LCOV - code coverage report
Current view: top level - kernel - exit.c (source / functions) Hit Total Coverage
Test: coverage.info Lines: 205 671 30.6 %
Date: 2023-04-06 08:38:28 Functions: 18 49 36.7 %

          Line data    Source code
       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/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>
      71             : 
      72             : #include <linux/uaccess.h>
      73             : #include <asm/unistd.h>
      74             : #include <asm/mmu_context.h>
      75             : 
      76             : /*
      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.
      80             :  */
      81             : static unsigned int oops_limit = 10000;
      82             : 
      83             : #ifdef CONFIG_SYSCTL
      84             : static struct ctl_table kern_exit_table[] = {
      85             :         {
      86             :                 .procname       = "oops_limit",
      87             :                 .data           = &oops_limit,
      88             :                 .maxlen         = sizeof(oops_limit),
      89             :                 .mode           = 0644,
      90             :                 .proc_handler   = proc_douintvec,
      91             :         },
      92             :         { }
      93             : };
      94             : 
      95           1 : static __init int kernel_exit_sysctls_init(void)
      96             : {
      97           1 :         register_sysctl_init("kernel", kern_exit_table);
      98           1 :         return 0;
      99             : }
     100             : late_initcall(kernel_exit_sysctls_init);
     101             : #endif
     102             : 
     103             : static atomic_t oops_count = ATOMIC_INIT(0);
     104             : 
     105             : #ifdef CONFIG_SYSFS
     106           0 : static ssize_t oops_count_show(struct kobject *kobj, struct kobj_attribute *attr,
     107             :                                char *page)
     108             : {
     109           0 :         return sysfs_emit(page, "%d\n", atomic_read(&oops_count));
     110             : }
     111             : 
     112             : static struct kobj_attribute oops_count_attr = __ATTR_RO(oops_count);
     113             : 
     114           1 : static __init int kernel_exit_sysfs_init(void)
     115             : {
     116           1 :         sysfs_add_file_to_group(kernel_kobj, &oops_count_attr.attr, NULL);
     117           1 :         return 0;
     118             : }
     119             : late_initcall(kernel_exit_sysfs_init);
     120             : #endif
     121             : 
     122         333 : static void __unhash_process(struct task_struct *p, bool group_dead)
     123             : {
     124         333 :         nr_threads--;
     125         333 :         detach_pid(p, PIDTYPE_PID);
     126         333 :         if (group_dead) {
     127         333 :                 detach_pid(p, PIDTYPE_TGID);
     128         333 :                 detach_pid(p, PIDTYPE_PGID);
     129         333 :                 detach_pid(p, PIDTYPE_SID);
     130             : 
     131         666 :                 list_del_rcu(&p->tasks);
     132         666 :                 list_del_init(&p->sibling);
     133         333 :                 __this_cpu_dec(process_counts);
     134             :         }
     135         666 :         list_del_rcu(&p->thread_group);
     136         666 :         list_del_rcu(&p->thread_node);
     137         333 : }
     138             : 
     139             : /*
     140             :  * This function expects the tasklist_lock write-locked.
     141             :  */
     142         333 : static void __exit_signal(struct task_struct *tsk)
     143             : {
     144         333 :         struct signal_struct *sig = tsk->signal;
     145         666 :         bool group_dead = thread_group_leader(tsk);
     146             :         struct sighand_struct *sighand;
     147             :         struct tty_struct *tty;
     148             :         u64 utime, stime;
     149             : 
     150         333 :         sighand = rcu_dereference_check(tsk->sighand,
     151             :                                         lockdep_tasklist_lock_is_held());
     152         666 :         spin_lock(&sighand->siglock);
     153             : 
     154             : #ifdef CONFIG_POSIX_TIMERS
     155         333 :         posix_cpu_timers_exit(tsk);
     156         333 :         if (group_dead)
     157         333 :                 posix_cpu_timers_exit_group(tsk);
     158             : #endif
     159             : 
     160         333 :         if (group_dead) {
     161         333 :                 tty = sig->tty;
     162         333 :                 sig->tty = NULL;
     163             :         } else {
     164             :                 /*
     165             :                  * If there is any task waiting for the group exit
     166             :                  * then notify it:
     167             :                  */
     168           0 :                 if (sig->notify_count > 0 && !--sig->notify_count)
     169           0 :                         wake_up_process(sig->group_exec_task);
     170             : 
     171           0 :                 if (tsk == sig->curr_target)
     172           0 :                         sig->curr_target = next_thread(tsk);
     173             :         }
     174             : 
     175         333 :         add_device_randomness((const void*) &tsk->se.sum_exec_runtime,
     176             :                               sizeof(unsigned long long));
     177             : 
     178             :         /*
     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.
     183             :          */
     184         666 :         task_cputime(tsk, &utime, &stime);
     185         666 :         write_seqlock(&sig->stats_lock);
     186         333 :         sig->utime += utime;
     187         333 :         sig->stime += stime;
     188         333 :         sig->gtime += task_gtime(tsk);
     189         333 :         sig->min_flt += tsk->min_flt;
     190         333 :         sig->maj_flt += tsk->maj_flt;
     191         333 :         sig->nvcsw += tsk->nvcsw;
     192         333 :         sig->nivcsw += tsk->nivcsw;
     193         333 :         sig->inblock += task_io_get_inblock(tsk);
     194         333 :         sig->oublock += task_io_get_oublock(tsk);
     195         333 :         task_io_accounting_add(&sig->ioac, &tsk->ioac);
     196         333 :         sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
     197         333 :         sig->nr_threads--;
     198         333 :         __unhash_process(tsk, group_dead);
     199         666 :         write_sequnlock(&sig->stats_lock);
     200             : 
     201             :         /*
     202             :          * Do this under ->siglock, we can race with another thread
     203             :          * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
     204             :          */
     205         333 :         flush_sigqueue(&tsk->pending);
     206         333 :         tsk->sighand = NULL;
     207         666 :         spin_unlock(&sighand->siglock);
     208             : 
     209         333 :         __cleanup_sighand(sighand);
     210         666 :         clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
     211         333 :         if (group_dead) {
     212         333 :                 flush_sigqueue(&sig->shared_pending);
     213         333 :                 tty_kref_put(tty);
     214             :         }
     215         333 : }
     216             : 
     217         332 : static void delayed_put_task_struct(struct rcu_head *rhp)
     218             : {
     219         332 :         struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
     220             : 
     221         332 :         kprobe_flush_task(tsk);
     222             :         rethook_flush_task(tsk);
     223         332 :         perf_event_delayed_put(tsk);
     224         332 :         trace_sched_process_free(tsk);
     225         332 :         put_task_struct(tsk);
     226         332 : }
     227             : 
     228         666 : void put_task_struct_rcu_user(struct task_struct *task)
     229             : {
     230        1332 :         if (refcount_dec_and_test(&task->rcu_users))
     231         333 :                 call_rcu(&task->rcu, delayed_put_task_struct);
     232         666 : }
     233             : 
     234         333 : void __weak release_thread(struct task_struct *dead_task)
     235             : {
     236         333 : }
     237             : 
     238         333 : void release_task(struct task_struct *p)
     239             : {
     240             :         struct task_struct *leader;
     241             :         struct pid *thread_pid;
     242             :         int zap_leader;
     243             : repeat:
     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 */
     246             :         rcu_read_lock();
     247         666 :         dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1);
     248             :         rcu_read_unlock();
     249             : 
     250         333 :         cgroup_release(p);
     251             : 
     252         333 :         write_lock_irq(&tasklist_lock);
     253         333 :         ptrace_release_task(p);
     254         666 :         thread_pid = get_pid(p->thread_pid);
     255         333 :         __exit_signal(p);
     256             : 
     257             :         /*
     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.)
     261             :          */
     262         333 :         zap_leader = 0;
     263         333 :         leader = p->group_leader;
     264         333 :         if (leader != p && thread_group_empty(leader)
     265           0 :                         && leader->exit_state == EXIT_ZOMBIE) {
     266             :                 /*
     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.
     270             :                  */
     271           0 :                 zap_leader = do_notify_parent(leader, leader->exit_signal);
     272           0 :                 if (zap_leader)
     273           0 :                         leader->exit_state = EXIT_DEAD;
     274             :         }
     275             : 
     276         333 :         write_unlock_irq(&tasklist_lock);
     277         333 :         seccomp_filter_release(p);
     278         333 :         proc_flush_pid(thread_pid);
     279         333 :         put_pid(thread_pid);
     280         333 :         release_thread(p);
     281         333 :         put_task_struct_rcu_user(p);
     282             : 
     283         333 :         p = leader;
     284         333 :         if (unlikely(zap_leader))
     285             :                 goto repeat;
     286         333 : }
     287             : 
     288           2 : int rcuwait_wake_up(struct rcuwait *w)
     289             : {
     290           2 :         int ret = 0;
     291             :         struct task_struct *task;
     292             : 
     293             :         rcu_read_lock();
     294             : 
     295             :         /*
     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().
     300             :          *
     301             :          *    WAIT                WAKE
     302             :          *    [S] tsk = current   [S] cond = true
     303             :          *        MB (A)              MB (B)
     304             :          *    [L] cond            [L] tsk
     305             :          */
     306           2 :         smp_mb(); /* (B) */
     307             : 
     308           2 :         task = rcu_dereference(w->task);
     309           2 :         if (task)
     310           0 :                 ret = wake_up_process(task);
     311             :         rcu_read_unlock();
     312             : 
     313           2 :         return ret;
     314             : }
     315             : EXPORT_SYMBOL_GPL(rcuwait_wake_up);
     316             : 
     317             : /*
     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.
