Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/kernel/exit.c |
| 3 | * |
| 4 | * Copyright (C) 1991, 1992 Linus Torvalds |
| 5 | */ |
| 6 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | #include <linux/mm.h> |
| 8 | #include <linux/slab.h> |
| 9 | #include <linux/interrupt.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 10 | #include <linux/module.h> |
Randy.Dunlap | c59ede7 | 2006-01-11 12:17:46 -0800 | [diff] [blame] | 11 | #include <linux/capability.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 12 | #include <linux/completion.h> |
| 13 | #include <linux/personality.h> |
| 14 | #include <linux/tty.h> |
Jens Axboe | da9cbc8 | 2008-06-30 20:42:08 +0200 | [diff] [blame] | 15 | #include <linux/iocontext.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 16 | #include <linux/key.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 17 | #include <linux/cpu.h> |
| 18 | #include <linux/acct.h> |
Jay Lan | 8f0ab51 | 2006-09-30 23:28:59 -0700 | [diff] [blame] | 19 | #include <linux/tsacct_kern.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 20 | #include <linux/file.h> |
Al Viro | 9f3acc3 | 2008-04-24 07:44:08 -0400 | [diff] [blame] | 21 | #include <linux/fdtable.h> |
Mandeep Singh Baines | 80d26af | 2013-02-27 17:03:20 -0800 | [diff] [blame] | 22 | #include <linux/freezer.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 23 | #include <linux/binfmts.h> |
Serge E. Hallyn | ab51601 | 2006-10-02 02:18:06 -0700 | [diff] [blame] | 24 | #include <linux/nsproxy.h> |
Sukadev Bhattiprolu | 84d7378 | 2006-12-08 02:38:01 -0800 | [diff] [blame] | 25 | #include <linux/pid_namespace.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 26 | #include <linux/ptrace.h> |
| 27 | #include <linux/profile.h> |
| 28 | #include <linux/mount.h> |
| 29 | #include <linux/proc_fs.h> |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 30 | #include <linux/kthread.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | #include <linux/mempolicy.h> |
Shailabh Nagar | c757249 | 2006-07-14 00:24:40 -0700 | [diff] [blame] | 32 | #include <linux/taskstats_kern.h> |
Shailabh Nagar | ca74e92 | 2006-07-14 00:24:36 -0700 | [diff] [blame] | 33 | #include <linux/delayacct.h> |
Paul Menage | b4f48b6 | 2007-10-18 23:39:33 -0700 | [diff] [blame] | 34 | #include <linux/cgroup.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 35 | #include <linux/syscalls.h> |
Jesper Juhl | 7ed20e1 | 2005-05-01 08:59:14 -0700 | [diff] [blame] | 36 | #include <linux/signal.h> |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 37 | #include <linux/posix-timers.h> |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 38 | #include <linux/cn_proc.h> |
Ingo Molnar | de5097c | 2006-01-09 15:59:21 -0800 | [diff] [blame] | 39 | #include <linux/mutex.h> |
Ingo Molnar | 0771dfe | 2006-03-27 01:16:22 -0800 | [diff] [blame] | 40 | #include <linux/futex.h> |
Jens Axboe | b92ce55 | 2006-04-11 13:52:07 +0200 | [diff] [blame] | 41 | #include <linux/pipe_fs_i.h> |
Al Viro | fa84cb9 | 2006-03-29 20:30:19 -0500 | [diff] [blame] | 42 | #include <linux/audit.h> /* for audit_free() */ |
Adrian Bunk | 83cc5ed | 2006-06-25 05:47:41 -0700 | [diff] [blame] | 43 | #include <linux/resource.h> |
David Howells | 0d67a46 | 2006-08-29 19:05:56 +0100 | [diff] [blame] | 44 | #include <linux/blkdev.h> |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 45 | #include <linux/task_io_accounting_ops.h> |
Roland McGrath | 30199f5 | 2008-07-25 19:45:46 -0700 | [diff] [blame] | 46 | #include <linux/tracehook.h> |
Al Viro | 5ad4e53 | 2009-03-29 19:50:06 -0400 | [diff] [blame] | 47 | #include <linux/fs_struct.h> |
David Howells | d84f4f9 | 2008-11-14 10:39:23 +1100 | [diff] [blame] | 48 | #include <linux/init_task.h> |
Ingo Molnar | cdd6c48 | 2009-09-21 12:02:48 +0200 | [diff] [blame] | 49 | #include <linux/perf_event.h> |
Steven Rostedt | ad8d75f | 2009-04-14 19:39:12 -0400 | [diff] [blame] | 50 | #include <trace/events/sched.h> |
Frederic Weisbecker | 24f1e32c | 2009-09-09 19:22:48 +0200 | [diff] [blame] | 51 | #include <linux/hw_breakpoint.h> |
Ying Han | 3d5992d | 2010-10-26 14:21:23 -0700 | [diff] [blame] | 52 | #include <linux/oom.h> |
Wu Fengguang | 54848d7 | 2011-04-05 13:21:19 -0600 | [diff] [blame] | 53 | #include <linux/writeback.h> |
Al Viro | 4040153 | 2012-02-13 03:58:52 +0000 | [diff] [blame] | 54 | #include <linux/shm.h> |
Dmitry Vyukov | 5c9a875 | 2016-03-22 14:27:30 -0700 | [diff] [blame] | 55 | #include <linux/kcov.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 56 | |
| 57 | #include <asm/uaccess.h> |
| 58 | #include <asm/unistd.h> |
| 59 | #include <asm/pgtable.h> |
| 60 | #include <asm/mmu_context.h> |
| 61 | |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 62 | static void __unhash_process(struct task_struct *p, bool group_dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 63 | { |
| 64 | nr_threads--; |
Oleg Nesterov | 50d75f8 | 2012-06-20 12:53:04 -0700 | [diff] [blame] | 65 | detach_pid(p, PIDTYPE_PID); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 66 | if (group_dead) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 67 | detach_pid(p, PIDTYPE_PGID); |
| 68 | detach_pid(p, PIDTYPE_SID); |
Oleg Nesterov | c97d989 | 2006-03-28 16:11:06 -0800 | [diff] [blame] | 69 | |
Eric W. Biederman | 5e85d4a | 2006-04-18 22:20:16 -0700 | [diff] [blame] | 70 | list_del_rcu(&p->tasks); |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 71 | list_del_init(&p->sibling); |
Christoph Lameter | 909ea96 | 2010-12-08 16:22:55 +0100 | [diff] [blame] | 72 | __this_cpu_dec(process_counts); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 73 | } |
Oleg Nesterov | 47e6532 | 2006-03-28 16:11:25 -0800 | [diff] [blame] | 74 | list_del_rcu(&p->thread_group); |
Oleg Nesterov | 0c740d0 | 2014-01-21 15:49:56 -0800 | [diff] [blame] | 75 | list_del_rcu(&p->thread_node); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 76 | } |
| 77 | |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 78 | /* |
| 79 | * This function expects the tasklist_lock write-locked. |
| 80 | */ |
| 81 | static void __exit_signal(struct task_struct *tsk) |
| 82 | { |
| 83 | struct signal_struct *sig = tsk->signal; |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 84 | bool group_dead = thread_group_leader(tsk); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 85 | struct sighand_struct *sighand; |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 86 | struct tty_struct *uninitialized_var(tty); |
Frederic Weisbecker | 6fac482 | 2012-11-13 14:20:55 +0100 | [diff] [blame] | 87 | cputime_t utime, stime; |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 88 | |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 89 | sighand = rcu_dereference_check(tsk->sighand, |
Paul E. McKenney | db1466b | 2010-03-03 07:46:56 -0800 | [diff] [blame] | 90 | lockdep_tasklist_lock_is_held()); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 91 | spin_lock(&sighand->siglock); |
| 92 | |
| 93 | posix_cpu_timers_exit(tsk); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 94 | if (group_dead) { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 95 | posix_cpu_timers_exit_group(tsk); |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 96 | tty = sig->tty; |
| 97 | sig->tty = NULL; |
Oleg Nesterov | 4a59994 | 2010-05-26 14:43:12 -0700 | [diff] [blame] | 98 | } else { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 99 | /* |
Oleg Nesterov | e0a7021 | 2010-11-05 16:53:42 +0100 | [diff] [blame] | 100 | * This can only happen if the caller is de_thread(). |
| 101 | * FIXME: this is the temporary hack, we should teach |
| 102 | * posix-cpu-timers to handle this case correctly. |
| 103 | */ |
| 104 | if (unlikely(has_group_leader_pid(tsk))) |
| 105 | posix_cpu_timers_exit_group(tsk); |
| 106 | |
| 107 | /* |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 108 | * If there is any task waiting for the group exit |
| 109 | * then notify it: |
| 110 | */ |
Oleg Nesterov | d344193 | 2010-05-26 14:43:11 -0700 | [diff] [blame] | 111 | if (sig->notify_count > 0 && !--sig->notify_count) |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 112 | wake_up_process(sig->group_exit_task); |
Oleg Nesterov | 6db840f | 2007-10-16 23:27:23 -0700 | [diff] [blame] | 113 | |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 114 | if (tsk == sig->curr_target) |
| 115 | sig->curr_target = next_thread(tsk); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 116 | } |
| 117 | |
Rik van Riel | 90ed9cb | 2014-08-15 16:05:36 -0400 | [diff] [blame] | 118 | /* |
Oleg Nesterov | 26e75b5 | 2014-12-10 15:54:54 -0800 | [diff] [blame] | 119 | * Accumulate here the counters for all threads as they die. We could |
| 120 | * skip the group leader because it is the last user of signal_struct, |
| 121 | * but we want to avoid the race with thread_group_cputime() which can |
| 122 | * see the empty ->thread_head list. |
Rik van Riel | 90ed9cb | 2014-08-15 16:05:36 -0400 | [diff] [blame] | 123 | */ |
| 124 | task_cputime(tsk, &utime, &stime); |
Rik van Riel | e78c349 | 2014-08-16 13:40:10 -0400 | [diff] [blame] | 125 | write_seqlock(&sig->stats_lock); |
Rik van Riel | 90ed9cb | 2014-08-15 16:05:36 -0400 | [diff] [blame] | 126 | sig->utime += utime; |
| 127 | sig->stime += stime; |
| 128 | sig->gtime += task_gtime(tsk); |
| 129 | sig->min_flt += tsk->min_flt; |
| 130 | sig->maj_flt += tsk->maj_flt; |
| 131 | sig->nvcsw += tsk->nvcsw; |
| 132 | sig->nivcsw += tsk->nivcsw; |
| 133 | sig->inblock += task_io_get_inblock(tsk); |
| 134 | sig->oublock += task_io_get_oublock(tsk); |
| 135 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
| 136 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
Oleg Nesterov | b3ac022 | 2010-05-26 14:43:24 -0700 | [diff] [blame] | 137 | sig->nr_threads--; |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 138 | __unhash_process(tsk, group_dead); |
Rik van Riel | e78c349 | 2014-08-16 13:40:10 -0400 | [diff] [blame] | 139 | write_sequnlock(&sig->stats_lock); |
Oleg Nesterov | 5876700 | 2006-03-28 16:11:20 -0800 | [diff] [blame] | 140 | |
Oleg Nesterov | da7978b | 2008-05-23 13:04:41 -0700 | [diff] [blame] | 141 | /* |
| 142 | * Do this under ->siglock, we can race with another thread |
| 143 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. |
| 144 | */ |
| 145 | flush_sigqueue(&tsk->pending); |
Oleg Nesterov | a7e5328 | 2006-03-28 16:11:27 -0800 | [diff] [blame] | 146 | tsk->sighand = NULL; |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 147 | spin_unlock(&sighand->siglock); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 148 | |
Oleg Nesterov | a7e5328 | 2006-03-28 16:11:27 -0800 | [diff] [blame] | 149 | __cleanup_sighand(sighand); |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 150 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 151 | if (group_dead) { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 152 | flush_sigqueue(&sig->shared_pending); |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 153 | tty_kref_put(tty); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 154 | } |
| 155 | } |
| 156 | |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 157 | static void delayed_put_task_struct(struct rcu_head *rhp) |
| 158 | { |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 159 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
| 160 | |
Peter Zijlstra | 4e231c7 | 2010-09-09 21:01:59 +0200 | [diff] [blame] | 161 | perf_event_delayed_put(tsk); |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 162 | trace_sched_process_free(tsk); |
| 163 | put_task_struct(tsk); |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 164 | } |
| 165 | |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 166 | |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 167 | void release_task(struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 168 | { |
Ingo Molnar | 36c8b58 | 2006-07-03 00:25:41 -0700 | [diff] [blame] | 169 | struct task_struct *leader; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 170 | int zap_leader; |
