1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1991, 1992 Linus Torvalds
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/task_io_accounting_ops.h>
52 #include <linux/blkdev.h>
53 #include <linux/task_work.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/kmsan.h>
64 #include <linux/random.h>
65 #include <linux/rcuwait.h>
66 #include <linux/compat.h>
67 #include <linux/io_uring.h>
68 #include <linux/kprobes.h>
69 #include <linux/rethook.h>
70 #include <linux/sysfs.h>
72 #include <linux/uaccess.h>
73 #include <asm/unistd.h>
74 #include <asm/mmu_context.h>
77 * The default value should be high enough to not crash a system that randomly
78 * crashes its kernel from time to time, but low enough to at least not permit
79 * overflowing 32-bit refcounts or the ldsem writer count.
81 static unsigned int oops_limit = 10000;
84 static struct ctl_table kern_exit_table[] = {
86 .procname = "oops_limit",
88 .maxlen = sizeof(oops_limit),
90 .proc_handler = proc_douintvec,
95 static __init int kernel_exit_sysctls_init(void)
97 register_sysctl_init("kernel", kern_exit_table);
100 late_initcall(kernel_exit_sysctls_init);
103 static atomic_t oops_count = ATOMIC_INIT(0);
106 static ssize_t oops_count_show(struct kobject *kobj, struct kobj_attribute *attr,
109 return sysfs_emit(page, "%d\n", atomic_read(&oops_count));
112 static struct kobj_attribute oops_count_attr = __ATTR_RO(oops_count);
114 static __init int kernel_exit_sysfs_init(void)
116 sysfs_add_file_to_group(kernel_kobj, &oops_count_attr.attr, NULL);
119 late_initcall(kernel_exit_sysfs_init);
122 static void __unhash_process(struct task_struct *p, bool group_dead)
125 detach_pid(p, PIDTYPE_PID);
127 detach_pid(p, PIDTYPE_TGID);
128 detach_pid(p, PIDTYPE_PGID);
129 detach_pid(p, PIDTYPE_SID);
131 list_del_rcu(&p->tasks);
132 list_del_init(&p->sibling);
133 __this_cpu_dec(process_counts);
135 list_del_rcu(&p->thread_group);
136 list_del_rcu(&p->thread_node);
140 * This function expects the tasklist_lock write-locked.
142 static void __exit_signal(struct task_struct *tsk)
144 struct signal_struct *sig = tsk->signal;
145 bool group_dead = thread_group_leader(tsk);
146 struct sighand_struct *sighand;
147 struct tty_struct *tty;
150 sighand = rcu_dereference_check(tsk->sighand,
151 lockdep_tasklist_lock_is_held());
152 spin_lock(&sighand->siglock);
154 #ifdef CONFIG_POSIX_TIMERS
155 posix_cpu_timers_exit(tsk);
157 posix_cpu_timers_exit_group(tsk);
165 * If there is any task waiting for the group exit
168 if (sig->notify_count > 0 && !--sig->notify_count)
169 wake_up_process(sig->group_exec_task);
171 if (tsk == sig->curr_target)
172 sig->curr_target = next_thread(tsk);
175 add_device_randomness((const void*) &tsk->se.sum_exec_runtime,
176 sizeof(unsigned long long));
179 * Accumulate here the counters for all threads as they die. We could
180 * skip the group leader because it is the last user of signal_struct,
181 * but we want to avoid the race with thread_group_cputime() which can
182 * see the empty ->thread_head list.
184 task_cputime(tsk, &utime, &stime);
185 write_seqlock(&sig->stats_lock);
188 sig->gtime += task_gtime(tsk);
189 sig->min_flt += tsk->min_flt;
190 sig->maj_flt += tsk->maj_flt;
191 sig->nvcsw += tsk->nvcsw;
192 sig->nivcsw += tsk->nivcsw;
193 sig->inblock += task_io_get_inblock(tsk);
194 sig->oublock += task_io_get_oublock(tsk);
195 task_io_accounting_add(&sig->ioac, &tsk->ioac);
196 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
198 __unhash_process(tsk, group_dead);
199 write_sequnlock(&sig->stats_lock);
202 * Do this under ->siglock, we can race with another thread
203 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
205 flush_sigqueue(&tsk->pending);
207 spin_unlock(&sighand->siglock);
209 __cleanup_sighand(sighand);
210 clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
212 flush_sigqueue(&sig->shared_pending);
217 static void delayed_put_task_struct(struct rcu_head *rhp)
219 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
221 kprobe_flush_task(tsk);
222 rethook_flush_task(tsk);
223 perf_event_delayed_put(tsk);
224 trace_sched_process_free(tsk);
225 put_task_struct(tsk);
228 void put_task_struct_rcu_user(struct task_struct *task)
230 if (refcount_dec_and_test(&task->rcu_users))
231 call_rcu(&task->rcu, delayed_put_task_struct);
234 void __weak release_thread(struct task_struct *dead_task)
238 void release_task(struct task_struct *p)
240 struct task_struct *leader;
241 struct pid *thread_pid;
244 /* don't need to get the RCU readlock here - the process is dead and
245 * can't be modifying its own credentials. But shut RCU-lockdep up */
247 dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1);
252 write_lock_irq(&tasklist_lock);
253 ptrace_release_task(p);
254 thread_pid = get_pid(p->thread_pid);
258 * If we are the last non-leader member of the thread
259 * group, and the leader is zombie, then notify the
260 * group leader's parent process. (if it wants notification.)
