linux/kernel/signal.c
<<
>>
Prefs
   1/*
   2 *  linux/kernel/signal.c
   3 *
   4 *  Copyright (C) 1991, 1992  Linus Torvalds
   5 *
   6 *  1997-11-02  Modified for POSIX.1b signals by Richard Henderson
   7 *
   8 *  2003-06-02  Jim Houston - Concurrent Computer Corp.
   9 *              Changes to use preallocated sigqueue structures
  10 *              to allow signals to be sent reliably.
  11 */
  12
  13#include <linux/slab.h>
  14#include <linux/export.h>
  15#include <linux/init.h>
  16#include <linux/sched.h>
  17#include <linux/fs.h>
  18#include <linux/tty.h>
  19#include <linux/binfmts.h>
  20#include <linux/coredump.h>
  21#include <linux/security.h>
  22#include <linux/syscalls.h>
  23#include <linux/ptrace.h>
  24#include <linux/signal.h>
  25#include <linux/signalfd.h>
  26#include <linux/ratelimit.h>
  27#include <linux/tracehook.h>
  28#include <linux/capability.h>
  29#include <linux/freezer.h>
  30#include <linux/pid_namespace.h>
  31#include <linux/nsproxy.h>
  32#include <linux/user_namespace.h>
  33#include <linux/uprobes.h>
  34#define CREATE_TRACE_POINTS
  35#include <trace/events/signal.h>
  36
  37#include <asm/param.h>
  38#include <asm/uaccess.h>
  39#include <asm/unistd.h>
  40#include <asm/siginfo.h>
  41#include <asm/cacheflush.h>
  42#include "audit.h"      /* audit_signal_info() */
  43
  44/*
  45 * SLAB caches for signal bits.
  46 */
  47
  48static struct kmem_cache *sigqueue_cachep;
  49
  50int print_fatal_signals __read_mostly;
  51
  52static void __user *sig_handler(struct task_struct *t, int sig)
  53{
  54        return t->sighand->action[sig - 1].sa.sa_handler;
  55}
  56
  57static int sig_handler_ignored(void __user *handler, int sig)
  58{
  59        /* Is it explicitly or implicitly ignored? */
  60        return handler == SIG_IGN ||
  61                (handler == SIG_DFL && sig_kernel_ignore(sig));
  62}
  63
  64static int sig_task_ignored(struct task_struct *t, int sig, bool force)
  65{
  66        void __user *handler;
  67
  68        handler = sig_handler(t, sig);
  69
  70        if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
  71                        handler == SIG_DFL && !force)
  72                return 1;
  73
  74        return sig_handler_ignored(handler, sig);
  75}
  76
  77static int sig_ignored(struct task_struct *t, int sig, bool force)
  78{
  79        /*
  80         * Blocked signals are never ignored, since the
  81         * signal handler may change by the time it is
  82         * unblocked.
  83         */
  84        if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
  85                return 0;
  86
  87        if (!sig_task_ignored(t, sig, force))
  88                return 0;
  89
  90        /*
  91         * Tracers may want to know about even ignored signals.
  92         */
  93        return !t->ptrace;
  94}
  95
  96/*
  97 * Re-calculate pending state from the set of locally pending
  98 * signals, globally pending signals, and blocked signals.
  99 */
 100static inline int has_pending_signals(sigset_t *signal, sigset_t *blocked)
 101{
 102        unsigned long ready;
 103        long i;
 104
 105        switch (_NSIG_WORDS) {
 106        default:
 107                for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
 108                        ready |= signal->sig[i] &~ blocked->sig[i];
 109                break;
 110
 111        case 4: ready  = signal->sig[3] &~ blocked->sig[3];
 112                ready |= signal->sig[2] &~ blocked->sig[2];
 113                ready |= signal->sig[1] &~ blocked->sig[1];
 114                ready |= signal->sig[0] &~ blocked->sig[0];
 115                break;
 116
 117        case 2: ready  = signal->sig[1] &~ blocked->sig[1];
 118                ready |= signal->sig[0] &~ blocked->sig[0];
 119                break;
 120
 121        case 1: ready  = signal->sig[0] &~ blocked->sig[0];
 122        }
 123        return ready != 0;
 124}
 125
 126#define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
 127
 128static int recalc_sigpending_tsk(struct task_struct *t)
 129{
 130        if ((t->jobctl & JOBCTL_PENDING_MASK) ||
 131            PENDING(&t->pending, &t->blocked) ||
 132            PENDING(&t->signal->shared_pending, &t->blocked)) {
 133                set_tsk_thread_flag(t, TIF_SIGPENDING);
 134                return 1;
 135        }
 136        /*
 137         * We must never clear the flag in another thread, or in current
 138         * when it's possible the current syscall is returning -ERESTART*.
 139         * So we don't clear it here, and only callers who know they should do.
 140         */
 141        return 0;
 142}
 143
 144/*
 145 * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
 146 * This is superfluous when called on current, the wakeup is a harmless no-op.
 147 */
 148void recalc_sigpending_and_wake(struct task_struct *t)
 149{
 150        if (recalc_sigpending_tsk(t))
 151                signal_wake_up(t, 0);
 152}
 153
 154void recalc_sigpending(void)
 155{
 156        if (!recalc_sigpending_tsk(current) && !freezing(current))
 157                clear_thread_flag(TIF_SIGPENDING);
 158
 159}
 160
 161/* Given the mask, find the first available signal that should be serviced. */
 162
 163#define SYNCHRONOUS_MASK \
 164        (sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
 165         sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
 166
 167int next_signal(struct sigpending *pending, sigset_t *mask)
 168{
 169        unsigned long i, *s, *m, x;
 170        int sig = 0;
 171
 172        s = pending->signal.sig;
 173        m = mask->sig;
 174
 175        /*
 176         * Handle the first word specially: it contains the
 177         * synchronous signals that need to be dequeued first.
 178         */
 179        x = *s &~ *m;
 180        if (x) {
 181                if (x & SYNCHRONOUS_MASK)
 182                        x &= SYNCHRONOUS_MASK;
 183                sig = ffz(~x) + 1;
 184                return sig;
 185        }
 186
 187        switch (_NSIG_WORDS) {
 188        default:
 189                for (i = 1; i < _NSIG_WORDS; ++i) {
 190                        x = *++s &~ *++m;
 191                        if (!x)
 192                                continue;
 193                        sig = ffz(~x) + i*_NSIG_BPW + 1;
 194                        break;
 195                }
 196                break;
 197
 198        case 2:
 199                x = s[1] &~ m[1];
 200                if (!x)
 201                        break;
 202                sig = ffz(~x) + _NSIG_BPW + 1;
 203                break;
 204
 205        case 1:
 206                /* Nothing to do */
 207                break;
 208        }
 209
 210        return sig;
 211}
 212
 213static inline void print_dropped_signal(int sig)
 214{
 215        static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
 216
 217        if (!print_fatal_signals)
 218                return;
 219
 220        if (!__ratelimit(&ratelimit_state))
 221                return;
 222
 223        printk(KERN_INFO "%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
 224                                current->comm, current->pid, sig);
 225}
 226
 227/**
 228 * task_set_jobctl_pending - set jobctl pending bits
 229 * @task: target task
 230 * @mask: pending bits to set
 231 *
 232 * Clear @mask from @task->jobctl.  @mask must be subset of
 233 * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
 234 * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
 235 * cleared.  If @task is already being killed or exiting, this function
 236 * becomes noop.
 237 *
 238 * CONTEXT:
 239 * Must be called with @task->sighand->siglock held.
 240 *
 241 * RETURNS:
 242 * %true if @mask is set, %false if made noop because @task was dying.
 243 */
 244bool task_set_jobctl_pending(struct task_struct *task, unsigned int mask)
 245{
 246        BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
 247                        JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
 248        BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
 249
 250        if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
 251                return false;
 252
 253        if (mask & JOBCTL_STOP_SIGMASK)
 254                task->jobctl &= ~JOBCTL_STOP_SIGMASK;
 255
 256        task->jobctl |= mask;
 257        return true;
 258}
 259
 260/**
 261 * task_clear_jobctl_trapping - clear jobctl trapping bit
 262 * @task: target task
 263 *
 264 * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
 265 * Clear it and wake up the ptracer.  Note that we don't need any further
 266 * locking.  @task->siglock guarantees that @task->parent points to the
 267 * ptracer.
 268 *
 269 * CONTEXT:
 270 * Must be called with @task->sighand->siglock held.
 271 */
 272void task_clear_jobctl_trapping(struct task_struct *task)
 273{
 274        if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
 275                task->jobctl &= ~JOBCTL_TRAPPING;
 276                wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
 277        }
 278}
 279
 280/**
 281 * task_clear_jobctl_pending - clear jobctl pending bits
 282 * @task: target task
 283 * @mask: pending bits to clear
 284 *
 285 * Clear @mask from @task->jobctl.  @mask must be subset of
 286 * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
 287 * STOP bits are cleared together.
 288 *
 289 * If clearing of @mask leaves no stop or trap pending, this function calls
 290 * task_clear_jobctl_trapping().
 291 *
 292 * CONTEXT:
 293 * Must be called with @task->sighand->siglock held.
 294 */
 295void task_clear_jobctl_pending(struct task_struct *task, unsigned int mask)
 296{
 297        BUG_ON(mask & ~JOBCTL_PENDING_MASK);
 298
 299        if (mask & JOBCTL_STOP_PENDING)
 300                mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
 301
 302        task->jobctl &= ~mask;
 303
 304        if (!(task->jobctl & JOBCTL_PENDING_MASK))
 305                task_clear_jobctl_trapping(task);
 306}
 307
 308/**
 309 * task_participate_group_stop - participate in a group stop
 310 * @task: task participating in a group stop
 311 *
 312 * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
 313 * Group stop states are cleared and the group stop count is consumed if
 314 * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
 315 * stop, the appropriate %SIGNAL_* flags are set.
 316 *
 317 * CONTEXT:
 318 * Must be called with @task->sighand->siglock held.
 319 *
 320 * RETURNS:
 321 * %true if group stop completion should be notified to the parent, %false
 322 * otherwise.
 323 */
 324static bool task_participate_group_stop(struct task_struct *task)
 325{
 326        struct signal_struct *sig = task->signal;
 327        bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
 328
 329        WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
 330
 331        task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
 332
 333        if (!consume)
 334                return false;
 335
 336        if (!WARN_ON_ONCE(sig->group_stop_count == 0))
 337                sig->group_stop_count--;
 338
 339        /*
 340         * Tell the caller to notify completion iff we are entering into a
 341         * fresh group stop.  Read comment in do_signal_stop() for details.
 342         */
 343        if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
 344                sig->flags = SIGNAL_STOP_STOPPED;
 345                return true;
 346        }
 347        return false;
 348}
 349
 350/*
 351 * allocate a new signal queue record
 352 * - this may be called without locks if and only if t == current, otherwise an
 353 *   appropriate lock must be held to stop the target task from exiting
 354 */
 355static struct sigqueue *
 356__sigqueue_alloc(int sig, struct task_struct *t, gfp_t flags, int override_rlimit)
 357{
 358        struct sigqueue *q = NULL;
 359        struct user_struct *user;
 360
 361        /*
 362         * Protect access to @t credentials. This can go away when all
 363         * callers hold rcu read lock.
 364         */
 365        rcu_read_lock();
 366        user = get_uid(__task_cred(t)->user);
 367        atomic_inc(&user->sigpending);
 368        rcu_read_unlock();
 369
 370        if (override_rlimit ||
 371            atomic_read(&user->sigpending) <=
 372                        task_rlimit(t, RLIMIT_SIGPENDING)) {
 373                q = kmem_cache_alloc(sigqueue_cachep, flags);
 374        } else {
 375                print_dropped_signal(sig);
 376        }
 377
 378        if (unlikely(q == NULL)) {
 379                atomic_dec(&user->sigpending);
 380                free_uid(user);
 381        } else {
 382                INIT_LIST_HEAD(&q->list);
 383                q->flags = 0;
 384                q->user = user;
 385        }
 386
 387        return q;
 388}
 389
 390static void __sigqueue_free(struct sigqueue *q)
 391{
 392        if (q->flags & SIGQUEUE_PREALLOC)
 393                return;
 394        atomic_dec(&q->user->sigpending);
 395        free_uid(q->user);
 396        kmem_cache_free(sigqueue_cachep, q);
 397}
 398
 399void flush_sigqueue(struct sigpending *queue)
 400{
 401        struct sigqueue *q;
 402
 403        sigemptyset(&queue->signal);
 404        while (!list_empty(&queue->list)) {
 405                q = list_entry(queue->list.next, struct sigqueue , list);
 406                list_del_init(&q->list);
 407                __sigqueue_free(q);
 408        }
 409}
 410
 411/*
 412 * Flush all pending signals for a task.
 413 */
 414void __flush_signals(struct task_struct *t)
 415{
 416        clear_tsk_thread_flag(t, TIF_SIGPENDING);
 417        flush_sigqueue(&t->pending);
 418        flush_sigqueue(&t->signal->shared_pending);
 419}
 420
 421void flush_signals(struct task_struct *t)
 422{
 423        unsigned long flags;
 424
 425        spin_lock_irqsave(&t->sighand->siglock, flags);
 426        __flush_signals(t);
 427        spin_unlock_irqrestore(&t->sighand->siglock, flags);
 428}
 429
 430static void __flush_itimer_signals(struct sigpending *pending)
 431{
 432        sigset_t signal, retain;
 433        struct sigqueue *q, *n;
 434
 435        signal = pending->signal;
 436        sigemptyset(&retain);
 437
 438        list_for_each_entry_safe(q, n, &pending->list, list) {
 439                int sig = q->info.si_signo;
 440
 441                if (likely(q->info.si_code != SI_TIMER)) {
 442                        sigaddset(&retain, sig);
 443                } else {
 444                        sigdelset(&signal, sig);
 445                        list_del_init(&q->list);
 446                        __sigqueue_free(q);
 447                }
 448        }
 449
 450        sigorsets(&pending->signal, &signal, &retain);
 451}
 452
 453void flush_itimer_signals(void)
 454{
 455        struct task_struct *tsk = current;
 456        unsigned long flags;
 457
 458        spin_lock_irqsave(&tsk->sighand->siglock, flags);
 459        __flush_itimer_signals(&tsk->pending);
 460        __flush_itimer_signals(&tsk->signal->shared_pending);
 461        spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
 462}
 463
 464void ignore_signals(struct task_struct *t)
 465{
 466        int i;
 467
 468        for (i = 0; i < _NSIG; ++i)
 469                t->sighand->action[i].sa.sa_handler = SIG_IGN;
 470
 471        flush_signals(t);
 472}
 473
 474/*
 475 * Flush all handlers for a task.