Oleg Nesterov | 1f09f97 | 2006-03-28 16:11:11 -0800 | [diff] [blame] | 171 | repeat: |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 172 | /* don't need to get the RCU readlock here - the process is dead and |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 173 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
| 174 | rcu_read_lock(); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 175 | atomic_dec(&__task_cred(p)->user->processes); |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 176 | rcu_read_unlock(); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 177 | |
Pavel Emelyanov | 60347f6 | 2007-10-18 23:40:03 -0700 | [diff] [blame] | 178 | proc_flush_task(p); |
Ingo Molnar | 0203026 | 2009-05-17 11:24:08 +0200 | [diff] [blame] | 179 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 180 | write_lock_irq(&tasklist_lock); |
Tejun Heo | a288eec | 2011-06-17 16:50:37 +0200 | [diff] [blame] | 181 | ptrace_release_task(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | __exit_signal(p); |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 183 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 184 | /* |
| 185 | * If we are the last non-leader member of the thread |
| 186 | * group, and the leader is zombie, then notify the |
| 187 | * group leader's parent process. (if it wants notification.) |
| 188 | */ |
| 189 | zap_leader = 0; |
| 190 | leader = p->group_leader; |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 191 | if (leader != p && thread_group_empty(leader) |
| 192 | && leader->exit_state == EXIT_ZOMBIE) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 193 | /* |
| 194 | * If we were the last child thread and the leader has |
| 195 | * exited already, and the leader's parent ignores SIGCHLD, |
| 196 | * then we are the one who should release the leader. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 197 | */ |
Oleg Nesterov | 8677347 | 2011-06-22 23:09:09 +0200 | [diff] [blame] | 198 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
Roland McGrath | dae3357 | 2008-07-25 19:45:48 -0700 | [diff] [blame] | 199 | if (zap_leader) |
| 200 | leader->exit_state = EXIT_DEAD; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 201 | } |
| 202 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 203 | write_unlock_irq(&tasklist_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 204 | release_thread(p); |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 205 | call_rcu(&p->rcu, delayed_put_task_struct); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 206 | |
| 207 | p = leader; |
| 208 | if (unlikely(zap_leader)) |
| 209 | goto repeat; |
| 210 | } |
| 211 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 212 | /* |
Oleg Nesterov | 150593b | 2016-05-18 19:02:18 +0200 | [diff] [blame] | 213 | * Note that if this function returns a valid task_struct pointer (!NULL) |
| 214 | * task->usage must remain >0 for the duration of the RCU critical section. |
| 215 | */ |
| 216 | struct task_struct *task_rcu_dereference(struct task_struct **ptask) |
| 217 | { |
| 218 | struct sighand_struct *sighand; |
| 219 | struct task_struct *task; |
| 220 | |
| 221 | /* |
| 222 | * We need to verify that release_task() was not called and thus |
| 223 | * delayed_put_task_struct() can't run and drop the last reference |
| 224 | * before rcu_read_unlock(). We check task->sighand != NULL, |
| 225 | * but we can read the already freed and reused memory. |
| 226 | */ |
| 227 | retry: |
| 228 | task = rcu_dereference(*ptask); |
| 229 | if (!task) |
| 230 | return NULL; |
| 231 | |
| 232 | probe_kernel_address(&task->sighand, sighand); |
| 233 | |
| 234 | /* |
| 235 | * Pairs with atomic_dec_and_test() in put_task_struct(). If this task |
| 236 | * was already freed we can not miss the preceding update of this |
| 237 | * pointer. |
| 238 | */ |
| 239 | smp_rmb(); |
| 240 | if (unlikely(task != READ_ONCE(*ptask))) |
| 241 | goto retry; |
| 242 | |
| 243 | /* |
| 244 | * We've re-checked that "task == *ptask", now we have two different |
| 245 | * cases: |
| 246 | * |
| 247 | * 1. This is actually the same task/task_struct. In this case |
| 248 | * sighand != NULL tells us it is still alive. |
| 249 | * |
| 250 | * 2. This is another task which got the same memory for task_struct. |
| 251 | * We can't know this of course, and we can not trust |
| 252 | * sighand != NULL. |
| 253 | * |
| 254 | * In this case we actually return a random value, but this is |
| 255 | * correct. |
| 256 | * |
| 257 | * If we return NULL - we can pretend that we actually noticed that |
| 258 | * *ptask was updated when the previous task has exited. Or pretend |
| 259 | * that probe_slab_address(&sighand) reads NULL. |
| 260 | * |
| 261 | * If we return the new task (because sighand is not NULL for any |
| 262 | * reason) - this is fine too. This (new) task can't go away before |
| 263 | * another gp pass. |
| 264 | * |
| 265 | * And note: We could even eliminate the false positive if re-read |
| 266 | * task->sighand once again to avoid the falsely NULL. But this case |
| 267 | * is very unlikely so we don't care. |
| 268 | */ |
| 269 | if (!sighand) |
| 270 | return NULL; |
| 271 | |
| 272 | return task; |
| 273 | } |
| 274 | |
| 275 | struct task_struct *try_get_task_struct(struct task_struct **ptask) |
| 276 | { |
| 277 | struct task_struct *task; |
| 278 | |
| 279 | rcu_read_lock(); |
| 280 | task = task_rcu_dereference(ptask); |
| 281 | if (task) |
| 282 | get_task_struct(task); |
| 283 | rcu_read_unlock(); |
| 284 | |
| 285 | return task; |
| 286 | } |
| 287 | |
| 288 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 289 | * Determine if a process group is "orphaned", according to the POSIX |
| 290 | * definition in 2.2.2.52. Orphaned process groups are not to be affected |
| 291 | * by terminal-generated stop signals. Newly orphaned process groups are |
| 292 | * to receive a SIGHUP and a SIGCONT. |
| 293 | * |
| 294 | * "I ask you, have you ever known what it is to be an orphan?" |
| 295 | */ |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 296 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
| 297 | struct task_struct *ignored_task) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 298 | { |
| 299 | struct task_struct *p; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 300 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 301 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 302 | if ((p == ignored_task) || |
| 303 | (p->exit_state && thread_group_empty(p)) || |
| 304 | is_global_init(p->real_parent)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 305 | continue; |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 306 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 307 | if (task_pgrp(p->real_parent) != pgrp && |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 308 | task_session(p->real_parent) == task_session(p)) |
| 309 | return 0; |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 310 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 311 | |
| 312 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 313 | } |
| 314 | |
Eric W. Biederman | 3e7cd6c | 2007-02-12 00:52:58 -0800 | [diff] [blame] | 315 | int is_current_pgrp_orphaned(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 316 | { |
| 317 | int retval; |
| 318 | |
| 319 | read_lock(&tasklist_lock); |
Eric W. Biederman | 3e7cd6c | 2007-02-12 00:52:58 -0800 | [diff] [blame] | 320 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 321 | read_unlock(&tasklist_lock); |
| 322 | |
| 323 | return retval; |
| 324 | } |
| 325 | |
Oleg Nesterov | 961c467 | 2011-07-07 21:33:54 +0200 | [diff] [blame] | 326 | static bool has_stopped_jobs(struct pid *pgrp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 327 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 328 | struct task_struct *p; |
| 329 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 330 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Oleg Nesterov | 961c467 | 2011-07-07 21:33:54 +0200 | [diff] [blame] | 331 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
| 332 | return true; |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 333 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Oleg Nesterov | 961c467 | 2011-07-07 21:33:54 +0200 | [diff] [blame] | 334 | |
| 335 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 336 | } |
| 337 | |
Oleg Nesterov | f49ee50 | 2008-03-02 21:44:40 +0300 | [diff] [blame] | 338 | /* |
| 339 | * Check to see if any process groups have become orphaned as |
| 340 | * a result of our exiting, and if they have any stopped jobs, |
| 341 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
| 342 | */ |
| 343 | static void |
| 344 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) |
| 345 | { |
| 346 | struct pid *pgrp = task_pgrp(tsk); |
| 347 | struct task_struct *ignored_task = tsk; |
| 348 | |
| 349 | if (!parent) |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 350 | /* exit: our father is in a different pgrp than |
| 351 | * we are and we were the only connection outside. |
| 352 | */ |
Oleg Nesterov | f49ee50 | 2008-03-02 21:44:40 +0300 | [diff] [blame] | 353 | parent = tsk->real_parent; |
| 354 | else |
| 355 | /* reparent: our child is in a different pgrp than |
| 356 | * we are, and it was the only connection outside. |
| 357 | */ |
| 358 | ignored_task = NULL; |
| 359 | |
| 360 | if (task_pgrp(parent) != pgrp && |
| 361 | task_session(parent) == task_session(tsk) && |
| 362 | will_become_orphaned_pgrp(pgrp, ignored_task) && |
| 363 | has_stopped_jobs(pgrp)) { |
| 364 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); |
| 365 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); |
| 366 | } |
| 367 | } |
| 368 | |
Oleg Nesterov | f98bafa | 2014-06-04 16:07:34 -0700 | [diff] [blame] | 369 | #ifdef CONFIG_MEMCG |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 370 | /* |
KAMEZAWA Hiroyuki | 733eda7 | 2011-06-15 15:08:43 -0700 | [diff] [blame] | 371 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 372 | */ |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 373 | void mm_update_next_owner(struct mm_struct *mm) |
| 374 | { |
| 375 | struct task_struct *c, *g, *p = current; |
| 376 | |
| 377 | retry: |
KAMEZAWA Hiroyuki | 733eda7 | 2011-06-15 15:08:43 -0700 | [diff] [blame] | 378 | /* |
| 379 | * If the exiting or execing task is not the owner, it's |
| 380 | * someone else's problem. |
| 381 | */ |
| 382 | if (mm->owner != p) |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 383 | return; |
KAMEZAWA Hiroyuki | 733eda7 | 2011-06-15 15:08:43 -0700 | [diff] [blame] | 384 | /* |
| 385 | * The current owner is exiting/execing and there are no other |
| 386 | * candidates. Do not leave the mm pointing to a possibly |
| 387 | * freed task structure. |
| 388 | */ |
| 389 | if (atomic_read(&mm->mm_users) <= 1) { |
| 390 | mm->owner = NULL; |
| 391 | return; |
| 392 | } |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 393 | |
| 394 | read_lock(&tasklist_lock); |
| 395 | /* |
| 396 | * Search in the children |
| 397 | */ |
| 398 | list_for_each_entry(c, &p->children, sibling) { |
| 399 | if (c->mm == mm) |
| 400 | goto assign_new_owner; |
| 401 | } |
| 402 | |
| 403 | /* |
| 404 | * Search in the siblings |
| 405 | */ |
Oleg Nesterov | dea33cf | 2009-06-17 16:27:29 -0700 | [diff] [blame] | 406 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 407 | if (c->mm == mm) |
| 408 | goto assign_new_owner; |
| 409 | } |
| 410 | |
| 411 | /* |
Oleg Nesterov | f87fb59 | 2014-06-04 16:07:52 -0700 | [diff] [blame] | 412 | * Search through everything else, we should not get here often. |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 413 | */ |
Oleg Nesterov | 39af176 | 2014-06-04 16:07:54 -0700 | [diff] [blame] | 414 | for_each_process(g) { |
| 415 | if (g->flags & PF_KTHREAD) |
| 416 | continue; |
| 417 | for_each_thread(g, c) { |
| 418 | if (c->mm == mm) |
| 419 | goto assign_new_owner; |
| 420 | if (c->mm) |
| 421 | break; |
| 422 | } |
Oleg Nesterov | f87fb59 | 2014-06-04 16:07:52 -0700 | [diff] [blame] | 423 | } |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 424 | read_unlock(&tasklist_lock); |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 425 | /* |
