linux/kernel/signal.c
<<
>>
Prefs
   1/*
   2 *  linux/kernel/signal.c
   3 *
   4 *  Copyright (C) 1991, 1992  Linus Torvalds
   5 *
   6 *  1997-11-02  Modified for POSIX.1b signals by Richard Henderson
   7 *
   8 *  2003-06-02  Jim Houston - Concurrent Computer Corp.
   9 *              Changes to use preallocated sigqueue structures
  10 *              to allow signals to be sent reliably.
  11 */
  12
  13#include <linux/slab.h>
  14#include <linux/export.h>
  15#include <linux/init.h>
  16#include <linux/sched.h>
  17#include <linux/fs.h>
  18#include <linux/tty.h>
  19#include <linux/binfmts.h>
  20#include <linux/coredump.h>
  21#include <linux/security.h>
  22#include <linux/syscalls.h>
  23#include <linux/ptrace.h>
  24#include <linux/signal.h>
  25#include <linux/signalfd.h>
  26#include <linux/ratelimit.h>
  27#include <linux/tracehook.h>
  28#include <linux/capability.h>
  29#include <linux/freezer.h>
  30#include <linux/pid_namespace.h>
  31#include <linux/nsproxy.h>
  32#include <linux/user_namespace.h>
  33#include <linux/uprobes.h>
  34#include <linux/compat.h>
  35#define CREATE_TRACE_POINTS
  36#include <trace/events/signal.h>
  37
  38#include <asm/param.h>
  39#include <asm/uaccess.h>
  40#include <asm/unistd.h>
  41#include <asm/siginfo.h>
  42#include <asm/cacheflush.h>
  43#include "audit.h"      /* audit_signal_info() */
  44
  45/*
  46 * SLAB caches for signal bits.
  47 */
  48
  49static struct kmem_cache *sigqueue_cachep;
  50
  51int print_fatal_signals __read_mostly;
  52
  53static void __user *sig_handler(struct task_struct *t, int sig)
  54{
  55        return t->sighand->action[sig - 1].sa.sa_handler;
  56}
  57
  58static int sig_handler_ignored(void __user *handler, int sig)
  59{
  60        /* Is it explicitly or implicitly ignored? */
  61        return handler == SIG_IGN ||
  62                (handler == SIG_DFL && sig_kernel_ignore(sig));
  63}
  64
  65static int sig_task_ignored(struct task_struct *t, int sig, bool force)
  66{
  67        void __user *handler;
  68
  69        handler = sig_handler(t, sig);
  70
  71        if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
  72                        handler == SIG_DFL && !force)
  73                return 1;
  74
  75        return sig_handler_ignored(handler, sig);
  76}
  77
  78static int sig_ignored(struct task_struct *t, int sig, bool force)
  79{
  80        /*
  81         * Blocked signals are never ignored, since the
  82         * signal handler may change by the time it is
  83         * unblocked.
  84         */
  85        if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
  86                return 0;
  87
  88        if (!sig_task_ignored(t, sig, force))
  89                return 0;
  90
  91        /*
  92         * Tracers may want to know about even ignored signals.
  93         */
  94        return !t->ptrace;
  95}
  96
  97/*
  98 * Re-calculate pending state from the set of locally pending
  99 * signals, globally pending signals, and blocked signals.
 100 */
 101static inline int has_pending_signals(sigset_t *signal, sigset_t *blocked)
 102{
 103        unsigned long ready;
 104        long i;
 105
 106        switch (_NSIG_WORDS) {
 107        default:
 108                for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
 109                        ready |= signal->sig[i] &~ blocked->sig[i];
 110                break;
 111
 112        case 4: ready  = signal->sig[3] &~ blocked->sig[3];
 113                ready |= signal->sig[2] &~ blocked->sig[2];
 114                ready |= signal->sig[1] &~ blocked->sig[1];
 115                ready |= signal->sig[0] &~ blocked->sig[0];
 116                break;
 117
 118        case 2: ready  = signal->sig[1] &~ blocked->sig[1];
 119                ready |= signal->sig[0] &~ blocked->sig[0];
 120                break;
 121
 122        case 1: ready  = signal->sig[0] &~ blocked->sig[0];
 123        }
 124        return ready != 0;
 125}
 126
 127#define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
 128
 129static int recalc_sigpending_tsk(struct task_struct *t)
 130{
 131        if ((t->jobctl & JOBCTL_PENDING_MASK) ||
 132            PENDING(&t->pending, &t->blocked) ||
 133            PENDING(&t->signal->shared_pending, &t->blocked)) {
 134                set_tsk_thread_flag(t, TIF_SIGPENDING);
 135                return 1;
 136        }
 137        /*
 138         * We must never clear the flag in another thread, or in current
 139         * when it's possible the current syscall is returning -ERESTART*.
 140         * So we don't clear it here, and only callers who know they should do.
 141         */
 142        return 0;
 143}
 144
 145/*
 146 * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
 147 * This is superfluous when called on current, the wakeup is a harmless no-op.
 148 */
 149void recalc_sigpending_and_wake(struct task_struct *t)
 150{
 151        if (recalc_sigpending_tsk(t))
 152                signal_wake_up(t, 0);
 153}
 154
 155void recalc_sigpending(void)
 156{
 157        if (!recalc_sigpending_tsk(current) && !freezing(current))
 158                clear_thread_flag(TIF_SIGPENDING);
 159
 160}
 161
 162/* Given the mask, find the first available signal that should be serviced. */
 163
 164#define SYNCHRONOUS_MASK \
 165        (sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
 166         sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
 167
 168int next_signal(struct sigpending *pending, sigset_t *mask)
 169{
 170        unsigned long i, *s, *m, x;
 171        int sig = 0;
 172
 173        s = pending->signal.sig;
 174        m = mask->sig;
 175
 176        /*
 177         * Handle the first word specially: it contains the
 178         * synchronous signals that need to be dequeued first.
 179         */
 180        x = *s &~ *m;
 181        if (x) {
 182                if (x & SYNCHRONOUS_MASK)
 183                        x &= SYNCHRONOUS_MASK;
 184                sig = ffz(~x) + 1;
 185                return sig;
 186        }
 187
 188        switch (_NSIG_WORDS) {
 189        default:
 190                for (i = 1; i < _NSIG_WORDS; ++i) {
 191                        x = *++s &~ *++m;
 192                        if (!x)
 193                                continue;
 194                        sig = ffz(~x) + i*_NSIG_BPW + 1;
 195                        break;
 196                }
 197                break;
 198
 199        case 2:
 200                x = s[1] &~ m[1];
 201                if (!x)
 202                        break;
 203                sig = ffz(~x) + _NSIG_BPW + 1;
 204                break;
 205
 206        case 1:
 207                /* Nothing to do */
 208                break;
 209        }
 210
 211        return sig;
 212}
 213
 214static inline void print_dropped_signal(int sig)
 215{
 216        static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
 217
 218        if (!print_fatal_signals)
 219                return;
 220
 221        if (!__ratelimit(&ratelimit_state))
 222                return;
 223
 224        printk(KERN_INFO "%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
 225                                current->comm, current->pid, sig);
 226}
 227
 228/**
 229 * task_set_jobctl_pending - set jobctl pending bits
 230 * @task: target task
 231 * @mask: pending bits to set
 232 *
 233 * Clear @mask from @task->jobctl.  @mask must be subset of
 234 * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
 235 * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
 236 * cleared.  If @task is already being killed or exiting, this function
 237 * becomes noop.
 238 *
 239 * CONTEXT:
 240 * Must be called with @task->sighand->siglock held.
 241 *
 242 * RETURNS:
 243 * %true if @mask is set, %false if made noop because @task was dying.
 244 */
 245bool task_set_jobctl_pending(struct task_struct *task, unsigned int mask)
 246{
 247        BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
 248                        JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
 249        BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
 250
 251        if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
 252                return false;
 253
 254        if (mask & JOBCTL_STOP_SIGMASK)
 255                task->jobctl &= ~JOBCTL_STOP_SIGMASK;
 256
 257        task->jobctl |= mask;
 258        return true;
 259}
 260
 261/**
 262 * task_clear_jobctl_trapping - clear jobctl trapping bit
 263 * @task: target task
 264 *
 265 * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
 266 * Clear it and wake up the ptracer.  Note that we don't need any further
 267 * locking.  @task->siglock guarantees that @task->parent points to the
 268 * ptracer.
 269 *
 270 * CONTEXT:
 271 * Must be called with @task->sighand->siglock held.
 272 */
 273void task_clear_jobctl_trapping(struct task_struct *task)
 274{
 275        if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
 276                task->jobctl &= ~JOBCTL_TRAPPING;
 277                wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
 278        }
 279}
 280
 281/**
 282 * task_clear_jobctl_pending - clear jobctl pending bits
 283 * @task: target task
 284 * @mask: pending bits to clear
 285 *
 286 * Clear @mask from @task->jobctl.  @mask must be subset of
 287 * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
 288 * STOP bits are cleared together.
 289 *
 290 * If clearing of @mask leaves no stop or trap pending, this function calls
 291 * task_clear_jobctl_trapping().
 292 *
 293 * CONTEXT:
 294 * Must be called with @task->sighand->siglock held.
 295 */
 296void task_clear_jobctl_pending(struct task_struct *task, unsigned int mask)
 297{
 298        BUG_ON(mask & ~JOBCTL_PENDING_MASK);
 299
 300        if (mask & JOBCTL_STOP_PENDING)
 301                mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
 302
 303        task->jobctl &= ~mask;
 304
 305        if (!(task->jobctl & JOBCTL_PENDING_MASK))
 306                task_clear_jobctl_trapping(task);
 307}
 308
 309/**
 310 * task_participate_group_stop - participate in a group stop
 311 * @task: task participating in a group stop
 312 *
 313 * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
 314 * Group stop states are cleared and the group stop count is consumed if
 315 * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
 316 * stop, the appropriate %SIGNAL_* flags are set.
 317 *
 318 * CONTEXT:
 319 * Must be called with @task->sighand->siglock held.
 320 *
 321 * RETURNS:
 322 * %true if group stop completion should be notified to the parent, %false
 323 * otherwise.
 324 */
 325static bool task_participate_group_stop(struct task_struct *task)
 326{
 327        struct signal_struct *sig = task->signal;
 328        bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
 329
 330        WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
 331
 332        task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
 333
 334        if (!consume)
 335                return false;
 336
 337        if (!WARN_ON_ONCE(sig->group_stop_count == 0))
 338                sig->group_stop_count--;
 339
 340        /*
 341         * Tell the caller to notify completion iff we are entering into a
 342         * fresh group stop.  Read comment in do_signal_stop() for details.
 343         */
 344        if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
 345                sig->flags = SIGNAL_STOP_STOPPED;
 346                return true;
 347        }
 348        return false;
 349}
 350
 351/*
 352 * allocate a new signal queue record
 353 * - this may be called without locks if and only if t == current, otherwise an
 354 *   appropriate lock must be held to stop the target task from exiting
 355 */
 356static struct sigqueue *
 357__sigqueue_alloc(int sig, struct task_struct *t, gfp_t flags, int override_rlimit)
 358{
 359        struct sigqueue *q = NULL;
 360        struct user_struct *user;
 361
 362        /*
 363         * Protect access to @t credentials. This can go away when all
 364         * callers hold rcu read lock.
 365         */
 366        rcu_read_lock();
 367        user = get_uid(__task_cred(t)->user);
 368        atomic_inc(&user->sigpending);
 369        rcu_read_unlock();
 370
 371        if (override_rlimit ||
 372            atomic_read(&user->sigpending) <=
 373                        task_rlimit(t, RLIMIT_SIGPENDING)) {
 374                q = kmem_cache_alloc(sigqueue_cachep, flags);
 375        } else {
 376                print_dropped_signal(sig);
 377        }
 378
 379        if (unlikely(q == NULL)) {
 380                atomic_dec(&user->sigpending);
 381                free_uid(user);
 382        } else {
 383                INIT_LIST_HEAD(&q->list);
 384                q->flags = 0;
 385                q->user = user;
 386        }
 387
 388        return q;
 389}
 390
 391static void __sigqueue_free(struct sigqueue *q)
 392{
 393        if (q->flags & SIGQUEUE_PREALLOC)
 394                return;
 395        atomic_dec(&q->user->sigpending);
 396        free_uid(q->user);
 397        kmem_cache_free(sigqueue_cachep, q);
 398}
 399
 400void flush_sigqueue(struct sigpending *queue)
 401{
 402        struct sigqueue *q;
 403
 404        sigemptyset(&queue->signal);
 405        while (!list_empty(&queue->list)) {
 406                q = list_entry(queue->list.next, struct sigqueue , list);
 407                list_del_init(&q->list);
 408                __sigqueue_free(q);
 409        }
 410}
 411
 412/*
 413 * Flush all pending signals for a task.
 414 */
 415void __flush_signals(struct task_struct *t)
 416{
 417        clear_tsk_thread_flag(t, TIF_SIGPENDING);
 418        flush_sigqueue(&t->pending);
 419        flush_sigqueue(&t->signal->shared_pending);
 420}
 421
 422void flush_signals(struct task_struct *t)
 423{
 424        unsigned long flags;
 425
 426        spin_lock_irqsave(&t->sighand->siglock, flags);
 427        __flush_signals(t);
 428        spin_unlock_irqrestore(&t->sighand->siglock, flags);
 429}
 430
 431static void __flush_itimer_signals(struct sigpending *pending)
 432{
 433        sigset_t signal, retain;
 434        struct sigqueue *q, *n;
 435
 436        signal = pending->signal;
 437        sigemptyset(&retain);
 438
 439        list_for_each_entry_safe(q, n, &pending->list, list) {
 440                int sig = q->info.si_signo;
 441
 442                if (likely(q->info.si_code != SI_TIMER)) {
 443                        sigaddset(&retain, sig);
 444                } else {
 445                        sigdelset(&signal, sig);
 446                        list_del_init(&q->list);
 447                        __sigqueue_free(q);
 448                }
 449        }
 450
 451        sigorsets(&pending->signal, &signal, &retain);
 452}
 453
 454void flush_itimer_signals(void)
 455{
 456        struct task_struct *tsk = current;
 457        unsigned long flags;
 458
 459        spin_lock_irqsave(&tsk->sighand->siglock, flags);
 460        __flush_itimer_signals(&tsk->pending);
 461        __flush_itimer_signals(&tsk->signal->shared_pending);
 462        spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
 463}
 464
 465void ignore_signals(struct task_struct *t)
 466{
 467        int i;
 468
 469        for (i = 0; i < _NSIG; ++i)
 470                t->sighand->action[i].sa.sa_handler = SIG_IGN;
 471
 472        flush_signals(t);
 473}
 474
 475/*
 476 * Flush all handlers for a task.