 476 */
 477
 478void
 479flush_signal_handlers(struct task_struct *t, int force_default)
 480{
 481        int i;
 482        struct k_sigaction *ka = &t->sighand->action[0];
 483        for (i = _NSIG ; i != 0 ; i--) {
 484                if (force_default || ka->sa.sa_handler != SIG_IGN)
 485                        ka->sa.sa_handler = SIG_DFL;
 486                ka->sa.sa_flags = 0;
 487                sigemptyset(&ka->sa.sa_mask);
 488                ka++;
 489        }
 490}
 491
 492int unhandled_signal(struct task_struct *tsk, int sig)
 493{
 494        void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
 495        if (is_global_init(tsk))
 496                return 1;
 497        if (handler != SIG_IGN && handler != SIG_DFL)
 498                return 0;
 499        /* if ptraced, let the tracer determine */
 500        return !tsk->ptrace;
 501}
 502
 503/*
 504 * Notify the system that a driver wants to block all signals for this
 505 * process, and wants to be notified if any signals at all were to be
 506 * sent/acted upon.  If the notifier routine returns non-zero, then the
 507 * signal will be acted upon after all.  If the notifier routine returns 0,
 508 * then then signal will be blocked.  Only one block per process is
 509 * allowed.  priv is a pointer to private data that the notifier routine
 510 * can use to determine if the signal should be blocked or not.
 511 */
 512void
 513block_all_signals(int (*notifier)(void *priv), void *priv, sigset_t *mask)
 514{
 515        unsigned long flags;
 516
 517        spin_lock_irqsave(&current->sighand->siglock, flags);
 518        current->notifier_mask = mask;
 519        current->notifier_data = priv;
 520        current->notifier = notifier;
 521        spin_unlock_irqrestore(&current->sighand->siglock, flags);
 522}
 523
 524/* Notify the system that blocking has ended. */
 525
 526void
 527unblock_all_signals(void)
 528{
 529        unsigned long flags;
 530
 531        spin_lock_irqsave(&current->sighand->siglock, flags);
 532        current->notifier = NULL;
 533        current->notifier_data = NULL;
 534        recalc_sigpending();
 535        spin_unlock_irqrestore(&current->sighand->siglock, flags);
 536}
 537
 538static void collect_signal(int sig, struct sigpending *list, siginfo_t *info)
 539{
 540        struct sigqueue *q, *first = NULL;
 541
 542        /*
 543         * Collect the siginfo appropriate to this signal.  Check if
 544         * there is another siginfo for the same signal.
 545        */
 546        list_for_each_entry(q, &list->list, list) {
 547                if (q->info.si_signo == sig) {
 548                        if (first)
 549                                goto still_pending;
 550                        first = q;
 551                }
 552        }
 553
 554        sigdelset(&list->signal, sig);
 555
 556        if (first) {
 557still_pending:
 558                list_del_init(&first->list);
 559                copy_siginfo(info, &first->info);
 560                __sigqueue_free(first);
 561        } else {
 562                /*
 563                 * Ok, it wasn't in the queue.  This must be
 564                 * a fast-pathed signal or we must have been
 565                 * out of queue space.  So zero out the info.
 566                 */
 567                info->si_signo = sig;
 568                info->si_errno = 0;
 569                info->si_code = SI_USER;
 570                info->si_pid = 0;
 571                info->si_uid = 0;
 572        }
 573}
 574
 575static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
 576                        siginfo_t *info)
 577{
 578        int sig = next_signal(pending, mask);
 579
 580        if (sig) {
 581                if (current->notifier) {
 582                        if (sigismember(current->notifier_mask, sig)) {
 583                                if (!(current->notifier)(current->notifier_data)) {
 584                                        clear_thread_flag(TIF_SIGPENDING);
 585                                        return 0;
 586                                }
 587                        }
 588                }
 589
 590                collect_signal(sig, pending, info);
 591        }
 592
 593        return sig;
 594}
 595
 596/*
 597 * Dequeue a signal and return the element to the caller, which is
 598 * expected to free it.
 599 *
 600 * All callers have to hold the siglock.
 601 */
 602int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info)
 603{
 604        int signr;
 605
 606        /* We only dequeue private signals from ourselves, we don't let
 607         * signalfd steal them
 608         */
 609        signr = __dequeue_signal(&tsk->pending, mask, info);
 610        if (!signr) {
 611                signr = __dequeue_signal(&tsk->signal->shared_pending,
 612                                         mask, info);
 613                /*
 614                 * itimer signal ?
 615                 *
 616                 * itimers are process shared and we restart periodic
 617                 * itimers in the signal delivery path to prevent DoS
 618                 * attacks in the high resolution timer case. This is
 619                 * compliant with the old way of self-restarting
 620                 * itimers, as the SIGALRM is a legacy signal and only
 621                 * queued once. Changing the restart behaviour to
 622                 * restart the timer in the signal dequeue path is
 623                 * reducing the timer noise on heavy loaded !highres
 624                 * systems too.
 625                 */
 626                if (unlikely(signr == SIGALRM)) {
 627                        struct hrtimer *tmr = &tsk->signal->real_timer;
 628
 629                        if (!hrtimer_is_queued(tmr) &&
 630                            tsk->signal->it_real_incr.tv64 != 0) {
 631                                hrtimer_forward(tmr, tmr->base->get_time(),
 632                                                tsk->signal->it_real_incr);
 633                                hrtimer_restart(tmr);
 634                        }
 635                }
 636        }
 637
 638        recalc_sigpending();
 639        if (!signr)
 640                return 0;
 641
 642        if (unlikely(sig_kernel_stop(signr))) {
 643                /*
 644                 * Set a marker that we have dequeued a stop signal.  Our
 645                 * caller might release the siglock and then the pending
 646                 * stop signal it is about to process is no longer in the
 647                 * pending bitmasks, but must still be cleared by a SIGCONT
 648                 * (and overruled by a SIGKILL).  So those cases clear this
 649                 * shared flag after we've set it.  Note that this flag may
 650                 * remain set after the signal we return is ignored or
 651                 * handled.  That doesn't matter because its only purpose
 652                 * is to alert stop-signal processing code when another
 653                 * processor has come along and cleared the flag.
 654                 */
 655                current->jobctl |= JOBCTL_STOP_DEQUEUED;
 656        }
 657        if ((info->si_code & __SI_MASK) == __SI_TIMER && info->si_sys_private) {
 658                /*
 659                 * Release the siglock to ensure proper locking order
 660                 * of timer locks outside of siglocks.  Note, we leave
 661                 * irqs disabled here, since the posix-timers code is
 662                 * about to disable them again anyway.
 663                 */
 664                spin_unlock(&tsk->sighand->siglock);
 665                do_schedule_next_timer(info);
 666                spin_lock(&tsk->sighand->siglock);
 667        }
 668        return signr;
 669}
 670
 671/*
 672 * Tell a process that it has a new active signal..
 673 *
 674 * NOTE! we rely on the previous spin_lock to
 675 * lock interrupts for us! We can only be called with
 676 * "siglock" held, and the local interrupt must
 677 * have been disabled when that got acquired!
 678 *
 679 * No need to set need_resched since signal event passing
 680 * goes through ->blocked
 681 */
 682void signal_wake_up(struct task_struct *t, int resume)
 683{
 684        unsigned int mask;
 685
 686        set_tsk_thread_flag(t, TIF_SIGPENDING);
 687
 688        /*
 689         * For SIGKILL, we want to wake it up in the stopped/traced/killable
 690         * case. We don't check t->state here because there is a race with it
 691         * executing another processor and just now entering stopped state.
 692         * By using wake_up_state, we ensure the process will wake up and
 693         * handle its death signal.
 694         */
 695        mask = TASK_INTERRUPTIBLE;
 696        if (resume)
 697                mask |= TASK_WAKEKILL;
 698        if (!wake_up_state(t, mask))
 699                kick_process(t);
 700}
 701
 702/*
 703 * Remove signals in mask from the pending set and queue.
 704 * Returns 1 if any signals were found.
 705 *
 706 * All callers must be holding the siglock.
 707 *
 708 * This version takes a sigset mask and looks at all signals,
 709 * not just those in the first mask word.
 710 */
 711static int rm_from_queue_full(sigset_t *mask, struct sigpending *s)
 712{
 713        struct sigqueue *q, *n;
 714        sigset_t m;
 715
 716        sigandsets(&m, mask, &s->signal);
 717        if (sigisemptyset(&m))
 718                return 0;
 719
 720        sigandnsets(&s->signal, &s->signal, mask);
 721        list_for_each_entry_safe(q, n, &s->list, list) {
 722                if (sigismember(mask, q->info.si_signo)) {
 723                        list_del_init(&q->list);
 724                        __sigqueue_free(q);
 725                }
 726        }
 727        return 1;
 728}
 729/*
 730 * Remove signals in mask from the pending set and queue.
 731 * Returns 1 if any signals were found.
 732 *
 733 * All callers must be holding the siglock.
 734 */
 735static int rm_from_queue(unsigned long mask, struct sigpending *s)
 736{
 737        struct sigqueue *q, *n;
 738
 739        if (!sigtestsetmask(&s->signal, mask))
 740                return 0;
 741
 742        sigdelsetmask(&s->signal, mask);
 743        list_for_each_entry_safe(q, n, &s->list, list) {
 744                if (q->info.si_signo < SIGRTMIN &&
 745                    (mask & sigmask(q->info.si_signo))) {
 746                        list_del_init(&q->list);
 747                        __sigqueue_free(q);
 748                }
 749        }
 750        return 1;
 751}
 752
 753static inline int is_si_special(const struct siginfo *info)
 754{
 755        return info <= SEND_SIG_FORCED;
 756}
 757
 758static inline bool si_fromuser(const struct siginfo *info)
 759{
 760        return info == SEND_SIG_NOINFO ||
 761                (!is_si_special(info) && SI_FROMUSER(info));
 762}
 763
 764/*
 765 * called with RCU read lock from check_kill_permission()
 766 */
 767static int kill_ok_by_cred(struct task_struct *t)
 768{
 769        const struct cred *cred = current_cred();
 770        const struct cred *tcred = __task_cred(t);
 771
 772        if (uid_eq(cred->euid, tcred->suid) ||
 773            uid_eq(cred->euid, tcred->uid)  ||
 774            uid_eq(cred->uid,  tcred->suid) ||
 775            uid_eq(cred->uid,  tcred->uid))
 776                return 1;
 777
 778        if (ns_capable(tcred->user_ns, CAP_KILL))
 779                return 1;
 780
 781        return 0;
 782}
 783
 784/*
 785 * Bad permissions for sending the signal
 786 * - the caller must hold the RCU read lock
 787 */
 788static int check_kill_permission(int sig, struct siginfo *info,
 789                                 struct task_struct *t)
 790{
 791        struct pid *sid;
 792        int error;
 793
 794        if (!valid_signal(sig))
 795                return -EINVAL;
 796
 797        if (!si_fromuser(info))
 798                return 0;
 799
 800        error = audit_signal_info(sig, t); /* Let audit system see the signal */
 801        if (error)
 802                return error;
 803
 804        if (!same_thread_group(current, t) &&
 805            !kill_ok_by_cred(t)) {
 806                switch (sig) {
 807                case SIGCONT:
 808                        sid = task_session(t);
 809                        /*
 810                         * We don't return the error if sid == NULL. The
 811                         * task was unhashed, the caller must notice this.
 812                         */
 813                        if (!sid || sid == task_session(current))
 814                                break;
 815                default:
 816                        return -EPERM;
 817                }
 818        }
 819
 820        return security_task_kill(t, info, sig, 0);
 821}
 822
 823/**
 824 * ptrace_trap_notify - schedule trap to notify ptracer
 825 * @t: tracee wanting to notify tracer
 826 *
 827 * This function schedules sticky ptrace trap which is cleared on the next
 828 * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
 829 * ptracer.
 830 *
 831 * If @t is running, STOP trap will be taken.  If trapped for STOP and
 832 * ptracer is listening for events, tracee is woken up so that it can
 833 * re-trap for the new event.  If trapped otherwise, STOP trap will be
 834 * eventually taken without returning to userland after the existing traps
 835 * are finished by PTRACE_CONT.
 836 *
 837 * CONTEXT:
 838 * Must be called with @task->sighand->siglock held.
 839 */
 840static void ptrace_trap_notify(struct task_struct *t)
 841{
 842        WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
 843        assert_spin_locked(&t->sighand->siglock);
 844
 845        task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
 846        signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
 847}
 848
 849/*
 850 * Handle magic process-wide effects of stop/continue signals. Unlike
 851 * the signal actions, these happen immediately at signal-generation
 852 * time regardless of blocking, ignoring, or handling.  This does the
 853 * actual continuing for SIGCONT, but not the actual stopping for stop
 854 * signals. The process stop is done as a signal action for SIG_DFL.
 855 *
 856 * Returns true if the signal should be actually delivered, otherwise
 857 * it should be dropped.
 858 */
 859static int prepare_signal(int sig, struct task_struct *p, bool force)
 860{
 861        struct signal_struct *signal = p->signal;
 862        struct task_struct *t;
 863
 864        if (unlikely(signal->flags & SIGNAL_GROUP_EXIT)) {
 865                /*
 866                 * The process is in the middle of dying, nothing to do.
