blob: 3d93325e0b1af77ebb2d3fa48a4eff2f24a622fc [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/binfmts.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070024#include <linux/nsproxy.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080025#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/ptrace.h>
27#include <linux/profile.h>
28#include <linux/mount.h>
29#include <linux/proc_fs.h>
Eric W. Biederman49d769d2007-05-09 02:34:33 -070030#include <linux/kthread.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/mempolicy.h>
Shailabh Nagarc7572492006-07-14 00:24:40 -070032#include <linux/taskstats_kern.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070033#include <linux/delayacct.h>
Rafael J. Wysocki83144182007-07-17 04:03:35 -070034#include <linux/freezer.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
Adrian Bunk408b6642005-05-01 08:59:29 -070062static void exit_mm(struct task_struct * tsk);
63
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--;
67 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);
Linus Torvalds1da177e2005-04-16 15:20:36 -070077}
78
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080079/*
80 * This function expects the tasklist_lock write-locked.
81 */
82static void __exit_signal(struct task_struct *tsk)
83{
84 struct signal_struct *sig = tsk->signal;
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070085 bool group_dead = thread_group_leader(tsk);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080086 struct sighand_struct *sighand;
Oleg Nesterov4ada8562010-05-26 14:43:17 -070087 struct tty_struct *uninitialized_var(tty);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080088
Paul E. McKenneyd11c5632010-02-22 17:04:50 -080089 sighand = rcu_dereference_check(tsk->sighand,
Paul E. McKenneydb1466b2010-03-03 07:46:56 -080090 lockdep_tasklist_lock_is_held());
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080091 spin_lock(&sighand->siglock);
92
93 posix_cpu_timers_exit(tsk);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070094 if (group_dead) {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080095 posix_cpu_timers_exit_group(tsk);
Oleg Nesterov4ada8562010-05-26 14:43:17 -070096 tty = sig->tty;
97 sig->tty = NULL;
Oleg Nesterov4a599942010-05-26 14:43:12 -070098 } else {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080099 /*
Oleg Nesterove0a70212010-11-05 16:53:42 +0100100 * 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 Nesterov6a14c5c2006-03-28 16:11:18 -0800108 * If there is any task waiting for the group exit
109 * then notify it:
110 */
Oleg Nesterovd3441932010-05-26 14:43:11 -0700111 if (sig->notify_count > 0 && !--sig->notify_count)
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800112 wake_up_process(sig->group_exit_task);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700113
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800114 if (tsk == sig->curr_target)
115 sig->curr_target = next_thread(tsk);
116 /*
117 * Accumulate here the counters for all threads but the
118 * group leader as they die, so they can be added into
119 * the process-wide totals when those are taken.
120 * The group leader stays around as a zombie as long
121 * as there are other threads. When it gets reaped,
122 * the exit.c code will add its counts into these totals.
123 * We won't ever get here for the group leader, since it
124 * will have been the last reference on the signal_struct.
125 */
Martin Schwidefsky64861632011-12-15 14:56:09 +0100126 sig->utime += tsk->utime;
127 sig->stime += tsk->stime;
128 sig->gtime += tsk->gtime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800129 sig->min_flt += tsk->min_flt;
130 sig->maj_flt += tsk->maj_flt;
131 sig->nvcsw += tsk->nvcsw;
132 sig->nivcsw += tsk->nivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700133 sig->inblock += task_io_get_inblock(tsk);
134 sig->oublock += task_io_get_oublock(tsk);
Andrea Righi59954772008-07-27 17:29:15 +0200135 task_io_accounting_add(&sig->ioac, &tsk->ioac);
Peter Zijlstra32bd6712009-02-05 12:24:15 +0100136 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800137 }
138
Oleg Nesterovb3ac0222010-05-26 14:43:24 -0700139 sig->nr_threads--;
Oleg Nesterovd40e48e2010-05-26 14:43:19 -0700140 __unhash_process(tsk, group_dead);
Oleg Nesterov58767002006-03-28 16:11:20 -0800141
Oleg Nesterovda7978b2008-05-23 13:04:41 -0700142 /*
143 * Do this under ->siglock, we can race with another thread
144 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
145 */
146 flush_sigqueue(&tsk->pending);
Oleg Nesterova7e53282006-03-28 16:11:27 -0800147 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800148 spin_unlock(&sighand->siglock);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800149
Oleg Nesterova7e53282006-03-28 16:11:27 -0800150 __cleanup_sighand(sighand);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800151 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -0700152 if (group_dead) {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800153 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov4ada8562010-05-26 14:43:17 -0700154 tty_kref_put(tty);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800155 }
156}
157
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800158static void delayed_put_task_struct(struct rcu_head *rhp)
159{
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400160 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
161
Peter Zijlstra4e231c72010-09-09 21:01:59 +0200162 perf_event_delayed_put(tsk);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400163 trace_sched_process_free(tsk);
164 put_task_struct(tsk);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800165}
166
Roland McGrathf4700212008-03-24 18:36:23 -0700167
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168void release_task(struct task_struct * p)
169{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700170 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800172repeat:
David Howellsc69e8d92008-11-14 10:39:19 +1100173 /* don't need to get the RCU readlock here - the process is dead and
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800174 * can't be modifying its own credentials. But shut RCU-lockdep up */
175 rcu_read_lock();
David Howellsc69e8d92008-11-14 10:39:19 +1100176 atomic_dec(&__task_cred(p)->user->processes);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800177 rcu_read_unlock();
David Howellsc69e8d92008-11-14 10:39:19 +1100178
Pavel Emelyanov60347f62007-10-18 23:40:03 -0700179 proc_flush_task(p);
Ingo Molnar02030262009-05-17 11:24:08 +0200180
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 write_lock_irq(&tasklist_lock);
Tejun Heoa288eec2011-06-17 16:50:37 +0200182 ptrace_release_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 __exit_signal(p);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -0800184
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 /*
186 * If we are the last non-leader member of the thread
187 * group, and the leader is zombie, then notify the
188 * group leader's parent process. (if it wants notification.)
189 */
190 zap_leader = 0;
191 leader = p->group_leader;
192 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 /*
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 Torvalds1da177e2005-04-16 15:20:36 -0700197 */
Oleg Nesterov86773472011-06-22 23:09:09 +0200198 zap_leader = do_notify_parent(leader, leader->exit_signal);
Roland McGrathdae33572008-07-25 19:45:48 -0700199 if (zap_leader)
200 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 }
202
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 release_thread(p);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800205 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206
207 p = leader;
208 if (unlikely(zap_leader))
209 goto repeat;
210}
211
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212/*
213 * This checks not only the pgrp, but falls back on the pid if no
214 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
215 * without this...
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800216 *
217 * The caller must hold rcu lock or the tasklist lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 */
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800219struct pid *session_of_pgrp(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220{
221 struct task_struct *p;
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800222 struct pid *sid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800224 p = pid_task(pgrp, PIDTYPE_PGID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800225 if (p == NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800226 p = pid_task(pgrp, PIDTYPE_PID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800227 if (p != NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800228 sid = task_session(p);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800229
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 return sid;
231}
232
233/*
234 * Determine if a process group is "orphaned", according to the POSIX
235 * definition in 2.2.2.52. Orphaned process groups are not to be affected
236 * by terminal-generated stop signals. Newly orphaned process groups are
237 * to receive a SIGHUP and a SIGCONT.
238 *
239 * "I ask you, have you ever known what it is to be an orphan?"
240 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800241static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242{
243 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800245 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300246 if ((p == ignored_task) ||
247 (p->exit_state && thread_group_empty(p)) ||
248 is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 continue;
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300250
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800251 if (task_pgrp(p->real_parent) != pgrp &&
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300252 task_session(p->real_parent) == task_session(p))
253 return 0;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800254 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300255
256 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257}
258
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800259int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260{
261 int retval;
262
263 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800264 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 read_unlock(&tasklist_lock);
266
267 return retval;
268}
269
Oleg Nesterov961c4672011-07-07 21:33:54 +0200270static bool has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 struct task_struct *p;
273
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800274 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov961c4672011-07-07 21:33:54 +0200275 if (p->signal->flags & SIGNAL_STOP_STOPPED)
276 return true;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800277 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov961c4672011-07-07 21:33:54 +0200278
279 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280}
281
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300282/*
283 * Check to see if any process groups have become orphaned as
284 * a result of our exiting, and if they have any stopped jobs,
285 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
286 */
287static void
288kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
289{
290 struct pid *pgrp = task_pgrp(tsk);
291 struct task_struct *ignored_task = tsk;
292
293 if (!parent)
294 /* exit: our father is in a different pgrp than
295 * we are and we were the only connection outside.