 477 */
 478
 479void
 480flush_signal_handlers(struct task_struct *t, int force_default)
 481{
 482        int i;
 483        struct k_sigaction *ka = &t->sighand->action[0];
 484        for (i = _NSIG ; i != 0 ; i--) {
 485                if (force_default || ka->sa.sa_handler != SIG_IGN)
 486                        ka->sa.sa_handler = SIG_DFL;
 487                ka->sa.sa_flags = 0;
 488                sigemptyset(&ka->sa.sa_mask);
 489                ka++;
 490        }
 491}
 492
 493int unhandled_signal(struct task_struct *tsk, int sig)
 494{
 495        void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
 496        if (is_global_init(tsk))
 497                return 1;
 498        if (handler != SIG_IGN && handler != SIG_DFL)
 499                return 0;
 500        /* if ptraced, let the tracer determine */
 501        return !tsk->ptrace;
 502}
 503
 504/*
 505 * Notify the system that a driver wants to block all signals for this
 506 * process, and wants to be notified if any signals at all were to be
 507 * sent/acted upon.  If the notifier routine returns non-zero, then the
 508 * signal will be acted upon after all.  If the notifier routine returns 0,
 509 * then then signal will be blocked.  Only one block per process is
 510 * allowed.  priv is a pointer to private data that the notifier routine
 511 * can use to determine if the signal should be blocked or not.
 512 */
 513void
 514block_all_signals(int (*notifier)(void *priv), void *priv, sigset_t *mask)
 515{
 516        unsigned long flags;
 517
 518        spin_lock_irqsave(&current->sighand->siglock, flags);
 519        current->notifier_mask = mask;
 520        current->notifier_data = priv;
 521        current->notifier = notifier;
 522        spin_unlock_irqrestore(&current->sighand->siglock, flags);
 523}
 524
 525/* Notify the system that blocking has ended. */
 526
 527void
 528unblock_all_signals(void)
 529{
 530        unsigned long flags;
 531
 532        spin_lock_irqsave(&current->sighand->siglock, flags);
 533        current->notifier = NULL;
 534        current->notifier_data = NULL;
 535        recalc_sigpending();
 536        spin_unlock_irqrestore(&current->sighand->siglock, flags);
 537}
 538
 539static void collect_signal(int sig, struct sigpending *list, siginfo_t *info)
 540{
 541        struct sigqueue *q, *first = NULL;
 542
 543        /*
 544         * Collect the siginfo appropriate to this signal.  Check if
 545         * there is another siginfo for the same signal.
 546        */
 547        list_for_each_entry(q, &list->list, list) {
 548                if (q->info.si_signo == sig) {
 549                        if (first)
 550                                goto still_pending;
 551                        first = q;
 552                }
 553        }
 554
 555        sigdelset(&list->signal, sig);
 556
 557        if (first) {
 558still_pending:
 559                list_del_init(&first->list);
 560                copy_siginfo(info, &first->info);
 561                __sigqueue_free(first);
 562        } else {
 563                /*
 564                 * Ok, it wasn't in the queue.  This must be
 565                 * a fast-pathed signal or we must have been
 566                 * out of queue space.  So zero out the info.
 567                 */
 568                info->si_signo = sig;
 569                info->si_errno = 0;
 570                info->si_code = SI_USER;
 571                info->si_pid = 0;
 572                info->si_uid = 0;
 573        }
 574}
 575
 576static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
 577                        siginfo_t *info)
 578{
 579        int sig = next_signal(pending, mask);
 580
 581        if (sig) {
 582                if (current->notifier) {
 583                        if (sigismember(current->notifier_mask, sig)) {
 584                                if (!(current->notifier)(current->notifier_data)) {
 585                                        clear_thread_flag(TIF_SIGPENDING);
 586                                        return 0;
 587                                }
 588                        }
 589                }
 590
 591                collect_signal(sig, pending, info);
 592        }
 593
 594        return sig;
 595}
 596
 597/*
 598 * Dequeue a signal and return the element to the caller, which is
 599 * expected to free it.
 600 *
 601 * All callers have to hold the siglock.
 602 */
 603int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info)
 604{
 605        int signr;
 606
 607        /* We only dequeue private signals from ourselves, we don't let
 608         * signalfd steal them
 609         */
 610        signr = __dequeue_signal(&tsk->pending, mask, info);
 611        if (!signr) {
 612                signr = __dequeue_signal(&tsk->signal->shared_pending,
 613                                         mask, info);
 614                /*
 615                 * itimer signal ?
 616                 *
 617                 * itimers are process shared and we restart periodic
 618                 * itimers in the signal delivery path to prevent DoS
 619                 * attacks in the high resolution timer case. This is
 620                 * compliant with the old way of self-restarting
 621                 * itimers, as the SIGALRM is a legacy signal and only
 622                 * queued once. Changing the restart behaviour to
 623                 * restart the timer in the signal dequeue path is
 624                 * reducing the timer noise on heavy loaded !highres
 625                 * systems too.
 626                 */
 627                if (unlikely(signr == SIGALRM)) {
 628                        struct hrtimer *tmr = &tsk->signal->real_timer;
 629
 630                        if (!hrtimer_is_queued(tmr) &&
 631                            tsk->signal->it_real_incr.tv64 != 0) {
 632                                hrtimer_forward(tmr, tmr->base->get_time(),
 633                                                tsk->signal->it_real_incr);
 634                                hrtimer_restart(tmr);
 635                        }
 636                }
 637        }
 638
 639        recalc_sigpending();
 640        if (!signr)
 641                return 0;
 642
 643        if (unlikely(sig_kernel_stop(signr))) {
 644                /*
 645                 * Set a marker that we have dequeued a stop signal.  Our
 646                 * caller might release the siglock and then the pending
 647                 * stop signal it is about to process is no longer in the
 648                 * pending bitmasks, but must still be cleared by a SIGCONT
 649                 * (and overruled by a SIGKILL).  So those cases clear this
 650                 * shared flag after we've set it.  Note that this flag may
 651                 * remain set after the signal we return is ignored or
 652                 * handled.  That doesn't matter because its only purpose
 653                 * is to alert stop-signal processing code when another
 654                 * processor has come along and cleared the flag.
 655                 */
 656                current->jobctl |= JOBCTL_STOP_DEQUEUED;
 657        }
 658        if ((info->si_code & __SI_MASK) == __SI_TIMER && info->si_sys_private) {
 659                /*
 660                 * Release the siglock to ensure proper locking order
 661                 * of timer locks outside of siglocks.  Note, we leave
 662                 * irqs disabled here, since the posix-timers code is
 663                 * about to disable them again anyway.
 664                 */
 665                spin_unlock(&tsk->sighand->siglock);
 666                do_schedule_next_timer(info);
 667                spin_lock(&tsk->sighand->siglock);
 668        }
 669        return signr;
 670}
 671
 672/*
 673 * Tell a process that it has a new active signal..
 674 *
 675 * NOTE! we rely on the previous spin_lock to
 676 * lock interrupts for us! We can only be called with
 677 * "siglock" held, and the local interrupt must
 678 * have been disabled when that got acquired!
 679 *
 680 * No need to set need_resched since signal event passing
 681 * goes through ->blocked
 682 */
 683void signal_wake_up_state(struct task_struct *t, unsigned int state)
 684{
 685        set_tsk_thread_flag(t, TIF_SIGPENDING);
 686        /*
 687         * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
 688         * case. We don't check t->state here because there is a race with it
 689         * executing another processor and just now entering stopped state.
 690         * By using wake_up_state, we ensure the process will wake up and
 691         * handle its death signal.
 692         */
 693        if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
 694                kick_process(t);
 695}
 696
 697/*
 698 * Remove signals in mask from the pending set and queue.
 699 * Returns 1 if any signals were found.
 700 *
 701 * All callers must be holding the siglock.
 702 *
 703 * This version takes a sigset mask and looks at all signals,
 704 * not just those in the first mask word.
 705 */
 706static int rm_from_queue_full(sigset_t *mask, struct sigpending *s)
 707{
 708        struct sigqueue *q, *n;
 709        sigset_t m;
 710
 711        sigandsets(&m, mask, &s->signal);
 712        if (sigisemptyset(&m))
 713                return 0;
 714
 715        sigandnsets(&s->signal, &s->signal, mask);
 716        list_for_each_entry_safe(q, n, &s->list, list) {
 717                if (sigismember(mask, q->info.si_signo)) {
 718                        list_del_init(&q->list);
 719                        __sigqueue_free(q);
 720                }
 721        }
 722        return 1;
 723}
 724/*
 725 * Remove signals in mask from the pending set and queue.
 726 * Returns 1 if any signals were found.
 727 *
 728 * All callers must be holding the siglock.
 729 */
 730static int rm_from_queue(unsigned long mask, struct sigpending *s)
 731{
 732        struct sigqueue *q, *n;
 733
 734        if (!sigtestsetmask(&s->signal, mask))
 735                return 0;
 736
 737        sigdelsetmask(&s->signal, mask);
 738        list_for_each_entry_safe(q, n, &s->list, list) {
 739                if (q->info.si_signo < SIGRTMIN &&
 740                    (mask & sigmask(q->info.si_signo))) {
 741                        list_del_init(&q->list);
 742                        __sigqueue_free(q);
 743                }
 744        }
 745        return 1;
 746}
 747
 748static inline int is_si_special(const struct siginfo *info)
 749{
 750        return info <= SEND_SIG_FORCED;
 751}
 752
 753static inline bool si_fromuser(const struct siginfo *info)
 754{
 755        return info == SEND_SIG_NOINFO ||
 756                (!is_si_special(info) && SI_FROMUSER(info));
 757}
 758
 759/*
 760 * called with RCU read lock from check_kill_permission()
 761 */
 762static int kill_ok_by_cred(struct task_struct *t)
 763{
 764        const struct cred *cred = current_cred();
 765        const struct cred *tcred = __task_cred(t);
 766
 767        if (uid_eq(cred->euid, tcred->suid) ||
 768            uid_eq(cred->euid, tcred->uid)  ||
 769            uid_eq(cred->uid,  tcred->suid) ||
 770            uid_eq(cred->uid,  tcred->uid))
 771                return 1;
 772
 773        if (ns_capable(tcred->user_ns, CAP_KILL))
 774                return 1;
 775
 776        return 0;
 777}
 778
 779/*
 780 * Bad permissions for sending the signal
 781 * - the caller must hold the RCU read lock
 782 */
 783static int check_kill_permission(int sig, struct siginfo *info,
 784                                 struct task_struct *t)
 785{
 786        struct pid *sid;
 787        int error;
 788
 789        if (!valid_signal(sig))
 790                return -EINVAL;
 791
 792        if (!si_fromuser(info))
 793                return 0;
 794
 795        error = audit_signal_info(sig, t); /* Let audit system see the signal */
 796        if (error)
 797                return error;
 798
 799        if (!same_thread_group(current, t) &&
 800            !kill_ok_by_cred(t)) {
 801                switch (sig) {
 802                case SIGCONT:
 803                        sid = task_session(t);
 804                        /*
 805                         * We don't return the error if sid == NULL. The
 806                         * task was unhashed, the caller must notice this.
 807                         */
 808                        if (!sid || sid == task_session(current))
 809                                break;
 810                default:
 811                        return -EPERM;
 812                }
 813        }
 814
 815        return security_task_kill(t, info, sig, 0);
 816}
 817
 818/**
 819 * ptrace_trap_notify - schedule trap to notify ptracer
 820 * @t: tracee wanting to notify tracer
 821 *
 822 * This function schedules sticky ptrace trap which is cleared on the next
 823 * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
 824 * ptracer.
 825 *
 826 * If @t is running, STOP trap will be taken.  If trapped for STOP and
 827 * ptracer is listening for events, tracee is woken up so that it can
 828 * re-trap for the new event.  If trapped otherwise, STOP trap will be
 829 * eventually taken without returning to userland after the existing traps
 830 * are finished by PTRACE_CONT.
 831 *
 832 * CONTEXT:
 833 * Must be called with @task->sighand->siglock held.
 834 */
 835static void ptrace_trap_notify(struct task_struct *t)
 836{
 837        WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
 838        assert_spin_locked(&t->sighand->siglock);
 839
 840        task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
 841        ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
 842}
 843
 844/*
 845 * Handle magic process-wide effects of stop/continue signals. Unlike
 846 * the signal actions, these happen immediately at signal-generation
 847 * time regardless of blocking, ignoring, or handling.  This does the
 848 * actual continuing for SIGCONT, but not the actual stopping for stop
 849 * signals. The process stop is done as a signal action for SIG_DFL.
 850 *
 851 * Returns true if the signal should be actually delivered, otherwise
 852 * it should be dropped.
 853 */
 854static int prepare_signal(int sig, struct task_struct *p, bool force)
 855{
 856        struct signal_struct *signal = p->signal;
 857        struct task_struct *t;
 858
 859        if (unlikely(signal->flags & SIGNAL_GROUP_EXIT)) {
 860                /*
 861                 * The process is in the middle of dying, nothing to do.
 862                 */
 863        } else if (sig_kernel_stop(sig)) {
 864                /*
 865                 * This is a stop signal.  Remove SIGCONT from all queues.
 866                 */
 867                rm_from_queue(sigmask(SIGCONT), &signal->shared_pending);
 868                t = p;
 869                do {
 870                        rm_from_queue(sigmask(SIGCONT), &t->pending);
 871                } while_each_thread(p, t);
 872        } else if (sig == SIGCONT) {
 873                unsigned int why;
 874                /*
 875                 * Remove all stop signals from all queues, wake all threads.