| 426 | * We found no owner yet mm_users > 1: this implies that we are |
| 427 | * most likely racing with swapoff (try_to_unuse()) or /proc or |
Hugh Dickins | e599137 | 2009-01-06 14:39:22 -0800 | [diff] [blame] | 428 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 429 | */ |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 430 | mm->owner = NULL; |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 431 | return; |
| 432 | |
| 433 | assign_new_owner: |
| 434 | BUG_ON(c == p); |
| 435 | get_task_struct(c); |
| 436 | /* |
| 437 | * The task_lock protects c->mm from changing. |
| 438 | * We always want mm->owner->mm == mm |
| 439 | */ |
| 440 | task_lock(c); |
Hugh Dickins | e599137 | 2009-01-06 14:39:22 -0800 | [diff] [blame] | 441 | /* |
| 442 | * Delay read_unlock() till we have the task_lock() |
| 443 | * to ensure that c does not slip away underneath us |
| 444 | */ |
| 445 | read_unlock(&tasklist_lock); |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 446 | if (c->mm != mm) { |
| 447 | task_unlock(c); |
| 448 | put_task_struct(c); |
| 449 | goto retry; |
| 450 | } |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 451 | mm->owner = c; |
| 452 | task_unlock(c); |
| 453 | put_task_struct(c); |
| 454 | } |
Oleg Nesterov | f98bafa | 2014-06-04 16:07:34 -0700 | [diff] [blame] | 455 | #endif /* CONFIG_MEMCG */ |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 456 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 457 | /* |
| 458 | * Turn us into a lazy TLB process if we |
| 459 | * aren't already.. |
| 460 | */ |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 461 | static void exit_mm(struct task_struct *tsk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 462 | { |
| 463 | struct mm_struct *mm = tsk->mm; |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 464 | struct core_state *core_state; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 465 | |
Thomas Gleixner | 394ff12 | 2021-02-01 10:01:35 +0000 | [diff] [blame] | 466 | exit_mm_release(tsk, mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 467 | if (!mm) |
| 468 | return; |
Konstantin Khlebnikov | 4fe7efd | 2012-06-20 12:53:01 -0700 | [diff] [blame] | 469 | sync_mm_rss(mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 470 | /* |
| 471 | * Serialize with any possible pending coredump. |
Oleg Nesterov | 999d9fc | 2008-07-25 01:47:41 -0700 | [diff] [blame] | 472 | * We must hold mmap_sem around checking core_state |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 473 | * and clearing tsk->mm. The core-inducing thread |
Oleg Nesterov | 999d9fc | 2008-07-25 01:47:41 -0700 | [diff] [blame] | 474 | * will increment ->nr_threads for each thread in the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 475 | * group with ->mm != NULL. |
| 476 | */ |
| 477 | down_read(&mm->mmap_sem); |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 478 | core_state = mm->core_state; |
| 479 | if (core_state) { |
| 480 | struct core_thread self; |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 481 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 482 | up_read(&mm->mmap_sem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 483 | |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 484 | self.task = tsk; |
Al Viro | 951cb4f | 2020-10-28 16:39:49 -0400 | [diff] [blame] | 485 | if (self.task->flags & PF_SIGNALED) |
| 486 | self.next = xchg(&core_state->dumper.next, &self); |
| 487 | else |
| 488 | self.task = NULL; |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 489 | /* |
| 490 | * Implies mb(), the result of xchg() must be visible |
| 491 | * to core_state->dumper. |
| 492 | */ |
| 493 | if (atomic_dec_and_test(&core_state->nr_threads)) |
| 494 | complete(&core_state->startup); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 495 | |
Oleg Nesterov | a94e2d4 | 2008-07-25 01:47:46 -0700 | [diff] [blame] | 496 | for (;;) { |
| 497 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
| 498 | if (!self.task) /* see coredump_finish() */ |
| 499 | break; |
Mandeep Singh Baines | 80d26af | 2013-02-27 17:03:20 -0800 | [diff] [blame] | 500 | freezable_schedule(); |
Oleg Nesterov | a94e2d4 | 2008-07-25 01:47:46 -0700 | [diff] [blame] | 501 | } |
| 502 | __set_task_state(tsk, TASK_RUNNING); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 503 | down_read(&mm->mmap_sem); |
| 504 | } |
| 505 | atomic_inc(&mm->mm_count); |
Eric Sesterhenn | 125e187 | 2006-06-23 02:06:06 -0700 | [diff] [blame] | 506 | BUG_ON(mm != tsk->active_mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 507 | /* more a memory barrier than a real lock */ |
| 508 | task_lock(tsk); |
| 509 | tsk->mm = NULL; |
| 510 | up_read(&mm->mmap_sem); |
| 511 | enter_lazy_tlb(mm, current); |
| 512 | task_unlock(tsk); |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 513 | mm_update_next_owner(mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 514 | mmput(mm); |
Michal Hocko | c32b3cb | 2015-02-11 15:26:24 -0800 | [diff] [blame] | 515 | if (test_thread_flag(TIF_MEMDIE)) |
Tetsuo Handa | 3853120 | 2016-10-07 16:59:03 -0700 | [diff] [blame] | 516 | exit_oom_victim(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 517 | } |
| 518 | |
Oleg Nesterov | c9dc05b | 2014-12-10 15:55:14 -0800 | [diff] [blame] | 519 | static struct task_struct *find_alive_thread(struct task_struct *p) |
| 520 | { |
| 521 | struct task_struct *t; |
| 522 | |
| 523 | for_each_thread(p, t) { |
| 524 | if (!(t->flags & PF_EXITING)) |
| 525 | return t; |
| 526 | } |
| 527 | return NULL; |
| 528 | } |
| 529 | |
Andrei Vagin | 44ccc0c | 2019-02-01 14:20:24 -0800 | [diff] [blame] | 530 | static struct task_struct *find_child_reaper(struct task_struct *father, |
| 531 | struct list_head *dead) |
Oleg Nesterov | 1109909 | 2014-12-10 15:55:11 -0800 | [diff] [blame] | 532 | __releases(&tasklist_lock) |
| 533 | __acquires(&tasklist_lock) |
| 534 | { |
| 535 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
| 536 | struct task_struct *reaper = pid_ns->child_reaper; |
Andrei Vagin | 44ccc0c | 2019-02-01 14:20:24 -0800 | [diff] [blame] | 537 | struct task_struct *p, *n; |
Oleg Nesterov | 1109909 | 2014-12-10 15:55:11 -0800 | [diff] [blame] | 538 | |
| 539 | if (likely(reaper != father)) |
| 540 | return reaper; |
| 541 | |
Oleg Nesterov | c9dc05b | 2014-12-10 15:55:14 -0800 | [diff] [blame] | 542 | reaper = find_alive_thread(father); |
| 543 | if (reaper) { |
Oleg Nesterov | 1109909 | 2014-12-10 15:55:11 -0800 | [diff] [blame] | 544 | pid_ns->child_reaper = reaper; |
| 545 | return reaper; |
| 546 | } |
| 547 | |
| 548 | write_unlock_irq(&tasklist_lock); |
| 549 | if (unlikely(pid_ns == &init_pid_ns)) { |
| 550 | panic("Attempted to kill init! exitcode=0x%08x\n", |
| 551 | father->signal->group_exit_code ?: father->exit_code); |
| 552 | } |
Andrei Vagin | 44ccc0c | 2019-02-01 14:20:24 -0800 | [diff] [blame] | 553 | |
| 554 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { |
| 555 | list_del_init(&p->ptrace_entry); |
| 556 | release_task(p); |
| 557 | } |
| 558 | |
Oleg Nesterov | 1109909 | 2014-12-10 15:55:11 -0800 | [diff] [blame] | 559 | zap_pid_ns_processes(pid_ns); |
| 560 | write_lock_irq(&tasklist_lock); |
| 561 | |
| 562 | return father; |
| 563 | } |
| 564 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 565 | /* |
Lennart Poettering | ebec18a | 2012-03-23 15:01:54 -0700 | [diff] [blame] | 566 | * When we die, we re-parent all our children, and try to: |
| 567 | * 1. give them to another thread in our thread group, if such a member exists |
| 568 | * 2. give it to the first ancestor process which prctl'd itself as a |
| 569 | * child_subreaper for its children (like a service manager) |
| 570 | * 3. give it to the init process (PID 1) in our pid namespace |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 571 | */ |
Oleg Nesterov | 1109909 | 2014-12-10 15:55:11 -0800 | [diff] [blame] | 572 | static struct task_struct *find_new_reaper(struct task_struct *father, |
| 573 | struct task_struct *child_reaper) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 574 | { |
Oleg Nesterov | c9dc05b | 2014-12-10 15:55:14 -0800 | [diff] [blame] | 575 | struct task_struct *thread, *reaper; |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 576 | |
Oleg Nesterov | c9dc05b | 2014-12-10 15:55:14 -0800 | [diff] [blame] | 577 | thread = find_alive_thread(father); |
| 578 | if (thread) |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 579 | return thread; |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 580 | |
Oleg Nesterov | 7d24e2d | 2014-12-10 15:55:02 -0800 | [diff] [blame] | 581 | if (father->signal->has_child_subreaper) { |
Lennart Poettering | ebec18a | 2012-03-23 15:01:54 -0700 | [diff] [blame] | 582 | /* |
Oleg Nesterov | 175aed3 | 2014-12-10 15:55:08 -0800 | [diff] [blame] | 583 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
| 584 | * We start from father to ensure we can not look into another |
| 585 | * namespace, this is safe because all its threads are dead. |
Lennart Poettering | ebec18a | 2012-03-23 15:01:54 -0700 | [diff] [blame] | 586 | */ |
Oleg Nesterov | 7d24e2d | 2014-12-10 15:55:02 -0800 | [diff] [blame] | 587 | for (reaper = father; |
Oleg Nesterov | 1109909 | 2014-12-10 15:55:11 -0800 | [diff] [blame] | 588 | !same_thread_group(reaper, child_reaper); |
Lennart Poettering | ebec18a | 2012-03-23 15:01:54 -0700 | [diff] [blame] | 589 | reaper = reaper->real_parent) { |
Oleg Nesterov | 175aed3 | 2014-12-10 15:55:08 -0800 | [diff] [blame] | 590 | /* call_usermodehelper() descendants need this check */ |
| 591 | if (reaper == &init_task) |
Lennart Poettering | ebec18a | 2012-03-23 15:01:54 -0700 | [diff] [blame] | 592 | break; |
| 593 | if (!reaper->signal->is_child_subreaper) |
| 594 | continue; |
Oleg Nesterov | c9dc05b | 2014-12-10 15:55:14 -0800 | [diff] [blame] | 595 | thread = find_alive_thread(reaper); |
| 596 | if (thread) |
| 597 | return thread; |
Lennart Poettering | ebec18a | 2012-03-23 15:01:54 -0700 | [diff] [blame] | 598 | } |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 599 | } |
| 600 | |
Oleg Nesterov | 1109909 | 2014-12-10 15:55:11 -0800 | [diff] [blame] | 601 | return child_reaper; |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 602 | } |
| 603 | |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 604 | /* |
| 605 | * Any that need to be release_task'd are put on the @dead list. |
| 606 | */ |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 607 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 608 | struct list_head *dead) |
| 609 | { |
Oleg Nesterov | 2831096 | 2014-12-10 15:45:30 -0800 | [diff] [blame] | 610 | if (unlikely(p->exit_state == EXIT_DEAD)) |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 611 | return; |
| 612 | |
Oleg Nesterov | abd50b3 | 2014-04-07 15:38:42 -0700 | [diff] [blame] | 613 | /* We don't want people slaying init. */ |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 614 | p->exit_signal = SIGCHLD; |
| 615 | |
| 616 | /* If it has exited notify the new parent about this child's death. */ |
Tejun Heo | d21142e | 2011-06-17 16:50:34 +0200 | [diff] [blame] | 617 | if (!p->ptrace && |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 618 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
Oleg Nesterov | 8677347 | 2011-06-22 23:09:09 +0200 | [diff] [blame] | 619 | if (do_notify_parent(p, p->exit_signal)) { |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 620 | p->exit_state = EXIT_DEAD; |
Oleg Nesterov | dc2fd4b | 2014-12-10 15:45:24 -0800 | [diff] [blame] | 621 | list_add(&p->ptrace_entry, dead); |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 622 | } |
| 623 | } |
| 624 | |
| 625 | kill_orphaned_pgrp(p, father); |
| 626 | } |
| 627 | |
Oleg Nesterov | 482a376 | 2014-12-10 15:55:20 -0800 | [diff] [blame] | 628 | /* |
| 629 | * This does two things: |
| 630 | * |
| 631 | * A. Make init inherit all the child processes |
| 632 | * B. Check to see if any process groups have become orphaned |
