kernel/exit.c

/* [<][>][^][v][top][bottom][index][help] */
This source file includes following definitions.
__unhash_process
__exit_signal
delayed_put_task_struct
put_task_struct_rcu_user
release_task
rcuwait_wake_up
will_become_orphaned_pgrp
is_current_pgrp_orphaned
has_stopped_jobs
kill_orphaned_pgrp
mm_update_next_owner
exit_mm
find_alive_thread
find_child_reaper
find_new_reaper
reparent_leader
forget_original_parent
exit_notify
check_stack_usage
check_stack_usage
do_exit
complete_and_exit
SYSCALL_DEFINE1
do_group_exit
SYSCALL_DEFINE1
eligible_pid
eligible_child
wait_task_zombie
task_stopped_code
wait_task_stopped
wait_task_continued
wait_consider_task
do_wait_thread
ptrace_do_wait
child_wait_callback
__wake_up_parent
do_wait
pidfd_get_pid
kernel_waitid
SYSCALL_DEFINE5
kernel_wait4
SYSCALL_DEFINE4
SYSCALL_DEFINE3
COMPAT_SYSCALL_DEFINE4
COMPAT_SYSCALL_DEFINE5
abort
   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *  linux/kernel/exit.c
   4  *
   5  *  Copyright (C) 1991, 1992  Linus Torvalds
   6  */
   7 
   8 #include <linux/mm.h>
   9 #include <linux/slab.h>
  10 #include <linux/sched/autogroup.h>
  11 #include <linux/sched/mm.h>
  12 #include <linux/sched/stat.h>
  13 #include <linux/sched/task.h>
  14 #include <linux/sched/task_stack.h>
  15 #include <linux/sched/cputime.h>
  16 #include <linux/interrupt.h>
  17 #include <linux/module.h>
  18 #include <linux/capability.h>
  19 #include <linux/completion.h>
  20 #include <linux/personality.h>
  21 #include <linux/tty.h>
  22 #include <linux/iocontext.h>
  23 #include <linux/key.h>
  24 #include <linux/cpu.h>
  25 #include <linux/acct.h>
  26 #include <linux/tsacct_kern.h>
  27 #include <linux/file.h>
  28 #include <linux/fdtable.h>
  29 #include <linux/freezer.h>
  30 #include <linux/binfmts.h>
  31 #include <linux/nsproxy.h>
  32 #include <linux/pid_namespace.h>
  33 #include <linux/ptrace.h>
  34 #include <linux/profile.h>
  35 #include <linux/mount.h>
  36 #include <linux/proc_fs.h>
  37 #include <linux/kthread.h>
  38 #include <linux/mempolicy.h>
  39 #include <linux/taskstats_kern.h>
  40 #include <linux/delayacct.h>
  41 #include <linux/cgroup.h>
  42 #include <linux/syscalls.h>
  43 #include <linux/signal.h>
  44 #include <linux/posix-timers.h>
  45 #include <linux/cn_proc.h>
  46 #include <linux/mutex.h>
  47 #include <linux/futex.h>
  48 #include <linux/pipe_fs_i.h>
  49 #include <linux/audit.h> /* for audit_free() */
  50 #include <linux/resource.h>
  51 #include <linux/blkdev.h>
  52 #include <linux/task_io_accounting_ops.h>
  53 #include <linux/tracehook.h>
  54 #include <linux/fs_struct.h>
  55 #include <linux/init_task.h>
  56 #include <linux/perf_event.h>
  57 #include <trace/events/sched.h>
  58 #include <linux/hw_breakpoint.h>
  59 #include <linux/oom.h>
  60 #include <linux/writeback.h>
  61 #include <linux/shm.h>
  62 #include <linux/kcov.h>
  63 #include <linux/random.h>
  64 #include <linux/rcuwait.h>
  65 #include <linux/compat.h>
  66 
  67 #include <linux/uaccess.h>
  68 #include <asm/unistd.h>
  69 #include <asm/pgtable.h>
  70 #include <asm/mmu_context.h>
  71 
  72 static void __unhash_process(struct task_struct *p, bool group_dead)
  73 {
  74         nr_threads--;
  75         detach_pid(p, PIDTYPE_PID);
  76         if (group_dead) {
  77                 detach_pid(p, PIDTYPE_TGID);
  78                 detach_pid(p, PIDTYPE_PGID);
  79                 detach_pid(p, PIDTYPE_SID);
  80 
  81                 list_del_rcu(&p->tasks);
  82                 list_del_init(&p->sibling);
  83                 __this_cpu_dec(process_counts);
  84         }
  85         list_del_rcu(&p->thread_group);
  86         list_del_rcu(&p->thread_node);
  87 }
  88 
  89 /*
  90  * This function expects the tasklist_lock write-locked.
  91  */
  92 static void __exit_signal(struct task_struct *tsk)
  93 {
  94         struct signal_struct *sig = tsk->signal;
  95         bool group_dead = thread_group_leader(tsk);
  96         struct sighand_struct *sighand;
  97         struct tty_struct *uninitialized_var(tty);
  98         u64 utime, stime;
  99 
 100         sighand = rcu_dereference_check(tsk->sighand,
 101                                         lockdep_tasklist_lock_is_held());
 102         spin_lock(&sighand->siglock);
 103 
 104 #ifdef CONFIG_POSIX_TIMERS
 105         posix_cpu_timers_exit(tsk);
 106         if (group_dead) {
 107                 posix_cpu_timers_exit_group(tsk);
 108         } else {
 109                 /*
 110                  * This can only happen if the caller is de_thread().
 111                  * FIXME: this is the temporary hack, we should teach
 112                  * posix-cpu-timers to handle this case correctly.
 113                  */
 114                 if (unlikely(has_group_leader_pid(tsk)))
 115                         posix_cpu_timers_exit_group(tsk);
 116         }
 117 #endif
 118 
 119         if (group_dead) {
 120                 tty = sig->tty;
 121                 sig->tty = NULL;
 122         } else {
 123                 /*
 124                  * If there is any task waiting for the group exit
 125                  * then notify it:
 126                  */
 127                 if (sig->notify_count > 0 && !--sig->notify_count)
 128                         wake_up_process(sig->group_exit_task);
 129 
 130                 if (tsk == sig->curr_target)
 131                         sig->curr_target = next_thread(tsk);
 132         }
 133 
 134         add_device_randomness((const void*) &tsk->se.sum_exec_runtime,
 135                               sizeof(unsigned long long));
 136 
 137         /*
 138          * Accumulate here the counters for all threads as they die. We could
 139          * skip the group leader because it is the last user of signal_struct,
 140          * but we want to avoid the race with thread_group_cputime() which can
 141          * see the empty ->thread_head list.
