signal - Signal a process through a pidfd
4051 * @pidfd:  file descriptor of the process
4052 * @sig:    signal to send
4053 * @info:   signal info
4054 * @flags:  future flags
4055 *
4056 * Send the signal to the thread group or to the individual thread depending
4057 * on PIDFD_THREAD.
4058 * In the future extension to @flags may be used to override the default scope
4059 * of @pidfd.
4060 *
4061 * Return: 0 on success, negative errno on failure
4062 */
4063SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
4064		siginfo_t __user *, info, unsigned int, flags)
4065{
4066	struct pid *pid;
4067	enum pid_type type;
4068	int ret;
4069
4070	/* Enforce flags be set to 0 until we add an extension. */
4071	if (flags & ~PIDFD_SEND_SIGNAL_FLAGS)
4072		return -EINVAL;
4073
4074	/* Ensure that only a single signal scope determining flag is set. */
4075	if (hweight32(flags & PIDFD_SEND_SIGNAL_FLAGS) > 1)
4076		return -EINVAL;
4077
4078	switch (pidfd) {
4079	case PIDFD_SELF_THREAD:
4080		pid = get_task_pid(current, PIDTYPE_PID);
4081		type = PIDTYPE_PID;
4082		break;
4083	case PIDFD_SELF_THREAD_GROUP:
4084		pid = get_task_pid(current, PIDTYPE_TGID);
4085		type = PIDTYPE_TGID;
4086		break;
4087	default: {
4088		CLASS(fd, f)(pidfd);
4089		if (fd_empty(f))
4090			return -EBADF;
4091
4092		/* Is this a pidfd? */
4093		pid = pidfd_to_pid(fd_file(f));
4094		if (IS_ERR(pid))
4095			return PTR_ERR(pid);
4096
4097		if (!access_pidfd_pidns(pid))
4098			return -EINVAL;
4099
4100		/* Infer scope from the type of pidfd. */
4101		if (fd_file(f)->f_flags & PIDFD_THREAD)
4102			type = PIDTYPE_PID;
4103		else
4104			type = PIDTYPE_TGID;
4105
4106		return do_pidfd_send_signal(pid, sig, type, info, flags);
4107	}
4108	}
4109
4110	ret = do_pidfd_send_signal(pid, sig, type, info, flags);
4111	put_pid(pid);
4112
4113	return ret;
4114}
4115
4116static int
4117do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
4118{
4119	struct task_struct *p;
4120	int error = -ESRCH;
4121
4122	rcu_read_lock();
4123	p = find_task_by_vpid(pid);
4124	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
4125		error = check_kill_permission(sig, info, p);
4126		/*
4127		 * The null signal is a permissions and process existence
4128		 * probe.  No signal is actually delivered.
4129		 */
4130		if (!error && sig) {
4131			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
4132			/*
4133			 * If lock_task_sighand() failed we pretend the task
4134			 * dies after receiving the signal. The window is tiny,
4135			 * and the signal is private anyway.
4136			 */
4137			if (unlikely(error == -ESRCH))
4138				error = 0;
4139		}
4140	}
4141	rcu_read_unlock();
4142
4143	return error;
4144}
4145
4146static int do_tkill(pid_t tgid, pid_t pid, int sig)
4147{
4148	struct kernel_siginfo info;
4149
4150	prepare_kill_siginfo(sig, &info, PIDTYPE_PID);
4151
4152	return do_send_specific(tgid, pid, sig, &info);
4153}
4154
4155/**
4156 *  sys_tgkill - send signal to one specific thread
4157 *  @tgid: the thread group ID of the thread
4158 *  @pid: the PID of the thread
4159 *  @sig: signal to be sent
4160 *
4161 *  This syscall also checks the @tgid and returns -ESRCH even if the PID
4162 *  exists but it's not belonging to the target process anymore. This
4163 *  method solves the problem of threads exiting and PIDs getting reused.