     322             :  *
     323             :  * "I ask you, have you ever known what it is to be an orphan?"
     324             :  */
     325           0 : static int will_become_orphaned_pgrp(struct pid *pgrp,
     326             :                                         struct task_struct *ignored_task)
     327             : {
     328             :         struct task_struct *p;
     329             : 
     330           0 :         do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
     331           0 :                 if ((p == ignored_task) ||
     332           0 :                     (p->exit_state && thread_group_empty(p)) ||
     333           0 :                     is_global_init(p->real_parent))
     334           0 :                         continue;
     335             : 
     336           0 :                 if (task_pgrp(p->real_parent) != pgrp &&
     337           0 :                     task_session(p->real_parent) == task_session(p))
     338             :                         return 0;
     339             :         } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
     340             : 
     341             :         return 1;
     342             : }
     343             : 
     344           0 : int is_current_pgrp_orphaned(void)
     345             : {
     346             :         int retval;
     347             : 
     348           0 :         read_lock(&tasklist_lock);
     349           0 :         retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
     350           0 :         read_unlock(&tasklist_lock);
     351             : 
     352           0 :         return retval;
     353             : }
     354             : 
     355             : static bool has_stopped_jobs(struct pid *pgrp)
     356             : {
     357             :         struct task_struct *p;
     358             : 
     359           0 :         do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
     360           0 :                 if (p->signal->flags & SIGNAL_STOP_STOPPED)
     361             :                         return true;
     362             :         } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
     363             : 
     364             :         return false;
     365             : }
     366             : 
     367             : /*
     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)
     371             :  */
     372             : static void
     373         333 : kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
     374             : {
     375         666 :         struct pid *pgrp = task_pgrp(tsk);
     376         333 :         struct task_struct *ignored_task = tsk;
     377             : 
     378         333 :         if (!parent)
     379             :                 /* exit: our father is in a different pgrp than
     380             :                  * we are and we were the only connection outside.
     381             :                  */
     382         333 :                 parent = tsk->real_parent;
     383             :         else
     384             :                 /* reparent: our child is in a different pgrp than
     385             :                  * we are, and it was the only connection outside.
     386             :                  */
     387             :                 ignored_task = NULL;
     388             : 
     389         666 :         if (task_pgrp(parent) != pgrp &&
     390           0 :             task_session(parent) == task_session(tsk) &&
     391           0 :             will_become_orphaned_pgrp(pgrp, ignored_task) &&
     392           0 :             has_stopped_jobs(pgrp)) {
     393           0 :                 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
     394           0 :                 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
     395             :         }
     396         333 : }
     397             : 
     398         333 : static void coredump_task_exit(struct task_struct *tsk)
     399             : {
     400             :         struct core_state *core_state;
     401             : 
     402             :         /*
     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.
     408             :          */
     409         666 :         spin_lock_irq(&tsk->sighand->siglock);
     410         333 :         tsk->flags |= PF_POSTCOREDUMP;
     411         333 :         core_state = tsk->signal->core_state;
     412         666 :         spin_unlock_irq(&tsk->sighand->siglock);
     413         333 :         if (core_state) {
     414             :                 struct core_thread self;
     415             : 
     416           0 :                 self.task = current;
     417           0 :                 if (self.task->flags & PF_SIGNALED)
     418           0 :                         self.next = xchg(&core_state->dumper.next, &self);
     419             :                 else
     420           0 :                         self.task = NULL;
     421             :                 /*
     422             :                  * Implies mb(), the result of xchg() must be visible
     423             :                  * to core_state->dumper.
     424             :                  */
     425           0 :                 if (atomic_dec_and_test(&core_state->nr_threads))
     426           0 :                         complete(&core_state->startup);
     427             : 
     428             :                 for (;;) {
     429           0 :                         set_current_state(TASK_UNINTERRUPTIBLE|TASK_FREEZABLE);
     430           0 :                         if (!self.task) /* see coredump_finish() */
     431             :                                 break;
     432           0 :                         schedule();
     433             :                 }
     434           0 :                 __set_current_state(TASK_RUNNING);
     435             :         }
     436         333 : }
     437             : 
     438             : #ifdef CONFIG_MEMCG
     439             : /*
     440             :  * A task is exiting.   If it owned this mm, find a new owner for the mm.
     441             :  */
     442             : void mm_update_next_owner(struct mm_struct *mm)
     443             : {
     444             :         struct task_struct *c, *g, *p = current;
     445             : 
     446             : retry:
     447             :         /*
     448             :          * If the exiting or execing task is not the owner, it's
     449             :          * someone else's problem.
     450             :          */
     451             :         if (mm->owner != p)
     452             :                 return;
     453             :         /*
     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.
     457             :          */
     458             :         if (atomic_read(&mm->mm_users) <= 1) {
     459             :                 WRITE_ONCE(mm->owner, NULL);
     460             :                 return;
     461             :         }
     462             : 
     463             :         read_lock(&tasklist_lock);
     464             :         /*
     465             :          * Search in the children
     466             :          */
     467             :         list_for_each_entry(c, &p->children, sibling) {
     468             :                 if (c->mm == mm)
     469             :                         goto assign_new_owner;
     470             :         }
     471             : 
     472             :         /*
     473             :          * Search in the siblings
     474             :          */
     475             :         list_for_each_entry(c, &p->real_parent->children, sibling) {
     476             :                 if (c->mm == mm)
     477             :                         goto assign_new_owner;
     478             :         }
     479             : 
     480             :         /*
     481             :          * Search through everything else, we should not get here often.
     482             :          */
     483             :         for_each_process(g) {
     484             :                 if (g->flags & PF_KTHREAD)
     485             :                         continue;
     486             :                 for_each_thread(g, c) {
     487             :                         if (c->mm == mm)
     488             :                                 goto assign_new_owner;
     489             :                         if (c->mm)
     490             :                                 break;
     491             :                 }
     492             :         }
     493             :         read_unlock(&tasklist_lock);
     494             :         /*
     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.
     498             :          */
     499             :         WRITE_ONCE(mm->owner, NULL);
     500             :         return;
     501             : 
     502             : assign_new_owner:
     503             :         BUG_ON(c == p);
     504             :         get_task_struct(c);
     505             :         /*
     506             :          * The task_lock protects c->mm from changing.
     507             :          * We always want mm->owner->mm == mm
     508             :          */
     509             :         task_lock(c);
     510             :         /*
     511             :          * Delay read_unlock() till we have the task_lock()
     512             :          * to ensure that c does not slip away underneath us
     513             :          */
     514             :         read_unlock(&tasklist_lock);
     515             :         if (c->mm != mm) {
     516             :                 task_unlock(c);
     517             :                 put_task_struct(c);
     518             :                 goto retry;
     519             :         }
     520             :         WRITE_ONCE(mm->owner, c);
     521             :         lru_gen_migrate_mm(mm);
     522             :         task_unlock(c);
     523             :         put_task_struct(c);
     524             : }
     525             : #endif /* CONFIG_MEMCG */
     526             : 
     527             : /*
     528             :  * Turn us into a lazy TLB process if we
     529             :  * aren't already..
     530             :  */
     531         333 : static void exit_mm(void)
     532             : {
     533         333 :         struct mm_struct *mm = current->mm;
     534             : 
     535         333 :         exit_mm_release(current, mm);
     536         333 :         if (!mm)
     537             :                 return;
     538           0 :         sync_mm_rss(mm);
     539           0 :         mmap_read_lock(mm);
     540           0 :         mmgrab(mm);
     541           0 :         BUG_ON(mm != current->active_mm);
     542             :         /* more a memory barrier than a real lock */
     543           0 :         task_lock(current);
     544             :         /*
     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.
     553             :          */
     554             :         smp_mb__after_spinlock();
     555             :         local_irq_disable();
     556           0 :         current->mm = NULL;
     557           0 :         membarrier_update_current_mm(NULL);
     558           0 :         enter_lazy_tlb(mm, current);
     559             :         local_irq_enable();
     560           0 :         task_unlock(current);
     561           0 :         mmap_read_unlock(mm);
     562           0 :         mm_update_next_owner(mm);
     563           0 :         mmput(mm);
     564           0 :         if (test_thread_flag(TIF_MEMDIE))
     565           0 :                 exit_oom_victim();
     566             : }
     567             : 
     568             : static struct task_struct *find_alive_thread(struct task_struct *p)
     569             : {
     570             :         struct task_struct *t;
     571             : 
     572           0 :         for_each_thread(p, t) {
     573           0 :                 if (!(t->flags & PF_EXITING))
     574             :                         return t;
     575             :         }
     576             :         return NULL;
     577             : }
     578             : 
     579         333 : static struct task_struct *find_child_reaper(struct task_struct *father,
     580             :                                                 struct list_head *dead)
     581             :         __releases(&tasklist_lock)
     582             :         __acquires(&tasklist_lock)
     583             : {
     584         333 :         struct pid_namespace *pid_ns = task_active_pid_ns(father);
     585         333 :         struct task_struct *reaper = pid_ns->child_reaper;
     586             :         struct task_struct *p, *n;
     587             : 
     588         333 :         if (likely(reaper != father))
     589             :                 return reaper;
     590             : 
     591           0 :         reaper = find_alive_thread(father);
     592           0 :         if (reaper) {
     593           0 :                 pid_ns->child_reaper = reaper;
     594           0 :                 return reaper;
     595             :         }
     596             : 
     597           0 :         write_unlock_irq(&tasklist_lock);
     598             : 
     599           0 :         list_for_each_entry_safe(p, n, dead, ptrace_entry) {
     600           0 :                 list_del_init(&p->ptrace_entry);
     601           0 :                 release_task(p);
     602             :         }
     603             : 
     604           0 :         zap_pid_ns_processes(pid_ns);
     605           0 :         write_lock_irq(&tasklist_lock);
     606             : 
     607           0 :         return father;
     608             : }
     609             : 
     610             : /*
     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
     616             :  */
     617           0 : static struct task_struct *find_new_reaper(struct task_struct *father,
     618             :                                            struct task_struct *child_reaper)
     619             : {
     620             :         struct task_struct *thread, *reaper;
     621             : 
     622           0 :         thread = find_alive_thread(father);
     623           0 :         if (thread)
     624             :                 return thread;
     625             : 
     626           0 :         if (father->signal->has_child_subreaper) {
     627           0 :                 unsigned int ns_level = task_pid(father)->level;
     628             :                 /*
     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).