| 633 | * as a result of our exiting, and if they have any stopped |
| 634 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
| 635 | */ |
| 636 | static void forget_original_parent(struct task_struct *father, |
| 637 | struct list_head *dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 638 | { |
Oleg Nesterov | 482a376 | 2014-12-10 15:55:20 -0800 | [diff] [blame] | 639 | struct task_struct *p, *t, *reaper; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 640 | |
Oleg Nesterov | 7c8bd23 | 2014-12-10 15:45:33 -0800 | [diff] [blame] | 641 | if (unlikely(!list_empty(&father->ptraced))) |
Oleg Nesterov | 482a376 | 2014-12-10 15:55:20 -0800 | [diff] [blame] | 642 | exit_ptrace(father, dead); |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 643 | |
Oleg Nesterov | 7c8bd23 | 2014-12-10 15:45:33 -0800 | [diff] [blame] | 644 | /* Can drop and reacquire tasklist_lock */ |
Andrei Vagin | 44ccc0c | 2019-02-01 14:20:24 -0800 | [diff] [blame] | 645 | reaper = find_child_reaper(father, dead); |
Oleg Nesterov | ad9e206 | 2014-12-10 15:55:17 -0800 | [diff] [blame] | 646 | if (list_empty(&father->children)) |
Oleg Nesterov | 482a376 | 2014-12-10 15:55:20 -0800 | [diff] [blame] | 647 | return; |
Oleg Nesterov | 1109909 | 2014-12-10 15:55:11 -0800 | [diff] [blame] | 648 | |
| 649 | reaper = find_new_reaper(father, reaper); |
Oleg Nesterov | 2831096 | 2014-12-10 15:45:30 -0800 | [diff] [blame] | 650 | list_for_each_entry(p, &father->children, sibling) { |
Oleg Nesterov | 57a0591 | 2014-12-10 15:45:27 -0800 | [diff] [blame] | 651 | for_each_thread(p, t) { |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 652 | t->real_parent = reaper; |
Oleg Nesterov | 57a0591 | 2014-12-10 15:45:27 -0800 | [diff] [blame] | 653 | BUG_ON((!t->ptrace) != (t->parent == father)); |
| 654 | if (likely(!t->ptrace)) |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 655 | t->parent = t->real_parent; |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 656 | if (t->pdeath_signal) |
| 657 | group_send_sig_info(t->pdeath_signal, |
| 658 | SEND_SIG_NOINFO, t); |
Oleg Nesterov | 57a0591 | 2014-12-10 15:45:27 -0800 | [diff] [blame] | 659 | } |
Oleg Nesterov | 2831096 | 2014-12-10 15:45:30 -0800 | [diff] [blame] | 660 | /* |
| 661 | * If this is a threaded reparent there is no need to |
| 662 | * notify anyone anything has happened. |
| 663 | */ |
| 664 | if (!same_thread_group(reaper, father)) |
Oleg Nesterov | 482a376 | 2014-12-10 15:55:20 -0800 | [diff] [blame] | 665 | reparent_leader(father, p, dead); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 666 | } |
Oleg Nesterov | 2831096 | 2014-12-10 15:45:30 -0800 | [diff] [blame] | 667 | list_splice_tail_init(&father->children, &reaper->children); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 668 | } |
| 669 | |
| 670 | /* |
| 671 | * Send signals to all our closest relatives so that they know |
| 672 | * to properly mourn us.. |
| 673 | */ |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 674 | static void exit_notify(struct task_struct *tsk, int group_dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 675 | { |
Oleg Nesterov | 53c8f9f | 2011-06-22 23:08:18 +0200 | [diff] [blame] | 676 | bool autoreap; |
Oleg Nesterov | 482a376 | 2014-12-10 15:55:20 -0800 | [diff] [blame] | 677 | struct task_struct *p, *n; |
| 678 | LIST_HEAD(dead); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 679 | |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 680 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 482a376 | 2014-12-10 15:55:20 -0800 | [diff] [blame] | 681 | forget_original_parent(tsk, &dead); |
| 682 | |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 683 | if (group_dead) |
| 684 | kill_orphaned_pgrp(tsk->group_leader, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 685 | |
Oleg Nesterov | 45cdf5c | 2011-06-23 19:06:50 +0200 | [diff] [blame] | 686 | if (unlikely(tsk->ptrace)) { |
| 687 | int sig = thread_group_leader(tsk) && |
| 688 | thread_group_empty(tsk) && |
| 689 | !ptrace_reparented(tsk) ? |
| 690 | tsk->exit_signal : SIGCHLD; |
| 691 | autoreap = do_notify_parent(tsk, sig); |
| 692 | } else if (thread_group_leader(tsk)) { |
| 693 | autoreap = thread_group_empty(tsk) && |
| 694 | do_notify_parent(tsk, tsk->exit_signal); |
| 695 | } else { |
| 696 | autoreap = true; |
| 697 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 698 | |
Oleg Nesterov | 53c8f9f | 2011-06-22 23:08:18 +0200 | [diff] [blame] | 699 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
Oleg Nesterov | 6c66e7d | 2014-12-10 15:55:23 -0800 | [diff] [blame] | 700 | if (tsk->exit_state == EXIT_DEAD) |
| 701 | list_add(&tsk->ptrace_entry, &dead); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 702 | |
Oleg Nesterov | 9c33916 | 2010-05-26 14:43:10 -0700 | [diff] [blame] | 703 | /* mt-exec, de_thread() is waiting for group leader */ |
| 704 | if (unlikely(tsk->signal->notify_count < 0)) |
Oleg Nesterov | 6db840f | 2007-10-16 23:27:23 -0700 | [diff] [blame] | 705 | wake_up_process(tsk->signal->group_exit_task); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 706 | write_unlock_irq(&tasklist_lock); |
| 707 | |
Oleg Nesterov | 482a376 | 2014-12-10 15:55:20 -0800 | [diff] [blame] | 708 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
| 709 | list_del_init(&p->ptrace_entry); |
| 710 | release_task(p); |
| 711 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 712 | } |
| 713 | |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 714 | #ifdef CONFIG_DEBUG_STACK_USAGE |
| 715 | static void check_stack_usage(void) |
| 716 | { |
| 717 | static DEFINE_SPINLOCK(low_water_lock); |
| 718 | static int lowest_to_date = THREAD_SIZE; |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 719 | unsigned long free; |
| 720 | |
Eric Sandeen | 7c9f886 | 2008-04-22 16:38:23 -0500 | [diff] [blame] | 721 | free = stack_not_used(current); |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 722 | |
| 723 | if (free >= lowest_to_date) |
| 724 | return; |
| 725 | |
| 726 | spin_lock(&low_water_lock); |
| 727 | if (free < lowest_to_date) { |
Anton Blanchard | 627393d | 2016-08-02 14:05:40 -0700 | [diff] [blame] | 728 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 729 | current->comm, task_pid_nr(current), free); |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 730 | lowest_to_date = free; |
| 731 | } |
| 732 | spin_unlock(&low_water_lock); |
| 733 | } |
| 734 | #else |
| 735 | static inline void check_stack_usage(void) {} |
| 736 | #endif |
| 737 | |
Peter Zijlstra | 9af6528 | 2016-09-13 18:37:29 +0200 | [diff] [blame] | 738 | void __noreturn do_exit(long code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 739 | { |
| 740 | struct task_struct *tsk = current; |
| 741 | int group_dead; |
Paul E. McKenney | 3f95aa8 | 2014-08-04 06:10:23 -0700 | [diff] [blame] | 742 | TASKS_RCU(int tasks_rcu_i); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 743 | |
Jann Horn | 1e587ce | 2020-03-05 23:06:57 +0100 | [diff] [blame] | 744 | /* |
| 745 | * We can get here from a kernel oops, sometimes with preemption off. |
| 746 | * Start by checking for critical errors. |
| 747 | * Then fix up important state like USER_DS and preemption. |
| 748 | * Then do everything else. |
| 749 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 750 | |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 751 | WARN_ON(blk_needs_flush_plug(tsk)); |
Jens Axboe | 22e2c50 | 2005-06-27 10:55:12 +0200 | [diff] [blame] | 752 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 753 | if (unlikely(in_interrupt())) |
| 754 | panic("Aiee, killing interrupt handler!"); |
| 755 | if (unlikely(!tsk->pid)) |
| 756 | panic("Attempted to kill the idle task!"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 757 | |
Nelson Elhage | 33dd94a | 2010-12-02 14:31:21 -0800 | [diff] [blame] | 758 | /* |
| 759 | * If do_exit is called because this processes oopsed, it's possible |
| 760 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before |
| 761 | * continuing. Amongst other possible reasons, this is to prevent |
| 762 | * mm_release()->clear_child_tid() from writing to a user-controlled |
| 763 | * kernel address. |
| 764 | */ |
| 765 | set_fs(USER_DS); |
| 766 | |
Jann Horn | 1e587ce | 2020-03-05 23:06:57 +0100 | [diff] [blame] | 767 | if (unlikely(in_atomic())) { |
| 768 | pr_info("note: %s[%d] exited with preempt_count %d\n", |
| 769 | current->comm, task_pid_nr(current), |
| 770 | preempt_count()); |
| 771 | preempt_count_set(PREEMPT_ENABLED); |
| 772 | } |
| 773 | |
| 774 | profile_task_exit(tsk); |
| 775 | kcov_task_exit(tsk); |
| 776 | |
Tejun Heo | a288eec | 2011-06-17 16:50:37 +0200 | [diff] [blame] | 777 | ptrace_event(PTRACE_EVENT_EXIT, code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 778 | |
David Howells | e0e8173 | 2009-09-02 09:13:40 +0100 | [diff] [blame] | 779 | validate_creds_for_do_exit(tsk); |
| 780 | |
Alexander Nyberg | df164db | 2005-06-23 00:09:13 -0700 | [diff] [blame] | 781 | /* |
| 782 | * We're taking recursive faults here in do_exit. Safest is to just |
| 783 | * leave this task alone and wait for reboot. |
| 784 | */ |
| 785 | if (unlikely(tsk->flags & PF_EXITING)) { |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 786 | pr_alert("Fixing recursive fault but reboot is needed!\n"); |
Thomas Gleixner | 32d7828 | 2021-02-01 10:01:38 +0000 | [diff] [blame] | 787 | futex_exit_recursive(tsk); |
Alexander Nyberg | df164db | 2005-06-23 00:09:13 -0700 | [diff] [blame] | 788 | set_current_state(TASK_UNINTERRUPTIBLE); |
| 789 | schedule(); |
| 790 | } |
| 791 | |
Oleg Nesterov | d12619b | 2008-02-08 04:19:12 -0800 | [diff] [blame] | 792 | exit_signals(tsk); /* sets PF_EXITING */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 793 | |
Linus Torvalds | 48d212a | 2012-06-07 17:54:07 -0700 | [diff] [blame] | 794 | /* sync mm's RSS info before statistics gathering */ |
| 795 | if (tsk->mm) |
| 796 | sync_mm_rss(tsk->mm); |
Rik van Riel | 51229b4 | 2015-06-25 15:03:56 -0700 | [diff] [blame] | 797 | acct_update_integrals(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 798 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
Andrew Morton | c306895 | 2005-08-04 16:49:32 -0700 | [diff] [blame] | 799 | if (group_dead) { |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 800 | hrtimer_cancel(&tsk->signal->real_timer); |
Roland McGrath | 25f407f | 2005-10-21 15:03:29 -0700 | [diff] [blame] | 801 | exit_itimers(tsk->signal); |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 802 | if (tsk->mm) |
| 803 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); |
Andrew Morton | c306895 | 2005-08-04 16:49:32 -0700 | [diff] [blame] | 804 | } |
KaiGai Kohei | f6ec29a | 2006-06-25 05:49:25 -0700 | [diff] [blame] | 805 | acct_collect(code, group_dead); |
Miloslav Trmac | 522ed77 | 2007-07-15 23:40:56 -0700 | [diff] [blame] | 806 | if (group_dead) |
| 807 | tty_audit_exit(); |
Eric Paris | a4ff8db | 2012-01-03 14:23:07 -0500 | [diff] [blame] | 808 | audit_free(tsk); |
Oleg Nesterov | 115085e | 2006-12-06 20:36:51 -0800 | [diff] [blame] | 809 | |
Linus Torvalds | 48d212a | 2012-06-07 17:54:07 -0700 | [diff] [blame] | 810 | tsk->exit_code = code; |
Oleg Nesterov | 115085e | 2006-12-06 20:36:51 -0800 | [diff] [blame] | 811 | taskstats_exit(tsk, group_dead); |
Shailabh Nagar | c757249 | 2006-07-14 00:24:40 -0700 | [diff] [blame] | 812 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 813 | exit_mm(tsk); |
| 814 | |
KaiGai Kohei | 0e46481 | 2006-06-25 05:49:24 -0700 | [diff] [blame] | 815 | if (group_dead) |
KaiGai Kohei | f6ec29a | 2006-06-25 05:49:25 -0700 | [diff] [blame] | 816 | acct_process(); |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 817 | trace_sched_process_exit(tsk); |
| 818 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 819 | exit_sem(tsk); |
Vasiliy Kulikov | b34a6b1 | 2011-07-26 16:08:48 -0700 | [diff] [blame] | 820 | exit_shm(tsk); |
Al Viro | 1ec7f1d | 2008-04-22 05:35:42 -0400 | [diff] [blame] | 821 | exit_files(tsk); |
| 822 | exit_fs(tsk); |
Oleg Nesterov | c39df5f | 2014-04-07 15:38:29 -0700 | [diff] [blame] | 823 | if (group_dead) |
