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#include <linux/compat.h>
  35#define CREATE_TRACE_POINTS
  36#include <trace/events/signal.h>
  37
  38#include <asm/param.h>
  39#include <asm/uaccess.h>
  40#include <asm/unistd.h>
  41#include <asm/siginfo.h>
  42#include <asm/cacheflush.h>
  43#include "audit.h"      /* audit_signal_info() */
  44
  45/*
  46 * SLAB caches for signal bits.
  47 */
  48
  49static struct kmem_cache *sigqueue_cachep;
  50
  51int print_fatal_signals __read_mostly;
  52
  53static void __user *sig_handler(struct task_struct *t, int sig)
  54{
  55        return t->sighand->action[sig - 1].sa.sa_handler;
  56}
  57
  58static int sig_handler_ignored(void __user *handler, int sig)
  59{
  60        /* Is it explicitly or implicitly ignored? */
  61        return handler == SIG_IGN ||
  62                (handler == SIG_DFL && sig_kernel_ignore(sig));
  63}
  64
  65static int sig_task_ignored(struct task_struct *t, int sig, bool force)
  66{
  67        void __user *handler;
  68
  69        handler = sig_handler(t, sig);
  70
  71        if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
  72                        handler == SIG_DFL && !force)
  73                return 1;
  74
  75        return sig_handler_ignored(handler, sig);
  76}
  77
  78static int sig_ignored(struct task_struct *t, int sig, bool force)
  79{
  80        /*
  81         * Blocked signals are never ignored, since the
  82         * signal handler may change by the time it is
  83         * unblocked.
  84         */
  85        if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
  86                return 0;
  87
  88        if (!sig_task_ignored(t, sig, force))
  89                return 0;
  90
  91        /*
  92         * Tracers may want to know about even ignored signals.
  93         */
  94        return !t->ptrace;
  95}
  96
  97/*
  98 * Re-calculate pending state from the set of locally pending
  99 * signals, globally pending signals, and blocked signals.
 100 */
 101static inline int has_pending_signals(sigset_t *signal, sigset_t *blocked)
 102{
 103        unsigned long ready;
 104        long i;
 105
 106        switch (_NSIG_WORDS) {
 107        default:
 108                for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
 109                        ready |= signal->sig[i] &~ blocked->sig[i];
 110                break;
 111
 112        case 4: ready  = signal->sig[3] &~ blocked->sig[3];
 113                ready |= signal->sig[2] &~ blocked->sig[2];
 114                ready |= signal->sig[1] &~ blocked->sig[1];
 115                ready |= signal->sig[0] &~ blocked->sig[0];
 116                break;
 117
 118        case 2: ready  = signal->sig[1] &~ blocked->sig[1];
 119                ready |= signal->sig[0] &~ blocked->sig[0];
 120                break;
 121
 122        case 1: ready  = signal->sig[0] &~ blocked->sig[0];
 123        }
 124        return ready != 0;
 125}
 126
 127#define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
 128
 129static int recalc_sigpending_tsk(struct task_struct *t)
 130{
 131        if ((t->jobctl & JOBCTL_PENDING_MASK) ||
 132            PENDING(&t->pending, &t->blocked) ||
 133            PENDING(&t->signal->shared_pending, &t->blocked)) {
 134                set_tsk_thread_flag(t, TIF_SIGPENDING);
 135                return 1;
 136        }
 137        /*
 138         * We must never clear the flag in another thread, or in current
 139         * when it's possible the current syscall is returning -ERESTART*.
 140         * So we don't clear it here, and only callers who know they should do.
 141         */
 142        return 0;
 143}
 144
 145/*
 146 * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
 147 * This is superfluous when called on current, the wakeup is a harmless no-op.
 148 */
 149void recalc_sigpending_and_wake(struct task_struct *t)
 150{
 151        if (recalc_sigpending_tsk(t))
 152                signal_wake_up(t, 0);
 153}
 154
 155void recalc_sigpending(void)
 156{
 157        if (!recalc_sigpending_tsk(current) && !freezing(current))
 158                clear_thread_flag(TIF_SIGPENDING);
 159
 160}
 161
 162/* Given the mask, find the first available signal that should be serviced. */
 163
 164#define SYNCHRONOUS_MASK \
 165        (sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
 166         sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
 167
 168int next_signal(struct sigpending *pending, sigset_t *mask)
 169{
 170        unsigned long i, *s, *m, x;
 171        int sig = 0;
 172
 173        s = pending->signal.sig;
 174        m = mask->sig;
 175
 176        /*
 177         * Handle the first word specially: it contains the
 178         * synchronous signals that need to be dequeued first.
 179         */
 180        x = *s &~ *m;
 181        if (x) {
 182                if (x & SYNCHRONOUS_MASK)
 183                        x &= SYNCHRONOUS_MASK;
 184                sig = ffz(~x) + 1;
 185                return sig;
 186        }
 187
 188        switch (_NSIG_WORDS) {
 189        default:
 190                for (i = 1; i < _NSIG_WORDS; ++i) {
 191                        x = *++s &~ *++m;
 192                        if (!x)
 193                                continue;
 194                        sig = ffz(~x) + i*_NSIG_BPW + 1;
 195                        break;
 196                }
 197                break;
 198
 199        case 2:
 200                x = s[1] &~ m[1];
 201                if (!x)
 202                        break;
 203                sig = ffz(~x) + _NSIG_BPW + 1;
 204                break;
 205
 206        case 1:
 207                /* Nothing to do */
 208                break;
 209        }
 210
 211        return sig;
 212}
 213
 214static inline void print_dropped_signal(int sig)
 215{
 216        static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
 217
 218        if (!print_fatal_signals)
 219                return;
 220
 221        if (!__ratelimit(&ratelimit_state))
 222                return;
 223
 224        printk(KERN_INFO "%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
 225                                current->comm, current->pid, sig);
 226}
 227
 228/**
 229 * task_set_jobctl_pending - set jobctl pending bits
 230 * @task: target task
 231 * @mask: pending bits to set
 232 *
 233 * Clear @mask from @task->jobctl.  @mask must be subset of
 234 * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
 235 * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
 236 * cleared.  If @task is already being killed or exiting, this function
 237 * becomes noop.
 238 *
 239 * CONTEXT:
 240 * Must be called with @task->sighand->siglock held.
 241 *
 242 * RETURNS:
 243 * %true if @mask is set, %false if made noop because @task was dying.
 244 */
 245bool task_set_jobctl_pending(struct task_struct *task, unsigned int mask)
 246{
 247        BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
 248                        JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
 249        BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
 250
 251        if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
 252                return false;
 253
 254        if (mask & JOBCTL_STOP_SIGMASK)
 255                task->jobctl &= ~JOBCTL_STOP_SIGMASK;
 256
 257        task->jobctl |= mask;
 258        return true;
 259}
 260
 261/**
 262 * task_clear_jobctl_trapping - clear jobctl trapping bit
 263 * @task: target task
 264 *
 265 * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
 266 * Clear it and wake up the ptracer.  Note that we don't need any further
 267 * locking.  @task->siglock guarantees that @task->parent points to the
 268 * ptracer.
 269 *
 270 * CONTEXT:
 271 * Must be called with @task->sighand->siglock held.
 272 */
 273void task_clear_jobctl_trapping(struct task_struct *task)
 274{
 275        if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
 276                task->jobctl &= ~JOBCTL_TRAPPING;
 277                wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
 278        }
 279}
 280
 281/**
 282 * task_clear_jobctl_pending - clear jobctl pending bits
 283 * @task: target task
 284 * @mask: pending bits to clear
 285 *
 286 * Clear @mask from @task->jobctl.  @mask must be subset of
 287 * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
 288 * STOP bits are cleared together.
 289 *
 290 * If clearing of @mask leaves no stop or trap pending, this function calls
 291 * task_clear_jobctl_trapping().
 292 *
 293 * CONTEXT:
 294 * Must be called with @task->sighand->siglock held.
 295 */
 296void task_clear_jobctl_pending(struct task_struct *task, unsigned int mask)
 297{
 298        BUG_ON(mask & ~JOBCTL_PENDING_MASK);
 299
 300        if (mask & JOBCTL_STOP_PENDING)
 301                mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
 302
 303        task->jobctl &= ~mask;
 304
 305        if (!(task->jobctl & JOBCTL_PENDING_MASK))
 306                task_clear_jobctl_trapping(task);
 307}
 308
 309/**
 310 * task_participate_group_stop - participate in a group stop
 311 * @task: task participating in a group stop
 312 *
 313 * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
 314 * Group stop states are cleared and the group stop count is consumed if
 315 * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
 316 * stop, the appropriate %SIGNAL_* flags are set.
 317 *
 318 * CONTEXT:
 319 * Must be called with @task->sighand->siglock held.
 320 *
 321 * RETURNS:
 322 * %true if group stop completion should be notified to the parent, %false
 323 * otherwise.
 324 */
 325static bool task_participate_group_stop(struct task_struct *task)
 326{
 327        struct signal_struct *sig = task->signal;
 328        bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
 329
 330        WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
 331
 332        task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
 333
 334        if (!consume)
 335                return false;
 336
 337        if (!WARN_ON_ONCE(sig->group_stop_count == 0))
 338                sig->group_stop_count--;
 339
 340        /*
 341         * Tell the caller to notify completion iff we are entering into a
 342         * fresh group stop.  Read comment in do_signal_stop() for details.
 343         */
 344        if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
 345                sig->flags = SIGNAL_STOP_STOPPED;
 346                return true;
 347        }
 348        return false;
 349}
 350
 351/*
 352 * allocate a new signal queue record
 353 * - this may be called without locks if and only if t == current, otherwise an
 354 *   appropriate lock must be held to stop the target task from exiting
 355 */
 356static struct sigqueue *
 357__sigqueue_alloc(int sig, struct task_struct *t, gfp_t flags, int override_rlimit)
 358{
 359        struct sigqueue *q = NULL;
 360        struct user_struct *user;
 361
 362        /*
 363         * Protect access to @t credentials. This can go away when all
 364         * callers hold rcu read lock.
 365         */
 366        rcu_read_lock();
 367        user = get_uid(__task_cred(t)->user);
 368        atomic_inc(&user->sigpending);
 369        rcu_read_unlock();
 370
 371        if (override_rlimit ||
 372            atomic_read(&user->sigpending) <=
 373                        task_rlimit(t, RLIMIT_SIGPENDING)) {
 374                q = kmem_cache_alloc(sigqueue_cachep, flags);
 375        } else {
 376                print_dropped_signal(sig);
 377        }
 378
 379        if (unlikely(q == NULL)) {
 380                atomic_dec(&user->sigpending);
 381                free_uid(user);
 382        } else {
 383                INIT_LIST_HEAD(&q->list);
 384                q->flags = 0;
 385                q->user = user;
 386        }
 387
 388        return q;
 389}
 390
 391static void __sigqueue_free(struct sigqueue *q)
 392{
 393        if (q->flags & SIGQUEUE_PREALLOC)
 394                return;
 395        atomic_dec(&q->user->sigpending);
 396        free_uid(q->user);
 397        kmem_cache_free(sigqueue_cachep, q);
 398}
 399
 400void flush_sigqueue(struct sigpending *queue)
 401{
 402        struct sigqueue *q;
 403
 404        sigemptyset(&queue->signal);
 405        while (!list_empty(&queue->list)) {
 406                q = list_entry(queue->list.next, struct sigqueue , list);
 407                list_del_init(&q->list);
 408                __sigqueue_free(q);
 409        }
 410}
 411
 412/*
 413 * Flush all pending signals for a task.
 414 */
 415void __flush_signals(struct task_struct *t)
 416{
 417        clear_tsk_thread_flag(t, TIF_SIGPENDING);
 418        flush_sigqueue(&t->pending);
 419        flush_sigqueue(&t->signal->shared_pending);
 420}
 421
 422void flush_signals(struct task_struct *t)
 423{
 424        unsigned long flags;
 425
 426        spin_lock_irqsave(&t->sighand->siglock, flags);
 427        __flush_signals(t);
 428        spin_unlock_irqrestore(&t->sighand->siglock, flags);
 429}
 430
 431static void __flush_itimer_signals(struct sigpending *pending)
 432{
 433        sigset_t signal, retain;
 434        struct sigqueue *q, *n;
 435
 436        signal = pending->signal;
 437        sigemptyset(&retain);
 438
 439        list_for_each_entry_safe(q, n, &pending->list, list) {
 440                int sig = q->info.si_signo;
 441
 442                if (likely(q->info.si_code != SI_TIMER)) {
 443                        sigaddset(&retain, sig);
 444                } else {
 445                        sigdelset(&signal, sig);
 446                        list_del_init(&q->list);
 447                        __sigqueue_free(q);
 448                }
 449        }
 450
 451        sigorsets(&pending->signal, &signal, &retain);
 452}
 453
 454void flush_itimer_signals(void)
 455{
 456        struct task_struct *tsk = current;
 457        unsigned long flags;
 458
 459        spin_lock_irqsave(&tsk->sighand->siglock, flags);
 460        __flush_itimer_signals(&tsk->pending);
 461        __flush_itimer_signals(&tsk->signal->shared_pending);
 462        spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
 463}
 464
 465void ignore_signals(struct task_struct *t)
 466{
 467        int i;
 468
 469        for (i = 0; i < _NSIG; ++i)
 470                t->sighand->action[i].sa.sa_handler = SIG_IGN;
 471
 472        flush_signals(t);
 473}
 474
 475/*
 476 * Flush all handlers for a task.