4164 */
4165SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
4166{
4167	/* This is only valid for single tasks */
4168	if (pid <= 0 || tgid <= 0)
4169		return -EINVAL;
4170
4171	return do_tkill(tgid, pid, sig);
4172}
4173
4174/**
4175 *  sys_tkill - send signal to one specific task
4176 *  @pid: the PID of the task
4177 *  @sig: signal to be sent
4178 *
4179 *  Send a signal to only one task, even if it's a CLONE_THREAD task.
4180 */
4181SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
4182{
4183	/* This is only valid for single tasks */
4184	if (pid <= 0)
4185		return -EINVAL;
4186
4187	return do_tkill(0, pid, sig);
4188}
4189
4190static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
4191{
4192	/* Not even root can pretend to send signals from the kernel.
4193	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4194	 */
4195	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4196	    (task_pid_vnr(current) != pid))
4197		return -EPERM;
4198
4199	/* POSIX.1b doesn't mention process groups.  */
4200	return kill_proc_info(sig, info, pid);
4201}
4202
4203/**
4204 *  sys_rt_sigqueueinfo - send signal information to a signal
4205 *  @pid: the PID of the thread
4206 *  @sig: signal to be sent
4207 *  @uinfo: signal info to be sent
4208 */
4209SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
4210		siginfo_t __user *, uinfo)
4211{
4212	kernel_siginfo_t info;
4213	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4214	if (unlikely(ret))
4215		return ret;
4216	return do_rt_sigqueueinfo(pid, sig, &info);
4217}
4218
4219#ifdef CONFIG_COMPAT
4220COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
4221			compat_pid_t, pid,
4222			int, sig,
4223			struct compat_siginfo __user *, uinfo)
4224{
4225	kernel_siginfo_t info;
4226	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4227	if (unlikely(ret))
4228		return ret;
4229	return do_rt_sigqueueinfo(pid, sig, &info);
4230}
4231#endif
4232
4233static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4234{
4235	/* This is only valid for single tasks */
4236	if (pid <= 0 || tgid <= 0)
4237		return -EINVAL;
4238
4239	/* Not even root can pretend to send signals from the kernel.
4240	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4241	 */
4242	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4243	    (task_pid_vnr(current) != pid))
4244		return -EPERM;
4245
4246	return do_send_specific(tgid, pid, sig, info);
4247}
4248
4249SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4250		siginfo_t __user *, uinfo)
4251{
4252	kernel_siginfo_t info;
4253	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4254	if (unlikely(ret))
4255		return ret;
4256	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4257}
4258
4259#ifdef CONFIG_COMPAT
4260COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4261			compat_pid_t, tgid,
4262			compat_pid_t, pid,
4263			int, sig,
4264			struct compat_siginfo __user *, uinfo)
4265{
4266	kernel_siginfo_t info;
4267	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4268	if (unlikely(ret))
4269		return ret;
4270	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4271}
4272#endif
4273
4274/*
4275 * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4276 */
4277void kernel_sigaction(int sig, __sighandler_t action)
4278{
4279	spin_lock_irq(&current->sighand->siglock);
4280	current->sighand->action[sig - 1].sa.sa_handler = action;
4281	if (action == SIG_IGN) {
4282		sigset_t mask;
4283
4284		sigemptyset(&mask);
4285		sigaddset(&mask, sig);
4286
4287		flush_sigqueue_mask(current, &mask, &current->signal->shared_pending);
4288		flush_sigqueue_mask(current, &mask, &current->pending);
4289		recalc_sigpending();
4290	}
4291	spin_unlock_irq(&current->sighand->siglock);
4292}
4293EXPORT_SYMBOL(kernel_sigaction);
4294
4295void __weak sigaction_compat_abi(struct k_sigaction *act,
4296		struct k_sigaction *oact)
4297{
4298}
4299
4300int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4301{
4302	struct task_struct *p = current, *t;
4303	struct k_sigaction *k;
4304	sigset_t mask;
4305
4306	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4307		return -EINVAL;
4308
4309	k = &p->sighand->action[sig-1];
4310
4311	spin_lock_irq(&p->sighand->siglock);
4312	if (k->sa.sa_flags & SA_IMMUTABLE) {
4313		spin_unlock_irq(&p->sighand->siglock);
4314		return -EINVAL;
4315	}
4316	if (oact)
4317		*oact = *k;
4318
4319	/*
4320	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4321	 * e.g. by having an architecture use the bit in their uapi.