| 824 | disassociate_ctty(1); |
Oleg Nesterov | 8aac627 | 2013-06-14 21:09:49 +0200 | [diff] [blame] | 825 | exit_task_namespaces(tsk); |
Al Viro | ed3e694 | 2012-06-27 11:31:24 +0400 | [diff] [blame] | 826 | exit_task_work(tsk); |
Jiri Slaby | e646469 | 2016-05-20 17:00:20 -0700 | [diff] [blame] | 827 | exit_thread(tsk); |
Stephane Eranian | 0b3fcf1 | 2011-01-03 18:20:01 +0200 | [diff] [blame] | 828 | |
| 829 | /* |
| 830 | * Flush inherited counters to the parent - before the parent |
| 831 | * gets woken up by child-exit notifications. |
| 832 | * |
| 833 | * because of cgroup mode, must be called before cgroup_exit() |
| 834 | */ |
| 835 | perf_event_exit_task(tsk); |
| 836 | |
Oleg Nesterov | 8e5bfa8 | 2016-11-14 19:46:12 +0100 | [diff] [blame] | 837 | sched_autogroup_exit_task(tsk); |
Li Zefan | 1ec4183 | 2014-03-28 15:22:19 +0800 | [diff] [blame] | 838 | cgroup_exit(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 839 | |
Ingo Molnar | 33b2fb3 | 2009-05-17 11:08:41 +0200 | [diff] [blame] | 840 | /* |
Frederic Weisbecker | 24f1e32c | 2009-09-09 19:22:48 +0200 | [diff] [blame] | 841 | * FIXME: do that only when needed, using sched_exit tracepoint |
| 842 | */ |
Oleg Nesterov | 7c8df28 | 2013-07-08 16:00:54 -0700 | [diff] [blame] | 843 | flush_ptrace_hw_breakpoint(tsk); |
Ingo Molnar | 33b2fb3 | 2009-05-17 11:08:41 +0200 | [diff] [blame] | 844 | |
Paul E. McKenney | 49f5903 | 2015-09-01 00:42:57 -0700 | [diff] [blame] | 845 | TASKS_RCU(preempt_disable()); |
Paul E. McKenney | 3f95aa8 | 2014-08-04 06:10:23 -0700 | [diff] [blame] | 846 | TASKS_RCU(tasks_rcu_i = __srcu_read_lock(&tasks_rcu_exit_srcu)); |
Paul E. McKenney | 49f5903 | 2015-09-01 00:42:57 -0700 | [diff] [blame] | 847 | TASKS_RCU(preempt_enable()); |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 848 | exit_notify(tsk, group_dead); |
Guillaume Morin | ef98239 | 2014-04-07 15:38:31 -0700 | [diff] [blame] | 849 | proc_exit_connector(tsk); |
David Rientjes | c11600e | 2016-09-01 16:15:07 -0700 | [diff] [blame] | 850 | mpol_put_task_policy(tsk); |
Alexey Dobriyan | 42b2dd0 | 2007-10-16 23:27:30 -0700 | [diff] [blame] | 851 | #ifdef CONFIG_FUTEX |
Ingo Molnar | c87e283 | 2006-06-27 02:54:58 -0700 | [diff] [blame] | 852 | if (unlikely(current->pi_state_cache)) |
| 853 | kfree(current->pi_state_cache); |
Alexey Dobriyan | 42b2dd0 | 2007-10-16 23:27:30 -0700 | [diff] [blame] | 854 | #endif |
Ingo Molnar | c87e283 | 2006-06-27 02:54:58 -0700 | [diff] [blame] | 855 | /* |
Ingo Molnar | 9a11b49a | 2006-07-03 00:24:33 -0700 | [diff] [blame] | 856 | * Make sure we are holding no locks: |
Ingo Molnar | de5097c | 2006-01-09 15:59:21 -0800 | [diff] [blame] | 857 | */ |
Colin Cross | 1b1d2fb | 2013-05-06 23:50:08 +0000 | [diff] [blame] | 858 | debug_check_no_locks_held(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 859 | |
Al Viro | afc847b | 2006-02-28 12:51:55 -0500 | [diff] [blame] | 860 | if (tsk->io_context) |
Louis Rilling | b69f229 | 2009-12-04 14:52:42 +0100 | [diff] [blame] | 861 | exit_io_context(tsk); |
Al Viro | afc847b | 2006-02-28 12:51:55 -0500 | [diff] [blame] | 862 | |
Jens Axboe | b92ce55 | 2006-04-11 13:52:07 +0200 | [diff] [blame] | 863 | if (tsk->splice_pipe) |
Al Viro | 4b8a8f1 | 2013-03-21 11:06:46 -0400 | [diff] [blame] | 864 | free_pipe_info(tsk->splice_pipe); |
Jens Axboe | b92ce55 | 2006-04-11 13:52:07 +0200 | [diff] [blame] | 865 | |
Eric Dumazet | 5640f76 | 2012-09-23 23:04:42 +0000 | [diff] [blame] | 866 | if (tsk->task_frag.page) |
| 867 | put_page(tsk->task_frag.page); |
| 868 | |
David Howells | e0e8173 | 2009-09-02 09:13:40 +0100 | [diff] [blame] | 869 | validate_creds_for_do_exit(tsk); |
| 870 | |
Oleg Nesterov | 4bcb823 | 2014-04-07 15:38:30 -0700 | [diff] [blame] | 871 | check_stack_usage(); |
Coywolf Qi Hunt | 7407251 | 2005-10-30 15:02:47 -0800 | [diff] [blame] | 872 | preempt_disable(); |
Wu Fengguang | 54848d7 | 2011-04-05 13:21:19 -0600 | [diff] [blame] | 873 | if (tsk->nr_dirtied) |
| 874 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); |
Paul E. McKenney | f41d911 | 2009-08-22 13:56:52 -0700 | [diff] [blame] | 875 | exit_rcu(); |
Paul E. McKenney | 3f95aa8 | 2014-08-04 06:10:23 -0700 | [diff] [blame] | 876 | TASKS_RCU(__srcu_read_unlock(&tasks_rcu_exit_srcu, tasks_rcu_i)); |
Yasunori Goto | b5740f4 | 2012-01-17 17:40:31 +0900 | [diff] [blame] | 877 | |
Peter Zijlstra | 9af6528 | 2016-09-13 18:37:29 +0200 | [diff] [blame] | 878 | do_task_dead(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 879 | } |
Russ Anderson | 012914d | 2005-04-23 00:08:00 -0700 | [diff] [blame] | 880 | EXPORT_SYMBOL_GPL(do_exit); |
| 881 | |
Joe Perches | 9402c95 | 2012-01-12 17:17:17 -0800 | [diff] [blame] | 882 | void complete_and_exit(struct completion *comp, long code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 883 | { |
| 884 | if (comp) |
| 885 | complete(comp); |
Oleg Nesterov | 55a101f | 2006-09-29 02:01:10 -0700 | [diff] [blame] | 886 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 887 | do_exit(code); |
| 888 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 889 | EXPORT_SYMBOL(complete_and_exit); |
| 890 | |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 891 | SYSCALL_DEFINE1(exit, int, error_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 892 | { |
| 893 | do_exit((error_code&0xff)<<8); |
| 894 | } |
| 895 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 896 | /* |
| 897 | * Take down every thread in the group. This is called by fatal signals |
| 898 | * as well as by sys_exit_group (below). |
| 899 | */ |
Joe Perches | 9402c95 | 2012-01-12 17:17:17 -0800 | [diff] [blame] | 900 | void |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 901 | do_group_exit(int exit_code) |
| 902 | { |
Oleg Nesterov | bfc4b08 | 2008-04-30 00:52:36 -0700 | [diff] [blame] | 903 | struct signal_struct *sig = current->signal; |
| 904 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 905 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
| 906 | |
Oleg Nesterov | bfc4b08 | 2008-04-30 00:52:36 -0700 | [diff] [blame] | 907 | if (signal_group_exit(sig)) |
| 908 | exit_code = sig->group_exit_code; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 909 | else if (!thread_group_empty(current)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 910 | struct sighand_struct *const sighand = current->sighand; |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 911 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 912 | spin_lock_irq(&sighand->siglock); |
Oleg Nesterov | ed5d2ca | 2008-02-04 22:27:24 -0800 | [diff] [blame] | 913 | if (signal_group_exit(sig)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 914 | /* Another thread got here before we took the lock. */ |
| 915 | exit_code = sig->group_exit_code; |
| 916 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 917 | sig->group_exit_code = exit_code; |
Oleg Nesterov | ed5d2ca | 2008-02-04 22:27:24 -0800 | [diff] [blame] | 918 | sig->flags = SIGNAL_GROUP_EXIT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 919 | zap_other_threads(current); |
| 920 | } |
| 921 | spin_unlock_irq(&sighand->siglock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 922 | } |
| 923 | |
| 924 | do_exit(exit_code); |
| 925 | /* NOTREACHED */ |
| 926 | } |
| 927 | |
| 928 | /* |
| 929 | * this kills every thread in the thread group. Note that any externally |
| 930 | * wait4()-ing process will get the correct exit code - even if this |
| 931 | * thread is not the thread group leader. |
| 932 | */ |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 933 | SYSCALL_DEFINE1(exit_group, int, error_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 934 | { |
| 935 | do_group_exit((error_code & 0xff) << 8); |
Heiko Carstens | 2ed7c03 | 2009-01-14 14:13:54 +0100 | [diff] [blame] | 936 | /* NOTREACHED */ |
| 937 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 938 | } |
| 939 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 940 | struct wait_opts { |
| 941 | enum pid_type wo_type; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 942 | int wo_flags; |
Richard Kennedy | e1eb1eb | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 943 | struct pid *wo_pid; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 944 | |
| 945 | struct siginfo __user *wo_info; |
| 946 | int __user *wo_stat; |
| 947 | struct rusage __user *wo_rusage; |
| 948 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 949 | wait_queue_t child_wait; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 950 | int notask_error; |
| 951 | }; |
| 952 | |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 953 | static inline |
| 954 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 955 | { |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 956 | if (type != PIDTYPE_PID) |
| 957 | task = task->group_leader; |
| 958 | return task->pids[type].pid; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 959 | } |
| 960 | |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 961 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 962 | { |
| 963 | return wo->wo_type == PIDTYPE_MAX || |
| 964 | task_pid_type(p, wo->wo_type) == wo->wo_pid; |
| 965 | } |
| 966 | |
Oleg Nesterov | bf95993 | 2016-05-23 16:23:50 -0700 | [diff] [blame] | 967 | static int |
| 968 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 969 | { |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 970 | if (!eligible_pid(wo, p)) |
| 971 | return 0; |
Oleg Nesterov | bf95993 | 2016-05-23 16:23:50 -0700 | [diff] [blame] | 972 | |
| 973 | /* |
| 974 | * Wait for all children (clone and not) if __WALL is set or |
| 975 | * if it is traced by us. |
| 976 | */ |
| 977 | if (ptrace || (wo->wo_flags & __WALL)) |
| 978 | return 1; |
| 979 | |
| 980 | /* |
| 981 | * Otherwise, wait for clone children *only* if __WCLONE is set; |
| 982 | * otherwise, wait for non-clone children *only*. |
| 983 | * |
| 984 | * Note: a "clone" child here is one that reports to its parent |
| 985 | * using a signal other than SIGCHLD, or a non-leader thread which |
| 986 | * we can only see if it is traced by us. |
| 987 | */ |
| 988 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 989 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 990 | |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 991 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 992 | } |
| 993 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 994 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
| 995 | pid_t pid, uid_t uid, int why, int status) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 996 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 997 | struct siginfo __user *infop; |
| 998 | int retval = wo->wo_rusage |
| 999 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Ingo Molnar | 36c8b58 | 2006-07-03 00:25:41 -0700 | [diff] [blame] | 1000 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1001 | put_task_struct(p); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1002 | infop = wo->wo_info; |
Vitaly Mayatskikh | b6fe2d1 | 2009-09-23 15:56:52 -0700 | [diff] [blame] | 1003 | if (infop) { |
| 1004 | if (!retval) |
| 1005 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1006 | if (!retval) |
| 1007 | retval = put_user(0, &infop->si_errno); |
| 1008 | if (!retval) |
| 1009 | retval = put_user((short)why, &infop->si_code); |
| 1010 | if (!retval) |
| 1011 | retval = put_user(pid, &infop->si_pid); |
| 1012 | if (!retval) |
| 1013 | retval = put_user(uid, &infop->si_uid); |
| 1014 | if (!retval) |
| 1015 | retval = put_user(status, &infop->si_status); |
| 1016 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1017 | if (!retval) |
| 1018 | retval = pid; |
| 1019 | return retval; |
| 1020 | } |
| 1021 | |
| 1022 | /* |
| 1023 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold |
| 1024 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
| 1025 | * the lock and this task is uninteresting. If we return nonzero, we have |
| 1026 | * released the lock and the system call should return. |
| 1027 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1028 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1029 | { |