 477 */
 478
 479void
 480flush_signal_handlers(struct task_struct *t, int force_default)
 481{
 482        int i;
 483        struct k_sigaction *ka = &t->sighand->action[0];
 484        for (i = _NSIG ; i != 0 ; i--) {
 485                if (force_default || ka->sa.sa_handler != SIG_IGN)
 486                        ka->sa.sa_handler = SIG_DFL;
 487                ka->sa.sa_flags = 0;
 488#ifdef SA_RESTORER
 489                ka->sa.sa_restorer = NULL;
 490#endif
 491                sigemptyset(&ka->sa.sa_mask);
 492                ka++;
 493        }
 494}
 495
 496int unhandled_signal(struct task_struct *tsk, int sig)
 497{
 498        void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
 499        if (is_global_init(tsk))
 500                return 1;
 501        if (handler != SIG_IGN && handler != SIG_DFL)
 502                return 0;
 503        /* if ptraced, let the tracer determine */
 504        return !tsk->ptrace;
 505}
 506
 507/*
 508 * Notify the system that a driver wants to block all signals for this
 509 * process, and wants to be notified if any signals at all were to be
 510 * sent/acted upon.  If the notifier routine returns non-zero, then the
 511 * signal will be acted upon after all.  If the notifier routine returns 0,
 512 * then then signal will be blocked.  Only one block per process is
 513 * allowed.  priv is a pointer to private data that the notifier routine
 514 * can use to determine if the signal should be blocked or not.
 515 */
 516void
 517block_all_signals(int (*notifier)(void *priv), void *priv, sigset_t *mask)
 518{
 519        unsigned long flags;
 520
 521        spin_lock_irqsave(&current->sighand->siglock, flags);
 522        current->notifier_mask = mask;
 523        current->notifier_data = priv;
 524        current->notifier = notifier;
 525        spin_unlock_irqrestore(&current->sighand->siglock, flags);
 526}
 527
 528/* Notify the system that blocking has ended. */
 529
 530void
 531unblock_all_signals(void)
 532{
 533        unsigned long flags;
 534
 535        spin_lock_irqsave(&current->sighand->siglock, flags);
 536        current->notifier = NULL;
 537        current->notifier_data = NULL;
 538        recalc_sigpending();
 539        spin_unlock_irqrestore(&current->sighand->siglock, flags);
 540}
 541
 542static void collect_signal(int sig, struct sigpending *list, siginfo_t *info)
 543{
 544        struct sigqueue *q, *first = NULL;
 545
 546        /*
 547         * Collect the siginfo appropriate to this signal.  Check if
 548         * there is another siginfo for the same signal.
 549        */
 550        list_for_each_entry(q, &list->list, list) {
 551                if (q->info.si_signo == sig) {
 552                        if (first)
 553                                goto still_pending;
 554                        first = q;
 555                }
 556        }
 557
 558        sigdelset(&list->signal, sig);
 559
 560        if (first) {
 561still_pending:
 562                list_del_init(&first->list);
 563                copy_siginfo(info, &first->info);
 564                __sigqueue_free(first);
 565        } else {
 566                /*
 567                 * Ok, it wasn't in the queue.  This must be
 568                 * a fast-pathed signal or we must have been
 569                 * out of queue space.  So zero out the info.
 570                 */
 571                info->si_signo = sig;
 572                info->si_errno = 0;
 573                info->si_code = SI_USER;
 574                info->si_pid = 0;
 575                info->si_uid = 0;
 576        }
 577}
 578
 579static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
 580                        siginfo_t *info)
 581{
 582        int sig = next_signal(pending, mask);
 583
 584        if (sig) {
 585                if (current->notifier) {
 586                        if (sigismember(current->notifier_mask, sig)) {
 587                                if (!(current->notifier)(current->notifier_data)) {
 588                                        clear_thread_flag(TIF_SIGPENDING);
 589                                        return 0;
 590                                }
 591                        }
 592                }
 593
 594                collect_signal(sig, pending, info);
 595        }
 596
 597        return sig;
 598}
 599
 600/*
 601 * Dequeue a signal and return the element to the caller, which is
 602 * expected to free it.
 603 *
 604 * All callers have to hold the siglock.
 605 */
 606int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info)
 607{
 608        int signr;
 609
 610        /* We only dequeue private signals from ourselves, we don't let
 611         * signalfd steal them
 612         */
 613        signr = __dequeue_signal(&tsk->pending, mask, info);
 614        if (!signr) {
 615                signr = __dequeue_signal(&tsk->signal->shared_pending,
 616                                         mask, info);
 617                /*
 618                 * itimer signal ?
 619                 *
 620                 * itimers are process shared and we restart periodic
 621                 * itimers in the signal delivery path to prevent DoS
 622                 * attacks in the high resolution timer case. This is
 623                 * compliant with the old way of self-restarting
 624                 * itimers, as the SIGALRM is a legacy signal and only
 625                 * queued once. Changing the restart behaviour to
 626                 * restart the timer in the signal dequeue path is
 627                 * reducing the timer noise on heavy loaded !highres
 628                 * systems too.
 629                 */
 630                if (unlikely(signr == SIGALRM)) {
 631                        struct hrtimer *tmr = &tsk->signal->real_timer;
 632
 633                        if (!hrtimer_is_queued(tmr) &&
 634                            tsk->signal->it_real_incr.tv64 != 0) {
 635                                hrtimer_forward(tmr, tmr->base->get_time(),
 636                                                tsk->signal->it_real_incr);
 637                                hrtimer_restart(tmr);
 638                        }
 639                }
 640        }
 641
 642        recalc_sigpending();
 643        if (!signr)
 644                return 0;
 645
 646        if (unlikely(sig_kernel_stop(signr))) {
 647                /*
 648                 * Set a marker that we have dequeued a stop signal.  Our
 649                 * caller might release the siglock and then the pending
 650                 * stop signal it is about to process is no longer in the
 651                 * pending bitmasks, but must still be cleared by a SIGCONT
 652                 * (and overruled by a SIGKILL).  So those cases clear this
 653                 * shared flag after we've set it.  Note that this flag may
 654                 * remain set after the signal we return is ignored or
 655                 * handled.  That doesn't matter because its only purpose
 656                 * is to alert stop-signal processing code when another
 657                 * processor has come along and cleared the flag.
 658                 */
 659                current->jobctl |= JOBCTL_STOP_DEQUEUED;
 660        }
 661        if ((info->si_code & __SI_MASK) == __SI_TIMER && info->si_sys_private) {
 662                /*
 663                 * Release the siglock to ensure proper locking order
 664                 * of timer locks outside of siglocks.  Note, we leave
 665                 * irqs disabled here, since the posix-timers code is
 666                 * about to disable them again anyway.
 667                 */
 668                spin_unlock(&tsk->sighand->siglock);
 669                do_schedule_next_timer(info);
 670                spin_lock(&tsk->sighand->siglock);
 671        }
 672        return signr;
 673}
 674
 675/*
 676 * Tell a process that it has a new active signal..
 677 *
 678 * NOTE! we rely on the previous spin_lock to
 679 * lock interrupts for us! We can only be called with
 680 * "siglock" held, and the local interrupt must
 681 * have been disabled when that got acquired!
 682 *
 683 * No need to set need_resched since signal event passing
 684 * goes through ->blocked
 685 */
 686void signal_wake_up_state(struct task_struct *t, unsigned int state)
 687{
 688        set_tsk_thread_flag(t, TIF_SIGPENDING);
 689        /*
 690         * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
 691         * case. We don't check t->state here because there is a race with it
 692         * executing another processor and just now entering stopped state.
 693         * By using wake_up_state, we ensure the process will wake up and
 694         * handle its death signal.
 695         */
 696        if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
 697                kick_process(t);
 698}
 699
 700/*
 701 * Remove signals in mask from the pending set and queue.
 702 * Returns 1 if any signals were found.
 703 *
 704 * All callers must be holding the siglock.
 705 *
 706 * This version takes a sigset mask and looks at all signals,
 707 * not just those in the first mask word.
 708 */
 709static int rm_from_queue_full(sigset_t *mask, struct sigpending *s)
 710{
 711        struct sigqueue *q, *n;
 712        sigset_t m;
 713
 714        sigandsets(&m, mask, &s->signal);
 715        if (sigisemptyset(&m))
 716                return 0;
 717
 718        sigandnsets(&s->signal, &s->signal, mask);
 719        list_for_each_entry_safe(q, n, &s->list, list) {
 720                if (sigismember(mask, q->info.si_signo)) {
 721                        list_del_init(&q->list);
 722                        __sigqueue_free(q);
 723                }
 724        }
 725        return 1;
 726}
 727/*
 728 * Remove signals in mask from the pending set and queue.
 729 * Returns 1 if any signals were found.
 730 *
 731 * All callers must be holding the siglock.
 732 */
 733static int rm_from_queue(unsigned long mask, struct sigpending *s)
 734{
 735        struct sigqueue *q, *n;
 736
 737        if (!sigtestsetmask(&s->signal, mask))
 738                return 0;
 739
 740        sigdelsetmask(&s->signal, mask);
 741        list_for_each_entry_safe(q, n, &s->list, list) {
 742                if (q->info.si_signo < SIGRTMIN &&
 743                    (mask & sigmask(q->info.si_signo))) {
 744                        list_del_init(&q->list);
 745                        __sigqueue_free(q);
 746                }
 747        }
 748        return 1;
 749}
 750
 751static inline int is_si_special(const struct siginfo *info)
 752{
 753        return info <= SEND_SIG_FORCED;
 754}
 755
 756static inline bool si_fromuser(const struct siginfo *info)
 757{
 758        return info == SEND_SIG_NOINFO ||
 759                (!is_si_special(info) && SI_FROMUSER(info));
 760}
 761
 762/*
 763 * called with RCU read lock from check_kill_permission()
 764 */
 765static int kill_ok_by_cred(struct task_struct *t)
 766{
 767        const struct cred *cred = current_cred();
 768        const struct cred *tcred = __task_cred(t);
 769
 770        if (uid_eq(cred->euid, tcred->suid) ||
 771            uid_eq(cred->euid, tcred->uid)  ||
 772            uid_eq(cred->uid,  tcred->suid) ||
 773            uid_eq(cred->uid,  tcred->uid))
 774                return 1;
 775
 776        if (ns_capable(tcred->user_ns, CAP_KILL))
 777                return 1;
 778
 779        return 0;
 780}
 781
 782/*
 783 * Bad permissions for sending the signal
 784 * - the caller must hold the RCU read lock
 785 */
 786static int check_kill_permission(int sig, struct siginfo *info,
 787                                 struct task_struct *t)
 788{
 789        struct pid *sid;
 790        int error;
 791
 792        if (!valid_signal(sig))
 793                return -EINVAL;
 794
 795        if (!si_fromuser(info))
 796                return 0;
 797
 798        error = audit_signal_info(sig, t); /* Let audit system see the signal */
 799        if (error)
 800                return error;
 801
 802        if (!same_thread_group(current, t) &&
 803            !kill_ok_by_cred(t)) {
 804                switch (sig) {
 805                case SIGCONT:
 806                        sid = task_session(t);
 807                        /*
 808                         * We don't return the error if sid == NULL. The
 809                         * task was unhashed, the caller must notice this.
 810                         */
 811                        if (!sid || sid == task_session(current))
 812                                break;
 813                default:
 814                        return -EPERM;
 815                }
 816        }
 817
 818        return security_task_kill(t, info, sig, 0);
 819}
 820
 821/**
 822 * ptrace_trap_notify - schedule trap to notify ptracer
 823 * @t: tracee wanting to notify tracer
 824 *
 825 * This function schedules sticky ptrace trap which is cleared on the next
 826 * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
 827 * ptracer.
 828 *
 829 * If @t is running, STOP trap will be taken.  If trapped for STOP and
 830 * ptracer is listening for events, tracee is woken up so that it can
 831 * re-trap for the new event.  If trapped otherwise, STOP trap will be
 832 * eventually taken without returning to userland after the existing traps
 833 * are finished by PTRACE_CONT.
 834 *
 835 * CONTEXT:
 836 * Must be called with @task->sighand->siglock held.
 837 */
 838static void ptrace_trap_notify(struct task_struct *t)
 839{
 840        WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
 841        assert_spin_locked(&t->sighand->siglock);
 842
 843        task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
 844        ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
 845}
 846
 847/*
 848 * Handle magic process-wide effects of stop/continue signals. Unlike
 849 * the signal actions, these happen immediately at signal-generation
 850 * time regardless of blocking, ignoring, or handling.  This does the
 851 * actual continuing for SIGCONT, but not the actual stopping for stop
 852 * signals. The process stop is done as a signal action for SIG_DFL.
 853 *
 854 * Returns true if the signal should be actually delivered, otherwise
 855 * it should be dropped.
 856 */
 857static int prepare_signal(int sig, struct task_struct *p, bool force)
 858{
 859        struct signal_struct *signal = p->signal;
 860        struct task_struct *t;
 861
 862        if (unlikely(signal->flags & SIGNAL_GROUP_EXIT)) {
 863                /*
 864                 * The process is in the middle of dying, nothing to do.
 865                 */
 866        } else if (sig_kernel_stop(sig)) {
 867                /*
 868                 * This is a stop signal.  Remove SIGCONT from all queues.
 869                 */
 870                rm_from_queue(sigmask(SIGCONT), &signal->shared_pending);
 871                t = p;
 872                do {
 873                        rm_from_queue(sigmask(SIGCONT), &t->pending);
 874                } while_each_thread(p, t);
 875        } else if (sig == SIGCONT) {
 876                unsigned int why;
 877                /*
 878                 * Remove all stop signals from all queues, wake all threads.
 879                 */
 880                rm_from_queue(SIG_KERNEL_STOP_MASK, &signal->shared_pending);
 881                t = p;
 882                do {
 883                        task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
 884                        rm_from_queue(SIG_KERNEL_STOP_MASK, &t->pending);
 885                        if (likely(!(t->ptrace & PT_SEIZED)))
 886                                wake_up_state(t, __TASK_STOPPED);
 887                        else
 888                                ptrace_trap_notify(t);
 889                } while_each_thread(p, t);
 890
 891                /*
 892                 * Notify the parent with CLD_CONTINUED if we were stopped.
 893                 *
 894                 * If we were in the middle of a group stop, we pretend it
 895                 * was already finished, and then continued. Since SIGCHLD
 896                 * doesn't queue we report only CLD_STOPPED, as if the next
 897                 * CLD_CONTINUED was dropped.