263 leader = p->group_leader;
264 if (leader != p && thread_group_empty(leader)
265 && leader->exit_state == EXIT_ZOMBIE) {
267 * If we were the last child thread and the leader has
268 * exited already, and the leader's parent ignores SIGCHLD,
269 * then we are the one who should release the leader.
271 zap_leader = do_notify_parent(leader, leader->exit_signal);
273 leader->exit_state = EXIT_DEAD;
276 write_unlock_irq(&tasklist_lock);
277 seccomp_filter_release(p);
278 proc_flush_pid(thread_pid);
281 put_task_struct_rcu_user(p);
284 if (unlikely(zap_leader))
288 int rcuwait_wake_up(struct rcuwait *w)
291 struct task_struct *task;
296 * Order condition vs @task, such that everything prior to the load
297 * of @task is visible. This is the condition as to why the user called
298 * rcuwait_wake() in the first place. Pairs with set_current_state()
299 * barrier (A) in rcuwait_wait_event().
302 * [S] tsk = current [S] cond = true
308 task = rcu_dereference(w->task);
310 ret = wake_up_process(task);
315 EXPORT_SYMBOL_GPL(rcuwait_wake_up);
318 * Determine if a process group is "orphaned", according to the POSIX
319 * definition in 2.2.2.52. Orphaned process groups are not to be affected
320 * by terminal-generated stop signals. Newly orphaned process groups are
321 * to receive a SIGHUP and a SIGCONT.
323 * "I ask you, have you ever known what it is to be an orphan?"
325 static int will_become_orphaned_pgrp(struct pid *pgrp,
326 struct task_struct *ignored_task)
328 struct task_struct *p;
330 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
331 if ((p == ignored_task) ||
332 (p->exit_state && thread_group_empty(p)) ||
333 is_global_init(p->real_parent))
336 if (task_pgrp(p->real_parent) != pgrp &&
337 task_session(p->real_parent) == task_session(p))
339 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
344 int is_current_pgrp_orphaned(void)
348 read_lock(&tasklist_lock);
349 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
350 read_unlock(&tasklist_lock);
355 static bool has_stopped_jobs(struct pid *pgrp)
357 struct task_struct *p;
359 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
360 if (p->signal->flags & SIGNAL_STOP_STOPPED)
362 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
368 * Check to see if any process groups have become orphaned as
369 * a result of our exiting, and if they have any stopped jobs,
370 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
373 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
375 struct pid *pgrp = task_pgrp(tsk);
376 struct task_struct *ignored_task = tsk;
379 /* exit: our father is in a different pgrp than
380 * we are and we were the only connection outside.
382 parent = tsk->real_parent;
384 /* reparent: our child is in a different pgrp than
385 * we are, and it was the only connection outside.
389 if (task_pgrp(parent) != pgrp &&
390 task_session(parent) == task_session(tsk) &&
391 will_become_orphaned_pgrp(pgrp, ignored_task) &&
392 has_stopped_jobs(pgrp)) {
393 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
394 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
398 static void coredump_task_exit(struct task_struct *tsk)
400 struct core_state *core_state;
403 * Serialize with any possible pending coredump.
404 * We must hold siglock around checking core_state
405 * and setting PF_POSTCOREDUMP. The core-inducing thread
406 * will increment ->nr_threads for each thread in the
407 * group without PF_POSTCOREDUMP set.
409 spin_lock_irq(&tsk->sighand->siglock);
410 tsk->flags |= PF_POSTCOREDUMP;
411 core_state = tsk->signal->core_state;
412 spin_unlock_irq(&tsk->sighand->siglock);
414 struct core_thread self;
417 if (self.task->flags & PF_SIGNALED)
418 self.next = xchg(&core_state->dumper.next, &self);
422 * Implies mb(), the result of xchg() must be visible
423 * to core_state->dumper.
425 if (atomic_dec_and_test(&core_state->nr_threads))
426 complete(&core_state->startup);
429 set_current_state(TASK_UNINTERRUPTIBLE|TASK_FREEZABLE);
430 if (!self.task) /* see coredump_finish() */
434 __set_current_state(TASK_RUNNING);
440 * A task is exiting. If it owned this mm, find a new owner for the mm.
442 void mm_update_next_owner(struct mm_struct *mm)
444 struct task_struct *c, *g, *p = current;
448 * If the exiting or execing task is not the owner, it's
449 * someone else's problem.
454 * The current owner is exiting/execing and there are no other
455 * candidates. Do not leave the mm pointing to a possibly
456 * freed task structure.
458 if (atomic_read(&mm->mm_users) <= 1) {
459 WRITE_ONCE(mm->owner, NULL);
463 read_lock(&tasklist_lock);
465 * Search in the children
467 list_for_each_entry(c, &p->children, sibling) {
469 goto assign_new_owner;
473 * Search in the siblings
475 list_for_each_entry(c, &p->real_parent->children, sibling) {
477 goto assign_new_owner;
481 * Search through everything else, we should not get here often.
483 for_each_process(g) {
484 if (g->flags & PF_KTHREAD)
486 for_each_thread(g, c) {
488 goto assign_new_owner;
493 read_unlock(&tasklist_lock);
495 * We found no owner yet mm_users > 1: this implies that we are
496 * most likely racing with swapoff (try_to_unuse()) or /proc or
497 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
499 WRITE_ONCE(mm->owner, NULL);
506 * The task_lock protects c->mm from changing.