 867                 */
 868        } else if (sig_kernel_stop(sig)) {
 869                /*
 870                 * This is a stop signal.  Remove SIGCONT from all queues.
 871                 */
 872                rm_from_queue(sigmask(SIGCONT), &signal->shared_pending);
 873                t = p;
 874                do {
 875                        rm_from_queue(sigmask(SIGCONT), &t->pending);
 876                } while_each_thread(p, t);
 877        } else if (sig == SIGCONT) {
 878                unsigned int why;
 879                /*
 880                 * Remove all stop signals from all queues, wake all threads.
 881                 */
 882                rm_from_queue(SIG_KERNEL_STOP_MASK, &signal->shared_pending);
 883                t = p;
 884                do {
 885                        task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
 886                        rm_from_queue(SIG_KERNEL_STOP_MASK, &t->pending);
 887                        if (likely(!(t->ptrace & PT_SEIZED)))
 888                                wake_up_state(t, __TASK_STOPPED);
 889                        else
 890                                ptrace_trap_notify(t);
 891                } while_each_thread(p, t);
 892
 893                /*
 894                 * Notify the parent with CLD_CONTINUED if we were stopped.
 895                 *
 896                 * If we were in the middle of a group stop, we pretend it
 897                 * was already finished, and then continued. Since SIGCHLD
 898                 * doesn't queue we report only CLD_STOPPED, as if the next
 899                 * CLD_CONTINUED was dropped.
 900                 */
 901                why = 0;
 902                if (signal->flags & SIGNAL_STOP_STOPPED)
 903                        why |= SIGNAL_CLD_CONTINUED;
 904                else if (signal->group_stop_count)
 905                        why |= SIGNAL_CLD_STOPPED;
 906
 907                if (why) {
 908                        /*
 909                         * The first thread which returns from do_signal_stop()
 910                         * will take ->siglock, notice SIGNAL_CLD_MASK, and
 911                         * notify its parent. See get_signal_to_deliver().
 912                         */
 913                        signal->flags = why | SIGNAL_STOP_CONTINUED;
 914                        signal->group_stop_count = 0;
 915                        signal->group_exit_code = 0;
 916                }
 917        }
 918
 919        return !sig_ignored(p, sig, force);
 920}
 921
 922/*
 923 * Test if P wants to take SIG.  After we've checked all threads with this,
 924 * it's equivalent to finding no threads not blocking SIG.  Any threads not
 925 * blocking SIG were ruled out because they are not running and already
 926 * have pending signals.  Such threads will dequeue from the shared queue
 927 * as soon as they're available, so putting the signal on the shared queue
 928 * will be equivalent to sending it to one such thread.
 929 */
 930static inline int wants_signal(int sig, struct task_struct *p)
 931{
 932        if (sigismember(&p->blocked, sig))
 933                return 0;
 934        if (p->flags & PF_EXITING)
 935                return 0;
 936        if (sig == SIGKILL)
 937                return 1;
 938        if (task_is_stopped_or_traced(p))
 939                return 0;
 940        return task_curr(p) || !signal_pending(p);
 941}
 942
 943static void complete_signal(int sig, struct task_struct *p, int group)
 944{
 945        struct signal_struct *signal = p->signal;
 946        struct task_struct *t;
 947
 948        /*
 949         * Now find a thread we can wake up to take the signal off the queue.
 950         *
 951         * If the main thread wants the signal, it gets first crack.
 952         * Probably the least surprising to the average bear.
 953         */
 954        if (wants_signal(sig, p))
 955                t = p;
 956        else if (!group || thread_group_empty(p))
 957                /*
 958                 * There is just one thread and it does not need to be woken.
 959                 * It will dequeue unblocked signals before it runs again.
 960                 */
 961                return;
 962        else {
 963                /*
 964                 * Otherwise try to find a suitable thread.
 965                 */
 966                t = signal->curr_target;
 967                while (!wants_signal(sig, t)) {
 968                        t = next_thread(t);
 969                        if (t == signal->curr_target)
 970                                /*
 971                                 * No thread needs to be woken.
 972                                 * Any eligible threads will see
 973                                 * the signal in the queue soon.
 974                                 */
 975                                return;
 976                }
 977                signal->curr_target = t;
 978        }
 979
 980        /*
 981         * Found a killable thread.  If the signal will be fatal,
 982         * then start taking the whole group down immediately.
 983         */
 984        if (sig_fatal(p, sig) &&
 985            !(signal->flags & (SIGNAL_UNKILLABLE | SIGNAL_GROUP_EXIT)) &&
 986            !sigismember(&t->real_blocked, sig) &&
 987            (sig == SIGKILL || !t->ptrace)) {
 988                /*
 989                 * This signal will be fatal to the whole group.
 990                 */
 991                if (!sig_kernel_coredump(sig)) {
 992                        /*
 993                         * Start a group exit and wake everybody up.
 994                         * This way we don't have other threads
 995                         * running and doing things after a slower
 996                         * thread has the fatal signal pending.
 997                         */
 998                        signal->flags = SIGNAL_GROUP_EXIT;
 999                        signal->group_exit_code = sig;
1000                        signal->group_stop_count = 0;
1001                        t = p;
1002                        do {
1003                                task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1004                                sigaddset(&t->pending.signal, SIGKILL);
1005                                signal_wake_up(t, 1);
1006                        } while_each_thread(p, t);
1007                        return;
1008                }
1009        }
1010
1011        /*
1012         * The signal is already in the shared-pending queue.
1013         * Tell the chosen thread to wake up and dequeue it.
1014         */
1015        signal_wake_up(t, sig == SIGKILL);
1016        return;
1017}
1018
1019static inline int legacy_queue(struct sigpending *signals, int sig)
1020{
1021        return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1022}
1023
1024#ifdef CONFIG_USER_NS
1025static inline void userns_fixup_signal_uid(struct siginfo *info, struct task_struct *t)
1026{
1027        if (current_user_ns() == task_cred_xxx(t, user_ns))
1028                return;
1029
1030        if (SI_FROMKERNEL(info))
1031                return;
1032
1033        rcu_read_lock();
1034        info->si_uid = from_kuid_munged(task_cred_xxx(t, user_ns),
1035                                        make_kuid(current_user_ns(), info->si_uid));
1036        rcu_read_unlock();
1037}
1038#else
1039static inline void userns_fixup_signal_uid(struct siginfo *info, struct task_struct *t)
1040{
1041        return;
1042}
1043#endif
1044
1045static int __send_signal(int sig, struct siginfo *info, struct task_struct *t,
1046                        int group, int from_ancestor_ns)
1047{
1048        struct sigpending *pending;
1049        struct sigqueue *q;
1050        int override_rlimit;
1051        int ret = 0, result;
1052
1053        assert_spin_locked(&t->sighand->siglock);
1054
1055        result = TRACE_SIGNAL_IGNORED;
1056        if (!prepare_signal(sig, t,
1057                        from_ancestor_ns || (info == SEND_SIG_FORCED)))
1058                goto ret;
1059
1060        pending = group ? &t->signal->shared_pending : &t->pending;
1061        /*
1062         * Short-circuit ignored signals and support queuing
1063         * exactly one non-rt signal, so that we can get more
1064         * detailed information about the cause of the signal.
1065         */
1066        result = TRACE_SIGNAL_ALREADY_PENDING;
1067        if (legacy_queue(pending, sig))
1068                goto ret;
1069
1070        result = TRACE_SIGNAL_DELIVERED;
1071        /*
1072         * fast-pathed signals for kernel-internal things like SIGSTOP
1073         * or SIGKILL.
1074         */
1075        if (info == SEND_SIG_FORCED)
1076                goto out_set;
1077
1078        /*
1079         * Real-time signals must be queued if sent by sigqueue, or
1080         * some other real-time mechanism.  It is implementation
1081         * defined whether kill() does so.  We attempt to do so, on
1082         * the principle of least surprise, but since kill is not
1083         * allowed to fail with EAGAIN when low on memory we just
1084         * make sure at least one signal gets delivered and don't
1085         * pass on the info struct.
1086         */
1087        if (sig < SIGRTMIN)
1088                override_rlimit = (is_si_special(info) || info->si_code >= 0);
1089        else
1090                override_rlimit = 0;
1091
1092        q = __sigqueue_alloc(sig, t, GFP_ATOMIC | __GFP_NOTRACK_FALSE_POSITIVE,
1093                override_rlimit);
1094        if (q) {
1095                list_add_tail(&q->list, &pending->list);
1096                switch ((unsigned long) info) {
1097                case (unsigned long) SEND_SIG_NOINFO:
1098                        q->info.si_signo = sig;
1099                        q->info.si_errno = 0;
1100                        q->info.si_code = SI_USER;
1101                        q->info.si_pid = task_tgid_nr_ns(current,
1102                                                        task_active_pid_ns(t));
1103                        q->info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1104                        break;
1105                case (unsigned long) SEND_SIG_PRIV:
1106                        q->info.si_signo = sig;
1107                        q->info.si_errno = 0;
1108                        q->info.si_code = SI_KERNEL;
1109                        q->info.si_pid = 0;
1110                        q->info.si_uid = 0;
1111                        break;
1112                default:
1113                        copy_siginfo(&q->info, info);
1114                        if (from_ancestor_ns)
1115                                q->info.si_pid = 0;
1116                        break;
1117                }
1118
1119                userns_fixup_signal_uid(&q->info, t);
1120
1121        } else if (!is_si_special(info)) {
1122                if (sig >= SIGRTMIN && info->si_code != SI_USER) {
1123                        /*
1124                         * Queue overflow, abort.  We may abort if the
1125                         * signal was rt and sent by user using something
1126                         * other than kill().
1127                         */
1128                        result = TRACE_SIGNAL_OVERFLOW_FAIL;
1129                        ret = -EAGAIN;
1130                        goto ret;
1131                } else {
1132                        /*
1133                         * This is a silent loss of information.  We still
1134                         * send the signal, but the *info bits are lost.
1135                         */
1136                        result = TRACE_SIGNAL_LOSE_INFO;
1137                }
1138        }
1139
1140out_set:
1141        signalfd_notify(t, sig);
1142        sigaddset(&pending->signal, sig);
1143        complete_signal(sig, t, group);
1144ret:
1145        trace_signal_generate(sig, info, t, group, result);
1146        return ret;
1147}
1148
1149static int send_signal(int sig, struct siginfo *info, struct task_struct *t,
1150                        int group)
1151{
1152        int from_ancestor_ns = 0;
1153
1154#ifdef CONFIG_PID_NS
1155        from_ancestor_ns = si_fromuser(info) &&
1156                           !task_pid_nr_ns(current, task_active_pid_ns(t));
1157#endif
1158
1159        return __send_signal(sig, info, t, group, from_ancestor_ns);
1160}
1161
1162static void print_fatal_signal(struct pt_regs *regs, int signr)
1163{
1164        printk("%s/%d: potentially unexpected fatal signal %d.\n",
1165                current->comm, task_pid_nr(current), signr);
1166
1167#if defined(__i386__) && !defined(__arch_um__)
1168        printk("code at %08lx: ", regs->ip);
1169        {
1170                int i;
1171                for (i = 0; i < 16; i++) {
1172                        unsigned char insn;
1173
1174                        if (get_user(insn, (unsigned char *)(regs->ip + i)))
1175                                break;
1176                        printk("%02x ", insn);
1177                }
1178        }
1179#endif
1180        printk("\n");
1181        preempt_disable();
1182        show_regs(regs);
1183        preempt_enable();
1184}
1185
1186static int __init setup_print_fatal_signals(char *str)
1187{
1188        get_option (&str, &print_fatal_signals);
1189
1190        return 1;
1191}
1192
1193__setup("print-fatal-signals=", setup_print_fatal_signals);
1194
1195int
1196__group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1197{
1198        return send_signal(sig, info, p, 1);
1199}
1200
1201static int
1202specific_send_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1203{
1204        return send_signal(sig, info, t, 0);
1205}
1206
1207int do_send_sig_info(int sig, struct siginfo *info, struct task_struct *p,
1208                        bool group)
1209{
1210        unsigned long flags;
1211        int ret = -ESRCH;
1212
1213        if (lock_task_sighand(p, &flags)) {
1214                ret = send_signal(sig, info, p, group);
1215                unlock_task_sighand(p, &flags);
1216        }
1217
1218        return ret;
1219}
1220
1221/*
1222 * Force a signal that the process can't ignore: if necessary
1223 * we unblock the signal and change any SIG_IGN to SIG_DFL.
1224 *
1225 * Note: If we unblock the signal, we always reset it to SIG_DFL,
1226 * since we do not want to have a signal handler that was blocked
1227 * be invoked when user space had explicitly blocked it.
1228 *
1229 * We don't want to have recursive SIGSEGV's etc, for example,
1230 * that is why we also clear SIGNAL_UNKILLABLE.
1231 */
1232int
1233force_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1234{
1235        unsigned long int flags;
1236        int ret, blocked, ignored;
1237        struct k_sigaction *action;
1238
1239        spin_lock_irqsave(&t->sighand->siglock, flags);
1240        action = &t->sighand->action[sig-1];
1241        ignored = action->sa.sa_handler == SIG_IGN;
1242        blocked = sigismember(&t->blocked, sig);
1243        if (blocked || ignored) {
1244                action->sa.sa_handler = SIG_DFL;
1245                if (blocked) {
1246                        sigdelset(&t->blocked, sig);
1247                        recalc_sigpending_and_wake(t);
1248                }
1249        }
1250        if (action->sa.sa_handler == SIG_DFL)
1251                t->signal->flags &= ~SIGNAL_UNKILLABLE;
1252        ret = specific_send_sig_info(sig, info, t);
1253        spin_unlock_irqrestore(&t->sighand->siglock, flags);
1254
1255        return ret;
1256}
1257
1258/*
1259 * Nuke all other threads in the group.