296 */
297 parent = tsk->real_parent;
298 else
299 /* reparent: our child is in a different pgrp than
300 * we are, and it was the only connection outside.
301 */
302 ignored_task = NULL;
303
304 if (task_pgrp(parent) != pgrp &&
305 task_session(parent) == task_session(tsk) &&
306 will_become_orphaned_pgrp(pgrp, ignored_task) &&
307 has_stopped_jobs(pgrp)) {
308 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
309 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
310 }
311}
312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313/**
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700314 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 *
316 * If a kernel thread is launched as a result of a system call, or if
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700317 * it ever exits, it should generally reparent itself to kthreadd so it
318 * isn't in the way of other processes and is correctly cleaned up on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 *
320 * The various task state such as scheduling policy and priority may have
321 * been inherited from a user process, so we reset them to sane values here.
322 *
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700323 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700325static void reparent_to_kthreadd(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326{
327 write_lock_irq(&tasklist_lock);
328
329 ptrace_unlink(current);
330 /* Reparent to init */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700331 current->real_parent = current->parent = kthreadd_task;
Roland McGrathf4700212008-03-24 18:36:23 -0700332 list_move_tail(&current->sibling, &current->real_parent->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
334 /* Set the exit signal to SIGCHLD so we signal init on exit */
335 current->exit_signal = SIGCHLD;
336
Ingo Molnare05606d2007-07-09 18:51:59 +0200337 if (task_nice(current) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 set_user_nice(current, 0);
339 /* cpus_allowed? */
340 /* rt_priority? */
341 /* signals? */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 memcpy(current->signal->rlim, init_task.signal->rlim,
343 sizeof(current->signal->rlim));
David Howellsd84f4f92008-11-14 10:39:23 +1100344
345 atomic_inc(&init_cred.usage);
346 commit_creds(&init_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348}
349
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800350void __set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351{
Oren Laadane19f2472006-01-08 01:03:58 -0800352 struct task_struct *curr = current->group_leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353
Christian Borntraeger0d0df592009-10-26 16:49:34 -0700354 if (task_session(curr) != pid)
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700355 change_pid(curr, PIDTYPE_SID, pid);
Oleg Nesterov1b0f7ff2009-04-02 16:58:39 -0700356
357 if (task_pgrp(curr) != pid)
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700358 change_pid(curr, PIDTYPE_PGID, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359}
360
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800361static void set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362{
363 write_lock_irq(&tasklist_lock);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800364 __set_special_pids(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 write_unlock_irq(&tasklist_lock);
366}
367
368/*
Oleg Nesterov87245132009-06-17 16:27:23 -0700369 * Let kernel threads use this to say that they allow a certain signal.
370 * Must not be used if kthread was cloned with CLONE_SIGHAND.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 */
372int allow_signal(int sig)
373{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700374 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 return -EINVAL;
376
377 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov87245132009-06-17 16:27:23 -0700378 /* This is only needed for daemonize()'ed kthreads */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 sigdelset(&current->blocked, sig);
Oleg Nesterov87245132009-06-17 16:27:23 -0700380 /*
381 * Kernel threads handle their own signals. Let the signal code
382 * know it'll be handled, so that they don't get converted to
383 * SIGKILL or just silently dropped.
384 */
385 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 recalc_sigpending();
387 spin_unlock_irq(&current->sighand->siglock);
388 return 0;
389}
390
391EXPORT_SYMBOL(allow_signal);
392
393int disallow_signal(int sig)
394{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700395 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 return -EINVAL;
397
398 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov10ab8252007-05-09 02:34:37 -0700399 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 recalc_sigpending();
401 spin_unlock_irq(&current->sighand->siglock);
402 return 0;
403}
404
405EXPORT_SYMBOL(disallow_signal);
406
407/*
408 * Put all the gunge required to become a kernel thread without
409 * attached user resources in one place where it belongs.
410 */
411
412void daemonize(const char *name, ...)
413{
414 va_list args;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 sigset_t blocked;
416
417 va_start(args, name);
418 vsnprintf(current->comm, sizeof(current->comm), name, args);
419 va_end(args);
420
421 /*
422 * If we were started as result of loading a module, close all of the
423 * user space pages. We don't need them, and if we didn't close them
424 * they would be locked into memory.
425 */
426 exit_mm(current);
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700427 /*
Marcos Paulo de Souza37f08be2012-02-21 23:57:47 +0100428 * We don't want to get frozen, in case system-wide hibernation
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700429 * or suspend transition begins right now.
430 */
Oleg Nesterov7b34e422008-07-25 01:47:37 -0700431 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800433 if (current->nsproxy != &init_nsproxy) {
434 get_nsproxy(&init_nsproxy);
435 switch_task_namespaces(current, &init_nsproxy);
436 }
Oleg Nesterov297bd422008-02-08 04:19:10 -0800437 set_special_pids(&init_struct_pid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -0800438 proc_clear_tty(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439
440 /* Block and flush all signals */
441 sigfillset(&blocked);
442 sigprocmask(SIG_BLOCK, &blocked, NULL);
443 flush_signals(current);
444
445 /* Become as one with the init task */
446
Al Viro3e93cd62009-03-29 19:00:13 -0400447 daemonize_fs_struct();
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700448 exit_files(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 current->files = init_task.files;
450 atomic_inc(&current->files->count);
451
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700452 reparent_to_kthreadd();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453}
454
455EXPORT_SYMBOL(daemonize);
456
Arjan van de Ven858119e2006-01-14 13:20:43 -0800457static void close_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458{
459 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700460 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461
462 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700463
464 /*
465 * It is safe to dereference the fd table without RCU or
466 * ->file_lock because this is the last reference to the
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800467 * files structure. But use RCU to shut RCU-lockdep up.
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700468 */
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800469 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700470 fdt = files_fdtable(files);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800471 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 for (;;) {
473 unsigned long set;
474 i = j * __NFDBITS;
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800475 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 break;
David Howells1fd36ad2012-02-16 17:49:54 +0000477 set = fdt->open_fds[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 while (set) {
479 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700480 struct file * file = xchg(&fdt->fd[i], NULL);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800481 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 filp_close(file, files);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800483 cond_resched();
484 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 }
486 i++;
487 set >>= 1;
488 }
489 }
490}
491
492struct files_struct *get_files_struct(struct task_struct *task)
493{
494 struct files_struct *files;
495
496 task_lock(task);
497 files = task->files;
498 if (files)
499 atomic_inc(&files->count);
500 task_unlock(task);
501
502 return files;
503}
504
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800505void put_files_struct(struct files_struct *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700507 struct fdtable *fdt;
508
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 if (atomic_dec_and_test(&files->count)) {
510 close_files(files);
511 /*
512 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700513 * If the fdtable was embedded, pass files for freeing
514 * at the end of the RCU grace period. Otherwise,
515 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 */
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800517 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700518 fdt = files_fdtable(files);
Vadim Lobanov4fd45812006-12-10 02:21:17 -0800519 if (fdt != &files->fdtab)
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700520 kmem_cache_free(files_cachep, files);
Vadim Lobanov01b2d932006-12-22 01:10:43 -0800521 free_fdtable(fdt);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800522 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 }
524}
525
Al Viro3b125382008-04-22 05:31:30 -0400526void reset_files_struct(struct files_struct *files)
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700527{
Al Viro3b125382008-04-22 05:31:30 -0400528 struct task_struct *tsk = current;
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700529 struct files_struct *old;
530
531 old = tsk->files;
532 task_lock(tsk);
533 tsk->files = files;
534 task_unlock(tsk);
535 put_files_struct(old);
536}
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700537
Al Viro1ec7f1d2008-04-22 05:35:42 -0400538void exit_files(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539{
540 struct files_struct * files = tsk->files;
541
542 if (files) {
543 task_lock(tsk);
544 tsk->files = NULL;
545 task_unlock(tsk);
546 put_files_struct(files);
547 }
548}
549
Balbir Singhcf475ad2008-04-29 01:00:16 -0700550#ifdef CONFIG_MM_OWNER
551/*
KAMEZAWA Hiroyuki733eda72011-06-15 15:08:43 -0700552 * A task is exiting. If it owned this mm, find a new owner for the mm.