507 * We always want mm->owner->mm == mm
511 * Delay read_unlock() till we have the task_lock()
512 * to ensure that c does not slip away underneath us
514 read_unlock(&tasklist_lock);
520 WRITE_ONCE(mm->owner, c);
521 lru_gen_migrate_mm(mm);
525 #endif /* CONFIG_MEMCG */
528 * Turn us into a lazy TLB process if we
531 static void exit_mm(void)
533 struct mm_struct *mm = current->mm;
535 exit_mm_release(current, mm);
541 BUG_ON(mm != current->active_mm);
542 /* more a memory barrier than a real lock */
545 * When a thread stops operating on an address space, the loop
546 * in membarrier_private_expedited() may not observe that
547 * tsk->mm, and the loop in membarrier_global_expedited() may
548 * not observe a MEMBARRIER_STATE_GLOBAL_EXPEDITED
549 * rq->membarrier_state, so those would not issue an IPI.
550 * Membarrier requires a memory barrier after accessing
551 * user-space memory, before clearing tsk->mm or the
552 * rq->membarrier_state.
554 smp_mb__after_spinlock();
557 membarrier_update_current_mm(NULL);
558 enter_lazy_tlb(mm, current);
560 task_unlock(current);
561 mmap_read_unlock(mm);
562 mm_update_next_owner(mm);
564 if (test_thread_flag(TIF_MEMDIE))
568 static struct task_struct *find_alive_thread(struct task_struct *p)
570 struct task_struct *t;
572 for_each_thread(p, t) {
573 if (!(t->flags & PF_EXITING))
579 static struct task_struct *find_child_reaper(struct task_struct *father,
580 struct list_head *dead)
581 __releases(&tasklist_lock)
582 __acquires(&tasklist_lock)
584 struct pid_namespace *pid_ns = task_active_pid_ns(father);
585 struct task_struct *reaper = pid_ns->child_reaper;
586 struct task_struct *p, *n;
588 if (likely(reaper != father))
591 reaper = find_alive_thread(father);
593 pid_ns->child_reaper = reaper;
597 write_unlock_irq(&tasklist_lock);
599 list_for_each_entry_safe(p, n, dead, ptrace_entry) {
600 list_del_init(&p->ptrace_entry);
604 zap_pid_ns_processes(pid_ns);
605 write_lock_irq(&tasklist_lock);
611 * When we die, we re-parent all our children, and try to:
612 * 1. give them to another thread in our thread group, if such a member exists
613 * 2. give it to the first ancestor process which prctl'd itself as a
614 * child_subreaper for its children (like a service manager)
615 * 3. give it to the init process (PID 1) in our pid namespace
617 static struct task_struct *find_new_reaper(struct task_struct *father,
618 struct task_struct *child_reaper)
620 struct task_struct *thread, *reaper;
622 thread = find_alive_thread(father);
626 if (father->signal->has_child_subreaper) {
627 unsigned int ns_level = task_pid(father)->level;
629 * Find the first ->is_child_subreaper ancestor in our pid_ns.
630 * We can't check reaper != child_reaper to ensure we do not
631 * cross the namespaces, the exiting parent could be injected
632 * by setns() + fork().
633 * We check pid->level, this is slightly more efficient than
634 * task_active_pid_ns(reaper) != task_active_pid_ns(father).
636 for (reaper = father->real_parent;
637 task_pid(reaper)->level == ns_level;
638 reaper = reaper->real_parent) {
639 if (reaper == &init_task)
641 if (!reaper->signal->is_child_subreaper)
643 thread = find_alive_thread(reaper);
653 * Any that need to be release_task'd are put on the @dead list.
655 static void reparent_leader(struct task_struct *father, struct task_struct *p,
656 struct list_head *dead)
658 if (unlikely(p->exit_state == EXIT_DEAD))
661 /* We don't want people slaying init. */
662 p->exit_signal = SIGCHLD;
664 /* If it has exited notify the new parent about this child's death. */
666 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
667 if (do_notify_parent(p, p->exit_signal)) {
668 p->exit_state = EXIT_DEAD;
669 list_add(&p->ptrace_entry, dead);
673 kill_orphaned_pgrp(p, father);
677 * This does two things:
679 * A. Make init inherit all the child processes
680 * B. Check to see if any process groups have become orphaned
681 * as a result of our exiting, and if they have any stopped
682 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
684 static void forget_original_parent(struct task_struct *father,
685 struct list_head *dead)
687 struct task_struct *p, *t, *reaper;
689 if (unlikely(!list_empty(&father->ptraced)))
690 exit_ptrace(father, dead);
692 /* Can drop and reacquire tasklist_lock */
693 reaper = find_child_reaper(father, dead);
694 if (list_empty(&father->children))
697 reaper = find_new_reaper(father, reaper);
698 list_for_each_entry(p, &father->children, sibling) {
699 for_each_thread(p, t) {
700 RCU_INIT_POINTER(t->real_parent, reaper);
701 BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father));
702 if (likely(!t->ptrace))
703 t->parent = t->real_parent;
704 if (t->pdeath_signal)
705 group_send_sig_info(t->pdeath_signal,
710 * If this is a threaded reparent there is no need to
711 * notify anyone anything has happened.
713 if (!same_thread_group(reaper, father))
714 reparent_leader(father, p, dead);
716 list_splice_tail_init(&father->children, &reaper->children);
720 * Send signals to all our closest relatives so that they know
721 * to properly mourn us..