     635             :                  */
     636           0 :                 for (reaper = father->real_parent;
     637           0 :                      task_pid(reaper)->level == ns_level;
     638           0 :                      reaper = reaper->real_parent) {
     639           0 :                         if (reaper == &init_task)
     640             :                                 break;
     641           0 :                         if (!reaper->signal->is_child_subreaper)
     642           0 :                                 continue;
     643           0 :                         thread = find_alive_thread(reaper);
     644           0 :                         if (thread)
     645             :                                 return thread;
     646             :                 }
     647             :         }
     648             : 
     649             :         return child_reaper;
     650             : }
     651             : 
     652             : /*
     653             : * Any that need to be release_task'd are put on the @dead list.
     654             :  */
     655           0 : static void reparent_leader(struct task_struct *father, struct task_struct *p,
     656             :                                 struct list_head *dead)
     657             : {
     658           0 :         if (unlikely(p->exit_state == EXIT_DEAD))
     659             :                 return;
     660             : 
     661             :         /* We don't want people slaying init. */
     662           0 :         p->exit_signal = SIGCHLD;
     663             : 
     664             :         /* If it has exited notify the new parent about this child's death. */
     665           0 :         if (!p->ptrace &&
     666           0 :             p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
     667           0 :                 if (do_notify_parent(p, p->exit_signal)) {
     668           0 :                         p->exit_state = EXIT_DEAD;
     669           0 :                         list_add(&p->ptrace_entry, dead);
     670             :                 }
     671             :         }
     672             : 
     673           0 :         kill_orphaned_pgrp(p, father);
     674             : }
     675             : 
     676             : /*
     677             :  * This does two things:
     678             :  *
     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)
     683             :  */
     684         333 : static void forget_original_parent(struct task_struct *father,
     685             :                                         struct list_head *dead)
     686             : {
     687             :         struct task_struct *p, *t, *reaper;
     688             : 
     689         666 :         if (unlikely(!list_empty(&father->ptraced)))
     690           0 :                 exit_ptrace(father, dead);
     691             : 
     692             :         /* Can drop and reacquire tasklist_lock */
     693         333 :         reaper = find_child_reaper(father, dead);
     694         666 :         if (list_empty(&father->children))
     695             :                 return;
     696             : 
     697           0 :         reaper = find_new_reaper(father, reaper);
     698           0 :         list_for_each_entry(p, &father->children, sibling) {
     699           0 :                 for_each_thread(p, t) {
     700           0 :                         RCU_INIT_POINTER(t->real_parent, reaper);
     701           0 :                         BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father));
     702           0 :                         if (likely(!t->ptrace))
     703           0 :                                 t->parent = t->real_parent;
     704           0 :                         if (t->pdeath_signal)
     705           0 :                                 group_send_sig_info(t->pdeath_signal,
     706             :                                                     SEND_SIG_NOINFO, t,
     707             :                                                     PIDTYPE_TGID);
     708             :                 }
     709             :                 /*
     710             :                  * If this is a threaded reparent there is no need to
     711             :                  * notify anyone anything has happened.
     712             :                  */
     713           0 :                 if (!same_thread_group(reaper, father))
     714           0 :                         reparent_leader(father, p, dead);
     715             :         }
     716           0 :         list_splice_tail_init(&father->children, &reaper->children);
     717             : }
     718             : 
     719             : /*
     720             :  * Send signals to all our closest relatives so that they know
     721             :  * to properly mourn us..
     722             :  */
     723         333 : static void exit_notify(struct task_struct *tsk, int group_dead)
     724             : {
     725             :         bool autoreap;
     726             :         struct task_struct *p, *n;
     727         333 :         LIST_HEAD(dead);
     728             : 
     729         333 :         write_lock_irq(&tasklist_lock);
     730         333 :         forget_original_parent(tsk, &dead);
     731             : 
     732         333 :         if (group_dead)
     733         333 :                 kill_orphaned_pgrp(tsk->group_leader, NULL);
     734             : 
     735         333 :         tsk->exit_state = EXIT_ZOMBIE;
     736         333 :         if (unlikely(tsk->ptrace)) {
     737           0 :                 int sig = thread_group_leader(tsk) &&
     738           0 :                                 thread_group_empty(tsk) &&
     739           0 :                                 !ptrace_reparented(tsk) ?
     740           0 :                         tsk->exit_signal : SIGCHLD;
     741           0 :                 autoreap = do_notify_parent(tsk, sig);
     742         333 :         } else if (thread_group_leader(tsk)) {
     743         666 :                 autoreap = thread_group_empty(tsk) &&
     744         333 :                         do_notify_parent(tsk, tsk->exit_signal);
     745             :         } else {
     746             :                 autoreap = true;
     747             :         }
     748             : 
     749         333 :         if (autoreap) {
     750         333 :                 tsk->exit_state = EXIT_DEAD;
     751         333 :                 list_add(&tsk->ptrace_entry, &dead);
     752             :         }
     753             : 
     754             :         /* mt-exec, de_thread() is waiting for group leader */
     755         333 :         if (unlikely(tsk->signal->notify_count < 0))
     756           0 :                 wake_up_process(tsk->signal->group_exec_task);
     757         333 :         write_unlock_irq(&tasklist_lock);
     758             : 
     759         666 :         list_for_each_entry_safe(p, n, &dead, ptrace_entry) {
     760         666 :                 list_del_init(&p->ptrace_entry);
     761         333 :                 release_task(p);
     762             :         }
     763         333 : }
     764             : 
     765             : #ifdef CONFIG_DEBUG_STACK_USAGE
     766             : static void check_stack_usage(void)
     767             : {
     768             :         static DEFINE_SPINLOCK(low_water_lock);
     769             :         static int lowest_to_date = THREAD_SIZE;
     770             :         unsigned long free;
     771             : 
     772             :         free = stack_not_used(current);
     773             : 
     774             :         if (free >= lowest_to_date)
     775             :                 return;
     776             : 
     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;
     782             :         }
     783             :         spin_unlock(&low_water_lock);
     784             : }
     785             : #else
     786             : static inline void check_stack_usage(void) {}
     787             : #endif
     788             : 
     789         333 : static void synchronize_group_exit(struct task_struct *tsk, long code)
     790             : {
     791         333 :         struct sighand_struct *sighand = tsk->sighand;
     792         333 :         struct signal_struct *signal = tsk->signal;
     793             : 
     794         666 :         spin_lock_irq(&sighand->siglock);
     795         333 :         signal->quick_threads--;
     796         666 :         if ((signal->quick_threads == 0) &&
     797         333 :             !(signal->flags & SIGNAL_GROUP_EXIT)) {
     798         333 :                 signal->flags = SIGNAL_GROUP_EXIT;
     799         333 :                 signal->group_exit_code = code;
     800         333 :                 signal->group_stop_count = 0;
     801             :         }
     802         666 :         spin_unlock_irq(&sighand->siglock);
     803         333 : }
     804             : 
     805         333 : void __noreturn do_exit(long code)
     806             : {
     807         333 :         struct task_struct *tsk = current;
     808             :         int group_dead;
     809             : 
     810         666 :         WARN_ON(irqs_disabled());
     811             : 
     812         333 :         synchronize_group_exit(tsk, code);
     813             : 
     814         333 :         WARN_ON(tsk->plug);
     815             : 
     816         333 :         kcov_task_exit(tsk);
     817         333 :         kmsan_task_exit(tsk);
     818             : 
     819         333 :         coredump_task_exit(tsk);
     820         333 :         ptrace_event(PTRACE_EVENT_EXIT, code);
     821             : 
     822         333 :         validate_creds_for_do_exit(tsk);
     823             : 
     824         333 :         io_uring_files_cancel();
     825         333 :         exit_signals(tsk);  /* sets PF_EXITING */
     826             : 
     827             :         /* sync mm's RSS info before statistics gathering */
     828             :         if (tsk->mm)
     829             :                 sync_mm_rss(tsk->mm);
     830         333 :         acct_update_integrals(tsk);
     831         666 :         group_dead = atomic_dec_and_test(&tsk->signal->live);
     832         333 :         if (group_dead) {
     833             :                 /*
     834             :                  * If the last thread of global init has exited, panic
     835             :                  * immediately to get a useable coredump.
