blob: feff10bbb30777f93c33369a186650a6eb1a2e6a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/kernel/exit.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#include <linux/mm.h>
8#include <linux/slab.h>
9#include <linux/interrupt.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/module.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080011#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/completion.h>
13#include <linux/personality.h>
14#include <linux/tty.h>
Jens Axboeda9cbc82008-06-30 20:42:08 +020015#include <linux/iocontext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/key.h>
17#include <linux/security.h>
18#include <linux/cpu.h>
19#include <linux/acct.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070020#include <linux/tsacct_kern.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040022#include <linux/fdtable.h>
Mandeep Singh Baines80d26af2013-02-27 17:03:20 -080023#include <linux/freezer.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/binfmts.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070025#include <linux/nsproxy.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080026#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/ptrace.h>
28#include <linux/profile.h>
29#include <linux/mount.h>
30#include <linux/proc_fs.h>
Eric W. Biederman49d769d2007-05-09 02:34:33 -070031#include <linux/kthread.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <linux/mempolicy.h>
Shailabh Nagarc7572492006-07-14 00:24:40 -070033#include <linux/taskstats_kern.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070034#include <linux/delayacct.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070035#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/syscalls.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070037#include <linux/signal.h>
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080038#include <linux/posix-timers.h>
Matt Helsley9f460802005-11-07 00:59:16 -080039#include <linux/cn_proc.h>
Ingo Molnarde5097c2006-01-09 15:59:21 -080040#include <linux/mutex.h>
Ingo Molnar0771dfe2006-03-27 01:16:22 -080041#include <linux/futex.h>
Jens Axboeb92ce552006-04-11 13:52:07 +020042#include <linux/pipe_fs_i.h>
Al Virofa84cb92006-03-29 20:30:19 -050043#include <linux/audit.h> /* for audit_free() */
Adrian Bunk83cc5ed2006-06-25 05:47:41 -070044#include <linux/resource.h>
David Howells0d67a462006-08-29 19:05:56 +010045#include <linux/blkdev.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070046#include <linux/task_io_accounting_ops.h>
Roland McGrath30199f52008-07-25 19:45:46 -070047#include <linux/tracehook.h>
Al Viro5ad4e532009-03-29 19:50:06 -040048#include <linux/fs_struct.h>
David Howellsd84f4f92008-11-14 10:39:23 +110049#include <linux/init_task.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020050#include <linux/perf_event.h>
Steven Rostedtad8d75f2009-04-14 19:39:12 -040051#include <trace/events/sched.h>
Frederic Weisbecker24f1e32c2009-09-09 19:22:48 +020052#include <linux/hw_breakpoint.h>
Ying Han3d5992d2010-10-26 14:21:23 -070053#include <linux/oom.h>
Wu Fengguang54848d72011-04-05 13:21:19 -060054#include <linux/writeback.h>
Al Viro40401532012-02-13 03:58:52 +000055#include <linux/shm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070056
57#include <asm/uaccess.h>
58#include <asm/unistd.h>
59#include <asm/pgtable.h>
60#include <asm/mmu_context.h>
61
Ionut Alexaa0be55d2014-08-08 14:21:18 -070062static void exit_mm(struct task_struct *tsk);
Adrian Bunk408b6642005-05-01 08:59:29 -070063
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070064static void __unhash_process(struct task_struct *p, bool group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -070065{
66 nr_threads--;
Oleg Nesterov50d75f82012-06-20 12:53:04 -070067 detach_pid(p, PIDTYPE_PID);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070068 if (group_dead) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 detach_pid(p, PIDTYPE_PGID);
70 detach_pid(p, PIDTYPE_SID);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080071
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -070072 list_del_rcu(&p->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -080073 list_del_init(&p->sibling);
Christoph Lameter909ea962010-12-08 16:22:55 +010074 __this_cpu_dec(process_counts);
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 }
Oleg Nesterov47e65322006-03-28 16:11:25 -080076 list_del_rcu(&p->thread_group);
Oleg Nesterov0c740d02014-01-21 15:49:56 -080077 list_del_rcu(&p->thread_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078}
79
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080080/*
81 * This function expects the tasklist_lock write-locked.
82 */
83static void __exit_signal(struct task_struct *tsk)
84{
85 struct signal_struct *sig = tsk->signal;
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070086 bool group_dead = thread_group_leader(tsk);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080087 struct sighand_struct *sighand;
Oleg Nesterov4ada8562010-05-26 14:43:17 -070088 struct tty_struct *uninitialized_var(tty);
Frederic Weisbecker6fac4822012-11-13 14:20:55 +010089 cputime_t utime, stime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080090
Paul E. McKenneyd11c5632010-02-22 17:04:50 -080091 sighand = rcu_dereference_check(tsk->sighand,
Paul E. McKenneydb1466b2010-03-03 07:46:56 -080092 lockdep_tasklist_lock_is_held());
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080093 spin_lock(&sighand->siglock);
94
95 posix_cpu_timers_exit(tsk);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070096 if (group_dead) {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080097 posix_cpu_timers_exit_group(tsk);
Oleg Nesterov4ada8562010-05-26 14:43:17 -070098 tty = sig->tty;
99 sig->tty = NULL;
Oleg Nesterov4a599942010-05-26 14:43:12 -0700100 } else {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800101 /*
Oleg Nesterove0a70212010-11-05 16:53:42 +0100102 * This can only happen if the caller is de_thread().
103 * FIXME: this is the temporary hack, we should teach
104 * posix-cpu-timers to handle this case correctly.
105 */
106 if (unlikely(has_group_leader_pid(tsk)))
107 posix_cpu_timers_exit_group(tsk);
108
109 /*
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800110 * If there is any task waiting for the group exit
111 * then notify it:
112 */
Oleg Nesterovd3441932010-05-26 14:43:11 -0700113 if (sig->notify_count > 0 && !--sig->notify_count)
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800114 wake_up_process(sig->group_exit_task);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700115
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800116 if (tsk == sig->curr_target)
117 sig->curr_target = next_thread(tsk);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800118 }
119
Rik van Riel90ed9cb2014-08-15 16:05:36 -0400120 /*
Oleg Nesterov26e75b52014-12-10 15:54:54 -0800121 * Accumulate here the counters for all threads as they die. We could
122 * skip the group leader because it is the last user of signal_struct,
123 * but we want to avoid the race with thread_group_cputime() which can
124 * see the empty ->thread_head list.
Rik van Riel90ed9cb2014-08-15 16:05:36 -0400125 */
126 task_cputime(tsk, &utime, &stime);
Rik van Riele78c3492014-08-16 13:40:10 -0400127 write_seqlock(&sig->stats_lock);
Rik van Riel90ed9cb2014-08-15 16:05:36 -0400128 sig->utime += utime;
129 sig->stime += stime;
130 sig->gtime += task_gtime(tsk);
131 sig->min_flt += tsk->min_flt;
132 sig->maj_flt += tsk->maj_flt;
133 sig->nvcsw += tsk->nvcsw;
134 sig->nivcsw += tsk->nivcsw;
135 sig->inblock += task_io_get_inblock(tsk);
136 sig->oublock += task_io_get_oublock(tsk);
137 task_io_accounting_add(&sig->ioac, &tsk->ioac);
138 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
Oleg Nesterovb3ac0222010-05-26 14:43:24 -0700139 sig->nr_threads--;
Oleg Nesterovd40e48e2010-05-26 14:43:19 -0700140 __unhash_process(tsk, group_dead);
Rik van Riele78c3492014-08-16 13:40:10 -0400141 write_sequnlock(&sig->stats_lock);
Oleg Nesterov58767002006-03-28 16:11:20 -0800142
Oleg Nesterovda7978b2008-05-23 13:04:41 -0700143 /*
144 * Do this under ->siglock, we can race with another thread
145 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
146 */
147 flush_sigqueue(&tsk->pending);
Oleg Nesterova7e53282006-03-28 16:11:27 -0800148 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800149 spin_unlock(&sighand->siglock);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800150
Oleg Nesterova7e53282006-03-28 16:11:27 -0800151 __cleanup_sighand(sighand);
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700152 clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -0700153 if (group_dead) {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800154 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov4ada8562010-05-26 14:43:17 -0700155 tty_kref_put(tty);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800156 }
157}
158
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800159static void delayed_put_task_struct(struct rcu_head *rhp)
160{
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400161 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
162
Peter Zijlstra4e231c72010-09-09 21:01:59 +0200163 perf_event_delayed_put(tsk);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400164 trace_sched_process_free(tsk);
165 put_task_struct(tsk);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800166}
167
Roland McGrathf4700212008-03-24 18:36:23 -0700168
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700169void release_task(struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700171 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800173repeat:
David Howellsc69e8d92008-11-14 10:39:19 +1100174 /* don't need to get the RCU readlock here - the process is dead and
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800175 * can't be modifying its own credentials. But shut RCU-lockdep up */
176 rcu_read_lock();
David Howellsc69e8d92008-11-14 10:39:19 +1100177 atomic_dec(&__task_cred(p)->user->processes);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800178 rcu_read_unlock();
David Howellsc69e8d92008-11-14 10:39:19 +1100179
Pavel Emelyanov60347f62007-10-18 23:40:03 -0700180 proc_flush_task(p);
Ingo Molnar02030262009-05-17 11:24:08 +0200181
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 write_lock_irq(&tasklist_lock);
Tejun Heoa288eec2011-06-17 16:50:37 +0200183 ptrace_release_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 __exit_signal(p);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -0800185
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 /*
187 * If we are the last non-leader member of the thread
188 * group, and the leader is zombie, then notify the
189 * group leader's parent process. (if it wants notification.)
