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