723 static void exit_notify(struct task_struct *tsk, int group_dead)
726 struct task_struct *p, *n;
729 write_lock_irq(&tasklist_lock);
730 forget_original_parent(tsk, &dead);
733 kill_orphaned_pgrp(tsk->group_leader, NULL);
735 tsk->exit_state = EXIT_ZOMBIE;
736 if (unlikely(tsk->ptrace)) {
737 int sig = thread_group_leader(tsk) &&
738 thread_group_empty(tsk) &&
739 !ptrace_reparented(tsk) ?
740 tsk->exit_signal : SIGCHLD;
741 autoreap = do_notify_parent(tsk, sig);
742 } else if (thread_group_leader(tsk)) {
743 autoreap = thread_group_empty(tsk) &&
744 do_notify_parent(tsk, tsk->exit_signal);
750 tsk->exit_state = EXIT_DEAD;
751 list_add(&tsk->ptrace_entry, &dead);
754 /* mt-exec, de_thread() is waiting for group leader */
755 if (unlikely(tsk->signal->notify_count < 0))
756 wake_up_process(tsk->signal->group_exec_task);
757 write_unlock_irq(&tasklist_lock);
759 list_for_each_entry_safe(p, n, &dead, ptrace_entry) {
760 list_del_init(&p->ptrace_entry);
765 #ifdef CONFIG_DEBUG_STACK_USAGE
766 static void check_stack_usage(void)
768 static DEFINE_SPINLOCK(low_water_lock);
769 static int lowest_to_date = THREAD_SIZE;
772 free = stack_not_used(current);
774 if (free >= lowest_to_date)
777 spin_lock(&low_water_lock);
778 if (free < lowest_to_date) {
779 pr_info("%s (%d) used greatest stack depth: %lu bytes left\n",
780 current->comm, task_pid_nr(current), free);
781 lowest_to_date = free;
783 spin_unlock(&low_water_lock);
786 static inline void check_stack_usage(void) {}
789 static void synchronize_group_exit(struct task_struct *tsk, long code)
791 struct sighand_struct *sighand = tsk->sighand;
792 struct signal_struct *signal = tsk->signal;
794 spin_lock_irq(&sighand->siglock);
795 signal->quick_threads--;
796 if ((signal->quick_threads == 0) &&
797 !(signal->flags & SIGNAL_GROUP_EXIT)) {
798 signal->flags = SIGNAL_GROUP_EXIT;
799 signal->group_exit_code = code;
800 signal->group_stop_count = 0;
802 spin_unlock_irq(&sighand->siglock);
805 void __noreturn do_exit(long code)
807 struct task_struct *tsk = current;
810 synchronize_group_exit(tsk, code);
815 kmsan_task_exit(tsk);
817 coredump_task_exit(tsk);
818 ptrace_event(PTRACE_EVENT_EXIT, code);
820 validate_creds_for_do_exit(tsk);
822 io_uring_files_cancel();
823 exit_signals(tsk); /* sets PF_EXITING */
825 /* sync mm's RSS info before statistics gathering */
827 sync_mm_rss(tsk->mm);
828 acct_update_integrals(tsk);
829 group_dead = atomic_dec_and_test(&tsk->signal->live);
832 * If the last thread of global init has exited, panic
833 * immediately to get a useable coredump.
835 if (unlikely(is_global_init(tsk)))
836 panic("Attempted to kill init! exitcode=0x%08x\n",
837 tsk->signal->group_exit_code ?: (int)code);
839 #ifdef CONFIG_POSIX_TIMERS
840 hrtimer_cancel(&tsk->signal->real_timer);
844 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
846 acct_collect(code, group_dead);
851 tsk->exit_code = code;
852 taskstats_exit(tsk, group_dead);
858 trace_sched_process_exit(tsk);
865 disassociate_ctty(1);
866 exit_task_namespaces(tsk);
871 * Flush inherited counters to the parent - before the parent
872 * gets woken up by child-exit notifications.
874 * because of cgroup mode, must be called before cgroup_exit()
876 perf_event_exit_task(tsk);
878 sched_autogroup_exit_task(tsk);
882 * FIXME: do that only when needed, using sched_exit tracepoint
884 flush_ptrace_hw_breakpoint(tsk);
886 exit_tasks_rcu_start();
887 exit_notify(tsk, group_dead);
888 proc_exit_connector(tsk);
889 mpol_put_task_policy(tsk);
891 if (unlikely(current->pi_state_cache))
892 kfree(current->pi_state_cache);
895 * Make sure we are holding no locks:
897 debug_check_no_locks_held();
900 exit_io_context(tsk);
902 if (tsk->splice_pipe)
903 free_pipe_info(tsk->splice_pipe);
905 if (tsk->task_frag.page)
906 put_page(tsk->task_frag.page);
908 validate_creds_for_do_exit(tsk);
909 exit_task_stack_account(tsk);
914 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
916 exit_tasks_rcu_finish();
918 lockdep_free_task(tsk);
922 void __noreturn make_task_dead(int signr)
925 * Take the task off the cpu after something catastrophic has
928 * We can get here from a kernel oops, sometimes with preemption off.
929 * Start by checking for critical errors.
930 * Then fix up important state like USER_DS and preemption.
931 * Then do everything else.