 142          */
 143         task_cputime(tsk, &utime, &stime);
 144         write_seqlock(&sig->stats_lock);
 145         sig->utime += utime;
 146         sig->stime += stime;
 147         sig->gtime += task_gtime(tsk);
 148         sig->min_flt += tsk->min_flt;
 149         sig->maj_flt += tsk->maj_flt;
 150         sig->nvcsw += tsk->nvcsw;
 151         sig->nivcsw += tsk->nivcsw;
 152         sig->inblock += task_io_get_inblock(tsk);
 153         sig->oublock += task_io_get_oublock(tsk);
 154         task_io_accounting_add(&sig->ioac, &tsk->ioac);
 155         sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
 156         sig->nr_threads--;
 157         __unhash_process(tsk, group_dead);
 158         write_sequnlock(&sig->stats_lock);
 159 
 160         /*
 161          * Do this under ->siglock, we can race with another thread
 162          * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
 163          */
 164         flush_sigqueue(&tsk->pending);
 165         tsk->sighand = NULL;
 166         spin_unlock(&sighand->siglock);
 167 
 168         __cleanup_sighand(sighand);
 169         clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
 170         if (group_dead) {
 171                 flush_sigqueue(&sig->shared_pending);
 172                 tty_kref_put(tty);
 173         }
 174 }
 175 
 176 static void delayed_put_task_struct(struct rcu_head *rhp)
 177 {
 178         struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
 179 
 180         perf_event_delayed_put(tsk);
 181         trace_sched_process_free(tsk);
 182         put_task_struct(tsk);
 183 }
 184 
 185 void put_task_struct_rcu_user(struct task_struct *task)
 186 {
 187         if (refcount_dec_and_test(&task->rcu_users))
 188                 call_rcu(&task->rcu, delayed_put_task_struct);
 189 }
 190 
 191 void release_task(struct task_struct *p)
 192 {
 193         struct task_struct *leader;
 194         int zap_leader;
 195 repeat:
 196         /* don't need to get the RCU readlock here - the process is dead and
 197          * can't be modifying its own credentials. But shut RCU-lockdep up */
 198         rcu_read_lock();
 199         atomic_dec(&__task_cred(p)->user->processes);
 200         rcu_read_unlock();
 201 
 202         proc_flush_task(p);
 203         cgroup_release(p);
 204 
 205         write_lock_irq(&tasklist_lock);
 206         ptrace_release_task(p);
 207         __exit_signal(p);
 208 
 209         /*
 210          * If we are the last non-leader member of the thread
 211          * group, and the leader is zombie, then notify the
 212          * group leader's parent process. (if it wants notification.)
 213          */
 214         zap_leader = 0;
 215         leader = p->group_leader;
 216         if (leader != p && thread_group_empty(leader)
 217                         && leader->exit_state == EXIT_ZOMBIE) {
 218                 /*
 219                  * If we were the last child thread and the leader has
 220                  * exited already, and the leader's parent ignores SIGCHLD,
 221                  * then we are the one who should release the leader.
 222                  */
 223                 zap_leader = do_notify_parent(leader, leader->exit_signal);
 224                 if (zap_leader)
 225                         leader->exit_state = EXIT_DEAD;
 226         }
 227 
 228         write_unlock_irq(&tasklist_lock);
 229         release_thread(p);
 230         put_task_struct_rcu_user(p);
 231 
 232         p = leader;
 233         if (unlikely(zap_leader))
 234                 goto repeat;
 235 }
 236 
 237 void rcuwait_wake_up(struct rcuwait *w)
 238 {
 239         struct task_struct *task;
 240 
 241         rcu_read_lock();
 242 
 243         /*
 244          * Order condition vs @task, such that everything prior to the load
 245          * of @task is visible. This is the condition as to why the user called
 246          * rcuwait_trywake() in the first place. Pairs with set_current_state()
 247          * barrier (A) in rcuwait_wait_event().
 248          *
 249          *    WAIT                WAKE
 250          *    [S] tsk = current   [S] cond = true
 251          *        MB (A)              MB (B)
 252          *    [L] cond            [L] tsk
 253          */
 254         smp_mb(); /* (B) */
 255 
 256         task = rcu_dereference(w->task);
 257         if (task)
 258                 wake_up_process(task);
 259         rcu_read_unlock();
 260 }
 261 
 262 /*
 263  * Determine if a process group is "orphaned", according to the POSIX
 264  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
 265  * by terminal-generated stop signals.  Newly orphaned process groups are
 266  * to receive a SIGHUP and a SIGCONT.
 267  *
 268  * "I ask you, have you ever known what it is to be an orphan?"
 269  */
 270 static int will_become_orphaned_pgrp(struct pid *pgrp,
 271                                         struct task_struct *ignored_task)
 272 {
 273         struct task_struct *p;
 274 
 275         do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
 276                 if ((p == ignored_task) ||
 277                     (p->exit_state && thread_group_empty(p)) ||
 278                     is_global_init(p->real_parent))
 279                         continue;
 280 
 281                 if (task_pgrp(p->real_parent) != pgrp &&
 282                     task_session(p->real_parent) == task_session(p))
 283                         return 0;
 284         } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
 285 
 286         return 1;
 287 }
 288 
 289 int is_current_pgrp_orphaned(void)
 290 {
 291         int retval;
 292 
 293         read_lock(&tasklist_lock);
 294         retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
 295         read_unlock(&tasklist_lock);
 296 
 297         return retval;
 298 }
 299 
 300 static bool has_stopped_jobs(struct pid *pgrp)
 301 {
 302         struct task_struct *p;
 303 
 304         do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
 305                 if (p->signal->flags & SIGNAL_STOP_STOPPED)
 306                         return true;
 307         } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
 308 
 309         return false;
 310 }
 311 
 312 /*
 313  * Check to see if any process groups have become orphaned as
 314  * a result of our exiting, and if they have any stopped jobs,
 315  * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
 316  */
 317 static void
 318 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
 319 {
 320         struct pid *pgrp = task_pgrp(tsk);
 321         struct task_struct *ignored_task = tsk;
 322 
 323         if (!parent)
 324                 /* exit: our father is in a different pgrp than
 325                  * we are and we were the only connection outside.
 326                  */
 327                 parent = tsk->real_parent;
 328         else
 329                 /* reparent: our child is in a different pgrp than
 330                  * we are, and it was the only connection outside.
 331                  */
 332                 ignored_task = NULL;
 333 
 334         if (task_pgrp(parent) != pgrp &&
 335             task_session(parent) == task_session(tsk) &&
 336             will_become_orphaned_pgrp(pgrp, ignored_task) &&
 337             has_stopped_jobs(pgrp)) {
 338                 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
 339                 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
 340         }
 341 }
 342 
 343 #ifdef CONFIG_MEMCG
 344 /*
 345  * A task is exiting.   If it owned this mm, find a new owner for the mm.
 346  */
 347 void mm_update_next_owner(struct mm_struct *mm)
 348 {
 349         struct task_struct *c, *g, *p = current;
 350 
 351 retry:
 352         /*
 353          * If the exiting or execing task is not the owner, it's
 354          * someone else's problem.
 355          */
 356         if (mm->owner != p)
 357                 return;
 358         /*
 359          * The current owner is exiting/execing and there are no other
 360          * candidates.  Do not leave the mm pointing to a possibly
 361          * freed task structure.
 362          */
 363         if (atomic_read(&mm->mm_users) <= 1) {
 364                 WRITE_ONCE(mm->owner, NULL);
 365                 return;
 366         }
 367 
 368         read_lock(&tasklist_lock);
 369         /*
 370          * Search in the children
 371          */
 372         list_for_each_entry(c, &p->children, sibling) {
 373                 if (c->mm == mm)
 374                         goto assign_new_owner;
 375         }
 376 
 377         /*
 378          * Search in the siblings
 379          */
 380         list_for_each_entry(c, &p->real_parent->children, sibling) {
 381                 if (c->mm == mm)
 382                         goto assign_new_owner;
 383         }
 384 
 385         /*
 386          * Search through everything else, we should not get here often.