 898                 */
 899                why = 0;
 900                if (signal->flags & SIGNAL_STOP_STOPPED)
 901                        why |= SIGNAL_CLD_CONTINUED;
 902                else if (signal->group_stop_count)
 903                        why |= SIGNAL_CLD_STOPPED;
 904
 905                if (why) {
 906                        /*
 907                         * The first thread which returns from do_signal_stop()
 908                         * will take ->siglock, notice SIGNAL_CLD_MASK, and
 909                         * notify its parent. See get_signal_to_deliver().
 910                         */
 911                        signal->flags = why | SIGNAL_STOP_CONTINUED;
 912                        signal->group_stop_count = 0;
 913                        signal->group_exit_code = 0;
 914                }
 915        }
 916
 917        return !sig_ignored(p, sig, force);
 918}
 919
 920/*
 921 * Test if P wants to take SIG.  After we've checked all threads with this,
 922 * it's equivalent to finding no threads not blocking SIG.  Any threads not
 923 * blocking SIG were ruled out because they are not running and already
 924 * have pending signals.  Such threads will dequeue from the shared queue
 925 * as soon as they're available, so putting the signal on the shared queue
 926 * will be equivalent to sending it to one such thread.
 927 */
 928static inline int wants_signal(int sig, struct task_struct *p)
 929{
 930        if (sigismember(&p->blocked, sig))
 931                return 0;
 932        if (p->flags & PF_EXITING)
 933                return 0;
 934        if (sig == SIGKILL)
 935                return 1;
 936        if (task_is_stopped_or_traced(p))
 937                return 0;
 938        return task_curr(p) || !signal_pending(p);
 939}
 940
 941static void complete_signal(int sig, struct task_struct *p, int group)
 942{
 943        struct signal_struct *signal = p->signal;
 944        struct task_struct *t;
 945
 946        /*
 947         * Now find a thread we can wake up to take the signal off the queue.
 948         *
 949         * If the main thread wants the signal, it gets first crack.
 950         * Probably the least surprising to the average bear.
 951         */
 952        if (wants_signal(sig, p))
 953                t = p;
 954        else if (!group || thread_group_empty(p))
 955                /*
 956                 * There is just one thread and it does not need to be woken.
 957                 * It will dequeue unblocked signals before it runs again.
 958                 */
 959                return;
 960        else {
 961                /*
 962                 * Otherwise try to find a suitable thread.
 963                 */
 964                t = signal->curr_target;
 965                while (!wants_signal(sig, t)) {
 966                        t = next_thread(t);
 967                        if (t == signal->curr_target)
 968                                /*
 969                                 * No thread needs to be woken.
 970                                 * Any eligible threads will see
 971                                 * the signal in the queue soon.
 972                                 */
 973                                return;
 974                }
 975                signal->curr_target = t;
 976        }
 977
 978        /*
 979         * Found a killable thread.  If the signal will be fatal,
 980         * then start taking the whole group down immediately.
 981         */
 982        if (sig_fatal(p, sig) &&
 983            !(signal->flags & (SIGNAL_UNKILLABLE | SIGNAL_GROUP_EXIT)) &&
 984            !sigismember(&t->real_blocked, sig) &&
 985            (sig == SIGKILL || !t->ptrace)) {
 986                /*
 987                 * This signal will be fatal to the whole group.
 988                 */
 989                if (!sig_kernel_coredump(sig)) {
 990                        /*
 991                         * Start a group exit and wake everybody up.
 992                         * This way we don't have other threads
 993                         * running and doing things after a slower
 994                         * thread has the fatal signal pending.
 995                         */
 996                        signal->flags = SIGNAL_GROUP_EXIT;
 997                        signal->group_exit_code = sig;
 998                        signal->group_stop_count = 0;
 999                        t = p;
1000                        do {
1001                                task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1002                                sigaddset(&t->pending.signal, SIGKILL);
1003                                signal_wake_up(t, 1);
1004                        } while_each_thread(p, t);
1005                        return;
1006                }
1007        }
1008
1009        /*
1010         * The signal is already in the shared-pending queue.
1011         * Tell the chosen thread to wake up and dequeue it.
1012         */
1013        signal_wake_up(t, sig == SIGKILL);
1014        return;
1015}
1016
1017static inline int legacy_queue(struct sigpending *signals, int sig)
1018{
1019        return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1020}
1021
1022#ifdef CONFIG_USER_NS
1023static inline void userns_fixup_signal_uid(struct siginfo *info, struct task_struct *t)
1024{
1025        if (current_user_ns() == task_cred_xxx(t, user_ns))
1026                return;
1027
1028        if (SI_FROMKERNEL(info))
1029                return;
1030
1031        rcu_read_lock();
1032        info->si_uid = from_kuid_munged(task_cred_xxx(t, user_ns),
1033                                        make_kuid(current_user_ns(), info->si_uid));
1034        rcu_read_unlock();
1035}
1036#else
1037static inline void userns_fixup_signal_uid(struct siginfo *info, struct task_struct *t)
1038{
1039        return;
1040}
1041#endif
1042
1043static int __send_signal(int sig, struct siginfo *info, struct task_struct *t,
1044                        int group, int from_ancestor_ns)
1045{
1046        struct sigpending *pending;
1047        struct sigqueue *q;
1048        int override_rlimit;
1049        int ret = 0, result;
1050
1051        assert_spin_locked(&t->sighand->siglock);
1052
1053        result = TRACE_SIGNAL_IGNORED;
1054        if (!prepare_signal(sig, t,
1055                        from_ancestor_ns || (info == SEND_SIG_FORCED)))
1056                goto ret;
1057
1058        pending = group ? &t->signal->shared_pending : &t->pending;
1059        /*
1060         * Short-circuit ignored signals and support queuing
1061         * exactly one non-rt signal, so that we can get more
1062         * detailed information about the cause of the signal.
1063         */
1064        result = TRACE_SIGNAL_ALREADY_PENDING;
1065        if (legacy_queue(pending, sig))
1066                goto ret;
1067
1068        result = TRACE_SIGNAL_DELIVERED;
1069        /*
1070         * fast-pathed signals for kernel-internal things like SIGSTOP
1071         * or SIGKILL.
1072         */
1073        if (info == SEND_SIG_FORCED)
1074                goto out_set;
1075
1076        /*
1077         * Real-time signals must be queued if sent by sigqueue, or
1078         * some other real-time mechanism.  It is implementation
1079         * defined whether kill() does so.  We attempt to do so, on
1080         * the principle of least surprise, but since kill is not
1081         * allowed to fail with EAGAIN when low on memory we just
1082         * make sure at least one signal gets delivered and don't
1083         * pass on the info struct.
1084         */
1085        if (sig < SIGRTMIN)
1086                override_rlimit = (is_si_special(info) || info->si_code >= 0);
1087        else
1088                override_rlimit = 0;
1089
1090        q = __sigqueue_alloc(sig, t, GFP_ATOMIC | __GFP_NOTRACK_FALSE_POSITIVE,
1091                override_rlimit);
1092        if (q) {
1093                list_add_tail(&q->list, &pending->list);
1094                switch ((unsigned long) info) {
1095                case (unsigned long) SEND_SIG_NOINFO:
1096                        q->info.si_signo = sig;
1097                        q->info.si_errno = 0;
1098                        q->info.si_code = SI_USER;
1099                        q->info.si_pid = task_tgid_nr_ns(current,
1100                                                        task_active_pid_ns(t));
1101                        q->info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1102                        break;
1103                case (unsigned long) SEND_SIG_PRIV:
1104                        q->info.si_signo = sig;
1105                        q->info.si_errno = 0;
1106                        q->info.si_code = SI_KERNEL;
1107                        q->info.si_pid = 0;
1108                        q->info.si_uid = 0;
1109                        break;
1110                default:
1111                        copy_siginfo(&q->info, info);
1112                        if (from_ancestor_ns)
1113                                q->info.si_pid = 0;
1114                        break;
1115                }
1116
1117                userns_fixup_signal_uid(&q->info, t);
1118
1119        } else if (!is_si_special(info)) {
1120                if (sig >= SIGRTMIN && info->si_code != SI_USER) {
1121                        /*
1122                         * Queue overflow, abort.  We may abort if the
1123                         * signal was rt and sent by user using something
1124                         * other than kill().
1125                         */
1126                        result = TRACE_SIGNAL_OVERFLOW_FAIL;
1127                        ret = -EAGAIN;
1128                        goto ret;
1129                } else {
1130                        /*
1131                         * This is a silent loss of information.  We still
1132                         * send the signal, but the *info bits are lost.
1133                         */
1134                        result = TRACE_SIGNAL_LOSE_INFO;
1135                }
1136        }
1137
1138out_set:
1139        signalfd_notify(t, sig);
1140        sigaddset(&pending->signal, sig);
1141        complete_signal(sig, t, group);
1142ret:
1143        trace_signal_generate(sig, info, t, group, result);
1144        return ret;
1145}
1146
1147static int send_signal(int sig, struct siginfo *info, struct task_struct *t,
1148                        int group)
1149{
1150        int from_ancestor_ns = 0;
1151
1152#ifdef CONFIG_PID_NS
1153        from_ancestor_ns = si_fromuser(info) &&
1154                           !task_pid_nr_ns(current, task_active_pid_ns(t));
1155#endif
1156
1157        return __send_signal(sig, info, t, group, from_ancestor_ns);
1158}
1159
1160static void print_fatal_signal(int signr)
1161{
1162        struct pt_regs *regs = signal_pt_regs();
1163        printk("%s/%d: potentially unexpected fatal signal %d.\n",
1164                current->comm, task_pid_nr(current), signr);
1165
1166#if defined(__i386__) && !defined(__arch_um__)
1167        printk("code at %08lx: ", regs->ip);
1168        {
1169                int i;
1170                for (i = 0; i < 16; i++) {
1171                        unsigned char insn;
1172
1173                        if (get_user(insn, (unsigned char *)(regs->ip + i)))
1174                                break;
1175                        printk("%02x ", insn);
1176                }
1177        }
1178#endif
1179        printk("\n");
1180        preempt_disable();
1181        show_regs(regs);
1182        preempt_enable();
1183}
1184
1185static int __init setup_print_fatal_signals(char *str)
1186{
1187        get_option (&str, &print_fatal_signals);
1188
1189        return 1;
1190}
1191
1192__setup("print-fatal-signals=", setup_print_fatal_signals);
1193
1194int
1195__group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1196{
1197        return send_signal(sig, info, p, 1);
1198}
1199
1200static int
1201specific_send_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1202{
1203        return send_signal(sig, info, t, 0);
1204}
1205
1206int do_send_sig_info(int sig, struct siginfo *info, struct task_struct *p,
1207                        bool group)
1208{
1209        unsigned long flags;
1210        int ret = -ESRCH;
1211
1212        if (lock_task_sighand(p, &flags)) {
1213                ret = send_signal(sig, info, p, group);
1214                unlock_task_sighand(p, &flags);
1215        }
1216
1217        return ret;
1218}
1219
1220/*
1221 * Force a signal that the process can't ignore: if necessary
1222 * we unblock the signal and change any SIG_IGN to SIG_DFL.
1223 *
1224 * Note: If we unblock the signal, we always reset it to SIG_DFL,
1225 * since we do not want to have a signal handler that was blocked
1226 * be invoked when user space had explicitly blocked it.
1227 *
1228 * We don't want to have recursive SIGSEGV's etc, for example,
1229 * that is why we also clear SIGNAL_UNKILLABLE.
1230 */
1231int
1232force_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1233{
1234        unsigned long int flags;
1235        int ret, blocked, ignored;
1236        struct k_sigaction *action;
1237
1238        spin_lock_irqsave(&t->sighand->siglock, flags);
1239        action = &t->sighand->action[sig-1];
1240        ignored = action->sa.sa_handler == SIG_IGN;
1241        blocked = sigismember(&t->blocked, sig);
1242        if (blocked || ignored) {
1243                action->sa.sa_handler = SIG_DFL;
1244                if (blocked) {
1245                        sigdelset(&t->blocked, sig);
1246                        recalc_sigpending_and_wake(t);
1247                }
1248        }
1249        if (action->sa.sa_handler == SIG_DFL)
1250                t->signal->flags &= ~SIGNAL_UNKILLABLE;
1251        ret = specific_send_sig_info(sig, info, t);
1252        spin_unlock_irqrestore(&t->sighand->siglock, flags);
1253
1254        return ret;
1255}
1256
1257/*
1258 * Nuke all other threads in the group.
1259 */
1260int zap_other_threads(struct task_struct *p)
1261{
1262        struct task_struct *t = p;
1263        int count = 0;
1264
1265        p->signal->group_stop_count = 0;
1266
1267        while_each_thread(p, t) {
1268                task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1269                count++;
1270
1271                /* Don't bother with already dead threads */
1272                if (t->exit_state)
1273                        continue;
1274                sigaddset(&t->pending.signal, SIGKILL);
1275                signal_wake_up(t, 1);
1276        }
1277
1278        return count;
1279}
1280
1281struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1282                                           unsigned long *flags)
1283{
1284        struct sighand_struct *sighand;
1285
1286        for (;;) {
1287                local_irq_save(*flags);
1288                rcu_read_lock();
1289                sighand = rcu_dereference(tsk->sighand);
1290                if (unlikely(sighand == NULL)) {
1291                        rcu_read_unlock();
1292                        local_irq_restore(*flags);
1293                        break;
1294                }
1295
1296                spin_lock(&sighand->siglock);
1297                if (likely(sighand == tsk->sighand)) {
1298                        rcu_read_unlock();
1299                        break;
1300                }
1301                spin_unlock(&sighand->siglock);
1302                rcu_read_unlock();
1303                local_irq_restore(*flags);
1304        }
1305
1306        return sighand;
1307}
1308
1309/*
1310 * send signal info to all the members of a group
1311 */
1312int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1313{
1314        int ret;
1315
1316        rcu_read_lock();
1317        ret = check_kill_permission(sig, info, p);
1318        rcu_read_unlock();
1319
1320        if (!ret && sig)
1321                ret = do_send_sig_info(sig, info, p, true);
1322
1323        return ret;
1324}
1325
1326/*
1327 * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1328 * control characters do (^C, ^Z etc)
1329 * - the caller must hold at least a readlock on tasklist_lock
1330 */
1331int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp)
1332{
1333        struct task_struct *p = NULL;
1334        int retval, success;
1335
1336        success = 0;
1337        retval = -ESRCH;
1338        do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1339                int err = group_send_sig_info(sig, info, p);
1340                success |= !err;
1341                retval = err;
1342        } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1343        return success ? 0 : retval;
1344}
1345
1346int kill_pid_info(int sig, struct siginfo *info, struct pid *pid)
1347{
1348        int error = -ESRCH;
1349        struct task_struct *p;
1350
1351        rcu_read_lock();
1352retry:
1353        p = pid_task(pid, PIDTYPE_PID);
1354        if (p) {
1355                error = group_send_sig_info(sig, info, p);
1356                if (unlikely(error == -ESRCH))
1357                        /*
1358                         * The task was unhashed in between, try again.