933 struct task_struct *tsk = current;
935 if (unlikely(in_interrupt()))
936 panic("Aiee, killing interrupt handler!");
937 if (unlikely(!tsk->pid))
938 panic("Attempted to kill the idle task!");
940 if (unlikely(in_atomic())) {
941 pr_info("note: %s[%d] exited with preempt_count %d\n",
942 current->comm, task_pid_nr(current),
944 preempt_count_set(PREEMPT_ENABLED);
948 * Every time the system oopses, if the oops happens while a reference
949 * to an object was held, the reference leaks.
950 * If the oops doesn't also leak memory, repeated oopsing can cause
951 * reference counters to wrap around (if they're not using refcount_t).
952 * This means that repeated oopsing can make unexploitable-looking bugs
953 * exploitable through repeated oopsing.
954 * To make sure this can't happen, place an upper bound on how often the
955 * kernel may oops without panic().
957 if (atomic_inc_return(&oops_count) >= READ_ONCE(oops_limit) && oops_limit)
958 panic("Oopsed too often (kernel.oops_limit is %d)", oops_limit);
961 * We're taking recursive faults here in make_task_dead. Safest is to just
962 * leave this task alone and wait for reboot.
964 if (unlikely(tsk->flags & PF_EXITING)) {
965 pr_alert("Fixing recursive fault but reboot is needed!\n");
966 futex_exit_recursive(tsk);
967 tsk->exit_state = EXIT_DEAD;
968 refcount_inc(&tsk->rcu_users);
975 SYSCALL_DEFINE1(exit, int, error_code)
977 do_exit((error_code&0xff)<<8);
981 * Take down every thread in the group. This is called by fatal signals
982 * as well as by sys_exit_group (below).
985 do_group_exit(int exit_code)
987 struct signal_struct *sig = current->signal;
989 if (sig->flags & SIGNAL_GROUP_EXIT)
990 exit_code = sig->group_exit_code;
991 else if (sig->group_exec_task)
994 struct sighand_struct *const sighand = current->sighand;
996 spin_lock_irq(&sighand->siglock);
997 if (sig->flags & SIGNAL_GROUP_EXIT)
998 /* Another thread got here before we took the lock. */
999 exit_code = sig->group_exit_code;
1000 else if (sig->group_exec_task)
1003 sig->group_exit_code = exit_code;
1004 sig->flags = SIGNAL_GROUP_EXIT;
1005 zap_other_threads(current);
1007 spin_unlock_irq(&sighand->siglock);
1015 * this kills every thread in the thread group. Note that any externally
1016 * wait4()-ing process will get the correct exit code - even if this
1017 * thread is not the thread group leader.
1019 SYSCALL_DEFINE1(exit_group, int, error_code)
1021 do_group_exit((error_code & 0xff) << 8);
1026 struct waitid_info {
1034 enum pid_type wo_type;
1038 struct waitid_info *wo_info;
1040 struct rusage *wo_rusage;
1042 wait_queue_entry_t child_wait;
1046 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1048 return wo->wo_type == PIDTYPE_MAX ||
1049 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1053 eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p)
1055 if (!eligible_pid(wo, p))
1059 * Wait for all children (clone and not) if __WALL is set or
1060 * if it is traced by us.
1062 if (ptrace || (wo->wo_flags & __WALL))
1066 * Otherwise, wait for clone children *only* if __WCLONE is set;
1067 * otherwise, wait for non-clone children *only*.
1069 * Note: a "clone" child here is one that reports to its parent
1070 * using a signal other than SIGCHLD, or a non-leader thread which
1071 * we can only see if it is traced by us.
1073 if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1080 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1081 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1082 * the lock and this task is uninteresting. If we return nonzero, we have
1083 * released the lock and the system call should return.
1085 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1088 pid_t pid = task_pid_vnr(p);
1089 uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p));
1090 struct waitid_info *infop;
1092 if (!likely(wo->wo_flags & WEXITED))
1095 if (unlikely(wo->wo_flags & WNOWAIT)) {
1096 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1097 ? p->signal->group_exit_code : p->exit_code;
1099 read_unlock(&tasklist_lock);
1100 sched_annotate_sleep();
1102 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1107 * Move the task's state to DEAD/TRACE, only one thread can do this.
1109 state = (ptrace_reparented(p) && thread_group_leader(p)) ?
1110 EXIT_TRACE : EXIT_DEAD;
1111 if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE)
1114 * We own this thread, nobody else can reap it.
1116 read_unlock(&tasklist_lock);
1117 sched_annotate_sleep();
1120 * Check thread_group_leader() to exclude the traced sub-threads.
1122 if (state == EXIT_DEAD && thread_group_leader(p)) {
1123 struct signal_struct *sig = p->signal;
1124 struct signal_struct *psig = current->signal;
1125 unsigned long maxrss;
1126 u64 tgutime, tgstime;
1129 * The resource counters for the group leader are in its
1130 * own task_struct. Those for dead threads in the group
1131 * are in its signal_struct, as are those for the child
1132 * processes it has previously reaped. All these
1133 * accumulate in the parent's signal_struct c* fields.
1135 * We don't bother to take a lock here to protect these
1136 * p->signal fields because the whole thread group is dead
1137 * and nobody can change them.
1139 * psig->stats_lock also protects us from our sub-threads
1140 * which can reap other children at the same time. Until
1141 * we change k_getrusage()-like users to rely on this lock
1142 * we have to take ->siglock as well.