     836             :                  */
     837         333 :                 if (unlikely(is_global_init(tsk)))
     838           0 :                         panic("Attempted to kill init! exitcode=0x%08x\n",
     839           0 :                                 tsk->signal->group_exit_code ?: (int)code);
     840             : 
     841             : #ifdef CONFIG_POSIX_TIMERS
     842         333 :                 hrtimer_cancel(&tsk->signal->real_timer);
     843         333 :                 exit_itimers(tsk);
     844             : #endif
     845         333 :                 if (tsk->mm)
     846           0 :                         setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
     847             :         }
     848             :         acct_collect(code, group_dead);
     849             :         if (group_dead)
     850             :                 tty_audit_exit();
     851         333 :         audit_free(tsk);
     852             : 
     853         333 :         tsk->exit_code = code;
     854         333 :         taskstats_exit(tsk, group_dead);
     855             : 
     856         333 :         exit_mm();
     857             : 
     858             :         if (group_dead)
     859             :                 acct_process();
     860         333 :         trace_sched_process_exit(tsk);
     861             : 
     862         333 :         exit_sem(tsk);
     863         333 :         exit_shm(tsk);
     864         333 :         exit_files(tsk);
     865         333 :         exit_fs(tsk);
     866         333 :         if (group_dead)
     867         333 :                 disassociate_ctty(1);
     868         333 :         exit_task_namespaces(tsk);
     869         333 :         exit_task_work(tsk);
     870         333 :         exit_thread(tsk);
     871             : 
     872             :         /*
     873             :          * Flush inherited counters to the parent - before the parent
     874             :          * gets woken up by child-exit notifications.
     875             :          *
     876             :          * because of cgroup mode, must be called before cgroup_exit()
     877             :          */
     878         333 :         perf_event_exit_task(tsk);
     879             : 
     880         333 :         sched_autogroup_exit_task(tsk);
     881         333 :         cgroup_exit(tsk);
     882             : 
     883             :         /*
     884             :          * FIXME: do that only when needed, using sched_exit tracepoint
     885             :          */
     886         333 :         flush_ptrace_hw_breakpoint(tsk);
     887             : 
     888             :         exit_tasks_rcu_start();
     889         333 :         exit_notify(tsk, group_dead);
     890         333 :         proc_exit_connector(tsk);
     891         333 :         mpol_put_task_policy(tsk);
     892             : #ifdef CONFIG_FUTEX
     893         333 :         if (unlikely(current->pi_state_cache))
     894           0 :                 kfree(current->pi_state_cache);
     895             : #endif
     896             :         /*
     897             :          * Make sure we are holding no locks:
     898             :          */
     899             :         debug_check_no_locks_held();
     900             : 
     901         333 :         if (tsk->io_context)
     902           0 :                 exit_io_context(tsk);
     903             : 
     904         333 :         if (tsk->splice_pipe)
     905           0 :                 free_pipe_info(tsk->splice_pipe);
     906             : 
     907         333 :         if (tsk->task_frag.page)
     908           0 :                 put_page(tsk->task_frag.page);
     909             : 
     910         333 :         validate_creds_for_do_exit(tsk);
     911         333 :         exit_task_stack_account(tsk);
     912             : 
     913             :         check_stack_usage();
     914         333 :         preempt_disable();
     915         333 :         if (tsk->nr_dirtied)
     916           0 :                 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
     917             :         exit_rcu();
     918             :         exit_tasks_rcu_finish();
     919             : 
     920         333 :         lockdep_free_task(tsk);
     921         333 :         do_task_dead();
     922             : }
     923             : 
     924           0 : void __noreturn make_task_dead(int signr)
     925             : {
     926             :         /*
     927             :          * Take the task off the cpu after something catastrophic has
     928             :          * happened.
     929             :          *
     930             :          * We can get here from a kernel oops, sometimes with preemption off.
     931             :          * Start by checking for critical errors.
     932             :          * Then fix up important state like USER_DS and preemption.
     933             :          * Then do everything else.
     934             :          */
     935           0 :         struct task_struct *tsk = current;
     936             :         unsigned int limit;
     937             : 
     938           0 :         if (unlikely(in_interrupt()))
     939           0 :                 panic("Aiee, killing interrupt handler!");
     940           0 :         if (unlikely(!tsk->pid))
     941           0 :                 panic("Attempted to kill the idle task!");
     942             : 
     943           0 :         if (unlikely(irqs_disabled())) {
     944           0 :                 pr_info("note: %s[%d] exited with irqs disabled\n",
     945             :                         current->comm, task_pid_nr(current));
     946             :                 local_irq_enable();
     947             :         }
     948           0 :         if (unlikely(in_atomic())) {
     949           0 :                 pr_info("note: %s[%d] exited with preempt_count %d\n",
     950             :                         current->comm, task_pid_nr(current),
     951             :                         preempt_count());
     952             :                 preempt_count_set(PREEMPT_ENABLED);
     953             :         }
     954             : 
     955             :         /*
     956             :          * Every time the system oopses, if the oops happens while a reference
     957             :          * to an object was held, the reference leaks.
     958             :          * If the oops doesn't also leak memory, repeated oopsing can cause
     959             :          * reference counters to wrap around (if they're not using refcount_t).
     960             :          * This means that repeated oopsing can make unexploitable-looking bugs
     961             :          * exploitable through repeated oopsing.
     962             :          * To make sure this can't happen, place an upper bound on how often the
     963             :          * kernel may oops without panic().
     964             :          */
     965           0 :         limit = READ_ONCE(oops_limit);
     966           0 :         if (atomic_inc_return(&oops_count) >= limit && limit)
     967           0 :                 panic("Oopsed too often (kernel.oops_limit is %d)", limit);
     968             : 
     969             :         /*
     970             :          * We're taking recursive faults here in make_task_dead. Safest is to just
     971             :          * leave this task alone and wait for reboot.
     972             :          */
     973           0 :         if (unlikely(tsk->flags & PF_EXITING)) {
     974           0 :                 pr_alert("Fixing recursive fault but reboot is needed!\n");
     975           0 :                 futex_exit_recursive(tsk);
     976           0 :                 tsk->exit_state = EXIT_DEAD;
     977           0 :                 refcount_inc(&tsk->rcu_users);
     978           0 :                 do_task_dead();
     979             :         }
     980             : 
     981           0 :         do_exit(signr);
     982             : }
     983             : 
     984           0 : SYSCALL_DEFINE1(exit, int, error_code)
     985             : {
     986           0 :         do_exit((error_code&0xff)<<8);
     987             : }
     988             : 
     989             : /*
     990             :  * Take down every thread in the group.  This is called by fatal signals
     991             :  * as well as by sys_exit_group (below).
     992             :  */
     993             : void __noreturn
     994           0 : do_group_exit(int exit_code)
     995             : {
     996           0 :         struct signal_struct *sig = current->signal;
     997             : 
     998           0 :         if (sig->flags & SIGNAL_GROUP_EXIT)
     999           0 :                 exit_code = sig->group_exit_code;
    1000           0 :         else if (sig->group_exec_task)
    1001             :                 exit_code = 0;
    1002             :         else {
    1003           0 :                 struct sighand_struct *const sighand = current->sighand;
    1004             : 
    1005           0 :                 spin_lock_irq(&sighand->siglock);
    1006           0 :                 if (sig->flags & SIGNAL_GROUP_EXIT)
    1007             :                         /* Another thread got here before we took the lock.  */
    1008           0 :                         exit_code = sig->group_exit_code;
    1009           0 :                 else if (sig->group_exec_task)
    1010             :                         exit_code = 0;
    1011             :                 else {
    1012           0 :                         sig->group_exit_code = exit_code;
    1013           0 :                         sig->flags = SIGNAL_GROUP_EXIT;
    1014           0 :                         zap_other_threads(current);
    1015             :                 }
    1016           0 :                 spin_unlock_irq(&sighand->siglock);
    1017             :         }
    1018             : 
    1019           0 :         do_exit(exit_code);
    1020             :         /* NOTREACHED */
    1021             : }
    1022             : 
    1023             : /*
    1024             :  * this kills every thread in the thread group. Note that any externally
    1025             :  * wait4()-ing process will get the correct exit code - even if this
    1026             :  * thread is not the thread group leader.
    1027             :  */
    1028           0 : SYSCALL_DEFINE1(exit_group, int, error_code)
    1029             : {
    1030           0 :         do_group_exit((error_code & 0xff) << 8);
    1031             :         /* NOTREACHED */
    1032             :         return 0;
    1033             : }
    1034             : 
    1035             : struct waitid_info {
    1036             :         pid_t pid;
    1037             :         uid_t uid;
    1038             :         int status;
    1039             :         int cause;
    1040             : };
    1041             : 
    1042             : struct wait_opts {
    1043             :         enum pid_type           wo_type;
    1044             :         int                     wo_flags;
    1045             :         struct pid              *wo_pid;
    1046             : 
    1047             :         struct waitid_info      *wo_info;
    1048             :         int                     wo_stat;
    1049             :         struct rusage           *wo_rusage;
    1050             : 
    1051             :         wait_queue_entry_t              child_wait;
    1052             :         int                     notask_error;
    1053             : };
    1054             : 
    1055             : static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
    1056             : {
    1057           0 :         return  wo->wo_type == PIDTYPE_MAX ||
    1058           0 :                 task_pid_type(p, wo->wo_type) == wo->wo_pid;
    1059             : }
    1060             : 
    1061             : static int
    1062           0 : eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p)
    1063             : {
    1064           0 :         if (!eligible_pid(wo, p))
    1065             :                 return 0;
    1066             : 
    1067             :         /*
    1068             :          * Wait for all children (clone and not) if __WALL is set or
    1069             :          * if it is traced by us.