4322	 */
4323	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4324
4325	/*
4326	 * Clear unknown flag bits in order to allow userspace to detect missing
4327	 * support for flag bits and to allow the kernel to use non-uapi bits
4328	 * internally.
4329	 */
4330	if (act)
4331		act->sa.sa_flags &= UAPI_SA_FLAGS;
4332	if (oact)
4333		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4334
4335	sigaction_compat_abi(act, oact);
4336
4337	if (act) {
4338		bool was_ignored = k->sa.sa_handler == SIG_IGN;
4339
4340		sigdelsetmask(&act->sa.sa_mask,
4341			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4342		*k = *act;
4343		/*
4344		 * POSIX 3.3.1.3:
4345		 *  "Setting a signal action to SIG_IGN for a signal that is
4346		 *   pending shall cause the pending signal to be discarded,
4347		 *   whether or not it is blocked."
4348		 *
4349		 *  "Setting a signal action to SIG_DFL for a signal that is
4350		 *   pending and whose default action is to ignore the signal
4351		 *   (for example, SIGCHLD), shall cause the pending signal to
4352		 *   be discarded, whether or not it is blocked"
4353		 */
4354		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4355			sigemptyset(&mask);
4356			sigaddset(&mask, sig);
4357			flush_sigqueue_mask(p, &mask, &p->signal->shared_pending);
4358			for_each_thread(p, t)
4359				flush_sigqueue_mask(p, &mask, &t->pending);
4360		} else if (was_ignored) {
4361			posixtimer_sig_unignore(p, sig);
4362		}
4363	}
4364
4365	spin_unlock_irq(&p->sighand->siglock);
4366	return 0;
4367}
4368
4369#ifdef CONFIG_DYNAMIC_SIGFRAME
4370static inline void sigaltstack_lock(void)
4371	__acquires(&current->sighand->siglock)
4372{
4373	spin_lock_irq(&current->sighand->siglock);
4374}
4375
4376static inline void sigaltstack_unlock(void)
4377	__releases(&current->sighand->siglock)
4378{
4379	spin_unlock_irq(&current->sighand->siglock);
4380}
4381#else
4382static inline void sigaltstack_lock(void) { }
4383static inline void sigaltstack_unlock(void) { }
4384#endif
4385
4386static int
4387do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4388		size_t min_ss_size)
4389{
4390	struct task_struct *t = current;
4391	int ret = 0;
4392
4393	if (oss) {
4394		memset(oss, 0, sizeof(stack_t));
4395		oss->ss_sp = (void __user *) t->sas_ss_sp;
4396		oss->ss_size = t->sas_ss_size;
4397		oss->ss_flags = sas_ss_flags(sp) |
4398			(current->sas_ss_flags & SS_FLAG_BITS);
4399	}
4400
4401	if (ss) {
4402		void __user *ss_sp = ss->ss_sp;
4403		size_t ss_size = ss->ss_size;
4404		unsigned ss_flags = ss->ss_flags;
4405		int ss_mode;
4406
4407		if (unlikely(on_sig_stack(sp)))
4408			return -EPERM;
4409
4410		ss_mode = ss_flags & ~SS_FLAG_BITS;
4411		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4412				ss_mode != 0))
4413			return -EINVAL;
4414
4415		/*
4416		 * Return before taking any locks if no actual
4417		 * sigaltstack changes were requested.