 876                 */
 877                rm_from_queue(SIG_KERNEL_STOP_MASK, &signal->shared_pending);
 878                t = p;
 879                do {
 880                        task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
 881                        rm_from_queue(SIG_KERNEL_STOP_MASK, &t->pending);
 882                        if (likely(!(t->ptrace & PT_SEIZED)))
 883                                wake_up_state(t, __TASK_STOPPED);
 884                        else
 885                                ptrace_trap_notify(t);
 886                } while_each_thread(p, t);
 887
 888                /*
 889                 * Notify the parent with CLD_CONTINUED if we were stopped.
 890                 *
 891                 * If we were in the middle of a group stop, we pretend it
 892                 * was already finished, and then continued. Since SIGCHLD
 893                 * doesn't queue we report only CLD_STOPPED, as if the next
 894                 * CLD_CONTINUED was dropped.
 895                 */
 896                why = 0;
 897                if (signal->flags & SIGNAL_STOP_STOPPED)
 898                        why |= SIGNAL_CLD_CONTINUED;
 899                else if (signal->group_stop_count)
 900                        why |= SIGNAL_CLD_STOPPED;
 901
 902                if (why) {
 903                        /*
 904                         * The first thread which returns from do_signal_stop()
 905                         * will take ->siglock, notice SIGNAL_CLD_MASK, and
 906                         * notify its parent. See get_signal_to_deliver().
 907                         */
 908                        signal->flags = why | SIGNAL_STOP_CONTINUED;
 909                        signal->group_stop_count = 0;
 910                        signal->group_exit_code = 0;
 911                }
 912        }
 913
 914        return !sig_ignored(p, sig, force);
 915}
 916
 917/*
 918 * Test if P wants to take SIG.  After we've checked all threads with this,
 919 * it's equivalent to finding no threads not blocking SIG.  Any threads not
 920 * blocking SIG were ruled out because they are not running and already
 921 * have pending signals.  Such threads will dequeue from the shared queue
 922 * as soon as they're available, so putting the signal on the shared queue
 923 * will be equivalent to sending it to one such thread.
 924 */
 925static inline int wants_signal(int sig, struct task_struct *p)
 926{
 927        if (sigismember(&p->blocked, sig))
 928                return 0;
 929        if (p->flags & PF_EXITING)
 930                return 0;
 931        if (sig == SIGKILL)
 932                return 1;
 933        if (task_is_stopped_or_traced(p))
 934                return 0;
 935        return task_curr(p) || !signal_pending(p);
 936}
 937
 938static void complete_signal(int sig, struct task_struct *p, int group)
 939{
 940        struct signal_struct *signal = p->signal;
 941        struct task_struct *t;
 942
 943        /*
 944         * Now find a thread we can wake up to take the signal off the queue.
 945         *
 946         * If the main thread wants the signal, it gets first crack.
 947         * Probably the least surprising to the average bear.
 948         */
 949        if (wants_signal(sig, p))
 950                t = p;
 951        else if (!group || thread_group_empty(p))
 952                /*
 953                 * There is just one thread and it does not need to be woken.
 954                 * It will dequeue unblocked signals before it runs again.
 955                 */
 956                return;
 957        else {
 958                /*
 959                 * Otherwise try to find a suitable thread.
 960                 */
 961                t = signal->curr_target;
 962                while (!wants_signal(sig, t)) {
 963                        t = next_thread(t);
 964                        if (t == signal->curr_target)
 965                                /*
 966                                 * No thread needs to be woken.
 967                                 * Any eligible threads will see
 968                                 * the signal in the queue soon.
 969                                 */
 970                                return;
 971                }
 972                signal->curr_target = t;
 973        }
 974
 975        /*
 976         * Found a killable thread.  If the signal will be fatal,
 977         * then start taking the whole group down immediately.
 978         */
 979        if (sig_fatal(p, sig) &&
 980            !(signal->flags & (SIGNAL_UNKILLABLE | SIGNAL_GROUP_EXIT)) &&
 981            !sigismember(&t->real_blocked, sig) &&
 982            (sig == SIGKILL || !t->ptrace)) {
 983                /*
 984                 * This signal will be fatal to the whole group.
 985                 */
 986                if (!sig_kernel_coredump(sig)) {
 987                        /*
 988                         * Start a group exit and wake everybody up.
 989                         * This way we don't have other threads
 990                         * running and doing things after a slower
 991                         * thread has the fatal signal pending.
 992                         */
 993                        signal->flags = SIGNAL_GROUP_EXIT;
 994                        signal->group_exit_code = sig;
 995                        signal->group_stop_count = 0;
 996                        t = p;
 997                        do {
 998                                task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
 999                                sigaddset(&t->pending.signal, SIGKILL);
1000                                signal_wake_up(t, 1);
1001                        } while_each_thread(p, t);
1002                        return;
1003                }
1004        }
1005
1006        /*
1007         * The signal is already in the shared-pending queue.
1008         * Tell the chosen thread to wake up and dequeue it.
1009         */
1010        signal_wake_up(t, sig == SIGKILL);
1011        return;
1012}
1013
1014static inline int legacy_queue(struct sigpending *signals, int sig)
1015{
1016        return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1017}
1018
1019#ifdef CONFIG_USER_NS
1020static inline void userns_fixup_signal_uid(struct siginfo *info, struct task_struct *t)
1021{
1022        if (current_user_ns() == task_cred_xxx(t, user_ns))
1023                return;
1024
1025        if (SI_FROMKERNEL(info))
1026                return;
1027
1028        rcu_read_lock();
1029        info->si_uid = from_kuid_munged(task_cred_xxx(t, user_ns),
1030                                        make_kuid(current_user_ns(), info->si_uid));
1031        rcu_read_unlock();
1032}
1033#else
1034static inline void userns_fixup_signal_uid(struct siginfo *info, struct task_struct *t)
1035{
1036        return;
1037}
1038#endif
1039
1040static int __send_signal(int sig, struct siginfo *info, struct task_struct *t,
1041                        int group, int from_ancestor_ns)
1042{
1043        struct sigpending *pending;
1044        struct sigqueue *q;
1045        int override_rlimit;
1046        int ret = 0, result;
1047
1048        assert_spin_locked(&t->sighand->siglock);
1049
1050        result = TRACE_SIGNAL_IGNORED;
1051        if (!prepare_signal(sig, t,
1052                        from_ancestor_ns || (info == SEND_SIG_FORCED)))
1053                goto ret;
1054
1055        pending = group ? &t->signal->shared_pending : &t->pending;
1056        /*
1057         * Short-circuit ignored signals and support queuing
1058         * exactly one non-rt signal, so that we can get more
1059         * detailed information about the cause of the signal.
1060         */
1061        result = TRACE_SIGNAL_ALREADY_PENDING;
1062        if (legacy_queue(pending, sig))
1063                goto ret;
1064
1065        result = TRACE_SIGNAL_DELIVERED;
1066        /*
1067         * fast-pathed signals for kernel-internal things like SIGSTOP
1068         * or SIGKILL.
1069         */
1070        if (info == SEND_SIG_FORCED)
1071                goto out_set;
1072
1073        /*
1074         * Real-time signals must be queued if sent by sigqueue, or
1075         * some other real-time mechanism.  It is implementation
1076         * defined whether kill() does so.  We attempt to do so, on
1077         * the principle of least surprise, but since kill is not
1078         * allowed to fail with EAGAIN when low on memory we just
1079         * make sure at least one signal gets delivered and don't
1080         * pass on the info struct.
1081         */
1082        if (sig < SIGRTMIN)
1083                override_rlimit = (is_si_special(info) || info->si_code >= 0);
1084        else
1085                override_rlimit = 0;
1086
1087        q = __sigqueue_alloc(sig, t, GFP_ATOMIC | __GFP_NOTRACK_FALSE_POSITIVE,
1088                override_rlimit);
1089        if (q) {
1090                list_add_tail(&q->list, &pending->list);
1091                switch ((unsigned long) info) {
1092                case (unsigned long) SEND_SIG_NOINFO:
1093                        q->info.si_signo = sig;
1094                        q->info.si_errno = 0;
1095                        q->info.si_code = SI_USER;
1096                        q->info.si_pid = task_tgid_nr_ns(current,
1097                                                        task_active_pid_ns(t));
1098                        q->info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1099                        break;
1100                case (unsigned long) SEND_SIG_PRIV:
1101                        q->info.si_signo = sig;
1102                        q->info.si_errno = 0;
1103                        q->info.si_code = SI_KERNEL;
1104                        q->info.si_pid = 0;
1105                        q->info.si_uid = 0;
1106                        break;
1107                default:
1108                        copy_siginfo(&q->info, info);
1109                        if (from_ancestor_ns)
1110                                q->info.si_pid = 0;
1111                        break;
1112                }
1113
1114                userns_fixup_signal_uid(&q->info, t);
1115
1116        } else if (!is_si_special(info)) {
1117                if (sig >= SIGRTMIN && info->si_code != SI_USER) {
1118                        /*
1119                         * Queue overflow, abort.  We may abort if the
1120                         * signal was rt and sent by user using something
1121                         * other than kill().
1122                         */
1123                        result = TRACE_SIGNAL_OVERFLOW_FAIL;
1124                        ret = -EAGAIN;
1125                        goto ret;
1126                } else {
1127                        /*
1128                         * This is a silent loss of information.  We still
1129                         * send the signal, but the *info bits are lost.
1130                         */
1131                        result = TRACE_SIGNAL_LOSE_INFO;
1132                }
1133        }
1134
1135out_set:
1136        signalfd_notify(t, sig);
1137        sigaddset(&pending->signal, sig);
1138        complete_signal(sig, t, group);
1139ret:
1140        trace_signal_generate(sig, info, t, group, result);
1141        return ret;
1142}
1143
1144static int send_signal(int sig, struct siginfo *info, struct task_struct *t,
1145                        int group)
1146{
1147        int from_ancestor_ns = 0;
1148
1149#ifdef CONFIG_PID_NS
1150        from_ancestor_ns = si_fromuser(info) &&
1151                           !task_pid_nr_ns(current, task_active_pid_ns(t));
1152#endif
1153
1154        return __send_signal(sig, info, t, group, from_ancestor_ns);
1155}
1156
1157static void print_fatal_signal(int signr)
1158{
1159        struct pt_regs *regs = signal_pt_regs();
1160        printk("%s/%d: potentially unexpected fatal signal %d.\n",
1161                current->comm, task_pid_nr(current), signr);
1162
1163#if defined(__i386__) && !defined(__arch_um__)
1164        printk("code at %08lx: ", regs->ip);
1165        {
1166                int i;
1167                for (i = 0; i < 16; i++) {
1168                        unsigned char insn;
1169
1170                        if (get_user(insn, (unsigned char *)(regs->ip + i)))
1171                                break;
1172                        printk("%02x ", insn);
1173                }
1174        }
1175#endif
1176        printk("\n");
1177        preempt_disable();
1178        show_regs(regs);
1179        preempt_enable();
1180}
1181
1182static int __init setup_print_fatal_signals(char *str)
1183{
1184        get_option (&str, &print_fatal_signals);
1185
1186        return 1;
1187}
1188
1189__setup("print-fatal-signals=", setup_print_fatal_signals);
1190
1191int
1192__group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1193{
1194        return send_signal(sig, info, p, 1);
1195}
1196
1197static int
1198specific_send_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1199{
1200        return send_signal(sig, info, t, 0);
1201}
1202
1203int do_send_sig_info(int sig, struct siginfo *info, struct task_struct *p,
1204                        bool group)
1205{
1206        unsigned long flags;
1207        int ret = -ESRCH;
1208
1209        if (lock_task_sighand(p, &flags)) {
1210                ret = send_signal(sig, info, p, group);
1211                unlock_task_sighand(p, &flags);
1212        }
1213
1214        return ret;
1215}
1216
1217/*
1218 * Force a signal that the process can't ignore: if necessary
1219 * we unblock the signal and change any SIG_IGN to SIG_DFL.
1220 *
1221 * Note: If we unblock the signal, we always reset it to SIG_DFL,
1222 * since we do not want to have a signal handler that was blocked
1223 * be invoked when user space had explicitly blocked it.
1224 *
1225 * We don't want to have recursive SIGSEGV's etc, for example,
1226 * that is why we also clear SIGNAL_UNKILLABLE.
1227 */
1228int
1229force_sig_info(int sig, struct siginfo *info, struct task_struct *t)
1230{
1231        unsigned long int flags;
1232        int ret, blocked, ignored;
1233        struct k_sigaction *action;
1234
1235        spin_lock_irqsave(&t->sighand->siglock, flags);
1236        action = &t->sighand->action[sig-1];
1237        ignored = action->sa.sa_handler == SIG_IGN;
1238        blocked = sigismember(&t->blocked, sig);
1239        if (blocked || ignored) {
1240                action->sa.sa_handler = SIG_DFL;
1241                if (blocked) {
1242                        sigdelset(&t->blocked, sig);
1243                        recalc_sigpending_and_wake(t);
1244                }
1245        }
1246        if (action->sa.sa_handler == SIG_DFL)
1247                t->signal->flags &= ~SIGNAL_UNKILLABLE;
1248        ret = specific_send_sig_info(sig, info, t);
1249        spin_unlock_irqrestore(&t->sighand->siglock, flags);
1250
1251        return ret;
1252}
1253
1254/*
1255 * Nuke all other threads in the group.