    1070             :          */
    1071           0 :         if (ptrace || (wo->wo_flags & __WALL))
    1072             :                 return 1;
    1073             : 
    1074             :         /*
    1075             :          * Otherwise, wait for clone children *only* if __WCLONE is set;
    1076             :          * otherwise, wait for non-clone children *only*.
    1077             :          *
    1078             :          * Note: a "clone" child here is one that reports to its parent
    1079             :          * using a signal other than SIGCHLD, or a non-leader thread which
    1080             :          * we can only see if it is traced by us.
    1081             :          */
    1082           0 :         if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
    1083             :                 return 0;
    1084             : 
    1085           0 :         return 1;
    1086             : }
    1087             : 
    1088             : /*
    1089             :  * Handle sys_wait4 work for one task in state EXIT_ZOMBIE.  We hold
    1090             :  * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
    1091             :  * the lock and this task is uninteresting.  If we return nonzero, we have
    1092             :  * released the lock and the system call should return.
    1093             :  */
    1094           0 : static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
    1095             : {
    1096             :         int state, status;
    1097           0 :         pid_t pid = task_pid_vnr(p);
    1098           0 :         uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p));
    1099             :         struct waitid_info *infop;
    1100             : 
    1101           0 :         if (!likely(wo->wo_flags & WEXITED))
    1102             :                 return 0;
    1103             : 
    1104           0 :         if (unlikely(wo->wo_flags & WNOWAIT)) {
    1105           0 :                 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
    1106           0 :                         ? p->signal->group_exit_code : p->exit_code;
    1107           0 :                 get_task_struct(p);
    1108           0 :                 read_unlock(&tasklist_lock);
    1109             :                 sched_annotate_sleep();
    1110           0 :                 if (wo->wo_rusage)
    1111           0 :                         getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
    1112           0 :                 put_task_struct(p);
    1113           0 :                 goto out_info;
    1114             :         }
    1115             :         /*
    1116             :          * Move the task's state to DEAD/TRACE, only one thread can do this.
    1117             :          */
    1118           0 :         state = (ptrace_reparented(p) && thread_group_leader(p)) ?
    1119           0 :                 EXIT_TRACE : EXIT_DEAD;
    1120           0 :         if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE)
    1121             :                 return 0;
    1122             :         /*
    1123             :          * We own this thread, nobody else can reap it.
    1124             :          */
    1125           0 :         read_unlock(&tasklist_lock);
    1126             :         sched_annotate_sleep();
    1127             : 
    1128             :         /*
    1129             :          * Check thread_group_leader() to exclude the traced sub-threads.
    1130             :          */
    1131           0 :         if (state == EXIT_DEAD && thread_group_leader(p)) {
    1132           0 :                 struct signal_struct *sig = p->signal;
    1133           0 :                 struct signal_struct *psig = current->signal;
    1134             :                 unsigned long maxrss;
    1135             :                 u64 tgutime, tgstime;
    1136             : 
    1137             :                 /*
    1138             :                  * The resource counters for the group leader are in its
    1139             :                  * own task_struct.  Those for dead threads in the group
    1140             :                  * are in its signal_struct, as are those for the child
    1141             :                  * processes it has previously reaped.  All these
    1142             :                  * accumulate in the parent's signal_struct c* fields.
    1143             :                  *
    1144             :                  * We don't bother to take a lock here to protect these
    1145             :                  * p->signal fields because the whole thread group is dead
    1146             :                  * and nobody can change them.
    1147             :                  *
    1148             :                  * psig->stats_lock also protects us from our sub-threads
    1149             :                  * which can reap other children at the same time. Until
    1150             :                  * we change k_getrusage()-like users to rely on this lock
    1151             :                  * we have to take ->siglock as well.
    1152             :                  *
    1153             :                  * We use thread_group_cputime_adjusted() to get times for
    1154             :                  * the thread group, which consolidates times for all threads
    1155             :                  * in the group including the group leader.
    1156             :                  */
    1157           0 :                 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
    1158           0 :                 spin_lock_irq(&current->sighand->siglock);
    1159           0 :                 write_seqlock(&psig->stats_lock);
    1160           0 :                 psig->cutime += tgutime + sig->cutime;
    1161           0 :                 psig->cstime += tgstime + sig->cstime;
    1162           0 :                 psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime;
    1163           0 :                 psig->cmin_flt +=
    1164           0 :                         p->min_flt + sig->min_flt + sig->cmin_flt;
    1165           0 :                 psig->cmaj_flt +=
    1166           0 :                         p->maj_flt + sig->maj_flt + sig->cmaj_flt;
    1167           0 :                 psig->cnvcsw +=
    1168           0 :                         p->nvcsw + sig->nvcsw + sig->cnvcsw;
    1169           0 :                 psig->cnivcsw +=
    1170           0 :                         p->nivcsw + sig->nivcsw + sig->cnivcsw;
    1171           0 :                 psig->cinblock +=
    1172           0 :                         task_io_get_inblock(p) +
    1173           0 :                         sig->inblock + sig->cinblock;
    1174           0 :                 psig->coublock +=
    1175           0 :                         task_io_get_oublock(p) +
    1176           0 :                         sig->oublock + sig->coublock;
    1177           0 :                 maxrss = max(sig->maxrss, sig->cmaxrss);
    1178           0 :                 if (psig->cmaxrss < maxrss)
    1179           0 :                         psig->cmaxrss = maxrss;
    1180           0 :                 task_io_accounting_add(&psig->ioac, &p->ioac);
    1181           0 :                 task_io_accounting_add(&psig->ioac, &sig->ioac);
    1182           0 :                 write_sequnlock(&psig->stats_lock);
    1183           0 :                 spin_unlock_irq(&current->sighand->siglock);
    1184             :         }
    1185             : 
    1186           0 :         if (wo->wo_rusage)
    1187           0 :                 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
    1188           0 :         status = (p->signal->flags & SIGNAL_GROUP_EXIT)
    1189           0 :                 ? p->signal->group_exit_code : p->exit_code;
    1190           0 :         wo->wo_stat = status;
    1191             : 
    1192           0 :         if (state == EXIT_TRACE) {
    1193           0 :                 write_lock_irq(&tasklist_lock);
    1194             :                 /* We dropped tasklist, ptracer could die and untrace */
    1195           0 :                 ptrace_unlink(p);
    1196             : 
    1197             :                 /* If parent wants a zombie, don't release it now */
    1198           0 :                 state = EXIT_ZOMBIE;
    1199           0 :                 if (do_notify_parent(p, p->exit_signal))
    1200           0 :                         state = EXIT_DEAD;
    1201           0 :                 p->exit_state = state;
    1202           0 :                 write_unlock_irq(&tasklist_lock);
    1203             :         }
    1204           0 :         if (state == EXIT_DEAD)
    1205           0 :                 release_task(p);
    1206             : 
    1207             : out_info:
    1208           0 :         infop = wo->wo_info;
    1209           0 :         if (infop) {
    1210           0 :                 if ((status & 0x7f) == 0) {
    1211           0 :                         infop->cause = CLD_EXITED;
    1212           0 :                         infop->status = status >> 8;
    1213             :                 } else {
    1214           0 :                         infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
    1215           0 :                         infop->status = status & 0x7f;
    1216             :                 }
    1217           0 :                 infop->pid = pid;
    1218           0 :                 infop->uid = uid;
    1219             :         }
    1220             : 
    1221             :         return pid;
    1222             : }
    1223             : 
    1224             : static int *task_stopped_code(struct task_struct *p, bool ptrace)
    1225             : {
    1226           0 :         if (ptrace) {
    1227           0 :                 if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING))
    1228           0 :                         return &p->exit_code;
    1229             :         } else {
    1230           0 :                 if (p->signal->flags & SIGNAL_STOP_STOPPED)
    1231           0 :                         return &p->signal->group_exit_code;
    1232             :         }
    1233             :         return NULL;
    1234             : }
    1235             : 
    1236             : /**
    1237             :  * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
    1238             :  * @wo: wait options
    1239             :  * @ptrace: is the wait for ptrace
    1240             :  * @p: task to wait for
    1241             :  *
    1242             :  * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
    1243             :  *
    1244             :  * CONTEXT:
    1245             :  * read_lock(&tasklist_lock), which is released if return value is
    1246             :  * non-zero.  Also, grabs and releases @p->sighand->siglock.
    1247             :  *
    1248             :  * RETURNS:
    1249             :  * 0 if wait condition didn't exist and search for other wait conditions
    1250             :  * should continue.  Non-zero return, -errno on failure and @p's pid on
    1251             :  * success, implies that tasklist_lock is released and wait condition
    1252             :  * search should terminate.