190 */
191 zap_leader = 0;
192 leader = p->group_leader;
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700193 if (leader != p && thread_group_empty(leader)
194 && leader->exit_state == EXIT_ZOMBIE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 /*
196 * If we were the last child thread and the leader has
197 * exited already, and the leader's parent ignores SIGCHLD,
198 * then we are the one who should release the leader.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 */
Oleg Nesterov86773472011-06-22 23:09:09 +0200200 zap_leader = do_notify_parent(leader, leader->exit_signal);
Roland McGrathdae33572008-07-25 19:45:48 -0700201 if (zap_leader)
202 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 }
204
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 release_thread(p);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800207 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208
209 p = leader;
210 if (unlikely(zap_leader))
211 goto repeat;
212}
213
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 * Determine if a process group is "orphaned", according to the POSIX
216 * definition in 2.2.2.52. Orphaned process groups are not to be affected
217 * by terminal-generated stop signals. Newly orphaned process groups are
218 * to receive a SIGHUP and a SIGCONT.
219 *
220 * "I ask you, have you ever known what it is to be an orphan?"
221 */
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700222static int will_become_orphaned_pgrp(struct pid *pgrp,
223 struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224{
225 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800227 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300228 if ((p == ignored_task) ||
229 (p->exit_state && thread_group_empty(p)) ||
230 is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 continue;
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300232
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800233 if (task_pgrp(p->real_parent) != pgrp &&
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300234 task_session(p->real_parent) == task_session(p))
235 return 0;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800236 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300237
238 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239}
240
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800241int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242{
243 int retval;
244
245 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800246 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 read_unlock(&tasklist_lock);
248
249 return retval;
250}
251
Oleg Nesterov961c4672011-07-07 21:33:54 +0200252static bool has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 struct task_struct *p;
255
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800256 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov961c4672011-07-07 21:33:54 +0200257 if (p->signal->flags & SIGNAL_STOP_STOPPED)
258 return true;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800259 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov961c4672011-07-07 21:33:54 +0200260
261 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262}
263
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300264/*
265 * Check to see if any process groups have become orphaned as
266 * a result of our exiting, and if they have any stopped jobs,
267 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
268 */
269static void
270kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
271{
272 struct pid *pgrp = task_pgrp(tsk);
273 struct task_struct *ignored_task = tsk;
274
275 if (!parent)
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700276 /* exit: our father is in a different pgrp than
277 * we are and we were the only connection outside.
278 */
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300279 parent = tsk->real_parent;
280 else
281 /* reparent: our child is in a different pgrp than
282 * we are, and it was the only connection outside.
283 */
284 ignored_task = NULL;
285
286 if (task_pgrp(parent) != pgrp &&
287 task_session(parent) == task_session(tsk) &&
288 will_become_orphaned_pgrp(pgrp, ignored_task) &&
289 has_stopped_jobs(pgrp)) {
290 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
291 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
292 }
293}
294
Oleg Nesterovf98bafa2014-06-04 16:07:34 -0700295#ifdef CONFIG_MEMCG
Balbir Singhcf475ad2008-04-29 01:00:16 -0700296/*
KAMEZAWA Hiroyuki733eda72011-06-15 15:08:43 -0700297 * A task is exiting. If it owned this mm, find a new owner for the mm.
Balbir Singhcf475ad2008-04-29 01:00:16 -0700298 */
Balbir Singhcf475ad2008-04-29 01:00:16 -0700299void mm_update_next_owner(struct mm_struct *mm)
300{
301 struct task_struct *c, *g, *p = current;
302
303retry:
KAMEZAWA Hiroyuki733eda72011-06-15 15:08:43 -0700304 /*
305 * If the exiting or execing task is not the owner, it's
306 * someone else's problem.
307 */
308 if (mm->owner != p)
Balbir Singhcf475ad2008-04-29 01:00:16 -0700309 return;
KAMEZAWA Hiroyuki733eda72011-06-15 15:08:43 -0700310 /*
311 * The current owner is exiting/execing and there are no other
312 * candidates. Do not leave the mm pointing to a possibly
313 * freed task structure.
314 */
315 if (atomic_read(&mm->mm_users) <= 1) {
316 mm->owner = NULL;
317 return;
318 }
Balbir Singhcf475ad2008-04-29 01:00:16 -0700319
320 read_lock(&tasklist_lock);
321 /*
322 * Search in the children
323 */
324 list_for_each_entry(c, &p->children, sibling) {
325 if (c->mm == mm)
326 goto assign_new_owner;
327 }
328
329 /*
330 * Search in the siblings
331 */
Oleg Nesterovdea33cf2009-06-17 16:27:29 -0700332 list_for_each_entry(c, &p->real_parent->children, sibling) {
Balbir Singhcf475ad2008-04-29 01:00:16 -0700333 if (c->mm == mm)
334 goto assign_new_owner;
335 }
336
337 /*
Oleg Nesterovf87fb592014-06-04 16:07:52 -0700338 * Search through everything else, we should not get here often.
Balbir Singhcf475ad2008-04-29 01:00:16 -0700339 */
Oleg Nesterov39af1762014-06-04 16:07:54 -0700340 for_each_process(g) {
341 if (g->flags & PF_KTHREAD)
342 continue;
343 for_each_thread(g, c) {
344 if (c->mm == mm)
345 goto assign_new_owner;
346 if (c->mm)
347 break;
348 }
Oleg Nesterovf87fb592014-06-04 16:07:52 -0700349 }
Balbir Singhcf475ad2008-04-29 01:00:16 -0700350 read_unlock(&tasklist_lock);
Balbir Singh31a78f22008-09-28 23:09:31 +0100351 /*
352 * We found no owner yet mm_users > 1: this implies that we are
353 * most likely racing with swapoff (try_to_unuse()) or /proc or
Hugh Dickinse5991372009-01-06 14:39:22 -0800354 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
Balbir Singh31a78f22008-09-28 23:09:31 +0100355 */
Balbir Singh31a78f22008-09-28 23:09:31 +0100356 mm->owner = NULL;
Balbir Singhcf475ad2008-04-29 01:00:16 -0700357 return;
358
359assign_new_owner:
360 BUG_ON(c == p);
361 get_task_struct(c);
362 /*
363 * The task_lock protects c->mm from changing.
364 * We always want mm->owner->mm == mm
365 */
366 task_lock(c);
Hugh Dickinse5991372009-01-06 14:39:22 -0800367 /*
368 * Delay read_unlock() till we have the task_lock()
369 * to ensure that c does not slip away underneath us
370 */
371 read_unlock(&tasklist_lock);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700372 if (c->mm != mm) {
373 task_unlock(c);
374 put_task_struct(c);
375 goto retry;
376 }
Balbir Singhcf475ad2008-04-29 01:00:16 -0700377 mm->owner = c;
378 task_unlock(c);
379 put_task_struct(c);
380}
Oleg Nesterovf98bafa2014-06-04 16:07:34 -0700381#endif /* CONFIG_MEMCG */
Balbir Singhcf475ad2008-04-29 01:00:16 -0700382
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383/*
384 * Turn us into a lazy TLB process if we
385 * aren't already..
386 */
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700387static void exit_mm(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388{
389 struct mm_struct *mm = tsk->mm;
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700390 struct core_state *core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391
Linus Torvalds48d212a2012-06-07 17:54:07 -0700392 mm_release(tsk, mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 if (!mm)
394 return;
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -0700395 sync_mm_rss(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 /*
397 * Serialize with any possible pending coredump.
