blob: 21aa7b3001fb49edce43446dd85898f62ab33d98 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
55#include <asm/uaccess.h>
56#include <asm/unistd.h>
57#include <asm/pgtable.h>
58#include <asm/mmu_context.h>
59
Adrian Bunk408b6642005-05-01 08:59:29 -070060static void exit_mm(struct task_struct * tsk);
61
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070062static void __unhash_process(struct task_struct *p, bool group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -070063{
64 nr_threads--;
65 detach_pid(p, PIDTYPE_PID);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070066 if (group_dead) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 detach_pid(p, PIDTYPE_PGID);
68 detach_pid(p, PIDTYPE_SID);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080069
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -070070 list_del_rcu(&p->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -080071 list_del_init(&p->sibling);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -080072 __get_cpu_var(process_counts)--;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 }
Oleg Nesterov47e65322006-03-28 16:11:25 -080074 list_del_rcu(&p->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -070075}
76
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080077/*
78 * This function expects the tasklist_lock write-locked.
79 */
80static void __exit_signal(struct task_struct *tsk)
81{
82 struct signal_struct *sig = tsk->signal;
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070083 bool group_dead = thread_group_leader(tsk);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080084 struct sighand_struct *sighand;
Oleg Nesterov4ada8562010-05-26 14:43:17 -070085 struct tty_struct *uninitialized_var(tty);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080086
Paul E. McKenneyd11c5632010-02-22 17:04:50 -080087 sighand = rcu_dereference_check(tsk->sighand,
88 rcu_read_lock_held() ||
Paul E. McKenneydb1466b2010-03-03 07:46:56 -080089 lockdep_tasklist_lock_is_held());
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080090 spin_lock(&sighand->siglock);
91
92 posix_cpu_timers_exit(tsk);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070093 if (group_dead) {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080094 posix_cpu_timers_exit_group(tsk);
Oleg Nesterov4ada8562010-05-26 14:43:17 -070095 tty = sig->tty;
96 sig->tty = NULL;
Oleg Nesterov4a599942010-05-26 14:43:12 -070097 } else {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080098 /*
Oleg Nesterove0a70212010-11-05 16:53:42 +010099 * This can only happen if the caller is de_thread().
100 * FIXME: this is the temporary hack, we should teach
101 * posix-cpu-timers to handle this case correctly.
102 */
103 if (unlikely(has_group_leader_pid(tsk)))
104 posix_cpu_timers_exit_group(tsk);
105
106 /*
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800107 * If there is any task waiting for the group exit
108 * then notify it:
109 */
Oleg Nesterovd3441932010-05-26 14:43:11 -0700110 if (sig->notify_count > 0 && !--sig->notify_count)
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800111 wake_up_process(sig->group_exit_task);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700112
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800113 if (tsk == sig->curr_target)
114 sig->curr_target = next_thread(tsk);
115 /*
116 * Accumulate here the counters for all threads but the
117 * group leader as they die, so they can be added into
118 * the process-wide totals when those are taken.
119 * The group leader stays around as a zombie as long
120 * as there are other threads. When it gets reaped,
121 * the exit.c code will add its counts into these totals.
122 * We won't ever get here for the group leader, since it
123 * will have been the last reference on the signal_struct.