1260 */
1261int zap_other_threads(struct task_struct *p)
1262{
1263        struct task_struct *t = p;
1264        int count = 0;
1265
1266        p->signal->group_stop_count = 0;
1267
1268        while_each_thread(p, t) {
1269                task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1270                count++;
1271
1272                /* Don't bother with already dead threads */
1273                if (t->exit_state)
1274                        continue;
1275                sigaddset(&t->pending.signal, SIGKILL);
1276                signal_wake_up(t, 1);
1277        }
1278
1279        return count;
1280}
1281
1282struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1283                                           unsigned long *flags)
1284{
1285        struct sighand_struct *sighand;
1286
1287        for (;;) {
1288                local_irq_save(*flags);
1289                rcu_read_lock();
1290                sighand = rcu_dereference(tsk->sighand);
1291                if (unlikely(sighand == NULL)) {
1292                        rcu_read_unlock();
1293                        local_irq_restore(*flags);
1294                        break;
1295                }
1296
1297                spin_lock(&sighand->siglock);
1298                if (likely(sighand == tsk->sighand)) {
1299                        rcu_read_unlock();
1300                        break;
1301                }
1302                spin_unlock(&sighand->siglock);
1303                rcu_read_unlock();
1304                local_irq_restore(*flags);
1305        }
1306
1307        return sighand;
1308}
1309
1310/*
1311 * send signal info to all the members of a group
1312 */
1313int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1314{
1315        int ret;
1316
1317        rcu_read_lock();
1318        ret = check_kill_permission(sig, info, p);
1319        rcu_read_unlock();
1320
1321        if (!ret && sig)
1322                ret = do_send_sig_info(sig, info, p, true);
1323
1324        return ret;
1325}
1326
1327/*
1328 * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1329 * control characters do (^C, ^Z etc)
1330 * - the caller must hold at least a readlock on tasklist_lock
1331 */
1332int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp)
1333{
1334        struct task_struct *p = NULL;
1335        int retval, success;
1336
1337        success = 0;
1338        retval = -ESRCH;
1339        do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1340                int err = group_send_sig_info(sig, info, p);
1341                success |= !err;
1342                retval = err;
1343        } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1344        return success ? 0 : retval;
1345}
1346
1347int kill_pid_info(int sig, struct siginfo *info, struct pid *pid)
1348{
1349        int error = -ESRCH;
1350        struct task_struct *p;
1351
1352        rcu_read_lock();
1353retry:
1354        p = pid_task(pid, PIDTYPE_PID);
1355        if (p) {
1356                error = group_send_sig_info(sig, info, p);
1357                if (unlikely(error == -ESRCH))
1358                        /*
1359                         * The task was unhashed in between, try again.
1360                         * If it is dead, pid_task() will return NULL,
1361                         * if we race with de_thread() it will find the
1362                         * new leader.
1363                         */
1364                        goto retry;
1365        }
1366        rcu_read_unlock();
1367
1368        return error;
1369}
1370
1371int kill_proc_info(int sig, struct siginfo *info, pid_t pid)
1372{
1373        int error;
1374        rcu_read_lock();
1375        error = kill_pid_info(sig, info, find_vpid(pid));
1376        rcu_read_unlock();
1377        return error;
1378}
1379
1380static int kill_as_cred_perm(const struct cred *cred,
1381                             struct task_struct *target)
1382{
1383        const struct cred *pcred = __task_cred(target);
1384        if (!uid_eq(cred->euid, pcred->suid) && !uid_eq(cred->euid, pcred->uid) &&
1385            !uid_eq(cred->uid,  pcred->suid) && !uid_eq(cred->uid,  pcred->uid))
1386                return 0;
1387        return 1;
1388}
1389
1390/* like kill_pid_info(), but doesn't use uid/euid of "current" */
1391int kill_pid_info_as_cred(int sig, struct siginfo *info, struct pid *pid,
1392                         const struct cred *cred, u32 secid)
1393{
1394        int ret = -EINVAL;
1395        struct task_struct *p;
1396        unsigned long flags;
1397
1398        if (!valid_signal(sig))
1399                return ret;
1400
1401        rcu_read_lock();
1402        p = pid_task(pid, PIDTYPE_PID);
1403        if (!p) {
1404                ret = -ESRCH;
1405                goto out_unlock;
1406        }
1407        if (si_fromuser(info) && !kill_as_cred_perm(cred, p)) {
1408                ret = -EPERM;
1409                goto out_unlock;
1410        }
1411        ret = security_task_kill(p, info, sig, secid);
1412        if (ret)
1413                goto out_unlock;
1414
1415        if (sig) {
1416                if (lock_task_sighand(p, &flags)) {
1417                        ret = __send_signal(sig, info, p, 1, 0);
1418                        unlock_task_sighand(p, &flags);
1419                } else
1420                        ret = -ESRCH;
1421        }
1422out_unlock:
1423        rcu_read_unlock();
1424        return ret;
1425}
1426EXPORT_SYMBOL_GPL(kill_pid_info_as_cred);
1427
1428/*
1429 * kill_something_info() interprets pid in interesting ways just like kill(2).
1430 *
1431 * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1432 * is probably wrong.  Should make it like BSD or SYSV.
1433 */
1434
1435static int kill_something_info(int sig, struct siginfo *info, pid_t pid)
1436{
1437        int ret;
1438
1439        if (pid > 0) {
1440                rcu_read_lock();
1441                ret = kill_pid_info(sig, info, find_vpid(pid));
1442                rcu_read_unlock();
1443                return ret;
1444        }
1445
1446        read_lock(&tasklist_lock);
1447        if (pid != -1) {
1448                ret = __kill_pgrp_info(sig, info,
1449                                pid ? find_vpid(-pid) : task_pgrp(current));
1450        } else {
1451                int retval = 0, count = 0;
1452                struct task_struct * p;
1453
1454                for_each_process(p) {
1455                        if (task_pid_vnr(p) > 1 &&
1456                                        !same_thread_group(p, current)) {
1457                                int err = group_send_sig_info(sig, info, p);
1458                                ++count;
1459                                if (err != -EPERM)
1460                                        retval = err;
1461                        }
1462                }
1463                ret = count ? retval : -ESRCH;
1464        }
1465        read_unlock(&tasklist_lock);
1466
1467        return ret;
1468}
1469
1470/*
1471 * These are for backward compatibility with the rest of the kernel source.
1472 */
1473
1474int send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1475{
1476        /*
1477         * Make sure legacy kernel users don't send in bad values
1478         * (normal paths check this in check_kill_permission).
1479         */
1480        if (!valid_signal(sig))
1481                return -EINVAL;
1482
1483        return do_send_sig_info(sig, info, p, false);
1484}
1485
1486#define __si_special(priv) \
1487        ((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1488
1489int
1490send_sig(int sig, struct task_struct *p, int priv)
1491{
1492        return send_sig_info(sig, __si_special(priv), p);
1493}
1494
1495void
1496force_sig(int sig, struct task_struct *p)
1497{
1498        force_sig_info(sig, SEND_SIG_PRIV, p);
1499}
1500
1501/*
1502 * When things go south during signal handling, we
1503 * will force a SIGSEGV. And if the signal that caused
1504 * the problem was already a SIGSEGV, we'll want to
1505 * make sure we don't even try to deliver the signal..
1506 */
1507int
1508force_sigsegv(int sig, struct task_struct *p)
1509{
1510        if (sig == SIGSEGV) {
1511                unsigned long flags;
1512                spin_lock_irqsave(&p->sighand->siglock, flags);
1513                p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1514                spin_unlock_irqrestore(&p->sighand->siglock, flags);
1515        }
1516        force_sig(SIGSEGV, p);
1517        return 0;
1518}
1519
1520int kill_pgrp(struct pid *pid, int sig, int priv)
1521{
1522        int ret;
1523
1524        read_lock(&tasklist_lock);
1525        ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1526        read_unlock(&tasklist_lock);
1527
1528        return ret;
1529}
1530EXPORT_SYMBOL(kill_pgrp);
1531
1532int kill_pid(struct pid *pid, int sig, int priv)
1533{
1534        return kill_pid_info(sig, __si_special(priv), pid);
1535}
1536EXPORT_SYMBOL(kill_pid);
1537
1538/*
1539 * These functions support sending signals using preallocated sigqueue
1540 * structures.  This is needed "because realtime applications cannot
1541 * afford to lose notifications of asynchronous events, like timer
1542 * expirations or I/O completions".  In the case of POSIX Timers
1543 * we allocate the sigqueue structure from the timer_create.  If this
1544 * allocation fails we are able to report the failure to the application
1545 * with an EAGAIN error.
1546 */
1547struct sigqueue *sigqueue_alloc(void)
1548{
1549        struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1550
1551        if (q)
1552                q->flags |= SIGQUEUE_PREALLOC;
1553
1554        return q;
1555}
1556
1557void sigqueue_free(struct sigqueue *q)
1558{
1559        unsigned long flags;
1560        spinlock_t *lock = &current->sighand->siglock;
1561
1562        BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1563        /*
1564         * We must hold ->siglock while testing q->list
1565         * to serialize with collect_signal() or with
1566         * __exit_signal()->flush_sigqueue().
1567         */
1568        spin_lock_irqsave(lock, flags);
1569        q->flags &= ~SIGQUEUE_PREALLOC;
1570        /*
1571         * If it is queued it will be freed when dequeued,
1572         * like the "regular" sigqueue.
1573         */
1574        if (!list_empty(&q->list))
1575                q = NULL;
1576        spin_unlock_irqrestore(lock, flags);
1577
1578        if (q)
1579                __sigqueue_free(q);
1580}
1581
1582int send_sigqueue(struct sigqueue *q, struct task_struct *t, int group)
1583{
1584        int sig = q->info.si_signo;
1585        struct sigpending *pending;
1586        unsigned long flags;
1587        int ret, result;
1588
1589        BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1590
1591        ret = -1;
1592        if (!likely(lock_task_sighand(t, &flags)))
1593                goto ret;
1594
1595        ret = 1; /* the signal is ignored */
1596        result = TRACE_SIGNAL_IGNORED;
1597        if (!prepare_signal(sig, t, false))
1598                goto out;
1599
1600        ret = 0;
1601        if (unlikely(!list_empty(&q->list))) {
1602                /*
1603                 * If an SI_TIMER entry is already queue just increment
1604                 * the overrun count.
1605                 */
1606                BUG_ON(q->info.si_code != SI_TIMER);
1607                q->info.si_overrun++;
1608                result = TRACE_SIGNAL_ALREADY_PENDING;
1609                goto out;
1610        }
1611        q->info.si_overrun = 0;
1612
1613        signalfd_notify(t, sig);
1614        pending = group ? &t->signal->shared_pending : &t->pending;
1615        list_add_tail(&q->list, &pending->list);
1616        sigaddset(&pending->signal, sig);
1617        complete_signal(sig, t, group);
1618        result = TRACE_SIGNAL_DELIVERED;
1619out:
1620        trace_signal_generate(sig, &q->info, t, group, result);
1621        unlock_task_sighand(t, &flags);
1622ret:
1623        return ret;
1624}
1625
1626/*
1627 * Let a parent know about the death of a child.
1628 * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1629 *
1630 * Returns true if our parent ignored us and so we've switched to
1631 * self-reaping.
1632 */
1633bool do_notify_parent(struct task_struct *tsk, int sig)
1634{
1635        struct siginfo info;
1636        unsigned long flags;
1637        struct sighand_struct *psig;
1638        bool autoreap = false;
1639
1640        BUG_ON(sig == -1);
1641
1642        /* do_notify_parent_cldstop should have been called instead.  */
1643        BUG_ON(task_is_stopped_or_traced(tsk));
1644
1645        BUG_ON(!tsk->ptrace &&
1646               (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1647
1648        if (sig != SIGCHLD) {
1649                /*
1650                 * This is only possible if parent == real_parent.
1651                 * Check if it has changed security domain.
1652                 */
1653                if (tsk->parent_exec_id != tsk->parent->self_exec_id)
1654                        sig = SIGCHLD;
1655        }
1656
1657        info.si_signo = sig;
1658        info.si_errno = 0;
1659        /*
1660         * We are under tasklist_lock here so our parent is tied to
1661         * us and cannot change.
1662         *
1663         * task_active_pid_ns will always return the same pid namespace
1664         * until a task passes through release_task.
1665         *
1666         * write_lock() currently calls preempt_disable() which is the
1667         * same as rcu_read_lock(), but according to Oleg, this is not
1668         * correct to rely on this
1669         */
1670        rcu_read_lock();
1671        info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1672        info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1673                                       task_uid(tsk));
1674        rcu_read_unlock();
1675
1676        info.si_utime = cputime_to_clock_t(tsk->utime + tsk->signal->utime);
1677        info.si_stime = cputime_to_clock_t(tsk->stime + tsk->signal->stime);
1678
1679        info.si_status = tsk->exit_code & 0x7f;
1680        if (tsk->exit_code & 0x80)
1681                info.si_code = CLD_DUMPED;
1682        else if (tsk->exit_code & 0x7f)
1683                info.si_code = CLD_KILLED;
1684        else {
1685                info.si_code = CLD_EXITED;
1686                info.si_status = tsk->exit_code >> 8;
1687        }
1688
1689        psig = tsk->parent->sighand;
1690        spin_lock_irqsave(&psig->siglock, flags);
1691        if (!tsk->ptrace && sig == SIGCHLD &&
1692            (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1693             (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1694                /*
1695                 * We are exiting and our parent doesn't care.  POSIX.1
1696                 * defines special semantics for setting SIGCHLD to SIG_IGN
1697                 * or setting the SA_NOCLDWAIT flag: we should be reaped
1698                 * automatically and not left for our parent's wait4 call.