1359                         * If it is dead, pid_task() will return NULL,
1360                         * if we race with de_thread() it will find the
1361                         * new leader.
1362                         */
1363                        goto retry;
1364        }
1365        rcu_read_unlock();
1366
1367        return error;
1368}
1369
1370int kill_proc_info(int sig, struct siginfo *info, pid_t pid)
1371{
1372        int error;
1373        rcu_read_lock();
1374        error = kill_pid_info(sig, info, find_vpid(pid));
1375        rcu_read_unlock();
1376        return error;
1377}
1378
1379static int kill_as_cred_perm(const struct cred *cred,
1380                             struct task_struct *target)
1381{
1382        const struct cred *pcred = __task_cred(target);
1383        if (!uid_eq(cred->euid, pcred->suid) && !uid_eq(cred->euid, pcred->uid) &&
1384            !uid_eq(cred->uid,  pcred->suid) && !uid_eq(cred->uid,  pcred->uid))
1385                return 0;
1386        return 1;
1387}
1388
1389/* like kill_pid_info(), but doesn't use uid/euid of "current" */
1390int kill_pid_info_as_cred(int sig, struct siginfo *info, struct pid *pid,
1391                         const struct cred *cred, u32 secid)
1392{
1393        int ret = -EINVAL;
1394        struct task_struct *p;
1395        unsigned long flags;
1396
1397        if (!valid_signal(sig))
1398                return ret;
1399
1400        rcu_read_lock();
1401        p = pid_task(pid, PIDTYPE_PID);
1402        if (!p) {
1403                ret = -ESRCH;
1404                goto out_unlock;
1405        }
1406        if (si_fromuser(info) && !kill_as_cred_perm(cred, p)) {
1407                ret = -EPERM;
1408                goto out_unlock;
1409        }
1410        ret = security_task_kill(p, info, sig, secid);
1411        if (ret)
1412                goto out_unlock;
1413
1414        if (sig) {
1415                if (lock_task_sighand(p, &flags)) {
1416                        ret = __send_signal(sig, info, p, 1, 0);
1417                        unlock_task_sighand(p, &flags);
1418                } else
1419                        ret = -ESRCH;
1420        }
1421out_unlock:
1422        rcu_read_unlock();
1423        return ret;
1424}
1425EXPORT_SYMBOL_GPL(kill_pid_info_as_cred);
1426
1427/*
1428 * kill_something_info() interprets pid in interesting ways just like kill(2).
1429 *
1430 * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1431 * is probably wrong.  Should make it like BSD or SYSV.
1432 */
1433
1434static int kill_something_info(int sig, struct siginfo *info, pid_t pid)
1435{
1436        int ret;
1437
1438        if (pid > 0) {
1439                rcu_read_lock();
1440                ret = kill_pid_info(sig, info, find_vpid(pid));
1441                rcu_read_unlock();
1442                return ret;
1443        }
1444
1445        read_lock(&tasklist_lock);
1446        if (pid != -1) {
1447                ret = __kill_pgrp_info(sig, info,
1448                                pid ? find_vpid(-pid) : task_pgrp(current));
1449        } else {
1450                int retval = 0, count = 0;
1451                struct task_struct * p;
1452
1453                for_each_process(p) {
1454                        if (task_pid_vnr(p) > 1 &&
1455                                        !same_thread_group(p, current)) {
1456                                int err = group_send_sig_info(sig, info, p);
1457                                ++count;
1458                                if (err != -EPERM)
1459                                        retval = err;
1460                        }
1461                }
1462                ret = count ? retval : -ESRCH;
1463        }
1464        read_unlock(&tasklist_lock);
1465
1466        return ret;
1467}
1468
1469/*
1470 * These are for backward compatibility with the rest of the kernel source.
1471 */
1472
1473int send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1474{
1475        /*
1476         * Make sure legacy kernel users don't send in bad values
1477         * (normal paths check this in check_kill_permission).
1478         */
1479        if (!valid_signal(sig))
1480                return -EINVAL;
1481
1482        return do_send_sig_info(sig, info, p, false);
1483}
1484
1485#define __si_special(priv) \
1486        ((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1487
1488int
1489send_sig(int sig, struct task_struct *p, int priv)
1490{
1491        return send_sig_info(sig, __si_special(priv), p);
1492}
1493
1494void
1495force_sig(int sig, struct task_struct *p)
1496{
1497        force_sig_info(sig, SEND_SIG_PRIV, p);
1498}
1499
1500/*
1501 * When things go south during signal handling, we
1502 * will force a SIGSEGV. And if the signal that caused
1503 * the problem was already a SIGSEGV, we'll want to
1504 * make sure we don't even try to deliver the signal..
1505 */
1506int
1507force_sigsegv(int sig, struct task_struct *p)
1508{
1509        if (sig == SIGSEGV) {
1510                unsigned long flags;
1511                spin_lock_irqsave(&p->sighand->siglock, flags);
1512                p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1513                spin_unlock_irqrestore(&p->sighand->siglock, flags);
1514        }
1515        force_sig(SIGSEGV, p);
1516        return 0;
1517}
1518
1519int kill_pgrp(struct pid *pid, int sig, int priv)
1520{
1521        int ret;
1522
1523        read_lock(&tasklist_lock);
1524        ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1525        read_unlock(&tasklist_lock);
1526
1527        return ret;
1528}
1529EXPORT_SYMBOL(kill_pgrp);
1530
1531int kill_pid(struct pid *pid, int sig, int priv)
1532{
1533        return kill_pid_info(sig, __si_special(priv), pid);
1534}
1535EXPORT_SYMBOL(kill_pid);
1536
1537/*
1538 * These functions support sending signals using preallocated sigqueue
1539 * structures.  This is needed "because realtime applications cannot
1540 * afford to lose notifications of asynchronous events, like timer
1541 * expirations or I/O completions".  In the case of POSIX Timers
1542 * we allocate the sigqueue structure from the timer_create.  If this
1543 * allocation fails we are able to report the failure to the application
1544 * with an EAGAIN error.
1545 */
1546struct sigqueue *sigqueue_alloc(void)
1547{
1548        struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1549
1550        if (q)
1551                q->flags |= SIGQUEUE_PREALLOC;
1552
1553        return q;
1554}
1555
1556void sigqueue_free(struct sigqueue *q)
1557{
1558        unsigned long flags;
1559        spinlock_t *lock = &current->sighand->siglock;
1560
1561        BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1562        /*
1563         * We must hold ->siglock while testing q->list
1564         * to serialize with collect_signal() or with
1565         * __exit_signal()->flush_sigqueue().
1566         */
1567        spin_lock_irqsave(lock, flags);
1568        q->flags &= ~SIGQUEUE_PREALLOC;
1569        /*
1570         * If it is queued it will be freed when dequeued,
1571         * like the "regular" sigqueue.
1572         */
1573        if (!list_empty(&q->list))
1574                q = NULL;
1575        spin_unlock_irqrestore(lock, flags);
1576
1577        if (q)
1578                __sigqueue_free(q);
1579}
1580
1581int send_sigqueue(struct sigqueue *q, struct task_struct *t, int group)
1582{
1583        int sig = q->info.si_signo;
1584        struct sigpending *pending;
1585        unsigned long flags;
1586        int ret, result;
1587
1588        BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1589
1590        ret = -1;
1591        if (!likely(lock_task_sighand(t, &flags)))
1592                goto ret;
1593
1594        ret = 1; /* the signal is ignored */
1595        result = TRACE_SIGNAL_IGNORED;
1596        if (!prepare_signal(sig, t, false))
1597                goto out;
1598
1599        ret = 0;
1600        if (unlikely(!list_empty(&q->list))) {
1601                /*
1602                 * If an SI_TIMER entry is already queue just increment
1603                 * the overrun count.
1604                 */
1605                BUG_ON(q->info.si_code != SI_TIMER);
1606                q->info.si_overrun++;
1607                result = TRACE_SIGNAL_ALREADY_PENDING;
1608                goto out;
1609        }
1610        q->info.si_overrun = 0;
1611
1612        signalfd_notify(t, sig);
1613        pending = group ? &t->signal->shared_pending : &t->pending;
1614        list_add_tail(&q->list, &pending->list);
1615        sigaddset(&pending->signal, sig);
1616        complete_signal(sig, t, group);
1617        result = TRACE_SIGNAL_DELIVERED;
1618out:
1619        trace_signal_generate(sig, &q->info, t, group, result);
1620        unlock_task_sighand(t, &flags);
1621ret:
1622        return ret;
1623}
1624
1625/*
1626 * Let a parent know about the death of a child.
1627 * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1628 *
1629 * Returns true if our parent ignored us and so we've switched to
1630 * self-reaping.
1631 */
1632bool do_notify_parent(struct task_struct *tsk, int sig)
1633{
1634        struct siginfo info;
1635        unsigned long flags;
1636        struct sighand_struct *psig;
1637        bool autoreap = false;
1638
1639        BUG_ON(sig == -1);
1640
1641        /* do_notify_parent_cldstop should have been called instead.  */
1642        BUG_ON(task_is_stopped_or_traced(tsk));
1643
1644        BUG_ON(!tsk->ptrace &&
1645               (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1646
1647        if (sig != SIGCHLD) {
1648                /*
1649                 * This is only possible if parent == real_parent.
1650                 * Check if it has changed security domain.
1651                 */
1652                if (tsk->parent_exec_id != tsk->parent->self_exec_id)
1653                        sig = SIGCHLD;
1654        }
1655
1656        info.si_signo = sig;
1657        info.si_errno = 0;
1658        /*
1659         * We are under tasklist_lock here so our parent is tied to
1660         * us and cannot change.
1661         *
1662         * task_active_pid_ns will always return the same pid namespace
1663         * until a task passes through release_task.
1664         *
1665         * write_lock() currently calls preempt_disable() which is the
1666         * same as rcu_read_lock(), but according to Oleg, this is not
1667         * correct to rely on this
1668         */
1669        rcu_read_lock();
1670        info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1671        info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1672                                       task_uid(tsk));
1673        rcu_read_unlock();
1674
1675        info.si_utime = cputime_to_clock_t(tsk->utime + tsk->signal->utime);
1676        info.si_stime = cputime_to_clock_t(tsk->stime + tsk->signal->stime);
1677
1678        info.si_status = tsk->exit_code & 0x7f;
1679        if (tsk->exit_code & 0x80)
1680                info.si_code = CLD_DUMPED;
1681        else if (tsk->exit_code & 0x7f)
1682                info.si_code = CLD_KILLED;
1683        else {
1684                info.si_code = CLD_EXITED;
1685                info.si_status = tsk->exit_code >> 8;
1686        }
1687
1688        psig = tsk->parent->sighand;
1689        spin_lock_irqsave(&psig->siglock, flags);
1690        if (!tsk->ptrace && sig == SIGCHLD &&
1691            (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1692             (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1693                /*
1694                 * We are exiting and our parent doesn't care.  POSIX.1
1695                 * defines special semantics for setting SIGCHLD to SIG_IGN
1696                 * or setting the SA_NOCLDWAIT flag: we should be reaped
1697                 * automatically and not left for our parent's wait4 call.
1698                 * Rather than having the parent do it as a magic kind of
1699                 * signal handler, we just set this to tell do_exit that we
1700                 * can be cleaned up without becoming a zombie.  Note that
1701                 * we still call __wake_up_parent in this case, because a
1702                 * blocked sys_wait4 might now return -ECHILD.
1703                 *
1704                 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1705                 * is implementation-defined: we do (if you don't want
1706                 * it, just use SIG_IGN instead).
1707                 */
1708                autoreap = true;
1709                if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
1710                        sig = 0;
1711        }
1712        if (valid_signal(sig) && sig)
1713                __group_send_sig_info(sig, &info, tsk->parent);
1714        __wake_up_parent(tsk, tsk->parent);
1715        spin_unlock_irqrestore(&psig->siglock, flags);
1716
1717        return autoreap;
1718}
1719
1720/**
1721 * do_notify_parent_cldstop - notify parent of stopped/continued state change
1722 * @tsk: task reporting the state change
1723 * @for_ptracer: the notification is for ptracer
1724 * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
1725 *
1726 * Notify @tsk's parent that the stopped/continued state has changed.  If
1727 * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
1728 * If %true, @tsk reports to @tsk->parent which should be the ptracer.
1729 *
1730 * CONTEXT:
1731 * Must be called with tasklist_lock at least read locked.