4418		 */
4419		if (t->sas_ss_sp == (unsigned long)ss_sp &&
4420		    t->sas_ss_size == ss_size &&
4421		    t->sas_ss_flags == ss_flags)
4422			return 0;
4423
4424		sigaltstack_lock();
4425		if (ss_mode == SS_DISABLE) {
4426			ss_size = 0;
4427			ss_sp = NULL;
4428		} else {
4429			if (unlikely(ss_size < min_ss_size))
4430				ret = -ENOMEM;
4431			if (!sigaltstack_size_valid(ss_size))
4432				ret = -ENOMEM;
4433		}
4434		if (!ret) {
4435			t->sas_ss_sp = (unsigned long) ss_sp;
4436			t->sas_ss_size = ss_size;
4437			t->sas_ss_flags = ss_flags;
4438		}
4439		sigaltstack_unlock();
4440	}
4441	return ret;
4442}
4443
4444SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4445{
4446	stack_t new, old;
4447	int err;
4448	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4449		return -EFAULT;
4450	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4451			      current_user_stack_pointer(),
4452			      MINSIGSTKSZ);
4453	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4454		err = -EFAULT;
4455	return err;
4456}
4457
4458int restore_altstack(const stack_t __user *uss)
4459{
4460	stack_t new;
4461	if (copy_from_user(&new, uss, sizeof(stack_t)))
4462		return -EFAULT;
4463	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4464			     MINSIGSTKSZ);
4465	/* squash all but EFAULT for now */
4466	return 0;
4467}
4468
4469int __save_altstack(stack_t __user *uss, unsigned long sp)
4470{
4471	struct task_struct *t = current;
4472	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4473		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4474		__put_user(t->sas_ss_size, &uss->ss_size);
4475	return err;
4476}
4477
4478#ifdef CONFIG_COMPAT
4479static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4480				 compat_stack_t __user *uoss_ptr)
4481{
4482	stack_t uss, uoss;
4483	int ret;
4484
4485	if (uss_ptr) {
4486		compat_stack_t uss32;
4487		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4488			return -EFAULT;
4489		uss.ss_sp = compat_ptr(uss32.ss_sp);
4490		uss.ss_flags = uss32.ss_flags;
4491		uss.ss_size = uss32.ss_size;
4492	}
4493	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4494			     compat_user_stack_pointer(),
4495			     COMPAT_MINSIGSTKSZ);
4496	if (ret >= 0 && uoss_ptr)  {
4497		compat_stack_t old;
4498		memset(&old, 0, sizeof(old));
4499		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4500		old.ss_flags = uoss.ss_flags;
4501		old.ss_size = uoss.ss_size;
4502		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4503			ret = -EFAULT;
4504	}
4505	return ret;
4506}
4507
4508COMPAT_SYSCALL_DEFINE2(sigaltstack,
4509			const compat_stack_t __user *, uss_ptr,
4510			compat_stack_t __user *, uoss_ptr)
4511{
4512	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4513}
4514
4515int compat_restore_altstack(const compat_stack_t __user *uss)
4516{
4517	int err = do_compat_sigaltstack(uss, NULL);
4518	/* squash all but -EFAULT for now */
4519	return err == -EFAULT ? err : 0;
4520}
4521
4522int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4523{
4524	int err;
4525	struct task_struct *t = current;
4526	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4527			 &uss->ss_sp) |
4528		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4529		__put_user(t->sas_ss_size, &uss->ss_size);
4530	return err;
4531}
4532#endif
4533
4534#ifdef __ARCH_WANT_SYS_SIGPENDING
4535
4536/**
4537 *  sys_sigpending - examine pending signals
4538 *  @uset: where mask of pending signal is returned
4539 */
4540SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4541{
4542	sigset_t set;
4543
4544	if (sizeof(old_sigset_t) > sizeof(*uset))
4545		return -EINVAL;
4546
4547	do_sigpending(&set);
4548
4549	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4550		return -EFAULT;
4551
4552	return 0;
4553}
4554
4555#ifdef CONFIG_COMPAT
4556COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4557{
4558	sigset_t set;
4559
4560	do_sigpending(&set);
4561
4562	return put_user(set.sig[0], set32);
4563}
4564#endif
4565
4566#endif
4567
4568#ifdef __ARCH_WANT_SYS_SIGPROCMASK
4569/**
4570 *  sys_sigprocmask - examine and change blocked signals
4571 *  @how: whether to add, remove, or set signals
4572 *  @nset: signals to add or remove (if non-null)
4573 *  @oset: previous value of signal mask if non-null
4574 *
4575 * Some platforms have their own version with special arguments;
4576 * others support only sys_rt_sigprocmask.