Balbir Singhcf475ad2008-04-29 01:00:16 -0700553 */
Balbir Singhcf475ad2008-04-29 01:00:16 -0700554void mm_update_next_owner(struct mm_struct *mm)
555{
556 struct task_struct *c, *g, *p = current;
557
558retry:
KAMEZAWA Hiroyuki733eda72011-06-15 15:08:43 -0700559 /*
560 * If the exiting or execing task is not the owner, it's
561 * someone else's problem.
562 */
563 if (mm->owner != p)
Balbir Singhcf475ad2008-04-29 01:00:16 -0700564 return;
KAMEZAWA Hiroyuki733eda72011-06-15 15:08:43 -0700565 /*
566 * The current owner is exiting/execing and there are no other
567 * candidates. Do not leave the mm pointing to a possibly
568 * freed task structure.
569 */
570 if (atomic_read(&mm->mm_users) <= 1) {
571 mm->owner = NULL;
572 return;
573 }
Balbir Singhcf475ad2008-04-29 01:00:16 -0700574
575 read_lock(&tasklist_lock);
576 /*
577 * Search in the children
578 */
579 list_for_each_entry(c, &p->children, sibling) {
580 if (c->mm == mm)
581 goto assign_new_owner;
582 }
583
584 /*
585 * Search in the siblings
586 */
Oleg Nesterovdea33cf2009-06-17 16:27:29 -0700587 list_for_each_entry(c, &p->real_parent->children, sibling) {
Balbir Singhcf475ad2008-04-29 01:00:16 -0700588 if (c->mm == mm)
589 goto assign_new_owner;
590 }
591
592 /*
593 * Search through everything else. We should not get
594 * here often
595 */
596 do_each_thread(g, c) {
597 if (c->mm == mm)
598 goto assign_new_owner;
599 } while_each_thread(g, c);
600
601 read_unlock(&tasklist_lock);
Balbir Singh31a78f22008-09-28 23:09:31 +0100602 /*
603 * We found no owner yet mm_users > 1: this implies that we are
604 * most likely racing with swapoff (try_to_unuse()) or /proc or
Hugh Dickinse5991372009-01-06 14:39:22 -0800605 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
Balbir Singh31a78f22008-09-28 23:09:31 +0100606 */
Balbir Singh31a78f22008-09-28 23:09:31 +0100607 mm->owner = NULL;
Balbir Singhcf475ad2008-04-29 01:00:16 -0700608 return;
609
610assign_new_owner:
611 BUG_ON(c == p);
612 get_task_struct(c);
613 /*
614 * The task_lock protects c->mm from changing.
615 * We always want mm->owner->mm == mm
616 */
617 task_lock(c);
Hugh Dickinse5991372009-01-06 14:39:22 -0800618 /*
619 * Delay read_unlock() till we have the task_lock()
620 * to ensure that c does not slip away underneath us
621 */
622 read_unlock(&tasklist_lock);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700623 if (c->mm != mm) {
624 task_unlock(c);
625 put_task_struct(c);
626 goto retry;
627 }
Balbir Singhcf475ad2008-04-29 01:00:16 -0700628 mm->owner = c;
629 task_unlock(c);
630 put_task_struct(c);
631}
632#endif /* CONFIG_MM_OWNER */
633
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634/*
635 * Turn us into a lazy TLB process if we
636 * aren't already..
637 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700638static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639{
640 struct mm_struct *mm = tsk->mm;
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700641 struct core_state *core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642
643 mm_release(tsk, mm);
644 if (!mm)
645 return;
646 /*
647 * Serialize with any possible pending coredump.
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700648 * We must hold mmap_sem around checking core_state
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 * and clearing tsk->mm. The core-inducing thread
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700650 * will increment ->nr_threads for each thread in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 * group with ->mm != NULL.
652 */
653 down_read(&mm->mmap_sem);
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700654 core_state = mm->core_state;
655 if (core_state) {
656 struct core_thread self;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 up_read(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700659 self.task = tsk;
660 self.next = xchg(&core_state->dumper.next, &self);
661 /*
662 * Implies mb(), the result of xchg() must be visible
663 * to core_state->dumper.
664 */
665 if (atomic_dec_and_test(&core_state->nr_threads))
666 complete(&core_state->startup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Oleg Nesterova94e2d42008-07-25 01:47:46 -0700668 for (;;) {
669 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
670 if (!self.task) /* see coredump_finish() */
671 break;
672 schedule();
673 }
674 __set_task_state(tsk, TASK_RUNNING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 down_read(&mm->mmap_sem);
676 }
677 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700678 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 /* more a memory barrier than a real lock */
680 task_lock(tsk);
681 tsk->mm = NULL;
682 up_read(&mm->mmap_sem);
683 enter_lazy_tlb(mm, current);
684 task_unlock(tsk);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700685 mm_update_next_owner(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 mmput(mm);
687}
688
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689/*
Lennart Poetteringebec18a2012-03-23 15:01:54 -0700690 * When we die, we re-parent all our children, and try to:
691 * 1. give them to another thread in our thread group, if such a member exists
692 * 2. give it to the first ancestor process which prctl'd itself as a
693 * child_subreaper for its children (like a service manager)
694 * 3. give it to the init process (PID 1) in our pid namespace
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 */
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700696static struct task_struct *find_new_reaper(struct task_struct *father)
Namhyung Kimd16e15f2010-10-27 15:34:10 -0700697 __releases(&tasklist_lock)
698 __acquires(&tasklist_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700700 struct pid_namespace *pid_ns = task_active_pid_ns(father);
701 struct task_struct *thread;
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700702
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700703 thread = father;
704 while_each_thread(father, thread) {
705 if (thread->flags & PF_EXITING)
706 continue;
707 if (unlikely(pid_ns->child_reaper == father))
708 pid_ns->child_reaper = thread;
709 return thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700711
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700712 if (unlikely(pid_ns->child_reaper == father)) {
713 write_unlock_irq(&tasklist_lock);
Denys Vlasenko397a21f2012-03-23 15:01:54 -0700714 if (unlikely(pid_ns == &init_pid_ns)) {
715 panic("Attempted to kill init! exitcode=0x%08x\n",
716 father->signal->group_exit_code ?:
717 father->exit_code);
718 }
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700719
720 zap_pid_ns_processes(pid_ns);
721 write_lock_irq(&tasklist_lock);
722 /*
723 * We can not clear ->child_reaper or leave it alone.
724 * There may by stealth EXIT_DEAD tasks on ->children,
725 * forget_original_parent() must move them somewhere.
726 */
727 pid_ns->child_reaper = init_pid_ns.child_reaper;
Lennart Poetteringebec18a2012-03-23 15:01:54 -0700728 } else if (father->signal->has_child_subreaper) {
729 struct task_struct *reaper;
730
731 /*
732 * Find the first ancestor marked as child_subreaper.
733 * Note that the code below checks same_thread_group(reaper,
734 * pid_ns->child_reaper). This is what we need to DTRT in a
735 * PID namespace. However we still need the check above, see
736 * http://marc.info/?l=linux-kernel&m=131385460420380
737 */
738 for (reaper = father->real_parent;
739 reaper != &init_task;
740 reaper = reaper->real_parent) {
741 if (same_thread_group(reaper, pid_ns->child_reaper))
742 break;
743 if (!reaper->signal->is_child_subreaper)
744 continue;
745 thread = reaper;
746 do {
747 if (!(thread->flags & PF_EXITING))
748 return reaper;
749 } while_each_thread(reaper, thread);
750 }
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700751 }
752
753 return pid_ns->child_reaper;
754}
755
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700756/*
757* Any that need to be release_task'd are put on the @dead list.
758 */
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800759static void reparent_leader(struct task_struct *father, struct task_struct *p,
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700760 struct list_head *dead)
761{
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700762 list_move_tail(&p->sibling, &p->real_parent->children);
763
Oleg Nesterov0976a032011-06-22 23:09:39 +0200764 if (p->exit_state == EXIT_DEAD)
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700765 return;
766 /*
767 * If this is a threaded reparent there is no need to
768 * notify anyone anything has happened.