1732 */
1733static void do_notify_parent_cldstop(struct task_struct *tsk,
1734                                     bool for_ptracer, int why)
1735{
1736        struct siginfo info;
1737        unsigned long flags;
1738        struct task_struct *parent;
1739        struct sighand_struct *sighand;
1740
1741        if (for_ptracer) {
1742                parent = tsk->parent;
1743        } else {
1744                tsk = tsk->group_leader;
1745                parent = tsk->real_parent;
1746        }
1747
1748        info.si_signo = SIGCHLD;
1749        info.si_errno = 0;
1750        /*
1751         * see comment in do_notify_parent() about the following 4 lines
1752         */
1753        rcu_read_lock();
1754        info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
1755        info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
1756        rcu_read_unlock();
1757
1758        info.si_utime = cputime_to_clock_t(tsk->utime);
1759        info.si_stime = cputime_to_clock_t(tsk->stime);
1760
1761        info.si_code = why;
1762        switch (why) {
1763        case CLD_CONTINUED:
1764                info.si_status = SIGCONT;
1765                break;
1766        case CLD_STOPPED:
1767                info.si_status = tsk->signal->group_exit_code & 0x7f;
1768                break;
1769        case CLD_TRAPPED:
1770                info.si_status = tsk->exit_code & 0x7f;
1771                break;
1772        default:
1773                BUG();
1774        }
1775
1776        sighand = parent->sighand;
1777        spin_lock_irqsave(&sighand->siglock, flags);
1778        if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
1779            !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
1780                __group_send_sig_info(SIGCHLD, &info, parent);
1781        /*
1782         * Even if SIGCHLD is not generated, we must wake up wait4 calls.
1783         */
1784        __wake_up_parent(tsk, parent);
1785        spin_unlock_irqrestore(&sighand->siglock, flags);
1786}
1787
1788static inline int may_ptrace_stop(void)
1789{
1790        if (!likely(current->ptrace))
1791                return 0;
1792        /*
1793         * Are we in the middle of do_coredump?
1794         * If so and our tracer is also part of the coredump stopping
1795         * is a deadlock situation, and pointless because our tracer
1796         * is dead so don't allow us to stop.
1797         * If SIGKILL was already sent before the caller unlocked
1798         * ->siglock we must see ->core_state != NULL. Otherwise it
1799         * is safe to enter schedule().
1800         *
1801         * This is almost outdated, a task with the pending SIGKILL can't
1802         * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
1803         * after SIGKILL was already dequeued.
1804         */
1805        if (unlikely(current->mm->core_state) &&
1806            unlikely(current->mm == current->parent->mm))
1807                return 0;
1808
1809        return 1;
1810}
1811
1812/*
1813 * Return non-zero if there is a SIGKILL that should be waking us up.
1814 * Called with the siglock held.
1815 */
1816static int sigkill_pending(struct task_struct *tsk)
1817{
1818        return  sigismember(&tsk->pending.signal, SIGKILL) ||
1819                sigismember(&tsk->signal->shared_pending.signal, SIGKILL);
1820}
1821
1822/*
1823 * This must be called with current->sighand->siglock held.
1824 *
1825 * This should be the path for all ptrace stops.
1826 * We always set current->last_siginfo while stopped here.
1827 * That makes it a way to test a stopped process for
1828 * being ptrace-stopped vs being job-control-stopped.
1829 *
1830 * If we actually decide not to stop at all because the tracer
1831 * is gone, we keep current->exit_code unless clear_code.
1832 */
1833static void ptrace_stop(int exit_code, int why, int clear_code, siginfo_t *info)
1834        __releases(&current->sighand->siglock)
1835        __acquires(&current->sighand->siglock)
1836{
1837        bool gstop_done = false;
1838
1839        if (arch_ptrace_stop_needed(exit_code, info)) {
1840                /*
1841                 * The arch code has something special to do before a
1842                 * ptrace stop.  This is allowed to block, e.g. for faults
1843                 * on user stack pages.  We can't keep the siglock while
1844                 * calling arch_ptrace_stop, so we must release it now.
1845                 * To preserve proper semantics, we must do this before
1846                 * any signal bookkeeping like checking group_stop_count.
1847                 * Meanwhile, a SIGKILL could come in before we retake the
1848                 * siglock.  That must prevent us from sleeping in TASK_TRACED.
1849                 * So after regaining the lock, we must check for SIGKILL.
1850                 */
1851                spin_unlock_irq(&current->sighand->siglock);
1852                arch_ptrace_stop(exit_code, info);
1853                spin_lock_irq(&current->sighand->siglock);
1854                if (sigkill_pending(current))
1855                        return;
1856        }
1857
1858        /*
1859         * We're committing to trapping.  TRACED should be visible before
1860         * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
1861         * Also, transition to TRACED and updates to ->jobctl should be
1862         * atomic with respect to siglock and should be done after the arch
1863         * hook as siglock is released and regrabbed across it.
1864         */
1865        set_current_state(TASK_TRACED);
1866
1867        current->last_siginfo = info;
1868        current->exit_code = exit_code;
1869
1870        /*
1871         * If @why is CLD_STOPPED, we're trapping to participate in a group
1872         * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
1873         * across siglock relocks since INTERRUPT was scheduled, PENDING
1874         * could be clear now.  We act as if SIGCONT is received after
1875         * TASK_TRACED is entered - ignore it.
1876         */
1877        if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
1878                gstop_done = task_participate_group_stop(current);
1879
1880        /* any trap clears pending STOP trap, STOP trap clears NOTIFY */
1881        task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
1882        if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
1883                task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
1884
1885        /* entering a trap, clear TRAPPING */
1886        task_clear_jobctl_trapping(current);
1887
1888        spin_unlock_irq(&current->sighand->siglock);
1889        read_lock(&tasklist_lock);
1890        if (may_ptrace_stop()) {
1891                /*
1892                 * Notify parents of the stop.
1893                 *
1894                 * While ptraced, there are two parents - the ptracer and
1895                 * the real_parent of the group_leader.  The ptracer should
1896                 * know about every stop while the real parent is only
1897                 * interested in the completion of group stop.  The states
1898                 * for the two don't interact with each other.  Notify
1899                 * separately unless they're gonna be duplicates.
1900                 */
1901                do_notify_parent_cldstop(current, true, why);
1902                if (gstop_done && ptrace_reparented(current))
1903                        do_notify_parent_cldstop(current, false, why);
1904
1905                /*
1906                 * Don't want to allow preemption here, because
1907                 * sys_ptrace() needs this task to be inactive.
1908                 *
1909                 * XXX: implement read_unlock_no_resched().
1910                 */
1911                preempt_disable();
1912                read_unlock(&tasklist_lock);
1913                preempt_enable_no_resched();
1914                freezable_schedule();
1915        } else {
1916                /*
1917                 * By the time we got the lock, our tracer went away.
1918                 * Don't drop the lock yet, another tracer may come.
1919                 *
1920                 * If @gstop_done, the ptracer went away between group stop
1921                 * completion and here.  During detach, it would have set
1922                 * JOBCTL_STOP_PENDING on us and we'll re-enter
1923                 * TASK_STOPPED in do_signal_stop() on return, so notifying
1924                 * the real parent of the group stop completion is enough.
1925                 */
1926                if (gstop_done)
1927                        do_notify_parent_cldstop(current, false, why);
1928
1929                /* tasklist protects us from ptrace_freeze_traced() */
1930                __set_current_state(TASK_RUNNING);
1931                if (clear_code)
1932                        current->exit_code = 0;
1933                read_unlock(&tasklist_lock);
1934        }
1935
1936        /*
1937         * We are back.  Now reacquire the siglock before touching
1938         * last_siginfo, so that we are sure to have synchronized with
1939         * any signal-sending on another CPU that wants to examine it.
1940         */
1941        spin_lock_irq(&current->sighand->siglock);
1942        current->last_siginfo = NULL;
1943
1944        /* LISTENING can be set only during STOP traps, clear it */
1945        current->jobctl &= ~JOBCTL_LISTENING;
1946
1947        /*
1948         * Queued signals ignored us while we were stopped for tracing.
1949         * So check for any that we should take before resuming user mode.
1950         * This sets TIF_SIGPENDING, but never clears it.
1951         */
1952        recalc_sigpending_tsk(current);
1953}
1954
1955static void ptrace_do_notify(int signr, int exit_code, int why)
1956{
1957        siginfo_t info;
1958
1959        memset(&info, 0, sizeof info);
1960        info.si_signo = signr;
1961        info.si_code = exit_code;
1962        info.si_pid = task_pid_vnr(current);
1963        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1964
1965        /* Let the debugger run.  */
1966        ptrace_stop(exit_code, why, 1, &info);
1967}
1968
1969void ptrace_notify(int exit_code)
1970{
1971        BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
1972        if (unlikely(current->task_works))
1973                task_work_run();
1974
1975        spin_lock_irq(&current->sighand->siglock);
1976        ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
1977        spin_unlock_irq(&current->sighand->siglock);
1978}
1979
1980/**
1981 * do_signal_stop - handle group stop for SIGSTOP and other stop signals
1982 * @signr: signr causing group stop if initiating
1983 *
1984 * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
1985 * and participate in it.  If already set, participate in the existing
1986 * group stop.  If participated in a group stop (and thus slept), %true is
1987 * returned with siglock released.
1988 *
1989 * If ptraced, this function doesn't handle stop itself.  Instead,
1990 * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
1991 * untouched.  The caller must ensure that INTERRUPT trap handling takes
1992 * places afterwards.
1993 *
1994 * CONTEXT:
1995 * Must be called with @current->sighand->siglock held, which is released
1996 * on %true return.
1997 *
1998 * RETURNS:
1999 * %false if group stop is already cancelled or ptrace trap is scheduled.
2000 * %true if participated in group stop.
2001 */
2002static bool do_signal_stop(int signr)
2003        __releases(&current->sighand->siglock)
2004{
2005        struct signal_struct *sig = current->signal;
2006
2007        if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2008                unsigned int gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2009                struct task_struct *t;
2010
2011                /* signr will be recorded in task->jobctl for retries */
2012                WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2013
2014                if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2015                    unlikely(signal_group_exit(sig)))
2016                        return false;
2017                /*
2018                 * There is no group stop already in progress.  We must
2019                 * initiate one now.
2020                 *
2021                 * While ptraced, a task may be resumed while group stop is
2022                 * still in effect and then receive a stop signal and
2023                 * initiate another group stop.  This deviates from the
2024                 * usual behavior as two consecutive stop signals can't
2025                 * cause two group stops when !ptraced.  That is why we
2026                 * also check !task_is_stopped(t) below.
2027                 *
2028                 * The condition can be distinguished by testing whether
2029                 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2030                 * group_exit_code in such case.
2031                 *
2032                 * This is not necessary for SIGNAL_STOP_CONTINUED because
2033                 * an intervening stop signal is required to cause two
2034                 * continued events regardless of ptrace.
2035                 */
2036                if (!(sig->flags & SIGNAL_STOP_STOPPED))
2037                        sig->group_exit_code = signr;
2038
2039                sig->group_stop_count = 0;
2040
2041                if (task_set_jobctl_pending(current, signr | gstop))
2042                        sig->group_stop_count++;
2043
2044                for (t = next_thread(current); t != current;
2045                     t = next_thread(t)) {
2046                        /*
2047                         * Setting state to TASK_STOPPED for a group
2048                         * stop is always done with the siglock held,
2049                         * so this check has no races.
2050                         */
2051                        if (!task_is_stopped(t) &&
2052                            task_set_jobctl_pending(t, signr | gstop)) {
2053                                sig->group_stop_count++;
2054                                if (likely(!(t->ptrace & PT_SEIZED)))
2055                                        signal_wake_up(t, 0);
2056                                else
2057                                        ptrace_trap_notify(t);
2058                        }
2059                }
2060        }
2061
2062        if (likely(!current->ptrace)) {
2063                int notify = 0;
2064
2065                /*
2066                 * If there are no other threads in the group, or if there
2067                 * is a group stop in progress and we are the last to stop,
2068                 * report to the parent.
2069                 */
2070                if (task_participate_group_stop(current))
2071                        notify = CLD_STOPPED;
2072
2073                __set_current_state(TASK_STOPPED);
2074                spin_unlock_irq(&current->sighand->siglock);
2075
2076                /*
2077                 * Notify the parent of the group stop completion.  Because
2078                 * we're not holding either the siglock or tasklist_lock
2079                 * here, ptracer may attach inbetween; however, this is for
2080                 * group stop and should always be delivered to the real
2081                 * parent of the group leader.  The new ptracer will get
2082                 * its notification when this task transitions into
2083                 * TASK_TRACED.
2084                 */
2085                if (notify) {
2086                        read_lock(&tasklist_lock);
2087                        do_notify_parent_cldstop(current, false, notify);
2088                        read_unlock(&tasklist_lock);
2089                }
2090
2091                /* Now we don't run again until woken by SIGCONT or SIGKILL */
2092                freezable_schedule();
2093                return true;
2094        } else {
2095                /*
2096                 * While ptraced, group stop is handled by STOP trap.
2097                 * Schedule it and let the caller deal with it.
2098                 */
2099                task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2100                return false;
2101        }
2102}
2103
2104/**
2105 * do_jobctl_trap - take care of ptrace jobctl traps
2106 *
2107 * When PT_SEIZED, it's used for both group stop and explicit
2108 * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2109 * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2110 * the stop signal; otherwise, %SIGTRAP.
2111 *
2112 * When !PT_SEIZED, it's used only for group stop trap with stop signal
2113 * number as exit_code and no siginfo.
2114 *
2115 * CONTEXT:
2116 * Must be called with @current->sighand->siglock held, which may be
2117 * released and re-acquired before returning with intervening sleep.
2118 */
2119static void do_jobctl_trap(void)
2120{
2121        struct signal_struct *signal = current->signal;
2122        int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2123
2124        if (current->ptrace & PT_SEIZED) {
2125                if (!signal->group_stop_count &&
2126                    !(signal->flags & SIGNAL_STOP_STOPPED))
2127                        signr = SIGTRAP;
2128                WARN_ON_ONCE(!signr);
2129                ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2130                                 CLD_STOPPED);
2131        } else {
2132                WARN_ON_ONCE(!signr);
2133                ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2134                current->exit_code = 0;
2135        }
2136}
2137
2138static int ptrace_signal(int signr, siginfo_t *info)
2139{
2140        ptrace_signal_deliver();
2141        /*
2142         * We do not check sig_kernel_stop(signr) but set this marker
2143         * unconditionally because we do not know whether debugger will
2144         * change signr. This flag has no meaning unless we are going
2145         * to stop after return from ptrace_stop(). In this case it will
2146         * be checked in do_signal_stop(), we should only stop if it was
2147         * not cleared by SIGCONT while we were sleeping. See also the
2148         * comment in dequeue_signal().