1256 */
1257int zap_other_threads(struct task_struct *p)
1258{
1259        struct task_struct *t = p;
1260        int count = 0;
1261
1262        p->signal->group_stop_count = 0;
1263
1264        while_each_thread(p, t) {
1265                task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1266                count++;
1267
1268                /* Don't bother with already dead threads */
1269                if (t->exit_state)
1270                        continue;
1271                sigaddset(&t->pending.signal, SIGKILL);
1272                signal_wake_up(t, 1);
1273        }
1274
1275        return count;
1276}
1277
1278struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1279                                           unsigned long *flags)
1280{
1281        struct sighand_struct *sighand;
1282
1283        for (;;) {
1284                local_irq_save(*flags);
1285                rcu_read_lock();
1286                sighand = rcu_dereference(tsk->sighand);
1287                if (unlikely(sighand == NULL)) {
1288                        rcu_read_unlock();
1289                        local_irq_restore(*flags);
1290                        break;
1291                }
1292
1293                spin_lock(&sighand->siglock);
1294                if (likely(sighand == tsk->sighand)) {
1295                        rcu_read_unlock();
1296                        break;
1297                }
1298                spin_unlock(&sighand->siglock);
1299                rcu_read_unlock();
1300                local_irq_restore(*flags);
1301        }
1302
1303        return sighand;
1304}
1305
1306/*
1307 * send signal info to all the members of a group
1308 */
1309int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1310{
1311        int ret;
1312
1313        rcu_read_lock();
1314        ret = check_kill_permission(sig, info, p);
1315        rcu_read_unlock();
1316
1317        if (!ret && sig)
1318                ret = do_send_sig_info(sig, info, p, true);
1319
1320        return ret;
1321}
1322
1323/*
1324 * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1325 * control characters do (^C, ^Z etc)
1326 * - the caller must hold at least a readlock on tasklist_lock
1327 */
1328int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp)
1329{
1330        struct task_struct *p = NULL;
1331        int retval, success;
1332
1333        success = 0;
1334        retval = -ESRCH;
1335        do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1336                int err = group_send_sig_info(sig, info, p);
1337                success |= !err;
1338                retval = err;
1339        } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1340        return success ? 0 : retval;
1341}
1342
1343int kill_pid_info(int sig, struct siginfo *info, struct pid *pid)
1344{
1345        int error = -ESRCH;
1346        struct task_struct *p;
1347
1348        rcu_read_lock();
1349retry:
1350        p = pid_task(pid, PIDTYPE_PID);
1351        if (p) {
1352                error = group_send_sig_info(sig, info, p);
1353                if (unlikely(error == -ESRCH))
1354                        /*
1355                         * The task was unhashed in between, try again.
1356                         * If it is dead, pid_task() will return NULL,
1357                         * if we race with de_thread() it will find the
1358                         * new leader.
1359                         */
1360                        goto retry;
1361        }
1362        rcu_read_unlock();
1363
1364        return error;
1365}
1366
1367int kill_proc_info(int sig, struct siginfo *info, pid_t pid)
1368{
1369        int error;
1370        rcu_read_lock();
1371        error = kill_pid_info(sig, info, find_vpid(pid));
1372        rcu_read_unlock();
1373        return error;
1374}
1375
1376static int kill_as_cred_perm(const struct cred *cred,
1377                             struct task_struct *target)
1378{
1379        const struct cred *pcred = __task_cred(target);
1380        if (!uid_eq(cred->euid, pcred->suid) && !uid_eq(cred->euid, pcred->uid) &&
1381            !uid_eq(cred->uid,  pcred->suid) && !uid_eq(cred->uid,  pcred->uid))
1382                return 0;
1383        return 1;
1384}
1385
1386/* like kill_pid_info(), but doesn't use uid/euid of "current" */
1387int kill_pid_info_as_cred(int sig, struct siginfo *info, struct pid *pid,
1388                         const struct cred *cred, u32 secid)
1389{
1390        int ret = -EINVAL;
1391        struct task_struct *p;
1392        unsigned long flags;
1393
1394        if (!valid_signal(sig))
1395                return ret;
1396
1397        rcu_read_lock();
1398        p = pid_task(pid, PIDTYPE_PID);
1399        if (!p) {
1400                ret = -ESRCH;
1401                goto out_unlock;
1402        }
1403        if (si_fromuser(info) && !kill_as_cred_perm(cred, p)) {
1404                ret = -EPERM;
1405                goto out_unlock;
1406        }
1407        ret = security_task_kill(p, info, sig, secid);
1408        if (ret)
1409                goto out_unlock;
1410
1411        if (sig) {
1412                if (lock_task_sighand(p, &flags)) {
1413                        ret = __send_signal(sig, info, p, 1, 0);
1414                        unlock_task_sighand(p, &flags);
1415                } else
1416                        ret = -ESRCH;
1417        }
1418out_unlock:
1419        rcu_read_unlock();
1420        return ret;
1421}
1422EXPORT_SYMBOL_GPL(kill_pid_info_as_cred);
1423
1424/*
1425 * kill_something_info() interprets pid in interesting ways just like kill(2).
1426 *
1427 * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1428 * is probably wrong.  Should make it like BSD or SYSV.
1429 */
1430
1431static int kill_something_info(int sig, struct siginfo *info, pid_t pid)
1432{
1433        int ret;
1434
1435        if (pid > 0) {
1436                rcu_read_lock();
1437                ret = kill_pid_info(sig, info, find_vpid(pid));
1438                rcu_read_unlock();
1439                return ret;
1440        }
1441
1442        read_lock(&tasklist_lock);
1443        if (pid != -1) {
1444                ret = __kill_pgrp_info(sig, info,
1445                                pid ? find_vpid(-pid) : task_pgrp(current));
1446        } else {
1447                int retval = 0, count = 0;
1448                struct task_struct * p;
1449
1450                for_each_process(p) {
1451                        if (task_pid_vnr(p) > 1 &&
1452                                        !same_thread_group(p, current)) {
1453                                int err = group_send_sig_info(sig, info, p);
1454                                ++count;
1455                                if (err != -EPERM)
1456                                        retval = err;
1457                        }
1458                }
1459                ret = count ? retval : -ESRCH;
1460        }
1461        read_unlock(&tasklist_lock);
1462
1463        return ret;
1464}
1465
1466/*
1467 * These are for backward compatibility with the rest of the kernel source.
1468 */
1469
1470int send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1471{
1472        /*
1473         * Make sure legacy kernel users don't send in bad values
1474         * (normal paths check this in check_kill_permission).
1475         */
1476        if (!valid_signal(sig))
1477                return -EINVAL;
1478
1479        return do_send_sig_info(sig, info, p, false);
1480}
1481
1482#define __si_special(priv) \
1483        ((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1484
1485int
1486send_sig(int sig, struct task_struct *p, int priv)
1487{
1488        return send_sig_info(sig, __si_special(priv), p);
1489}
1490
1491void
1492force_sig(int sig, struct task_struct *p)
1493{
1494        force_sig_info(sig, SEND_SIG_PRIV, p);
1495}
1496
1497/*
1498 * When things go south during signal handling, we
1499 * will force a SIGSEGV. And if the signal that caused
1500 * the problem was already a SIGSEGV, we'll want to
1501 * make sure we don't even try to deliver the signal..
1502 */
1503int
1504force_sigsegv(int sig, struct task_struct *p)
1505{
1506        if (sig == SIGSEGV) {
1507                unsigned long flags;
1508                spin_lock_irqsave(&p->sighand->siglock, flags);
1509                p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1510                spin_unlock_irqrestore(&p->sighand->siglock, flags);
1511        }
1512        force_sig(SIGSEGV, p);
1513        return 0;
1514}
1515
1516int kill_pgrp(struct pid *pid, int sig, int priv)
1517{
1518        int ret;
1519
1520        read_lock(&tasklist_lock);
1521        ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1522        read_unlock(&tasklist_lock);
1523
1524        return ret;
1525}
1526EXPORT_SYMBOL(kill_pgrp);
1527
1528int kill_pid(struct pid *pid, int sig, int priv)
1529{
1530        return kill_pid_info(sig, __si_special(priv), pid);
1531}
1532EXPORT_SYMBOL(kill_pid);
1533
1534/*
1535 * These functions support sending signals using preallocated sigqueue
1536 * structures.  This is needed "because realtime applications cannot
1537 * afford to lose notifications of asynchronous events, like timer
1538 * expirations or I/O completions".  In the case of POSIX Timers
1539 * we allocate the sigqueue structure from the timer_create.  If this
1540 * allocation fails we are able to report the failure to the application
1541 * with an EAGAIN error.
1542 */
1543struct sigqueue *sigqueue_alloc(void)
1544{
1545        struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1546
1547        if (q)
1548                q->flags |= SIGQUEUE_PREALLOC;
1549
1550        return q;
1551}
1552
1553void sigqueue_free(struct sigqueue *q)
1554{
1555        unsigned long flags;
1556        spinlock_t *lock = &current->sighand->siglock;
1557
1558        BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1559        /*
1560         * We must hold ->siglock while testing q->list
1561         * to serialize with collect_signal() or with
1562         * __exit_signal()->flush_sigqueue().
1563         */
1564        spin_lock_irqsave(lock, flags);
1565        q->flags &= ~SIGQUEUE_PREALLOC;
1566        /*
1567         * If it is queued it will be freed when dequeued,
1568         * like the "regular" sigqueue.
1569         */
1570        if (!list_empty(&q->list))
1571                q = NULL;
1572        spin_unlock_irqrestore(lock, flags);
1573
1574        if (q)
1575                __sigqueue_free(q);
1576}
1577
1578int send_sigqueue(struct sigqueue *q, struct task_struct *t, int group)
1579{
1580        int sig = q->info.si_signo;
1581        struct sigpending *pending;
1582        unsigned long flags;
1583        int ret, result;
1584
1585        BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1586
1587        ret = -1;
1588        if (!likely(lock_task_sighand(t, &flags)))
1589                goto ret;
1590
1591        ret = 1; /* the signal is ignored */
1592        result = TRACE_SIGNAL_IGNORED;
1593        if (!prepare_signal(sig, t, false))
1594                goto out;
1595
1596        ret = 0;
1597        if (unlikely(!list_empty(&q->list))) {
1598                /*
1599                 * If an SI_TIMER entry is already queue just increment
1600                 * the overrun count.
1601                 */
1602                BUG_ON(q->info.si_code != SI_TIMER);
1603                q->info.si_overrun++;
1604                result = TRACE_SIGNAL_ALREADY_PENDING;
1605                goto out;
1606        }
1607        q->info.si_overrun = 0;
1608
1609        signalfd_notify(t, sig);
1610        pending = group ? &t->signal->shared_pending : &t->pending;
1611        list_add_tail(&q->list, &pending->list);
1612        sigaddset(&pending->signal, sig);
1613        complete_signal(sig, t, group);
1614        result = TRACE_SIGNAL_DELIVERED;
1615out:
1616        trace_signal_generate(sig, &q->info, t, group, result);
1617        unlock_task_sighand(t, &flags);
1618ret:
1619        return ret;
1620}
1621
1622/*
1623 * Let a parent know about the death of a child.
1624 * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1625 *
1626 * Returns true if our parent ignored us and so we've switched to
1627 * self-reaping.
1628 */
1629bool do_notify_parent(struct task_struct *tsk, int sig)
1630{
1631        struct siginfo info;
1632        unsigned long flags;
1633        struct sighand_struct *psig;
1634        bool autoreap = false;
1635
1636        BUG_ON(sig == -1);
1637
1638        /* do_notify_parent_cldstop should have been called instead.  */
1639        BUG_ON(task_is_stopped_or_traced(tsk));
1640
1641        BUG_ON(!tsk->ptrace &&
1642               (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1643
1644        if (sig != SIGCHLD) {
1645                /*
1646                 * This is only possible if parent == real_parent.
1647                 * Check if it has changed security domain.
1648                 */
1649                if (tsk->parent_exec_id != tsk->parent->self_exec_id)
1650                        sig = SIGCHLD;
1651        }
1652
1653        info.si_signo = sig;
1654        info.si_errno = 0;
1655        /*
1656         * We are under tasklist_lock here so our parent is tied to
1657         * us and cannot change.
1658         *
1659         * task_active_pid_ns will always return the same pid namespace
1660         * until a task passes through release_task.
1661         *
1662         * write_lock() currently calls preempt_disable() which is the
1663         * same as rcu_read_lock(), but according to Oleg, this is not
1664         * correct to rely on this
1665         */
1666        rcu_read_lock();
1667        info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1668        info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1669                                       task_uid(tsk));
1670        rcu_read_unlock();
1671
1672        info.si_utime = cputime_to_clock_t(tsk->utime + tsk->signal->utime);
1673        info.si_stime = cputime_to_clock_t(tsk->stime + tsk->signal->stime);
1674
1675        info.si_status = tsk->exit_code & 0x7f;
1676        if (tsk->exit_code & 0x80)
1677                info.si_code = CLD_DUMPED;
1678        else if (tsk->exit_code & 0x7f)
1679                info.si_code = CLD_KILLED;
1680        else {
1681                info.si_code = CLD_EXITED;
1682                info.si_status = tsk->exit_code >> 8;
1683        }
1684
1685        psig = tsk->parent->sighand;
1686        spin_lock_irqsave(&psig->siglock, flags);
1687        if (!tsk->ptrace && sig == SIGCHLD &&
1688            (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1689             (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1690                /*
1691                 * We are exiting and our parent doesn't care.  POSIX.1
1692                 * defines special semantics for setting SIGCHLD to SIG_IGN
1693                 * or setting the SA_NOCLDWAIT flag: we should be reaped
1694                 * automatically and not left for our parent's wait4 call.
1695                 * Rather than having the parent do it as a magic kind of
1696                 * signal handler, we just set this to tell do_exit that we
1697                 * can be cleaned up without becoming a zombie.  Note that
1698                 * we still call __wake_up_parent in this case, because a
1699                 * blocked sys_wait4 might now return -ECHILD.
1700                 *
1701                 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1702                 * is implementation-defined: we do (if you don't want
1703                 * it, just use SIG_IGN instead).
1704                 */
1705                autoreap = true;
1706                if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
1707                        sig = 0;
1708        }
1709        if (valid_signal(sig) && sig)
1710                __group_send_sig_info(sig, &info, tsk->parent);
1711        __wake_up_parent(tsk, tsk->parent);
1712        spin_unlock_irqrestore(&psig->siglock, flags);
1713
1714        return autoreap;
1715}
1716
1717/**
1718 * do_notify_parent_cldstop - notify parent of stopped/continued state change
1719 * @tsk: task reporting the state change
1720 * @for_ptracer: the notification is for ptracer
1721 * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
1722 *
1723 * Notify @tsk's parent that the stopped/continued state has changed.  If
1724 * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
1725 * If %true, @tsk reports to @tsk->parent which should be the ptracer.
1726 *
1727 * CONTEXT:
1728 * Must be called with tasklist_lock at least read locked.