124 */
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +0900125 sig->utime = cputime_add(sig->utime, tsk->utime);
126 sig->stime = cputime_add(sig->stime, tsk->stime);
Hidetoshi Setod5b7c782009-11-26 14:49:05 +0900127 sig->gtime = cputime_add(sig->gtime, tsk->gtime);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800128 sig->min_flt += tsk->min_flt;
129 sig->maj_flt += tsk->maj_flt;
130 sig->nvcsw += tsk->nvcsw;
131 sig->nivcsw += tsk->nivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700132 sig->inblock += task_io_get_inblock(tsk);
133 sig->oublock += task_io_get_oublock(tsk);
Andrea Righi59954772008-07-27 17:29:15 +0200134 task_io_accounting_add(&sig->ioac, &tsk->ioac);
Peter Zijlstra32bd6712009-02-05 12:24:15 +0100135 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800136 }
137
Oleg Nesterovb3ac0222010-05-26 14:43:24 -0700138 sig->nr_threads--;
Oleg Nesterovd40e48e2010-05-26 14:43:19 -0700139 __unhash_process(tsk, group_dead);
Oleg Nesterov58767002006-03-28 16:11:20 -0800140
Oleg Nesterovda7978b2008-05-23 13:04:41 -0700141 /*
142 * Do this under ->siglock, we can race with another thread
143 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
144 */
145 flush_sigqueue(&tsk->pending);
Oleg Nesterova7e53282006-03-28 16:11:27 -0800146 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800147 spin_unlock(&sighand->siglock);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800148
Oleg Nesterova7e53282006-03-28 16:11:27 -0800149 __cleanup_sighand(sighand);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800150 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -0700151 if (group_dead) {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800152 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov4ada8562010-05-26 14:43:17 -0700153 tty_kref_put(tty);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800154 }
155}
156
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800157static void delayed_put_task_struct(struct rcu_head *rhp)
158{
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400159 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
160
Peter Zijlstra4e231c72010-09-09 21:01:59 +0200161 perf_event_delayed_put(tsk);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400162 trace_sched_process_free(tsk);
163 put_task_struct(tsk);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800164}
165
Roland McGrathf4700212008-03-24 18:36:23 -0700166
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167void release_task(struct task_struct * p)
168{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700169 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800171repeat:
Roland McGrathdae33572008-07-25 19:45:48 -0700172 tracehook_prepare_release_task(p);
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);
Roland McGrathdae33572008-07-25 19:45:48 -0700182 tracehook_finish_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) {
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700193 BUG_ON(task_detached(leader));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 do_notify_parent(leader, leader->exit_signal);
195 /*
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.
199 *
200 * do_notify_parent() will have marked it self-reaping in
201 * that case.
202 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700203 zap_leader = task_detached(leader);
Roland McGrathdae33572008-07-25 19:45:48 -0700204
205 /*
206 * This maintains the invariant that release_task()
207 * only runs on a task in EXIT_DEAD, just for sanity.
208 */
209 if (zap_leader)
210 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 }
212
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 release_thread(p);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800215 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216
217 p = leader;
218 if (unlikely(zap_leader))
219 goto repeat;
220}
221
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222/*
223 * This checks not only the pgrp, but falls back on the pid if no
224 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
225 * without this...
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800226 *
227 * The caller must hold rcu lock or the tasklist lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 */
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800229struct pid *session_of_pgrp(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230{
231 struct task_struct *p;
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800232 struct pid *sid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800234 p = pid_task(pgrp, PIDTYPE_PGID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800235 if (p == NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800236 p = pid_task(pgrp, PIDTYPE_PID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800237 if (p != NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800238 sid = task_session(p);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800239
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 return sid;
241}
242
243/*
244 * Determine if a process group is "orphaned", according to the POSIX
245 * definition in 2.2.2.52. Orphaned process groups are not to be affected
246 * by terminal-generated stop signals. Newly orphaned process groups are
247 * to receive a SIGHUP and a SIGCONT.
248 *
249 * "I ask you, have you ever known what it is to be an orphan?"
250 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800251static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252{
253 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800255 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300256 if ((p == ignored_task) ||
257 (p->exit_state && thread_group_empty(p)) ||
258 is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 continue;
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300260
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800261 if (task_pgrp(p->real_parent) != pgrp &&
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300262 task_session(p->real_parent) == task_session(p))
263 return 0;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800264 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300265
266 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267}
268
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800269int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270{
271 int retval;
272
273 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800274 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 read_unlock(&tasklist_lock);
276
277 return retval;
278}
279
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800280static int has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281{
282 int retval = 0;
283 struct task_struct *p;
284
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800285 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Matthew Wilcox338077e2007-12-06 11:09:35 -0500286 if (!task_is_stopped(p))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 retval = 1;
289 break;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800290 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 return retval;
292}
293
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300294/*
295 * Check to see if any process groups have become orphaned as
296 * a result of our exiting, and if they have any stopped jobs,
297 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
298 */
299static void
300kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
301{
302 struct pid *pgrp = task_pgrp(tsk);
303 struct task_struct *ignored_task = tsk;
304
305 if (!parent)
306 /* exit: our father is in a different pgrp than
307 * we are and we were the only connection outside.
308 */
309 parent = tsk->real_parent;
310 else
311 /* reparent: our child is in a different pgrp than
312 * we are, and it was the only connection outside.