Oleg Nesterov | f6507f8 | 2014-12-10 15:54:45 -0800 | [diff] [blame] | 1030 | int state, retval, status; |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1031 | pid_t pid = task_pid_vnr(p); |
Oleg Nesterov | 43e13cc | 2012-05-31 16:26:16 -0700 | [diff] [blame] | 1032 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1033 | struct siginfo __user *infop; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1034 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1035 | if (!likely(wo->wo_flags & WEXITED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1036 | return 0; |
| 1037 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1038 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1039 | int exit_code = p->exit_code; |
Thiago Farina | f3abd4f | 2010-03-05 13:42:52 -0800 | [diff] [blame] | 1040 | int why; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1041 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1042 | get_task_struct(p); |
| 1043 | read_unlock(&tasklist_lock); |
Peter Zijlstra | 1029a2b | 2014-09-24 10:18:49 +0200 | [diff] [blame] | 1044 | sched_annotate_sleep(); |
| 1045 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1046 | if ((exit_code & 0x7f) == 0) { |
| 1047 | why = CLD_EXITED; |
| 1048 | status = exit_code >> 8; |
| 1049 | } else { |
| 1050 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; |
| 1051 | status = exit_code & 0x7f; |
| 1052 | } |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1053 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1054 | } |
Oleg Nesterov | befca96 | 2009-06-18 16:49:11 -0700 | [diff] [blame] | 1055 | /* |
Oleg Nesterov | abd50b3 | 2014-04-07 15:38:42 -0700 | [diff] [blame] | 1056 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
| 1057 | */ |
Oleg Nesterov | f6507f8 | 2014-12-10 15:54:45 -0800 | [diff] [blame] | 1058 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
| 1059 | EXIT_TRACE : EXIT_DEAD; |
Oleg Nesterov | abd50b3 | 2014-04-07 15:38:42 -0700 | [diff] [blame] | 1060 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
| 1061 | return 0; |
Oleg Nesterov | 986094d | 2014-12-10 15:54:51 -0800 | [diff] [blame] | 1062 | /* |
| 1063 | * We own this thread, nobody else can reap it. |
| 1064 | */ |
| 1065 | read_unlock(&tasklist_lock); |
| 1066 | sched_annotate_sleep(); |
Oleg Nesterov | f6507f8 | 2014-12-10 15:54:45 -0800 | [diff] [blame] | 1067 | |
Oleg Nesterov | abd50b3 | 2014-04-07 15:38:42 -0700 | [diff] [blame] | 1068 | /* |
Oleg Nesterov | f6507f8 | 2014-12-10 15:54:45 -0800 | [diff] [blame] | 1069 | * Check thread_group_leader() to exclude the traced sub-threads. |
Oleg Nesterov | befca96 | 2009-06-18 16:49:11 -0700 | [diff] [blame] | 1070 | */ |
Oleg Nesterov | f6507f8 | 2014-12-10 15:54:45 -0800 | [diff] [blame] | 1071 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
Oleg Nesterov | f953ccd | 2014-12-10 15:54:48 -0800 | [diff] [blame] | 1072 | struct signal_struct *sig = p->signal; |
| 1073 | struct signal_struct *psig = current->signal; |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 1074 | unsigned long maxrss; |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1075 | cputime_t tgutime, tgstime; |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1076 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1077 | /* |
| 1078 | * The resource counters for the group leader are in its |
| 1079 | * own task_struct. Those for dead threads in the group |
| 1080 | * are in its signal_struct, as are those for the child |
| 1081 | * processes it has previously reaped. All these |
| 1082 | * accumulate in the parent's signal_struct c* fields. |
| 1083 | * |
| 1084 | * We don't bother to take a lock here to protect these |
Oleg Nesterov | f953ccd | 2014-12-10 15:54:48 -0800 | [diff] [blame] | 1085 | * p->signal fields because the whole thread group is dead |
| 1086 | * and nobody can change them. |
| 1087 | * |
| 1088 | * psig->stats_lock also protects us from our sub-theads |
| 1089 | * which can reap other children at the same time. Until |
| 1090 | * we change k_getrusage()-like users to rely on this lock |
| 1091 | * we have to take ->siglock as well. |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1092 | * |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 1093 | * We use thread_group_cputime_adjusted() to get times for |
| 1094 | * the thread group, which consolidates times for all threads |
| 1095 | * in the group including the group leader. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1096 | */ |
Frederic Weisbecker | e80d0a1a | 2012-11-21 16:26:44 +0100 | [diff] [blame] | 1097 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
Oleg Nesterov | f953ccd | 2014-12-10 15:54:48 -0800 | [diff] [blame] | 1098 | spin_lock_irq(¤t->sighand->siglock); |
Rik van Riel | e78c349 | 2014-08-16 13:40:10 -0400 | [diff] [blame] | 1099 | write_seqlock(&psig->stats_lock); |
Martin Schwidefsky | 6486163 | 2011-12-15 14:56:09 +0100 | [diff] [blame] | 1100 | psig->cutime += tgutime + sig->cutime; |
| 1101 | psig->cstime += tgstime + sig->cstime; |
Frederic Weisbecker | 6fac482 | 2012-11-13 14:20:55 +0100 | [diff] [blame] | 1102 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1103 | psig->cmin_flt += |
| 1104 | p->min_flt + sig->min_flt + sig->cmin_flt; |
| 1105 | psig->cmaj_flt += |
| 1106 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; |
| 1107 | psig->cnvcsw += |
| 1108 | p->nvcsw + sig->nvcsw + sig->cnvcsw; |
| 1109 | psig->cnivcsw += |
| 1110 | p->nivcsw + sig->nivcsw + sig->cnivcsw; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 1111 | psig->cinblock += |
| 1112 | task_io_get_inblock(p) + |
| 1113 | sig->inblock + sig->cinblock; |
| 1114 | psig->coublock += |
| 1115 | task_io_get_oublock(p) + |
| 1116 | sig->oublock + sig->coublock; |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 1117 | maxrss = max(sig->maxrss, sig->cmaxrss); |
| 1118 | if (psig->cmaxrss < maxrss) |
| 1119 | psig->cmaxrss = maxrss; |
Andrea Righi | 5995477 | 2008-07-27 17:29:15 +0200 | [diff] [blame] | 1120 | task_io_accounting_add(&psig->ioac, &p->ioac); |
| 1121 | task_io_accounting_add(&psig->ioac, &sig->ioac); |
Rik van Riel | e78c349 | 2014-08-16 13:40:10 -0400 | [diff] [blame] | 1122 | write_sequnlock(&psig->stats_lock); |
Oleg Nesterov | f953ccd | 2014-12-10 15:54:48 -0800 | [diff] [blame] | 1123 | spin_unlock_irq(¤t->sighand->siglock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1124 | } |
| 1125 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1126 | retval = wo->wo_rusage |
| 1127 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1128 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
| 1129 | ? p->signal->group_exit_code : p->exit_code; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1130 | if (!retval && wo->wo_stat) |
| 1131 | retval = put_user(status, wo->wo_stat); |
| 1132 | |
| 1133 | infop = wo->wo_info; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1134 | if (!retval && infop) |
| 1135 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1136 | if (!retval && infop) |
| 1137 | retval = put_user(0, &infop->si_errno); |
| 1138 | if (!retval && infop) { |
| 1139 | int why; |
| 1140 | |
| 1141 | if ((status & 0x7f) == 0) { |
| 1142 | why = CLD_EXITED; |
| 1143 | status >>= 8; |
| 1144 | } else { |
| 1145 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; |
| 1146 | status &= 0x7f; |
| 1147 | } |
| 1148 | retval = put_user((short)why, &infop->si_code); |
| 1149 | if (!retval) |
| 1150 | retval = put_user(status, &infop->si_status); |
| 1151 | } |
| 1152 | if (!retval && infop) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1153 | retval = put_user(pid, &infop->si_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1154 | if (!retval && infop) |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 1155 | retval = put_user(uid, &infop->si_uid); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1156 | if (!retval) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1157 | retval = pid; |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1158 | |
Oleg Nesterov | b436069 | 2014-04-07 15:38:43 -0700 | [diff] [blame] | 1159 | if (state == EXIT_TRACE) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1160 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1161 | /* We dropped tasklist, ptracer could die and untrace */ |
| 1162 | ptrace_unlink(p); |
Oleg Nesterov | b436069 | 2014-04-07 15:38:43 -0700 | [diff] [blame] | 1163 | |
| 1164 | /* If parent wants a zombie, don't release it now */ |
| 1165 | state = EXIT_ZOMBIE; |
| 1166 | if (do_notify_parent(p, p->exit_signal)) |
| 1167 | state = EXIT_DEAD; |
Oleg Nesterov | abd50b3 | 2014-04-07 15:38:42 -0700 | [diff] [blame] | 1168 | p->exit_state = state; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1169 | write_unlock_irq(&tasklist_lock); |
| 1170 | } |
Oleg Nesterov | abd50b3 | 2014-04-07 15:38:42 -0700 | [diff] [blame] | 1171 | if (state == EXIT_DEAD) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1172 | release_task(p); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1173 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1174 | return retval; |
| 1175 | } |
| 1176 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1177 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
| 1178 | { |
| 1179 | if (ptrace) { |
Oleg Nesterov | 570ac93 | 2016-01-20 14:59:58 -0800 | [diff] [blame] | 1180 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1181 | return &p->exit_code; |
| 1182 | } else { |
| 1183 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
| 1184 | return &p->signal->group_exit_code; |
| 1185 | } |
| 1186 | return NULL; |
| 1187 | } |
| 1188 | |
Tejun Heo | 19e2746 | 2011-05-12 10:47:23 +0200 | [diff] [blame] | 1189 | /** |
| 1190 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED |
| 1191 | * @wo: wait options |
| 1192 | * @ptrace: is the wait for ptrace |
| 1193 | * @p: task to wait for |
| 1194 | * |
| 1195 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. |
| 1196 | * |
| 1197 | * CONTEXT: |
| 1198 | * read_lock(&tasklist_lock), which is released if return value is |
| 1199 | * non-zero. Also, grabs and releases @p->sighand->siglock. |
| 1200 | * |
| 1201 | * RETURNS: |
| 1202 | * 0 if wait condition didn't exist and search for other wait conditions |
| 1203 | * should continue. Non-zero return, -errno on failure and @p's pid on |
| 1204 | * success, implies that tasklist_lock is released and wait condition |
| 1205 | * search should terminate. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1206 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1207 | static int wait_task_stopped(struct wait_opts *wo, |
| 1208 | int ptrace, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1209 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1210 | struct siginfo __user *infop; |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1211 | int retval, exit_code, *p_code, why; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1212 | uid_t uid = 0; /* unneeded, required by compiler */ |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1213 | pid_t pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1214 | |
Oleg Nesterov | 4791802 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1215 | /* |
| 1216 | * Traditionally we see ptrace'd stopped tasks regardless of options. |
| 1217 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1218 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1219 | return 0; |
| 1220 | |
Tejun Heo | 19e2746 | 2011-05-12 10:47:23 +0200 | [diff] [blame] | 1221 | if (!task_stopped_code(p, ptrace)) |
| 1222 | return 0; |
| 1223 | |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1224 | exit_code = 0; |
| 1225 | spin_lock_irq(&p->sighand->siglock); |
| 1226 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1227 | p_code = task_stopped_code(p, ptrace); |
| 1228 | if (unlikely(!p_code)) |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1229 | goto unlock_sig; |
| 1230 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1231 | exit_code = *p_code; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1232 | if (!exit_code) |