 387          */
 388         for_each_process(g) {
 389                 if (g->flags & PF_KTHREAD)
 390                         continue;
 391                 for_each_thread(g, c) {
 392                         if (c->mm == mm)
 393                                 goto assign_new_owner;
 394                         if (c->mm)
 395                                 break;
 396                 }
 397         }
 398         read_unlock(&tasklist_lock);
 399         /*
 400          * We found no owner yet mm_users > 1: this implies that we are
 401          * most likely racing with swapoff (try_to_unuse()) or /proc or
 402          * ptrace or page migration (get_task_mm()).  Mark owner as NULL.
 403          */
 404         WRITE_ONCE(mm->owner, NULL);
 405         return;
 406 
 407 assign_new_owner:
 408         BUG_ON(c == p);
 409         get_task_struct(c);
 410         /*
 411          * The task_lock protects c->mm from changing.
 412          * We always want mm->owner->mm == mm
 413          */
 414         task_lock(c);
 415         /*
 416          * Delay read_unlock() till we have the task_lock()
 417          * to ensure that c does not slip away underneath us
 418          */
 419         read_unlock(&tasklist_lock);
 420         if (c->mm != mm) {
 421                 task_unlock(c);
 422                 put_task_struct(c);
 423                 goto retry;
 424         }
 425         WRITE_ONCE(mm->owner, c);
 426         task_unlock(c);
 427         put_task_struct(c);
 428 }
 429 #endif /* CONFIG_MEMCG */
 430 
 431 /*
 432  * Turn us into a lazy TLB process if we
 433  * aren't already..
 434  */
 435 static void exit_mm(void)
 436 {
 437         struct mm_struct *mm = current->mm;
 438         struct core_state *core_state;
 439 
 440         exit_mm_release(current, mm);
 441         if (!mm)
 442                 return;
 443         sync_mm_rss(mm);
 444         /*
 445          * Serialize with any possible pending coredump.
 446          * We must hold mmap_sem around checking core_state
 447          * and clearing tsk->mm.  The core-inducing thread
 448          * will increment ->nr_threads for each thread in the
 449          * group with ->mm != NULL.
 450          */
 451         down_read(&mm->mmap_sem);
 452         core_state = mm->core_state;
 453         if (core_state) {
 454                 struct core_thread self;
 455 
 456                 up_read(&mm->mmap_sem);
 457 
 458                 self.task = current;
 459                 self.next = xchg(&core_state->dumper.next, &self);
 460                 /*
 461                  * Implies mb(), the result of xchg() must be visible
 462                  * to core_state->dumper.
 463                  */
 464                 if (atomic_dec_and_test(&core_state->nr_threads))
 465                         complete(&core_state->startup);
 466 
 467                 for (;;) {
 468                         set_current_state(TASK_UNINTERRUPTIBLE);
 469                         if (!self.task) /* see coredump_finish() */
 470                                 break;
 471                         freezable_schedule();
 472                 }
 473                 __set_current_state(TASK_RUNNING);
 474                 down_read(&mm->mmap_sem);
 475         }
 476         mmgrab(mm);
 477         BUG_ON(mm != current->active_mm);
 478         /* more a memory barrier than a real lock */
 479         task_lock(current);
 480         current->mm = NULL;
 481         up_read(&mm->mmap_sem);
 482         enter_lazy_tlb(mm, current);
 483         task_unlock(current);
 484         mm_update_next_owner(mm);
 485         mmput(mm);
 486         if (test_thread_flag(TIF_MEMDIE))
 487                 exit_oom_victim();
 488 }
 489 
 490 static struct task_struct *find_alive_thread(struct task_struct *p)
 491 {
 492         struct task_struct *t;
 493 
 494         for_each_thread(p, t) {
 495                 if (!(t->flags & PF_EXITING))
 496                         return t;
 497         }
 498         return NULL;
 499 }
 500 
 501 static struct task_struct *find_child_reaper(struct task_struct *father,
 502                                                 struct list_head *dead)
 503         __releases(&tasklist_lock)
 504         __acquires(&tasklist_lock)
 505 {
 506         struct pid_namespace *pid_ns = task_active_pid_ns(father);
 507         struct task_struct *reaper = pid_ns->child_reaper;
 508         struct task_struct *p, *n;
 509 
 510         if (likely(reaper != father))
 511                 return reaper;
 512 
 513         reaper = find_alive_thread(father);
 514         if (reaper) {
 515                 pid_ns->child_reaper = reaper;
 516                 return reaper;
 517         }
 518 
 519         write_unlock_irq(&tasklist_lock);
 520 
 521         list_for_each_entry_safe(p, n, dead, ptrace_entry) {
 522                 list_del_init(&p->ptrace_entry);
 523                 release_task(p);
 524         }
 525 
 526         zap_pid_ns_processes(pid_ns);
 527         write_lock_irq(&tasklist_lock);
 528 
 529         return father;
 530 }
 531 
 532 /*
 533  * When we die, we re-parent all our children, and try to:
 534  * 1. give them to another thread in our thread group, if such a member exists
 535  * 2. give it to the first ancestor process which prctl'd itself as a
 536  *    child_subreaper for its children (like a service manager)
 537  * 3. give it to the init process (PID 1) in our pid namespace
 538  */
 539 static struct task_struct *find_new_reaper(struct task_struct *father,
 540                                            struct task_struct *child_reaper)
 541 {
 542         struct task_struct *thread, *reaper;
 543 
 544         thread = find_alive_thread(father);
 545         if (thread)
 546                 return thread;
 547 
 548         if (father->signal->has_child_subreaper) {
 549                 unsigned int ns_level = task_pid(father)->level;
 550                 /*
 551                  * Find the first ->is_child_subreaper ancestor in our pid_ns.
 552                  * We can't check reaper != child_reaper to ensure we do not
 553                  * cross the namespaces, the exiting parent could be injected
 554                  * by setns() + fork().
 555                  * We check pid->level, this is slightly more efficient than
 556                  * task_active_pid_ns(reaper) != task_active_pid_ns(father).
 557                  */
 558                 for (reaper = father->real_parent;
 559                      task_pid(reaper)->level == ns_level;
 560                      reaper = reaper->real_parent) {
 561                         if (reaper == &init_task)
 562                                 break;
 563                         if (!reaper->signal->is_child_subreaper)
 564                                 continue;
 565                         thread = find_alive_thread(reaper);
 566                         if (thread)
 567                                 return thread;
 568                 }
 569         }
 570 
 571         return child_reaper;
 572 }
 573 
 574 /*
 575 * Any that need to be release_task'd are put on the @dead list.
 576  */
 577 static void reparent_leader(struct task_struct *father, struct task_struct *p,
 578                                 struct list_head *dead)
 579 {
 580         if (unlikely(p->exit_state == EXIT_DEAD))
 581                 return;
 582 
 583         /* We don't want people slaying init. */
 584         p->exit_signal = SIGCHLD;
 585 
 586         /* If it has exited notify the new parent about this child's death. */
 587         if (!p->ptrace &&
 588             p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
 589                 if (do_notify_parent(p, p->exit_signal)) {
 590                         p->exit_state = EXIT_DEAD;
 591                         list_add(&p->ptrace_entry, dead);
 592                 }
 593         }
 594 
 595         kill_orphaned_pgrp(p, father);
 596 }
 597 
 598 /*
 599  * This does two things:
 600  *
 601  * A.  Make init inherit all the child processes
 602  * B.  Check to see if any process groups have become orphaned
 603  *      as a result of our exiting, and if they have any stopped
 604  *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 605  */
 606 static void forget_original_parent(struct task_struct *father,
 607                                         struct list_head *dead)
 608 {
 609         struct task_struct *p, *t, *reaper;
 610 
 611         if (unlikely(!list_empty(&father->ptraced)))
 612                 exit_ptrace(father, dead);
 613 
 614         /* Can drop and reacquire tasklist_lock */
 615         reaper = find_child_reaper(father, dead);
 616         if (list_empty(&father->children))
 617                 return;
 618 
 619         reaper = find_new_reaper(father, reaper);
 620         list_for_each_entry(p, &father->children, sibling) {
 621                 for_each_thread(p, t) {
 622                         t->real_parent = reaper;
 623                         BUG_ON((!t->ptrace) != (t->parent == father));
 624                         if (likely(!t->ptrace))
 625                                 t->parent = t->real_parent;
 626                         if (t->pdeath_signal)
 627                                 group_send_sig_info(t->pdeath_signal,
 628                                                     SEND_SIG_NOINFO, t,
 629                                                     PIDTYPE_TGID);
 630                 }
 631                 /*
 632                  * If this is a threaded reparent there is no need to
 633                  * notify anyone anything has happened.
