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