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