 634                  */
 635                 if (!same_thread_group(reaper, father))
 636                         reparent_leader(father, p, dead);
 637         }
 638         list_splice_tail_init(&father->children, &reaper->children);
 639 }
 640 
 641 /*
 642  * Send signals to all our closest relatives so that they know
 643  * to properly mourn us..
 644  */
 645 static void exit_notify(struct task_struct *tsk, int group_dead)
 646 {
 647         bool autoreap;
 648         struct task_struct *p, *n;
 649         LIST_HEAD(dead);
 650 
 651         write_lock_irq(&tasklist_lock);
 652         forget_original_parent(tsk, &dead);
 653 
 654         if (group_dead)
 655                 kill_orphaned_pgrp(tsk->group_leader, NULL);
 656 
 657         tsk->exit_state = EXIT_ZOMBIE;
 658         if (unlikely(tsk->ptrace)) {
 659                 int sig = thread_group_leader(tsk) &&
 660                                 thread_group_empty(tsk) &&
 661                                 !ptrace_reparented(tsk) ?
 662                         tsk->exit_signal : SIGCHLD;
 663                 autoreap = do_notify_parent(tsk, sig);
 664         } else if (thread_group_leader(tsk)) {
 665                 autoreap = thread_group_empty(tsk) &&
 666                         do_notify_parent(tsk, tsk->exit_signal);
 667         } else {
 668                 autoreap = true;
 669         }
 670 
 671         if (autoreap) {
 672                 tsk->exit_state = EXIT_DEAD;
 673                 list_add(&tsk->ptrace_entry, &dead);
 674         }
 675 
 676         /* mt-exec, de_thread() is waiting for group leader */
 677         if (unlikely(tsk->signal->notify_count < 0))
 678                 wake_up_process(tsk->signal->group_exit_task);
 679         write_unlock_irq(&tasklist_lock);
 680 
 681         list_for_each_entry_safe(p, n, &dead, ptrace_entry) {
 682                 list_del_init(&p->ptrace_entry);
 683                 release_task(p);
 684         }
 685 }
 686 
 687 #ifdef CONFIG_DEBUG_STACK_USAGE
 688 static void check_stack_usage(void)
 689 {
 690         static DEFINE_SPINLOCK(low_water_lock);
 691         static int lowest_to_date = THREAD_SIZE;
 692         unsigned long free;
 693 
 694         free = stack_not_used(current);
 695 
 696         if (free >= lowest_to_date)
 697                 return;
 698 
 699         spin_lock(&low_water_lock);
 700         if (free < lowest_to_date) {
 701                 pr_info("%s (%d) used greatest stack depth: %lu bytes left\n",
 702                         current->comm, task_pid_nr(current), free);
 703                 lowest_to_date = free;
 704         }
 705         spin_unlock(&low_water_lock);
 706 }
 707 #else
 708 static inline void check_stack_usage(void) {}
 709 #endif
 710 
 711 void __noreturn do_exit(long code)
 712 {
 713         struct task_struct *tsk = current;
 714         int group_dead;
 715 
 716         profile_task_exit(tsk);
 717         kcov_task_exit(tsk);
 718 
 719         WARN_ON(blk_needs_flush_plug(tsk));
 720 
 721         if (unlikely(in_interrupt()))
 722                 panic("Aiee, killing interrupt handler!");
 723         if (unlikely(!tsk->pid))
 724                 panic("Attempted to kill the idle task!");
 725 
 726         /*
 727          * If do_exit is called because this processes oopsed, it's possible
 728          * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
 729          * continuing. Amongst other possible reasons, this is to prevent
 730          * mm_release()->clear_child_tid() from writing to a user-controlled
 731          * kernel address.
 732          */
 733         set_fs(USER_DS);
 734 
 735         ptrace_event(PTRACE_EVENT_EXIT, code);
 736 
 737         validate_creds_for_do_exit(tsk);
 738 
 739         /*
 740          * We're taking recursive faults here in do_exit. Safest is to just
 741          * leave this task alone and wait for reboot.
 742          */
 743         if (unlikely(tsk->flags & PF_EXITING)) {
 744                 pr_alert("Fixing recursive fault but reboot is needed!\n");
 745                 futex_exit_recursive(tsk);
 746                 set_current_state(TASK_UNINTERRUPTIBLE);
 747                 schedule();
 748         }
 749 
 750         exit_signals(tsk);  /* sets PF_EXITING */
 751 
 752         if (unlikely(in_atomic())) {
 753                 pr_info("note: %s[%d] exited with preempt_count %d\n",
 754                         current->comm, task_pid_nr(current),
 755                         preempt_count());
 756                 preempt_count_set(PREEMPT_ENABLED);
 757         }
 758 
 759         /* sync mm's RSS info before statistics gathering */
 760         if (tsk->mm)
 761                 sync_mm_rss(tsk->mm);
 762         acct_update_integrals(tsk);
 763         group_dead = atomic_dec_and_test(&tsk->signal->live);
 764         if (group_dead) {
 765                 /*
 766                  * If the last thread of global init has exited, panic
 767                  * immediately to get a useable coredump.
 768                  */
 769                 if (unlikely(is_global_init(tsk)))
 770                         panic("Attempted to kill init! exitcode=0x%08x\n",
 771                                 tsk->signal->group_exit_code ?: (int)code);
 772 
 773 #ifdef CONFIG_POSIX_TIMERS
 774                 hrtimer_cancel(&tsk->signal->real_timer);
 775                 exit_itimers(tsk->signal);
 776 #endif
 777                 if (tsk->mm)
 778                         setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
 779         }
 780         acct_collect(code, group_dead);
 781         if (group_dead)
 782                 tty_audit_exit();
 783         audit_free(tsk);
 784 
 785         tsk->exit_code = code;
 786         taskstats_exit(tsk, group_dead);
 787 
 788         exit_mm();
 789 
 790         if (group_dead)
 791                 acct_process();
 792         trace_sched_process_exit(tsk);
 793 
 794         exit_sem(tsk);
 795         exit_shm(tsk);
 796         exit_files(tsk);
 797         exit_fs(tsk);
 798         if (group_dead)
 799                 disassociate_ctty(1);
 800         exit_task_namespaces(tsk);
 801         exit_task_work(tsk);
 802         exit_thread(tsk);
 803         exit_umh(tsk);
 804 
 805         /*
 806          * Flush inherited counters to the parent - before the parent
 807          * gets woken up by child-exit notifications.