313 */
314 ignored_task = NULL;
315
316 if (task_pgrp(parent) != pgrp &&
317 task_session(parent) == task_session(tsk) &&
318 will_become_orphaned_pgrp(pgrp, ignored_task) &&
319 has_stopped_jobs(pgrp)) {
320 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
321 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
322 }
323}
324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325/**
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700326 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 *
328 * If a kernel thread is launched as a result of a system call, or if
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700329 * it ever exits, it should generally reparent itself to kthreadd so it
330 * isn't in the way of other processes and is correctly cleaned up on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 *
332 * The various task state such as scheduling policy and priority may have
333 * been inherited from a user process, so we reset them to sane values here.
334 *
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700335 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700337static void reparent_to_kthreadd(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338{
339 write_lock_irq(&tasklist_lock);
340
341 ptrace_unlink(current);
342 /* Reparent to init */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700343 current->real_parent = current->parent = kthreadd_task;
Roland McGrathf4700212008-03-24 18:36:23 -0700344 list_move_tail(&current->sibling, &current->real_parent->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345
346 /* Set the exit signal to SIGCHLD so we signal init on exit */
347 current->exit_signal = SIGCHLD;
348
Ingo Molnare05606d2007-07-09 18:51:59 +0200349 if (task_nice(current) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 set_user_nice(current, 0);
351 /* cpus_allowed? */
352 /* rt_priority? */
353 /* signals? */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 memcpy(current->signal->rlim, init_task.signal->rlim,
355 sizeof(current->signal->rlim));
David Howellsd84f4f92008-11-14 10:39:23 +1100356
357 atomic_inc(&init_cred.usage);
358 commit_creds(&init_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360}
361
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800362void __set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363{
Oren Laadane19f2472006-01-08 01:03:58 -0800364 struct task_struct *curr = current->group_leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
Christian Borntraeger0d0df592009-10-26 16:49:34 -0700366 if (task_session(curr) != pid)
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700367 change_pid(curr, PIDTYPE_SID, pid);
Oleg Nesterov1b0f7ff2009-04-02 16:58:39 -0700368
369 if (task_pgrp(curr) != pid)
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700370 change_pid(curr, PIDTYPE_PGID, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371}
372
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800373static void set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374{
375 write_lock_irq(&tasklist_lock);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800376 __set_special_pids(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 write_unlock_irq(&tasklist_lock);
378}
379
380/*
Oleg Nesterov87245132009-06-17 16:27:23 -0700381 * Let kernel threads use this to say that they allow a certain signal.
382 * Must not be used if kthread was cloned with CLONE_SIGHAND.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 */
384int allow_signal(int sig)
385{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700386 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 return -EINVAL;
388
389 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov87245132009-06-17 16:27:23 -0700390 /* This is only needed for daemonize()'ed kthreads */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 sigdelset(&current->blocked, sig);
Oleg Nesterov87245132009-06-17 16:27:23 -0700392 /*
393 * Kernel threads handle their own signals. Let the signal code
394 * know it'll be handled, so that they don't get converted to
395 * SIGKILL or just silently dropped.
396 */
397 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 recalc_sigpending();
399 spin_unlock_irq(&current->sighand->siglock);
400 return 0;
401}
402
403EXPORT_SYMBOL(allow_signal);
404
405int disallow_signal(int sig)
406{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700407 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 return -EINVAL;
409
410 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov10ab8252007-05-09 02:34:37 -0700411 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 recalc_sigpending();
413 spin_unlock_irq(&current->sighand->siglock);
414 return 0;
415}
416
417EXPORT_SYMBOL(disallow_signal);
418
419/*
420 * Put all the gunge required to become a kernel thread without
421 * attached user resources in one place where it belongs.
422 */
423
424void daemonize(const char *name, ...)
425{
426 va_list args;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 sigset_t blocked;
428
429 va_start(args, name);
430 vsnprintf(current->comm, sizeof(current->comm), name, args);
431 va_end(args);
432
433 /*
434 * If we were started as result of loading a module, close all of the
435 * user space pages. We don't need them, and if we didn't close them
436 * they would be locked into memory.
437 */
438 exit_mm(current);
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700439 /*
440 * We don't want to have TIF_FREEZE set if the system-wide hibernation
441 * or suspend transition begins right now.