769 */
770 if (same_thread_group(p->real_parent, father))
771 return;
772
773 /* We don't want people slaying init. */
774 p->exit_signal = SIGCHLD;
775
776 /* If it has exited notify the new parent about this child's death. */
Tejun Heod21142e2011-06-17 16:50:34 +0200777 if (!p->ptrace &&
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700778 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
Oleg Nesterov86773472011-06-22 23:09:09 +0200779 if (do_notify_parent(p, p->exit_signal)) {
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700780 p->exit_state = EXIT_DEAD;
781 list_move_tail(&p->sibling, dead);
782 }
783 }
784
785 kill_orphaned_pgrp(p, father);
786}
787
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788static void forget_original_parent(struct task_struct *father)
789{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700790 struct task_struct *p, *n, *reaper;
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700791 LIST_HEAD(dead_children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
793 write_lock_irq(&tasklist_lock);
Oleg Nesterovc7e49c12010-08-10 18:03:07 -0700794 /*
795 * Note that exit_ptrace() and find_new_reaper() might
796 * drop tasklist_lock and reacquire it.
797 */
798 exit_ptrace(father);
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700799 reaper = find_new_reaper(father);
Roland McGrathf4700212008-03-24 18:36:23 -0700800
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 list_for_each_entry_safe(p, n, &father->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800802 struct task_struct *t = p;
803 do {
804 t->real_parent = reaper;
805 if (t->parent == father) {
Tejun Heod21142e2011-06-17 16:50:34 +0200806 BUG_ON(t->ptrace);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800807 t->parent = t->real_parent;
808 }
809 if (t->pdeath_signal)
810 group_send_sig_info(t->pdeath_signal,
811 SEND_SIG_NOINFO, t);
812 } while_each_thread(p, t);
813 reparent_leader(father, p, &dead_children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700815 write_unlock_irq(&tasklist_lock);
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700816
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700817 BUG_ON(!list_empty(&father->children));
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700818
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700819 list_for_each_entry_safe(p, n, &dead_children, sibling) {
820 list_del_init(&p->sibling);
Oleg Nesterov39c626a2009-04-02 16:58:18 -0700821 release_task(p);
822 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823}
824
825/*
826 * Send signals to all our closest relatives so that they know
827 * to properly mourn us..
828 */
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300829static void exit_notify(struct task_struct *tsk, int group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830{
Oleg Nesterov53c8f9f2011-06-22 23:08:18 +0200831 bool autoreap;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 /*
834 * This does two things:
835 *
836 * A. Make init inherit all the child processes
837 * B. Check to see if any process groups have become orphaned
838 * as a result of our exiting, and if they have any stopped
839 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
840 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700841 forget_original_parent(tsk);
Pavel Emelyanov2e4a7072007-10-18 23:40:01 -0700842 exit_task_namespaces(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700844 write_lock_irq(&tasklist_lock);
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300845 if (group_dead)
846 kill_orphaned_pgrp(tsk->group_leader, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
Oleg Nesterov45cdf5c2011-06-23 19:06:50 +0200848 if (unlikely(tsk->ptrace)) {
849 int sig = thread_group_leader(tsk) &&
850 thread_group_empty(tsk) &&
851 !ptrace_reparented(tsk) ?
852 tsk->exit_signal : SIGCHLD;
853 autoreap = do_notify_parent(tsk, sig);
854 } else if (thread_group_leader(tsk)) {
855 autoreap = thread_group_empty(tsk) &&
856 do_notify_parent(tsk, tsk->exit_signal);
857 } else {
858 autoreap = true;
859 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860
Oleg Nesterov53c8f9f2011-06-22 23:08:18 +0200861 tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862
Oleg Nesterov9c339162010-05-26 14:43:10 -0700863 /* mt-exec, de_thread() is waiting for group leader */
864 if (unlikely(tsk->signal->notify_count < 0))
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700865 wake_up_process(tsk->signal->group_exit_task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 write_unlock_irq(&tasklist_lock);
867
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 /* If the process is dead, release it - nobody will wait for it */
Oleg Nesterov53c8f9f2011-06-22 23:08:18 +0200869 if (autoreap)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 release_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871}
872
Jeff Dikee18eecb2007-07-15 23:38:48 -0700873#ifdef CONFIG_DEBUG_STACK_USAGE
874static void check_stack_usage(void)
875{
876 static DEFINE_SPINLOCK(low_water_lock);
877 static int lowest_to_date = THREAD_SIZE;
Jeff Dikee18eecb2007-07-15 23:38:48 -0700878 unsigned long free;
879
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500880 free = stack_not_used(current);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700881
882 if (free >= lowest_to_date)
883 return;
884
885 spin_lock(&low_water_lock);
886 if (free < lowest_to_date) {
887 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
888 "left\n",
889 current->comm, free);
890 lowest_to_date = free;
891 }
892 spin_unlock(&low_water_lock);
893}
894#else
895static inline void check_stack_usage(void) {}
896#endif
897
Joe Perches9402c952012-01-12 17:17:17 -0800898void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899{
900 struct task_struct *tsk = current;
901 int group_dead;
902
903 profile_task_exit(tsk);
904
Jens Axboe73c10102011-03-08 13:19:51 +0100905 WARN_ON(blk_needs_flush_plug(tsk));
Jens Axboe22e2c502005-06-27 10:55:12 +0200906
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 if (unlikely(in_interrupt()))
908 panic("Aiee, killing interrupt handler!");
909 if (unlikely(!tsk->pid))
910 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911
Nelson Elhage33dd94a2010-12-02 14:31:21 -0800912 /*
913 * If do_exit is called because this processes oopsed, it's possible
914 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
915 * continuing. Amongst other possible reasons, this is to prevent
916 * mm_release()->clear_child_tid() from writing to a user-controlled
917 * kernel address.
918 */
919 set_fs(USER_DS);
920
Tejun Heoa288eec2011-06-17 16:50:37 +0200921 ptrace_event(PTRACE_EVENT_EXIT, code);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922
David Howellse0e81732009-09-02 09:13:40 +0100923 validate_creds_for_do_exit(tsk);
924
Alexander Nybergdf164db2005-06-23 00:09:13 -0700925 /*
926 * We're taking recursive faults here in do_exit. Safest is to just
927 * leave this task alone and wait for reboot.
928 */
929 if (unlikely(tsk->flags & PF_EXITING)) {
930 printk(KERN_ALERT
931 "Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700932 /*
933 * We can do this unlocked here. The futex code uses
934 * this flag just to verify whether the pi state
935 * cleanup has been done or not. In the worst case it
936 * loops once more. We pretend that the cleanup was
937 * done as there is no way to return. Either the
938 * OWNER_DIED bit is set by now or we push the blocked
939 * task into the wait for ever nirwana as well.
940 */
941 tsk->flags |= PF_EXITPIDONE;
Alexander Nybergdf164db2005-06-23 00:09:13 -0700942 set_current_state(TASK_UNINTERRUPTIBLE);
943 schedule();
944 }
945
Oleg Nesterovd12619b2008-02-08 04:19:12 -0800946 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700947 /*
948 * tsk->flags are checked in the futex code to protect against
Oleg Nesterove73f8952012-05-11 10:59:07 +1000949 * an exiting task cleaning up the robust pi futexes, and in
950 * task_work_add() to avoid the race with exit_task_work().