1729 */
1730static void do_notify_parent_cldstop(struct task_struct *tsk,
1731                                     bool for_ptracer, int why)
1732{
1733        struct siginfo info;
1734        unsigned long flags;
1735        struct task_struct *parent;
1736        struct sighand_struct *sighand;
1737
1738        if (for_ptracer) {
1739                parent = tsk->parent;
1740        } else {
1741                tsk = tsk->group_leader;
1742                parent = tsk->real_parent;
1743        }
1744
1745        info.si_signo = SIGCHLD;
1746        info.si_errno = 0;
1747        /*
1748         * see comment in do_notify_parent() about the following 4 lines
1749         */
1750        rcu_read_lock();
1751        info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
1752        info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
1753        rcu_read_unlock();
1754
1755        info.si_utime = cputime_to_clock_t(tsk->utime);
1756        info.si_stime = cputime_to_clock_t(tsk->stime);
1757
1758        info.si_code = why;
1759        switch (why) {
1760        case CLD_CONTINUED:
1761                info.si_status = SIGCONT;
1762                break;
1763        case CLD_STOPPED:
1764                info.si_status = tsk->signal->group_exit_code & 0x7f;
1765                break;
1766        case CLD_TRAPPED:
1767                info.si_status = tsk->exit_code & 0x7f;
1768                break;
1769        default:
1770                BUG();
1771        }
1772
1773        sighand = parent->sighand;
1774        spin_lock_irqsave(&sighand->siglock, flags);
1775        if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
1776            !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
1777                __group_send_sig_info(SIGCHLD, &info, parent);
1778        /*
1779         * Even if SIGCHLD is not generated, we must wake up wait4 calls.
1780         */
1781        __wake_up_parent(tsk, parent);
1782        spin_unlock_irqrestore(&sighand->siglock, flags);
1783}
1784
1785static inline int may_ptrace_stop(void)
1786{
1787        if (!likely(current->ptrace))
1788                return 0;
1789        /*
1790         * Are we in the middle of do_coredump?
1791         * If so and our tracer is also part of the coredump stopping
1792         * is a deadlock situation, and pointless because our tracer
1793         * is dead so don't allow us to stop.
1794         * If SIGKILL was already sent before the caller unlocked
1795         * ->siglock we must see ->core_state != NULL. Otherwise it
1796         * is safe to enter schedule().
1797         *
1798         * This is almost outdated, a task with the pending SIGKILL can't
1799         * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
1800         * after SIGKILL was already dequeued.
1801         */
1802        if (unlikely(current->mm->core_state) &&
1803            unlikely(current->mm == current->parent->mm))
1804                return 0;
1805
1806        return 1;
1807}
1808
1809/*
1810 * Return non-zero if there is a SIGKILL that should be waking us up.
1811 * Called with the siglock held.
1812 */
1813static int sigkill_pending(struct task_struct *tsk)
1814{
1815        return  sigismember(&tsk->pending.signal, SIGKILL) ||
1816                sigismember(&tsk->signal->shared_pending.signal, SIGKILL);
1817}
1818
1819/*
1820 * This must be called with current->sighand->siglock held.
1821 *
1822 * This should be the path for all ptrace stops.
1823 * We always set current->last_siginfo while stopped here.
1824 * That makes it a way to test a stopped process for
1825 * being ptrace-stopped vs being job-control-stopped.
1826 *
1827 * If we actually decide not to stop at all because the tracer
1828 * is gone, we keep current->exit_code unless clear_code.
1829 */
1830static void ptrace_stop(int exit_code, int why, int clear_code, siginfo_t *info)
1831        __releases(&current->sighand->siglock)
1832        __acquires(&current->sighand->siglock)
1833{
1834        bool gstop_done = false;
1835
1836        if (arch_ptrace_stop_needed(exit_code, info)) {
1837                /*
1838                 * The arch code has something special to do before a
1839                 * ptrace stop.  This is allowed to block, e.g. for faults
1840                 * on user stack pages.  We can't keep the siglock while
1841                 * calling arch_ptrace_stop, so we must release it now.
1842                 * To preserve proper semantics, we must do this before
1843                 * any signal bookkeeping like checking group_stop_count.
1844                 * Meanwhile, a SIGKILL could come in before we retake the
1845                 * siglock.  That must prevent us from sleeping in TASK_TRACED.
1846                 * So after regaining the lock, we must check for SIGKILL.
1847                 */
1848                spin_unlock_irq(&current->sighand->siglock);
1849                arch_ptrace_stop(exit_code, info);
1850                spin_lock_irq(&current->sighand->siglock);
1851                if (sigkill_pending(current))
1852                        return;
1853        }
1854
1855        /*
1856         * We're committing to trapping.  TRACED should be visible before
1857         * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
1858         * Also, transition to TRACED and updates to ->jobctl should be
1859         * atomic with respect to siglock and should be done after the arch
1860         * hook as siglock is released and regrabbed across it.
1861         */
1862        set_current_state(TASK_TRACED);
1863
1864        current->last_siginfo = info;
1865        current->exit_code = exit_code;
1866
1867        /*
1868         * If @why is CLD_STOPPED, we're trapping to participate in a group
1869         * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
1870         * across siglock relocks since INTERRUPT was scheduled, PENDING
1871         * could be clear now.  We act as if SIGCONT is received after
1872         * TASK_TRACED is entered - ignore it.
1873         */
1874        if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
1875                gstop_done = task_participate_group_stop(current);
1876
1877        /* any trap clears pending STOP trap, STOP trap clears NOTIFY */
1878        task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
1879        if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
1880                task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
1881
1882        /* entering a trap, clear TRAPPING */
1883        task_clear_jobctl_trapping(current);
1884
1885        spin_unlock_irq(&current->sighand->siglock);
1886        read_lock(&tasklist_lock);
1887        if (may_ptrace_stop()) {
1888                /*
1889                 * Notify parents of the stop.
1890                 *
1891                 * While ptraced, there are two parents - the ptracer and
1892                 * the real_parent of the group_leader.  The ptracer should
1893                 * know about every stop while the real parent is only
1894                 * interested in the completion of group stop.  The states
1895                 * for the two don't interact with each other.  Notify
1896                 * separately unless they're gonna be duplicates.
1897                 */
1898                do_notify_parent_cldstop(current, true, why);
1899                if (gstop_done && ptrace_reparented(current))
1900                        do_notify_parent_cldstop(current, false, why);
1901
1902                /*
1903                 * Don't want to allow preemption here, because
1904                 * sys_ptrace() needs this task to be inactive.
1905                 *
1906                 * XXX: implement read_unlock_no_resched().
1907                 */
1908                preempt_disable();
1909                read_unlock(&tasklist_lock);
1910                preempt_enable_no_resched();
1911                freezable_schedule();
1912        } else {
1913                /*
1914                 * By the time we got the lock, our tracer went away.
1915                 * Don't drop the lock yet, another tracer may come.
1916                 *
1917                 * If @gstop_done, the ptracer went away between group stop
1918                 * completion and here.  During detach, it would have set
1919                 * JOBCTL_STOP_PENDING on us and we'll re-enter
1920                 * TASK_STOPPED in do_signal_stop() on return, so notifying
1921                 * the real parent of the group stop completion is enough.
1922                 */
1923                if (gstop_done)
1924                        do_notify_parent_cldstop(current, false, why);
1925
1926                /* tasklist protects us from ptrace_freeze_traced() */
1927                __set_current_state(TASK_RUNNING);
1928                if (clear_code)
1929                        current->exit_code = 0;
1930                read_unlock(&tasklist_lock);
1931        }
1932
1933        /*
1934         * We are back.  Now reacquire the siglock before touching
1935         * last_siginfo, so that we are sure to have synchronized with
1936         * any signal-sending on another CPU that wants to examine it.
1937         */
1938        spin_lock_irq(&current->sighand->siglock);
1939        current->last_siginfo = NULL;
1940
1941        /* LISTENING can be set only during STOP traps, clear it */
1942        current->jobctl &= ~JOBCTL_LISTENING;
1943
1944        /*
1945         * Queued signals ignored us while we were stopped for tracing.
1946         * So check for any that we should take before resuming user mode.
1947         * This sets TIF_SIGPENDING, but never clears it.
1948         */
1949        recalc_sigpending_tsk(current);
1950}
1951
1952static void ptrace_do_notify(int signr, int exit_code, int why)
1953{
1954        siginfo_t info;
1955
1956        memset(&info, 0, sizeof info);
1957        info.si_signo = signr;
1958        info.si_code = exit_code;
1959        info.si_pid = task_pid_vnr(current);
1960        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
1961
1962        /* Let the debugger run.  */
1963        ptrace_stop(exit_code, why, 1, &info);
1964}
1965
1966void ptrace_notify(int exit_code)
1967{
1968        BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
1969        if (unlikely(current->task_works))
1970                task_work_run();
1971
1972        spin_lock_irq(&current->sighand->siglock);
1973        ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
1974        spin_unlock_irq(&current->sighand->siglock);
1975}
1976
1977/**
1978 * do_signal_stop - handle group stop for SIGSTOP and other stop signals
1979 * @signr: signr causing group stop if initiating
1980 *
1981 * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
1982 * and participate in it.  If already set, participate in the existing
1983 * group stop.  If participated in a group stop (and thus slept), %true is
1984 * returned with siglock released.
1985 *
1986 * If ptraced, this function doesn't handle stop itself.  Instead,
1987 * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
1988 * untouched.  The caller must ensure that INTERRUPT trap handling takes
1989 * places afterwards.
1990 *
1991 * CONTEXT:
1992 * Must be called with @current->sighand->siglock held, which is released
1993 * on %true return.
1994 *
1995 * RETURNS:
1996 * %false if group stop is already cancelled or ptrace trap is scheduled.
1997 * %true if participated in group stop.
1998 */
1999static bool do_signal_stop(int signr)
2000        __releases(&current->sighand->siglock)
2001{
2002        struct signal_struct *sig = current->signal;
2003
2004        if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2005                unsigned int gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2006                struct task_struct *t;
2007
2008                /* signr will be recorded in task->jobctl for retries */
2009                WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2010
2011                if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2012                    unlikely(signal_group_exit(sig)))
2013                        return false;
2014                /*
2015                 * There is no group stop already in progress.  We must
2016                 * initiate one now.
2017                 *
2018                 * While ptraced, a task may be resumed while group stop is
2019                 * still in effect and then receive a stop signal and
2020                 * initiate another group stop.  This deviates from the
2021                 * usual behavior as two consecutive stop signals can't
2022                 * cause two group stops when !ptraced.  That is why we
2023                 * also check !task_is_stopped(t) below.
2024                 *
2025                 * The condition can be distinguished by testing whether
2026                 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2027                 * group_exit_code in such case.
2028                 *
2029                 * This is not necessary for SIGNAL_STOP_CONTINUED because
2030                 * an intervening stop signal is required to cause two
2031                 * continued events regardless of ptrace.
2032                 */
2033                if (!(sig->flags & SIGNAL_STOP_STOPPED))
2034                        sig->group_exit_code = signr;
2035
2036                sig->group_stop_count = 0;
2037
2038                if (task_set_jobctl_pending(current, signr | gstop))
2039                        sig->group_stop_count++;
2040
2041                for (t = next_thread(current); t != current;
2042                     t = next_thread(t)) {
2043                        /*
2044                         * Setting state to TASK_STOPPED for a group
2045                         * stop is always done with the siglock held,
2046                         * so this check has no races.
2047                         */
2048                        if (!task_is_stopped(t) &&
2049                            task_set_jobctl_pending(t, signr | gstop)) {
2050                                sig->group_stop_count++;
2051                                if (likely(!(t->ptrace & PT_SEIZED)))
2052                                        signal_wake_up(t, 0);
2053                                else
2054                                        ptrace_trap_notify(t);
2055                        }
2056                }
2057        }
2058
2059        if (likely(!current->ptrace)) {
2060                int notify = 0;
2061
2062                /*
2063                 * If there are no other threads in the group, or if there
2064                 * is a group stop in progress and we are the last to stop,
2065                 * report to the parent.
2066                 */
2067                if (task_participate_group_stop(current))
2068                        notify = CLD_STOPPED;
2069
2070                __set_current_state(TASK_STOPPED);
2071                spin_unlock_irq(&current->sighand->siglock);
2072
2073                /*
2074                 * Notify the parent of the group stop completion.  Because
2075                 * we're not holding either the siglock or tasklist_lock
2076                 * here, ptracer may attach inbetween; however, this is for
2077                 * group stop and should always be delivered to the real
2078                 * parent of the group leader.  The new ptracer will get
2079                 * its notification when this task transitions into
2080                 * TASK_TRACED.
2081                 */
2082                if (notify) {
2083                        read_lock(&tasklist_lock);
2084                        do_notify_parent_cldstop(current, false, notify);
2085                        read_unlock(&tasklist_lock);
2086                }
2087
2088                /* Now we don't run again until woken by SIGCONT or SIGKILL */
2089                freezable_schedule();
2090                return true;
2091        } else {
2092                /*
2093                 * While ptraced, group stop is handled by STOP trap.
2094                 * Schedule it and let the caller deal with it.
2095                 */
2096                task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2097                return false;
2098        }
2099}
2100
2101/**
2102 * do_jobctl_trap - take care of ptrace jobctl traps
2103 *
2104 * When PT_SEIZED, it's used for both group stop and explicit
2105 * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2106 * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2107 * the stop signal; otherwise, %SIGTRAP.
2108 *
2109 * When !PT_SEIZED, it's used only for group stop trap with stop signal
2110 * number as exit_code and no siginfo.
2111 *
2112 * CONTEXT:
2113 * Must be called with @current->sighand->siglock held, which may be
2114 * released and re-acquired before returning with intervening sleep.
2115 */
2116static void do_jobctl_trap(void)
2117{
2118        struct signal_struct *signal = current->signal;
2119        int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2120
2121        if (current->ptrace & PT_SEIZED) {
2122                if (!signal->group_stop_count &&
2123                    !(signal->flags & SIGNAL_STOP_STOPPED))
2124                        signr = SIGTRAP;
2125                WARN_ON_ONCE(!signr);
2126                ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2127                                 CLD_STOPPED);
2128        } else {
2129                WARN_ON_ONCE(!signr);
2130                ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2131                current->exit_code = 0;
2132        }
2133}
2134
2135static int ptrace_signal(int signr, siginfo_t *info)
2136{
2137        ptrace_signal_deliver();
2138        /*
2139         * We do not check sig_kernel_stop(signr) but set this marker
2140         * unconditionally because we do not know whether debugger will
2141         * change signr. This flag has no meaning unless we are going
2142         * to stop after return from ptrace_stop(). In this case it will
2143         * be checked in do_signal_stop(), we should only stop if it was
2144         * not cleared by SIGCONT while we were sleeping. See also the
2145         * comment in dequeue_signal().
