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>
71 #include <linux/uaccess.h>
72 #include <asm/unistd.h>
73 #include <asm/mmu_context.h>
76 * The default value should be high enough to not crash a system that randomly
77 * crashes its kernel from time to time, but low enough to at least not permit
78 * overflowing 32-bit refcounts or the ldsem writer count.
80 static unsigned int oops_limit = 10000;
83 static struct ctl_table kern_exit_table[] = {
85 .procname = "oops_limit",
87 .maxlen = sizeof(oops_limit),
89 .proc_handler = proc_douintvec,
94 static __init int kernel_exit_sysctls_init(void)
96 register_sysctl_init("kernel", kern_exit_table);
99 late_initcall(kernel_exit_sysctls_init);
102 static void __unhash_process(struct task_struct *p, bool group_dead)
105 detach_pid(p, PIDTYPE_PID);
107 detach_pid(p, PIDTYPE_TGID);
108 detach_pid(p, PIDTYPE_PGID);
109 detach_pid(p, PIDTYPE_SID);
111 list_del_rcu(&p->tasks);
112 list_del_init(&p->sibling);
113 __this_cpu_dec(process_counts);
115 list_del_rcu(&p->thread_group);
116 list_del_rcu(&p->thread_node);
120 * This function expects the tasklist_lock write-locked.
122 static void __exit_signal(struct task_struct *tsk)
124 struct signal_struct *sig = tsk->signal;
125 bool group_dead = thread_group_leader(tsk);
126 struct sighand_struct *sighand;
127 struct tty_struct *tty;
130 sighand = rcu_dereference_check(tsk->sighand,
131 lockdep_tasklist_lock_is_held());
132 spin_lock(&sighand->siglock);
134 #ifdef CONFIG_POSIX_TIMERS
135 posix_cpu_timers_exit(tsk);
137 posix_cpu_timers_exit_group(tsk);
145 * If there is any task waiting for the group exit
148 if (sig->notify_count > 0 && !--sig->notify_count)
149 wake_up_process(sig->group_exec_task);
151 if (tsk == sig->curr_target)
152 sig->curr_target = next_thread(tsk);
155 add_device_randomness((const void*) &tsk->se.sum_exec_runtime,
156 sizeof(unsigned long long));
159 * Accumulate here the counters for all threads as they die. We could
160 * skip the group leader because it is the last user of signal_struct,
161 * but we want to avoid the race with thread_group_cputime() which can
162 * see the empty ->thread_head list.
164 task_cputime(tsk, &utime, &stime);
165 write_seqlock(&sig->stats_lock);
168 sig->gtime += task_gtime(tsk);
169 sig->min_flt += tsk->min_flt;
170 sig->maj_flt += tsk->maj_flt;
171 sig->nvcsw += tsk->nvcsw;
172 sig->nivcsw += tsk->nivcsw;
173 sig->inblock += task_io_get_inblock(tsk);
174 sig->oublock += task_io_get_oublock(tsk);
175 task_io_accounting_add(&sig->ioac, &tsk->ioac);
176 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
178 __unhash_process(tsk, group_dead);
179 write_sequnlock(&sig->stats_lock);
182 * Do this under ->siglock, we can race with another thread
183 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
185 flush_sigqueue(&tsk->pending);
187 spin_unlock(&sighand->siglock);
189 __cleanup_sighand(sighand);
190 clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
192 flush_sigqueue(&sig->shared_pending);
197 static void delayed_put_task_struct(struct rcu_head *rhp)
199 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
201 kprobe_flush_task(tsk);
202 rethook_flush_task(tsk);
203 perf_event_delayed_put(tsk);
204 trace_sched_process_free(tsk);
205 put_task_struct(tsk);
208 void put_task_struct_rcu_user(struct task_struct *task)
210 if (refcount_dec_and_test(&task->rcu_users))
211 call_rcu(&task->rcu, delayed_put_task_struct);
214 void __weak release_thread(struct task_struct *dead_task)
218 void release_task(struct task_struct *p)
220 struct task_struct *leader;
221 struct pid *thread_pid;
224 /* don't need to get the RCU readlock here - the process is dead and
225 * can't be modifying its own credentials. But shut RCU-lockdep up */
227 dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1);
232 write_lock_irq(&tasklist_lock);
233 ptrace_release_task(p);
234 thread_pid = get_pid(p->thread_pid);
238 * If we are the last non-leader member of the thread
239 * group, and the leader is zombie, then notify the
240 * group leader's parent process. (if it wants notification.)
243 leader = p->group_leader;
244 if (leader != p && thread_group_empty(leader)
245 && leader->exit_state == EXIT_ZOMBIE) {
247 * If we were the last child thread and the leader has
248 * exited already, and the leader's parent ignores SIGCHLD,
249 * then we are the one who should release the leader.
251 zap_leader = do_notify_parent(leader, leader->exit_signal);
253 leader->exit_state = EXIT_DEAD;
256 write_unlock_irq(&tasklist_lock);
257 seccomp_filter_release(p);
258 proc_flush_pid(thread_pid);
261 put_task_struct_rcu_user(p);
264 if (unlikely(zap_leader))
268 int rcuwait_wake_up(struct rcuwait *w)
271 struct task_struct *task;
276 * Order condition vs @task, such that everything prior to the load
277 * of @task is visible. This is the condition as to why the user called
278 * rcuwait_wake() in the first place. Pairs with set_current_state()
279 * barrier (A) in rcuwait_wait_event().