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700951 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -0700952 smp_mb();
Thomas Gleixner1d615482009-11-17 14:54:03 +0100953 raw_spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954
Oleg Nesterove73f8952012-05-11 10:59:07 +1000955 exit_task_work(tsk);
956
Alexander Gordeev5234ffb2012-03-09 14:59:59 +0100957 exit_irq_thread();
958
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 if (unlikely(in_atomic()))
960 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -0700961 current->comm, task_pid_nr(current),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 preempt_count());
963
964 acct_update_integrals(tsk);
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -0800965 /* sync mm's RSS info before statistics gathering */
KAMEZAWA Hiroyukia3a2e762010-04-06 14:34:42 -0700966 if (tsk->mm)
David Rientjes05af2e12012-03-21 16:34:13 -0700967 sync_mm_rss(tsk->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -0700969 if (group_dead) {
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700970 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -0700971 exit_itimers(tsk->signal);
Jiri Pirko1f102062009-09-22 16:44:10 -0700972 if (tsk->mm)
973 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
Andrew Mortonc3068952005-08-04 16:49:32 -0700974 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700975 acct_collect(code, group_dead);
Miloslav Trmac522ed772007-07-15 23:40:56 -0700976 if (group_dead)
977 tty_audit_exit();
Eric Parisa4ff8db2012-01-03 14:23:07 -0500978 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -0800979
Jonathan Limf2ab6d82007-08-30 23:56:23 -0700980 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -0800981 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -0700982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 exit_mm(tsk);
984
KaiGai Kohei0e464812006-06-25 05:49:24 -0700985 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700986 acct_process();
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400987 trace_sched_process_exit(tsk);
988
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 exit_sem(tsk);
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700990 exit_shm(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400991 exit_files(tsk);
992 exit_fs(tsk);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700993 check_stack_usage();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 exit_thread();
Stephane Eranian0b3fcf12011-01-03 18:20:01 +0200995
996 /*
997 * Flush inherited counters to the parent - before the parent
998 * gets woken up by child-exit notifications.
999 *
1000 * because of cgroup mode, must be called before cgroup_exit()
1001 */
1002 perf_event_exit_task(tsk);
1003
Paul Menageb4f48b62007-10-18 23:39:33 -07001004 cgroup_exit(tsk, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
Alan Cox5ec93d12009-11-30 13:18:45 +00001006 if (group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 disassociate_ctty(1);
1008
Al Viroa1261f52005-11-13 16:06:55 -08001009 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Matt Helsley9f460802005-11-07 00:59:16 -08001011 proc_exit_connector(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +02001012
1013 /*
Frederic Weisbecker24f1e32c2009-09-09 19:22:48 +02001014 * FIXME: do that only when needed, using sched_exit tracepoint
1015 */
Frederic Weisbeckerbf26c012011-04-07 16:53:20 +02001016 ptrace_put_breakpoints(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +02001017
Oleg Nesterov821c7de2008-03-02 21:44:44 +03001018 exit_notify(tsk, group_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019#ifdef CONFIG_NUMA
Miao Xiec0ff7452010-05-24 14:32:08 -07001020 task_lock(tsk);
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001021 mpol_put(tsk->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 tsk->mempolicy = NULL;
Miao Xiec0ff7452010-05-24 14:32:08 -07001023 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001025#ifdef CONFIG_FUTEX
Ingo Molnarc87e2832006-06-27 02:54:58 -07001026 if (unlikely(current->pi_state_cache))
1027 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001028#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001029 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001030 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -08001031 */
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001032 debug_check_no_locks_held(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001033 /*
1034 * We can do this unlocked here. The futex code uses this flag
1035 * just to verify whether the pi state cleanup has been done
1036 * or not. In the worst case it loops once more.
1037 */
1038 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
Al Viroafc847b2006-02-28 12:51:55 -05001040 if (tsk->io_context)
Louis Rillingb69f2292009-12-04 14:52:42 +01001041 exit_io_context(tsk);
Al Viroafc847b2006-02-28 12:51:55 -05001042
Jens Axboeb92ce552006-04-11 13:52:07 +02001043 if (tsk->splice_pipe)
1044 __free_pipe_info(tsk->splice_pipe);
1045
David Howellse0e81732009-09-02 09:13:40 +01001046 validate_creds_for_do_exit(tsk);
1047
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001048 preempt_disable();
Wu Fengguang54848d72011-04-05 13:21:19 -06001049 if (tsk->nr_dirtied)
1050 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001051 exit_rcu();
Yasunori Gotob5740f42012-01-17 17:40:31 +09001052
1053 /*
1054 * The setting of TASK_RUNNING by try_to_wake_up() may be delayed
1055 * when the following two conditions become true.
1056 * - There is race condition of mmap_sem (It is acquired by
1057 * exit_mm()), and
1058 * - SMI occurs before setting TASK_RUNINNG.
1059 * (or hypervisor of virtual machine switches to other guest)
1060 * As a result, we may become TASK_RUNNING after becoming TASK_DEAD
1061 *
1062 * To avoid it, we have to wait for releasing tsk->pi_lock which
1063 * is held by try_to_wake_up()
1064 */
1065 smp_mb();
1066 raw_spin_unlock_wait(&tsk->pi_lock);
1067
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001068 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -07001069 tsk->state = TASK_DEAD;
Tejun Heoa5850422011-11-21 12:32:23 -08001070 tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 schedule();
1072 BUG();
1073 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -07001074 for (;;)
1075 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076}
1077
Russ Anderson012914d2005-04-23 00:08:00 -07001078EXPORT_SYMBOL_GPL(do_exit);
1079
Joe Perches9402c952012-01-12 17:17:17 -08001080void complete_and_exit(struct completion *comp, long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081{
1082 if (comp)
1083 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001084
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 do_exit(code);
1086}
1087
1088EXPORT_SYMBOL(complete_and_exit);
1089
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001090SYSCALL_DEFINE1(exit, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091{
1092 do_exit((error_code&0xff)<<8);
1093}
1094
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095/*
1096 * Take down every thread in the group. This is called by fatal signals
1097 * as well as by sys_exit_group (below).
1098 */
Joe Perches9402c952012-01-12 17:17:17 -08001099void
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100do_group_exit(int exit_code)
1101{
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001102 struct signal_struct *sig = current->signal;
1103
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1105
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001106 if (signal_group_exit(sig))
1107 exit_code = sig->group_exit_code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 else if (!thread_group_empty(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 struct sighand_struct *const sighand = current->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001111 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 /* Another thread got here before we took the lock. */
1113 exit_code = sig->group_exit_code;
1114 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001116 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 zap_other_threads(current);
1118 }
1119 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 }
1121
1122 do_exit(exit_code);
1123 /* NOTREACHED */
1124}
1125
1126/*
1127 * this kills every thread in the thread group. Note that any externally
1128 * wait4()-ing process will get the correct exit code - even if this
1129 * thread is not the thread group leader.
1130 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001131SYSCALL_DEFINE1(exit_group, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132{
1133 do_group_exit((error_code & 0xff) << 8);
Heiko Carstens2ed7c032009-01-14 14:13:54 +01001134 /* NOTREACHED */
1135 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136}
1137
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001138struct wait_opts {
1139 enum pid_type wo_type;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001140 int wo_flags;
Richard Kennedye1eb1eb2009-06-17 16:27:42 -07001141 struct pid *wo_pid;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001142
1143 struct siginfo __user *wo_info;
1144 int __user *wo_stat;
1145 struct rusage __user *wo_rusage;
1146
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001147 wait_queue_t child_wait;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001148 int notask_error;
1149};
1150
Oleg Nesterov989264f2009-09-23 15:56:49 -07001151static inline
1152struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
Eric W. Biederman161550d2008-02-08 04:19:14 -08001153{
Oleg Nesterov989264f2009-09-23 15:56:49 -07001154 if (type != PIDTYPE_PID)
1155 task = task->group_leader;
1156 return task->pids[type].pid;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001157}
1158
Oleg Nesterov989264f2009-09-23 15:56:49 -07001159static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001160{
1161 return wo->wo_type == PIDTYPE_MAX ||
1162 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1163}
1164
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001165static int eligible_child(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166{
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001167 if (!eligible_pid(wo, p))
1168 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 /* Wait for all children (clone and not) if __WALL is set;
1170 * otherwise, wait for clone children *only* if __WCLONE is
1171 * set; otherwise, wait for non-clone children *only*. (Note:
1172 * A "clone" child here is one that reports to its parent
1173 * using a signal other than SIGCHLD.) */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001174 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1175 && !(wo->wo_flags & __WALL))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177
Roland McGrath14dd0b82008-03-30 18:41:25 -07001178 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179}
1180
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001181static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1182 pid_t pid, uid_t uid, int why, int status)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001184 struct siginfo __user *infop;
1185 int retval = wo->wo_rusage
1186 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -07001187
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001189 infop = wo->wo_info;
Vitaly Mayatskikhb6fe2d12009-09-23 15:56:52 -07001190 if (infop) {
1191 if (!retval)
1192 retval = put_user(SIGCHLD, &infop->si_signo);
1193 if (!retval)
1194 retval = put_user(0, &infop->si_errno);
1195 if (!retval)
1196 retval = put_user((short)why, &infop->si_code);
1197 if (!retval)
1198 retval = put_user(pid, &infop->si_pid);
1199 if (!retval)
1200 retval = put_user(uid, &infop->si_uid);
1201 if (!retval)
1202 retval = put_user(status, &infop->si_status);
1203 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 if (!retval)
1205 retval = pid;
1206 return retval;
1207}
1208
1209/*
1210 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1211 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1212 * the lock and this task is uninteresting. If we return nonzero, we have
1213 * released the lock and the system call should return.