 808          *
 809          * because of cgroup mode, must be called before cgroup_exit()
 810          */
 811         perf_event_exit_task(tsk);
 812 
 813         sched_autogroup_exit_task(tsk);
 814         cgroup_exit(tsk);
 815 
 816         /*
 817          * FIXME: do that only when needed, using sched_exit tracepoint
 818          */
 819         flush_ptrace_hw_breakpoint(tsk);
 820 
 821         exit_tasks_rcu_start();
 822         exit_notify(tsk, group_dead);
 823         proc_exit_connector(tsk);
 824         mpol_put_task_policy(tsk);
 825 #ifdef CONFIG_FUTEX
 826         if (unlikely(current->pi_state_cache))
 827                 kfree(current->pi_state_cache);
 828 #endif
 829         /*
 830          * Make sure we are holding no locks:
 831          */
 832         debug_check_no_locks_held();
 833 
 834         if (tsk->io_context)
 835                 exit_io_context(tsk);
 836 
 837         if (tsk->splice_pipe)
 838                 free_pipe_info(tsk->splice_pipe);
 839 
 840         if (tsk->task_frag.page)
 841                 put_page(tsk->task_frag.page);
 842 
 843         validate_creds_for_do_exit(tsk);
 844 
 845         check_stack_usage();
 846         preempt_disable();
 847         if (tsk->nr_dirtied)
 848                 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
 849         exit_rcu();
 850         exit_tasks_rcu_finish();
 851 
 852         lockdep_free_task(tsk);
 853         do_task_dead();
 854 }
 855 EXPORT_SYMBOL_GPL(do_exit);
 856 
 857 void complete_and_exit(struct completion *comp, long code)
 858 {
 859         if (comp)
 860                 complete(comp);
 861 
 862         do_exit(code);
 863 }
 864 EXPORT_SYMBOL(complete_and_exit);
 865 
 866 SYSCALL_DEFINE1(exit, int, error_code)
 867 {
 868         do_exit((error_code&0xff)<<8);
 869 }
 870 
 871 /*
 872  * Take down every thread in the group.  This is called by fatal signals
 873  * as well as by sys_exit_group (below).
 874  */
 875 void
 876 do_group_exit(int exit_code)
 877 {
 878         struct signal_struct *sig = current->signal;
 879 
 880         BUG_ON(exit_code & 0x80); /* core dumps don't get here */
 881 
 882         if (signal_group_exit(sig))
 883                 exit_code = sig->group_exit_code;
 884         else if (!thread_group_empty(current)) {
 885                 struct sighand_struct *const sighand = current->sighand;
 886 
 887                 spin_lock_irq(&sighand->siglock);
 888                 if (signal_group_exit(sig))
 889                         /* Another thread got here before we took the lock.  */
 890                         exit_code = sig->group_exit_code;
 891                 else {
 892                         sig->group_exit_code = exit_code;
 893                         sig->flags = SIGNAL_GROUP_EXIT;
 894                         zap_other_threads(current);
 895                 }
 896                 spin_unlock_irq(&sighand->siglock);
 897         }
 898 
 899         do_exit(exit_code);
 900         /* NOTREACHED */
 901 }
 902 
 903 /*
 904  * this kills every thread in the thread group. Note that any externally
 905  * wait4()-ing process will get the correct exit code - even if this
 906  * thread is not the thread group leader.
 907  */
 908 SYSCALL_DEFINE1(exit_group, int, error_code)
 909 {
 910         do_group_exit((error_code & 0xff) << 8);
 911         /* NOTREACHED */
 912         return 0;
 913 }
 914 
 915 struct waitid_info {
 916         pid_t pid;
 917         uid_t uid;
 918         int status;
 919         int cause;
 920 };
 921 
 922 struct wait_opts {
 923         enum pid_type           wo_type;
 924         int                     wo_flags;
 925         struct pid              *wo_pid;
 926 
 927         struct waitid_info      *wo_info;
 928         int                     wo_stat;
 929         struct rusage           *wo_rusage;
 930 
 931         wait_queue_entry_t              child_wait;
 932         int                     notask_error;
 933 };
 934 
 935 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
 936 {
 937         return  wo->wo_type == PIDTYPE_MAX ||
 938                 task_pid_type(p, wo->wo_type) == wo->wo_pid;
 939 }
 940 
 941 static int
 942 eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p)
 943 {
 944         if (!eligible_pid(wo, p))
 945                 return 0;
 946 
 947         /*
 948          * Wait for all children (clone and not) if __WALL is set or
 949          * if it is traced by us.
 950          */
 951         if (ptrace || (wo->wo_flags & __WALL))
 952                 return 1;
 953 
 954         /*
 955          * Otherwise, wait for clone children *only* if __WCLONE is set;
 956          * otherwise, wait for non-clone children *only*.
 957          *
 958          * Note: a "clone" child here is one that reports to its parent
 959          * using a signal other than SIGCHLD, or a non-leader thread which
 960          * we can only see if it is traced by us.
 961          */
 962         if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
 963                 return 0;
 964 
 965         return 1;
 966 }
 967 
 968 /*
 969  * Handle sys_wait4 work for one task in state EXIT_ZOMBIE.  We hold
 970  * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
 971  * the lock and this task is uninteresting.  If we return nonzero, we have
 972  * released the lock and the system call should return.
 973  */
 974 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
 975 {
 976         int state, status;
 977         pid_t pid = task_pid_vnr(p);
 978         uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p));
 979         struct waitid_info *infop;
 980 
 981         if (!likely(wo->wo_flags & WEXITED))
 982                 return 0;
 983 
 984         if (unlikely(wo->wo_flags & WNOWAIT)) {
 985                 status = p->exit_code;
 986                 get_task_struct(p);
 987                 read_unlock(&tasklist_lock);
 988                 sched_annotate_sleep();
 989                 if (wo->wo_rusage)
 990                         getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
 991                 put_task_struct(p);
 992                 goto out_info;
 993         }
 994         /*
 995          * Move the task's state to DEAD/TRACE, only one thread can do this.
 996          */
 997         state = (ptrace_reparented(p) && thread_group_leader(p)) ?
 998                 EXIT_TRACE : EXIT_DEAD;
 999         if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE)
1000                 return 0;
1001         /*
1002          * We own this thread, nobody else can reap it.
1003          */
1004         read_unlock(&tasklist_lock);
1005         sched_annotate_sleep();
1006 
1007         /*
1008          * Check thread_group_leader() to exclude the traced sub-threads.
1009          */
1010         if (state == EXIT_DEAD && thread_group_leader(p)) {
1011                 struct signal_struct *sig = p->signal;
1012                 struct signal_struct *psig = current->signal;
1013                 unsigned long maxrss;
1014                 u64 tgutime, tgstime;
1015 
1016                 /*
1017                  * The resource counters for the group leader are in its
1018                  * own task_struct.  Those for dead threads in the group
1019                  * are in its signal_struct, as are those for the child
1020                  * processes it has previously reaped.  All these
1021                  * accumulate in the parent's signal_struct c* fields.
1022                  *
1023                  * We don't bother to take a lock here to protect these
1024                  * p->signal fields because the whole thread group is dead
1025                  * and nobody can change them.
1026                  *
1027                  * psig->stats_lock also protects us from our sub-theads
1028                  * which can reap other children at the same time. Until
1029                  * we change k_getrusage()-like users to rely on this lock
1030                  * we have to take ->siglock as well.
1031                  *
1032                  * We use thread_group_cputime_adjusted() to get times for
1033                  * the thread group, which consolidates times for all threads
1034                  * in the group including the group leader.