2149         */
2150        current->jobctl |= JOBCTL_STOP_DEQUEUED;
2151        ptrace_stop(signr, CLD_TRAPPED, 0, info);
2152
2153        /* We're back.  Did the debugger cancel the sig?  */
2154        signr = current->exit_code;
2155        if (signr == 0)
2156                return signr;
2157
2158        current->exit_code = 0;
2159
2160        /*
2161         * Update the siginfo structure if the signal has
2162         * changed.  If the debugger wanted something
2163         * specific in the siginfo structure then it should
2164         * have updated *info via PTRACE_SETSIGINFO.
2165         */
2166        if (signr != info->si_signo) {
2167                info->si_signo = signr;
2168                info->si_errno = 0;
2169                info->si_code = SI_USER;
2170                rcu_read_lock();
2171                info->si_pid = task_pid_vnr(current->parent);
2172                info->si_uid = from_kuid_munged(current_user_ns(),
2173                                                task_uid(current->parent));
2174                rcu_read_unlock();
2175        }
2176
2177        /* If the (new) signal is now blocked, requeue it.  */
2178        if (sigismember(&current->blocked, signr)) {
2179                specific_send_sig_info(signr, info, current);
2180                signr = 0;
2181        }
2182
2183        return signr;
2184}
2185
2186int get_signal_to_deliver(siginfo_t *info, struct k_sigaction *return_ka,
2187                          struct pt_regs *regs, void *cookie)
2188{
2189        struct sighand_struct *sighand = current->sighand;
2190        struct signal_struct *signal = current->signal;
2191        int signr;
2192
2193        if (unlikely(current->task_works))
2194                task_work_run();
2195
2196        if (unlikely(uprobe_deny_signal()))
2197                return 0;
2198
2199        /*
2200         * Do this once, we can't return to user-mode if freezing() == T.
2201         * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2202         * thus do not need another check after return.
2203         */
2204        try_to_freeze();
2205
2206relock:
2207        spin_lock_irq(&sighand->siglock);
2208        /*
2209         * Every stopped thread goes here after wakeup. Check to see if
2210         * we should notify the parent, prepare_signal(SIGCONT) encodes
2211         * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2212         */
2213        if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2214                int why;
2215
2216                if (signal->flags & SIGNAL_CLD_CONTINUED)
2217                        why = CLD_CONTINUED;
2218                else
2219                        why = CLD_STOPPED;
2220
2221                signal->flags &= ~SIGNAL_CLD_MASK;
2222
2223                spin_unlock_irq(&sighand->siglock);
2224
2225                /*
2226                 * Notify the parent that we're continuing.  This event is
2227                 * always per-process and doesn't make whole lot of sense
2228                 * for ptracers, who shouldn't consume the state via
2229                 * wait(2) either, but, for backward compatibility, notify
2230                 * the ptracer of the group leader too unless it's gonna be
2231                 * a duplicate.
2232                 */
2233                read_lock(&tasklist_lock);
2234                do_notify_parent_cldstop(current, false, why);
2235
2236                if (ptrace_reparented(current->group_leader))
2237                        do_notify_parent_cldstop(current->group_leader,
2238                                                true, why);
2239                read_unlock(&tasklist_lock);
2240
2241                goto relock;
2242        }
2243
2244        for (;;) {
2245                struct k_sigaction *ka;
2246
2247                if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2248                    do_signal_stop(0))
2249                        goto relock;
2250
2251                if (unlikely(current->jobctl & JOBCTL_TRAP_MASK)) {
2252                        do_jobctl_trap();
2253                        spin_unlock_irq(&sighand->siglock);
2254                        goto relock;
2255                }
2256
2257                signr = dequeue_signal(current, &current->blocked, info);
2258
2259                if (!signr)
2260                        break; /* will return 0 */
2261
2262                if (unlikely(current->ptrace) && signr != SIGKILL) {
2263                        signr = ptrace_signal(signr, info);
2264                        if (!signr)
2265                                continue;
2266                }
2267
2268                ka = &sighand->action[signr-1];
2269
2270                /* Trace actually delivered signals. */
2271                trace_signal_deliver(signr, info, ka);
2272
2273                if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2274                        continue;
2275                if (ka->sa.sa_handler != SIG_DFL) {
2276                        /* Run the handler.  */
2277                        *return_ka = *ka;
2278
2279                        if (ka->sa.sa_flags & SA_ONESHOT)
2280                                ka->sa.sa_handler = SIG_DFL;
2281
2282                        break; /* will return non-zero "signr" value */
2283                }
2284
2285                /*
2286                 * Now we are doing the default action for this signal.
2287                 */
2288                if (sig_kernel_ignore(signr)) /* Default is nothing. */
2289                        continue;
2290
2291                /*
2292                 * Global init gets no signals it doesn't want.
2293                 * Container-init gets no signals it doesn't want from same
2294                 * container.
2295                 *
2296                 * Note that if global/container-init sees a sig_kernel_only()
2297                 * signal here, the signal must have been generated internally
2298                 * or must have come from an ancestor namespace. In either
2299                 * case, the signal cannot be dropped.
2300                 */
2301                if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2302                                !sig_kernel_only(signr))
2303                        continue;
2304
2305                if (sig_kernel_stop(signr)) {
2306                        /*
2307                         * The default action is to stop all threads in
2308                         * the thread group.  The job control signals
2309                         * do nothing in an orphaned pgrp, but SIGSTOP
2310                         * always works.  Note that siglock needs to be
2311                         * dropped during the call to is_orphaned_pgrp()
2312                         * because of lock ordering with tasklist_lock.
2313                         * This allows an intervening SIGCONT to be posted.
2314                         * We need to check for that and bail out if necessary.
2315                         */
2316                        if (signr != SIGSTOP) {
2317                                spin_unlock_irq(&sighand->siglock);
2318
2319                                /* signals can be posted during this window */
2320
2321                                if (is_current_pgrp_orphaned())
2322                                        goto relock;
2323
2324                                spin_lock_irq(&sighand->siglock);
2325                        }
2326
2327                        if (likely(do_signal_stop(info->si_signo))) {
2328                                /* It released the siglock.  */
2329                                goto relock;
2330                        }
2331
2332                        /*
2333                         * We didn't actually stop, due to a race
2334                         * with SIGCONT or something like that.
2335                         */
2336                        continue;
2337                }
2338
2339                spin_unlock_irq(&sighand->siglock);
2340
2341                /*
2342                 * Anything else is fatal, maybe with a core dump.
2343                 */
2344                current->flags |= PF_SIGNALED;
2345
2346                if (sig_kernel_coredump(signr)) {
2347                        if (print_fatal_signals)
2348                                print_fatal_signal(info->si_signo);
2349                        /*
2350                         * If it was able to dump core, this kills all
2351                         * other threads in the group and synchronizes with
2352                         * their demise.  If we lost the race with another
2353                         * thread getting here, it set group_exit_code
2354                         * first and our do_group_exit call below will use
2355                         * that value and ignore the one we pass it.
2356                         */
2357                        do_coredump(info);
2358                }
2359
2360                /*
2361                 * Death signals, no core dump.
2362                 */
2363                do_group_exit(info->si_signo);
2364                /* NOTREACHED */
2365        }
2366        spin_unlock_irq(&sighand->siglock);
2367        return signr;
2368}
2369
2370/**
2371 * signal_delivered - 
2372 * @sig:                number of signal being delivered
2373 * @info:               siginfo_t of signal being delivered
2374 * @ka:                 sigaction setting that chose the handler
2375 * @regs:               user register state
2376 * @stepping:           nonzero if debugger single-step or block-step in use
2377 *
2378 * This function should be called when a signal has succesfully been
2379 * delivered. It updates the blocked signals accordingly (@ka->sa.sa_mask
2380 * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2381 * is set in @ka->sa.sa_flags.  Tracing is notified.
2382 */
2383void signal_delivered(int sig, siginfo_t *info, struct k_sigaction *ka,
2384                        struct pt_regs *regs, int stepping)
2385{
2386        sigset_t blocked;
2387
2388        /* A signal was successfully delivered, and the
2389           saved sigmask was stored on the signal frame,
2390           and will be restored by sigreturn.  So we can
2391           simply clear the restore sigmask flag.  */
2392        clear_restore_sigmask();
2393
2394        sigorsets(&blocked, &current->blocked, &ka->sa.sa_mask);
2395        if (!(ka->sa.sa_flags & SA_NODEFER))
2396                sigaddset(&blocked, sig);
2397        set_current_blocked(&blocked);
2398        tracehook_signal_handler(sig, info, ka, regs, stepping);
2399}
2400
2401/*
2402 * It could be that complete_signal() picked us to notify about the
2403 * group-wide signal. Other threads should be notified now to take
2404 * the shared signals in @which since we will not.
2405 */
2406static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2407{
2408        sigset_t retarget;
2409        struct task_struct *t;
2410
2411        sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2412        if (sigisemptyset(&retarget))
2413                return;
2414
2415        t = tsk;
2416        while_each_thread(tsk, t) {
2417                if (t->flags & PF_EXITING)
2418                        continue;
2419
2420                if (!has_pending_signals(&retarget, &t->blocked))
2421                        continue;
2422                /* Remove the signals this thread can handle. */
2423                sigandsets(&retarget, &retarget, &t->blocked);
2424
2425                if (!signal_pending(t))
2426                        signal_wake_up(t, 0);
2427
2428                if (sigisemptyset(&retarget))
2429                        break;
2430        }
2431}
2432
2433void exit_signals(struct task_struct *tsk)
2434{
2435        int group_stop = 0;
2436        sigset_t unblocked;
2437
2438        /*
2439         * @tsk is about to have PF_EXITING set - lock out users which
2440         * expect stable threadgroup.
2441         */
2442        threadgroup_change_begin(tsk);
2443
2444        if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2445                tsk->flags |= PF_EXITING;
2446                threadgroup_change_end(tsk);
2447                return;
2448        }
2449
2450        spin_lock_irq(&tsk->sighand->siglock);
2451        /*
2452         * From now this task is not visible for group-wide signals,
2453         * see wants_signal(), do_signal_stop().
2454         */
2455        tsk->flags |= PF_EXITING;
2456
2457        threadgroup_change_end(tsk);
2458
2459        if (!signal_pending(tsk))
2460                goto out;
2461
2462        unblocked = tsk->blocked;
2463        signotset(&unblocked);
2464        retarget_shared_pending(tsk, &unblocked);
2465
2466        if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2467            task_participate_group_stop(tsk))
2468                group_stop = CLD_STOPPED;
2469out:
2470        spin_unlock_irq(&tsk->sighand->siglock);
2471
2472        /*
2473         * If group stop has completed, deliver the notification.  This
2474         * should always go to the real parent of the group leader.
2475         */
2476        if (unlikely(group_stop)) {
2477                read_lock(&tasklist_lock);
2478                do_notify_parent_cldstop(tsk, false, group_stop);
2479                read_unlock(&tasklist_lock);
2480        }
2481}
2482
2483EXPORT_SYMBOL(recalc_sigpending);
2484EXPORT_SYMBOL_GPL(dequeue_signal);
2485EXPORT_SYMBOL(flush_signals);
2486EXPORT_SYMBOL(force_sig);
2487EXPORT_SYMBOL(send_sig);
2488EXPORT_SYMBOL(send_sig_info);
2489EXPORT_SYMBOL(sigprocmask);
2490EXPORT_SYMBOL(block_all_signals);
2491EXPORT_SYMBOL(unblock_all_signals);
2492
2493
2494/*
2495 * System call entry points.
2496 */
2497
2498/**
2499 *  sys_restart_syscall - restart a system call
2500 */
2501SYSCALL_DEFINE0(restart_syscall)
2502{
2503        struct restart_block *restart = &current_thread_info()->restart_block;
2504        return restart->fn(restart);
2505}
2506
2507long do_no_restart_syscall(struct restart_block *param)
2508{
2509        return -EINTR;
2510}
2511
2512static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2513{
2514        if (signal_pending(tsk) && !thread_group_empty(tsk)) {
2515                sigset_t newblocked;
2516                /* A set of now blocked but previously unblocked signals. */
2517                sigandnsets(&newblocked, newset, &current->blocked);
2518                retarget_shared_pending(tsk, &newblocked);
2519        }
2520        tsk->blocked = *newset;
2521        recalc_sigpending();
2522}
2523
2524/**
2525 * set_current_blocked - change current->blocked mask
2526 * @newset: new mask
2527 *
2528 * It is wrong to change ->blocked directly, this helper should be used
2529 * to ensure the process can't miss a shared signal we are going to block.
2530 */
2531void set_current_blocked(sigset_t *newset)
2532{
2533        sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2534        __set_current_blocked(newset);
2535}
2536
2537void __set_current_blocked(const sigset_t *newset)
2538{
2539        struct task_struct *tsk = current;
2540
2541        spin_lock_irq(&tsk->sighand->siglock);
2542        __set_task_blocked(tsk, newset);
2543        spin_unlock_irq(&tsk->sighand->siglock);
2544}
2545
2546/*
2547 * This is also useful for kernel threads that want to temporarily
2548 * (or permanently) block certain signals.
2549 *
2550 * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2551 * interface happily blocks "unblockable" signals like SIGKILL
2552 * and friends.