1144 * We use thread_group_cputime_adjusted() to get times for
1145 * the thread group, which consolidates times for all threads
1146 * in the group including the group leader.
1148 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1149 spin_lock_irq(¤t->sighand->siglock);
1150 write_seqlock(&psig->stats_lock);
1151 psig->cutime += tgutime + sig->cutime;
1152 psig->cstime += tgstime + sig->cstime;
1153 psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime;
1155 p->min_flt + sig->min_flt + sig->cmin_flt;
1157 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1159 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1161 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1163 task_io_get_inblock(p) +
1164 sig->inblock + sig->cinblock;
1166 task_io_get_oublock(p) +
1167 sig->oublock + sig->coublock;
1168 maxrss = max(sig->maxrss, sig->cmaxrss);
1169 if (psig->cmaxrss < maxrss)
1170 psig->cmaxrss = maxrss;
1171 task_io_accounting_add(&psig->ioac, &p->ioac);
1172 task_io_accounting_add(&psig->ioac, &sig->ioac);
1173 write_sequnlock(&psig->stats_lock);
1174 spin_unlock_irq(¤t->sighand->siglock);
1178 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1179 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1180 ? p->signal->group_exit_code : p->exit_code;
1181 wo->wo_stat = status;
1183 if (state == EXIT_TRACE) {
1184 write_lock_irq(&tasklist_lock);
1185 /* We dropped tasklist, ptracer could die and untrace */
1188 /* If parent wants a zombie, don't release it now */
1189 state = EXIT_ZOMBIE;
1190 if (do_notify_parent(p, p->exit_signal))
1192 p->exit_state = state;
1193 write_unlock_irq(&tasklist_lock);
1195 if (state == EXIT_DEAD)
1199 infop = wo->wo_info;
1201 if ((status & 0x7f) == 0) {
1202 infop->cause = CLD_EXITED;
1203 infop->status = status >> 8;
1205 infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1206 infop->status = status & 0x7f;
1215 static int *task_stopped_code(struct task_struct *p, bool ptrace)
1218 if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING))
1219 return &p->exit_code;
1221 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1222 return &p->signal->group_exit_code;
1228 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1230 * @ptrace: is the wait for ptrace
1231 * @p: task to wait for
1233 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1236 * read_lock(&tasklist_lock), which is released if return value is
1237 * non-zero. Also, grabs and releases @p->sighand->siglock.
1240 * 0 if wait condition didn't exist and search for other wait conditions
1241 * should continue. Non-zero return, -errno on failure and @p's pid on
1242 * success, implies that tasklist_lock is released and wait condition
1243 * search should terminate.
1245 static int wait_task_stopped(struct wait_opts *wo,
1246 int ptrace, struct task_struct *p)
1248 struct waitid_info *infop;
1249 int exit_code, *p_code, why;
1250 uid_t uid = 0; /* unneeded, required by compiler */
1254 * Traditionally we see ptrace'd stopped tasks regardless of options.
1256 if (!ptrace && !(wo->wo_flags & WUNTRACED))
1259 if (!task_stopped_code(p, ptrace))
1263 spin_lock_irq(&p->sighand->siglock);
1265 p_code = task_stopped_code(p, ptrace);
1266 if (unlikely(!p_code))
1269 exit_code = *p_code;
1273 if (!unlikely(wo->wo_flags & WNOWAIT))
1276 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1278 spin_unlock_irq(&p->sighand->siglock);
1283 * Now we are pretty sure this task is interesting.
1284 * Make sure it doesn't get reaped out from under us while we
1285 * give up the lock and then examine it below. We don't want to
1286 * keep holding onto the tasklist_lock while we call getrusage and
1287 * possibly take page faults for user memory.
1290 pid = task_pid_vnr(p);
1291 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1292 read_unlock(&tasklist_lock);
1293 sched_annotate_sleep();
1295 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1298 if (likely(!(wo->wo_flags & WNOWAIT)))
1299 wo->wo_stat = (exit_code << 8) | 0x7f;
1301 infop = wo->wo_info;
1304 infop->status = exit_code;
1312 * Handle do_wait work for one task in a live, non-stopped state.
1313 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1314 * the lock and this task is uninteresting. If we return nonzero, we have
1315 * released the lock and the system call should return.
1317 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1319 struct waitid_info *infop;
1323 if (!unlikely(wo->wo_flags & WCONTINUED))
1326 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1329 spin_lock_irq(&p->sighand->siglock);
1330 /* Re-check with the lock held. */
1331 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1332 spin_unlock_irq(&p->sighand->siglock);
1335 if (!unlikely(wo->wo_flags & WNOWAIT))
1336 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1337 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1338 spin_unlock_irq(&p->sighand->siglock);
1340 pid = task_pid_vnr(p);
1342 read_unlock(&tasklist_lock);
1343 sched_annotate_sleep();
1345 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1348 infop = wo->wo_info;
1350 wo->wo_stat = 0xffff;
1352 infop->cause = CLD_CONTINUED;
1355 infop->status = SIGCONT;
1361 * Consider @p for a wait by @parent.
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;
1366 * then ->notask_error is 0 if @p is an eligible child,
1369 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1370 struct task_struct *p)
1373 * We can race with wait_task_zombie() from another thread.
1374 * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition
1375 * can't confuse the checks below.