442 */
Oleg Nesterov7b34e422008-07-25 01:47:37 -0700443 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800445 if (current->nsproxy != &init_nsproxy) {
446 get_nsproxy(&init_nsproxy);
447 switch_task_namespaces(current, &init_nsproxy);
448 }
Oleg Nesterov297bd422008-02-08 04:19:10 -0800449 set_special_pids(&init_struct_pid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -0800450 proc_clear_tty(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451
452 /* Block and flush all signals */
453 sigfillset(&blocked);
454 sigprocmask(SIG_BLOCK, &blocked, NULL);
455 flush_signals(current);
456
457 /* Become as one with the init task */
458
Al Viro3e93cd62009-03-29 19:00:13 -0400459 daemonize_fs_struct();
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700460 exit_files(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 current->files = init_task.files;
462 atomic_inc(&current->files->count);
463
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700464 reparent_to_kthreadd();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465}
466
467EXPORT_SYMBOL(daemonize);
468
Arjan van de Ven858119e2006-01-14 13:20:43 -0800469static void close_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470{
471 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700472 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473
474 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700475
476 /*
477 * It is safe to dereference the fd table without RCU or
478 * ->file_lock because this is the last reference to the
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800479 * files structure. But use RCU to shut RCU-lockdep up.
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700480 */
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800481 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700482 fdt = files_fdtable(files);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800483 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 for (;;) {
485 unsigned long set;
486 i = j * __NFDBITS;
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800487 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700489 set = fdt->open_fds->fds_bits[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 while (set) {
491 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700492 struct file * file = xchg(&fdt->fd[i], NULL);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800493 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 filp_close(file, files);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800495 cond_resched();
496 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 }
498 i++;
499 set >>= 1;
500 }
501 }
502}
503
504struct files_struct *get_files_struct(struct task_struct *task)
505{
506 struct files_struct *files;
507
508 task_lock(task);
509 files = task->files;
510 if (files)
511 atomic_inc(&files->count);
512 task_unlock(task);
513
514 return files;
515}
516
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800517void put_files_struct(struct files_struct *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700519 struct fdtable *fdt;
520
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 if (atomic_dec_and_test(&files->count)) {
522 close_files(files);
523 /*
524 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700525 * If the fdtable was embedded, pass files for freeing
526 * at the end of the RCU grace period. Otherwise,
527 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 */
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800529 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700530 fdt = files_fdtable(files);
Vadim Lobanov4fd45812006-12-10 02:21:17 -0800531 if (fdt != &files->fdtab)
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700532 kmem_cache_free(files_cachep, files);
Vadim Lobanov01b2d932006-12-22 01:10:43 -0800533 free_fdtable(fdt);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800534 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 }
536}
537
Al Viro3b125382008-04-22 05:31:30 -0400538void reset_files_struct(struct files_struct *files)
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700539{
Al Viro3b125382008-04-22 05:31:30 -0400540 struct task_struct *tsk = current;
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700541 struct files_struct *old;
542
543 old = tsk->files;
544 task_lock(tsk);
545 tsk->files = files;
546 task_unlock(tsk);
547 put_files_struct(old);
548}
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700549
Al Viro1ec7f1d2008-04-22 05:35:42 -0400550void exit_files(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551{
552 struct files_struct * files = tsk->files;
553
554 if (files) {
555 task_lock(tsk);
556 tsk->files = NULL;
557 task_unlock(tsk);
558 put_files_struct(files);
559 }
560}
561
Balbir Singhcf475ad2008-04-29 01:00:16 -0700562#ifdef CONFIG_MM_OWNER
563/*
564 * Task p is exiting and it owned mm, lets find a new owner for it
565 */
566static inline int
567mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
568{
569 /*
570 * If there are other users of the mm and the owner (us) is exiting
571 * we need to find a new owner to take on the responsibility.