282 * [S] tsk = current [S] cond = true
288 task = rcu_dereference(w->task);
290 ret = wake_up_process(task);
295 EXPORT_SYMBOL_GPL(rcuwait_wake_up);
298 * Determine if a process group is "orphaned", according to the POSIX
299 * definition in 2.2.2.52. Orphaned process groups are not to be affected
300 * by terminal-generated stop signals. Newly orphaned process groups are
301 * to receive a SIGHUP and a SIGCONT.
303 * "I ask you, have you ever known what it is to be an orphan?"
305 static int will_become_orphaned_pgrp(struct pid *pgrp,
306 struct task_struct *ignored_task)
308 struct task_struct *p;
310 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
311 if ((p == ignored_task) ||
312 (p->exit_state && thread_group_empty(p)) ||
313 is_global_init(p->real_parent))
316 if (task_pgrp(p->real_parent) != pgrp &&
317 task_session(p->real_parent) == task_session(p))
319 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
324 int is_current_pgrp_orphaned(void)
328 read_lock(&tasklist_lock);
329 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
330 read_unlock(&tasklist_lock);
335 static bool has_stopped_jobs(struct pid *pgrp)
337 struct task_struct *p;
339 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
340 if (p->signal->flags & SIGNAL_STOP_STOPPED)
342 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
348 * Check to see if any process groups have become orphaned as
349 * a result of our exiting, and if they have any stopped jobs,
350 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
353 kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
355 struct pid *pgrp = task_pgrp(tsk);
356 struct task_struct *ignored_task = tsk;
359 /* exit: our father is in a different pgrp than
360 * we are and we were the only connection outside.
362 parent = tsk->real_parent;
364 /* reparent: our child is in a different pgrp than
365 * we are, and it was the only connection outside.
369 if (task_pgrp(parent) != pgrp &&
370 task_session(parent) == task_session(tsk) &&
371 will_become_orphaned_pgrp(pgrp, ignored_task) &&
372 has_stopped_jobs(pgrp)) {
373 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
374 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
378 static void coredump_task_exit(struct task_struct *tsk)
380 struct core_state *core_state;
383 * Serialize with any possible pending coredump.
384 * We must hold siglock around checking core_state
385 * and setting PF_POSTCOREDUMP. The core-inducing thread
386 * will increment ->nr_threads for each thread in the
387 * group without PF_POSTCOREDUMP set.
389 spin_lock_irq(&tsk->sighand->siglock);
390 tsk->flags |= PF_POSTCOREDUMP;
391 core_state = tsk->signal->core_state;
392 spin_unlock_irq(&tsk->sighand->siglock);
394 struct core_thread self;
397 if (self.task->flags & PF_SIGNALED)
398 self.next = xchg(&core_state->dumper.next, &self);
402 * Implies mb(), the result of xchg() must be visible
403 * to core_state->dumper.
405 if (atomic_dec_and_test(&core_state->nr_threads))
406 complete(&core_state->startup);
409 set_current_state(TASK_UNINTERRUPTIBLE|TASK_FREEZABLE);
410 if (!self.task) /* see coredump_finish() */
414 __set_current_state(TASK_RUNNING);
420 * A task is exiting. If it owned this mm, find a new owner for the mm.
422 void mm_update_next_owner(struct mm_struct *mm)
424 struct task_struct *c, *g, *p = current;
428 * If the exiting or execing task is not the owner, it's
429 * someone else's problem.
434 * The current owner is exiting/execing and there are no other
435 * candidates. Do not leave the mm pointing to a possibly
436 * freed task structure.
438 if (atomic_read(&mm->mm_users) <= 1) {
439 WRITE_ONCE(mm->owner, NULL);
443 read_lock(&tasklist_lock);
445 * Search in the children
447 list_for_each_entry(c, &p->children, sibling) {
449 goto assign_new_owner;
453 * Search in the siblings
455 list_for_each_entry(c, &p->real_parent->children, sibling) {
457 goto assign_new_owner;
461 * Search through everything else, we should not get here often.
463 for_each_process(g) {
464 if (g->flags & PF_KTHREAD)
466 for_each_thread(g, c) {
468 goto assign_new_owner;
473 read_unlock(&tasklist_lock);
475 * We found no owner yet mm_users > 1: this implies that we are
476 * most likely racing with swapoff (try_to_unuse()) or /proc or
477 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
479 WRITE_ONCE(mm->owner, NULL);
486 * The task_lock protects c->mm from changing.
487 * We always want mm->owner->mm == mm
491 * Delay read_unlock() till we have the task_lock()
492 * to ensure that c does not slip away underneath us
494 read_unlock(&tasklist_lock);
500 WRITE_ONCE(mm->owner, c);
501 lru_gen_migrate_mm(mm);
505 #endif /* CONFIG_MEMCG */
508 * Turn us into a lazy TLB process if we
511 static void exit_mm(void)
513 struct mm_struct *mm = current->mm;
515 exit_mm_release(current, mm);
521 BUG_ON(mm != current->active_mm);
522 /* more a memory barrier than a real lock */
525 * When a thread stops operating on an address space, the loop
526 * in membarrier_private_expedited() may not observe that
527 * tsk->mm, and the loop in membarrier_global_expedited() may
528 * not observe a MEMBARRIER_STATE_GLOBAL_EXPEDITED
529 * rq->membarrier_state, so those would not issue an IPI.