1699                 * Rather than having the parent do it as a magic kind of
1700                 * signal handler, we just set this to tell do_exit that we
1701                 * can be cleaned up without becoming a zombie.  Note that
1702                 * we still call __wake_up_parent in this case, because a
1703                 * blocked sys_wait4 might now return -ECHILD.
1704                 *
1705                 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1706                 * is implementation-defined: we do (if you don't want
1707                 * it, just use SIG_IGN instead).
1708                 */
1709                autoreap = true;
1710                if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
1711                        sig = 0;
1712        }
1713        if (valid_signal(sig) && sig)
1714                __group_send_sig_info(sig, &info, tsk->parent);
1715        __wake_up_parent(tsk, tsk->parent);
1716        spin_unlock_irqrestore(&psig->siglock, flags);
1717
1718        return autoreap;
1719}
1720
1721/**
1722 * do_notify_parent_cldstop - notify parent of stopped/continued state change
1723 * @tsk: task reporting the state change
1724 * @for_ptracer: the notification is for ptracer
1725 * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
1726 *
1727 * Notify @tsk's parent that the stopped/continued state has changed.  If
1728 * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
1729 * If %true, @tsk reports to @tsk->parent which should be the ptracer.
1730 *
1731 * CONTEXT:
1732 * Must be called with tasklist_lock at least read locked.
1733 */
1734static void do_notify_parent_cldstop(struct task_struct *tsk,
1735                                     bool for_ptracer, int why)
1736{
1737        struct siginfo info;
1738        unsigned long flags;
1739        struct task_struct *parent;
1740        struct sighand_struct *sighand;
1741
1742        if (for_ptracer) {
1743                parent = tsk->parent;
1744        } else {
1745                tsk = tsk->group_leader;
1746                parent = tsk->real_parent;
1747        }
1748
1749        info.si_signo = SIGCHLD;
1750        info.si_errno = 0;
1751        /*
1752         * see comment in do_notify_parent() about the following 4 lines
1753         */
1754        rcu_read_lock();
1755        info.si_pid = task_pid_nr_ns(tsk, parent->nsproxy->pid_ns);
1756        info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
1757        rcu_read_unlock();
1758
1759        info.si_utime = cputime_to_clock_t(tsk->utime);
1760        info.si_stime = cputime_to_clock_t(tsk->stime);
1761
1762        info.si_code = why;
1763        switch (why) {
1764        case CLD_CONTINUED:
1765                info.si_status = SIGCONT;
1766                break;
1767        case CLD_STOPPED:
1768                info.si_status = tsk->signal->group_exit_code & 0x7f;
1769                break;
1770        case CLD_TRAPPED:
1771                info.si_status = tsk->exit_code & 0x7f;
1772                break;
1773        default:
1774                BUG();
1775        }
1776
1777        sighand = parent->sighand;
1778        spin_lock_irqsave(&sighand->siglock, flags);
1779        if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
1780            !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
1781                __group_send_sig_info(SIGCHLD, &info, parent);
1782        /*
1783         * Even if SIGCHLD is not generated, we must wake up wait4 calls.
1784         */
1785        __wake_up_parent(tsk, parent);
1786        spin_unlock_irqrestore(&sighand->siglock, flags);
1787}
1788
1789static inline int may_ptrace_stop(void)
1790{
1791        if (!likely(current->ptrace))
1792                return 0;
1793        /*
1794         * Are we in the middle of do_coredump?
1795         * If so and our tracer is also part of the coredump stopping
1796         * is a deadlock situation, and pointless because our tracer
1797         * is dead so don't allow us to stop.
1798         * If SIGKILL was already sent before the caller unlocked
1799         * ->siglock we must see ->core_state != NULL. Otherwise it
1800         * is safe to enter schedule().
1801         */
1802        if (unlikely(current->mm->core_state) &&
1803            unlikely(current->mm == current->parent->mm))
1804                return 0;
1805
1806        return 1;
1807}
1808
1809/*
1810 * Return non-zero if there is a SIGKILL that should be waking us up.
1811 * Called with the siglock held.
1812 */
1813static int sigkill_pending(struct task_struct *tsk)
1814{
1815        return  sigismember(&tsk->pending.signal, SIGKILL) ||
1816                sigismember(&tsk->signal->shared_pending.signal, SIGKILL);
1817}
1818
1819/*
1820 * This must be called with current->sighand->siglock held.
1821 *
1822 * This should be the path for all ptrace stops.
1823 * We always set current->last_siginfo while stopped here.
1824 * That makes it a way to test a stopped process for
1825 * being ptrace-stopped vs being job-control-stopped.
1826 *
1827 * If we actually decide not to stop at all because the tracer
1828 * is gone, we keep current->exit_code unless clear_code.
1829 */
1830static void ptrace_stop(int exit_code, int why, int clear_code, siginfo_t *info)
1831        __releases(&current->sighand->siglock)
1832        __acquires(&current->sighand->siglock)
1833{
1834        bool gstop_done = false;
1835
1836        if (arch_ptrace_stop_needed(exit_code, info)) {
1837                /*
1838                 * The arch code has something special to do before a
1839                 * ptrace stop.  This is allowed to block, e.g. for faults
1840                 * on user stack pages.  We can't keep the siglock while
1841                 * calling arch_ptrace_stop, so we must release it now.
1842                 * To preserve proper semantics, we must do this before
1843                 * any signal bookkeeping like checking group_stop_count.
1844                 * Meanwhile, a SIGKILL could come in before we retake the
1845                 * siglock.  That must prevent us from sleeping in TASK_TRACED.
1846                 * So after regaining the lock, we must check for SIGKILL.
1847                 */
1848                spin_unlock_irq(&current->sighand->siglock);
1849                arch_ptrace_stop(exit_code, info);
1850                spin_lock_irq(&current->sighand->siglock);
1851                if (sigkill_pending(current))
1852                        return;
1853        }
1854
1855        /*
1856         * We're committing to trapping.  TRACED should be visible before
1857         * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
1858         * Also, transition to TRACED and updates to ->jobctl should be
1859         * atomic with respect to siglock and should be done after the arch
1860         * hook as siglock is released and regrabbed across it.
1861         */
1862        set_current_state(TASK_TRACED);
1863
1864        current->last_siginfo = info;
1865        current->exit_code = exit_code;
1866
1867        /*
1868         * If @why is CLD_STOPPED, we're trapping to participate in a group
1869         * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
1870         * across siglock relocks since INTERRUPT was scheduled, PENDING
1871         * could be clear now.  We act as if SIGCONT is received after
1872         * TASK_TRACED is entered - ignore it.
1873         */
1874        if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
1875                gstop_done = task_participate_group_stop(current);
1876
1877        /* any trap clears pending STOP trap, STOP trap clears NOTIFY */
1878        task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
1879        if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
1880                task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
1881
1882        /* entering a trap, clear TRAPPING */
1883        task_clear_jobctl_trapping(current);
1884
1885        spin_unlock_irq(&current->sighand->siglock);
1886        read_lock(&tasklist_lock);
1887        if (may_ptrace_stop()) {
1888                /*
1889                 * Notify parents of the stop.
1890                 *
1891                 * While ptraced, there are two parents - the ptracer and
1892                 * the real_parent of the group_leader.  The ptracer should
1893                 * know about every stop while the real parent is only
1894                 * interested in the completion of group stop.  The states
1895                 * for the two don't interact with each other.  Notify
1896                 * separately unless they're gonna be duplicates.
1897                 */
1898                do_notify_parent_cldstop(current, true, why);
1899                if (gstop_done && ptrace_reparented(current))
1900                        do_notify_parent_cldstop(current, false, why);
1901
1902                /*
1903                 * Don't want to allow preemption here, because
1904                 * sys_ptrace() needs this task to be inactive.
1905                 *
1906                 * XXX: implement read_unlock_no_resched().
1907                 */
1908                preempt_disable();
1909                read_unlock(&tasklist_lock);
1910                preempt_enable_no_resched();
1911                schedule();
1912        } else {
1913                /*
1914                 * By the time we got the lock, our tracer went away.
1915                 * Don't drop the lock yet, another tracer may come.
1916                 *
1917                 * If @gstop_done, the ptracer went away between group stop
1918                 * completion and here.  During detach, it would have set
1919                 * JOBCTL_STOP_PENDING on us and we'll re-enter
1920                 * TASK_STOPPED in do_signal_stop() on return, so notifying
1921                 * the real parent of the group stop completion is enough.
1922                 */
1923                if (gstop_done)
1924                        do_notify_parent_cldstop(current, false, why);
1925
1926                __set_current_state(TASK_RUNNING);
1927                if (clear_code)
1928                        current->exit_code = 0;
1929                read_unlock(&tasklist_lock);
1930        }
1931
1932        /*
1933         * While in TASK_TRACED, we were considered "frozen enough".
1934         * Now that we woke up, it's crucial if we're supposed to be
1935         * frozen that we freeze now before running anything substantial.
1936         */
1937        try_to_freeze();
1938
1939        /*
1940         * We are back.  Now reacquire the siglock before touching
1941         * last_siginfo, so that we are sure to have synchronized with
1942         * any signal-sending on another CPU that wants to examine it.
1943         */
1944        spin_lock_irq(&current->sighand->siglock);
1945        current->last_siginfo = NULL;
1946
1947        /* LISTENING can be set only during STOP traps, clear it */
1948        current->jobctl &= ~JOBCTL_LISTENING;
1949
1950        /*
1951         * Queued signals ignored us while we were stopped for tracing.
1952         * So check for any that we should take before resuming user mode.
1953         * This sets TIF_SIGPENDING, but never clears it.
1954         */
1955        recalc_sigpending_tsk(current);
1956}
1957
1958static void ptrace_do_notify(int signr, int exit_code, int why)
1959{
1960        siginfo_t info;
1961
1962        memset(&info, 0, sizeof info);
1963        info.si_signo = signr;
1964        info.si_code = exit_code;
1965        info.si_pid = task_pid_vnr(current);
1966        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1967
1968        /* Let the debugger run.  */
1969        ptrace_stop(exit_code, why, 1, &info);
1970}
1971
1972void ptrace_notify(int exit_code)
1973{
1974        BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
1975        if (unlikely(current->task_works))
1976                task_work_run();
1977
1978        spin_lock_irq(&current->sighand->siglock);
1979        ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
1980        spin_unlock_irq(&current->sighand->siglock);
1981}
1982
1983/**
1984 * do_signal_stop - handle group stop for SIGSTOP and other stop signals
1985 * @signr: signr causing group stop if initiating
1986 *
1987 * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
1988 * and participate in it.  If already set, participate in the existing
1989 * group stop.  If participated in a group stop (and thus slept), %true is
1990 * returned with siglock released.
1991 *
1992 * If ptraced, this function doesn't handle stop itself.  Instead,
1993 * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
1994 * untouched.  The caller must ensure that INTERRUPT trap handling takes
1995 * places afterwards.
1996 *
1997 * CONTEXT:
1998 * Must be called with @current->sighand->siglock held, which is released
1999 * on %true return.
2000 *
2001 * RETURNS:
2002 * %false if group stop is already cancelled or ptrace trap is scheduled.
2003 * %true if participated in group stop.
2004 */
2005static bool do_signal_stop(int signr)
2006        __releases(&current->sighand->siglock)
2007{
2008        struct signal_struct *sig = current->signal;
2009
2010        if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2011                unsigned int gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2012                struct task_struct *t;
2013
2014                /* signr will be recorded in task->jobctl for retries */
2015                WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2016
2017                if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2018                    unlikely(signal_group_exit(sig)))
2019                        return false;
2020                /*
2021                 * There is no group stop already in progress.  We must
2022                 * initiate one now.
2023                 *
2024                 * While ptraced, a task may be resumed while group stop is
2025                 * still in effect and then receive a stop signal and
2026                 * initiate another group stop.  This deviates from the
2027                 * usual behavior as two consecutive stop signals can't
2028                 * cause two group stops when !ptraced.  That is why we
2029                 * also check !task_is_stopped(t) below.
2030                 *
2031                 * The condition can be distinguished by testing whether
2032                 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2033                 * group_exit_code in such case.
2034                 *
2035                 * This is not necessary for SIGNAL_STOP_CONTINUED because
2036                 * an intervening stop signal is required to cause two
2037                 * continued events regardless of ptrace.
2038                 */
2039                if (!(sig->flags & SIGNAL_STOP_STOPPED))
2040                        sig->group_exit_code = signr;
2041
2042                sig->group_stop_count = 0;
2043
2044                if (task_set_jobctl_pending(current, signr | gstop))
2045                        sig->group_stop_count++;
2046
2047                for (t = next_thread(current); t != current;
2048                     t = next_thread(t)) {
2049                        /*
2050                         * Setting state to TASK_STOPPED for a group
2051                         * stop is always done with the siglock held,
2052                         * so this check has no races.
2053                         */
2054                        if (!task_is_stopped(t) &&
2055                            task_set_jobctl_pending(t, signr | gstop)) {
2056                                sig->group_stop_count++;
2057                                if (likely(!(t->ptrace & PT_SEIZED)))
2058                                        signal_wake_up(t, 0);
2059                                else
2060                                        ptrace_trap_notify(t);
2061                        }
2062                }
2063        }
2064
2065        if (likely(!current->ptrace)) {
2066                int notify = 0;
2067
2068                /*
2069                 * If there are no other threads in the group, or if there
2070                 * is a group stop in progress and we are the last to stop,
2071                 * report to the parent.
2072                 */
2073                if (task_participate_group_stop(current))
2074                        notify = CLD_STOPPED;
2075
2076                __set_current_state(TASK_STOPPED);
2077                spin_unlock_irq(&current->sighand->siglock);
2078
2079                /*
2080                 * Notify the parent of the group stop completion.  Because
2081                 * we're not holding either the siglock or tasklist_lock
2082                 * here, ptracer may attach inbetween; however, this is for
2083                 * group stop and should always be delivered to the real
2084                 * parent of the group leader.  The new ptracer will get
2085                 * its notification when this task transitions into
2086                 * TASK_TRACED.