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700398 * We must hold mmap_sem around checking core_state
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 * and clearing tsk->mm. The core-inducing thread
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700400 * will increment ->nr_threads for each thread in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 * group with ->mm != NULL.
402 */
403 down_read(&mm->mmap_sem);
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700404 core_state = mm->core_state;
405 if (core_state) {
406 struct core_thread self;
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700407
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 up_read(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700410 self.task = tsk;
411 self.next = xchg(&core_state->dumper.next, &self);
412 /*
413 * Implies mb(), the result of xchg() must be visible
414 * to core_state->dumper.
415 */
416 if (atomic_dec_and_test(&core_state->nr_threads))
417 complete(&core_state->startup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418
Oleg Nesterova94e2d42008-07-25 01:47:46 -0700419 for (;;) {
420 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
421 if (!self.task) /* see coredump_finish() */
422 break;
Mandeep Singh Baines80d26af2013-02-27 17:03:20 -0800423 freezable_schedule();
Oleg Nesterova94e2d42008-07-25 01:47:46 -0700424 }
425 __set_task_state(tsk, TASK_RUNNING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 down_read(&mm->mmap_sem);
427 }
428 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700429 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 /* more a memory barrier than a real lock */
431 task_lock(tsk);
432 tsk->mm = NULL;
433 up_read(&mm->mmap_sem);
434 enter_lazy_tlb(mm, current);
435 task_unlock(tsk);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700436 mm_update_next_owner(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 mmput(mm);
Michal Hockoc32b3cb2015-02-11 15:26:24 -0800438 if (test_thread_flag(TIF_MEMDIE))
439 unmark_oom_victim();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440}
441
Oleg Nesterovc9dc05b2014-12-10 15:55:14 -0800442static struct task_struct *find_alive_thread(struct task_struct *p)
443{
444 struct task_struct *t;
445
446 for_each_thread(p, t) {
447 if (!(t->flags & PF_EXITING))
448 return t;
449 }
450 return NULL;
451}
452
Oleg Nesterov11099092014-12-10 15:55:11 -0800453static struct task_struct *find_child_reaper(struct task_struct *father)
454 __releases(&tasklist_lock)
455 __acquires(&tasklist_lock)
456{
457 struct pid_namespace *pid_ns = task_active_pid_ns(father);
458 struct task_struct *reaper = pid_ns->child_reaper;
459
460 if (likely(reaper != father))
461 return reaper;
462
Oleg Nesterovc9dc05b2014-12-10 15:55:14 -0800463 reaper = find_alive_thread(father);
464 if (reaper) {
Oleg Nesterov11099092014-12-10 15:55:11 -0800465 pid_ns->child_reaper = reaper;
466 return reaper;
467 }
468
469 write_unlock_irq(&tasklist_lock);
470 if (unlikely(pid_ns == &init_pid_ns)) {
471 panic("Attempted to kill init! exitcode=0x%08x\n",
472 father->signal->group_exit_code ?: father->exit_code);
473 }
474 zap_pid_ns_processes(pid_ns);
475 write_lock_irq(&tasklist_lock);
476
477 return father;
478}
479
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480/*
Lennart Poetteringebec18a2012-03-23 15:01:54 -0700481 * When we die, we re-parent all our children, and try to:
482 * 1. give them to another thread in our thread group, if such a member exists
483 * 2. give it to the first ancestor process which prctl'd itself as a
484 * child_subreaper for its children (like a service manager)
485 * 3. give it to the init process (PID 1) in our pid namespace
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 */
Oleg Nesterov11099092014-12-10 15:55:11 -0800487static struct task_struct *find_new_reaper(struct task_struct *father,
488 struct task_struct *child_reaper)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
Oleg Nesterovc9dc05b2014-12-10 15:55:14 -0800490 struct task_struct *thread, *reaper;
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700491
Oleg Nesterovc9dc05b2014-12-10 15:55:14 -0800492 thread = find_alive_thread(father);
493 if (thread)
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700494 return thread;
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700495
Oleg Nesterov7d24e2d2014-12-10 15:55:02 -0800496 if (father->signal->has_child_subreaper) {
Lennart Poetteringebec18a2012-03-23 15:01:54 -0700497 /*
Oleg Nesterov175aed32014-12-10 15:55:08 -0800498 * Find the first ->is_child_subreaper ancestor in our pid_ns.
499 * We start from father to ensure we can not look into another
500 * namespace, this is safe because all its threads are dead.
Lennart Poetteringebec18a2012-03-23 15:01:54 -0700501 */
Oleg Nesterov7d24e2d2014-12-10 15:55:02 -0800502 for (reaper = father;
Oleg Nesterov11099092014-12-10 15:55:11 -0800503 !same_thread_group(reaper, child_reaper);
Lennart Poetteringebec18a2012-03-23 15:01:54 -0700504 reaper = reaper->real_parent) {
Oleg Nesterov175aed32014-12-10 15:55:08 -0800505 /* call_usermodehelper() descendants need this check */
506 if (reaper == &init_task)
Lennart Poetteringebec18a2012-03-23 15:01:54 -0700507 break;
508 if (!reaper->signal->is_child_subreaper)
509 continue;
Oleg Nesterovc9dc05b2014-12-10 15:55:14 -0800510 thread = find_alive_thread(reaper);
511 if (thread)
512 return thread;
Lennart Poetteringebec18a2012-03-23 15:01:54 -0700513 }
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700514 }
515
Oleg Nesterov11099092014-12-10 15:55:11 -0800516 return child_reaper;
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700517}
518
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700519/*
520* Any that need to be release_task'd are put on the @dead list.
521 */
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800522static void reparent_leader(struct task_struct *father, struct task_struct *p,
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700523 struct list_head *dead)
524{
Oleg Nesterov28310962014-12-10 15:45:30 -0800525 if (unlikely(p->exit_state == EXIT_DEAD))
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700526 return;
527
Oleg Nesterovabd50b32014-04-07 15:38:42 -0700528 /* We don't want people slaying init. */
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700529 p->exit_signal = SIGCHLD;
530
531 /* If it has exited notify the new parent about this child's death. */
Tejun Heod21142e2011-06-17 16:50:34 +0200532 if (!p->ptrace &&
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700533 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
Oleg Nesterov86773472011-06-22 23:09:09 +0200534 if (do_notify_parent(p, p->exit_signal)) {
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700535 p->exit_state = EXIT_DEAD;
Oleg Nesterovdc2fd4b2014-12-10 15:45:24 -0800536 list_add(&p->ptrace_entry, dead);
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700537 }
538 }
539
540 kill_orphaned_pgrp(p, father);
541}
542
Oleg Nesterov482a3762014-12-10 15:55:20 -0800543/*
544 * This does two things:
545 *
546 * A. Make init inherit all the child processes
547 * B. Check to see if any process groups have become orphaned
548 * as a result of our exiting, and if they have any stopped
549 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
550 */
551static void forget_original_parent(struct task_struct *father,
552 struct list_head *dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553{
Oleg Nesterov482a3762014-12-10 15:55:20 -0800554 struct task_struct *p, *t, *reaper;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555
Oleg Nesterov7c8bd232014-12-10 15:45:33 -0800556 if (unlikely(!list_empty(&father->ptraced)))
Oleg Nesterov482a3762014-12-10 15:55:20 -0800557 exit_ptrace(father, dead);
Roland McGrathf4700212008-03-24 18:36:23 -0700558
Oleg Nesterov7c8bd232014-12-10 15:45:33 -0800559 /* Can drop and reacquire tasklist_lock */
Oleg Nesterov11099092014-12-10 15:55:11 -0800560 reaper = find_child_reaper(father);
Oleg Nesterovad9e2062014-12-10 15:55:17 -0800561 if (list_empty(&father->children))
Oleg Nesterov482a3762014-12-10 15:55:20 -0800562 return;
Oleg Nesterov11099092014-12-10 15:55:11 -0800563
564 reaper = find_new_reaper(father, reaper);
Oleg Nesterov28310962014-12-10 15:45:30 -0800565 list_for_each_entry(p, &father->children, sibling) {
Oleg Nesterov57a05912014-12-10 15:45:27 -0800566 for_each_thread(p, t) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800567 t->real_parent = reaper;
Oleg Nesterov57a05912014-12-10 15:45:27 -0800568 BUG_ON((!t->ptrace) != (t->parent == father));
569 if (likely(!t->ptrace))
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800570 t->parent = t->real_parent;
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800571 if (t->pdeath_signal)
572 group_send_sig_info(t->pdeath_signal,
573 SEND_SIG_NOINFO, t);
Oleg Nesterov57a05912014-12-10 15:45:27 -0800574 }
Oleg Nesterov28310962014-12-10 15:45:30 -0800575 /*
576 * If this is a threaded reparent there is no need to
577 * notify anyone anything has happened.