530 * Membarrier requires a memory barrier after accessing
531 * user-space memory, before clearing tsk->mm or the
532 * rq->membarrier_state.
534 smp_mb__after_spinlock();
537 membarrier_update_current_mm(NULL);
538 enter_lazy_tlb(mm, current);
540 task_unlock(current);
541 mmap_read_unlock(mm);
542 mm_update_next_owner(mm);
544 if (test_thread_flag(TIF_MEMDIE))
548 static struct task_struct *find_alive_thread(struct task_struct *p)
550 struct task_struct *t;
552 for_each_thread(p, t) {
553 if (!(t->flags & PF_EXITING))
559 static struct task_struct *find_child_reaper(struct task_struct *father,
560 struct list_head *dead)
561 __releases(&tasklist_lock)
562 __acquires(&tasklist_lock)
564 struct pid_namespace *pid_ns = task_active_pid_ns(father);
565 struct task_struct *reaper = pid_ns->child_reaper;
566 struct task_struct *p, *n;
568 if (likely(reaper != father))
571 reaper = find_alive_thread(father);
573 pid_ns->child_reaper = reaper;
577 write_unlock_irq(&tasklist_lock);
579 list_for_each_entry_safe(p, n, dead, ptrace_entry) {
580 list_del_init(&p->ptrace_entry);
584 zap_pid_ns_processes(pid_ns);
585 write_lock_irq(&tasklist_lock);
591 * When we die, we re-parent all our children, and try to:
592 * 1. give them to another thread in our thread group, if such a member exists
593 * 2. give it to the first ancestor process which prctl'd itself as a
594 * child_subreaper for its children (like a service manager)
595 * 3. give it to the init process (PID 1) in our pid namespace
597 static struct task_struct *find_new_reaper(struct task_struct *father,
598 struct task_struct *child_reaper)
600 struct task_struct *thread, *reaper;
602 thread = find_alive_thread(father);
606 if (father->signal->has_child_subreaper) {
607 unsigned int ns_level = task_pid(father)->level;
609 * Find the first ->is_child_subreaper ancestor in our pid_ns.
610 * We can't check reaper != child_reaper to ensure we do not
611 * cross the namespaces, the exiting parent could be injected
612 * by setns() + fork().
613 * We check pid->level, this is slightly more efficient than
614 * task_active_pid_ns(reaper) != task_active_pid_ns(father).
616 for (reaper = father->real_parent;
617 task_pid(reaper)->level == ns_level;
618 reaper = reaper->real_parent) {
619 if (reaper == &init_task)
621 if (!reaper->signal->is_child_subreaper)
623 thread = find_alive_thread(reaper);
633 * Any that need to be release_task'd are put on the @dead list.
635 static void reparent_leader(struct task_struct *father, struct task_struct *p,
636 struct list_head *dead)
638 if (unlikely(p->exit_state == EXIT_DEAD))
641 /* We don't want people slaying init. */
642 p->exit_signal = SIGCHLD;
644 /* If it has exited notify the new parent about this child's death. */
646 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
647 if (do_notify_parent(p, p->exit_signal)) {
648 p->exit_state = EXIT_DEAD;
649 list_add(&p->ptrace_entry, dead);
653 kill_orphaned_pgrp(p, father);
657 * This does two things:
659 * A. Make init inherit all the child processes
660 * B. Check to see if any process groups have become orphaned
661 * as a result of our exiting, and if they have any stopped
662 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
664 static void forget_original_parent(struct task_struct *father,
665 struct list_head *dead)
667 struct task_struct *p, *t, *reaper;
669 if (unlikely(!list_empty(&father->ptraced)))
670 exit_ptrace(father, dead);
672 /* Can drop and reacquire tasklist_lock */
673 reaper = find_child_reaper(father, dead);
674 if (list_empty(&father->children))
677 reaper = find_new_reaper(father, reaper);
678 list_for_each_entry(p, &father->children, sibling) {
679 for_each_thread(p, t) {
680 RCU_INIT_POINTER(t->real_parent, reaper);
681 BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father));
682 if (likely(!t->ptrace))
683 t->parent = t->real_parent;
684 if (t->pdeath_signal)
685 group_send_sig_info(t->pdeath_signal,
690 * If this is a threaded reparent there is no need to
691 * notify anyone anything has happened.
693 if (!same_thread_group(reaper, father))
694 reparent_leader(father, p, dead);
696 list_splice_tail_init(&father->children, &reaper->children);
700 * Send signals to all our closest relatives so that they know
701 * to properly mourn us..
703 static void exit_notify(struct task_struct *tsk, int group_dead)
706 struct task_struct *p, *n;
709 write_lock_irq(&tasklist_lock);
710 forget_original_parent(tsk, &dead);
713 kill_orphaned_pgrp(tsk->group_leader, NULL);
715 tsk->exit_state = EXIT_ZOMBIE;
716 if (unlikely(tsk->ptrace)) {
717 int sig = thread_group_leader(tsk) &&
718 thread_group_empty(tsk) &&
719 !ptrace_reparented(tsk) ?