4577 */
4578
4579SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4580		old_sigset_t __user *, oset)
4581{
4582	old_sigset_t old_set, new_set;
4583	sigset_t new_blocked;
4584
4585	old_set = current->blocked.sig[0];
4586
4587	if (nset) {
4588		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4589			return -EFAULT;
4590
4591		new_blocked = current->blocked;
4592
4593		switch (how) {
4594		case SIG_BLOCK:
4595			sigaddsetmask(&new_blocked, new_set);
4596			break;
4597		case SIG_UNBLOCK:
4598			sigdelsetmask(&new_blocked, new_set);
4599			break;
4600		case SIG_SETMASK:
4601			new_blocked.sig[0] = new_set;
4602			break;
4603		default:
4604			return -EINVAL;
4605		}
4606
4607		set_current_blocked(&new_blocked);
4608	}
4609
4610	if (oset) {
4611		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4612			return -EFAULT;
4613	}
4614
4615	return 0;
4616}
4617#endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4618
4619#ifndef CONFIG_ODD_RT_SIGACTION
4620/**
4621 *  sys_rt_sigaction - alter an action taken by a process
4622 *  @sig: signal to be sent
4623 *  @act: new sigaction
4624 *  @oact: used to save the previous sigaction
4625 *  @sigsetsize: size of sigset_t type
4626 */
4627SYSCALL_DEFINE4(rt_sigaction, int, sig,
4628		const struct sigaction __user *, act,
4629		struct sigaction __user *, oact,
4630		size_t, sigsetsize)
4631{
4632	struct k_sigaction new_sa, old_sa;
4633	int ret;
4634
4635	/* XXX: Don't preclude handling different sized sigset_t's.  */
4636	if (sigsetsize != sizeof(sigset_t))
4637		return -EINVAL;
4638
4639	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4640		return -EFAULT;
4641
4642	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4643	if (ret)
4644		return ret;
4645
4646	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4647		return -EFAULT;
4648
4649	return 0;
4650}
4651#ifdef CONFIG_COMPAT
4652COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4653		const struct compat_sigaction __user *, act,
4654		struct compat_sigaction __user *, oact,
4655		compat_size_t, sigsetsize)
4656{
4657	struct k_sigaction new_ka, old_ka;
4658#ifdef __ARCH_HAS_SA_RESTORER
4659	compat_uptr_t restorer;
4660#endif
4661	int ret;
4662
4663	/* XXX: Don't preclude handling different sized sigset_t's.  */
4664	if (sigsetsize != sizeof(compat_sigset_t))
4665		return -EINVAL;
4666
4667	if (act) {
4668		compat_uptr_t handler;
4669		ret = get_user(handler, &act->sa_handler);
4670		new_ka.sa.sa_handler = compat_ptr(handler);
4671#ifdef __ARCH_HAS_SA_RESTORER
4672		ret |= get_user(restorer, &act->sa_restorer);
4673		new_ka.sa.sa_restorer = compat_ptr(restorer);
4674#endif
4675		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4676		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4677		if (ret)
4678			return -EFAULT;
4679	}
4680
4681	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4682	if (!ret && oact) {
4683		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler), 
4684			       &oact->sa_handler);
4685		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4686					 sizeof(oact->sa_mask));
4687		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4688#ifdef __ARCH_HAS_SA_RESTORER
4689		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4690				&oact->sa_restorer);