2146         */
2147        current->jobctl |= JOBCTL_STOP_DEQUEUED;
2148        ptrace_stop(signr, CLD_TRAPPED, 0, info);
2149
2150        /* We're back.  Did the debugger cancel the sig?  */
2151        signr = current->exit_code;
2152        if (signr == 0)
2153                return signr;
2154
2155        current->exit_code = 0;
2156
2157        /*
2158         * Update the siginfo structure if the signal has
2159         * changed.  If the debugger wanted something
2160         * specific in the siginfo structure then it should
2161         * have updated *info via PTRACE_SETSIGINFO.
2162         */
2163        if (signr != info->si_signo) {
2164                info->si_signo = signr;
2165                info->si_errno = 0;
2166                info->si_code = SI_USER;
2167                rcu_read_lock();
2168                info->si_pid = task_pid_vnr(current->parent);
2169                info->si_uid = from_kuid_munged(current_user_ns(),
2170                                                task_uid(current->parent));
2171                rcu_read_unlock();
2172        }
2173
2174        /* If the (new) signal is now blocked, requeue it.  */
2175        if (sigismember(&current->blocked, signr)) {
2176                specific_send_sig_info(signr, info, current);
2177                signr = 0;
2178        }
2179
2180        return signr;
2181}
2182
2183int get_signal_to_deliver(siginfo_t *info, struct k_sigaction *return_ka,
2184                          struct pt_regs *regs, void *cookie)
2185{
2186        struct sighand_struct *sighand = current->sighand;
2187        struct signal_struct *signal = current->signal;
2188        int signr;
2189
2190        if (unlikely(current->task_works))
2191                task_work_run();
2192
2193        if (unlikely(uprobe_deny_signal()))
2194                return 0;
2195
2196        /*
2197         * Do this once, we can't return to user-mode if freezing() == T.
2198         * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2199         * thus do not need another check after return.
2200         */
2201        try_to_freeze();
2202
2203relock:
2204        spin_lock_irq(&sighand->siglock);
2205        /*
2206         * Every stopped thread goes here after wakeup. Check to see if
2207         * we should notify the parent, prepare_signal(SIGCONT) encodes
2208         * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2209         */
2210        if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2211                int why;
2212
2213                if (signal->flags & SIGNAL_CLD_CONTINUED)
2214                        why = CLD_CONTINUED;
2215                else
2216                        why = CLD_STOPPED;
2217
2218                signal->flags &= ~SIGNAL_CLD_MASK;
2219
2220                spin_unlock_irq(&sighand->siglock);
2221
2222                /*
2223                 * Notify the parent that we're continuing.  This event is
2224                 * always per-process and doesn't make whole lot of sense
2225                 * for ptracers, who shouldn't consume the state via
2226                 * wait(2) either, but, for backward compatibility, notify
2227                 * the ptracer of the group leader too unless it's gonna be
2228                 * a duplicate.
2229                 */
2230                read_lock(&tasklist_lock);
2231                do_notify_parent_cldstop(current, false, why);
2232
2233                if (ptrace_reparented(current->group_leader))
2234                        do_notify_parent_cldstop(current->group_leader,
2235                                                true, why);
2236                read_unlock(&tasklist_lock);
2237
2238                goto relock;
2239        }
2240
2241        for (;;) {
2242                struct k_sigaction *ka;
2243
2244                if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2245                    do_signal_stop(0))
2246                        goto relock;
2247
2248                if (unlikely(current->jobctl & JOBCTL_TRAP_MASK)) {
2249                        do_jobctl_trap();
2250                        spin_unlock_irq(&sighand->siglock);
2251                        goto relock;
2252                }
2253
2254                signr = dequeue_signal(current, &current->blocked, info);
2255
2256                if (!signr)
2257                        break; /* will return 0 */
2258
2259                if (unlikely(current->ptrace) && signr != SIGKILL) {
2260                        signr = ptrace_signal(signr, info);
2261                        if (!signr)
2262                                continue;
2263                }
2264
2265                ka = &sighand->action[signr-1];
2266
2267                /* Trace actually delivered signals. */
2268                trace_signal_deliver(signr, info, ka);
2269
2270                if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2271                        continue;
2272                if (ka->sa.sa_handler != SIG_DFL) {
2273                        /* Run the handler.  */
2274                        *return_ka = *ka;
2275
2276                        if (ka->sa.sa_flags & SA_ONESHOT)
2277                                ka->sa.sa_handler = SIG_DFL;
2278
2279                        break; /* will return non-zero "signr" value */
2280                }
2281
2282                /*
2283                 * Now we are doing the default action for this signal.
2284                 */
2285                if (sig_kernel_ignore(signr)) /* Default is nothing. */
2286                        continue;
2287
2288                /*
2289                 * Global init gets no signals it doesn't want.
2290                 * Container-init gets no signals it doesn't want from same
2291                 * container.
2292                 *
2293                 * Note that if global/container-init sees a sig_kernel_only()
2294                 * signal here, the signal must have been generated internally
2295                 * or must have come from an ancestor namespace. In either
2296                 * case, the signal cannot be dropped.
2297                 */
2298                if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2299                                !sig_kernel_only(signr))
2300                        continue;
2301
2302                if (sig_kernel_stop(signr)) {
2303                        /*
2304                         * The default action is to stop all threads in
2305                         * the thread group.  The job control signals
2306                         * do nothing in an orphaned pgrp, but SIGSTOP
2307                         * always works.  Note that siglock needs to be
2308                         * dropped during the call to is_orphaned_pgrp()
2309                         * because of lock ordering with tasklist_lock.
2310                         * This allows an intervening SIGCONT to be posted.
2311                         * We need to check for that and bail out if necessary.
2312                         */
2313                        if (signr != SIGSTOP) {
2314                                spin_unlock_irq(&sighand->siglock);
2315
2316                                /* signals can be posted during this window */
2317
2318                                if (is_current_pgrp_orphaned())
2319                                        goto relock;
2320
2321                                spin_lock_irq(&sighand->siglock);
2322                        }
2323
2324                        if (likely(do_signal_stop(info->si_signo))) {
2325                                /* It released the siglock.  */
2326                                goto relock;
2327                        }
2328
2329                        /*
2330                         * We didn't actually stop, due to a race
2331                         * with SIGCONT or something like that.
2332                         */
2333                        continue;
2334                }
2335
2336                spin_unlock_irq(&sighand->siglock);
2337
2338                /*
2339                 * Anything else is fatal, maybe with a core dump.
2340                 */
2341                current->flags |= PF_SIGNALED;
2342
2343                if (sig_kernel_coredump(signr)) {
2344                        if (print_fatal_signals)
2345                                print_fatal_signal(info->si_signo);
2346                        /*
2347                         * If it was able to dump core, this kills all
2348                         * other threads in the group and synchronizes with
2349                         * their demise.  If we lost the race with another
2350                         * thread getting here, it set group_exit_code
2351                         * first and our do_group_exit call below will use
2352                         * that value and ignore the one we pass it.
2353                         */
2354                        do_coredump(info);
2355                }
2356
2357                /*
2358                 * Death signals, no core dump.
2359                 */
2360                do_group_exit(info->si_signo);
2361                /* NOTREACHED */
2362        }
2363        spin_unlock_irq(&sighand->siglock);
2364        return signr;
2365}
2366
2367/**
2368 * signal_delivered - 
2369 * @sig:                number of signal being delivered
2370 * @info:               siginfo_t of signal being delivered
2371 * @ka:                 sigaction setting that chose the handler
2372 * @regs:               user register state
2373 * @stepping:           nonzero if debugger single-step or block-step in use
2374 *
2375 * This function should be called when a signal has succesfully been
2376 * delivered. It updates the blocked signals accordingly (@ka->sa.sa_mask
2377 * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2378 * is set in @ka->sa.sa_flags.  Tracing is notified.
2379 */
2380void signal_delivered(int sig, siginfo_t *info, struct k_sigaction *ka,
2381                        struct pt_regs *regs, int stepping)
2382{
2383        sigset_t blocked;
2384
2385        /* A signal was successfully delivered, and the
2386           saved sigmask was stored on the signal frame,
2387           and will be restored by sigreturn.  So we can
2388           simply clear the restore sigmask flag.  */
2389        clear_restore_sigmask();
2390
2391        sigorsets(&blocked, &current->blocked, &ka->sa.sa_mask);
2392        if (!(ka->sa.sa_flags & SA_NODEFER))
2393                sigaddset(&blocked, sig);
2394        set_current_blocked(&blocked);
2395        tracehook_signal_handler(sig, info, ka, regs, stepping);
2396}
2397
2398/*
2399 * It could be that complete_signal() picked us to notify about the
2400 * group-wide signal. Other threads should be notified now to take
2401 * the shared signals in @which since we will not.
2402 */
2403static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2404{
2405        sigset_t retarget;
2406        struct task_struct *t;
2407
2408        sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2409        if (sigisemptyset(&retarget))
2410                return;
2411
2412        t = tsk;
2413        while_each_thread(tsk, t) {
2414                if (t->flags & PF_EXITING)
2415                        continue;
2416
2417                if (!has_pending_signals(&retarget, &t->blocked))
2418                        continue;
2419                /* Remove the signals this thread can handle. */
2420                sigandsets(&retarget, &retarget, &t->blocked);
2421
2422                if (!signal_pending(t))
2423                        signal_wake_up(t, 0);
2424
2425                if (sigisemptyset(&retarget))
2426                        break;
2427        }
2428}
2429
2430void exit_signals(struct task_struct *tsk)
2431{
2432        int group_stop = 0;
2433        sigset_t unblocked;
2434
2435        /*
2436         * @tsk is about to have PF_EXITING set - lock out users which
2437         * expect stable threadgroup.
2438         */
2439        threadgroup_change_begin(tsk);
2440
2441        if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2442                tsk->flags |= PF_EXITING;
2443                threadgroup_change_end(tsk);
2444                return;
2445        }
2446
2447        spin_lock_irq(&tsk->sighand->siglock);
2448        /*
2449         * From now this task is not visible for group-wide signals,
2450         * see wants_signal(), do_signal_stop().
2451         */
2452        tsk->flags |= PF_EXITING;
2453
2454        threadgroup_change_end(tsk);
2455
2456        if (!signal_pending(tsk))
2457                goto out;
2458
2459        unblocked = tsk->blocked;
2460        signotset(&unblocked);
2461        retarget_shared_pending(tsk, &unblocked);
2462
2463        if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2464            task_participate_group_stop(tsk))
2465                group_stop = CLD_STOPPED;
2466out:
2467        spin_unlock_irq(&tsk->sighand->siglock);
2468
2469        /*
2470         * If group stop has completed, deliver the notification.  This
2471         * should always go to the real parent of the group leader.
2472         */
2473        if (unlikely(group_stop)) {
2474                read_lock(&tasklist_lock);
2475                do_notify_parent_cldstop(tsk, false, group_stop);
2476                read_unlock(&tasklist_lock);
2477        }
2478}
2479
2480EXPORT_SYMBOL(recalc_sigpending);
2481EXPORT_SYMBOL_GPL(dequeue_signal);
2482EXPORT_SYMBOL(flush_signals);
2483EXPORT_SYMBOL(force_sig);
2484EXPORT_SYMBOL(send_sig);
2485EXPORT_SYMBOL(send_sig_info);
2486EXPORT_SYMBOL(sigprocmask);
2487EXPORT_SYMBOL(block_all_signals);
2488EXPORT_SYMBOL(unblock_all_signals);
2489
2490
2491/*
2492 * System call entry points.
2493 */
2494
2495/**
2496 *  sys_restart_syscall - restart a system call
2497 */
2498SYSCALL_DEFINE0(restart_syscall)
2499{
2500        struct restart_block *restart = &current_thread_info()->restart_block;
2501        return restart->fn(restart);
2502}
2503
2504long do_no_restart_syscall(struct restart_block *param)
2505{
2506        return -EINTR;
2507}
2508
2509static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2510{
2511        if (signal_pending(tsk) && !thread_group_empty(tsk)) {
2512                sigset_t newblocked;
2513                /* A set of now blocked but previously unblocked signals. */
2514                sigandnsets(&newblocked, newset, &current->blocked);
2515                retarget_shared_pending(tsk, &newblocked);
2516        }
2517        tsk->blocked = *newset;
2518        recalc_sigpending();
2519}
2520
2521/**
2522 * set_current_blocked - change current->blocked mask
2523 * @newset: new mask
2524 *
2525 * It is wrong to change ->blocked directly, this helper should be used
2526 * to ensure the process can't miss a shared signal we are going to block.
2527 */
2528void set_current_blocked(sigset_t *newset)
2529{
2530        sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2531        __set_current_blocked(newset);
2532}
2533
2534void __set_current_blocked(const sigset_t *newset)
2535{
2536        struct task_struct *tsk = current;
2537
2538        spin_lock_irq(&tsk->sighand->siglock);
2539        __set_task_blocked(tsk, newset);
2540        spin_unlock_irq(&tsk->sighand->siglock);
2541}
2542
2543/*
2544 * This is also useful for kernel threads that want to temporarily
2545 * (or permanently) block certain signals.
2546 *
2547 * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2548 * interface happily blocks "unblockable" signals like SIGKILL
2549 * and friends.