2087                 */
2088                if (notify) {
2089                        read_lock(&tasklist_lock);
2090                        do_notify_parent_cldstop(current, false, notify);
2091                        read_unlock(&tasklist_lock);
2092                }
2093
2094                /* Now we don't run again until woken by SIGCONT or SIGKILL */
2095                schedule();
2096                return true;
2097        } else {
2098                /*
2099                 * While ptraced, group stop is handled by STOP trap.
2100                 * Schedule it and let the caller deal with it.
2101                 */
2102                task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2103                return false;
2104        }
2105}
2106
2107/**
2108 * do_jobctl_trap - take care of ptrace jobctl traps
2109 *
2110 * When PT_SEIZED, it's used for both group stop and explicit
2111 * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2112 * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2113 * the stop signal; otherwise, %SIGTRAP.
2114 *
2115 * When !PT_SEIZED, it's used only for group stop trap with stop signal
2116 * number as exit_code and no siginfo.
2117 *
2118 * CONTEXT:
2119 * Must be called with @current->sighand->siglock held, which may be
2120 * released and re-acquired before returning with intervening sleep.
2121 */
2122static void do_jobctl_trap(void)
2123{
2124        struct signal_struct *signal = current->signal;
2125        int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2126
2127        if (current->ptrace & PT_SEIZED) {
2128                if (!signal->group_stop_count &&
2129                    !(signal->flags & SIGNAL_STOP_STOPPED))
2130                        signr = SIGTRAP;
2131                WARN_ON_ONCE(!signr);
2132                ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2133                                 CLD_STOPPED);
2134        } else {
2135                WARN_ON_ONCE(!signr);
2136                ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2137                current->exit_code = 0;
2138        }
2139}
2140
2141static int ptrace_signal(int signr, siginfo_t *info,
2142                         struct pt_regs *regs, void *cookie)
2143{
2144        ptrace_signal_deliver(regs, cookie);
2145        /*
2146         * We do not check sig_kernel_stop(signr) but set this marker
2147         * unconditionally because we do not know whether debugger will
2148         * change signr. This flag has no meaning unless we are going
2149         * to stop after return from ptrace_stop(). In this case it will
2150         * be checked in do_signal_stop(), we should only stop if it was
2151         * not cleared by SIGCONT while we were sleeping. See also the
2152         * comment in dequeue_signal().
2153         */
2154        current->jobctl |= JOBCTL_STOP_DEQUEUED;
2155        ptrace_stop(signr, CLD_TRAPPED, 0, info);
2156
2157        /* We're back.  Did the debugger cancel the sig?  */
2158        signr = current->exit_code;
2159        if (signr == 0)
2160                return signr;
2161
2162        current->exit_code = 0;
2163
2164        /*
2165         * Update the siginfo structure if the signal has
2166         * changed.  If the debugger wanted something
2167         * specific in the siginfo structure then it should
2168         * have updated *info via PTRACE_SETSIGINFO.
2169         */
2170        if (signr != info->si_signo) {
2171                info->si_signo = signr;
2172                info->si_errno = 0;
2173                info->si_code = SI_USER;
2174                rcu_read_lock();
2175                info->si_pid = task_pid_vnr(current->parent);
2176                info->si_uid = from_kuid_munged(current_user_ns(),
2177                                                task_uid(current->parent));
2178                rcu_read_unlock();
2179        }
2180
2181        /* If the (new) signal is now blocked, requeue it.  */
2182        if (sigismember(&current->blocked, signr)) {
2183                specific_send_sig_info(signr, info, current);
2184                signr = 0;
2185        }
2186
2187        return signr;
2188}
2189
2190int get_signal_to_deliver(siginfo_t *info, struct k_sigaction *return_ka,
2191                          struct pt_regs *regs, void *cookie)
2192{
2193        struct sighand_struct *sighand = current->sighand;
2194        struct signal_struct *signal = current->signal;
2195        int signr;
2196
2197        if (unlikely(current->task_works))
2198                task_work_run();
2199
2200        if (unlikely(uprobe_deny_signal()))
2201                return 0;
2202
2203relock:
2204        /*
2205         * We'll jump back here after any time we were stopped in TASK_STOPPED.
2206         * While in TASK_STOPPED, we were considered "frozen enough".
2207         * Now that we woke up, it's crucial if we're supposed to be
2208         * frozen that we freeze now before running anything substantial.
2209         */
2210        try_to_freeze();
2211
2212        spin_lock_irq(&sighand->siglock);
2213        /*
2214         * Every stopped thread goes here after wakeup. Check to see if
2215         * we should notify the parent, prepare_signal(SIGCONT) encodes
2216         * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2217         */
2218        if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2219                int why;
2220
2221                if (signal->flags & SIGNAL_CLD_CONTINUED)
2222                        why = CLD_CONTINUED;
2223                else
2224                        why = CLD_STOPPED;
2225
2226                signal->flags &= ~SIGNAL_CLD_MASK;
2227
2228                spin_unlock_irq(&sighand->siglock);
2229
2230                /*
2231                 * Notify the parent that we're continuing.  This event is
2232                 * always per-process and doesn't make whole lot of sense
2233                 * for ptracers, who shouldn't consume the state via
2234                 * wait(2) either, but, for backward compatibility, notify
2235                 * the ptracer of the group leader too unless it's gonna be
2236                 * a duplicate.
2237                 */
2238                read_lock(&tasklist_lock);
2239                do_notify_parent_cldstop(current, false, why);
2240
2241                if (ptrace_reparented(current->group_leader))
2242                        do_notify_parent_cldstop(current->group_leader,
2243                                                true, why);
2244                read_unlock(&tasklist_lock);
2245
2246                goto relock;
2247        }
2248
2249        for (;;) {
2250                struct k_sigaction *ka;
2251
2252                if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2253                    do_signal_stop(0))
2254                        goto relock;
2255
2256                if (unlikely(current->jobctl & JOBCTL_TRAP_MASK)) {
2257                        do_jobctl_trap();
2258                        spin_unlock_irq(&sighand->siglock);
2259                        goto relock;
2260                }
2261
2262                signr = dequeue_signal(current, &current->blocked, info);
2263
2264                if (!signr)
2265                        break; /* will return 0 */
2266
2267                if (unlikely(current->ptrace) && signr != SIGKILL) {
2268                        signr = ptrace_signal(signr, info,
2269                                              regs, cookie);
2270                        if (!signr)
2271                                continue;
2272                }
2273
2274                ka = &sighand->action[signr-1];
2275
2276                /* Trace actually delivered signals. */
2277                trace_signal_deliver(signr, info, ka);
2278
2279                if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2280                        continue;
2281                if (ka->sa.sa_handler != SIG_DFL) {
2282                        /* Run the handler.  */
2283                        *return_ka = *ka;
2284
2285                        if (ka->sa.sa_flags & SA_ONESHOT)
2286                                ka->sa.sa_handler = SIG_DFL;
2287
2288                        break; /* will return non-zero "signr" value */
2289                }
2290
2291                /*
2292                 * Now we are doing the default action for this signal.
2293                 */
2294                if (sig_kernel_ignore(signr)) /* Default is nothing. */
2295                        continue;
2296
2297                /*
2298                 * Global init gets no signals it doesn't want.
2299                 * Container-init gets no signals it doesn't want from same
2300                 * container.
2301                 *
2302                 * Note that if global/container-init sees a sig_kernel_only()
2303                 * signal here, the signal must have been generated internally
2304                 * or must have come from an ancestor namespace. In either
2305                 * case, the signal cannot be dropped.
2306                 */
2307                if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2308                                !sig_kernel_only(signr))
2309                        continue;
2310
2311                if (sig_kernel_stop(signr)) {
2312                        /*
2313                         * The default action is to stop all threads in
2314                         * the thread group.  The job control signals
2315                         * do nothing in an orphaned pgrp, but SIGSTOP
2316                         * always works.  Note that siglock needs to be
2317                         * dropped during the call to is_orphaned_pgrp()
2318                         * because of lock ordering with tasklist_lock.
2319                         * This allows an intervening SIGCONT to be posted.
2320                         * We need to check for that and bail out if necessary.
2321                         */
2322                        if (signr != SIGSTOP) {
2323                                spin_unlock_irq(&sighand->siglock);
2324
2325                                /* signals can be posted during this window */
2326
2327                                if (is_current_pgrp_orphaned())
2328                                        goto relock;
2329
2330                                spin_lock_irq(&sighand->siglock);
2331                        }
2332
2333                        if (likely(do_signal_stop(info->si_signo))) {
2334                                /* It released the siglock.  */
2335                                goto relock;
2336                        }
2337
2338                        /*
2339                         * We didn't actually stop, due to a race
2340                         * with SIGCONT or something like that.
2341                         */
2342                        continue;
2343                }
2344
2345                spin_unlock_irq(&sighand->siglock);
2346
2347                /*
2348                 * Anything else is fatal, maybe with a core dump.
2349                 */
2350                current->flags |= PF_SIGNALED;
2351
2352                if (sig_kernel_coredump(signr)) {
2353                        if (print_fatal_signals)
2354                                print_fatal_signal(regs, info->si_signo);
2355                        /*
2356                         * If it was able to dump core, this kills all
2357                         * other threads in the group and synchronizes with
2358                         * their demise.  If we lost the race with another
2359                         * thread getting here, it set group_exit_code
2360                         * first and our do_group_exit call below will use
2361                         * that value and ignore the one we pass it.
2362                         */
2363                        do_coredump(info, regs);
2364                }
2365
2366                /*
2367                 * Death signals, no core dump.
2368                 */
2369                do_group_exit(info->si_signo);
2370                /* NOTREACHED */
2371        }
2372        spin_unlock_irq(&sighand->siglock);
2373        return signr;
2374}
2375
2376/**
2377 * signal_delivered - 
2378 * @sig:                number of signal being delivered
2379 * @info:               siginfo_t of signal being delivered
2380 * @ka:                 sigaction setting that chose the handler
2381 * @regs:               user register state
2382 * @stepping:           nonzero if debugger single-step or block-step in use
2383 *
2384 * This function should be called when a signal has succesfully been
2385 * delivered. It updates the blocked signals accordingly (@ka->sa.sa_mask
2386 * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2387 * is set in @ka->sa.sa_flags.  Tracing is notified.
2388 */
2389void signal_delivered(int sig, siginfo_t *info, struct k_sigaction *ka,
2390                        struct pt_regs *regs, int stepping)
2391{
2392        sigset_t blocked;
2393
2394        /* A signal was successfully delivered, and the
2395           saved sigmask was stored on the signal frame,
2396           and will be restored by sigreturn.  So we can
2397           simply clear the restore sigmask flag.  */
2398        clear_restore_sigmask();
2399
2400        sigorsets(&blocked, &current->blocked, &ka->sa.sa_mask);
2401        if (!(ka->sa.sa_flags & SA_NODEFER))
2402                sigaddset(&blocked, sig);
2403        set_current_blocked(&blocked);
2404        tracehook_signal_handler(sig, info, ka, regs, stepping);
2405}
2406
2407/*
2408 * It could be that complete_signal() picked us to notify about the
2409 * group-wide signal. Other threads should be notified now to take
2410 * the shared signals in @which since we will not.
2411 */
2412static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2413{
2414        sigset_t retarget;
2415        struct task_struct *t;
2416
2417        sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2418        if (sigisemptyset(&retarget))
2419                return;
2420
2421        t = tsk;
2422        while_each_thread(tsk, t) {
2423                if (t->flags & PF_EXITING)
2424                        continue;
2425
2426                if (!has_pending_signals(&retarget, &t->blocked))
2427                        continue;
2428                /* Remove the signals this thread can handle. */
2429                sigandsets(&retarget, &retarget, &t->blocked);
2430
2431                if (!signal_pending(t))
2432                        signal_wake_up(t, 0);
2433
2434                if (sigisemptyset(&retarget))
2435                        break;
2436        }
2437}
2438
2439void exit_signals(struct task_struct *tsk)
2440{
2441        int group_stop = 0;
2442        sigset_t unblocked;
2443
2444        /*
2445         * @tsk is about to have PF_EXITING set - lock out users which
2446         * expect stable threadgroup.
2447         */
2448        threadgroup_change_begin(tsk);
2449
2450        if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2451                tsk->flags |= PF_EXITING;
2452                threadgroup_change_end(tsk);
2453                return;
2454        }
2455
2456        spin_lock_irq(&tsk->sighand->siglock);
2457        /*
2458         * From now this task is not visible for group-wide signals,
2459         * see wants_signal(), do_signal_stop().
2460         */
2461        tsk->flags |= PF_EXITING;
2462
2463        threadgroup_change_end(tsk);
2464
2465        if (!signal_pending(tsk))
2466                goto out;
2467
2468        unblocked = tsk->blocked;
2469        signotset(&unblocked);
2470        retarget_shared_pending(tsk, &unblocked);
2471
2472        if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2473            task_participate_group_stop(tsk))
2474                group_stop = CLD_STOPPED;
2475out:
2476        spin_unlock_irq(&tsk->sighand->siglock);
2477
2478        /*
2479         * If group stop has completed, deliver the notification.  This
2480         * should always go to the real parent of the group leader.
2481         */
2482        if (unlikely(group_stop)) {
2483                read_lock(&tasklist_lock);
2484                do_notify_parent_cldstop(tsk, false, group_stop);
2485                read_unlock(&tasklist_lock);
2486        }
2487}
2488
2489EXPORT_SYMBOL(recalc_sigpending);
2490EXPORT_SYMBOL_GPL(dequeue_signal);
2491EXPORT_SYMBOL(flush_signals);
2492EXPORT_SYMBOL(force_sig);
2493EXPORT_SYMBOL(send_sig);
2494EXPORT_SYMBOL(send_sig_info);
2495EXPORT_SYMBOL(sigprocmask);
2496EXPORT_SYMBOL(block_all_signals);
2497EXPORT_SYMBOL(unblock_all_signals);
2498
2499
2500/*
2501 * System call entry points.