| 1233 | goto unlock_sig; |
| 1234 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1235 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1236 | *p_code = 0; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1237 | |
Sasha Levin | 8ca937a | 2012-05-17 23:31:39 +0200 | [diff] [blame] | 1238 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1239 | unlock_sig: |
| 1240 | spin_unlock_irq(&p->sighand->siglock); |
| 1241 | if (!exit_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1242 | return 0; |
| 1243 | |
| 1244 | /* |
| 1245 | * Now we are pretty sure this task is interesting. |
| 1246 | * Make sure it doesn't get reaped out from under us while we |
| 1247 | * give up the lock and then examine it below. We don't want to |
| 1248 | * keep holding onto the tasklist_lock while we call getrusage and |
| 1249 | * possibly take page faults for user memory. |
| 1250 | */ |
| 1251 | get_task_struct(p); |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1252 | pid = task_pid_vnr(p); |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1253 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1254 | read_unlock(&tasklist_lock); |
Peter Zijlstra | 1029a2b | 2014-09-24 10:18:49 +0200 | [diff] [blame] | 1255 | sched_annotate_sleep(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1256 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1257 | if (unlikely(wo->wo_flags & WNOWAIT)) |
| 1258 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1259 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1260 | retval = wo->wo_rusage |
| 1261 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
| 1262 | if (!retval && wo->wo_stat) |
| 1263 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); |
| 1264 | |
| 1265 | infop = wo->wo_info; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1266 | if (!retval && infop) |
| 1267 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1268 | if (!retval && infop) |
| 1269 | retval = put_user(0, &infop->si_errno); |
| 1270 | if (!retval && infop) |
Roland McGrath | 6efcae4 | 2008-03-08 11:41:22 -0800 | [diff] [blame] | 1271 | retval = put_user((short)why, &infop->si_code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1272 | if (!retval && infop) |
| 1273 | retval = put_user(exit_code, &infop->si_status); |
| 1274 | if (!retval && infop) |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1275 | retval = put_user(pid, &infop->si_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1276 | if (!retval && infop) |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1277 | retval = put_user(uid, &infop->si_uid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1278 | if (!retval) |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1279 | retval = pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1280 | put_task_struct(p); |
| 1281 | |
| 1282 | BUG_ON(!retval); |
| 1283 | return retval; |
| 1284 | } |
| 1285 | |
| 1286 | /* |
| 1287 | * Handle do_wait work for one task in a live, non-stopped state. |
| 1288 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
| 1289 | * the lock and this task is uninteresting. If we return nonzero, we have |
| 1290 | * released the lock and the system call should return. |
| 1291 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1292 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1293 | { |
| 1294 | int retval; |
| 1295 | pid_t pid; |
| 1296 | uid_t uid; |
| 1297 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1298 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1299 | return 0; |
| 1300 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1301 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
| 1302 | return 0; |
| 1303 | |
| 1304 | spin_lock_irq(&p->sighand->siglock); |
| 1305 | /* Re-check with the lock held. */ |
| 1306 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { |
| 1307 | spin_unlock_irq(&p->sighand->siglock); |
| 1308 | return 0; |
| 1309 | } |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1310 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1311 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
Sasha Levin | 8ca937a | 2012-05-17 23:31:39 +0200 | [diff] [blame] | 1312 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1313 | spin_unlock_irq(&p->sighand->siglock); |
| 1314 | |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1315 | pid = task_pid_vnr(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1316 | get_task_struct(p); |
| 1317 | read_unlock(&tasklist_lock); |
Peter Zijlstra | 1029a2b | 2014-09-24 10:18:49 +0200 | [diff] [blame] | 1318 | sched_annotate_sleep(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1319 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1320 | if (!wo->wo_info) { |
| 1321 | retval = wo->wo_rusage |
| 1322 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1323 | put_task_struct(p); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1324 | if (!retval && wo->wo_stat) |
| 1325 | retval = put_user(0xffff, wo->wo_stat); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1326 | if (!retval) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1327 | retval = pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1328 | } else { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1329 | retval = wait_noreap_copyout(wo, p, pid, uid, |
| 1330 | CLD_CONTINUED, SIGCONT); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1331 | BUG_ON(retval == 0); |
| 1332 | } |
| 1333 | |
| 1334 | return retval; |
| 1335 | } |
| 1336 | |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1337 | /* |
| 1338 | * Consider @p for a wait by @parent. |
| 1339 | * |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1340 | * -ECHILD should be in ->notask_error before the first call. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1341 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
| 1342 | * Returns zero if the search for a child should continue; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1343 | * then ->notask_error is 0 if @p is an eligible child, |
Stephen Smalley | ab83798 | 2017-01-10 12:28:32 -0500 | [diff] [blame] | 1344 | * or still -ECHILD. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1345 | */ |
Oleg Nesterov | b6e763f | 2009-09-23 15:56:50 -0700 | [diff] [blame] | 1346 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
| 1347 | struct task_struct *p) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1348 | { |
Oleg Nesterov | 3245d6a | 2015-01-08 14:32:12 -0800 | [diff] [blame] | 1349 | /* |
| 1350 | * We can race with wait_task_zombie() from another thread. |
| 1351 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition |
| 1352 | * can't confuse the checks below. |
| 1353 | */ |
| 1354 | int exit_state = ACCESS_ONCE(p->exit_state); |
Oleg Nesterov | b3ab031 | 2014-04-07 15:38:45 -0700 | [diff] [blame] | 1355 | int ret; |
| 1356 | |
Oleg Nesterov | 3245d6a | 2015-01-08 14:32:12 -0800 | [diff] [blame] | 1357 | if (unlikely(exit_state == EXIT_DEAD)) |
Oleg Nesterov | b3ab031 | 2014-04-07 15:38:45 -0700 | [diff] [blame] | 1358 | return 0; |
| 1359 | |
Oleg Nesterov | bf95993 | 2016-05-23 16:23:50 -0700 | [diff] [blame] | 1360 | ret = eligible_child(wo, ptrace, p); |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1361 | if (!ret) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1362 | return ret; |
| 1363 | |
Oleg Nesterov | 3245d6a | 2015-01-08 14:32:12 -0800 | [diff] [blame] | 1364 | if (unlikely(exit_state == EXIT_TRACE)) { |
Oleg Nesterov | 50b8d25 | 2012-01-04 17:29:02 +0100 | [diff] [blame] | 1365 | /* |
Oleg Nesterov | abd50b3 | 2014-04-07 15:38:42 -0700 | [diff] [blame] | 1366 | * ptrace == 0 means we are the natural parent. In this case |
| 1367 | * we should clear notask_error, debugger will notify us. |
Oleg Nesterov | 50b8d25 | 2012-01-04 17:29:02 +0100 | [diff] [blame] | 1368 | */ |
Oleg Nesterov | abd50b3 | 2014-04-07 15:38:42 -0700 | [diff] [blame] | 1369 | if (likely(!ptrace)) |
Oleg Nesterov | 50b8d25 | 2012-01-04 17:29:02 +0100 | [diff] [blame] | 1370 | wo->notask_error = 0; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1371 | return 0; |
Oleg Nesterov | 50b8d25 | 2012-01-04 17:29:02 +0100 | [diff] [blame] | 1372 | } |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1373 | |
Oleg Nesterov | 377d75d | 2014-04-07 15:38:47 -0700 | [diff] [blame] | 1374 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
| 1375 | /* |
| 1376 | * If it is traced by its real parent's group, just pretend |
| 1377 | * the caller is ptrace_do_wait() and reap this child if it |
| 1378 | * is zombie. |
| 1379 | * |
| 1380 | * This also hides group stop state from real parent; otherwise |
| 1381 | * a single stop can be reported twice as group and ptrace stop. |
| 1382 | * If a ptracer wants to distinguish these two events for its |
| 1383 | * own children it should create a separate process which takes |
| 1384 | * the role of real parent. |
| 1385 | */ |
| 1386 | if (!ptrace_reparented(p)) |
| 1387 | ptrace = 1; |
| 1388 | } |
| 1389 | |
Tejun Heo | 9b84cca | 2011-03-23 10:37:01 +0100 | [diff] [blame] | 1390 | /* slay zombie? */ |
Oleg Nesterov | 3245d6a | 2015-01-08 14:32:12 -0800 | [diff] [blame] | 1391 | if (exit_state == EXIT_ZOMBIE) { |
Tejun Heo | 9b84cca | 2011-03-23 10:37:01 +0100 | [diff] [blame] | 1392 | /* we don't reap group leaders with subthreads */ |
Oleg Nesterov | 7c733eb | 2014-04-07 15:38:49 -0700 | [diff] [blame] | 1393 | if (!delay_group_leader(p)) { |
| 1394 | /* |
| 1395 | * A zombie ptracee is only visible to its ptracer. |
| 1396 | * Notification and reaping will be cascaded to the |
| 1397 | * real parent when the ptracer detaches. |
| 1398 | */ |
| 1399 | if (unlikely(ptrace) || likely(!p->ptrace)) |
| 1400 | return wait_task_zombie(wo, p); |
| 1401 | } |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1402 | |
Tejun Heo | 9b84cca | 2011-03-23 10:37:01 +0100 | [diff] [blame] | 1403 | /* |
| 1404 | * Allow access to stopped/continued state via zombie by |
| 1405 | * falling through. Clearing of notask_error is complex. |
| 1406 | * |
| 1407 | * When !@ptrace: |
| 1408 | * |
| 1409 | * If WEXITED is set, notask_error should naturally be |
| 1410 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, |
| 1411 | * so, if there are live subthreads, there are events to |
| 1412 | * wait for. If all subthreads are dead, it's still safe |
| 1413 | * to clear - this function will be called again in finite |
| 1414 | * amount time once all the subthreads are released and |
| 1415 | * will then return without clearing. |
| 1416 | * |
| 1417 | * When @ptrace: |
| 1418 | * |
| 1419 | * Stopped state is per-task and thus can't change once the |
| 1420 | * target task dies. Only continued and exited can happen. |
| 1421 | * Clear notask_error if WCONTINUED | WEXITED. |
| 1422 | */ |
| 1423 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) |
| 1424 | wo->notask_error = 0; |
| 1425 | } else { |
| 1426 | /* |
| 1427 | * @p is alive and it's gonna stop, continue or exit, so |
| 1428 | * there always is something to wait for. |
| 1429 | */ |
| 1430 | wo->notask_error = 0; |
| 1431 | } |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1432 | |
| 1433 | /* |
Tejun Heo | 45cb24a | 2011-03-23 10:37:01 +0100 | [diff] [blame] | 1434 | * Wait for stopped. Depending on @ptrace, different stopped state |
| 1435 | * is used and the two don't interact with each other. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1436 | */ |
Tejun Heo | 19e2746 | 2011-05-12 10:47:23 +0200 | [diff] [blame] | 1437 | ret = wait_task_stopped(wo, ptrace, p); |
| 1438 | if (ret) |
| 1439 | return ret; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1440 | |
Tejun Heo | 45cb24a | 2011-03-23 10:37:01 +0100 | [diff] [blame] | 1441 | /* |
| 1442 | * Wait for continued. There's only one continued state and the |
| 1443 | * ptracer can consume it which can confuse the real parent. Don't |
| 1444 | * use WCONTINUED from ptracer. You don't need or want it. |