578 */
579 if (!same_thread_group(reaper, father))
Oleg Nesterov482a3762014-12-10 15:55:20 -0800580 reparent_leader(father, p, dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 }
Oleg Nesterov28310962014-12-10 15:45:30 -0800582 list_splice_tail_init(&father->children, &reaper->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583}
584
585/*
586 * Send signals to all our closest relatives so that they know
587 * to properly mourn us..
588 */
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300589static void exit_notify(struct task_struct *tsk, int group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590{
Oleg Nesterov53c8f9f2011-06-22 23:08:18 +0200591 bool autoreap;
Oleg Nesterov482a3762014-12-10 15:55:20 -0800592 struct task_struct *p, *n;
593 LIST_HEAD(dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700595 write_lock_irq(&tasklist_lock);
Oleg Nesterov482a3762014-12-10 15:55:20 -0800596 forget_original_parent(tsk, &dead);
597
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300598 if (group_dead)
599 kill_orphaned_pgrp(tsk->group_leader, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
Oleg Nesterov45cdf5c2011-06-23 19:06:50 +0200601 if (unlikely(tsk->ptrace)) {
602 int sig = thread_group_leader(tsk) &&
603 thread_group_empty(tsk) &&
604 !ptrace_reparented(tsk) ?
605 tsk->exit_signal : SIGCHLD;
606 autoreap = do_notify_parent(tsk, sig);
607 } else if (thread_group_leader(tsk)) {
608 autoreap = thread_group_empty(tsk) &&
609 do_notify_parent(tsk, tsk->exit_signal);
610 } else {
611 autoreap = true;
612 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
Oleg Nesterov53c8f9f2011-06-22 23:08:18 +0200614 tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE;
Oleg Nesterov6c66e7d2014-12-10 15:55:23 -0800615 if (tsk->exit_state == EXIT_DEAD)
616 list_add(&tsk->ptrace_entry, &dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617
Oleg Nesterov9c339162010-05-26 14:43:10 -0700618 /* mt-exec, de_thread() is waiting for group leader */
619 if (unlikely(tsk->signal->notify_count < 0))
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700620 wake_up_process(tsk->signal->group_exit_task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 write_unlock_irq(&tasklist_lock);
622
Oleg Nesterov482a3762014-12-10 15:55:20 -0800623 list_for_each_entry_safe(p, n, &dead, ptrace_entry) {
624 list_del_init(&p->ptrace_entry);
625 release_task(p);
626 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627}
628
Jeff Dikee18eecb2007-07-15 23:38:48 -0700629#ifdef CONFIG_DEBUG_STACK_USAGE
630static void check_stack_usage(void)
631{
632 static DEFINE_SPINLOCK(low_water_lock);
633 static int lowest_to_date = THREAD_SIZE;
Jeff Dikee18eecb2007-07-15 23:38:48 -0700634 unsigned long free;
635
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500636 free = stack_not_used(current);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700637
638 if (free >= lowest_to_date)
639 return;
640
641 spin_lock(&low_water_lock);
642 if (free < lowest_to_date) {
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700643 pr_warn("%s (%d) used greatest stack depth: %lu bytes left\n",
644 current->comm, task_pid_nr(current), free);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700645 lowest_to_date = free;
646 }
647 spin_unlock(&low_water_lock);
648}
649#else
650static inline void check_stack_usage(void) {}
651#endif
652
Joe Perches9402c952012-01-12 17:17:17 -0800653void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654{
655 struct task_struct *tsk = current;
656 int group_dead;
Paul E. McKenney3f95aa82014-08-04 06:10:23 -0700657 TASKS_RCU(int tasks_rcu_i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
659 profile_task_exit(tsk);
660
Jens Axboe73c10102011-03-08 13:19:51 +0100661 WARN_ON(blk_needs_flush_plug(tsk));
Jens Axboe22e2c502005-06-27 10:55:12 +0200662
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 if (unlikely(in_interrupt()))
664 panic("Aiee, killing interrupt handler!");
665 if (unlikely(!tsk->pid))
666 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Nelson Elhage33dd94a2010-12-02 14:31:21 -0800668 /*
669 * If do_exit is called because this processes oopsed, it's possible
670 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
671 * continuing. Amongst other possible reasons, this is to prevent
672 * mm_release()->clear_child_tid() from writing to a user-controlled
673 * kernel address.
674 */
675 set_fs(USER_DS);
676
Tejun Heoa288eec2011-06-17 16:50:37 +0200677 ptrace_event(PTRACE_EVENT_EXIT, code);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678
David Howellse0e81732009-09-02 09:13:40 +0100679 validate_creds_for_do_exit(tsk);
680
Alexander Nybergdf164db2005-06-23 00:09:13 -0700681 /*
682 * We're taking recursive faults here in do_exit. Safest is to just
683 * leave this task alone and wait for reboot.
684 */
685 if (unlikely(tsk->flags & PF_EXITING)) {
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700686 pr_alert("Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700687 /*
688 * We can do this unlocked here. The futex code uses
689 * this flag just to verify whether the pi state
690 * cleanup has been done or not. In the worst case it
691 * loops once more. We pretend that the cleanup was
692 * done as there is no way to return. Either the
693 * OWNER_DIED bit is set by now or we push the blocked
694 * task into the wait for ever nirwana as well.
695 */
696 tsk->flags |= PF_EXITPIDONE;
Alexander Nybergdf164db2005-06-23 00:09:13 -0700697 set_current_state(TASK_UNINTERRUPTIBLE);
698 schedule();
699 }
700
Oleg Nesterovd12619b2008-02-08 04:19:12 -0800701 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700702 /*
703 * tsk->flags are checked in the futex code to protect against
Al Viroed3e6942012-06-27 11:31:24 +0400704 * an exiting task cleaning up the robust pi futexes.
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700705 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -0700706 smp_mb();
Thomas Gleixner1d615482009-11-17 14:54:03 +0100707 raw_spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 if (unlikely(in_atomic()))
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700710 pr_info("note: %s[%d] exited with preempt_count %d\n",
711 current->comm, task_pid_nr(current),
712 preempt_count());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713
714 acct_update_integrals(tsk);
Linus Torvalds48d212a2012-06-07 17:54:07 -0700715 /* sync mm's RSS info before statistics gathering */
716 if (tsk->mm)
717 sync_mm_rss(tsk->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -0700719 if (group_dead) {
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700720 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -0700721 exit_itimers(tsk->signal);
Jiri Pirko1f102062009-09-22 16:44:10 -0700722 if (tsk->mm)
723 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
Andrew Mortonc3068952005-08-04 16:49:32 -0700724 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700725 acct_collect(code, group_dead);
Miloslav Trmac522ed772007-07-15 23:40:56 -0700726 if (group_dead)
727 tty_audit_exit();
Eric Parisa4ff8db2012-01-03 14:23:07 -0500728 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -0800729
Linus Torvalds48d212a2012-06-07 17:54:07 -0700730 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -0800731 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -0700732
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 exit_mm(tsk);
734
KaiGai Kohei0e464812006-06-25 05:49:24 -0700735 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700736 acct_process();
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400737 trace_sched_process_exit(tsk);
738
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 exit_sem(tsk);
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700740 exit_shm(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400741 exit_files(tsk);
742 exit_fs(tsk);
Oleg Nesterovc39df5f2014-04-07 15:38:29 -0700743 if (group_dead)
744 disassociate_ctty(1);
Oleg Nesterov8aac6272013-06-14 21:09:49 +0200745 exit_task_namespaces(tsk);
Al Viroed3e6942012-06-27 11:31:24 +0400746 exit_task_work(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 exit_thread();
Stephane Eranian0b3fcf12011-01-03 18:20:01 +0200748
749 /*
750 * Flush inherited counters to the parent - before the parent
751 * gets woken up by child-exit notifications.