    1253             :  */
    1254           0 : static int wait_task_stopped(struct wait_opts *wo,
    1255             :                                 int ptrace, struct task_struct *p)
    1256             : {
    1257             :         struct waitid_info *infop;
    1258             :         int exit_code, *p_code, why;
    1259           0 :         uid_t uid = 0; /* unneeded, required by compiler */
    1260             :         pid_t pid;
    1261             : 
    1262             :         /*
    1263             :          * Traditionally we see ptrace'd stopped tasks regardless of options.
    1264             :          */
    1265           0 :         if (!ptrace && !(wo->wo_flags & WUNTRACED))
    1266             :                 return 0;
    1267             : 
    1268           0 :         if (!task_stopped_code(p, ptrace))
    1269             :                 return 0;
    1270             : 
    1271           0 :         exit_code = 0;
    1272           0 :         spin_lock_irq(&p->sighand->siglock);
    1273             : 
    1274           0 :         p_code = task_stopped_code(p, ptrace);
    1275           0 :         if (unlikely(!p_code))
    1276             :                 goto unlock_sig;
    1277             : 
    1278           0 :         exit_code = *p_code;
    1279           0 :         if (!exit_code)
    1280             :                 goto unlock_sig;
    1281             : 
    1282           0 :         if (!unlikely(wo->wo_flags & WNOWAIT))
    1283           0 :                 *p_code = 0;
    1284             : 
    1285           0 :         uid = from_kuid_munged(current_user_ns(), task_uid(p));
    1286             : unlock_sig:
    1287           0 :         spin_unlock_irq(&p->sighand->siglock);
    1288           0 :         if (!exit_code)
    1289             :                 return 0;
    1290             : 
    1291             :         /*
    1292             :          * Now we are pretty sure this task is interesting.
    1293             :          * Make sure it doesn't get reaped out from under us while we
    1294             :          * give up the lock and then examine it below.  We don't want to
    1295             :          * keep holding onto the tasklist_lock while we call getrusage and
    1296             :          * possibly take page faults for user memory.
    1297             :          */
    1298           0 :         get_task_struct(p);
    1299           0 :         pid = task_pid_vnr(p);
    1300           0 :         why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
    1301           0 :         read_unlock(&tasklist_lock);
    1302             :         sched_annotate_sleep();
    1303           0 :         if (wo->wo_rusage)
    1304           0 :                 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
    1305           0 :         put_task_struct(p);
    1306             : 
    1307           0 :         if (likely(!(wo->wo_flags & WNOWAIT)))
    1308           0 :                 wo->wo_stat = (exit_code << 8) | 0x7f;
    1309             : 
    1310           0 :         infop = wo->wo_info;
    1311           0 :         if (infop) {
    1312           0 :                 infop->cause = why;
    1313           0 :                 infop->status = exit_code;
    1314           0 :                 infop->pid = pid;
    1315           0 :                 infop->uid = uid;
    1316             :         }
    1317             :         return pid;
    1318             : }
    1319             : 
    1320             : /*
    1321             :  * Handle do_wait work for one task in a live, non-stopped state.
    1322             :  * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
    1323             :  * the lock and this task is uninteresting.  If we return nonzero, we have
    1324             :  * released the lock and the system call should return.
    1325             :  */
    1326           0 : static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
    1327             : {
    1328             :         struct waitid_info *infop;
    1329             :         pid_t pid;
    1330             :         uid_t uid;
    1331             : 
    1332           0 :         if (!unlikely(wo->wo_flags & WCONTINUED))
    1333             :                 return 0;
    1334             : 
    1335           0 :         if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
    1336             :                 return 0;
    1337             : 
    1338           0 :         spin_lock_irq(&p->sighand->siglock);
    1339             :         /* Re-check with the lock held.  */
    1340           0 :         if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
    1341           0 :                 spin_unlock_irq(&p->sighand->siglock);
    1342           0 :                 return 0;
    1343             :         }
    1344           0 :         if (!unlikely(wo->wo_flags & WNOWAIT))
    1345           0 :                 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
    1346           0 :         uid = from_kuid_munged(current_user_ns(), task_uid(p));
    1347           0 :         spin_unlock_irq(&p->sighand->siglock);
    1348             : 
    1349           0 :         pid = task_pid_vnr(p);
    1350           0 :         get_task_struct(p);
    1351           0 :         read_unlock(&tasklist_lock);
    1352             :         sched_annotate_sleep();
    1353           0 :         if (wo->wo_rusage)
    1354           0 :                 getrusage(p, RUSAGE_BOTH, wo->wo_rusage);
    1355           0 :         put_task_struct(p);
    1356             : 
    1357           0 :         infop = wo->wo_info;
    1358           0 :         if (!infop) {
    1359           0 :                 wo->wo_stat = 0xffff;
    1360             :         } else {
    1361           0 :                 infop->cause = CLD_CONTINUED;
    1362           0 :                 infop->pid = pid;
    1363           0 :                 infop->uid = uid;
    1364           0 :                 infop->status = SIGCONT;
    1365             :         }
    1366             :         return pid;
    1367             : }
    1368             : 
    1369             : /*
    1370             :  * Consider @p for a wait by @parent.
    1371             :  *
    1372             :  * -ECHILD should be in ->notask_error before the first call.
    1373             :  * Returns nonzero for a final return, when we have unlocked tasklist_lock.
    1374             :  * Returns zero if the search for a child should continue;
    1375             :  * then ->notask_error is 0 if @p is an eligible child,
    1376             :  * or still -ECHILD.
    1377             :  */
    1378           0 : static int wait_consider_task(struct wait_opts *wo, int ptrace,
    1379             :                                 struct task_struct *p)
    1380             : {
    1381             :         /*
    1382             :          * We can race with wait_task_zombie() from another thread.
    1383             :          * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition
    1384             :          * can't confuse the checks below.
    1385             :          */
    1386           0 :         int exit_state = READ_ONCE(p->exit_state);
    1387             :         int ret;
    1388             : 
    1389           0 :         if (unlikely(exit_state == EXIT_DEAD))
    1390             :                 return 0;
    1391             : 
    1392           0 :         ret = eligible_child(wo, ptrace, p);
    1393           0 :         if (!ret)
    1394             :                 return ret;
    1395             : 
    1396           0 :         if (unlikely(exit_state == EXIT_TRACE)) {
    1397             :                 /*
    1398             :                  * ptrace == 0 means we are the natural parent. In this case
    1399             :                  * we should clear notask_error, debugger will notify us.
    1400             :                  */
    1401           0 :                 if (likely(!ptrace))
    1402           0 :                         wo->notask_error = 0;
    1403             :                 return 0;
    1404             :         }
    1405             : 
    1406           0 :         if (likely(!ptrace) && unlikely(p->ptrace)) {
    1407             :                 /*
    1408             :                  * If it is traced by its real parent's group, just pretend
    1409             :                  * the caller is ptrace_do_wait() and reap this child if it
    1410             :                  * is zombie.
    1411             :                  *
    1412             :                  * This also hides group stop state from real parent; otherwise
    1413             :                  * a single stop can be reported twice as group and ptrace stop.
    1414             :                  * If a ptracer wants to distinguish these two events for its
    1415             :                  * own children it should create a separate process which takes
    1416             :                  * the role of real parent.
    1417             :                  */
    1418           0 :                 if (!ptrace_reparented(p))
    1419           0 :                         ptrace = 1;
    1420             :         }
    1421             : 
    1422             :         /* slay zombie? */
    1423           0 :         if (exit_state == EXIT_ZOMBIE) {
    1424             :                 /* we don't reap group leaders with subthreads */
    1425           0 :                 if (!delay_group_leader(p)) {
    1426             :                         /*
    1427             :                          * A zombie ptracee is only visible to its ptracer.
    1428             :                          * Notification and reaping will be cascaded to the
    1429             :                          * real parent when the ptracer detaches.
    1430             :                          */
    1431           0 :                         if (unlikely(ptrace) || likely(!p->ptrace))
    1432           0 :                                 return wait_task_zombie(wo, p);
    1433             :                 }
    1434             : 
    1435             :                 /*
    1436             :                  * Allow access to stopped/continued state via zombie by
    1437             :                  * falling through.  Clearing of notask_error is complex.
    1438             :                  *
    1439             :                  * When !@ptrace:
    1440             :                  *
    1441             :                  * If WEXITED is set, notask_error should naturally be
    1442             :                  * cleared.  If not, subset of WSTOPPED|WCONTINUED is set,
    1443             :                  * so, if there are live subthreads, there are events to
    1444             :                  * wait for.  If all subthreads are dead, it's still safe
    1445             :                  * to clear - this function will be called again in finite
    1446             :                  * amount time once all the subthreads are released and
    1447             :                  * will then return without clearing.
    1448             :                  *
    1449             :                  * When @ptrace:
    1450             :                  *
    1451             :                  * Stopped state is per-task and thus can't change once the
    1452             :                  * target task dies.  Only continued and exited can happen.
    1453             :                  * Clear notask_error if WCONTINUED | WEXITED.
    1454             :                  */
    1455           0 :                 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
    1456           0 :                         wo->notask_error = 0;
    1457             :         } else {
    1458             :                 /*
    1459             :                  * @p is alive and it's gonna stop, continue or exit, so
    1460             :                  * there always is something to wait for.
    1461             :                  */
    1462           0 :                 wo->notask_error = 0;
    1463             :         }
    1464             : 
    1465             :         /*
    1466             :          * Wait for stopped.  Depending on @ptrace, different stopped state
    1467             :          * is used and the two don't interact with each other.