| 1445 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1446 | return wait_task_continued(wo, p); |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1447 | } |
| 1448 | |
| 1449 | /* |
| 1450 | * Do the work of do_wait() for one thread in the group, @tsk. |
| 1451 | * |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1452 | * -ECHILD should be in ->notask_error before the first call. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1453 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
| 1454 | * Returns zero if the search for a child should continue; then |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1455 | * ->notask_error is 0 if there were any eligible children, |
Stephen Smalley | ab83798 | 2017-01-10 12:28:32 -0500 | [diff] [blame] | 1456 | * or still -ECHILD. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1457 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1458 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1459 | { |
| 1460 | struct task_struct *p; |
| 1461 | |
| 1462 | list_for_each_entry(p, &tsk->children, sibling) { |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 1463 | int ret = wait_consider_task(wo, 0, p); |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 1464 | |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 1465 | if (ret) |
| 1466 | return ret; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1467 | } |
| 1468 | |
| 1469 | return 0; |
| 1470 | } |
| 1471 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1472 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1473 | { |
| 1474 | struct task_struct *p; |
| 1475 | |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1476 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
Oleg Nesterov | b6e763f | 2009-09-23 15:56:50 -0700 | [diff] [blame] | 1477 | int ret = wait_consider_task(wo, 1, p); |
Ionut Alexa | a0be55d | 2014-08-08 14:21:18 -0700 | [diff] [blame] | 1478 | |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1479 | if (ret) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1480 | return ret; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1481 | } |
| 1482 | |
| 1483 | return 0; |
| 1484 | } |
| 1485 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1486 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
| 1487 | int sync, void *key) |
| 1488 | { |
| 1489 | struct wait_opts *wo = container_of(wait, struct wait_opts, |
| 1490 | child_wait); |
| 1491 | struct task_struct *p = key; |
| 1492 | |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 1493 | if (!eligible_pid(wo, p)) |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1494 | return 0; |
| 1495 | |
Oleg Nesterov | b4fe518 | 2009-09-23 15:56:47 -0700 | [diff] [blame] | 1496 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
| 1497 | return 0; |
| 1498 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1499 | return default_wake_function(wait, mode, sync, key); |
| 1500 | } |
| 1501 | |
Oleg Nesterov | a7f0765 | 2009-09-23 15:56:44 -0700 | [diff] [blame] | 1502 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
| 1503 | { |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1504 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
| 1505 | TASK_INTERRUPTIBLE, 1, p); |
Oleg Nesterov | a7f0765 | 2009-09-23 15:56:44 -0700 | [diff] [blame] | 1506 | } |
| 1507 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1508 | static long do_wait(struct wait_opts *wo) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1509 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1510 | struct task_struct *tsk; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1511 | int retval; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1512 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1513 | trace_sched_process_wait(wo->wo_pid); |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 1514 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1515 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
| 1516 | wo->child_wait.private = current; |
| 1517 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1518 | repeat: |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1519 | /* |
Frans Klaver | 3da56d1 | 2015-05-21 22:35:57 +0200 | [diff] [blame] | 1520 | * If there is nothing that can match our criteria, just get out. |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1521 | * We will clear ->notask_error to zero if we see any child that |
| 1522 | * might later match our criteria, even if we are not able to reap |
| 1523 | * it yet. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1524 | */ |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1525 | wo->notask_error = -ECHILD; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1526 | if ((wo->wo_type < PIDTYPE_MAX) && |
| 1527 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1528 | goto notask; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1529 | |
Oleg Nesterov | f95d39d | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 1530 | set_current_state(TASK_INTERRUPTIBLE); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1531 | read_lock(&tasklist_lock); |
| 1532 | tsk = current; |
| 1533 | do { |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1534 | retval = do_wait_thread(wo, tsk); |
| 1535 | if (retval) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1536 | goto end; |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1537 | |
| 1538 | retval = ptrace_do_wait(wo, tsk); |
| 1539 | if (retval) |
| 1540 | goto end; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1541 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1542 | if (wo->wo_flags & __WNOTHREAD) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1543 | break; |
Oleg Nesterov | a3f6dfb | 2009-06-17 16:27:41 -0700 | [diff] [blame] | 1544 | } while_each_thread(current, tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1545 | read_unlock(&tasklist_lock); |
Oleg Nesterov | f2cc3eb | 2008-02-08 04:19:06 -0800 | [diff] [blame] | 1546 | |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1547 | notask: |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1548 | retval = wo->notask_error; |
| 1549 | if (!retval && !(wo->wo_flags & WNOHANG)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1550 | retval = -ERESTARTSYS; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1551 | if (!signal_pending(current)) { |
| 1552 | schedule(); |
| 1553 | goto repeat; |
| 1554 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1555 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1556 | end: |
Oleg Nesterov | f95d39d | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 1557 | __set_current_state(TASK_RUNNING); |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1558 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1559 | return retval; |
| 1560 | } |
| 1561 | |
Heiko Carstens | 17da2bd | 2009-01-14 14:14:10 +0100 | [diff] [blame] | 1562 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
| 1563 | infop, int, options, struct rusage __user *, ru) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1564 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1565 | struct wait_opts wo; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1566 | struct pid *pid = NULL; |
| 1567 | enum pid_type type; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1568 | long ret; |
| 1569 | |
Oleg Nesterov | 91c4e8e | 2016-05-23 16:23:53 -0700 | [diff] [blame] | 1570 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
| 1571 | __WNOTHREAD|__WCLONE|__WALL)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1572 | return -EINVAL; |
| 1573 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) |
| 1574 | return -EINVAL; |
| 1575 | |
| 1576 | switch (which) { |
| 1577 | case P_ALL: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1578 | type = PIDTYPE_MAX; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1579 | break; |
| 1580 | case P_PID: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1581 | type = PIDTYPE_PID; |
| 1582 | if (upid <= 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1583 | return -EINVAL; |
| 1584 | break; |
| 1585 | case P_PGID: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1586 | type = PIDTYPE_PGID; |
| 1587 | if (upid <= 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1588 | return -EINVAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1589 | break; |
| 1590 | default: |
| 1591 | return -EINVAL; |
| 1592 | } |
| 1593 | |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1594 | if (type < PIDTYPE_MAX) |
| 1595 | pid = find_get_pid(upid); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1596 | |
| 1597 | wo.wo_type = type; |
| 1598 | wo.wo_pid = pid; |
| 1599 | wo.wo_flags = options; |
| 1600 | wo.wo_info = infop; |
| 1601 | wo.wo_stat = NULL; |
| 1602 | wo.wo_rusage = ru; |
| 1603 | ret = do_wait(&wo); |
Vitaly Mayatskikh | dfe16df | 2009-09-23 15:56:51 -0700 | [diff] [blame] | 1604 | |
| 1605 | if (ret > 0) { |
| 1606 | ret = 0; |
| 1607 | } else if (infop) { |
| 1608 | /* |
| 1609 | * For a WNOHANG return, clear out all the fields |
| 1610 | * we would set so the user can easily tell the |
| 1611 | * difference. |
| 1612 | */ |
| 1613 | if (!ret) |
| 1614 | ret = put_user(0, &infop->si_signo); |
| 1615 | if (!ret) |
| 1616 | ret = put_user(0, &infop->si_errno); |
| 1617 | if (!ret) |
| 1618 | ret = put_user(0, &infop->si_code); |
| 1619 | if (!ret) |
| 1620 | ret = put_user(0, &infop->si_pid); |
| 1621 | if (!ret) |
| 1622 | ret = put_user(0, &infop->si_uid); |
| 1623 | if (!ret) |
| 1624 | ret = put_user(0, &infop->si_status); |
| 1625 | } |
| 1626 | |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1627 | put_pid(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1628 | return ret; |
| 1629 | } |
| 1630 | |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 1631 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
| 1632 | int, options, struct rusage __user *, ru) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1633 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1634 | struct wait_opts wo; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1635 | struct pid *pid = NULL; |
| 1636 | enum pid_type type; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1637 | long ret; |
| 1638 | |
| 1639 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| |
| 1640 | __WNOTHREAD|__WCLONE|__WALL)) |
| 1641 | return -EINVAL; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1642 | |
zhongjiang | 04103c2 | 2017-07-10 15:53:01 -0700 | [diff] [blame] | 1643 | /* -INT_MIN is not defined */ |
| 1644 | if (upid == INT_MIN) |
| 1645 | return -ESRCH; |
| 1646 | |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1647 | if (upid == -1) |
| 1648 | type = PIDTYPE_MAX; |
| 1649 | else if (upid < 0) { |
| 1650 | type = PIDTYPE_PGID; |
| 1651 | pid = find_get_pid(-upid); |
| 1652 | } else if (upid == 0) { |
| 1653 | type = PIDTYPE_PGID; |
Oleg Nesterov | 2ae448e | 2009-04-02 16:58:36 -0700 | [diff] [blame] | 1654 | pid = get_task_pid(current, PIDTYPE_PGID); |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1655 | } else /* upid > 0 */ { |
| 1656 | type = PIDTYPE_PID; |
| 1657 | pid = find_get_pid(upid); |
| 1658 | } |
| 1659 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1660 | wo.wo_type = type; |
| 1661 | wo.wo_pid = pid; |
| 1662 | wo.wo_flags = options | WEXITED; |
| 1663 | wo.wo_info = NULL; |
| 1664 | wo.wo_stat = stat_addr; |
| 1665 | wo.wo_rusage = ru; |
| 1666 | ret = do_wait(&wo); |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1667 | put_pid(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1668 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1669 | return ret; |
| 1670 | } |
| 1671 | |
| 1672 | #ifdef __ARCH_WANT_SYS_WAITPID |
| 1673 | |
| 1674 | /* |
| 1675 | * sys_waitpid() remains for compatibility. waitpid() should be |
| 1676 | * implemented by calling sys_wait4() from libc.a. |
| 1677 | */ |
Heiko Carstens | 17da2bd | 2009-01-14 14:14:10 +0100 | [diff] [blame] | 1678 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1679 | { |
| 1680 | return sys_wait4(pid, stat_addr, options, NULL); |
| 1681 | } |
| 1682 | |
| 1683 | #endif |