752 *
753 * because of cgroup mode, must be called before cgroup_exit()
754 */
755 perf_event_exit_task(tsk);
756
Li Zefan1ec41832014-03-28 15:22:19 +0800757 cgroup_exit(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
Al Viroa1261f52005-11-13 16:06:55 -0800759 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200761 /*
Frederic Weisbecker24f1e32c2009-09-09 19:22:48 +0200762 * FIXME: do that only when needed, using sched_exit tracepoint
763 */
Oleg Nesterov7c8df282013-07-08 16:00:54 -0700764 flush_ptrace_hw_breakpoint(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200765
Paul E. McKenney3f95aa82014-08-04 06:10:23 -0700766 TASKS_RCU(tasks_rcu_i = __srcu_read_lock(&tasks_rcu_exit_srcu));
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300767 exit_notify(tsk, group_dead);
Guillaume Morinef982392014-04-07 15:38:31 -0700768 proc_exit_connector(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769#ifdef CONFIG_NUMA
Miao Xiec0ff7452010-05-24 14:32:08 -0700770 task_lock(tsk);
Lee Schermerhornf0be3d32008-04-28 02:13:08 -0700771 mpol_put(tsk->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 tsk->mempolicy = NULL;
Miao Xiec0ff7452010-05-24 14:32:08 -0700773 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700775#ifdef CONFIG_FUTEX
Ingo Molnarc87e2832006-06-27 02:54:58 -0700776 if (unlikely(current->pi_state_cache))
777 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700778#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -0700779 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -0700780 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -0800781 */
Colin Cross1b1d2fb2013-05-06 23:50:08 +0000782 debug_check_no_locks_held();
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700783 /*
784 * We can do this unlocked here. The futex code uses this flag
785 * just to verify whether the pi state cleanup has been done
786 * or not. In the worst case it loops once more.
787 */
788 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789
Al Viroafc847b2006-02-28 12:51:55 -0500790 if (tsk->io_context)
Louis Rillingb69f2292009-12-04 14:52:42 +0100791 exit_io_context(tsk);
Al Viroafc847b2006-02-28 12:51:55 -0500792
Jens Axboeb92ce552006-04-11 13:52:07 +0200793 if (tsk->splice_pipe)
Al Viro4b8a8f12013-03-21 11:06:46 -0400794 free_pipe_info(tsk->splice_pipe);
Jens Axboeb92ce552006-04-11 13:52:07 +0200795
Eric Dumazet5640f762012-09-23 23:04:42 +0000796 if (tsk->task_frag.page)
797 put_page(tsk->task_frag.page);
798
David Howellse0e81732009-09-02 09:13:40 +0100799 validate_creds_for_do_exit(tsk);
800
Oleg Nesterov4bcb8232014-04-07 15:38:30 -0700801 check_stack_usage();
Coywolf Qi Hunt74072512005-10-30 15:02:47 -0800802 preempt_disable();
Wu Fengguang54848d72011-04-05 13:21:19 -0600803 if (tsk->nr_dirtied)
804 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
Paul E. McKenneyf41d9112009-08-22 13:56:52 -0700805 exit_rcu();
Paul E. McKenney3f95aa82014-08-04 06:10:23 -0700806 TASKS_RCU(__srcu_read_unlock(&tasks_rcu_exit_srcu, tasks_rcu_i));
Yasunori Gotob5740f42012-01-17 17:40:31 +0900807
808 /*
809 * The setting of TASK_RUNNING by try_to_wake_up() may be delayed
810 * when the following two conditions become true.
811 * - There is race condition of mmap_sem (It is acquired by
812 * exit_mm()), and
813 * - SMI occurs before setting TASK_RUNINNG.
814 * (or hypervisor of virtual machine switches to other guest)
815 * As a result, we may become TASK_RUNNING after becoming TASK_DEAD
816 *
817 * To avoid it, we have to wait for releasing tsk->pi_lock which
818 * is held by try_to_wake_up()
819 */
820 smp_mb();
821 raw_spin_unlock_wait(&tsk->pi_lock);
822
Oleg Nesterov55a101f2006-09-29 02:01:10 -0700823 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -0700824 tsk->state = TASK_DEAD;
Tejun Heoa5850422011-11-21 12:32:23 -0800825 tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 schedule();
827 BUG();
828 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -0700829 for (;;)
830 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831}
Russ Anderson012914d2005-04-23 00:08:00 -0700832EXPORT_SYMBOL_GPL(do_exit);
833
Joe Perches9402c952012-01-12 17:17:17 -0800834void complete_and_exit(struct completion *comp, long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
836 if (comp)
837 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -0700838
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 do_exit(code);
840}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841EXPORT_SYMBOL(complete_and_exit);
842
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100843SYSCALL_DEFINE1(exit, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844{
845 do_exit((error_code&0xff)<<8);
846}
847
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848/*
849 * Take down every thread in the group. This is called by fatal signals
850 * as well as by sys_exit_group (below).
851 */
Joe Perches9402c952012-01-12 17:17:17 -0800852void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853do_group_exit(int exit_code)
854{
Oleg Nesterovbfc4b082008-04-30 00:52:36 -0700855 struct signal_struct *sig = current->signal;
856
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
858
Oleg Nesterovbfc4b082008-04-30 00:52:36 -0700859 if (signal_group_exit(sig))
860 exit_code = sig->group_exit_code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 else if (!thread_group_empty(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 struct sighand_struct *const sighand = current->sighand;
Ionut Alexaa0be55d2014-08-08 14:21:18 -0700863
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800865 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 /* Another thread got here before we took the lock. */
867 exit_code = sig->group_exit_code;
868 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800870 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 zap_other_threads(current);
872 }
873 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 }
875
876 do_exit(exit_code);
877 /* NOTREACHED */
878}
879
880/*
881 * this kills every thread in the thread group. Note that any externally
882 * wait4()-ing process will get the correct exit code - even if this
883 * thread is not the thread group leader.
884 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100885SYSCALL_DEFINE1(exit_group, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886{
887 do_group_exit((error_code & 0xff) << 8);
Heiko Carstens2ed7c032009-01-14 14:13:54 +0100888 /* NOTREACHED */
889 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890}
891
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700892struct wait_opts {
893 enum pid_type wo_type;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700894 int wo_flags;
Richard Kennedye1eb1eb2009-06-17 16:27:42 -0700895 struct pid *wo_pid;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700896
897 struct siginfo __user *wo_info;
898 int __user *wo_stat;
899 struct rusage __user *wo_rusage;
900
Oleg Nesterov0b7570e2009-09-23 15:56:46 -0700901 wait_queue_t child_wait;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700902 int notask_error;
903};
904
Oleg Nesterov989264f2009-09-23 15:56:49 -0700905static inline
906struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
Eric W. Biederman161550d2008-02-08 04:19:14 -0800907{
Oleg Nesterov989264f2009-09-23 15:56:49 -0700908 if (type != PIDTYPE_PID)
909 task = task->group_leader;
910 return task->pids[type].pid;
Eric W. Biederman161550d2008-02-08 04:19:14 -0800911}
912
Oleg Nesterov989264f2009-09-23 15:56:49 -0700913static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
Oleg Nesterov5c01ba42009-09-23 15:56:48 -0700914{
915 return wo->wo_type == PIDTYPE_MAX ||
916 task_pid_type(p, wo->wo_type) == wo->wo_pid;
917}
918
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700919static int eligible_child(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920{
Oleg Nesterov5c01ba42009-09-23 15:56:48 -0700921 if (!eligible_pid(wo, p))
922 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 /* Wait for all children (clone and not) if __WALL is set;
924 * otherwise, wait for clone children *only* if __WCLONE is
925 * set; otherwise, wait for non-clone children *only*. (Note:
926 * A "clone" child here is one that reports to its parent
927 * using a signal other than SIGCHLD.) */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700928 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
929 && !(wo->wo_flags & __WALL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
Roland McGrath14dd0b82008-03-30 18:41:25 -0700932 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933}
934
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700935static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
936 pid_t pid, uid_t uid, int why, int status)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700938 struct siginfo __user *infop;
939 int retval = wo->wo_rusage
940 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -0700941
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700943 infop = wo->wo_info;
Vitaly Mayatskikhb6fe2d12009-09-23 15:56:52 -0700944 if (infop) {
945 if (!retval)
946 retval = put_user(SIGCHLD, &infop->si_signo);
947 if (!retval)
948 retval = put_user(0, &infop->si_errno);
949 if (!retval)
950 retval = put_user((short)why, &infop->si_code);
951 if (!retval)
952 retval = put_user(pid, &infop->si_pid);
953 if (!retval)
954 retval = put_user(uid, &infop->si_uid);
955 if (!retval)
956 retval = put_user(status, &infop->si_status);
957 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 if (!retval)
959 retval = pid;
960 return retval;
961}
962
963/*
964 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
965 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
966 * the lock and this task is uninteresting. If we return nonzero, we have
967 * released the lock and the system call should return.