4691#endif
4692	}
4693	return ret;
4694}
4695#endif
4696#endif /* !CONFIG_ODD_RT_SIGACTION */
4697
4698#ifdef CONFIG_OLD_SIGACTION
4699SYSCALL_DEFINE3(sigaction, int, sig,
4700		const struct old_sigaction __user *, act,
4701	        struct old_sigaction __user *, oact)
4702{
4703	struct k_sigaction new_ka, old_ka;
4704	int ret;
4705
4706	if (act) {
4707		old_sigset_t mask;
4708		if (!access_ok(act, sizeof(*act)) ||
4709		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4710		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4711		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4712		    __get_user(mask, &act->sa_mask))
4713			return -EFAULT;
4714#ifdef __ARCH_HAS_KA_RESTORER
4715		new_ka.ka_restorer = NULL;
4716#endif
4717		siginitset(&new_ka.sa.sa_mask, mask);
4718	}
4719
4720	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4721
4722	if (!ret && oact) {
4723		if (!access_ok(oact, sizeof(*oact)) ||
4724		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4725		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4726		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4727		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4728			return -EFAULT;
4729	}
4730
4731	return ret;
4732}
4733#endif
4734#ifdef CONFIG_COMPAT_OLD_SIGACTION
4735COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4736		const struct compat_old_sigaction __user *, act,
4737	        struct compat_old_sigaction __user *, oact)
4738{
4739	struct k_sigaction new_ka, old_ka;
4740	int ret;
4741	compat_old_sigset_t mask;
4742	compat_uptr_t handler, restorer;
4743
4744	if (act) {
4745		if (!access_ok(act, sizeof(*act)) ||
4746		    __get_user(handler, &act->sa_handler) ||
4747		    __get_user(restorer, &act->sa_restorer) ||
4748		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4749		    __get_user(mask, &act->sa_mask))
4750			return -EFAULT;
4751
4752#ifdef __ARCH_HAS_KA_RESTORER
4753		new_ka.ka_restorer = NULL;
4754#endif
4755		new_ka.sa.sa_handler = compat_ptr(handler);
4756		new_ka.sa.sa_restorer = compat_ptr(restorer);
4757		siginitset(&new_ka.sa.sa_mask, mask);
4758	}
4759
4760	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4761
4762	if (!ret && oact) {
4763		if (!access_ok(oact, sizeof(*oact)) ||
4764		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4765			       &oact->sa_handler) ||
4766		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4767			       &oact->sa_restorer) ||
4768		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4769		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4770			return -EFAULT;
4771	}
4772	return ret;
4773}
4774#endif
4775
4776#ifdef CONFIG_SGETMASK_SYSCALL
4777
4778/*
4779 * For backwards compatibility.  Functionality superseded by sigprocmask.
4780 */
4781SYSCALL_DEFINE0(sgetmask)
4782{
4783	/* SMP safe */
4784	return current->blocked.sig[0];
4785}
4786
4787SYSCALL_DEFINE1(ssetmask, int, newmask)
4788{
4789	int old = current->blocked.sig[0];
4790	sigset_t newset;
4791
4792	siginitset(&newset, newmask);
4793	set_current_blocked(&newset);
4794
4795	return old;
4796}
4797#endif /* CONFIG_SGETMASK_SYSCALL */
4798
4799#ifdef __ARCH_WANT_SYS_SIGNAL
4800/*
4801 * For backwards compatibility.  Functionality superseded by sigaction.