2553 */
2554int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2555{
2556        struct task_struct *tsk = current;
2557        sigset_t newset;
2558
2559        /* Lockless, only current can change ->blocked, never from irq */
2560        if (oldset)
2561                *oldset = tsk->blocked;
2562
2563        switch (how) {
2564        case SIG_BLOCK:
2565                sigorsets(&newset, &tsk->blocked, set);
2566                break;
2567        case SIG_UNBLOCK:
2568                sigandnsets(&newset, &tsk->blocked, set);
2569                break;
2570        case SIG_SETMASK:
2571                newset = *set;
2572                break;
2573        default:
2574                return -EINVAL;
2575        }
2576
2577        __set_current_blocked(&newset);
2578        return 0;
2579}
2580
2581/**
2582 *  sys_rt_sigprocmask - change the list of currently blocked signals
2583 *  @how: whether to add, remove, or set signals
2584 *  @nset: stores pending signals
2585 *  @oset: previous value of signal mask if non-null
2586 *  @sigsetsize: size of sigset_t type
2587 */
2588SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
2589                sigset_t __user *, oset, size_t, sigsetsize)
2590{
2591        sigset_t old_set, new_set;
2592        int error;
2593
2594        /* XXX: Don't preclude handling different sized sigset_t's.  */
2595        if (sigsetsize != sizeof(sigset_t))
2596                return -EINVAL;
2597
2598        old_set = current->blocked;
2599
2600        if (nset) {
2601                if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
2602                        return -EFAULT;
2603                sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2604
2605                error = sigprocmask(how, &new_set, NULL);
2606                if (error)
2607                        return error;
2608        }
2609
2610        if (oset) {
2611                if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
2612                        return -EFAULT;
2613        }
2614
2615        return 0;
2616}
2617
2618long do_sigpending(void __user *set, unsigned long sigsetsize)
2619{
2620        long error = -EINVAL;
2621        sigset_t pending;
2622
2623        if (sigsetsize > sizeof(sigset_t))
2624                goto out;
2625
2626        spin_lock_irq(&current->sighand->siglock);
2627        sigorsets(&pending, &current->pending.signal,
2628                  &current->signal->shared_pending.signal);
2629        spin_unlock_irq(&current->sighand->siglock);
2630
2631        /* Outside the lock because only this thread touches it.  */
2632        sigandsets(&pending, &current->blocked, &pending);
2633
2634        error = -EFAULT;
2635        if (!copy_to_user(set, &pending, sigsetsize))
2636                error = 0;
2637
2638out:
2639        return error;
2640}
2641
2642/**
2643 *  sys_rt_sigpending - examine a pending signal that has been raised
2644 *                      while blocked
2645 *  @set: stores pending signals
2646 *  @sigsetsize: size of sigset_t type or larger
2647 */
2648SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, set, size_t, sigsetsize)
2649{
2650        return do_sigpending(set, sigsetsize);
2651}
2652
2653#ifndef HAVE_ARCH_COPY_SIGINFO_TO_USER
2654
2655int copy_siginfo_to_user(siginfo_t __user *to, siginfo_t *from)
2656{
2657        int err;
2658
2659        if (!access_ok (VERIFY_WRITE, to, sizeof(siginfo_t)))
2660                return -EFAULT;
2661        if (from->si_code < 0)
2662                return __copy_to_user(to, from, sizeof(siginfo_t))
2663                        ? -EFAULT : 0;
2664        /*
2665         * If you change siginfo_t structure, please be sure
2666         * this code is fixed accordingly.
2667         * Please remember to update the signalfd_copyinfo() function
2668         * inside fs/signalfd.c too, in case siginfo_t changes.
2669         * It should never copy any pad contained in the structure
2670         * to avoid security leaks, but must copy the generic
2671         * 3 ints plus the relevant union member.
2672         */
2673        err = __put_user(from->si_signo, &to->si_signo);
2674        err |= __put_user(from->si_errno, &to->si_errno);
2675        err |= __put_user((short)from->si_code, &to->si_code);
2676        switch (from->si_code & __SI_MASK) {
2677        case __SI_KILL:
2678                err |= __put_user(from->si_pid, &to->si_pid);
2679                err |= __put_user(from->si_uid, &to->si_uid);
2680                break;
2681        case __SI_TIMER:
2682                 err |= __put_user(from->si_tid, &to->si_tid);
2683                 err |= __put_user(from->si_overrun, &to->si_overrun);
2684                 err |= __put_user(from->si_ptr, &to->si_ptr);
2685                break;
2686        case __SI_POLL:
2687                err |= __put_user(from->si_band, &to->si_band);
2688                err |= __put_user(from->si_fd, &to->si_fd);
2689                break;
2690        case __SI_FAULT:
2691                err |= __put_user(from->si_addr, &to->si_addr);
2692#ifdef __ARCH_SI_TRAPNO
2693                err |= __put_user(from->si_trapno, &to->si_trapno);
2694#endif
2695#ifdef BUS_MCEERR_AO
2696                /*
2697                 * Other callers might not initialize the si_lsb field,
2698                 * so check explicitly for the right codes here.
2699                 */
2700                if (from->si_code == BUS_MCEERR_AR || from->si_code == BUS_MCEERR_AO)
2701                        err |= __put_user(from->si_addr_lsb, &to->si_addr_lsb);
2702#endif
2703                break;
2704        case __SI_CHLD:
2705                err |= __put_user(from->si_pid, &to->si_pid);
2706                err |= __put_user(from->si_uid, &to->si_uid);
2707                err |= __put_user(from->si_status, &to->si_status);
2708                err |= __put_user(from->si_utime, &to->si_utime);
2709                err |= __put_user(from->si_stime, &to->si_stime);
2710                break;
2711        case __SI_RT: /* This is not generated by the kernel as of now. */
2712        case __SI_MESGQ: /* But this is */
2713                err |= __put_user(from->si_pid, &to->si_pid);
2714                err |= __put_user(from->si_uid, &to->si_uid);
2715                err |= __put_user(from->si_ptr, &to->si_ptr);
2716                break;
2717#ifdef __ARCH_SIGSYS
2718        case __SI_SYS:
2719                err |= __put_user(from->si_call_addr, &to->si_call_addr);
2720                err |= __put_user(from->si_syscall, &to->si_syscall);
2721                err |= __put_user(from->si_arch, &to->si_arch);
2722                break;
2723#endif
2724        default: /* this is just in case for now ... */
2725                err |= __put_user(from->si_pid, &to->si_pid);
2726                err |= __put_user(from->si_uid, &to->si_uid);
2727                break;
2728        }
2729        return err;
2730}
2731
2732#endif
2733
2734/**
2735 *  do_sigtimedwait - wait for queued signals specified in @which
2736 *  @which: queued signals to wait for
2737 *  @info: if non-null, the signal's siginfo is returned here
2738 *  @ts: upper bound on process time suspension
2739 */
2740int do_sigtimedwait(const sigset_t *which, siginfo_t *info,
2741                        const struct timespec *ts)
2742{
2743        struct task_struct *tsk = current;
2744        long timeout = MAX_SCHEDULE_TIMEOUT;
2745        sigset_t mask = *which;
2746        int sig;
2747
2748        if (ts) {
2749                if (!timespec_valid(ts))
2750                        return -EINVAL;
2751                timeout = timespec_to_jiffies(ts);
2752                /*
2753                 * We can be close to the next tick, add another one
2754                 * to ensure we will wait at least the time asked for.
2755                 */
2756                if (ts->tv_sec || ts->tv_nsec)
2757                        timeout++;
2758        }
2759
2760        /*
2761         * Invert the set of allowed signals to get those we want to block.
2762         */
2763        sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
2764        signotset(&mask);
2765
2766        spin_lock_irq(&tsk->sighand->siglock);
2767        sig = dequeue_signal(tsk, &mask, info);
2768        if (!sig && timeout) {
2769                /*
2770                 * None ready, temporarily unblock those we're interested
2771                 * while we are sleeping in so that we'll be awakened when
2772                 * they arrive. Unblocking is always fine, we can avoid
2773                 * set_current_blocked().
2774                 */
2775                tsk->real_blocked = tsk->blocked;
2776                sigandsets(&tsk->blocked, &tsk->blocked, &mask);
2777                recalc_sigpending();
2778                spin_unlock_irq(&tsk->sighand->siglock);
2779
2780                timeout = schedule_timeout_interruptible(timeout);
2781
2782                spin_lock_irq(&tsk->sighand->siglock);
2783                __set_task_blocked(tsk, &tsk->real_blocked);
2784                siginitset(&tsk->real_blocked, 0);
2785                sig = dequeue_signal(tsk, &mask, info);
2786        }
2787        spin_unlock_irq(&tsk->sighand->siglock);
2788
2789        if (sig)
2790                return sig;
2791        return timeout ? -EINTR : -EAGAIN;
2792}
2793
2794/**
2795 *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
2796 *                      in @uthese
2797 *  @uthese: queued signals to wait for
2798 *  @uinfo: if non-null, the signal's siginfo is returned here
2799 *  @uts: upper bound on process time suspension
2800 *  @sigsetsize: size of sigset_t type
2801 */
2802SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
2803                siginfo_t __user *, uinfo, const struct timespec __user *, uts,
2804                size_t, sigsetsize)
2805{
2806        sigset_t these;
2807        struct timespec ts;
2808        siginfo_t info;
2809        int ret;
2810
2811        /* XXX: Don't preclude handling different sized sigset_t's.  */
2812        if (sigsetsize != sizeof(sigset_t))
2813                return -EINVAL;
2814
2815        if (copy_from_user(&these, uthese, sizeof(these)))
2816                return -EFAULT;
2817
2818        if (uts) {
2819                if (copy_from_user(&ts, uts, sizeof(ts)))
2820                        return -EFAULT;
2821        }
2822
2823        ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
2824
2825        if (ret > 0 && uinfo) {
2826                if (copy_siginfo_to_user(uinfo, &info))
2827                        ret = -EFAULT;
2828        }
2829
2830        return ret;
2831}
2832
2833/**
2834 *  sys_kill - send a signal to a process
2835 *  @pid: the PID of the process
2836 *  @sig: signal to be sent
2837 */
2838SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
2839{
2840        struct siginfo info;
2841
2842        info.si_signo = sig;
2843        info.si_errno = 0;
2844        info.si_code = SI_USER;
2845        info.si_pid = task_tgid_vnr(current);
2846        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2847
2848        return kill_something_info(sig, &info, pid);
2849}
2850
2851static int
2852do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info)
2853{
2854        struct task_struct *p;
2855        int error = -ESRCH;
2856
2857        rcu_read_lock();
2858        p = find_task_by_vpid(pid);
2859        if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
2860                error = check_kill_permission(sig, info, p);
2861                /*
2862                 * The null signal is a permissions and process existence
2863                 * probe.  No signal is actually delivered.
2864                 */
2865                if (!error && sig) {
2866                        error = do_send_sig_info(sig, info, p, false);
2867                        /*
2868                         * If lock_task_sighand() failed we pretend the task
2869                         * dies after receiving the signal. The window is tiny,
2870                         * and the signal is private anyway.
2871                         */
2872                        if (unlikely(error == -ESRCH))
2873                                error = 0;
2874                }
2875        }
2876        rcu_read_unlock();
2877
2878        return error;
2879}
2880
2881static int do_tkill(pid_t tgid, pid_t pid, int sig)
2882{
2883        struct siginfo info;
2884
2885        info.si_signo = sig;
2886        info.si_errno = 0;
2887        info.si_code = SI_TKILL;
2888        info.si_pid = task_tgid_vnr(current);
2889        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2890
2891        return do_send_specific(tgid, pid, sig, &info);
2892}
2893
2894/**
2895 *  sys_tgkill - send signal to one specific thread
2896 *  @tgid: the thread group ID of the thread
2897 *  @pid: the PID of the thread
2898 *  @sig: signal to be sent
2899 *
2900 *  This syscall also checks the @tgid and returns -ESRCH even if the PID
2901 *  exists but it's not belonging to the target process anymore. This
2902 *  method solves the problem of threads exiting and PIDs getting reused.
2903 */
2904SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
2905{
2906        /* This is only valid for single tasks */
2907        if (pid <= 0 || tgid <= 0)
2908                return -EINVAL;
2909
2910        return do_tkill(tgid, pid, sig);
2911}
2912
2913/**
2914 *  sys_tkill - send signal to one specific task
2915 *  @pid: the PID of the task
2916 *  @sig: signal to be sent
2917 *
2918 *  Send a signal to only one task, even if it's a CLONE_THREAD task.
2919 */
2920SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
2921{
2922        /* This is only valid for single tasks */
2923        if (pid <= 0)
2924                return -EINVAL;
2925
2926        return do_tkill(0, pid, sig);
2927}
2928
2929/**
2930 *  sys_rt_sigqueueinfo - send signal information to a signal
2931 *  @pid: the PID of the thread
2932 *  @sig: signal to be sent
2933 *  @uinfo: signal info to be sent
2934 */
2935SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
2936                siginfo_t __user *, uinfo)
2937{
2938        siginfo_t info;
2939
2940        if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
2941                return -EFAULT;
2942
2943        /* Not even root can pretend to send signals from the kernel.
2944         * Nor can they impersonate a kill()/tgkill(), which adds source info.
2945         */
2946        if (info.si_code >= 0 || info.si_code == SI_TKILL) {
2947                /* We used to allow any < 0 si_code */
2948                WARN_ON_ONCE(info.si_code < 0);
2949                return -EPERM;
2950        }
2951        info.si_signo = sig;
2952
2953        /* POSIX.1b doesn't mention process groups.  */
2954        return kill_proc_info(sig, &info, pid);
2955}
2956
2957long do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, siginfo_t *info)
2958{
2959        /* This is only valid for single tasks */
2960        if (pid <= 0 || tgid <= 0)
2961                return -EINVAL;
2962
2963        /* Not even root can pretend to send signals from the kernel.
2964         * Nor can they impersonate a kill()/tgkill(), which adds source info.
2965         */
2966        if (info->si_code >= 0 || info->si_code == SI_TKILL) {
2967                /* We used to allow any < 0 si_code */
2968                WARN_ON_ONCE(info->si_code < 0);
2969                return -EPERM;
2970        }
2971        info->si_signo = sig;
2972
2973        return do_send_specific(tgid, pid, sig, info);
2974}
2975
2976SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
2977                siginfo_t __user *, uinfo)
2978{
2979        siginfo_t info;
2980
2981        if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
2982                return -EFAULT;
2983
2984        return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
2985}
2986
2987int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
2988{
2989        struct task_struct *t = current;
2990        struct k_sigaction *k;
2991        sigset_t mask;
2992
2993        if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
2994                return -EINVAL;
2995
2996        k = &t->sighand->action[sig-1];
2997
2998        spin_lock_irq(&current->sighand->siglock);
2999        if (oact)
3000                *oact = *k;
3001
3002        if (act) {
3003                sigdelsetmask(&act->sa.sa_mask,
3004                              sigmask(SIGKILL) | sigmask(SIGSTOP));
3005                *k = *act;
3006                /*
3007                 * POSIX 3.3.1.3:
3008                 *  "Setting a signal action to SIG_IGN for a signal that is
3009                 *   pending shall cause the pending signal to be discarded,
3010                 *   whether or not it is blocked."