720 tsk->exit_signal : SIGCHLD;
721 autoreap = do_notify_parent(tsk, sig);
722 } else if (thread_group_leader(tsk)) {
723 autoreap = thread_group_empty(tsk) &&
724 do_notify_parent(tsk, tsk->exit_signal);
730 tsk->exit_state = EXIT_DEAD;
731 list_add(&tsk->ptrace_entry, &dead);
734 /* mt-exec, de_thread() is waiting for group leader */
735 if (unlikely(tsk->signal->notify_count < 0))
736 wake_up_process(tsk->signal->group_exec_task);
737 write_unlock_irq(&tasklist_lock);
739 list_for_each_entry_safe(p, n, &dead, ptrace_entry) {
740 list_del_init(&p->ptrace_entry);
745 #ifdef CONFIG_DEBUG_STACK_USAGE
746 static void check_stack_usage(void)
748 static DEFINE_SPINLOCK(low_water_lock);
749 static int lowest_to_date = THREAD_SIZE;
752 free = stack_not_used(current);
754 if (free >= lowest_to_date)
757 spin_lock(&low_water_lock);
758 if (free < lowest_to_date) {
759 pr_info("%s (%d) used greatest stack depth: %lu bytes left\n",
760 current->comm, task_pid_nr(current), free);
761 lowest_to_date = free;
763 spin_unlock(&low_water_lock);
766 static inline void check_stack_usage(void) {}
769 static void synchronize_group_exit(struct task_struct *tsk, long code)
771 struct sighand_struct *sighand = tsk->sighand;
772 struct signal_struct *signal = tsk->signal;
774 spin_lock_irq(&sighand->siglock);
775 signal->quick_threads--;
776 if ((signal->quick_threads == 0) &&
777 !(signal->flags & SIGNAL_GROUP_EXIT)) {
778 signal->flags = SIGNAL_GROUP_EXIT;
779 signal->group_exit_code = code;
780 signal->group_stop_count = 0;
782 spin_unlock_irq(&sighand->siglock);
785 void __noreturn do_exit(long code)
787 struct task_struct *tsk = current;
790 synchronize_group_exit(tsk, code);
795 kmsan_task_exit(tsk);
797 coredump_task_exit(tsk);
798 ptrace_event(PTRACE_EVENT_EXIT, code);
800 validate_creds_for_do_exit(tsk);
802 io_uring_files_cancel();
803 exit_signals(tsk); /* sets PF_EXITING */
805 /* sync mm's RSS info before statistics gathering */
807 sync_mm_rss(tsk->mm);
808 acct_update_integrals(tsk);
809 group_dead = atomic_dec_and_test(&tsk->signal->live);
812 * If the last thread of global init has exited, panic
813 * immediately to get a useable coredump.
815 if (unlikely(is_global_init(tsk)))
816 panic("Attempted to kill init! exitcode=0x%08x\n",
817 tsk->signal->group_exit_code ?: (int)code);
819 #ifdef CONFIG_POSIX_TIMERS
820 hrtimer_cancel(&tsk->signal->real_timer);
824 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
826 acct_collect(code, group_dead);
831 tsk->exit_code = code;
832 taskstats_exit(tsk, group_dead);
838 trace_sched_process_exit(tsk);
845 disassociate_ctty(1);
846 exit_task_namespaces(tsk);
851 * Flush inherited counters to the parent - before the parent
852 * gets woken up by child-exit notifications.
854 * because of cgroup mode, must be called before cgroup_exit()
856 perf_event_exit_task(tsk);
858 sched_autogroup_exit_task(tsk);
862 * FIXME: do that only when needed, using sched_exit tracepoint
864 flush_ptrace_hw_breakpoint(tsk);
866 exit_tasks_rcu_start();
867 exit_notify(tsk, group_dead);
868 proc_exit_connector(tsk);
869 mpol_put_task_policy(tsk);
871 if (unlikely(current->pi_state_cache))
872 kfree(current->pi_state_cache);
875 * Make sure we are holding no locks:
877 debug_check_no_locks_held();
880 exit_io_context(tsk);
882 if (tsk->splice_pipe)
883 free_pipe_info(tsk->splice_pipe);
885 if (tsk->task_frag.page)
886 put_page(tsk->task_frag.page);
888 validate_creds_for_do_exit(tsk);
889 exit_task_stack_account(tsk);
894 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
896 exit_tasks_rcu_finish();
898 lockdep_free_task(tsk);
902 void __noreturn make_task_dead(int signr)
904 static atomic_t oops_count = ATOMIC_INIT(0);
907 * Take the task off the cpu after something catastrophic has
910 * We can get here from a kernel oops, sometimes with preemption off.
911 * Start by checking for critical errors.
912 * Then fix up important state like USER_DS and preemption.
913 * Then do everything else.
915 struct task_struct *tsk = current;
917 if (unlikely(in_interrupt()))
918 panic("Aiee, killing interrupt handler!");
919 if (unlikely(!tsk->pid))
920 panic("Attempted to kill the idle task!");
922 if (unlikely(in_atomic())) {
923 pr_info("note: %s[%d] exited with preempt_count %d\n",
924 current->comm, task_pid_nr(current),
926 preempt_count_set(PREEMPT_ENABLED);
930 * Every time the system oopses, if the oops happens while a reference
931 * to an object was held, the reference leaks.
932 * If the oops doesn't also leak memory, repeated oopsing can cause
933 * reference counters to wrap around (if they're not using refcount_t).