4802 */
4803SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4804{
4805	struct k_sigaction new_sa, old_sa;
4806	int ret;
4807
4808	new_sa.sa.sa_handler = handler;
4809	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4810	sigemptyset(&new_sa.sa.sa_mask);
4811
4812	ret = do_sigaction(sig, &new_sa, &old_sa);
4813
4814	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4815}
4816#endif /* __ARCH_WANT_SYS_SIGNAL */
4817
4818#ifdef __ARCH_WANT_SYS_PAUSE
4819
4820SYSCALL_DEFINE0(pause)
4821{
4822	while (!signal_pending(current)) {
4823		__set_current_state(TASK_INTERRUPTIBLE);
4824		schedule();
4825	}
4826	return -ERESTARTNOHAND;
4827}
4828
4829#endif
4830
4831static int sigsuspend(sigset_t *set)
4832{
4833	current->saved_sigmask = current->blocked;
4834	set_current_blocked(set);
4835
4836	while (!signal_pending(current)) {
4837		__set_current_state(TASK_INTERRUPTIBLE);
4838		schedule();
4839	}
4840	set_restore_sigmask();
4841	return -ERESTARTNOHAND;
4842}
4843
4844/**
4845 *  sys_rt_sigsuspend - replace the signal mask for a value with the
4846 *	@unewset value until a signal is received
4847 *  @unewset: new signal mask value
4848 *  @sigsetsize: size of sigset_t type
4849 */
4850SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4851{
4852	sigset_t newset;
4853
4854	/* XXX: Don't preclude handling different sized sigset_t's.  */
4855	if (sigsetsize != sizeof(sigset_t))
4856		return -EINVAL;
4857
4858	if (copy_from_user(&newset, unewset, sizeof(newset)))
4859		return -EFAULT;
4860	return sigsuspend(&newset);
4861}
4862 
4863#ifdef CONFIG_COMPAT
4864COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4865{
4866	sigset_t newset;
4867
4868	/* XXX: Don't preclude handling different sized sigset_t's.  */
4869	if (sigsetsize != sizeof(sigset_t))
4870		return -EINVAL;
4871
4872	if (get_compat_sigset(&newset, unewset))
4873		return -EFAULT;
4874	return sigsuspend(&newset);
4875}
4876#endif
4877
4878#ifdef CONFIG_OLD_SIGSUSPEND
4879SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4880{
4881	sigset_t blocked;
4882	siginitset(&blocked, mask);
4883	return sigsuspend(&blocked);
4884}
4885#endif
4886#ifdef CONFIG_OLD_SIGSUSPEND3
4887SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4888{
4889	sigset_t blocked;
4890	siginitset(&blocked, mask);
4891	return sigsuspend(&blocked);
4892}
4893#endif
4894
4895__weak const char *arch_vma_name(struct vm_area_struct *vma)
4896{
4897	return NULL;
4898}
4899
4900static inline void siginfo_buildtime_checks(void)
4901{
4902	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4903
4904	/* Verify the offsets in the two siginfos match */
4905#define CHECK_OFFSET(field) \
4906	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4907
4908	/* kill */
4909	CHECK_OFFSET(si_pid);
4910	CHECK_OFFSET(si_uid);
4911
4912	/* timer */
4913	CHECK_OFFSET(si_tid);
4914	CHECK_OFFSET(si_overrun);
4915	CHECK_OFFSET(si_value);
4916
4917	/* rt */
4918	CHECK_OFFSET(si_pid);
4919	CHECK_OFFSET(si_uid);
4920	CHECK_OFFSET(si_value);
4921
4922	/* sigchld */
4923	CHECK_OFFSET(si_pid);
4924	CHECK_OFFSET(si_uid);
4925	CHECK_OFFSET(si_status);
4926	CHECK_OFFSET(si_utime);
4927	CHECK_OFFSET(si_stime);
4928
4929	/* sigfault */
4930	CHECK_OFFSET(si_addr);
4931	CHECK_OFFSET(si_trapno);
4932	CHECK_OFFSET(si_addr_lsb);
4933	CHECK_OFFSET(si_lower);
4934	CHECK_OFFSET(si_upper);
4935	CHECK_OFFSET(si_pkey);
4936	CHECK_OFFSET(si_perf_data);
4937	CHECK_OFFSET(si_perf_type);
4938	CHECK_OFFSET(si_perf_flags);
4939
4940	/* sigpoll */