1035                  */
1036                 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1037                 spin_lock_irq(&current->sighand->siglock);
1038                 write_seqlock(&psig->stats_lock);
1039                 psig->cutime += tgutime + sig->cutime;
1040                 psig->cstime += tgstime + sig->cstime;
1041                 psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime;
1042                 psig->cmin_flt +=
1043                         p->min_flt + sig->min_flt + sig->cmin_flt;
1044                 psig->cmaj_flt +=
1045                         p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1046                 psig->cnvcsw +=
1047                         p->nvcsw + sig->nvcsw + sig->cnvcsw;
1048                 psig->cnivcsw +=
1049                         p->nivcsw + sig->nivcsw + sig->cnivcsw;
1050                 psig->cinblock +=
1051                         task_io_get_inblock(p) +
1052                         sig->inblock + sig->cinblock;
1053                 psig->coublock +=
1054                         task_io_get_oublock(p) +
1055                         sig->oublock + sig->coublock;
1056                 maxrss = max(sig->maxrss, sig->cmaxrss);
1057                 if (psig->cmaxrss < maxrss)
1058                         psig->cmaxrss = maxrss;
1059                 task_io_accounting_add(&psig->ioac, &p->ioac);
1060                 task_io_accounting_add(&psig->ioac, &sig->ioac);
1061                 write_sequnlock(&psig->stats_lock);
1062                 spin_unlock_irq(&current->sighand->siglock);
1063         }
1064 
1065         if (wo->wo_rusage)
1066                 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1067         status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1068                 ? p->signal->group_exit_code : p->exit_code;
1069         wo->wo_stat = status;
1070 
1071         if (state == EXIT_TRACE) {
1072                 write_lock_irq(&tasklist_lock);
1073                 /* We dropped tasklist, ptracer could die and untrace */
1074                 ptrace_unlink(p);
1075 
1076                 /* If parent wants a zombie, don't release it now */
1077                 state = EXIT_ZOMBIE;
1078                 if (do_notify_parent(p, p->exit_signal))
1079                         state = EXIT_DEAD;
1080                 p->exit_state = state;
1081                 write_unlock_irq(&tasklist_lock);
1082         }
1083         if (state == EXIT_DEAD)
1084                 release_task(p);
1085 
1086 out_info:
1087         infop = wo->wo_info;
1088         if (infop) {
1089                 if ((status & 0x7f) == 0) {
1090                         infop->cause = CLD_EXITED;
1091                         infop->status = status >> 8;
1092                 } else {
1093                         infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1094                         infop->status = status & 0x7f;
1095                 }
1096                 infop->pid = pid;
1097                 infop->uid = uid;
1098         }
1099 
1100         return pid;
1101 }
1102 
1103 static int *task_stopped_code(struct task_struct *p, bool ptrace)
1104 {
1105         if (ptrace) {
1106                 if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING))
1107                         return &p->exit_code;
1108         } else {
1109                 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1110                         return &p->signal->group_exit_code;
1111         }
1112         return NULL;
1113 }
1114 
1115 /**
1116  * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1117  * @wo: wait options
1118  * @ptrace: is the wait for ptrace
1119  * @p: task to wait for
1120  *
1121  * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1122  *
1123  * CONTEXT:
1124  * read_lock(&tasklist_lock), which is released if return value is
1125  * non-zero.  Also, grabs and releases @p->sighand->siglock.
1126  *
1127  * RETURNS:
1128  * 0 if wait condition didn't exist and search for other wait conditions
1129  * should continue.  Non-zero return, -errno on failure and @p's pid on
1130  * success, implies that tasklist_lock is released and wait condition
1131  * search should terminate.
1132  */
1133 static int wait_task_stopped(struct wait_opts *wo,
1134                                 int ptrace, struct task_struct *p)
1135 {
1136         struct waitid_info *infop;
1137         int exit_code, *p_code, why;
1138         uid_t uid = 0; /* unneeded, required by compiler */
1139         pid_t pid;
1140 
1141         /*
1142          * Traditionally we see ptrace'd stopped tasks regardless of options.
1143          */
1144         if (!ptrace && !(wo->wo_flags & WUNTRACED))
1145                 return 0;
1146 
1147         if (!task_stopped_code(p, ptrace))
1148                 return 0;
1149 
1150         exit_code = 0;
1151         spin_lock_irq(&p->sighand->siglock);
1152 
1153         p_code = task_stopped_code(p, ptrace);
1154         if (unlikely(!p_code))
1155                 goto unlock_sig;
1156 
1157         exit_code = *p_code;
1158         if (!exit_code)
1159                 goto unlock_sig;
1160 
1161         if (!unlikely(wo->wo_flags & WNOWAIT))
1162                 *p_code = 0;
1163 
1164         uid = from_kuid_munged(current_user_ns(), task_uid(p));
1165 unlock_sig:
1166         spin_unlock_irq(&p->sighand->siglock);
1167         if (!exit_code)
1168                 return 0;
1169 
1170         /*
1171          * Now we are pretty sure this task is interesting.
1172          * Make sure it doesn't get reaped out from under us while we
1173          * give up the lock and then examine it below.  We don't want to
1174          * keep holding onto the tasklist_lock while we call getrusage and
1175          * possibly take page faults for user memory.
1176          */
1177         get_task_struct(p);
1178         pid = task_pid_vnr(p);
1179         why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1180         read_unlock(&tasklist_lock);
1181         sched_annotate_sleep();
1182         if (wo->wo_rusage)
1183                 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1184         put_task_struct(p);
1185 
1186         if (likely(!(wo->wo_flags & WNOWAIT)))
1187                 wo->wo_stat = (exit_code << 8) | 0x7f;
1188 
1189         infop = wo->wo_info;
1190         if (infop) {
1191                 infop->cause = why;
1192                 infop->status = exit_code;
1193                 infop->pid = pid;
1194                 infop->uid = uid;
1195         }
1196         return pid;
1197 }
1198 
1199 /*
1200  * Handle do_wait work for one task in a live, non-stopped state.
1201  * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
1202  * the lock and this task is uninteresting.  If we return nonzero, we have
1203  * released the lock and the system call should return.
1204  */
1205 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1206 {
1207         struct waitid_info *infop;
1208         pid_t pid;
1209         uid_t uid;
1210 
1211         if (!unlikely(wo->wo_flags & WCONTINUED))
1212                 return 0;
1213 
1214         if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1215                 return 0;
1216 
1217         spin_lock_irq(&p->sighand->siglock);
1218         /* Re-check with the lock held.  */
1219         if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1220                 spin_unlock_irq(&p->sighand->siglock);
1221                 return 0;
1222         }
1223         if (!unlikely(wo->wo_flags & WNOWAIT))
1224                 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1225         uid = from_kuid_munged(current_user_ns(), task_uid(p));
1226         spin_unlock_irq(&p->sighand->siglock);
1227 
1228         pid = task_pid_vnr(p);
1229         get_task_struct(p);
1230         read_unlock(&tasklist_lock);
1231         sched_annotate_sleep();
1232         if (wo->wo_rusage)
1233                 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1234         put_task_struct(p);
1235 
1236         infop = wo->wo_info;
1237         if (!infop) {
1238                 wo->wo_stat = 0xffff;
1239         } else {
1240                 infop->cause = CLD_CONTINUED;
1241                 infop->pid = pid;
1242                 infop->uid = uid;
1243                 infop->status = SIGCONT;
1244         }
1245         return pid;
1246 }
1247 
1248 /*
1249  * Consider @p for a wait by @parent.