934 * This means that repeated oopsing can make unexploitable-looking bugs
935 * exploitable through repeated oopsing.
936 * To make sure this can't happen, place an upper bound on how often the
937 * kernel may oops without panic().
939 if (atomic_inc_return(&oops_count) >= READ_ONCE(oops_limit))
940 panic("Oopsed too often (kernel.oops_limit is %d)", oops_limit);
943 * We're taking recursive faults here in make_task_dead. Safest is to just
944 * leave this task alone and wait for reboot.
946 if (unlikely(tsk->flags & PF_EXITING)) {
947 pr_alert("Fixing recursive fault but reboot is needed!\n");
948 futex_exit_recursive(tsk);
949 tsk->exit_state = EXIT_DEAD;
950 refcount_inc(&tsk->rcu_users);
957 SYSCALL_DEFINE1(exit, int, error_code)
959 do_exit((error_code&0xff)<<8);
963 * Take down every thread in the group. This is called by fatal signals
964 * as well as by sys_exit_group (below).
967 do_group_exit(int exit_code)
969 struct signal_struct *sig = current->signal;
971 if (sig->flags & SIGNAL_GROUP_EXIT)
972 exit_code = sig->group_exit_code;
973 else if (sig->group_exec_task)
976 struct sighand_struct *const sighand = current->sighand;
978 spin_lock_irq(&sighand->siglock);
979 if (sig->flags & SIGNAL_GROUP_EXIT)
980 /* Another thread got here before we took the lock. */
981 exit_code = sig->group_exit_code;
982 else if (sig->group_exec_task)
985 sig->group_exit_code = exit_code;
986 sig->flags = SIGNAL_GROUP_EXIT;
987 zap_other_threads(current);
989 spin_unlock_irq(&sighand->siglock);
997 * this kills every thread in the thread group. Note that any externally
998 * wait4()-ing process will get the correct exit code - even if this
999 * thread is not the thread group leader.
1001 SYSCALL_DEFINE1(exit_group, int, error_code)
1003 do_group_exit((error_code & 0xff) << 8);
1008 struct waitid_info {
1016 enum pid_type wo_type;
1020 struct waitid_info *wo_info;
1022 struct rusage *wo_rusage;
1024 wait_queue_entry_t child_wait;
1028 static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1030 return wo->wo_type == PIDTYPE_MAX ||
1031 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1035 eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p)
1037 if (!eligible_pid(wo, p))
1041 * Wait for all children (clone and not) if __WALL is set or
1042 * if it is traced by us.
1044 if (ptrace || (wo->wo_flags & __WALL))
1048 * Otherwise, wait for clone children *only* if __WCLONE is set;
1049 * otherwise, wait for non-clone children *only*.
1051 * Note: a "clone" child here is one that reports to its parent
1052 * using a signal other than SIGCHLD, or a non-leader thread which
1053 * we can only see if it is traced by us.
1055 if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1062 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1063 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1064 * the lock and this task is uninteresting. If we return nonzero, we have
1065 * released the lock and the system call should return.
1067 static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1070 pid_t pid = task_pid_vnr(p);
1071 uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p));
1072 struct waitid_info *infop;
1074 if (!likely(wo->wo_flags & WEXITED))
1077 if (unlikely(wo->wo_flags & WNOWAIT)) {
1078 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1079 ? p->signal->group_exit_code : p->exit_code;
1081 read_unlock(&tasklist_lock);
1082 sched_annotate_sleep();
1084 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1089 * Move the task's state to DEAD/TRACE, only one thread can do this.
1091 state = (ptrace_reparented(p) && thread_group_leader(p)) ?
1092 EXIT_TRACE : EXIT_DEAD;
1093 if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE)
1096 * We own this thread, nobody else can reap it.
1098 read_unlock(&tasklist_lock);
1099 sched_annotate_sleep();
1102 * Check thread_group_leader() to exclude the traced sub-threads.
1104 if (state == EXIT_DEAD && thread_group_leader(p)) {
1105 struct signal_struct *sig = p->signal;
1106 struct signal_struct *psig = current->signal;
1107 unsigned long maxrss;
1108 u64 tgutime, tgstime;
1111 * The resource counters for the group leader are in its
1112 * own task_struct. Those for dead threads in the group
1113 * are in its signal_struct, as are those for the child
1114 * processes it has previously reaped. All these
1115 * accumulate in the parent's signal_struct c* fields.
1117 * We don't bother to take a lock here to protect these
1118 * p->signal fields because the whole thread group is dead
1119 * and nobody can change them.
1121 * psig->stats_lock also protects us from our sub-threads
1122 * which can reap other children at the same time. Until
1123 * we change k_getrusage()-like users to rely on this lock
1124 * we have to take ->siglock as well.
1126 * We use thread_group_cputime_adjusted() to get times for
1127 * the thread group, which consolidates times for all threads
1128 * in the group including the group leader.