2550 */
2551int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2552{
2553        struct task_struct *tsk = current;
2554        sigset_t newset;
2555
2556        /* Lockless, only current can change ->blocked, never from irq */
2557        if (oldset)
2558                *oldset = tsk->blocked;
2559
2560        switch (how) {
2561        case SIG_BLOCK:
2562                sigorsets(&newset, &tsk->blocked, set);
2563                break;
2564        case SIG_UNBLOCK:
2565                sigandnsets(&newset, &tsk->blocked, set);
2566                break;
2567        case SIG_SETMASK:
2568                newset = *set;
2569                break;
2570        default:
2571                return -EINVAL;
2572        }
2573
2574        __set_current_blocked(&newset);
2575        return 0;
2576}
2577
2578/**
2579 *  sys_rt_sigprocmask - change the list of currently blocked signals
2580 *  @how: whether to add, remove, or set signals
2581 *  @nset: stores pending signals
2582 *  @oset: previous value of signal mask if non-null
2583 *  @sigsetsize: size of sigset_t type
2584 */
2585SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
2586                sigset_t __user *, oset, size_t, sigsetsize)
2587{
2588        sigset_t old_set, new_set;
2589        int error;
2590
2591        /* XXX: Don't preclude handling different sized sigset_t's.  */
2592        if (sigsetsize != sizeof(sigset_t))
2593                return -EINVAL;
2594
2595        old_set = current->blocked;
2596
2597        if (nset) {
2598                if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
2599                        return -EFAULT;
2600                sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
2601
2602                error = sigprocmask(how, &new_set, NULL);
2603                if (error)
2604                        return error;
2605        }
2606
2607        if (oset) {
2608                if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
2609                        return -EFAULT;
2610        }
2611
2612        return 0;
2613}
2614
2615long do_sigpending(void __user *set, unsigned long sigsetsize)
2616{
2617        long error = -EINVAL;
2618        sigset_t pending;
2619
2620        if (sigsetsize > sizeof(sigset_t))
2621                goto out;
2622
2623        spin_lock_irq(&current->sighand->siglock);
2624        sigorsets(&pending, &current->pending.signal,
2625                  &current->signal->shared_pending.signal);
2626        spin_unlock_irq(&current->sighand->siglock);
2627
2628        /* Outside the lock because only this thread touches it.  */
2629        sigandsets(&pending, &current->blocked, &pending);
2630
2631        error = -EFAULT;
2632        if (!copy_to_user(set, &pending, sigsetsize))
2633                error = 0;
2634
2635out:
2636        return error;
2637}
2638
2639/**
2640 *  sys_rt_sigpending - examine a pending signal that has been raised
2641 *                      while blocked
2642 *  @set: stores pending signals
2643 *  @sigsetsize: size of sigset_t type or larger
2644 */
2645SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, set, size_t, sigsetsize)
2646{
2647        return do_sigpending(set, sigsetsize);
2648}
2649
2650#ifndef HAVE_ARCH_COPY_SIGINFO_TO_USER
2651
2652int copy_siginfo_to_user(siginfo_t __user *to, siginfo_t *from)
2653{
2654        int err;
2655
2656        if (!access_ok (VERIFY_WRITE, to, sizeof(siginfo_t)))
2657                return -EFAULT;
2658        if (from->si_code < 0)
2659                return __copy_to_user(to, from, sizeof(siginfo_t))
2660                        ? -EFAULT : 0;
2661        /*
2662         * If you change siginfo_t structure, please be sure
2663         * this code is fixed accordingly.
2664         * Please remember to update the signalfd_copyinfo() function
2665         * inside fs/signalfd.c too, in case siginfo_t changes.
2666         * It should never copy any pad contained in the structure
2667         * to avoid security leaks, but must copy the generic
2668         * 3 ints plus the relevant union member.
2669         */
2670        err = __put_user(from->si_signo, &to->si_signo);
2671        err |= __put_user(from->si_errno, &to->si_errno);
2672        err |= __put_user((short)from->si_code, &to->si_code);
2673        switch (from->si_code & __SI_MASK) {
2674        case __SI_KILL:
2675                err |= __put_user(from->si_pid, &to->si_pid);
2676                err |= __put_user(from->si_uid, &to->si_uid);
2677                break;
2678        case __SI_TIMER:
2679                 err |= __put_user(from->si_tid, &to->si_tid);
2680                 err |= __put_user(from->si_overrun, &to->si_overrun);
2681                 err |= __put_user(from->si_ptr, &to->si_ptr);
2682                break;
2683        case __SI_POLL:
2684                err |= __put_user(from->si_band, &to->si_band);
2685                err |= __put_user(from->si_fd, &to->si_fd);
2686                break;
2687        case __SI_FAULT:
2688                err |= __put_user(from->si_addr, &to->si_addr);
2689#ifdef __ARCH_SI_TRAPNO
2690                err |= __put_user(from->si_trapno, &to->si_trapno);
2691#endif
2692#ifdef BUS_MCEERR_AO
2693                /*
2694                 * Other callers might not initialize the si_lsb field,
2695                 * so check explicitly for the right codes here.
2696                 */
2697                if (from->si_code == BUS_MCEERR_AR || from->si_code == BUS_MCEERR_AO)
2698                        err |= __put_user(from->si_addr_lsb, &to->si_addr_lsb);
2699#endif
2700                break;
2701        case __SI_CHLD:
2702                err |= __put_user(from->si_pid, &to->si_pid);
2703                err |= __put_user(from->si_uid, &to->si_uid);
2704                err |= __put_user(from->si_status, &to->si_status);
2705                err |= __put_user(from->si_utime, &to->si_utime);
2706                err |= __put_user(from->si_stime, &to->si_stime);
2707                break;
2708        case __SI_RT: /* This is not generated by the kernel as of now. */
2709        case __SI_MESGQ: /* But this is */
2710                err |= __put_user(from->si_pid, &to->si_pid);
2711                err |= __put_user(from->si_uid, &to->si_uid);
2712                err |= __put_user(from->si_ptr, &to->si_ptr);
2713                break;
2714#ifdef __ARCH_SIGSYS
2715        case __SI_SYS:
2716                err |= __put_user(from->si_call_addr, &to->si_call_addr);
2717                err |= __put_user(from->si_syscall, &to->si_syscall);
2718                err |= __put_user(from->si_arch, &to->si_arch);
2719                break;
2720#endif
2721        default: /* this is just in case for now ... */
2722                err |= __put_user(from->si_pid, &to->si_pid);
2723                err |= __put_user(from->si_uid, &to->si_uid);
2724                break;
2725        }
2726        return err;
2727}
2728
2729#endif
2730
2731/**
2732 *  do_sigtimedwait - wait for queued signals specified in @which
2733 *  @which: queued signals to wait for
2734 *  @info: if non-null, the signal's siginfo is returned here
2735 *  @ts: upper bound on process time suspension
2736 */
2737int do_sigtimedwait(const sigset_t *which, siginfo_t *info,
2738                        const struct timespec *ts)
2739{
2740        struct task_struct *tsk = current;
2741        long timeout = MAX_SCHEDULE_TIMEOUT;
2742        sigset_t mask = *which;
2743        int sig;
2744
2745        if (ts) {
2746                if (!timespec_valid(ts))
2747                        return -EINVAL;
2748                timeout = timespec_to_jiffies(ts);
2749                /*
2750                 * We can be close to the next tick, add another one
2751                 * to ensure we will wait at least the time asked for.
2752                 */
2753                if (ts->tv_sec || ts->tv_nsec)
2754                        timeout++;
2755        }
2756
2757        /*
2758         * Invert the set of allowed signals to get those we want to block.
2759         */
2760        sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
2761        signotset(&mask);
2762
2763        spin_lock_irq(&tsk->sighand->siglock);
2764        sig = dequeue_signal(tsk, &mask, info);
2765        if (!sig && timeout) {
2766                /*
2767                 * None ready, temporarily unblock those we're interested
2768                 * while we are sleeping in so that we'll be awakened when
2769                 * they arrive. Unblocking is always fine, we can avoid
2770                 * set_current_blocked().
2771                 */
2772                tsk->real_blocked = tsk->blocked;
2773                sigandsets(&tsk->blocked, &tsk->blocked, &mask);
2774                recalc_sigpending();
2775                spin_unlock_irq(&tsk->sighand->siglock);
2776
2777                timeout = schedule_timeout_interruptible(timeout);
2778
2779                spin_lock_irq(&tsk->sighand->siglock);
2780                __set_task_blocked(tsk, &tsk->real_blocked);
2781                siginitset(&tsk->real_blocked, 0);
2782                sig = dequeue_signal(tsk, &mask, info);
2783        }
2784        spin_unlock_irq(&tsk->sighand->siglock);
2785
2786        if (sig)
2787                return sig;
2788        return timeout ? -EINTR : -EAGAIN;
2789}
2790
2791/**
2792 *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
2793 *                      in @uthese
2794 *  @uthese: queued signals to wait for
2795 *  @uinfo: if non-null, the signal's siginfo is returned here
2796 *  @uts: upper bound on process time suspension
2797 *  @sigsetsize: size of sigset_t type
2798 */
2799SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
2800                siginfo_t __user *, uinfo, const struct timespec __user *, uts,
2801                size_t, sigsetsize)
2802{
2803        sigset_t these;
2804        struct timespec ts;
2805        siginfo_t info;
2806        int ret;
2807
2808        /* XXX: Don't preclude handling different sized sigset_t's.  */
2809        if (sigsetsize != sizeof(sigset_t))
2810                return -EINVAL;
2811
2812        if (copy_from_user(&these, uthese, sizeof(these)))
2813                return -EFAULT;
2814
2815        if (uts) {
2816                if (copy_from_user(&ts, uts, sizeof(ts)))
2817                        return -EFAULT;
2818        }
2819
2820        ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
2821
2822        if (ret > 0 && uinfo) {
2823                if (copy_siginfo_to_user(uinfo, &info))
2824                        ret = -EFAULT;
2825        }
2826
2827        return ret;
2828}
2829
2830/**
2831 *  sys_kill - send a signal to a process
2832 *  @pid: the PID of the process
2833 *  @sig: signal to be sent
2834 */
2835SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
2836{
2837        struct siginfo info;
2838
2839        info.si_signo = sig;
2840        info.si_errno = 0;
2841        info.si_code = SI_USER;
2842        info.si_pid = task_tgid_vnr(current);
2843        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2844
2845        return kill_something_info(sig, &info, pid);
2846}
2847
2848static int
2849do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info)
2850{
2851        struct task_struct *p;
2852        int error = -ESRCH;
2853
2854        rcu_read_lock();
2855        p = find_task_by_vpid(pid);
2856        if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
2857                error = check_kill_permission(sig, info, p);
2858                /*
2859                 * The null signal is a permissions and process existence
2860                 * probe.  No signal is actually delivered.
2861                 */
2862                if (!error && sig) {
2863                        error = do_send_sig_info(sig, info, p, false);
2864                        /*
2865                         * If lock_task_sighand() failed we pretend the task
2866                         * dies after receiving the signal. The window is tiny,
2867                         * and the signal is private anyway.
2868                         */
2869                        if (unlikely(error == -ESRCH))
2870                                error = 0;
2871                }
2872        }
2873        rcu_read_unlock();
2874
2875        return error;
2876}
2877
2878static int do_tkill(pid_t tgid, pid_t pid, int sig)
2879{
2880        struct siginfo info;
2881
2882        info.si_signo = sig;
2883        info.si_errno = 0;
2884        info.si_code = SI_TKILL;
2885        info.si_pid = task_tgid_vnr(current);
2886        info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2887
2888        return do_send_specific(tgid, pid, sig, &info);
2889}
2890
2891/**
2892 *  sys_tgkill - send signal to one specific thread
2893 *  @tgid: the thread group ID of the thread
2894 *  @pid: the PID of the thread
2895 *  @sig: signal to be sent
2896 *
2897 *  This syscall also checks the @tgid and returns -ESRCH even if the PID
2898 *  exists but it's not belonging to the target process anymore. This
2899 *  method solves the problem of threads exiting and PIDs getting reused.
2900 */
2901SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
2902{
2903        /* This is only valid for single tasks */
2904        if (pid <= 0 || tgid <= 0)
2905                return -EINVAL;
2906
2907        return do_tkill(tgid, pid, sig);
2908}
2909
2910/**
2911 *  sys_tkill - send signal to one specific task
2912 *  @pid: the PID of the task
2913 *  @sig: signal to be sent
2914 *
2915 *  Send a signal to only one task, even if it's a CLONE_THREAD task.
2916 */
2917SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
2918{
2919        /* This is only valid for single tasks */
2920        if (pid <= 0)
2921                return -EINVAL;
2922
2923        return do_tkill(0, pid, sig);
2924}
2925
2926/**
2927 *  sys_rt_sigqueueinfo - send signal information to a signal
2928 *  @pid: the PID of the thread
2929 *  @sig: signal to be sent
2930 *  @uinfo: signal info to be sent
2931 */
2932SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
2933                siginfo_t __user *, uinfo)
2934{
2935        siginfo_t info;
2936
2937        if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
2938                return -EFAULT;
2939
2940        /* Not even root can pretend to send signals from the kernel.
2941         * Nor can they impersonate a kill()/tgkill(), which adds source info.
2942         */
2943        if (info.si_code >= 0 || info.si_code == SI_TKILL) {
2944                /* We used to allow any < 0 si_code */
2945                WARN_ON_ONCE(info.si_code < 0);
2946                return -EPERM;
2947        }
2948        info.si_signo = sig;
2949
2950        /* POSIX.1b doesn't mention process groups.  */
2951        return kill_proc_info(sig, &info, pid);
2952}
2953
2954long do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, siginfo_t *info)
2955{
2956        /* This is only valid for single tasks */
2957        if (pid <= 0 || tgid <= 0)
2958                return -EINVAL;
2959
2960        /* Not even root can pretend to send signals from the kernel.
2961         * Nor can they impersonate a kill()/tgkill(), which adds source info.
2962         */
2963        if (info->si_code >= 0 || info->si_code == SI_TKILL) {
2964                /* We used to allow any < 0 si_code */
2965                WARN_ON_ONCE(info->si_code < 0);
2966                return -EPERM;
2967        }
2968        info->si_signo = sig;
2969
2970        return do_send_specific(tgid, pid, sig, info);
2971}
2972
2973SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
2974                siginfo_t __user *, uinfo)
2975{
2976        siginfo_t info;
2977
2978        if (copy_from_user(&info, uinfo, sizeof(siginfo_t)))
2979                return -EFAULT;
2980
2981        return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
2982}
2983
2984int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
2985{
2986        struct task_struct *t = current;
2987        struct k_sigaction *k;
2988        sigset_t mask;
2989
2990        if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
2991                return -EINVAL;
2992
2993        k = &t->sighand->action[sig-1];
2994
2995        spin_lock_irq(&current->sighand->siglock);
2996        if (oact)
2997                *oact = *k;
2998
2999        if (act) {
3000                sigdelsetmask(&act->sa.sa_mask,
3001                              sigmask(SIGKILL) | sigmask(SIGSTOP));
3002                *k = *act;
3003                /*
3004                 * POSIX 3.3.1.3:
3005                 *  "Setting a signal action to SIG_IGN for a signal that is
3006                 *   pending shall cause the pending signal to be discarded,
3007                 *   whether or not it is blocked."