968 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700969static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970{
Oleg Nesterovf6507f82014-12-10 15:54:45 -0800971 int state, retval, status;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800972 pid_t pid = task_pid_vnr(p);
Oleg Nesterov43e13cc2012-05-31 16:26:16 -0700973 uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p));
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700974 struct siginfo __user *infop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700976 if (!likely(wo->wo_flags & WEXITED))
Roland McGrath98abed02008-03-19 19:24:59 -0700977 return 0;
978
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700979 if (unlikely(wo->wo_flags & WNOWAIT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 int exit_code = p->exit_code;
Thiago Farinaf3abd4f2010-03-05 13:42:52 -0800981 int why;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 get_task_struct(p);
984 read_unlock(&tasklist_lock);
Peter Zijlstra1029a2b2014-09-24 10:18:49 +0200985 sched_annotate_sleep();
986
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 if ((exit_code & 0x7f) == 0) {
988 why = CLD_EXITED;
989 status = exit_code >> 8;
990 } else {
991 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
992 status = exit_code & 0x7f;
993 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -0700994 return wait_noreap_copyout(wo, p, pid, uid, why, status);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 }
Oleg Nesterovbefca962009-06-18 16:49:11 -0700996 /*
Oleg Nesterovabd50b32014-04-07 15:38:42 -0700997 * Move the task's state to DEAD/TRACE, only one thread can do this.
998 */
Oleg Nesterovf6507f82014-12-10 15:54:45 -0800999 state = (ptrace_reparented(p) && thread_group_leader(p)) ?
1000 EXIT_TRACE : EXIT_DEAD;
Oleg Nesterovabd50b32014-04-07 15:38:42 -07001001 if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE)
1002 return 0;
Oleg Nesterov986094d2014-12-10 15:54:51 -08001003 /*
1004 * We own this thread, nobody else can reap it.
1005 */
1006 read_unlock(&tasklist_lock);
1007 sched_annotate_sleep();
Oleg Nesterovf6507f82014-12-10 15:54:45 -08001008
Oleg Nesterovabd50b32014-04-07 15:38:42 -07001009 /*
Oleg Nesterovf6507f82014-12-10 15:54:45 -08001010 * Check thread_group_leader() to exclude the traced sub-threads.
Oleg Nesterovbefca962009-06-18 16:49:11 -07001011 */
Oleg Nesterovf6507f82014-12-10 15:54:45 -08001012 if (state == EXIT_DEAD && thread_group_leader(p)) {
Oleg Nesterovf953ccd2014-12-10 15:54:48 -08001013 struct signal_struct *sig = p->signal;
1014 struct signal_struct *psig = current->signal;
Jiri Pirko1f102062009-09-22 16:44:10 -07001015 unsigned long maxrss;
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001016 cputime_t tgutime, tgstime;
Jesper Juhl3795e162006-01-09 20:54:39 -08001017
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 /*
1019 * The resource counters for the group leader are in its
1020 * own task_struct. Those for dead threads in the group
1021 * are in its signal_struct, as are those for the child
1022 * processes it has previously reaped. All these
1023 * accumulate in the parent's signal_struct c* fields.
1024 *
1025 * We don't bother to take a lock here to protect these
Oleg Nesterovf953ccd2014-12-10 15:54:48 -08001026 * p->signal fields because the whole thread group is dead
1027 * and nobody can change them.
1028 *
1029 * psig->stats_lock also protects us from our sub-theads
1030 * which can reap other children at the same time. Until
1031 * we change k_getrusage()-like users to rely on this lock
1032 * we have to take ->siglock as well.
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001033 *
Ionut Alexaa0be55d2014-08-08 14:21:18 -07001034 * We use thread_group_cputime_adjusted() to get times for
1035 * the thread group, which consolidates times for all threads
1036 * in the group including the group leader.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 */
Frederic Weisbeckere80d0a1a2012-11-21 16:26:44 +01001038 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
Oleg Nesterovf953ccd2014-12-10 15:54:48 -08001039 spin_lock_irq(&current->sighand->siglock);
Rik van Riele78c3492014-08-16 13:40:10 -04001040 write_seqlock(&psig->stats_lock);
Martin Schwidefsky64861632011-12-15 14:56:09 +01001041 psig->cutime += tgutime + sig->cutime;
1042 psig->cstime += tgstime + sig->cstime;
Frederic Weisbecker6fac4822012-11-13 14:20:55 +01001043 psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime;
Jesper Juhl3795e162006-01-09 20:54:39 -08001044 psig->cmin_flt +=
1045 p->min_flt + sig->min_flt + sig->cmin_flt;
1046 psig->cmaj_flt +=
1047 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1048 psig->cnvcsw +=
1049 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1050 psig->cnivcsw +=
1051 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001052 psig->cinblock +=
1053 task_io_get_inblock(p) +
1054 sig->inblock + sig->cinblock;
1055 psig->coublock +=
1056 task_io_get_oublock(p) +
1057 sig->oublock + sig->coublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001058 maxrss = max(sig->maxrss, sig->cmaxrss);
1059 if (psig->cmaxrss < maxrss)
1060 psig->cmaxrss = maxrss;
Andrea Righi59954772008-07-27 17:29:15 +02001061 task_io_accounting_add(&psig->ioac, &p->ioac);
1062 task_io_accounting_add(&psig->ioac, &sig->ioac);
Rik van Riele78c3492014-08-16 13:40:10 -04001063 write_sequnlock(&psig->stats_lock);
Oleg Nesterovf953ccd2014-12-10 15:54:48 -08001064 spin_unlock_irq(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 }
1066
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001067 retval = wo->wo_rusage
1068 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1070 ? p->signal->group_exit_code : p->exit_code;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001071 if (!retval && wo->wo_stat)
1072 retval = put_user(status, wo->wo_stat);
1073
1074 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 if (!retval && infop)
1076 retval = put_user(SIGCHLD, &infop->si_signo);
1077 if (!retval && infop)
1078 retval = put_user(0, &infop->si_errno);
1079 if (!retval && infop) {
1080 int why;
1081
1082 if ((status & 0x7f) == 0) {
1083 why = CLD_EXITED;
1084 status >>= 8;
1085 } else {
1086 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1087 status &= 0x7f;
1088 }
1089 retval = put_user((short)why, &infop->si_code);
1090 if (!retval)
1091 retval = put_user(status, &infop->si_status);
1092 }
1093 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001094 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 if (!retval && infop)
David Howellsc69e8d92008-11-14 10:39:19 +11001096 retval = put_user(uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001097 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001098 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001099
Oleg Nesterovb4360692014-04-07 15:38:43 -07001100 if (state == EXIT_TRACE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001102 /* We dropped tasklist, ptracer could die and untrace */
1103 ptrace_unlink(p);
Oleg Nesterovb4360692014-04-07 15:38:43 -07001104
1105 /* If parent wants a zombie, don't release it now */
1106 state = EXIT_ZOMBIE;
1107 if (do_notify_parent(p, p->exit_signal))
1108 state = EXIT_DEAD;
Oleg Nesterovabd50b32014-04-07 15:38:42 -07001109 p->exit_state = state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 write_unlock_irq(&tasklist_lock);
1111 }
Oleg Nesterovabd50b32014-04-07 15:38:42 -07001112 if (state == EXIT_DEAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001114
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 return retval;
1116}
1117
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001118static int *task_stopped_code(struct task_struct *p, bool ptrace)
1119{
1120 if (ptrace) {
Tejun Heo544b2c92011-06-14 11:20:18 +02001121 if (task_is_stopped_or_traced(p) &&
1122 !(p->jobctl & JOBCTL_LISTENING))
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001123 return &p->exit_code;
1124 } else {
1125 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1126 return &p->signal->group_exit_code;
1127 }
1128 return NULL;
1129}
1130
Tejun Heo19e27462011-05-12 10:47:23 +02001131/**
1132 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1133 * @wo: wait options
1134 * @ptrace: is the wait for ptrace
1135 * @p: task to wait for
1136 *
1137 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1138 *
1139 * CONTEXT:
1140 * read_lock(&tasklist_lock), which is released if return value is
1141 * non-zero. Also, grabs and releases @p->sighand->siglock.
1142 *
1143 * RETURNS:
1144 * 0 if wait condition didn't exist and search for other wait conditions
1145 * should continue. Non-zero return, -errno on failure and @p's pid on
1146 * success, implies that tasklist_lock is released and wait condition
1147 * search should terminate.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001149static int wait_task_stopped(struct wait_opts *wo,
1150 int ptrace, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001152 struct siginfo __user *infop;
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001153 int retval, exit_code, *p_code, why;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001154 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001155 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
Oleg Nesterov47918022009-06-17 16:27:39 -07001157 /*
1158 * Traditionally we see ptrace'd stopped tasks regardless of options.