1130 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1131 spin_lock_irq(¤t->sighand->siglock);
1132 write_seqlock(&psig->stats_lock);
1133 psig->cutime += tgutime + sig->cutime;
1134 psig->cstime += tgstime + sig->cstime;
1135 psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime;
1137 p->min_flt + sig->min_flt + sig->cmin_flt;
1139 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1141 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1143 p->nivcsw + sig->nivcsw + sig->cnivcsw;
1145 task_io_get_inblock(p) +
1146 sig->inblock + sig->cinblock;
1148 task_io_get_oublock(p) +
1149 sig->oublock + sig->coublock;
1150 maxrss = max(sig->maxrss, sig->cmaxrss);
1151 if (psig->cmaxrss < maxrss)
1152 psig->cmaxrss = maxrss;
1153 task_io_accounting_add(&psig->ioac, &p->ioac);
1154 task_io_accounting_add(&psig->ioac, &sig->ioac);
1155 write_sequnlock(&psig->stats_lock);
1156 spin_unlock_irq(¤t->sighand->siglock);
1160 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1161 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1162 ? p->signal->group_exit_code : p->exit_code;
1163 wo->wo_stat = status;
1165 if (state == EXIT_TRACE) {
1166 write_lock_irq(&tasklist_lock);
1167 /* We dropped tasklist, ptracer could die and untrace */
1170 /* If parent wants a zombie, don't release it now */
1171 state = EXIT_ZOMBIE;
1172 if (do_notify_parent(p, p->exit_signal))
1174 p->exit_state = state;
1175 write_unlock_irq(&tasklist_lock);
1177 if (state == EXIT_DEAD)
1181 infop = wo->wo_info;
1183 if ((status & 0x7f) == 0) {
1184 infop->cause = CLD_EXITED;
1185 infop->status = status >> 8;
1187 infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1188 infop->status = status & 0x7f;
1197 static int *task_stopped_code(struct task_struct *p, bool ptrace)
1200 if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING))
1201 return &p->exit_code;
1203 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1204 return &p->signal->group_exit_code;
1210 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1212 * @ptrace: is the wait for ptrace
1213 * @p: task to wait for
1215 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1218 * read_lock(&tasklist_lock), which is released if return value is
1219 * non-zero. Also, grabs and releases @p->sighand->siglock.
1222 * 0 if wait condition didn't exist and search for other wait conditions
1223 * should continue. Non-zero return, -errno on failure and @p's pid on
1224 * success, implies that tasklist_lock is released and wait condition
1225 * search should terminate.
1227 static int wait_task_stopped(struct wait_opts *wo,
1228 int ptrace, struct task_struct *p)
1230 struct waitid_info *infop;
1231 int exit_code, *p_code, why;
1232 uid_t uid = 0; /* unneeded, required by compiler */
1236 * Traditionally we see ptrace'd stopped tasks regardless of options.
1238 if (!ptrace && !(wo->wo_flags & WUNTRACED))
1241 if (!task_stopped_code(p, ptrace))
1245 spin_lock_irq(&p->sighand->siglock);
1247 p_code = task_stopped_code(p, ptrace);
1248 if (unlikely(!p_code))
1251 exit_code = *p_code;
1255 if (!unlikely(wo->wo_flags & WNOWAIT))
1258 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1260 spin_unlock_irq(&p->sighand->siglock);
1265 * Now we are pretty sure this task is interesting.
1266 * Make sure it doesn't get reaped out from under us while we
1267 * give up the lock and then examine it below. We don't want to
1268 * keep holding onto the tasklist_lock while we call getrusage and
1269 * possibly take page faults for user memory.
1272 pid = task_pid_vnr(p);
1273 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1274 read_unlock(&tasklist_lock);
1275 sched_annotate_sleep();
1277 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1280 if (likely(!(wo->wo_flags & WNOWAIT)))
1281 wo->wo_stat = (exit_code << 8) | 0x7f;
1283 infop = wo->wo_info;
1286 infop->status = exit_code;
1294 * Handle do_wait work for one task in a live, non-stopped state.
1295 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1296 * the lock and this task is uninteresting. If we return nonzero, we have
1297 * released the lock and the system call should return.
1299 static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1301 struct waitid_info *infop;
1305 if (!unlikely(wo->wo_flags & WCONTINUED))
1308 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1311 spin_lock_irq(&p->sighand->siglock);
1312 /* Re-check with the lock held. */
1313 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1314 spin_unlock_irq(&p->sighand->siglock);
1317 if (!unlikely(wo->wo_flags & WNOWAIT))
1318 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1319 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1320 spin_unlock_irq(&p->sighand->siglock);
1322 pid = task_pid_vnr(p);
1324 read_unlock(&tasklist_lock);
1325 sched_annotate_sleep();
1327 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
1330 infop = wo->wo_info;
1332 wo->wo_stat = 0xffff;
1334 infop->cause = CLD_CONTINUED;
1337 infop->status = SIGCONT;
1343 * Consider @p for a wait by @parent.
1345 * -ECHILD should be in ->notask_error before the first call.
1346 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1347 * Returns zero if the search for a child should continue;
1348 * then ->notask_error is 0 if @p is an eligible child,
1351 static int wait_consider_task(struct wait_opts *wo, int ptrace,
1352 struct task_struct *p)
1355 * We can race with wait_task_zombie() from another thread.
1356 * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition
1357 * can't confuse the checks below.
1359 int exit_state = READ_ONCE(p->exit_state);
1362 if (unlikely(exit_state == EXIT_DEAD))
1365 ret = eligible_child(wo, ptrace, p);
1369 if (unlikely(exit_state == EXIT_TRACE)) {
1371 * ptrace == 0 means we are the natural parent. In this case
1372 * we should clear notask_error, debugger will notify us.