2502 */
2503
2504/**
2505 *  sys_restart_syscall - restart a system call
2506 */
2507SYSCALL_DEFINE0(restart_syscall)
2508{
2509        struct restart_block *restart = &current_thread_info()->restart_block;
2510        return restart->fn(restart);
2511}
2512
2513long do_no_restart_syscall(struct restart_block *param)
2514{
2515        return -EINTR;
2516}
2517
2518static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2519{
2520        if (signal_pending(tsk) && !thread_group_empty(tsk)) {
2521                sigset_t newblocked;
2522                /* A set of now blocked but previously unblocked signals. */
2523                sigandnsets(&newblocked, newset, &current->blocked);
2524                retarget_shared_pending(tsk, &newblocked);
2525        }
2526        tsk->blocked = *newset;
2527        recalc_sigpending();
2528}
2529
2530/**
2531 * set_current_blocked - change current->blocked mask
2532 * @newset: new mask
2533 *
2534 * It is wrong to change ->blocked directly, this helper should be used
2535 * to ensure the process can't miss a shared signal we are going to block.
2536 */
2537void set_current_blocked(sigset_t *newset)
2538{
2539        struct task_struct *tsk = current;
2540        sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2541        spin_lock_irq(&tsk->sighand->siglock);
2542        __set_task_blocked(tsk, newset);
2543        spin_unlock_irq(&tsk->sighand->siglock);
2544}
2545
2546void __set_current_blocked(const sigset_t *newset)
2547{
2548        struct task_struct *tsk = current;
2549
2550        spin_lock_irq(&tsk->sighand->siglock);
2551        __set_task_blocked(tsk, newset);
2552        spin_unlock_irq(&tsk->sighand->siglock);
2553}
2554
2555/*
2556 * This is also useful for kernel threads that want to temporarily
2557 * (or permanently) block certain signals.
2558 *
2559 * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2560 * interface happily blocks "unblockable" signals like SIGKILL
2561 * and friends.
2562 */
2563int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2564{
2565        struct task_struct *tsk = current;
2566        sigset_t newset;
2567
2568        /* Lockless, only current can change ->blocked, never from irq */
2569        if (oldset)
2570                *oldset = tsk->blocked;
2571
2572        switch (how) {
2573        case SIG_BLOCK:
2574                sigorsets(&newset, &tsk->blocked, set);
2575                break;
2576        case SIG_UNBLOCK:
2577                sigandnsets(&newset, &tsk->blocked, set);
2578                break;
2579        case SIG_SETMASK:
2580                newset = *set;
2581                break;
2582        default:
2583                return -EINVAL;
2584        }
2585
2586        __set_current_blocked(&newset);
2587        return 0;
2588}
2589
2590/**
2591 *  sys_rt_sigprocmask - change the list of currently blocked signals
2592 *  @how: whether to add, remove, or set signals
2593 *  @nset: stores pending signals
2594 *  @oset: previous value of signal mask if non-null
2595 *  @sigsetsize: size of sigset_t type
2596 */
2597SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
2598                sigset_t __user *, oset, size_t, sigsetsize)
2599{
2600        sigset_t old_set, new_set;
2601        int error;
2602
2603        /* XXX: Don't preclude handling different sized sigset_t's.  */
2604        if (sigsetsize != sizeof(sigset_t))
2605                return -EINVAL;
2606
2607        old_set = current->blocked;
2608
2609        if (nset) {
2610                if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
2611                        return -EFAULT;
2612                sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2613
2614                error = sigprocmask(how, &new_set, NULL);
2615                if (error)
2616                        return error;
2617        }
2618
2619        if (oset) {
2620                if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
2621                        return -EFAULT;
2622        }
2623
2624        return 0;
2625}
2626
2627long do_sigpending(void __user *set, unsigned long sigsetsize)
2628{
2629        long error = -EINVAL;
2630        sigset_t pending;
2631
2632        if (sigsetsize > sizeof(sigset_t))
2633                goto out;
2634
2635        spin_lock_irq(&current->sighand->siglock);
2636        sigorsets(&pending, &current->pending.signal,
2637                  &current->signal->shared_pending.signal);
2638        spin_unlock_irq(&current->sighand->siglock);
2639
2640        /* Outside the lock because only this thread touches it.  */
2641        sigandsets(&pending, &current->blocked, &pending);
2642
2643        error = -EFAULT;
2644        if (!copy_to_user(set, &pending, sigsetsize))
2645                error = 0;
2646
2647out:
2648        return error;
2649}
2650
2651/**
2652 *  sys_rt_sigpending - examine a pending signal that has been raised
2653 *                      while blocked
2654 *  @set: stores pending signals
2655 *  @sigsetsize: size of sigset_t type or larger
2656 */
2657SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, set, size_t, sigsetsize)
2658{
2659        return do_sigpending(set, sigsetsize);
2660}
2661
2662#ifndef HAVE_ARCH_COPY_SIGINFO_TO_USER
2663
2664int copy_siginfo_to_user(siginfo_t __user *to, siginfo_t *from)
2665{
2666        int err;
2667
2668        if (!access_ok (VERIFY_WRITE, to, sizeof(siginfo_t)))
2669                return -EFAULT;
2670        if (from->si_code < 0)
2671                return __copy_to_user(to, from, sizeof(siginfo_t))
2672                        ? -EFAULT : 0;
2673        /*
2674         * If you change siginfo_t structure, please be sure
2675         * this code is fixed accordingly.
2676         * Please remember to update the signalfd_copyinfo() function
2677         * inside fs/signalfd.c too, in case siginfo_t changes.
2678         * It should never copy any pad contained in the structure
2679         * to avoid security leaks, but must copy the generic
2680         * 3 ints plus the relevant union member.
2681         */
2682        err = __put_user(from->si_signo, &to->si_signo);
2683        err |= __put_user(from->si_errno, &to->si_errno);
2684        err |= __put_user((short)from->si_code, &to->si_code);
2685        switch (from->si_code & __SI_MASK) {
2686        case __SI_KILL:
2687                err |= __put_user(from->si_pid, &to->si_pid);
2688                err |= __put_user(from->si_uid, &to->si_uid);
2689                break;
2690        case __SI_TIMER:
2691                 err |= __put_user(from->si_tid, &to->si_tid);
2692                 err |= __put_user(from->si_overrun, &to->si_overrun);
2693                 err |= __put_user(from->si_ptr, &to->si_ptr);
2694                break;
2695        case __SI_POLL:
2696                err |= __put_user(from->si_band, &to->si_band);
2697                err |= __put_user(from->si_fd, &to->si_fd);
2698                break;
2699        case __SI_FAULT:
2700                err |= __put_user(from->si_addr, &to->si_addr);
2701#ifdef __ARCH_SI_TRAPNO
2702                err |= __put_user(from->si_trapno, &to->si_trapno);
2703#endif
2704#ifdef BUS_MCEERR_AO
2705                /*
2706                 * Other callers might not initialize the si_lsb field,
2707                 * so check explicitly for the right codes here.
2708                 */
2709                if (from->si_code == BUS_MCEERR_AR || from->si_code == BUS_MCEERR_AO)
2710                        err |= __put_user(from->si_addr_lsb, &to->si_addr_lsb);
2711#endif
2712                break;
2713        case __SI_CHLD:
2714                err |= __put_user(from->si_pid, &to->si_pid);
2715                err |= __put_user(from->si_uid, &to->si_uid);
2716                err |= __put_user(from->si_status, &to->si_status);
2717                err |= __put_user(from->si_utime, &to->si_utime);
2718                err |= __put_user(from->si_stime, &to->si_stime);
2719                break;
2720        case __SI_RT: /* This is not generated by the kernel as of now. */
2721        case __SI_MESGQ: /* But this is */
2722                err |= __put_user(from->si_pid, &to->si_pid);
2723                err |= __put_user(from->si_uid, &to->si_uid);
2724                err |= __put_user(from->si_ptr, &to->si_ptr);
2725                break;
2726#ifdef __ARCH_SIGSYS
2727        case __SI_SYS:
2728                err |= __put_user(from->si_call_addr, &to->si_call_addr);
2729                err |= __put_user(from->si_syscall, &to->si_syscall);
2730                err |= __put_user(from->si_arch, &to->si_arch);
2731                break;
2732#endif
2733        default: /* this is just in case for now ... */
2734                err |= __put_user(from->si_pid, &to->si_pid);
2735                err |= __put_user(from->si_uid, &to->si_uid);
2736                break;
2737        }
2738        return err;
2739}
2740
2741#endif
2742
2743/**
2744 *  do_sigtimedwait - wait for queued signals specified in @which
2745 *  @which: queued signals to wait for
2746 *  @info: if non-null, the signal's siginfo is returned here
2747 *  @ts: upper bound on process time suspension
2748 */
2749int do_sigtimedwait(const sigset_t *which, siginfo_t *info,
2750                        const struct timespec *ts)
2751{
2752        struct task_struct *tsk = current;
2753        long timeout = MAX_SCHEDULE_TIMEOUT;
2754        sigset_t mask = *which;
2755        int sig;
2756
2757        if (ts) {
2758                if (!timespec_valid(ts))
2759                        return -EINVAL;
2760                timeout = timespec_to_jiffies(ts);
2761                /*
2762                 * We can be close to the next tick, add another one
2763                 * to ensure we will wait at least the time asked for.
2764                 */
2765                if (ts->tv_sec || ts->tv_nsec)
2766                        timeout++;
2767        }
2768
2769        /*
2770         * Invert the set of allowed signals to get those we want to block.
2771         */
2772        sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
2773        signotset(&mask);
2774
2775        spin_lock_irq(&tsk->sighand->siglock);
2776        sig = dequeue_signal(tsk, &mask, info);
2777        if (!sig && timeout) {
2778                /*
2779                 * None ready, temporarily unblock those we're interested
2780                 * while we are sleeping in so that we'll be awakened when
2781                 * they arrive. Unblocking is always fine, we can avoid
2782                 * set_current_blocked().
2783                 */
2784                tsk->real_blocked = tsk->blocked;
2785                sigandsets(&tsk->blocked, &tsk->blocked, &mask);
2786                recalc_sigpending();
2787                spin_unlock_irq(&tsk->sighand->siglock);
2788
2789                timeout = schedule_timeout_interruptible(timeout);
2790
2791                spin_lock_irq(&tsk->sighand->siglock);
2792                __set_task_blocked(tsk, &tsk->real_blocked);
2793                siginitset(&tsk->real_blocked, 0);
2794                sig = dequeue_signal(tsk, &mask, info);
2795        }
2796        spin_unlock_irq(&tsk->sighand->siglock);
2797
2798        if (sig)
2799                return sig;
2800        return timeout ? -EINTR : -EAGAIN;
2801}
2802
2803/**
2804 *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
2805 *                      in @uthese
2806 *  @uthese: queued signals to wait for
2807 *  @uinfo: if non-null, the signal's siginfo is returned here
2808 *  @uts: upper bound on process time suspension
2809 *  @sigsetsize: size of sigset_t type
2810 */
2811SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
2812                siginfo_t __user *, uinfo, const struct timespec __user *, uts,
2813                size_t, sigsetsize)
2814{
2815        sigset_t these;
2816        struct timespec ts;
2817        siginfo_t info;
2818        int ret;
2819
2820        /* XXX: Don't preclude handling different sized sigset_t's.  */
2821        if (sigsetsize != sizeof(sigset_t))
2822                return -EINVAL;
2823
2824        if (copy_from_user(&these, uthese, sizeof(these)))
2825                return -EFAULT;
2826
2827        if (uts) {
2828                if (copy_from_user(&ts, uts, sizeof(ts)))
2829                        return -EFAULT;
2830        }
2831
2832        ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
2833
2834        if (ret > 0 && uinfo) {
2835                if (copy_siginfo_to_user(uinfo, &info))
2836                        ret = -EFAULT;
2837        }
2838
2839        return ret;
2840}
2841
2842/**
2843 *  sys_kill - send a signal to a process
2844 *  @pid: the PID of the process
2845 *  @sig: signal to be sent
2846 */
2847SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
2848{
2849        struct siginfo info;
2850
2851        info.si_signo = sig;
2852        info.si_errno = 0;
2853        info.si_code = SI_USER;
2854        info.si_pid = task_tgid_vnr(current);
2855        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2856
2857        return kill_something_info(sig, &info, pid);
2858}
2859
2860static int
2861do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info)
2862{
2863        struct task_struct *p;
2864        int error = -ESRCH;
2865
2866        rcu_read_lock();
2867        p = find_task_by_vpid(pid);
2868        if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
2869                error = check_kill_permission(sig, info, p);
2870                /*
2871                 * The null signal is a permissions and process existence
2872                 * probe.  No signal is actually delivered.
2873                 */
2874                if (!error && sig) {
2875                        error = do_send_sig_info(sig, info, p, false);
2876                        /*
2877                         * If lock_task_sighand() failed we pretend the task
2878                         * dies after receiving the signal. The window is tiny,
2879                         * and the signal is private anyway.
2880                         */
2881                        if (unlikely(error == -ESRCH))
2882                                error = 0;
2883                }
2884        }
2885        rcu_read_unlock();
2886
2887        return error;
2888}
2889
2890static int do_tkill(pid_t tgid, pid_t pid, int sig)
2891{
2892        struct siginfo info;
2893
2894        info.si_signo = sig;
2895        info.si_errno = 0;
2896        info.si_code = SI_TKILL;
2897        info.si_pid = task_tgid_vnr(current);
2898        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2899
2900        return do_send_specific(tgid, pid, sig, &info);
2901}
2902
2903/**
2904 *  sys_tgkill - send signal to one specific thread
2905 *  @tgid: the thread group ID of the thread
2906 *  @pid: the PID of the thread
2907 *  @sig: signal to be sent
2908 *
2909 *  This syscall also checks the @tgid and returns -ESRCH even if the PID
2910 *  exists but it's not belonging to the target process anymore. This
2911 *  method solves the problem of threads exiting and PIDs getting reused.