3011                 *
3012                 *  "Setting a signal action to SIG_DFL for a signal that is
3013                 *   pending and whose default action is to ignore the signal
3014                 *   (for example, SIGCHLD), shall cause the pending signal to
3015                 *   be discarded, whether or not it is blocked"
3016                 */
3017                if (sig_handler_ignored(sig_handler(t, sig), sig)) {
3018                        sigemptyset(&mask);
3019                        sigaddset(&mask, sig);
3020                        rm_from_queue_full(&mask, &t->signal->shared_pending);
3021                        do {
3022                                rm_from_queue_full(&mask, &t->pending);
3023                                t = next_thread(t);
3024                        } while (t != current);
3025                }
3026        }
3027
3028        spin_unlock_irq(&current->sighand->siglock);
3029        return 0;
3030}
3031
3032int 
3033do_sigaltstack (const stack_t __user *uss, stack_t __user *uoss, unsigned long sp)
3034{
3035        stack_t oss;
3036        int error;
3037
3038        oss.ss_sp = (void __user *) current->sas_ss_sp;
3039        oss.ss_size = current->sas_ss_size;
3040        oss.ss_flags = sas_ss_flags(sp);
3041
3042        if (uss) {
3043                void __user *ss_sp;
3044                size_t ss_size;
3045                int ss_flags;
3046
3047                error = -EFAULT;
3048                if (!access_ok(VERIFY_READ, uss, sizeof(*uss)))
3049                        goto out;
3050                error = __get_user(ss_sp, &uss->ss_sp) |
3051                        __get_user(ss_flags, &uss->ss_flags) |
3052                        __get_user(ss_size, &uss->ss_size);
3053                if (error)
3054                        goto out;
3055
3056                error = -EPERM;
3057                if (on_sig_stack(sp))
3058                        goto out;
3059
3060                error = -EINVAL;
3061                /*
3062                 * Note - this code used to test ss_flags incorrectly:
3063                 *        old code may have been written using ss_flags==0
3064                 *        to mean ss_flags==SS_ONSTACK (as this was the only
3065                 *        way that worked) - this fix preserves that older
3066                 *        mechanism.
3067                 */
3068                if (ss_flags != SS_DISABLE && ss_flags != SS_ONSTACK && ss_flags != 0)
3069                        goto out;
3070
3071                if (ss_flags == SS_DISABLE) {
3072                        ss_size = 0;
3073                        ss_sp = NULL;
3074                } else {
3075                        error = -ENOMEM;
3076                        if (ss_size < MINSIGSTKSZ)
3077                                goto out;
3078                }
3079
3080                current->sas_ss_sp = (unsigned long) ss_sp;
3081                current->sas_ss_size = ss_size;
3082        }
3083
3084        error = 0;
3085        if (uoss) {
3086                error = -EFAULT;
3087                if (!access_ok(VERIFY_WRITE, uoss, sizeof(*uoss)))
3088                        goto out;
3089                error = __put_user(oss.ss_sp, &uoss->ss_sp) |
3090                        __put_user(oss.ss_size, &uoss->ss_size) |
3091                        __put_user(oss.ss_flags, &uoss->ss_flags);
3092        }
3093
3094out:
3095        return error;
3096}
3097#ifdef CONFIG_GENERIC_SIGALTSTACK
3098SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
3099{
3100        return do_sigaltstack(uss, uoss, current_user_stack_pointer());
3101}
3102#endif
3103
3104int restore_altstack(const stack_t __user *uss)
3105{
3106        int err = do_sigaltstack(uss, NULL, current_user_stack_pointer());
3107        /* squash all but EFAULT for now */
3108        return err == -EFAULT ? err : 0;
3109}
3110
3111int __save_altstack(stack_t __user *uss, unsigned long sp)
3112{
3113        struct task_struct *t = current;
3114        return  __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
3115                __put_user(sas_ss_flags(sp), &uss->ss_flags) |
3116                __put_user(t->sas_ss_size, &uss->ss_size);
3117}
3118
3119#ifdef CONFIG_COMPAT
3120#ifdef CONFIG_GENERIC_SIGALTSTACK
3121COMPAT_SYSCALL_DEFINE2(sigaltstack,
3122                        const compat_stack_t __user *, uss_ptr,
3123                        compat_stack_t __user *, uoss_ptr)
3124{
3125        stack_t uss, uoss;
3126        int ret;
3127        mm_segment_t seg;
3128
3129        if (uss_ptr) {
3130                compat_stack_t uss32;
3131
3132                memset(&uss, 0, sizeof(stack_t));
3133                if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
3134                        return -EFAULT;
3135                uss.ss_sp = compat_ptr(uss32.ss_sp);
3136                uss.ss_flags = uss32.ss_flags;
3137                uss.ss_size = uss32.ss_size;
3138        }
3139        seg = get_fs();
3140        set_fs(KERNEL_DS);
3141        ret = do_sigaltstack((stack_t __force __user *) (uss_ptr ? &uss : NULL),
3142                             (stack_t __force __user *) &uoss,
3143                             compat_user_stack_pointer());
3144        set_fs(seg);
3145        if (ret >= 0 && uoss_ptr)  {
3146                if (!access_ok(VERIFY_WRITE, uoss_ptr, sizeof(compat_stack_t)) ||
3147                    __put_user(ptr_to_compat(uoss.ss_sp), &uoss_ptr->ss_sp) ||
3148                    __put_user(uoss.ss_flags, &uoss_ptr->ss_flags) ||
3149                    __put_user(uoss.ss_size, &uoss_ptr->ss_size))
3150                        ret = -EFAULT;
3151        }
3152        return ret;
3153}
3154
3155int compat_restore_altstack(const compat_stack_t __user *uss)
3156{
3157        int err = compat_sys_sigaltstack(uss, NULL);
3158        /* squash all but -EFAULT for now */
3159        return err == -EFAULT ? err : 0;
3160}
3161
3162int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
3163{
3164        struct task_struct *t = current;
3165        return  __put_user(ptr_to_compat((void __user *)t->sas_ss_sp), &uss->ss_sp) |
3166                __put_user(sas_ss_flags(sp), &uss->ss_flags) |
3167                __put_user(t->sas_ss_size, &uss->ss_size);
3168}
3169#endif
3170#endif
3171
3172#ifdef __ARCH_WANT_SYS_SIGPENDING
3173
3174/**
3175 *  sys_sigpending - examine pending signals
3176 *  @set: where mask of pending signal is returned
3177 */
3178SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, set)
3179{
3180        return do_sigpending(set, sizeof(*set));
3181}
3182
3183#endif
3184
3185#ifdef __ARCH_WANT_SYS_SIGPROCMASK
3186/**
3187 *  sys_sigprocmask - examine and change blocked signals
3188 *  @how: whether to add, remove, or set signals
3189 *  @nset: signals to add or remove (if non-null)
3190 *  @oset: previous value of signal mask if non-null
3191 *
3192 * Some platforms have their own version with special arguments;
3193 * others support only sys_rt_sigprocmask.
3194 */
3195
3196SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
3197                old_sigset_t __user *, oset)
3198{
3199        old_sigset_t old_set, new_set;
3200        sigset_t new_blocked;
3201
3202        old_set = current->blocked.sig[0];
3203
3204        if (nset) {
3205                if (copy_from_user(&new_set, nset, sizeof(*nset)))
3206                        return -EFAULT;
3207
3208                new_blocked = current->blocked;
3209
3210                switch (how) {
3211                case SIG_BLOCK:
3212                        sigaddsetmask(&new_blocked, new_set);
3213                        break;
3214                case SIG_UNBLOCK:
3215                        sigdelsetmask(&new_blocked, new_set);
3216                        break;
3217                case SIG_SETMASK:
3218                        new_blocked.sig[0] = new_set;
3219                        break;
3220                default:
3221                        return -EINVAL;
3222                }
3223
3224                set_current_blocked(&new_blocked);
3225        }
3226
3227        if (oset) {
3228                if (copy_to_user(oset, &old_set, sizeof(*oset)))
3229                        return -EFAULT;
3230        }
3231
3232        return 0;
3233}
3234#endif /* __ARCH_WANT_SYS_SIGPROCMASK */
3235
3236#ifdef __ARCH_WANT_SYS_RT_SIGACTION
3237/**
3238 *  sys_rt_sigaction - alter an action taken by a process
3239 *  @sig: signal to be sent
3240 *  @act: new sigaction
3241 *  @oact: used to save the previous sigaction
3242 *  @sigsetsize: size of sigset_t type
3243 */
3244SYSCALL_DEFINE4(rt_sigaction, int, sig,
3245                const struct sigaction __user *, act,
3246                struct sigaction __user *, oact,
3247                size_t, sigsetsize)
3248{
3249        struct k_sigaction new_sa, old_sa;
3250        int ret = -EINVAL;
3251
3252        /* XXX: Don't preclude handling different sized sigset_t's.  */
3253        if (sigsetsize != sizeof(sigset_t))
3254                goto out;
3255
3256        if (act) {
3257                if (copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
3258                        return -EFAULT;
3259        }
3260
3261        ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
3262
3263        if (!ret && oact) {
3264                if (copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
3265                        return -EFAULT;
3266        }
3267out:
3268        return ret;
3269}
3270#endif /* __ARCH_WANT_SYS_RT_SIGACTION */
3271
3272#ifdef __ARCH_WANT_SYS_SGETMASK
3273
3274/*
3275 * For backwards compatibility.  Functionality superseded by sigprocmask.
3276 */
3277SYSCALL_DEFINE0(sgetmask)
3278{
3279        /* SMP safe */
3280        return current->blocked.sig[0];
3281}
3282
3283SYSCALL_DEFINE1(ssetmask, int, newmask)
3284{
3285        int old = current->blocked.sig[0];
3286        sigset_t newset;
3287
3288        siginitset(&newset, newmask);
3289        set_current_blocked(&newset);
3290
3291        return old;
3292}
3293#endif /* __ARCH_WANT_SGETMASK */
3294
3295#ifdef __ARCH_WANT_SYS_SIGNAL
3296/*
3297 * For backwards compatibility.  Functionality superseded by sigaction.
3298 */
3299SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
3300{
3301        struct k_sigaction new_sa, old_sa;
3302        int ret;
3303
3304        new_sa.sa.sa_handler = handler;
3305        new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
3306        sigemptyset(&new_sa.sa.sa_mask);
3307
3308        ret = do_sigaction(sig, &new_sa, &old_sa);
3309
3310        return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
3311}
3312#endif /* __ARCH_WANT_SYS_SIGNAL */
3313
3314#ifdef __ARCH_WANT_SYS_PAUSE
3315
3316SYSCALL_DEFINE0(pause)
3317{
3318        while (!signal_pending(current)) {
3319                current->state = TASK_INTERRUPTIBLE;
3320                schedule();
3321        }
3322        return -ERESTARTNOHAND;
3323}
3324
3325#endif
3326
3327int sigsuspend(sigset_t *set)
3328{
3329        current->saved_sigmask = current->blocked;
3330        set_current_blocked(set);
3331
3332        current->state = TASK_INTERRUPTIBLE;
3333        schedule();
3334        set_restore_sigmask();
3335        return -ERESTARTNOHAND;
3336}
3337
3338#ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
3339/**
3340 *  sys_rt_sigsuspend - replace the signal mask for a value with the
3341 *      @unewset value until a signal is received
3342 *  @unewset: new signal mask value
3343 *  @sigsetsize: size of sigset_t type
3344 */
3345SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
3346{
3347        sigset_t newset;
3348
3349        /* XXX: Don't preclude handling different sized sigset_t's.  */
3350        if (sigsetsize != sizeof(sigset_t))
3351                return -EINVAL;
3352
3353        if (copy_from_user(&newset, unewset, sizeof(newset)))
3354                return -EFAULT;
3355        return sigsuspend(&newset);
3356}
3357#endif /* __ARCH_WANT_SYS_RT_SIGSUSPEND */
3358
3359__attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma)
3360{
3361        return NULL;
3362}
3363
3364void __init signals_init(void)
3365{
3366        sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
3367}
3368
3369#ifdef CONFIG_KGDB_KDB
3370#include <linux/kdb.h>
3371/*
3372 * kdb_send_sig_info - Allows kdb to send signals without exposing
3373 * signal internals.  This function checks if the required locks are
3374 * available before calling the main signal code, to avoid kdb
3375 * deadlocks.
3376 */
3377void
3378kdb_send_sig_info(struct task_struct *t, struct siginfo *info)
3379{
3380        static struct task_struct *kdb_prev_t;
3381        int sig, new_t;
3382        if (!spin_trylock(&t->sighand->siglock)) {
3383                kdb_printf("Can't do kill command now.\n"
3384                           "The sigmask lock is held somewhere else in "
3385                           "kernel, try again later\n");
3386                return;
3387        }
3388        spin_unlock(&t->sighand->siglock);
3389        new_t = kdb_prev_t != t;
3390        kdb_prev_t = t;
3391        if (t->state != TASK_RUNNING && new_t) {
3392                kdb_printf("Process is not RUNNING, sending a signal from "
3393                           "kdb risks deadlock\n"
3394                           "on the run queue locks. "
3395                           "The signal has _not_ been sent.\n"
3396                           "Reissue the kill command if you want to risk "
3397                           "the deadlock.\n");
3398                return;
3399        }
3400        sig = info->si_signo;
3401        if (send_sig_info(sig, info, t))
3402                kdb_printf("Fail to deliver Signal %d to process %d.\n",
3403                           sig, t->pid);
3404        else
3405                kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
3406}
3407#endif  /* CONFIG_KGDB_KDB */
3408
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.