4941	CHECK_OFFSET(si_band);
4942	CHECK_OFFSET(si_fd);
4943
4944	/* sigsys */
4945	CHECK_OFFSET(si_call_addr);
4946	CHECK_OFFSET(si_syscall);
4947	CHECK_OFFSET(si_arch);
4948#undef CHECK_OFFSET
4949
4950	/* usb asyncio */
4951	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4952		     offsetof(struct siginfo, si_addr));
4953	if (sizeof(int) == sizeof(void __user *)) {
4954		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4955			     sizeof(void __user *));
4956	} else {
4957		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4958			      sizeof_field(struct siginfo, si_uid)) !=
4959			     sizeof(void __user *));
4960		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4961			     offsetof(struct siginfo, si_uid));
4962	}
4963#ifdef CONFIG_COMPAT
4964	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4965		     offsetof(struct compat_siginfo, si_addr));
4966	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4967		     sizeof(compat_uptr_t));
4968	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4969		     sizeof_field(struct siginfo, si_pid));
4970#endif
4971}
4972
4973#if defined(CONFIG_SYSCTL)
4974static const struct ctl_table signal_debug_table[] = {
4975#ifdef CONFIG_SYSCTL_EXCEPTION_TRACE
4976	{
4977		.procname	= "exception-trace",
4978		.data		= &show_unhandled_signals,
4979		.maxlen		= sizeof(int),
4980		.mode		= 0644,
4981		.proc_handler	= proc_dointvec
4982	},
4983#endif
4984};
4985
4986static const struct ctl_table signal_table[] = {
4987	{
4988		.procname	= "print-fatal-signals",
4989		.data		= &print_fatal_signals,
4990		.maxlen		= sizeof(int),
4991		.mode		= 0644,
4992		.proc_handler	= proc_dointvec,
4993	},
4994};
4995
4996static int __init init_signal_sysctls(void)
4997{
4998	register_sysctl_init("debug", signal_debug_table);
4999	register_sysctl_init("kernel", signal_table);
5000	return 0;
5001}
5002early_initcall(init_signal_sysctls);
5003#endif /* CONFIG_SYSCTL */
5004
5005void __init signals_init(void)
5006{
5007	siginfo_buildtime_checks();
5008
5009	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
5010}
5011
5012#ifdef CONFIG_KGDB_KDB
5013#include <linux/kdb.h>
5014/*
5015 * kdb_send_sig - Allows kdb to send signals without exposing
5016 * signal internals.  This function checks if the required locks are
5017 * available before calling the main signal code, to avoid kdb
5018 * deadlocks.
5019 */
5020void kdb_send_sig(struct task_struct *t, int sig)
5021{
5022	static struct task_struct *kdb_prev_t;
5023	int new_t, ret;
5024	if (!spin_trylock(&t->sighand->siglock)) {
5025		kdb_printf("Can't do kill command now.\n"
5026			   "The sigmask lock is held somewhere else in "
5027			   "kernel, try again later\n");
5028		return;
5029	}
5030	new_t = kdb_prev_t != t;
5031	kdb_prev_t = t;
5032	if (!task_is_running(t) && new_t) {
5033		spin_unlock(&t->sighand->siglock);
5034		kdb_printf("Process is not RUNNING, sending a signal from "
5035			   "kdb risks deadlock\n"
5036			   "on the run queue locks. "
5037			   "The signal has _not_ been sent.\n"
5038			   "Reissue the kill command if you want to risk "
5039			   "the deadlock.\n");
5040		return;
5041	}
5042	ret = send_signal_locked(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
5043	spin_unlock(&t->sighand->siglock);
5044	if (ret)
5045		kdb_printf("Fail to deliver Signal %d to process %d.\n",
5046			   sig, t->pid);
5047	else
5048		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
5049}
5050#endif	/* CONFIG_KGDB_KDB */