2912 */
2913SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
2914{
2915        /* This is only valid for single tasks */
2916        if (pid <= 0 || tgid <= 0)
2917                return -EINVAL;
2918
2919        return do_tkill(tgid, pid, sig);
2920}
2921
2922/**
2923 *  sys_tkill - send signal to one specific task
2924 *  @pid: the PID of the task
2925 *  @sig: signal to be sent
2926 *
2927 *  Send a signal to only one task, even if it's a CLONE_THREAD task.
2928 */
2929SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
2930{
2931        /* This is only valid for single tasks */
2932        if (pid <= 0)
2933                return -EINVAL;
2934
2935        return do_tkill(0, pid, sig);
2936}
2937
2938/**
2939 *  sys_rt_sigqueueinfo - send signal information to a signal
2940 *  @pid: the PID of the thread
2941 *  @sig: signal to be sent
2942 *  @uinfo: signal info to be sent
2943 */
2944SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
2945                siginfo_t __user *, uinfo)
2946{
2947        siginfo_t info;
2948
2949        if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
2950                return -EFAULT;
2951
2952        /* Not even root can pretend to send signals from the kernel.
2953         * Nor can they impersonate a kill()/tgkill(), which adds source info.
2954         */
2955        if (info.si_code >= 0 || info.si_code == SI_TKILL) {
2956                /* We used to allow any < 0 si_code */
2957                WARN_ON_ONCE(info.si_code < 0);
2958                return -EPERM;
2959        }
2960        info.si_signo = sig;
2961
2962        /* POSIX.1b doesn't mention process groups.  */
2963        return kill_proc_info(sig, &info, pid);
2964}
2965
2966long do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, siginfo_t *info)
2967{
2968        /* This is only valid for single tasks */
2969        if (pid <= 0 || tgid <= 0)
2970                return -EINVAL;
2971
2972        /* Not even root can pretend to send signals from the kernel.
2973         * Nor can they impersonate a kill()/tgkill(), which adds source info.
2974         */
2975        if (info->si_code >= 0 || info->si_code == SI_TKILL) {
2976                /* We used to allow any < 0 si_code */
2977                WARN_ON_ONCE(info->si_code < 0);
2978                return -EPERM;
2979        }
2980        info->si_signo = sig;
2981
2982        return do_send_specific(tgid, pid, sig, info);
2983}
2984
2985SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
2986                siginfo_t __user *, uinfo)
2987{
2988        siginfo_t info;
2989
2990        if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
2991                return -EFAULT;
2992
2993        return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
2994}
2995
2996int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
2997{
2998        struct task_struct *t = current;
2999        struct k_sigaction *k;
3000        sigset_t mask;
3001
3002        if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
3003                return -EINVAL;
3004
3005        k = &t->sighand->action[sig-1];
3006
3007        spin_lock_irq(&current->sighand->siglock);
3008        if (oact)
3009                *oact = *k;
3010
3011        if (act) {
3012                sigdelsetmask(&act->sa.sa_mask,
3013                              sigmask(SIGKILL) | sigmask(SIGSTOP));
3014                *k = *act;
3015                /*
3016                 * POSIX 3.3.1.3:
3017                 *  "Setting a signal action to SIG_IGN for a signal that is
3018                 *   pending shall cause the pending signal to be discarded,
3019                 *   whether or not it is blocked."
3020                 *
3021                 *  "Setting a signal action to SIG_DFL for a signal that is
3022                 *   pending and whose default action is to ignore the signal
3023                 *   (for example, SIGCHLD), shall cause the pending signal to
3024                 *   be discarded, whether or not it is blocked"
3025                 */
3026                if (sig_handler_ignored(sig_handler(t, sig), sig)) {
3027                        sigemptyset(&mask);
3028                        sigaddset(&mask, sig);
3029                        rm_from_queue_full(&mask, &t->signal->shared_pending);
3030                        do {
3031                                rm_from_queue_full(&mask, &t->pending);
3032                                t = next_thread(t);
3033                        } while (t != current);
3034                }
3035        }
3036
3037        spin_unlock_irq(&current->sighand->siglock);
3038        return 0;
3039}
3040
3041int 
3042do_sigaltstack (const stack_t __user *uss, stack_t __user *uoss, unsigned long sp)
3043{
3044        stack_t oss;
3045        int error;
3046
3047        oss.ss_sp = (void __user *) current->sas_ss_sp;
3048        oss.ss_size = current->sas_ss_size;
3049        oss.ss_flags = sas_ss_flags(sp);
3050
3051        if (uss) {
3052                void __user *ss_sp;
3053                size_t ss_size;
3054                int ss_flags;
3055
3056                error = -EFAULT;
3057                if (!access_ok(VERIFY_READ, uss, sizeof(*uss)))
3058                        goto out;
3059                error = __get_user(ss_sp, &uss->ss_sp) |
3060                        __get_user(ss_flags, &uss->ss_flags) |
3061                        __get_user(ss_size, &uss->ss_size);
3062                if (error)
3063                        goto out;
3064
3065                error = -EPERM;
3066                if (on_sig_stack(sp))
3067                        goto out;
3068
3069                error = -EINVAL;
3070                /*
3071                 * Note - this code used to test ss_flags incorrectly:
3072                 *        old code may have been written using ss_flags==0
3073                 *        to mean ss_flags==SS_ONSTACK (as this was the only
3074                 *        way that worked) - this fix preserves that older
3075                 *        mechanism.
3076                 */
3077                if (ss_flags != SS_DISABLE && ss_flags != SS_ONSTACK && ss_flags != 0)
3078                        goto out;
3079
3080                if (ss_flags == SS_DISABLE) {
3081                        ss_size = 0;
3082                        ss_sp = NULL;
3083                } else {
3084                        error = -ENOMEM;
3085                        if (ss_size < MINSIGSTKSZ)
3086                                goto out;
3087                }
3088
3089                current->sas_ss_sp = (unsigned long) ss_sp;
3090                current->sas_ss_size = ss_size;
3091        }
3092
3093        error = 0;
3094        if (uoss) {
3095                error = -EFAULT;
3096                if (!access_ok(VERIFY_WRITE, uoss, sizeof(*uoss)))
3097                        goto out;
3098                error = __put_user(oss.ss_sp, &uoss->ss_sp) |
3099                        __put_user(oss.ss_size, &uoss->ss_size) |
3100                        __put_user(oss.ss_flags, &uoss->ss_flags);
3101        }
3102
3103out:
3104        return error;
3105}
3106
3107#ifdef __ARCH_WANT_SYS_SIGPENDING
3108
3109/**
3110 *  sys_sigpending - examine pending signals
3111 *  @set: where mask of pending signal is returned
3112 */
3113SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, set)
3114{
3115        return do_sigpending(set, sizeof(*set));
3116}
3117
3118#endif
3119
3120#ifdef __ARCH_WANT_SYS_SIGPROCMASK
3121/**
3122 *  sys_sigprocmask - examine and change blocked signals
3123 *  @how: whether to add, remove, or set signals
3124 *  @nset: signals to add or remove (if non-null)
3125 *  @oset: previous value of signal mask if non-null
3126 *
3127 * Some platforms have their own version with special arguments;
3128 * others support only sys_rt_sigprocmask.
3129 */
3130
3131SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
3132                old_sigset_t __user *, oset)
3133{
3134        old_sigset_t old_set, new_set;
3135        sigset_t new_blocked;
3136
3137        old_set = current->blocked.sig[0];
3138
3139        if (nset) {
3140                if (copy_from_user(&new_set, nset, sizeof(*nset)))
3141                        return -EFAULT;
3142                new_set &= ~(sigmask(SIGKILL) | sigmask(SIGSTOP));
3143
3144                new_blocked = current->blocked;
3145
3146                switch (how) {
3147                case SIG_BLOCK:
3148                        sigaddsetmask(&new_blocked, new_set);
3149                        break;
3150                case SIG_UNBLOCK:
3151                        sigdelsetmask(&new_blocked, new_set);
3152                        break;
3153                case SIG_SETMASK:
3154                        new_blocked.sig[0] = new_set;
3155                        break;
3156                default:
3157                        return -EINVAL;
3158                }
3159
3160                __set_current_blocked(&new_blocked);
3161        }
3162
3163        if (oset) {
3164                if (copy_to_user(oset, &old_set, sizeof(*oset)))
3165                        return -EFAULT;
3166        }
3167
3168        return 0;
3169}
3170#endif /* __ARCH_WANT_SYS_SIGPROCMASK */
3171
3172#ifdef __ARCH_WANT_SYS_RT_SIGACTION
3173/**
3174 *  sys_rt_sigaction - alter an action taken by a process
3175 *  @sig: signal to be sent
3176 *  @act: new sigaction
3177 *  @oact: used to save the previous sigaction
3178 *  @sigsetsize: size of sigset_t type
3179 */
3180SYSCALL_DEFINE4(rt_sigaction, int, sig,
3181                const struct sigaction __user *, act,
3182                struct sigaction __user *, oact,
3183                size_t, sigsetsize)
3184{
3185        struct k_sigaction new_sa, old_sa;
3186        int ret = -EINVAL;
3187
3188        /* XXX: Don't preclude handling different sized sigset_t's.  */
3189        if (sigsetsize != sizeof(sigset_t))
3190                goto out;
3191
3192        if (act) {
3193                if (copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
3194                        return -EFAULT;
3195        }
3196
3197        ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
3198
3199        if (!ret && oact) {
3200                if (copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
3201                        return -EFAULT;
3202        }
3203out:
3204        return ret;
3205}
3206#endif /* __ARCH_WANT_SYS_RT_SIGACTION */
3207
3208#ifdef __ARCH_WANT_SYS_SGETMASK
3209
3210/*
3211 * For backwards compatibility.  Functionality superseded by sigprocmask.
3212 */
3213SYSCALL_DEFINE0(sgetmask)
3214{
3215        /* SMP safe */
3216        return current->blocked.sig[0];
3217}
3218
3219SYSCALL_DEFINE1(ssetmask, int, newmask)
3220{
3221        int old = current->blocked.sig[0];
3222        sigset_t newset;
3223
3224        set_current_blocked(&newset);
3225
3226        return old;
3227}
3228#endif /* __ARCH_WANT_SGETMASK */
3229
3230#ifdef __ARCH_WANT_SYS_SIGNAL
3231/*
3232 * For backwards compatibility.  Functionality superseded by sigaction.
3233 */
3234SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
3235{
3236        struct k_sigaction new_sa, old_sa;
3237        int ret;
3238
3239        new_sa.sa.sa_handler = handler;
3240        new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
3241        sigemptyset(&new_sa.sa.sa_mask);
3242
3243        ret = do_sigaction(sig, &new_sa, &old_sa);
3244
3245        return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
3246}
3247#endif /* __ARCH_WANT_SYS_SIGNAL */
3248
3249#ifdef __ARCH_WANT_SYS_PAUSE
3250
3251SYSCALL_DEFINE0(pause)
3252{
3253        while (!signal_pending(current)) {
3254                current->state = TASK_INTERRUPTIBLE;
3255                schedule();
3256        }
3257        return -ERESTARTNOHAND;
3258}
3259
3260#endif
3261
3262int sigsuspend(sigset_t *set)
3263{
3264        current->saved_sigmask = current->blocked;
3265        set_current_blocked(set);
3266
3267        current->state = TASK_INTERRUPTIBLE;
3268        schedule();
3269        set_restore_sigmask();
3270        return -ERESTARTNOHAND;
3271}
3272
3273#ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
3274/**
3275 *  sys_rt_sigsuspend - replace the signal mask for a value with the
3276 *      @unewset value until a signal is received
3277 *  @unewset: new signal mask value
3278 *  @sigsetsize: size of sigset_t type
3279 */
3280SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
3281{
3282        sigset_t newset;
3283
3284        /* XXX: Don't preclude handling different sized sigset_t's.  */
3285        if (sigsetsize != sizeof(sigset_t))
3286                return -EINVAL;
3287
3288        if (copy_from_user(&newset, unewset, sizeof(newset)))
3289                return -EFAULT;
3290        return sigsuspend(&newset);
3291}
3292#endif /* __ARCH_WANT_SYS_RT_SIGSUSPEND */
3293
3294__attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma)
3295{
3296        return NULL;
3297}
3298
3299void __init signals_init(void)
3300{
3301        sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
3302}
3303
3304#ifdef CONFIG_KGDB_KDB
3305#include <linux/kdb.h>
3306/*
3307 * kdb_send_sig_info - Allows kdb to send signals without exposing
3308 * signal internals.  This function checks if the required locks are
3309 * available before calling the main signal code, to avoid kdb
3310 * deadlocks.
3311 */
3312void
3313kdb_send_sig_info(struct task_struct *t, struct siginfo *info)
3314{
3315        static struct task_struct *kdb_prev_t;
3316        int sig, new_t;
3317        if (!spin_trylock(&t->sighand->siglock)) {
3318                kdb_printf("Can't do kill command now.\n"
3319                           "The sigmask lock is held somewhere else in "
3320                           "kernel, try again later\n");
3321                return;
3322        }
3323        spin_unlock(&t->sighand->siglock);
3324        new_t = kdb_prev_t != t;
3325        kdb_prev_t = t;
3326        if (t->state != TASK_RUNNING && new_t) {
3327                kdb_printf("Process is not RUNNING, sending a signal from "
3328                           "kdb risks deadlock\n"
3329                           "on the run queue locks. "
3330                           "The signal has _not_ been sent.\n"
3331                           "Reissue the kill command if you want to risk "
3332                           "the deadlock.\n");
3333                return;
3334        }
3335        sig = info->si_signo;
3336        if (send_sig_info(sig, info, t))
3337                kdb_printf("Fail to deliver Signal %d to process %d.\n",
3338                           sig, t->pid);
3339        else
3340                kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
3341}
3342#endif  /* CONFIG_KGDB_KDB */
3343