1374 if (likely(!ptrace))
1375 wo->notask_error = 0;
1379 if (likely(!ptrace) && unlikely(p->ptrace)) {
1381 * If it is traced by its real parent's group, just pretend
1382 * the caller is ptrace_do_wait() and reap this child if it
1385 * This also hides group stop state from real parent; otherwise
1386 * a single stop can be reported twice as group and ptrace stop.
1387 * If a ptracer wants to distinguish these two events for its
1388 * own children it should create a separate process which takes
1389 * the role of real parent.
1391 if (!ptrace_reparented(p))
1396 if (exit_state == EXIT_ZOMBIE) {
1397 /* we don't reap group leaders with subthreads */
1398 if (!delay_group_leader(p)) {
1400 * A zombie ptracee is only visible to its ptracer.
1401 * Notification and reaping will be cascaded to the
1402 * real parent when the ptracer detaches.
1404 if (unlikely(ptrace) || likely(!p->ptrace))
1405 return wait_task_zombie(wo, p);
1409 * Allow access to stopped/continued state via zombie by
1410 * falling through. Clearing of notask_error is complex.
1414 * If WEXITED is set, notask_error should naturally be
1415 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1416 * so, if there are live subthreads, there are events to
1417 * wait for. If all subthreads are dead, it's still safe
1418 * to clear - this function will be called again in finite
1419 * amount time once all the subthreads are released and
1420 * will then return without clearing.
1424 * Stopped state is per-task and thus can't change once the
1425 * target task dies. Only continued and exited can happen.
1426 * Clear notask_error if WCONTINUED | WEXITED.
1428 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1429 wo->notask_error = 0;
1432 * @p is alive and it's gonna stop, continue or exit, so
1433 * there always is something to wait for.
1435 wo->notask_error = 0;
1439 * Wait for stopped. Depending on @ptrace, different stopped state
1440 * is used and the two don't interact with each other.
1442 ret = wait_task_stopped(wo, ptrace, p);
1447 * Wait for continued. There's only one continued state and the
1448 * ptracer can consume it which can confuse the real parent. Don't
1449 * use WCONTINUED from ptracer. You don't need or want it.
1451 return wait_task_continued(wo, p);
1455 * Do the work of do_wait() for one thread in the group, @tsk.
1457 * -ECHILD should be in ->notask_error before the first call.
1458 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1459 * Returns zero if the search for a child should continue; then
1460 * ->notask_error is 0 if there were any eligible children,
1463 static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1465 struct task_struct *p;
1467 list_for_each_entry(p, &tsk->children, sibling) {
1468 int ret = wait_consider_task(wo, 0, p);
1477 static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1479 struct task_struct *p;
1481 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1482 int ret = wait_consider_task(wo, 1, p);
1491 static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode,
1492 int sync, void *key)
1494 struct wait_opts *wo = container_of(wait, struct wait_opts,
1496 struct task_struct *p = key;
1498 if (!eligible_pid(wo, p))
1501 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1504 return default_wake_function(wait, mode, sync, key);
1507 void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1509 __wake_up_sync_key(&parent->signal->wait_chldexit,
1510 TASK_INTERRUPTIBLE, p);
1513 static bool is_effectively_child(struct wait_opts *wo, bool ptrace,
1514 struct task_struct *target)
1516 struct task_struct *parent =
1517 !ptrace ? target->real_parent : target->parent;
1519 return current == parent || (!(wo->wo_flags & __WNOTHREAD) &&
1520 same_thread_group(current, parent));
1524 * Optimization for waiting on PIDTYPE_PID. No need to iterate through child
1525 * and tracee lists to find the target task.
1527 static int do_wait_pid(struct wait_opts *wo)
1530 struct task_struct *target;
1534 target = pid_task(wo->wo_pid, PIDTYPE_TGID);
1535 if (target && is_effectively_child(wo, ptrace, target)) {
1536 retval = wait_consider_task(wo, ptrace, target);
1542 target = pid_task(wo->wo_pid, PIDTYPE_PID);
1543 if (target && target->ptrace &&
1544 is_effectively_child(wo, ptrace, target)) {
1545 retval = wait_consider_task(wo, ptrace, target);
1553 static long do_wait(struct wait_opts *wo)
1557 trace_sched_process_wait(wo->wo_pid);
1559 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1560 wo->child_wait.private = current;
1561 add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1564 * If there is nothing that can match our criteria, just get out.