1214 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001215static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216{
1217 unsigned long state;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001218 int retval, status, traced;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001219 pid_t pid = task_pid_vnr(p);
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001220 uid_t uid = from_kuid_munged(current_user_ns(), __task_cred(p)->uid);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001221 struct siginfo __user *infop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001223 if (!likely(wo->wo_flags & WEXITED))
Roland McGrath98abed02008-03-19 19:24:59 -07001224 return 0;
1225
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001226 if (unlikely(wo->wo_flags & WNOWAIT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 int exit_code = p->exit_code;
Thiago Farinaf3abd4f2010-03-05 13:42:52 -08001228 int why;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 get_task_struct(p);
1231 read_unlock(&tasklist_lock);
1232 if ((exit_code & 0x7f) == 0) {
1233 why = CLD_EXITED;
1234 status = exit_code >> 8;
1235 } else {
1236 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1237 status = exit_code & 0x7f;
1238 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001239 return wait_noreap_copyout(wo, p, pid, uid, why, status);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 }
1241
1242 /*
1243 * Try to move the task's state to DEAD
1244 * only one thread is allowed to do this:
1245 */
1246 state = xchg(&p->exit_state, EXIT_DEAD);
1247 if (state != EXIT_ZOMBIE) {
1248 BUG_ON(state != EXIT_DEAD);
1249 return 0;
1250 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251
Oleg Nesterov53b6f9f2008-04-30 00:53:13 -07001252 traced = ptrace_reparented(p);
Oleg Nesterovbefca962009-06-18 16:49:11 -07001253 /*
1254 * It can be ptraced but not reparented, check
Oleg Nesterove550f142011-06-22 23:09:54 +02001255 * thread_group_leader() to filter out sub-threads.
Oleg Nesterovbefca962009-06-18 16:49:11 -07001256 */
Oleg Nesterove550f142011-06-22 23:09:54 +02001257 if (likely(!traced) && thread_group_leader(p)) {
Jesper Juhl3795e162006-01-09 20:54:39 -08001258 struct signal_struct *psig;
1259 struct signal_struct *sig;
Jiri Pirko1f102062009-09-22 16:44:10 -07001260 unsigned long maxrss;
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001261 cputime_t tgutime, tgstime;
Jesper Juhl3795e162006-01-09 20:54:39 -08001262
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 /*
1264 * The resource counters for the group leader are in its
1265 * own task_struct. Those for dead threads in the group
1266 * are in its signal_struct, as are those for the child
1267 * processes it has previously reaped. All these
1268 * accumulate in the parent's signal_struct c* fields.
1269 *
1270 * We don't bother to take a lock here to protect these
1271 * p->signal fields, because they are only touched by
1272 * __exit_signal, which runs with tasklist_lock
1273 * write-locked anyway, and so is excluded here. We do
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001274 * need to protect the access to parent->signal fields,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 * as other threads in the parent group can be right
1276 * here reaping other children at the same time.
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001277 *
1278 * We use thread_group_times() to get times for the thread
1279 * group, which consolidates times for all threads in the
1280 * group including the group leader.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 */
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001282 thread_group_times(p, &tgutime, &tgstime);
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001283 spin_lock_irq(&p->real_parent->sighand->siglock);
1284 psig = p->real_parent->signal;
Jesper Juhl3795e162006-01-09 20:54:39 -08001285 sig = p->signal;
Martin Schwidefsky64861632011-12-15 14:56:09 +01001286 psig->cutime += tgutime + sig->cutime;
1287 psig->cstime += tgstime + sig->cstime;
1288 psig->cgtime += p->gtime + sig->gtime + sig->cgtime;
Jesper Juhl3795e162006-01-09 20:54:39 -08001289 psig->cmin_flt +=
1290 p->min_flt + sig->min_flt + sig->cmin_flt;
1291 psig->cmaj_flt +=
1292 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1293 psig->cnvcsw +=
1294 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1295 psig->cnivcsw +=
1296 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001297 psig->cinblock +=
1298 task_io_get_inblock(p) +
1299 sig->inblock + sig->cinblock;
1300 psig->coublock +=
1301 task_io_get_oublock(p) +
1302 sig->oublock + sig->coublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001303 maxrss = max(sig->maxrss, sig->cmaxrss);
1304 if (psig->cmaxrss < maxrss)
1305 psig->cmaxrss = maxrss;
Andrea Righi59954772008-07-27 17:29:15 +02001306 task_io_accounting_add(&psig->ioac, &p->ioac);
1307 task_io_accounting_add(&psig->ioac, &sig->ioac);
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001308 spin_unlock_irq(&p->real_parent->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 }
1310
1311 /*
1312 * Now we are sure this task is interesting, and no other
1313 * thread can reap it because we set its state to EXIT_DEAD.
1314 */
1315 read_unlock(&tasklist_lock);
1316
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001317 retval = wo->wo_rusage
1318 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1320 ? p->signal->group_exit_code : p->exit_code;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001321 if (!retval && wo->wo_stat)
1322 retval = put_user(status, wo->wo_stat);
1323
1324 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 if (!retval && infop)
1326 retval = put_user(SIGCHLD, &infop->si_signo);
1327 if (!retval && infop)
1328 retval = put_user(0, &infop->si_errno);
1329 if (!retval && infop) {
1330 int why;
1331
1332 if ((status & 0x7f) == 0) {
1333 why = CLD_EXITED;
1334 status >>= 8;
1335 } else {
1336 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1337 status &= 0x7f;
1338 }
1339 retval = put_user((short)why, &infop->si_code);
1340 if (!retval)
1341 retval = put_user(status, &infop->si_status);
1342 }
1343 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001344 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 if (!retval && infop)
David Howellsc69e8d92008-11-14 10:39:19 +11001346 retval = put_user(uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001347 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001348 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001349
1350 if (traced) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001352 /* We dropped tasklist, ptracer could die and untrace */
1353 ptrace_unlink(p);
1354 /*
Oleg Nesterov86773472011-06-22 23:09:09 +02001355 * If this is not a sub-thread, notify the parent.
1356 * If parent wants a zombie, don't release it now.
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001357 */
Oleg Nesterov86773472011-06-22 23:09:09 +02001358 if (thread_group_leader(p) &&
1359 !do_notify_parent(p, p->exit_signal)) {
1360 p->exit_state = EXIT_ZOMBIE;
1361 p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 }
1363 write_unlock_irq(&tasklist_lock);
1364 }
1365 if (p != NULL)
1366 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001367
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 return retval;
1369}
1370
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001371static int *task_stopped_code(struct task_struct *p, bool ptrace)
1372{
1373 if (ptrace) {
Tejun Heo544b2c92011-06-14 11:20:18 +02001374 if (task_is_stopped_or_traced(p) &&
1375 !(p->jobctl & JOBCTL_LISTENING))
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001376 return &p->exit_code;
1377 } else {
1378 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1379 return &p->signal->group_exit_code;
1380 }
1381 return NULL;
1382}
1383
Tejun Heo19e27462011-05-12 10:47:23 +02001384/**
1385 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1386 * @wo: wait options
1387 * @ptrace: is the wait for ptrace
1388 * @p: task to wait for
1389 *
1390 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1391 *
1392 * CONTEXT:
1393 * read_lock(&tasklist_lock), which is released if return value is
1394 * non-zero. Also, grabs and releases @p->sighand->siglock.