1250  *
1251  * -ECHILD should be in ->notask_error before the first call.
1252  * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1253  * Returns zero if the search for a child should continue;
1254  * then ->notask_error is 0 if @p is an eligible child,
1255  * or still -ECHILD.
1256  */
1257 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1258                                 struct task_struct *p)
1259 {
1260         /*
1261          * We can race with wait_task_zombie() from another thread.
1262          * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition
1263          * can't confuse the checks below.
1264          */
1265         int exit_state = READ_ONCE(p->exit_state);
1266         int ret;
1267 
1268         if (unlikely(exit_state == EXIT_DEAD))
1269                 return 0;
1270 
1271         ret = eligible_child(wo, ptrace, p);
1272         if (!ret)
1273                 return ret;
1274 
1275         if (unlikely(exit_state == EXIT_TRACE)) {
1276                 /*
1277                  * ptrace == 0 means we are the natural parent. In this case
1278                  * we should clear notask_error, debugger will notify us.
1279                  */
1280                 if (likely(!ptrace))
1281                         wo->notask_error = 0;
1282                 return 0;
1283         }
1284 
1285         if (likely(!ptrace) && unlikely(p->ptrace)) {
1286                 /*
1287                  * If it is traced by its real parent's group, just pretend
1288                  * the caller is ptrace_do_wait() and reap this child if it
1289                  * is zombie.
1290                  *
1291                  * This also hides group stop state from real parent; otherwise
1292                  * a single stop can be reported twice as group and ptrace stop.
1293                  * If a ptracer wants to distinguish these two events for its
1294                  * own children it should create a separate process which takes
1295                  * the role of real parent.
1296                  */
1297                 if (!ptrace_reparented(p))
1298                         ptrace = 1;
1299         }
1300 
1301         /* slay zombie? */
1302         if (exit_state == EXIT_ZOMBIE) {
1303                 /* we don't reap group leaders with subthreads */
1304                 if (!delay_group_leader(p)) {
1305                         /*
1306                          * A zombie ptracee is only visible to its ptracer.
1307                          * Notification and reaping will be cascaded to the
1308                          * real parent when the ptracer detaches.
1309                          */
1310                         if (unlikely(ptrace) || likely(!p->ptrace))
1311                                 return wait_task_zombie(wo, p);
1312                 }
1313 
1314                 /*
1315                  * Allow access to stopped/continued state via zombie by
1316                  * falling through.  Clearing of notask_error is complex.
1317                  *
1318                  * When !@ptrace:
1319                  *
1320                  * If WEXITED is set, notask_error should naturally be
1321                  * cleared.  If not, subset of WSTOPPED|WCONTINUED is set,
1322                  * so, if there are live subthreads, there are events to
1323                  * wait for.  If all subthreads are dead, it's still safe
1324                  * to clear - this function will be called again in finite
1325                  * amount time once all the subthreads are released and
1326                  * will then return without clearing.
1327                  *
1328                  * When @ptrace:
1329                  *
1330                  * Stopped state is per-task and thus can't change once the
1331                  * target task dies.  Only continued and exited can happen.
1332                  * Clear notask_error if WCONTINUED | WEXITED.
1333                  */
1334                 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1335                         wo->notask_error = 0;
1336         } else {
1337                 /*
1338                  * @p is alive and it's gonna stop, continue or exit, so
1339                  * there always is something to wait for.
1340                  */
1341                 wo->notask_error = 0;
1342         }
1343 
1344         /*
1345          * Wait for stopped.  Depending on @ptrace, different stopped state
1346          * is used and the two don't interact with each other.
1347          */
1348         ret = wait_task_stopped(wo, ptrace, p);
1349         if (ret)
1350                 return ret;
1351 
1352         /*
1353          * Wait for continued.  There's only one continued state and the
1354          * ptracer can consume it which can confuse the real parent.  Don't
1355          * use WCONTINUED from ptracer.  You don't need or want it.
1356          */
1357         return wait_task_continued(wo, p);
1358 }
1359 
1360 /*
1361  * Do the work of do_wait() for one thread in the group, @tsk.
1362  *
1363  * -ECHILD should be in ->notask_error before the first call.
1364  * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1365  * Returns zero if the search for a child should continue; then
1366  * ->notask_error is 0 if there were any eligible children,
1367  * or still -ECHILD.
1368  */
1369 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1370 {
1371         struct task_struct *p;
1372 
1373         list_for_each_entry(p, &tsk->children, sibling) {
1374                 int ret = wait_consider_task(wo, 0, p);
1375 
1376                 if (ret)
1377                         return ret;
1378         }
1379 
1380         return 0;
1381 }
1382 
1383 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1384 {
1385         struct task_struct *p;
1386 
1387         list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1388                 int ret = wait_consider_task(wo, 1, p);
1389 
1390                 if (ret)
1391                         return ret;
1392         }
1393 
1394         return 0;
1395 }
1396 
1397 static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode,
1398                                 int sync, void *key)
1399 {
1400         struct wait_opts *wo = container_of(wait, struct wait_opts,
1401                                                 child_wait);
1402         struct task_struct *p = key;
1403 
1404         if (!eligible_pid(wo, p))
1405                 return 0;
1406 
1407         if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1408                 return 0;
1409 
1410         return default_wake_function(wait, mode, sync, key);
1411 }
1412 
1413 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1414 {
1415         __wake_up_sync_key(&parent->signal->wait_chldexit,
1416                                 TASK_INTERRUPTIBLE, 1, p);
1417 }
1418 
1419 static long do_wait(struct wait_opts *wo)
1420 {
1421         struct task_struct *tsk;
1422         int retval;
1423 
1424         trace_sched_process_wait(wo->wo_pid);
1425 
1426         init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1427         wo->child_wait.private = current;
1428         add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1429 repeat:
1430         /*
1431          * If there is nothing that can match our criteria, just get out.
1432          * We will clear ->notask_error to zero if we see any child that
1433          * might later match our criteria, even if we are not able to reap
1434          * it yet.
