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