1377 int exit_state = READ_ONCE(p->exit_state);
1380 if (unlikely(exit_state == EXIT_DEAD))
1383 ret = eligible_child(wo, ptrace, p);
1387 if (unlikely(exit_state == EXIT_TRACE)) {
1389 * ptrace == 0 means we are the natural parent. In this case
1390 * we should clear notask_error, debugger will notify us.
1392 if (likely(!ptrace))
1393 wo->notask_error = 0;
1397 if (likely(!ptrace) && unlikely(p->ptrace)) {
1399 * If it is traced by its real parent's group, just pretend
1400 * the caller is ptrace_do_wait() and reap this child if it
1403 * This also hides group stop state from real parent; otherwise
1404 * a single stop can be reported twice as group and ptrace stop.
1405 * If a ptracer wants to distinguish these two events for its
1406 * own children it should create a separate process which takes
1407 * the role of real parent.
1409 if (!ptrace_reparented(p))
1414 if (exit_state == EXIT_ZOMBIE) {
1415 /* we don't reap group leaders with subthreads */
1416 if (!delay_group_leader(p)) {
1418 * A zombie ptracee is only visible to its ptracer.
1419 * Notification and reaping will be cascaded to the
1420 * real parent when the ptracer detaches.
1422 if (unlikely(ptrace) || likely(!p->ptrace))
1423 return wait_task_zombie(wo, p);
1427 * Allow access to stopped/continued state via zombie by
1428 * falling through. Clearing of notask_error is complex.
1432 * If WEXITED is set, notask_error should naturally be
1433 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1434 * so, if there are live subthreads, there are events to
1435 * wait for. If all subthreads are dead, it's still safe
1436 * to clear - this function will be called again in finite
1437 * amount time once all the subthreads are released and
1438 * will then return without clearing.
1442 * Stopped state is per-task and thus can't change once the
1443 * target task dies. Only continued and exited can happen.
1444 * Clear notask_error if WCONTINUED | WEXITED.
1446 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1447 wo->notask_error = 0;
1450 * @p is alive and it's gonna stop, continue or exit, so
1451 * there always is something to wait for.
1453 wo->notask_error = 0;
1457 * Wait for stopped. Depending on @ptrace, different stopped state
1458 * is used and the two don't interact with each other.
1460 ret = wait_task_stopped(wo, ptrace, p);
1465 * Wait for continued. There's only one continued state and the
1466 * ptracer can consume it which can confuse the real parent. Don't
1467 * use WCONTINUED from ptracer. You don't need or want it.
1469 return wait_task_continued(wo, p);
1473 * Do the work of do_wait() for one thread in the group, @tsk.
1475 * -ECHILD should be in ->notask_error before the first call.
1476 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1477 * Returns zero if the search for a child should continue; then
1478 * ->notask_error is 0 if there were any eligible children,
1481 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1483 struct task_struct *p;
1485 list_for_each_entry(p, &tsk->children, sibling) {
1486 int ret = wait_consider_task(wo, 0, p);
1495 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1497 struct task_struct *p;
1499 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1500 int ret = wait_consider_task(wo, 1, p);
1509 static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode,
1510 int sync, void *key)
1512 struct wait_opts *wo = container_of(wait, struct wait_opts,
1514 struct task_struct *p = key;
1516 if (!eligible_pid(wo, p))
1519 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1522 return default_wake_function(wait, mode, sync, key);
1525 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1527 __wake_up_sync_key(&parent->signal->wait_chldexit,
1528 TASK_INTERRUPTIBLE, p);
1531 static bool is_effectively_child(struct wait_opts *wo, bool ptrace,
1532 struct task_struct *target)
1534 struct task_struct *parent =
1535 !ptrace ? target->real_parent : target->parent;
1537 return current == parent || (!(wo->wo_flags & __WNOTHREAD) &&
1538 same_thread_group(current, parent));
1542 * Optimization for waiting on PIDTYPE_PID. No need to iterate through child
1543 * and tracee lists to find the target task.
1545 static int do_wait_pid(struct wait_opts *wo)
1548 struct task_struct *target;
1552 target = pid_task(wo->wo_pid, PIDTYPE_TGID);
1553 if (target && is_effectively_child(wo, ptrace, target)) {
1554 retval = wait_consider_task(wo, ptrace, target);
1560 target = pid_task(wo->wo_pid, PIDTYPE_PID);
1561 if (target && target->ptrace &&
1562 is_effectively_child(wo, ptrace, target)) {
1563 retval = wait_consider_task(wo, ptrace, target);
1571 static long do_wait(struct wait_opts *wo)
1575 trace_sched_process_wait(wo->wo_pid);
1577 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1578 wo->child_wait.private = current;
1579 add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1582 * If there is nothing that can match our criteria, just get out.