1159 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001160 if (!ptrace && !(wo->wo_flags & WUNTRACED))
Roland McGrath98abed02008-03-19 19:24:59 -07001161 return 0;
1162
Tejun Heo19e27462011-05-12 10:47:23 +02001163 if (!task_stopped_code(p, ptrace))
1164 return 0;
1165
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001166 exit_code = 0;
1167 spin_lock_irq(&p->sighand->siglock);
1168
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001169 p_code = task_stopped_code(p, ptrace);
1170 if (unlikely(!p_code))
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001171 goto unlock_sig;
1172
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001173 exit_code = *p_code;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001174 if (!exit_code)
1175 goto unlock_sig;
1176
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001177 if (!unlikely(wo->wo_flags & WNOWAIT))
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001178 *p_code = 0;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001179
Sasha Levin8ca937a2012-05-17 23:31:39 +02001180 uid = from_kuid_munged(current_user_ns(), task_uid(p));
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001181unlock_sig:
1182 spin_unlock_irq(&p->sighand->siglock);
1183 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 return 0;
1185
1186 /*
1187 * Now we are pretty sure this task is interesting.
1188 * Make sure it doesn't get reaped out from under us while we
1189 * give up the lock and then examine it below. We don't want to
1190 * keep holding onto the tasklist_lock while we call getrusage and
1191 * possibly take page faults for user memory.
1192 */
1193 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001194 pid = task_pid_vnr(p);
Roland McGrathf4700212008-03-24 18:36:23 -07001195 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 read_unlock(&tasklist_lock);
Peter Zijlstra1029a2b2014-09-24 10:18:49 +02001197 sched_annotate_sleep();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001199 if (unlikely(wo->wo_flags & WNOWAIT))
1200 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001202 retval = wo->wo_rusage
1203 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1204 if (!retval && wo->wo_stat)
1205 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1206
1207 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208 if (!retval && infop)
1209 retval = put_user(SIGCHLD, &infop->si_signo);
1210 if (!retval && infop)
1211 retval = put_user(0, &infop->si_errno);
1212 if (!retval && infop)
Roland McGrath6efcae42008-03-08 11:41:22 -08001213 retval = put_user((short)why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 if (!retval && infop)
1215 retval = put_user(exit_code, &infop->si_status);
1216 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001217 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001219 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001221 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 put_task_struct(p);
1223
1224 BUG_ON(!retval);
1225 return retval;
1226}
1227
1228/*
1229 * Handle do_wait work for one task in a live, non-stopped state.
1230 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1231 * the lock and this task is uninteresting. If we return nonzero, we have
1232 * released the lock and the system call should return.
1233 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001234static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235{
1236 int retval;
1237 pid_t pid;
1238 uid_t uid;
1239
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001240 if (!unlikely(wo->wo_flags & WCONTINUED))
Roland McGrath98abed02008-03-19 19:24:59 -07001241 return 0;
1242
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1244 return 0;
1245
1246 spin_lock_irq(&p->sighand->siglock);
1247 /* Re-check with the lock held. */
1248 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1249 spin_unlock_irq(&p->sighand->siglock);
1250 return 0;
1251 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001252 if (!unlikely(wo->wo_flags & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
Sasha Levin8ca937a2012-05-17 23:31:39 +02001254 uid = from_kuid_munged(current_user_ns(), task_uid(p));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 spin_unlock_irq(&p->sighand->siglock);
1256
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001257 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 get_task_struct(p);
1259 read_unlock(&tasklist_lock);
Peter Zijlstra1029a2b2014-09-24 10:18:49 +02001260 sched_annotate_sleep();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001262 if (!wo->wo_info) {
1263 retval = wo->wo_rusage
1264 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001266 if (!retval && wo->wo_stat)
1267 retval = put_user(0xffff, wo->wo_stat);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001269 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 } else {
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001271 retval = wait_noreap_copyout(wo, p, pid, uid,
1272 CLD_CONTINUED, SIGCONT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 BUG_ON(retval == 0);
1274 }
1275
1276 return retval;
1277}
1278
Roland McGrath98abed02008-03-19 19:24:59 -07001279/*
1280 * Consider @p for a wait by @parent.
1281 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001282 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001283 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1284 * Returns zero if the search for a child should continue;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001285 * then ->notask_error is 0 if @p is an eligible child,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001286 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001287 */
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001288static int wait_consider_task(struct wait_opts *wo, int ptrace,
1289 struct task_struct *p)
Roland McGrath98abed02008-03-19 19:24:59 -07001290{
Oleg Nesterov3245d6a2015-01-08 14:32:12 -08001291 /*
1292 * We can race with wait_task_zombie() from another thread.
1293 * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition
1294 * can't confuse the checks below.
1295 */
1296 int exit_state = ACCESS_ONCE(p->exit_state);
Oleg Nesterovb3ab0312014-04-07 15:38:45 -07001297 int ret;
1298
Oleg Nesterov3245d6a2015-01-08 14:32:12 -08001299 if (unlikely(exit_state == EXIT_DEAD))
Oleg Nesterovb3ab0312014-04-07 15:38:45 -07001300 return 0;
1301
1302 ret = eligible_child(wo, p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001303 if (!ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001304 return ret;
1305
Oleg Nesterova2322e12009-09-23 15:56:45 -07001306 ret = security_task_wait(p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001307 if (unlikely(ret < 0)) {
1308 /*
1309 * If we have not yet seen any eligible child,
1310 * then let this error code replace -ECHILD.
1311 * A permission error will give the user a clue
1312 * to look for security policy problems, rather
1313 * than for mysterious wait bugs.
1314 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001315 if (wo->notask_error)
1316 wo->notask_error = ret;
Oleg Nesterov78a3d9d2009-04-29 18:01:23 +02001317 return 0;
Roland McGrath14dd0b82008-03-30 18:41:25 -07001318 }
1319
Oleg Nesterov3245d6a2015-01-08 14:32:12 -08001320 if (unlikely(exit_state == EXIT_TRACE)) {
Oleg Nesterov50b8d252012-01-04 17:29:02 +01001321 /*
Oleg Nesterovabd50b32014-04-07 15:38:42 -07001322 * ptrace == 0 means we are the natural parent. In this case
1323 * we should clear notask_error, debugger will notify us.
Oleg Nesterov50b8d252012-01-04 17:29:02 +01001324 */
Oleg Nesterovabd50b32014-04-07 15:38:42 -07001325 if (likely(!ptrace))
Oleg Nesterov50b8d252012-01-04 17:29:02 +01001326 wo->notask_error = 0;
Roland McGrath98abed02008-03-19 19:24:59 -07001327 return 0;
Oleg Nesterov50b8d252012-01-04 17:29:02 +01001328 }
Roland McGrath98abed02008-03-19 19:24:59 -07001329
Oleg Nesterov377d75d2014-04-07 15:38:47 -07001330 if (likely(!ptrace) && unlikely(p->ptrace)) {
1331 /*
1332 * If it is traced by its real parent's group, just pretend
1333 * the caller is ptrace_do_wait() and reap this child if it
1334 * is zombie.
1335 *
1336 * This also hides group stop state from real parent; otherwise
1337 * a single stop can be reported twice as group and ptrace stop.
1338 * If a ptracer wants to distinguish these two events for its
1339 * own children it should create a separate process which takes
1340 * the role of real parent.
1341 */
1342 if (!ptrace_reparented(p))
1343 ptrace = 1;
1344 }
1345
Tejun Heo9b84cca2011-03-23 10:37:01 +01001346 /* slay zombie? */
Oleg Nesterov3245d6a2015-01-08 14:32:12 -08001347 if (exit_state == EXIT_ZOMBIE) {
Tejun Heo9b84cca2011-03-23 10:37:01 +01001348 /* we don't reap group leaders with subthreads */
Oleg Nesterov7c733eb2014-04-07 15:38:49 -07001349 if (!delay_group_leader(p)) {
1350 /*
1351 * A zombie ptracee is only visible to its ptracer.
1352 * Notification and reaping will be cascaded to the
1353 * real parent when the ptracer detaches.
1354 */
1355 if (unlikely(ptrace) || likely(!p->ptrace))
1356 return wait_task_zombie(wo, p);
1357 }
Roland McGrath98abed02008-03-19 19:24:59 -07001358
Tejun Heo9b84cca2011-03-23 10:37:01 +01001359 /*
1360 * Allow access to stopped/continued state via zombie by
1361 * falling through. Clearing of notask_error is complex.