    1468             :          */
    1469           0 :         ret = wait_task_stopped(wo, ptrace, p);
    1470           0 :         if (ret)
    1471             :                 return ret;
    1472             : 
    1473             :         /*
    1474             :          * Wait for continued.  There's only one continued state and the
    1475             :          * ptracer can consume it which can confuse the real parent.  Don't
    1476             :          * use WCONTINUED from ptracer.  You don't need or want it.
    1477             :          */
    1478           0 :         return wait_task_continued(wo, p);
    1479             : }
    1480             : 
    1481             : /*
    1482             :  * Do the work of do_wait() for one thread in the group, @tsk.
    1483             :  *
    1484             :  * -ECHILD should be in ->notask_error before the first call.
    1485             :  * Returns nonzero for a final return, when we have unlocked tasklist_lock.
    1486             :  * Returns zero if the search for a child should continue; then
    1487             :  * ->notask_error is 0 if there were any eligible children,
    1488             :  * or still -ECHILD.
    1489             :  */
    1490           0 : static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
    1491             : {
    1492             :         struct task_struct *p;
    1493             : 
    1494           0 :         list_for_each_entry(p, &tsk->children, sibling) {
    1495           0 :                 int ret = wait_consider_task(wo, 0, p);
    1496             : 
    1497           0 :                 if (ret)
    1498             :                         return ret;
    1499             :         }
    1500             : 
    1501             :         return 0;
    1502             : }
    1503             : 
    1504           0 : static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
    1505             : {
    1506             :         struct task_struct *p;
    1507             : 
    1508           0 :         list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
    1509           0 :                 int ret = wait_consider_task(wo, 1, p);
    1510             : 
    1511           0 :                 if (ret)
    1512             :                         return ret;
    1513             :         }
    1514             : 
    1515             :         return 0;
    1516             : }
    1517             : 
    1518           0 : static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode,
    1519             :                                 int sync, void *key)
    1520             : {
    1521           0 :         struct wait_opts *wo = container_of(wait, struct wait_opts,
    1522             :                                                 child_wait);
    1523           0 :         struct task_struct *p = key;
    1524             : 
    1525           0 :         if (!eligible_pid(wo, p))
    1526             :                 return 0;
    1527             : 
    1528           0 :         if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
    1529             :                 return 0;
    1530             : 
    1531           0 :         return default_wake_function(wait, mode, sync, key);
    1532             : }
    1533             : 
    1534         333 : void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
    1535             : {
    1536         333 :         __wake_up_sync_key(&parent->signal->wait_chldexit,
    1537             :                            TASK_INTERRUPTIBLE, p);
    1538         333 : }
    1539             : 
    1540             : static bool is_effectively_child(struct wait_opts *wo, bool ptrace,
    1541             :                                  struct task_struct *target)
    1542             : {
    1543           0 :         struct task_struct *parent =
    1544           0 :                 !ptrace ? target->real_parent : target->parent;
    1545             : 
    1546           0 :         return current == parent || (!(wo->wo_flags & __WNOTHREAD) &&
    1547           0 :                                      same_thread_group(current, parent));
    1548             : }
    1549             : 
    1550             : /*
    1551             :  * Optimization for waiting on PIDTYPE_PID. No need to iterate through child
    1552             :  * and tracee lists to find the target task.
    1553             :  */
    1554           0 : static int do_wait_pid(struct wait_opts *wo)
    1555             : {
    1556             :         bool ptrace;
    1557             :         struct task_struct *target;
    1558             :         int retval;
    1559             : 
    1560           0 :         ptrace = false;
    1561           0 :         target = pid_task(wo->wo_pid, PIDTYPE_TGID);
    1562           0 :         if (target && is_effectively_child(wo, ptrace, target)) {
    1563           0 :                 retval = wait_consider_task(wo, ptrace, target);
    1564           0 :                 if (retval)
    1565             :                         return retval;
    1566             :         }
    1567             : 
    1568           0 :         ptrace = true;
    1569           0 :         target = pid_task(wo->wo_pid, PIDTYPE_PID);
    1570           0 :         if (target && target->ptrace &&
    1571           0 :             is_effectively_child(wo, ptrace, target)) {
    1572           0 :                 retval = wait_consider_task(wo, ptrace, target);
    1573           0 :                 if (retval)
    1574             :                         return retval;
    1575             :         }
    1576             : 
    1577             :         return 0;
    1578             : }
    1579             : 
    1580           0 : static long do_wait(struct wait_opts *wo)
    1581             : {
    1582             :         int retval;
    1583             : 
    1584           0 :         trace_sched_process_wait(wo->wo_pid);
    1585             : 
    1586           0 :         init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
    1587           0 :         wo->child_wait.private = current;
    1588           0 :         add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
    1589             : repeat:
    1590             :         /*
    1591             :          * If there is nothing that can match our criteria, just get out.
    1592             :          * We will clear ->notask_error to zero if we see any child that
    1593             :          * might later match our criteria, even if we are not able to reap
    1594             :          * it yet.
    1595             :          */
    1596           0 :         wo->notask_error = -ECHILD;
    1597           0 :         if ((wo->wo_type < PIDTYPE_MAX) &&
    1598           0 :            (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type)))
    1599             :                 goto notask;
    1600             : 
    1601           0 :         set_current_state(TASK_INTERRUPTIBLE);
    1602           0 :         read_lock(&tasklist_lock);
    1603             : 
    1604           0 :         if (wo->wo_type == PIDTYPE_PID) {
    1605           0 :                 retval = do_wait_pid(wo);
    1606           0 :                 if (retval)
    1607             :                         goto end;
    1608             :         } else {
    1609           0 :                 struct task_struct *tsk = current;
    1610             : 
    1611             :                 do {
    1612           0 :                         retval = do_wait_thread(wo, tsk);
    1613           0 :                         if (retval)
    1614             :                                 goto end;
    1615             : 
    1616           0 :                         retval = ptrace_do_wait(wo, tsk);
    1617           0 :                         if (retval)
    1618             :                                 goto end;
    1619             : 
    1620           0 :                         if (wo->wo_flags & __WNOTHREAD)
    1621             :                                 break;
    1622           0 :                 } while_each_thread(current, tsk);
    1623             :         }
    1624           0 :         read_unlock(&tasklist_lock);
    1625             : 
    1626             : notask:
    1627           0 :         retval = wo->notask_error;
    1628           0 :         if (!retval && !(wo->wo_flags & WNOHANG)) {
    1629           0 :                 retval = -ERESTARTSYS;
    1630           0 :                 if (!signal_pending(current)) {
    1631           0 :                         schedule();
    1632           0 :                         goto repeat;
    1633             :                 }
    1634             :         }
    1635             : end:
    1636           0 :         __set_current_state(TASK_RUNNING);
    1637           0 :         remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
    1638           0 :         return retval;
    1639             : }
    1640             : 
    1641           0 : static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop,
    1642             :                           int options, struct rusage *ru)
    1643             : {
    1644             :         struct wait_opts wo;
    1645           0 :         struct pid *pid = NULL;
    1646             :         enum pid_type type;
    1647             :         long ret;
    1648           0 :         unsigned int f_flags = 0;
    1649             : 
    1650           0 :         if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED|
    1651             :                         __WNOTHREAD|__WCLONE|__WALL))
    1652             :                 return -EINVAL;
    1653           0 :         if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
    1654             :                 return -EINVAL;
    1655             : 
    1656           0 :         switch (which) {
    1657             :         case P_ALL:
    1658             :                 type = PIDTYPE_MAX;
    1659             :                 break;
    1660             :         case P_PID:
    1661           0 :                 type = PIDTYPE_PID;
    1662           0 :                 if (upid <= 0)
    1663             :                         return -EINVAL;
    1664             : 
    1665           0 :                 pid = find_get_pid(upid);
    1666           0 :                 break;
    1667             :         case P_PGID:
    1668           0 :                 type = PIDTYPE_PGID;
    1669           0 :                 if (upid < 0)
    1670             :                         return -EINVAL;
    1671             : 
    1672           0 :                 if (upid)
    1673           0 :                         pid = find_get_pid(upid);
    1674             :                 else
    1675           0 :                         pid = get_task_pid(current, PIDTYPE_PGID);
    1676             :                 break;
    1677             :         case P_PIDFD:
    1678           0 :                 type = PIDTYPE_PID;
    1679           0 :                 if (upid < 0)
    1680             :                         return -EINVAL;
    1681             : 
    1682           0 :                 pid = pidfd_get_pid(upid, &f_flags);
    1683           0 :                 if (IS_ERR(pid))
    1684           0 :                         return PTR_ERR(pid);
    1685             : 
    1686             :                 break;
    1687             :         default:
    1688             :                 return -EINVAL;
    1689             :         }
    1690             : 
    1691           0 :         wo.wo_type      = type;
    1692           0 :         wo.wo_pid       = pid;
    1693           0 :         wo.wo_flags     = options;