1565 * We will clear ->notask_error to zero if we see any child that
1566 * might later match our criteria, even if we are not able to reap
1569 wo->notask_error = -ECHILD;
1570 if ((wo->wo_type < PIDTYPE_MAX) &&
1571 (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type)))
1574 set_current_state(TASK_INTERRUPTIBLE);
1575 read_lock(&tasklist_lock);
1577 if (wo->wo_type == PIDTYPE_PID) {
1578 retval = do_wait_pid(wo);
1582 struct task_struct *tsk = current;
1585 retval = do_wait_thread(wo, tsk);
1589 retval = ptrace_do_wait(wo, tsk);
1593 if (wo->wo_flags & __WNOTHREAD)
1595 } while_each_thread(current, tsk);
1597 read_unlock(&tasklist_lock);
1600 retval = wo->notask_error;
1601 if (!retval && !(wo->wo_flags & WNOHANG)) {
1602 retval = -ERESTARTSYS;
1603 if (!signal_pending(current)) {
1609 __set_current_state(TASK_RUNNING);
1610 remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait);
1614 static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop,
1615 int options, struct rusage *ru)
1617 struct wait_opts wo;
1618 struct pid *pid = NULL;
1621 unsigned int f_flags = 0;
1623 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED|
1624 __WNOTHREAD|__WCLONE|__WALL))
1626 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1638 pid = find_get_pid(upid);
1641 type = PIDTYPE_PGID;
1646 pid = find_get_pid(upid);
1648 pid = get_task_pid(current, PIDTYPE_PGID);
1655 pid = pidfd_get_pid(upid, &f_flags);
1657 return PTR_ERR(pid);
1666 wo.wo_flags = options;
1669 if (f_flags & O_NONBLOCK)
1670 wo.wo_flags |= WNOHANG;
1673 if (!ret && !(options & WNOHANG) && (f_flags & O_NONBLOCK))
1680 SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1681 infop, int, options, struct rusage __user *, ru)
1684 struct waitid_info info = {.status = 0};
1685 long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL);
1691 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
1697 if (!user_write_access_begin(infop, sizeof(*infop)))
1700 unsafe_put_user(signo, &infop->si_signo, Efault);
1701 unsafe_put_user(0, &infop->si_errno, Efault);
1702 unsafe_put_user(info.cause, &infop->si_code, Efault);
1703 unsafe_put_user(info.pid, &infop->si_pid, Efault);
1704 unsafe_put_user(info.uid, &infop->si_uid, Efault);
1705 unsafe_put_user(info.status, &infop->si_status, Efault);
1706 user_write_access_end();
1709 user_write_access_end();
1713 long kernel_wait4(pid_t upid, int __user *stat_addr, int options,
1716 struct wait_opts wo;
1717 struct pid *pid = NULL;
1721 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1722 __WNOTHREAD|__WCLONE|__WALL))
1725 /* -INT_MIN is not defined */
1726 if (upid == INT_MIN)
1731 else if (upid < 0) {
1732 type = PIDTYPE_PGID;
1733 pid = find_get_pid(-upid);
1734 } else if (upid == 0) {
1735 type = PIDTYPE_PGID;
1736 pid = get_task_pid(current, PIDTYPE_PGID);
1737 } else /* upid > 0 */ {
1739 pid = find_get_pid(upid);
1744 wo.wo_flags = options | WEXITED;
1750 if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr))
1756 int kernel_wait(pid_t pid, int *stat)
1758 struct wait_opts wo = {
1759 .wo_type = PIDTYPE_PID,
1760 .wo_pid = find_get_pid(pid),
1761 .wo_flags = WEXITED,
1766 if (ret > 0 && wo.wo_stat)
1772 SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1773 int, options, struct rusage __user *, ru)
1776 long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL);
1779 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
1785 #ifdef __ARCH_WANT_SYS_WAITPID
1788 * sys_waitpid() remains for compatibility. waitpid() should be
1789 * implemented by calling sys_wait4() from libc.a.
1791 SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1793 return kernel_wait4(pid, stat_addr, options, NULL);
1798 #ifdef CONFIG_COMPAT
1799 COMPAT_SYSCALL_DEFINE4(wait4,
1801 compat_uint_t __user *, stat_addr,
1803 struct compat_rusage __user *, ru)
1806 long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL);
1808 if (ru && put_compat_rusage(&r, ru))
1814 COMPAT_SYSCALL_DEFINE5(waitid,
1815 int, which, compat_pid_t, pid,
1816 struct compat_siginfo __user *, infop, int, options,
1817 struct compat_rusage __user *, uru)
1820 struct waitid_info info = {.status = 0};
1821 long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL);
1827 /* kernel_waitid() overwrites everything in ru */
1828 if (COMPAT_USE_64BIT_TIME)
1829 err = copy_to_user(uru, &ru, sizeof(ru));
1831 err = put_compat_rusage(&ru, uru);
1840 if (!user_write_access_begin(infop, sizeof(*infop)))
1843 unsafe_put_user(signo, &infop->si_signo, Efault);
1844 unsafe_put_user(0, &infop->si_errno, Efault);
1845 unsafe_put_user(info.cause, &infop->si_code, Efault);
1846 unsafe_put_user(info.pid, &infop->si_pid, Efault);
1847 unsafe_put_user(info.uid, &infop->si_uid, Efault);
1848 unsafe_put_user(info.status, &infop->si_status, Efault);
1849 user_write_access_end();
1852 user_write_access_end();
1858 * thread_group_exited - check that a thread group has exited
1859 * @pid: tgid of thread group to be checked.
1861 * Test if the thread group represented by tgid has exited (all
1862 * threads are zombies, dead or completely gone).
1864 * Return: true if the thread group has exited. false otherwise.
1866 bool thread_group_exited(struct pid *pid)
1868 struct task_struct *task;
1872 task = pid_task(pid, PIDTYPE_PID);
1874 (READ_ONCE(task->exit_state) && thread_group_empty(task));
1879 EXPORT_SYMBOL(thread_group_exited);
1881 __weak void abort(void)
1885 /* if that doesn't kill us, halt */
1886 panic("Oops failed to kill thread");
1888 EXPORT_SYMBOL(abort);