1395 *
1396 * RETURNS:
1397 * 0 if wait condition didn't exist and search for other wait conditions
1398 * should continue. Non-zero return, -errno on failure and @p's pid on
1399 * success, implies that tasklist_lock is released and wait condition
1400 * search should terminate.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001402static int wait_task_stopped(struct wait_opts *wo,
1403 int ptrace, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001405 struct siginfo __user *infop;
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001406 int retval, exit_code, *p_code, why;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001407 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001408 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409
Oleg Nesterov47918022009-06-17 16:27:39 -07001410 /*
1411 * Traditionally we see ptrace'd stopped tasks regardless of options.
1412 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001413 if (!ptrace && !(wo->wo_flags & WUNTRACED))
Roland McGrath98abed02008-03-19 19:24:59 -07001414 return 0;
1415
Tejun Heo19e27462011-05-12 10:47:23 +02001416 if (!task_stopped_code(p, ptrace))
1417 return 0;
1418
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001419 exit_code = 0;
1420 spin_lock_irq(&p->sighand->siglock);
1421
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001422 p_code = task_stopped_code(p, ptrace);
1423 if (unlikely(!p_code))
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001424 goto unlock_sig;
1425
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001426 exit_code = *p_code;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001427 if (!exit_code)
1428 goto unlock_sig;
1429
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001430 if (!unlikely(wo->wo_flags & WNOWAIT))
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001431 *p_code = 0;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001432
Sasha Levin8ca937a2012-05-17 23:31:39 +02001433 uid = from_kuid_munged(current_user_ns(), task_uid(p));
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001434unlock_sig:
1435 spin_unlock_irq(&p->sighand->siglock);
1436 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 return 0;
1438
1439 /*
1440 * Now we are pretty sure this task is interesting.
1441 * Make sure it doesn't get reaped out from under us while we
1442 * give up the lock and then examine it below. We don't want to
1443 * keep holding onto the tasklist_lock while we call getrusage and
1444 * possibly take page faults for user memory.
1445 */
1446 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001447 pid = task_pid_vnr(p);
Roland McGrathf4700212008-03-24 18:36:23 -07001448 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 read_unlock(&tasklist_lock);
1450
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001451 if (unlikely(wo->wo_flags & WNOWAIT))
1452 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001454 retval = wo->wo_rusage
1455 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1456 if (!retval && wo->wo_stat)
1457 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1458
1459 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 if (!retval && infop)
1461 retval = put_user(SIGCHLD, &infop->si_signo);
1462 if (!retval && infop)
1463 retval = put_user(0, &infop->si_errno);
1464 if (!retval && infop)
Roland McGrath6efcae42008-03-08 11:41:22 -08001465 retval = put_user((short)why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466 if (!retval && infop)
1467 retval = put_user(exit_code, &infop->si_status);
1468 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001469 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001471 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001473 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 put_task_struct(p);
1475
1476 BUG_ON(!retval);
1477 return retval;
1478}
1479
1480/*
1481 * Handle do_wait work for one task in a live, non-stopped state.
1482 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1483 * the lock and this task is uninteresting. If we return nonzero, we have
1484 * released the lock and the system call should return.
1485 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001486static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487{
1488 int retval;
1489 pid_t pid;
1490 uid_t uid;
1491
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001492 if (!unlikely(wo->wo_flags & WCONTINUED))
Roland McGrath98abed02008-03-19 19:24:59 -07001493 return 0;
1494
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1496 return 0;
1497
1498 spin_lock_irq(&p->sighand->siglock);
1499 /* Re-check with the lock held. */
1500 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1501 spin_unlock_irq(&p->sighand->siglock);
1502 return 0;
1503 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001504 if (!unlikely(wo->wo_flags & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
Sasha Levin8ca937a2012-05-17 23:31:39 +02001506 uid = from_kuid_munged(current_user_ns(), task_uid(p));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 spin_unlock_irq(&p->sighand->siglock);
1508
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001509 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 get_task_struct(p);
1511 read_unlock(&tasklist_lock);
1512
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001513 if (!wo->wo_info) {
1514 retval = wo->wo_rusage
1515 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001517 if (!retval && wo->wo_stat)
1518 retval = put_user(0xffff, wo->wo_stat);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001520 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 } else {
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001522 retval = wait_noreap_copyout(wo, p, pid, uid,
1523 CLD_CONTINUED, SIGCONT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 BUG_ON(retval == 0);
1525 }
1526
1527 return retval;
1528}
1529
Roland McGrath98abed02008-03-19 19:24:59 -07001530/*
1531 * Consider @p for a wait by @parent.
1532 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001533 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001534 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1535 * Returns zero if the search for a child should continue;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001536 * then ->notask_error is 0 if @p is an eligible child,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001537 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001538 */
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001539static int wait_consider_task(struct wait_opts *wo, int ptrace,
1540 struct task_struct *p)
Roland McGrath98abed02008-03-19 19:24:59 -07001541{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001542 int ret = eligible_child(wo, p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001543 if (!ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001544 return ret;
1545
Oleg Nesterova2322e12009-09-23 15:56:45 -07001546 ret = security_task_wait(p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001547 if (unlikely(ret < 0)) {
1548 /*
1549 * If we have not yet seen any eligible child,
1550 * then let this error code replace -ECHILD.
1551 * A permission error will give the user a clue
1552 * to look for security policy problems, rather
1553 * than for mysterious wait bugs.
1554 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001555 if (wo->notask_error)
1556 wo->notask_error = ret;
Oleg Nesterov78a3d9d2009-04-29 18:01:23 +02001557 return 0;
Roland McGrath14dd0b82008-03-30 18:41:25 -07001558 }
1559
Tejun Heo823b0182011-03-23 10:37:01 +01001560 /* dead body doesn't have much to contribute */
Oleg Nesterov50b8d252012-01-04 17:29:02 +01001561 if (unlikely(p->exit_state == EXIT_DEAD)) {
1562 /*
1563 * But do not ignore this task until the tracer does
1564 * wait_task_zombie()->do_notify_parent().
1565 */
1566 if (likely(!ptrace) && unlikely(ptrace_reparented(p)))
1567 wo->notask_error = 0;
Roland McGrath98abed02008-03-19 19:24:59 -07001568 return 0;
Oleg Nesterov50b8d252012-01-04 17:29:02 +01001569 }
Roland McGrath98abed02008-03-19 19:24:59 -07001570
Tejun Heo9b84cca2011-03-23 10:37:01 +01001571 /* slay zombie? */
1572 if (p->exit_state == EXIT_ZOMBIE) {
Tejun Heo45cb24a2011-03-23 10:37:01 +01001573 /*
1574 * A zombie ptracee is only visible to its ptracer.
1575 * Notification and reaping will be cascaded to the real
1576 * parent when the ptracer detaches.
1577 */
Tejun Heod21142e2011-06-17 16:50:34 +02001578 if (likely(!ptrace) && unlikely(p->ptrace)) {
Tejun Heo45cb24a2011-03-23 10:37:01 +01001579 /* it will become visible, clear notask_error */
1580 wo->notask_error = 0;
1581 return 0;
1582 }
1583
Tejun Heo9b84cca2011-03-23 10:37:01 +01001584 /* we don't reap group leaders with subthreads */
1585 if (!delay_group_leader(p))
1586 return wait_task_zombie(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001587
Tejun Heo9b84cca2011-03-23 10:37:01 +01001588 /*
1589 * Allow access to stopped/continued state via zombie by
1590 * falling through. Clearing of notask_error is complex.
1591 *
1592 * When !@ptrace:
1593 *
1594 * If WEXITED is set, notask_error should naturally be
1595 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1596 * so, if there are live subthreads, there are events to
1597 * wait for. If all subthreads are dead, it's still safe
1598 * to clear - this function will be called again in finite
1599 * amount time once all the subthreads are released and
1600 * will then return without clearing.
1601 *
1602 * When @ptrace:
1603 *
1604 * Stopped state is per-task and thus can't change once the
1605 * target task dies. Only continued and exited can happen.
1606 * Clear notask_error if WCONTINUED | WEXITED.
1607 */
1608 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1609 wo->notask_error = 0;
1610 } else {
1611 /*
Tejun Heo45cb24a2011-03-23 10:37:01 +01001612 * If @p is ptraced by a task in its real parent's group,
1613 * hide group stop/continued state when looking at @p as
1614 * the real parent; otherwise, a single stop can be
1615 * reported twice as group and ptrace stops.