    1694           0 :         wo.wo_info      = infop;
    1695           0 :         wo.wo_rusage    = ru;
    1696           0 :         if (f_flags & O_NONBLOCK)
    1697           0 :                 wo.wo_flags |= WNOHANG;
    1698             : 
    1699           0 :         ret = do_wait(&wo);
    1700           0 :         if (!ret && !(options & WNOHANG) && (f_flags & O_NONBLOCK))
    1701           0 :                 ret = -EAGAIN;
    1702             : 
    1703           0 :         put_pid(pid);
    1704           0 :         return ret;
    1705             : }
    1706             : 
    1707           0 : SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
    1708             :                 infop, int, options, struct rusage __user *, ru)
    1709             : {
    1710             :         struct rusage r;
    1711           0 :         struct waitid_info info = {.status = 0};
    1712           0 :         long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL);
    1713           0 :         int signo = 0;
    1714             : 
    1715           0 :         if (err > 0) {
    1716           0 :                 signo = SIGCHLD;
    1717           0 :                 err = 0;
    1718           0 :                 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
    1719             :                         return -EFAULT;
    1720             :         }
    1721           0 :         if (!infop)
    1722             :                 return err;
    1723             : 
    1724           0 :         if (!user_write_access_begin(infop, sizeof(*infop)))
    1725             :                 return -EFAULT;
    1726             : 
    1727           0 :         unsafe_put_user(signo, &infop->si_signo, Efault);
    1728           0 :         unsafe_put_user(0, &infop->si_errno, Efault);
    1729           0 :         unsafe_put_user(info.cause, &infop->si_code, Efault);
    1730           0 :         unsafe_put_user(info.pid, &infop->si_pid, Efault);
    1731           0 :         unsafe_put_user(info.uid, &infop->si_uid, Efault);
    1732           0 :         unsafe_put_user(info.status, &infop->si_status, Efault);
    1733             :         user_write_access_end();
    1734             :         return err;
    1735             : Efault:
    1736             :         user_write_access_end();
    1737             :         return -EFAULT;
    1738             : }
    1739             : 
    1740           0 : long kernel_wait4(pid_t upid, int __user *stat_addr, int options,
    1741             :                   struct rusage *ru)
    1742             : {
    1743             :         struct wait_opts wo;
    1744           0 :         struct pid *pid = NULL;
    1745             :         enum pid_type type;
    1746             :         long ret;
    1747             : 
    1748           0 :         if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
    1749             :                         __WNOTHREAD|__WCLONE|__WALL))
    1750             :                 return -EINVAL;
    1751             : 
    1752             :         /* -INT_MIN is not defined */
    1753           0 :         if (upid == INT_MIN)
    1754             :                 return -ESRCH;
    1755             : 
    1756           0 :         if (upid == -1)
    1757             :                 type = PIDTYPE_MAX;
    1758           0 :         else if (upid < 0) {
    1759           0 :                 type = PIDTYPE_PGID;
    1760           0 :                 pid = find_get_pid(-upid);
    1761           0 :         } else if (upid == 0) {
    1762           0 :                 type = PIDTYPE_PGID;
    1763           0 :                 pid = get_task_pid(current, PIDTYPE_PGID);
    1764             :         } else /* upid > 0 */ {
    1765           0 :                 type = PIDTYPE_PID;
    1766           0 :                 pid = find_get_pid(upid);
    1767             :         }
    1768             : 
    1769           0 :         wo.wo_type      = type;
    1770           0 :         wo.wo_pid       = pid;
    1771           0 :         wo.wo_flags     = options | WEXITED;
    1772           0 :         wo.wo_info      = NULL;
    1773           0 :         wo.wo_stat      = 0;
    1774           0 :         wo.wo_rusage    = ru;
    1775           0 :         ret = do_wait(&wo);
    1776           0 :         put_pid(pid);
    1777           0 :         if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr))
    1778           0 :                 ret = -EFAULT;
    1779             : 
    1780             :         return ret;
    1781             : }
    1782             : 
    1783           0 : int kernel_wait(pid_t pid, int *stat)
    1784             : {
    1785           0 :         struct wait_opts wo = {
    1786             :                 .wo_type        = PIDTYPE_PID,
    1787           0 :                 .wo_pid         = find_get_pid(pid),
    1788             :                 .wo_flags       = WEXITED,
    1789             :         };
    1790             :         int ret;
    1791             : 
    1792           0 :         ret = do_wait(&wo);
    1793           0 :         if (ret > 0 && wo.wo_stat)
    1794           0 :                 *stat = wo.wo_stat;
    1795           0 :         put_pid(wo.wo_pid);
    1796           0 :         return ret;
    1797             : }
    1798             : 
    1799           0 : SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
    1800             :                 int, options, struct rusage __user *, ru)
    1801             : {
    1802             :         struct rusage r;
    1803           0 :         long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL);
    1804             : 
    1805           0 :         if (err > 0) {
    1806           0 :                 if (ru && copy_to_user(ru, &r, sizeof(struct rusage)))
    1807             :                         return -EFAULT;
    1808             :         }
    1809             :         return err;
    1810             : }
    1811             : 
    1812             : #ifdef __ARCH_WANT_SYS_WAITPID
    1813             : 
    1814             : /*
    1815             :  * sys_waitpid() remains for compatibility. waitpid() should be
    1816             :  * implemented by calling sys_wait4() from libc.a.
    1817             :  */
    1818           0 : SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
    1819             : {
    1820           0 :         return kernel_wait4(pid, stat_addr, options, NULL);
    1821             : }
    1822             : 
    1823             : #endif
    1824             : 
    1825             : #ifdef CONFIG_COMPAT
    1826             : COMPAT_SYSCALL_DEFINE4(wait4,
    1827             :         compat_pid_t, pid,
    1828             :         compat_uint_t __user *, stat_addr,
    1829             :         int, options,
    1830             :         struct compat_rusage __user *, ru)
    1831             : {
    1832             :         struct rusage r;
    1833             :         long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL);
    1834             :         if (err > 0) {
    1835             :                 if (ru && put_compat_rusage(&r, ru))
    1836             :                         return -EFAULT;
    1837             :         }
    1838             :         return err;
    1839             : }
    1840             : 
    1841             : COMPAT_SYSCALL_DEFINE5(waitid,
    1842             :                 int, which, compat_pid_t, pid,
    1843             :                 struct compat_siginfo __user *, infop, int, options,
    1844             :                 struct compat_rusage __user *, uru)
    1845             : {
    1846             :         struct rusage ru;
    1847             :         struct waitid_info info = {.status = 0};
    1848             :         long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL);
    1849             :         int signo = 0;
    1850             :         if (err > 0) {
    1851             :                 signo = SIGCHLD;
    1852             :                 err = 0;
    1853             :                 if (uru) {
    1854             :                         /* kernel_waitid() overwrites everything in ru */
    1855             :                         if (COMPAT_USE_64BIT_TIME)
    1856             :                                 err = copy_to_user(uru, &ru, sizeof(ru));
    1857             :                         else
    1858             :                                 err = put_compat_rusage(&ru, uru);
    1859             :                         if (err)
    1860             :                                 return -EFAULT;
    1861             :                 }
    1862             :         }
    1863             : 
    1864             :         if (!infop)
    1865             :                 return err;
    1866             : 
    1867             :         if (!user_write_access_begin(infop, sizeof(*infop)))
    1868             :                 return -EFAULT;
    1869             : 
    1870             :         unsafe_put_user(signo, &infop->si_signo, Efault);
    1871             :         unsafe_put_user(0, &infop->si_errno, Efault);
    1872             :         unsafe_put_user(info.cause, &infop->si_code, Efault);
    1873             :         unsafe_put_user(info.pid, &infop->si_pid, Efault);
    1874             :         unsafe_put_user(info.uid, &infop->si_uid, Efault);
    1875             :         unsafe_put_user(info.status, &infop->si_status, Efault);
    1876             :         user_write_access_end();
    1877             :         return err;
    1878             : Efault:
    1879             :         user_write_access_end();
    1880             :         return -EFAULT;
    1881             : }
    1882             : #endif
    1883             : 
    1884             : /**
    1885             :  * thread_group_exited - check that a thread group has exited
    1886             :  * @pid: tgid of thread group to be checked.
    1887             :  *
    1888             :  * Test if the thread group represented by tgid has exited (all
    1889             :  * threads are zombies, dead or completely gone).
    1890             :  *
    1891             :  * Return: true if the thread group has exited. false otherwise.
    1892             :  */
    1893           0 : bool thread_group_exited(struct pid *pid)
    1894             : {
    1895             :         struct task_struct *task;
    1896             :         bool exited;
    1897             : 
    1898             :         rcu_read_lock();
    1899           0 :         task = pid_task(pid, PIDTYPE_PID);
    1900           0 :         exited = !task ||
    1901           0 :                 (READ_ONCE(task->exit_state) && thread_group_empty(task));
    1902             :         rcu_read_unlock();
    1903             : 
    1904           0 :         return exited;
    1905             : }
    1906             : EXPORT_SYMBOL(thread_group_exited);
    1907             : 
    1908             : /*
    1909             :  * This needs to be __function_aligned as GCC implicitly makes any
    1910             :  * implementation of abort() cold and drops alignment specified by
    1911             :  * -falign-functions=N.
    1912             :  *
    1913             :  * See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88345#c11
    1914             :  */
    1915           0 : __weak __function_aligned void abort(void)
    1916             : {
    1917           0 :         BUG();
    1918             : 
    1919             :         /* if that doesn't kill us, halt */
    1920             :         panic("Oops failed to kill thread");
    1921             : }
    1922             : EXPORT_SYMBOL(abort);

Generated by: LCOV version 1.14