3008                 *
3009                 *  "Setting a signal action to SIG_DFL for a signal that is
3010                 *   pending and whose default action is to ignore the signal
3011                 *   (for example, SIGCHLD), shall cause the pending signal to
3012                 *   be discarded, whether or not it is blocked"
3013                 */
3014                if (sig_handler_ignored(sig_handler(t, sig), sig)) {
3015                        sigemptyset(&mask);
3016                        sigaddset(&mask, sig);
3017                        rm_from_queue_full(&mask, &t->signal->shared_pending);
3018                        do {
3019                                rm_from_queue_full(&mask, &t->pending);
3020                                t = next_thread(t);
3021                        } while (t != current);
3022                }
3023        }
3024
3025        spin_unlock_irq(&current->sighand->siglock);
3026        return 0;
3027}
3028
3029int 
3030do_sigaltstack (const stack_t __user *uss, stack_t __user *uoss, unsigned long sp)
3031{
3032        stack_t oss;
3033        int error;
3034
3035        oss.ss_sp = (void __user *) current->sas_ss_sp;
3036        oss.ss_size = current->sas_ss_size;
3037        oss.ss_flags = sas_ss_flags(sp);
3038
3039        if (uss) {
3040                void __user *ss_sp;
3041                size_t ss_size;
3042                int ss_flags;
3043
3044                error = -EFAULT;
3045                if (!access_ok(VERIFY_READ, uss, sizeof(*uss)))
3046                        goto out;
3047                error = __get_user(ss_sp, &uss->ss_sp) |
3048                        __get_user(ss_flags, &uss->ss_flags) |
3049                        __get_user(ss_size, &uss->ss_size);
3050                if (error)
3051                        goto out;
3052
3053                error = -EPERM;
3054                if (on_sig_stack(sp))
3055                        goto out;
3056
3057                error = -EINVAL;
3058                /*
3059                 * Note - this code used to test ss_flags incorrectly:
3060                 *        old code may have been written using ss_flags==0
3061                 *        to mean ss_flags==SS_ONSTACK (as this was the only
3062                 *        way that worked) - this fix preserves that older
3063                 *        mechanism.
3064                 */
3065                if (ss_flags != SS_DISABLE && ss_flags != SS_ONSTACK && ss_flags != 0)
3066                        goto out;
3067
3068                if (ss_flags == SS_DISABLE) {
3069                        ss_size = 0;
3070                        ss_sp = NULL;
3071                } else {
3072                        error = -ENOMEM;
3073                        if (ss_size < MINSIGSTKSZ)
3074                                goto out;
3075                }
3076
3077                current->sas_ss_sp = (unsigned long) ss_sp;
3078                current->sas_ss_size = ss_size;
3079        }
3080
3081        error = 0;
3082        if (uoss) {
3083                error = -EFAULT;
3084                if (!access_ok(VERIFY_WRITE, uoss, sizeof(*uoss)))
3085                        goto out;
3086                error = __put_user(oss.ss_sp, &uoss->ss_sp) |
3087                        __put_user(oss.ss_size, &uoss->ss_size) |
3088                        __put_user(oss.ss_flags, &uoss->ss_flags);
3089        }
3090
3091out:
3092        return error;
3093}
3094#ifdef CONFIG_GENERIC_SIGALTSTACK
3095SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
3096{
3097        return do_sigaltstack(uss, uoss, current_user_stack_pointer());
3098}
3099#endif
3100
3101int restore_altstack(const stack_t __user *uss)
3102{
3103        int err = do_sigaltstack(uss, NULL, current_user_stack_pointer());
3104        /* squash all but EFAULT for now */
3105        return err == -EFAULT ? err : 0;
3106}
3107
3108int __save_altstack(stack_t __user *uss, unsigned long sp)
3109{
3110        struct task_struct *t = current;
3111        return  __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
3112                __put_user(sas_ss_flags(sp), &uss->ss_flags) |
3113                __put_user(t->sas_ss_size, &uss->ss_size);
3114}
3115
3116#ifdef CONFIG_COMPAT
3117#ifdef CONFIG_GENERIC_SIGALTSTACK
3118COMPAT_SYSCALL_DEFINE2(sigaltstack,
3119                        const compat_stack_t __user *, uss_ptr,
3120                        compat_stack_t __user *, uoss_ptr)
3121{
3122        stack_t uss, uoss;
3123        int ret;
3124        mm_segment_t seg;
3125
3126        if (uss_ptr) {
3127                compat_stack_t uss32;
3128
3129                memset(&uss, 0, sizeof(stack_t));
3130                if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
3131                        return -EFAULT;
3132                uss.ss_sp = compat_ptr(uss32.ss_sp);
3133                uss.ss_flags = uss32.ss_flags;
3134                uss.ss_size = uss32.ss_size;
3135        }
3136        seg = get_fs();
3137        set_fs(KERNEL_DS);
3138        ret = do_sigaltstack((stack_t __force __user *) (uss_ptr ? &uss : NULL),
3139                             (stack_t __force __user *) &uoss,
3140                             compat_user_stack_pointer());
3141        set_fs(seg);
3142        if (ret >= 0 && uoss_ptr)  {
3143                if (!access_ok(VERIFY_WRITE, uoss_ptr, sizeof(compat_stack_t)) ||
3144                    __put_user(ptr_to_compat(uoss.ss_sp), &uoss_ptr->ss_sp) ||
3145                    __put_user(uoss.ss_flags, &uoss_ptr->ss_flags) ||
3146                    __put_user(uoss.ss_size, &uoss_ptr->ss_size))
3147                        ret = -EFAULT;
3148        }
3149        return ret;
3150}
3151
3152int compat_restore_altstack(const compat_stack_t __user *uss)
3153{
3154        int err = compat_sys_sigaltstack(uss, NULL);
3155        /* squash all but -EFAULT for now */
3156        return err == -EFAULT ? err : 0;
3157}
3158
3159int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
3160{
3161        struct task_struct *t = current;
3162        return  __put_user(ptr_to_compat((void __user *)t->sas_ss_sp), &uss->ss_sp) |
3163                __put_user(sas_ss_flags(sp), &uss->ss_flags) |
3164                __put_user(t->sas_ss_size, &uss->ss_size);
3165}
3166#endif
3167#endif
3168
3169#ifdef __ARCH_WANT_SYS_SIGPENDING
3170
3171/**
3172 *  sys_sigpending - examine pending signals
3173 *  @set: where mask of pending signal is returned
3174 */
3175SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, set)
3176{
3177        return do_sigpending(set, sizeof(*set));
3178}
3179
3180#endif
3181
3182#ifdef __ARCH_WANT_SYS_SIGPROCMASK
3183/**
3184 *  sys_sigprocmask - examine and change blocked signals
3185 *  @how: whether to add, remove, or set signals
3186 *  @nset: signals to add or remove (if non-null)
3187 *  @oset: previous value of signal mask if non-null
3188 *
3189 * Some platforms have their own version with special arguments;
3190 * others support only sys_rt_sigprocmask.
3191 */
3192
3193SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
3194                old_sigset_t __user *, oset)
3195{
3196        old_sigset_t old_set, new_set;
3197        sigset_t new_blocked;
3198
3199        old_set = current->blocked.sig[0];
3200
3201        if (nset) {
3202                if (copy_from_user(&new_set, nset, sizeof(*nset)))
3203                        return -EFAULT;
3204
3205                new_blocked = current->blocked;
3206
3207                switch (how) {
3208                case SIG_BLOCK:
3209                        sigaddsetmask(&new_blocked, new_set);
3210                        break;
3211                case SIG_UNBLOCK:
3212                        sigdelsetmask(&new_blocked, new_set);
3213                        break;
3214                case SIG_SETMASK:
3215                        new_blocked.sig[0] = new_set;
3216                        break;
3217                default:
3218                        return -EINVAL;
3219                }
3220
3221                set_current_blocked(&new_blocked);
3222        }
3223
3224        if (oset) {
3225                if (copy_to_user(oset, &old_set, sizeof(*oset)))
3226                        return -EFAULT;
3227        }
3228
3229        return 0;
3230}
3231#endif /* __ARCH_WANT_SYS_SIGPROCMASK */
3232
3233#ifdef __ARCH_WANT_SYS_RT_SIGACTION
3234/**
3235 *  sys_rt_sigaction - alter an action taken by a process
3236 *  @sig: signal to be sent
3237 *  @act: new sigaction
3238 *  @oact: used to save the previous sigaction
3239 *  @sigsetsize: size of sigset_t type
3240 */
3241SYSCALL_DEFINE4(rt_sigaction, int, sig,
3242                const struct sigaction __user *, act,
3243                struct sigaction __user *, oact,
3244                size_t, sigsetsize)
3245{
3246        struct k_sigaction new_sa, old_sa;
3247        int ret = -EINVAL;
3248
3249        /* XXX: Don't preclude handling different sized sigset_t's.  */
3250        if (sigsetsize != sizeof(sigset_t))
3251                goto out;
3252
3253        if (act) {
3254                if (copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
3255                        return -EFAULT;
3256        }
3257
3258        ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
3259
3260        if (!ret && oact) {
3261                if (copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
3262                        return -EFAULT;
3263        }
3264out:
3265        return ret;
3266}
3267#endif /* __ARCH_WANT_SYS_RT_SIGACTION */
3268
3269#ifdef __ARCH_WANT_SYS_SGETMASK
3270
3271/*
3272 * For backwards compatibility.  Functionality superseded by sigprocmask.
3273 */
3274SYSCALL_DEFINE0(sgetmask)
3275{
3276        /* SMP safe */
3277        return current->blocked.sig[0];
3278}
3279
3280SYSCALL_DEFINE1(ssetmask, int, newmask)
3281{
3282        int old = current->blocked.sig[0];
3283        sigset_t newset;
3284
3285        siginitset(&newset, newmask);
3286        set_current_blocked(&newset);
3287
3288        return old;
3289}
3290#endif /* __ARCH_WANT_SGETMASK */
3291
3292#ifdef __ARCH_WANT_SYS_SIGNAL
3293/*
3294 * For backwards compatibility.  Functionality superseded by sigaction.
3295 */
3296SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
3297{
3298        struct k_sigaction new_sa, old_sa;
3299        int ret;
3300
3301        new_sa.sa.sa_handler = handler;
3302        new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
3303        sigemptyset(&new_sa.sa.sa_mask);
3304
3305        ret = do_sigaction(sig, &new_sa, &old_sa);
3306
3307        return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
3308}
3309#endif /* __ARCH_WANT_SYS_SIGNAL */
3310
3311#ifdef __ARCH_WANT_SYS_PAUSE
3312
3313SYSCALL_DEFINE0(pause)
3314{
3315        while (!signal_pending(current)) {
3316                current->state = TASK_INTERRUPTIBLE;
3317                schedule();
3318        }
3319        return -ERESTARTNOHAND;
3320}
3321
3322#endif
3323
3324int sigsuspend(sigset_t *set)
3325{
3326        current->saved_sigmask = current->blocked;
3327        set_current_blocked(set);
3328
3329        current->state = TASK_INTERRUPTIBLE;
3330        schedule();
3331        set_restore_sigmask();
3332        return -ERESTARTNOHAND;
3333}
3334
3335#ifdef __ARCH_WANT_SYS_RT_SIGSUSPEND
3336/**
3337 *  sys_rt_sigsuspend - replace the signal mask for a value with the
3338 *      @unewset value until a signal is received
3339 *  @unewset: new signal mask value
3340 *  @sigsetsize: size of sigset_t type
3341 */
3342SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
3343{
3344        sigset_t newset;
3345
3346        /* XXX: Don't preclude handling different sized sigset_t's.  */
3347        if (sigsetsize != sizeof(sigset_t))
3348                return -EINVAL;
3349
3350        if (copy_from_user(&newset, unewset, sizeof(newset)))
3351                return -EFAULT;
3352        return sigsuspend(&newset);
3353}
3354#endif /* __ARCH_WANT_SYS_RT_SIGSUSPEND */
3355
3356__attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma)
3357{
3358        return NULL;
3359}
3360
3361void __init signals_init(void)
3362{
3363        sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
3364}
3365
3366#ifdef CONFIG_KGDB_KDB
3367#include <linux/kdb.h>
3368/*
3369 * kdb_send_sig_info - Allows kdb to send signals without exposing
3370 * signal internals.  This function checks if the required locks are
3371 * available before calling the main signal code, to avoid kdb
3372 * deadlocks.
3373 */
3374void
3375kdb_send_sig_info(struct task_struct *t, struct siginfo *info)
3376{
3377        static struct task_struct *kdb_prev_t;
3378        int sig, new_t;
3379        if (!spin_trylock(&t->sighand->siglock)) {
3380                kdb_printf("Can't do kill command now.\n"
3381                           "The sigmask lock is held somewhere else in "
3382                           "kernel, try again later\n");
3383                return;
3384        }
3385        spin_unlock(&t->sighand->siglock);
3386        new_t = kdb_prev_t != t;
3387        kdb_prev_t = t;
3388        if (t->state != TASK_RUNNING && new_t) {
3389                kdb_printf("Process is not RUNNING, sending a signal from "
3390                           "kdb risks deadlock\n"
3391                           "on the run queue locks. "
3392                           "The signal has _not_ been sent.\n"
3393                           "Reissue the kill command if you want to risk "
3394                           "the deadlock.\n");
3395                return;
3396        }
3397        sig = info->si_signo;
3398        if (send_sig_info(sig, info, t))
3399                kdb_printf("Fail to deliver Signal %d to process %d.\n",
3400                           sig, t->pid);
3401        else
3402                kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
3403}
3404#endif  /* CONFIG_KGDB_KDB */
3405
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.