1616 *
1617 * If a ptracer wants to distinguish the two events for its
1618 * own children, it should create a separate process which
1619 * takes the role of real parent.
1620 */
Oleg Nesterov479bf982011-06-24 17:34:39 +02001621 if (likely(!ptrace) && p->ptrace && !ptrace_reparented(p))
Tejun Heo45cb24a2011-03-23 10:37:01 +01001622 return 0;
1623
1624 /*
Tejun Heo9b84cca2011-03-23 10:37:01 +01001625 * @p is alive and it's gonna stop, continue or exit, so
1626 * there always is something to wait for.
1627 */
1628 wo->notask_error = 0;
1629 }
Roland McGrath98abed02008-03-19 19:24:59 -07001630
1631 /*
Tejun Heo45cb24a2011-03-23 10:37:01 +01001632 * Wait for stopped. Depending on @ptrace, different stopped state
1633 * is used and the two don't interact with each other.
Roland McGrath98abed02008-03-19 19:24:59 -07001634 */
Tejun Heo19e27462011-05-12 10:47:23 +02001635 ret = wait_task_stopped(wo, ptrace, p);
1636 if (ret)
1637 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001638
Tejun Heo45cb24a2011-03-23 10:37:01 +01001639 /*
1640 * Wait for continued. There's only one continued state and the
1641 * ptracer can consume it which can confuse the real parent. Don't
1642 * use WCONTINUED from ptracer. You don't need or want it.
1643 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001644 return wait_task_continued(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001645}
1646
1647/*
1648 * Do the work of do_wait() for one thread in the group, @tsk.
1649 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001650 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001651 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1652 * Returns zero if the search for a child should continue; then
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001653 * ->notask_error is 0 if there were any eligible children,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001654 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001655 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001656static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001657{
1658 struct task_struct *p;
1659
1660 list_for_each_entry(p, &tsk->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001661 int ret = wait_consider_task(wo, 0, p);
1662 if (ret)
1663 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001664 }
1665
1666 return 0;
1667}
1668
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001669static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001670{
1671 struct task_struct *p;
1672
Roland McGrathf4700212008-03-24 18:36:23 -07001673 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001674 int ret = wait_consider_task(wo, 1, p);
Roland McGrathf4700212008-03-24 18:36:23 -07001675 if (ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001676 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001677 }
1678
1679 return 0;
1680}
1681
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001682static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1683 int sync, void *key)
1684{
1685 struct wait_opts *wo = container_of(wait, struct wait_opts,
1686 child_wait);
1687 struct task_struct *p = key;
1688
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001689 if (!eligible_pid(wo, p))
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001690 return 0;
1691
Oleg Nesterovb4fe5182009-09-23 15:56:47 -07001692 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1693 return 0;
1694
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001695 return default_wake_function(wait, mode, sync, key);
1696}
1697
Oleg Nesterova7f07652009-09-23 15:56:44 -07001698void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1699{
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001700 __wake_up_sync_key(&parent->signal->wait_chldexit,
1701 TASK_INTERRUPTIBLE, 1, p);
Oleg Nesterova7f07652009-09-23 15:56:44 -07001702}
1703
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001704static long do_wait(struct wait_opts *wo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 struct task_struct *tsk;
Roland McGrath98abed02008-03-19 19:24:59 -07001707 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001709 trace_sched_process_wait(wo->wo_pid);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001710
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001711 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1712 wo->child_wait.private = current;
1713 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714repeat:
Roland McGrath98abed02008-03-19 19:24:59 -07001715 /*
1716 * If there is nothing that can match our critiera just get out.
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001717 * We will clear ->notask_error to zero if we see any child that
1718 * might later match our criteria, even if we are not able to reap
1719 * it yet.
Roland McGrath98abed02008-03-19 19:24:59 -07001720 */
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001721 wo->notask_error = -ECHILD;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001722 if ((wo->wo_type < PIDTYPE_MAX) &&
1723 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001724 goto notask;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001725
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001726 set_current_state(TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 read_lock(&tasklist_lock);
1728 tsk = current;
1729 do {
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001730 retval = do_wait_thread(wo, tsk);
1731 if (retval)
Roland McGrath98abed02008-03-19 19:24:59 -07001732 goto end;
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001733
1734 retval = ptrace_do_wait(wo, tsk);
1735 if (retval)
1736 goto end;
Roland McGrath98abed02008-03-19 19:24:59 -07001737
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001738 if (wo->wo_flags & __WNOTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 break;
Oleg Nesterova3f6dfb2009-06-17 16:27:41 -07001740 } while_each_thread(current, tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001742
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001743notask:
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001744 retval = wo->notask_error;
1745 if (!retval && !(wo->wo_flags & WNOHANG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 retval = -ERESTARTSYS;
Roland McGrath98abed02008-03-19 19:24:59 -07001747 if (!signal_pending(current)) {
1748 schedule();
1749 goto repeat;
1750 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752end:
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001753 __set_current_state(TASK_RUNNING);
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001754 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 return retval;
1756}
1757
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001758SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1759 infop, int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001761 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001762 struct pid *pid = NULL;
1763 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 long ret;
1765
1766 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1767 return -EINVAL;
1768 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1769 return -EINVAL;
1770
1771 switch (which) {
1772 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001773 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 break;
1775 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001776 type = PIDTYPE_PID;
1777 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 return -EINVAL;
1779 break;
1780 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001781 type = PIDTYPE_PGID;
1782 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784 break;
1785 default:
1786 return -EINVAL;
1787 }
1788
Eric W. Biederman161550d2008-02-08 04:19:14 -08001789 if (type < PIDTYPE_MAX)
1790 pid = find_get_pid(upid);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001791
1792 wo.wo_type = type;
1793 wo.wo_pid = pid;
1794 wo.wo_flags = options;
1795 wo.wo_info = infop;
1796 wo.wo_stat = NULL;
1797 wo.wo_rusage = ru;
1798 ret = do_wait(&wo);
Vitaly Mayatskikhdfe16df2009-09-23 15:56:51 -07001799
1800 if (ret > 0) {
1801 ret = 0;
1802 } else if (infop) {
1803 /*
1804 * For a WNOHANG return, clear out all the fields
1805 * we would set so the user can easily tell the
1806 * difference.
1807 */
1808 if (!ret)
1809 ret = put_user(0, &infop->si_signo);
1810 if (!ret)
1811 ret = put_user(0, &infop->si_errno);
1812 if (!ret)
1813 ret = put_user(0, &infop->si_code);
1814 if (!ret)
1815 ret = put_user(0, &infop->si_pid);
1816 if (!ret)
1817 ret = put_user(0, &infop->si_uid);
1818 if (!ret)
1819 ret = put_user(0, &infop->si_status);
1820 }
1821
Eric W. Biederman161550d2008-02-08 04:19:14 -08001822 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823
1824 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001825 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 return ret;
1827}
1828
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001829SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1830 int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001832 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001833 struct pid *pid = NULL;
1834 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835 long ret;
1836
1837 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1838 __WNOTHREAD|__WCLONE|__WALL))
1839 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001840
1841 if (upid == -1)
1842 type = PIDTYPE_MAX;
1843 else if (upid < 0) {
1844 type = PIDTYPE_PGID;
1845 pid = find_get_pid(-upid);
1846 } else if (upid == 0) {
1847 type = PIDTYPE_PGID;
Oleg Nesterov2ae448e2009-04-02 16:58:36 -07001848 pid = get_task_pid(current, PIDTYPE_PGID);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001849 } else /* upid > 0 */ {
1850 type = PIDTYPE_PID;
1851 pid = find_get_pid(upid);
1852 }
1853
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001854 wo.wo_type = type;
1855 wo.wo_pid = pid;
1856 wo.wo_flags = options | WEXITED;
1857 wo.wo_info = NULL;
1858 wo.wo_stat = stat_addr;
1859 wo.wo_rusage = ru;
1860 ret = do_wait(&wo);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001861 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862
1863 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001864 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865 return ret;
1866}
1867
1868#ifdef __ARCH_WANT_SYS_WAITPID
1869
1870/*
1871 * sys_waitpid() remains for compatibility. waitpid() should be
1872 * implemented by calling sys_wait4() from libc.a.
1873 */
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001874SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875{
1876 return sys_wait4(pid, stat_addr, options, NULL);
1877}
1878
1879#endif