1435          */
1436         wo->notask_error = -ECHILD;
1437         if ((wo->wo_type < PIDTYPE_MAX) &&
1438            (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
1439                 goto notask;
1440 
1441         set_current_state(TASK_INTERRUPTIBLE);
1442         read_lock(&tasklist_lock);
1443         tsk = current;
1444         do {
1445                 retval = do_wait_thread(wo, tsk);
1446                 if (retval)
1447                         goto end;
1448 
1449                 retval = ptrace_do_wait(wo, tsk);
1450                 if (retval)
1451                         goto end;
1452 
1453                 if (wo->wo_flags & __WNOTHREAD)
1454                         break;
1455         } while_each_thread(current, tsk);
1456         read_unlock(&tasklist_lock);
1457 
1458 notask:
1459         retval = wo->notask_error;
1460         if (!retval && !(wo->wo_flags & WNOHANG)) {
1461                 retval = -ERESTARTSYS;
1462                 if (!signal_pending(current)) {
1463                         schedule();
1464                         goto repeat;
1465                 }
1466         }
1467 end:
1468         __set_current_state(TASK_RUNNING);
1469         remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1470         return retval;
1471 }
1472 
1473 static struct pid *pidfd_get_pid(unsigned int fd)
1474 {
1475         struct fd f;
1476         struct pid *pid;
1477 
1478         f = fdget(fd);
1479         if (!f.file)
1480                 return ERR_PTR(-EBADF);
1481 
1482         pid = pidfd_pid(f.file);
1483         if (!IS_ERR(pid))
1484                 get_pid(pid);
1485 
1486         fdput(f);
1487         return pid;
1488 }
1489 
1490 static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop,
1491                           int options, struct rusage *ru)
1492 {
1493         struct wait_opts wo;
1494         struct pid *pid = NULL;
1495         enum pid_type type;
1496         long ret;
1497 
1498         if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED|
1499                         __WNOTHREAD|__WCLONE|__WALL))
1500                 return -EINVAL;
1501         if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1502                 return -EINVAL;
1503 
1504         switch (which) {
1505         case P_ALL:
1506                 type = PIDTYPE_MAX;
1507                 break;
1508         case P_PID:
1509                 type = PIDTYPE_PID;
1510                 if (upid <= 0)
1511                         return -EINVAL;
1512 
1513                 pid = find_get_pid(upid);
1514                 break;
1515         case P_PGID:
1516                 type = PIDTYPE_PGID;
1517                 if (upid < 0)
1518                         return -EINVAL;
1519 
1520                 if (upid)
1521                         pid = find_get_pid(upid);
1522                 else
1523                         pid = get_task_pid(current, PIDTYPE_PGID);
1524                 break;
1525         case P_PIDFD:
1526                 type = PIDTYPE_PID;
1527                 if (upid < 0)
1528                         return -EINVAL;
1529 
1530                 pid = pidfd_get_pid(upid);
1531                 if (IS_ERR(pid))
1532                         return PTR_ERR(pid);
1533                 break;
1534         default:
1535                 return -EINVAL;
1536         }
1537 
1538         wo.wo_type      = type;
1539         wo.wo_pid       = pid;
1540         wo.wo_flags     = options;
1541         wo.wo_info      = infop;
1542         wo.wo_rusage    = ru;
1543         ret = do_wait(&wo);
1544 
1545         put_pid(pid);
1546         return ret;
1547 }
1548 
1549 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1550                 infop, int, options, struct rusage __user *, ru)
1551 {
1552         struct rusage r;
1553         struct waitid_info info = {.status = 0};
1554         long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL);
1555         int signo = 0;
1556 
1557         if (err > 0) {
1558                 signo = SIGCHLD;
1559                 err = 0;
1560                 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
1561                         return -EFAULT;
1562         }
1563         if (!infop)
1564                 return err;
1565 
1566         if (!user_access_begin(infop, sizeof(*infop)))
1567                 return -EFAULT;
1568 
1569         unsafe_put_user(signo, &infop->si_signo, Efault);
1570         unsafe_put_user(0, &infop->si_errno, Efault);
1571         unsafe_put_user(info.cause, &infop->si_code, Efault);
1572         unsafe_put_user(info.pid, &infop->si_pid, Efault);
1573         unsafe_put_user(info.uid, &infop->si_uid, Efault);
1574         unsafe_put_user(info.status, &infop->si_status, Efault);
1575         user_access_end();
1576         return err;
1577 Efault:
1578         user_access_end();
1579         return -EFAULT;
1580 }
1581 
1582 long kernel_wait4(pid_t upid, int __user *stat_addr, int options,
1583                   struct rusage *ru)
1584 {
1585         struct wait_opts wo;
1586         struct pid *pid = NULL;
1587         enum pid_type type;
1588         long ret;
1589 
1590         if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1591                         __WNOTHREAD|__WCLONE|__WALL))
1592                 return -EINVAL;
1593 
1594         /* -INT_MIN is not defined */
1595         if (upid == INT_MIN)
1596                 return -ESRCH;
1597 
1598         if (upid == -1)
1599                 type = PIDTYPE_MAX;
1600         else if (upid < 0) {
1601                 type = PIDTYPE_PGID;
1602                 pid = find_get_pid(-upid);
1603         } else if (upid == 0) {
1604                 type = PIDTYPE_PGID;
1605                 pid = get_task_pid(current, PIDTYPE_PGID);
1606         } else /* upid > 0 */ {
1607                 type = PIDTYPE_PID;
1608                 pid = find_get_pid(upid);
1609         }
1610 
1611         wo.wo_type      = type;
1612         wo.wo_pid       = pid;
1613         wo.wo_flags     = options | WEXITED;
1614         wo.wo_info      = NULL;
1615         wo.wo_stat      = 0;
1616         wo.wo_rusage    = ru;
1617         ret = do_wait(&wo);
1618         put_pid(pid);
1619         if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr))
1620                 ret = -EFAULT;
1621 
1622         return ret;
1623 }
1624 
1625 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1626                 int, options, struct rusage __user *, ru)
1627 {
1628         struct rusage r;
1629         long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL);
1630 
1631         if (err > 0) {
1632                 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
1633                         return -EFAULT;
1634         }
1635         return err;
1636 }
1637 
1638 #ifdef __ARCH_WANT_SYS_WAITPID
1639 
1640 /*
1641  * sys_waitpid() remains for compatibility. waitpid() should be
1642  * implemented by calling sys_wait4() from libc.a.
1643  */
1644 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1645 {
1646         return kernel_wait4(pid, stat_addr, options, NULL);
1647 }
1648 
1649 #endif
1650 
1651 #ifdef CONFIG_COMPAT
1652 COMPAT_SYSCALL_DEFINE4(wait4,
1653         compat_pid_t, pid,
1654         compat_uint_t __user *, stat_addr,
1655         int, options,
1656         struct compat_rusage __user *, ru)
1657 {
1658         struct rusage r;
1659         long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL);
1660         if (err > 0) {
1661                 if (ru && put_compat_rusage(&r, ru))
1662                         return -EFAULT;
1663         }
1664         return err;
1665 }
1666 
1667 COMPAT_SYSCALL_DEFINE5(waitid,
1668                 int, which, compat_pid_t, pid,
1669                 struct compat_siginfo __user *, infop, int, options,
1670                 struct compat_rusage __user *, uru)
1671 {
1672         struct rusage ru;
1673         struct waitid_info info = {.status = 0};
1674         long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL);
1675         int signo = 0;
1676         if (err > 0) {
1677                 signo = SIGCHLD;
1678                 err = 0;
1679                 if (uru) {
1680                         /* kernel_waitid() overwrites everything in ru */
1681                         if (COMPAT_USE_64BIT_TIME)
1682                                 err = copy_to_user(uru, &ru, sizeof(ru));
1683                         else
1684                                 err = put_compat_rusage(&ru, uru);
1685                         if (err)
1686                                 return -EFAULT;
1687                 }
1688         }
1689 
1690         if (!infop)
1691                 return err;
1692 
1693         if (!user_access_begin(infop, sizeof(*infop)))
1694                 return -EFAULT;
1695 
1696         unsafe_put_user(signo, &infop->si_signo, Efault);
1697         unsafe_put_user(0, &infop->si_errno, Efault);
1698         unsafe_put_user(info.cause, &infop->si_code, Efault);
1699         unsafe_put_user(info.pid, &infop->si_pid, Efault);
1700         unsafe_put_user(info.uid, &infop->si_uid, Efault);
1701         unsafe_put_user(info.status, &infop->si_status, Efault);
1702         user_access_end();
1703         return err;
1704 Efault:
1705         user_access_end();
1706         return -EFAULT;
1707 }
1708 #endif
1709 
1710 __weak void abort(void)
1711 {
1712         BUG();
1713 
1714         /* if that doesn't kill us, halt */
1715         panic("Oops failed to kill thread");
1716 }
1717 EXPORT_SYMBOL(abort);
/* [<][>][^][v][top][bottom][index][help] */
root/kernel/exit.c

DEFINITIONS