1583 * We will clear ->notask_error to zero if we see any child that
1584 * might later match our criteria, even if we are not able to reap
1587 wo->notask_error = -ECHILD;
1588 if ((wo->wo_type < PIDTYPE_MAX) &&
1589 (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type)))
1592 set_current_state(TASK_INTERRUPTIBLE);
1593 read_lock(&tasklist_lock);
1595 if (wo->wo_type == PIDTYPE_PID) {
1596 retval = do_wait_pid(wo);
1600 struct task_struct *tsk = current;
1603 retval = do_wait_thread(wo, tsk);
1607 retval = ptrace_do_wait(wo, tsk);
1611 if (wo->wo_flags & __WNOTHREAD)
1613 } while_each_thread(current, tsk);
1615 read_unlock(&tasklist_lock);
1618 retval = wo->notask_error;
1619 if (!retval && !(wo->wo_flags & WNOHANG)) {
1620 retval = -ERESTARTSYS;
1621 if (!signal_pending(current)) {
1627 __set_current_state(TASK_RUNNING);
1628 remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1632 static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop,
1633 int options, struct rusage *ru)
1635 struct wait_opts wo;
1636 struct pid *pid = NULL;
1639 unsigned int f_flags = 0;
1641 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED|
1642 __WNOTHREAD|__WCLONE|__WALL))
1644 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1656 pid = find_get_pid(upid);
1659 type = PIDTYPE_PGID;
1664 pid = find_get_pid(upid);
1666 pid = get_task_pid(current, PIDTYPE_PGID);
1673 pid = pidfd_get_pid(upid, &f_flags);
1675 return PTR_ERR(pid);
1684 wo.wo_flags = options;
1687 if (f_flags & O_NONBLOCK)
1688 wo.wo_flags |= WNOHANG;
1691 if (!ret && !(options & WNOHANG) && (f_flags & O_NONBLOCK))
1698 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1699 infop, int, options, struct rusage __user *, ru)
1702 struct waitid_info info = {.status = 0};
1703 long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL);
1709 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
1715 if (!user_write_access_begin(infop, sizeof(*infop)))
1718 unsafe_put_user(signo, &infop->si_signo, Efault);
1719 unsafe_put_user(0, &infop->si_errno, Efault);
1720 unsafe_put_user(info.cause, &infop->si_code, Efault);
1721 unsafe_put_user(info.pid, &infop->si_pid, Efault);
1722 unsafe_put_user(info.uid, &infop->si_uid, Efault);
1723 unsafe_put_user(info.status, &infop->si_status, Efault);
1724 user_write_access_end();
1727 user_write_access_end();
1731 long kernel_wait4(pid_t upid, int __user *stat_addr, int options,
1734 struct wait_opts wo;
1735 struct pid *pid = NULL;
1739 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1740 __WNOTHREAD|__WCLONE|__WALL))
1743 /* -INT_MIN is not defined */
1744 if (upid == INT_MIN)
1749 else if (upid < 0) {
1750 type = PIDTYPE_PGID;
1751 pid = find_get_pid(-upid);
1752 } else if (upid == 0) {
1753 type = PIDTYPE_PGID;
1754 pid = get_task_pid(current, PIDTYPE_PGID);
1755 } else /* upid > 0 */ {
1757 pid = find_get_pid(upid);
1762 wo.wo_flags = options | WEXITED;
1768 if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr))
1774 int kernel_wait(pid_t pid, int *stat)
1776 struct wait_opts wo = {
1777 .wo_type = PIDTYPE_PID,
1778 .wo_pid = find_get_pid(pid),
1779 .wo_flags = WEXITED,
1784 if (ret > 0 && wo.wo_stat)
1790 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1791 int, options, struct rusage __user *, ru)
1794 long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL);
1797 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
1803 #ifdef __ARCH_WANT_SYS_WAITPID
1806 * sys_waitpid() remains for compatibility. waitpid() should be
1807 * implemented by calling sys_wait4() from libc.a.
1809 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1811 return kernel_wait4(pid, stat_addr, options, NULL);
1816 #ifdef CONFIG_COMPAT
1817 COMPAT_SYSCALL_DEFINE4(wait4,
1819 compat_uint_t __user *, stat_addr,
1821 struct compat_rusage __user *, ru)
1824 long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL);
1826 if (ru && put_compat_rusage(&r, ru))
1832 COMPAT_SYSCALL_DEFINE5(waitid,
1833 int, which, compat_pid_t, pid,
1834 struct compat_siginfo __user *, infop, int, options,
1835 struct compat_rusage __user *, uru)
1838 struct waitid_info info = {.status = 0};
1839 long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL);
1845 /* kernel_waitid() overwrites everything in ru */
1846 if (COMPAT_USE_64BIT_TIME)
1847 err = copy_to_user(uru, &ru, sizeof(ru));
1849 err = put_compat_rusage(&ru, uru);
1858 if (!user_write_access_begin(infop, sizeof(*infop)))
1861 unsafe_put_user(signo, &infop->si_signo, Efault);
1862 unsafe_put_user(0, &infop->si_errno, Efault);
1863 unsafe_put_user(info.cause, &infop->si_code, Efault);
1864 unsafe_put_user(info.pid, &infop->si_pid, Efault);
1865 unsafe_put_user(info.uid, &infop->si_uid, Efault);
1866 unsafe_put_user(info.status, &infop->si_status, Efault);
1867 user_write_access_end();
1870 user_write_access_end();
1876 * thread_group_exited - check that a thread group has exited
1877 * @pid: tgid of thread group to be checked.
1879 * Test if the thread group represented by tgid has exited (all
1880 * threads are zombies, dead or completely gone).
1882 * Return: true if the thread group has exited. false otherwise.
1884 bool thread_group_exited(struct pid *pid)
1886 struct task_struct *task;
1890 task = pid_task(pid, PIDTYPE_PID);
1892 (READ_ONCE(task->exit_state) && thread_group_empty(task));
1897 EXPORT_SYMBOL(thread_group_exited);
1899 __weak void abort(void)
1903 /* if that doesn't kill us, halt */
1904 panic("Oops failed to kill thread");
1906 EXPORT_SYMBOL(abort);