1362 *
1363 * When !@ptrace:
1364 *
1365 * If WEXITED is set, notask_error should naturally be
1366 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1367 * so, if there are live subthreads, there are events to
1368 * wait for. If all subthreads are dead, it's still safe
1369 * to clear - this function will be called again in finite
1370 * amount time once all the subthreads are released and
1371 * will then return without clearing.
1372 *
1373 * When @ptrace:
1374 *
1375 * Stopped state is per-task and thus can't change once the
1376 * target task dies. Only continued and exited can happen.
1377 * Clear notask_error if WCONTINUED | WEXITED.
1378 */
1379 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1380 wo->notask_error = 0;
1381 } else {
1382 /*
1383 * @p is alive and it's gonna stop, continue or exit, so
1384 * there always is something to wait for.
1385 */
1386 wo->notask_error = 0;
1387 }
Roland McGrath98abed02008-03-19 19:24:59 -07001388
1389 /*
Tejun Heo45cb24a2011-03-23 10:37:01 +01001390 * Wait for stopped. Depending on @ptrace, different stopped state
1391 * is used and the two don't interact with each other.
Roland McGrath98abed02008-03-19 19:24:59 -07001392 */
Tejun Heo19e27462011-05-12 10:47:23 +02001393 ret = wait_task_stopped(wo, ptrace, p);
1394 if (ret)
1395 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001396
Tejun Heo45cb24a2011-03-23 10:37:01 +01001397 /*
1398 * Wait for continued. There's only one continued state and the
1399 * ptracer can consume it which can confuse the real parent. Don't
1400 * use WCONTINUED from ptracer. You don't need or want it.
1401 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001402 return wait_task_continued(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001403}
1404
1405/*
1406 * Do the work of do_wait() for one thread in the group, @tsk.
1407 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001408 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001409 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1410 * Returns zero if the search for a child should continue; then
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001411 * ->notask_error is 0 if there were any eligible children,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001412 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001413 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001414static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001415{
1416 struct task_struct *p;
1417
1418 list_for_each_entry(p, &tsk->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001419 int ret = wait_consider_task(wo, 0, p);
Ionut Alexaa0be55d2014-08-08 14:21:18 -07001420
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001421 if (ret)
1422 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001423 }
1424
1425 return 0;
1426}
1427
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001428static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001429{
1430 struct task_struct *p;
1431
Roland McGrathf4700212008-03-24 18:36:23 -07001432 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001433 int ret = wait_consider_task(wo, 1, p);
Ionut Alexaa0be55d2014-08-08 14:21:18 -07001434
Roland McGrathf4700212008-03-24 18:36:23 -07001435 if (ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001436 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001437 }
1438
1439 return 0;
1440}
1441
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001442static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1443 int sync, void *key)
1444{
1445 struct wait_opts *wo = container_of(wait, struct wait_opts,
1446 child_wait);
1447 struct task_struct *p = key;
1448
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001449 if (!eligible_pid(wo, p))
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001450 return 0;
1451
Oleg Nesterovb4fe5182009-09-23 15:56:47 -07001452 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1453 return 0;
1454
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001455 return default_wake_function(wait, mode, sync, key);
1456}
1457
Oleg Nesterova7f07652009-09-23 15:56:44 -07001458void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1459{
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001460 __wake_up_sync_key(&parent->signal->wait_chldexit,
1461 TASK_INTERRUPTIBLE, 1, p);
Oleg Nesterova7f07652009-09-23 15:56:44 -07001462}
1463
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001464static long do_wait(struct wait_opts *wo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466 struct task_struct *tsk;
Roland McGrath98abed02008-03-19 19:24:59 -07001467 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001469 trace_sched_process_wait(wo->wo_pid);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001470
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001471 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1472 wo->child_wait.private = current;
1473 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474repeat:
Roland McGrath98abed02008-03-19 19:24:59 -07001475 /*
1476 * If there is nothing that can match our critiera just get out.
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001477 * We will clear ->notask_error to zero if we see any child that
1478 * might later match our criteria, even if we are not able to reap
1479 * it yet.
Roland McGrath98abed02008-03-19 19:24:59 -07001480 */
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001481 wo->notask_error = -ECHILD;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001482 if ((wo->wo_type < PIDTYPE_MAX) &&
1483 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001484 goto notask;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001485
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001486 set_current_state(TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 read_lock(&tasklist_lock);
1488 tsk = current;
1489 do {
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001490 retval = do_wait_thread(wo, tsk);
1491 if (retval)
Roland McGrath98abed02008-03-19 19:24:59 -07001492 goto end;
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001493
1494 retval = ptrace_do_wait(wo, tsk);
1495 if (retval)
1496 goto end;
Roland McGrath98abed02008-03-19 19:24:59 -07001497
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001498 if (wo->wo_flags & __WNOTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 break;
Oleg Nesterova3f6dfb2009-06-17 16:27:41 -07001500 } while_each_thread(current, tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001502
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001503notask:
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001504 retval = wo->notask_error;
1505 if (!retval && !(wo->wo_flags & WNOHANG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 retval = -ERESTARTSYS;
Roland McGrath98abed02008-03-19 19:24:59 -07001507 if (!signal_pending(current)) {
1508 schedule();
1509 goto repeat;
1510 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512end:
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001513 __set_current_state(TASK_RUNNING);
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001514 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 return retval;
1516}
1517
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001518SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1519 infop, int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001521 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001522 struct pid *pid = NULL;
1523 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 long ret;
1525
1526 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1527 return -EINVAL;
1528 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1529 return -EINVAL;
1530
1531 switch (which) {
1532 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001533 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 break;
1535 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001536 type = PIDTYPE_PID;
1537 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 return -EINVAL;
1539 break;
1540 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001541 type = PIDTYPE_PGID;
1542 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 break;
1545 default:
1546 return -EINVAL;
1547 }
1548
Eric W. Biederman161550d2008-02-08 04:19:14 -08001549 if (type < PIDTYPE_MAX)
1550 pid = find_get_pid(upid);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001551
1552 wo.wo_type = type;
1553 wo.wo_pid = pid;
1554 wo.wo_flags = options;
1555 wo.wo_info = infop;
1556 wo.wo_stat = NULL;
1557 wo.wo_rusage = ru;
1558 ret = do_wait(&wo);
Vitaly Mayatskikhdfe16df2009-09-23 15:56:51 -07001559
1560 if (ret > 0) {
1561 ret = 0;
1562 } else if (infop) {
1563 /*
1564 * For a WNOHANG return, clear out all the fields
1565 * we would set so the user can easily tell the
1566 * difference.
1567 */
1568 if (!ret)
1569 ret = put_user(0, &infop->si_signo);
1570 if (!ret)
1571 ret = put_user(0, &infop->si_errno);
1572 if (!ret)
1573 ret = put_user(0, &infop->si_code);
1574 if (!ret)
1575 ret = put_user(0, &infop->si_pid);
1576 if (!ret)
1577 ret = put_user(0, &infop->si_uid);
1578 if (!ret)
1579 ret = put_user(0, &infop->si_status);
1580 }
1581
Eric W. Biederman161550d2008-02-08 04:19:14 -08001582 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 return ret;
1584}
1585
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001586SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1587 int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001589 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001590 struct pid *pid = NULL;
1591 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 long ret;
1593
1594 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1595 __WNOTHREAD|__WCLONE|__WALL))
1596 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001597
1598 if (upid == -1)
1599 type = PIDTYPE_MAX;
1600 else if (upid < 0) {
1601 type = PIDTYPE_PGID;
1602 pid = find_get_pid(-upid);
1603 } else if (upid == 0) {
1604 type = PIDTYPE_PGID;
Oleg Nesterov2ae448e2009-04-02 16:58:36 -07001605 pid = get_task_pid(current, PIDTYPE_PGID);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001606 } else /* upid > 0 */ {
1607 type = PIDTYPE_PID;
1608 pid = find_get_pid(upid);
1609 }
1610
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001611 wo.wo_type = type;
1612 wo.wo_pid = pid;
1613 wo.wo_flags = options | WEXITED;
1614 wo.wo_info = NULL;
1615 wo.wo_stat = stat_addr;
1616 wo.wo_rusage = ru;
1617 ret = do_wait(&wo);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001618 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 return ret;
1621}
1622
1623#ifdef __ARCH_WANT_SYS_WAITPID
1624
1625/*
1626 * sys_waitpid() remains for compatibility. waitpid() should be
1627 * implemented by calling sys_wait4() from libc.a.
1628 */
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001629SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630{
1631 return sys_wait4(pid, stat_addr, options, NULL);
1632}
1633
1634#endif