572 */
Balbir Singhcf475ad2008-04-29 01:00:16 -0700573 if (atomic_read(&mm->mm_users) <= 1)
574 return 0;
575 if (mm->owner != p)
576 return 0;
577 return 1;
578}
579
580void mm_update_next_owner(struct mm_struct *mm)
581{
582 struct task_struct *c, *g, *p = current;
583
584retry:
585 if (!mm_need_new_owner(mm, p))
586 return;
587
588 read_lock(&tasklist_lock);
589 /*
590 * Search in the children
591 */
592 list_for_each_entry(c, &p->children, sibling) {
593 if (c->mm == mm)
594 goto assign_new_owner;
595 }
596
597 /*
598 * Search in the siblings
599 */
Oleg Nesterovdea33cf2009-06-17 16:27:29 -0700600 list_for_each_entry(c, &p->real_parent->children, sibling) {
Balbir Singhcf475ad2008-04-29 01:00:16 -0700601 if (c->mm == mm)
602 goto assign_new_owner;
603 }
604
605 /*
606 * Search through everything else. We should not get
607 * here often
608 */
609 do_each_thread(g, c) {
610 if (c->mm == mm)
611 goto assign_new_owner;
612 } while_each_thread(g, c);
613
614 read_unlock(&tasklist_lock);
Balbir Singh31a78f22008-09-28 23:09:31 +0100615 /*
616 * We found no owner yet mm_users > 1: this implies that we are
617 * most likely racing with swapoff (try_to_unuse()) or /proc or
Hugh Dickinse5991372009-01-06 14:39:22 -0800618 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
Balbir Singh31a78f22008-09-28 23:09:31 +0100619 */
Balbir Singh31a78f22008-09-28 23:09:31 +0100620 mm->owner = NULL;
Balbir Singhcf475ad2008-04-29 01:00:16 -0700621 return;
622
623assign_new_owner:
624 BUG_ON(c == p);
625 get_task_struct(c);
626 /*
627 * The task_lock protects c->mm from changing.
628 * We always want mm->owner->mm == mm
629 */
630 task_lock(c);
Hugh Dickinse5991372009-01-06 14:39:22 -0800631 /*
632 * Delay read_unlock() till we have the task_lock()
633 * to ensure that c does not slip away underneath us
634 */
635 read_unlock(&tasklist_lock);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700636 if (c->mm != mm) {
637 task_unlock(c);
638 put_task_struct(c);
639 goto retry;
640 }
Balbir Singhcf475ad2008-04-29 01:00:16 -0700641 mm->owner = c;
642 task_unlock(c);
643 put_task_struct(c);
644}
645#endif /* CONFIG_MM_OWNER */
646
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647/*
648 * Turn us into a lazy TLB process if we
649 * aren't already..
650 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700651static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652{
653 struct mm_struct *mm = tsk->mm;
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700654 struct core_state *core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
656 mm_release(tsk, mm);
657 if (!mm)
658 return;
659 /*
660 * Serialize with any possible pending coredump.
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700661 * We must hold mmap_sem around checking core_state
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 * and clearing tsk->mm. The core-inducing thread
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700663 * will increment ->nr_threads for each thread in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 * group with ->mm != NULL.
665 */
666 down_read(&mm->mmap_sem);
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700667 core_state = mm->core_state;
668 if (core_state) {
669 struct core_thread self;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 up_read(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700672 self.task = tsk;
673 self.next = xchg(&core_state->dumper.next, &self);
674 /*
675 * Implies mb(), the result of xchg() must be visible
676 * to core_state->dumper.
677 */
678 if (atomic_dec_and_test(&core_state->nr_threads))
679 complete(&core_state->startup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
Oleg Nesterova94e2d42008-07-25 01:47:46 -0700681 for (;;) {
682 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
683 if (!self.task) /* see coredump_finish() */
684 break;
685 schedule();
686 }
687 __set_task_state(tsk, TASK_RUNNING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 down_read(&mm->mmap_sem);
689 }
690 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700691 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 /* more a memory barrier than a real lock */
693 task_lock(tsk);
694 tsk->mm = NULL;
695 up_read(&mm->mmap_sem);
696 enter_lazy_tlb(mm, current);
Rafael J. Wysocki0c1eecf2007-07-19 01:47:33 -0700697 /* We don't want this task to be frozen prematurely */
698 clear_freeze_flag(tsk);
Ying Han3d5992d2010-10-26 14:21:23 -0700699 if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
700 atomic_dec(&mm->oom_disable_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 task_unlock(tsk);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700702 mm_update_next_owner(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 mmput(mm);
704}
705
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706/*
707 * When we die, we re-parent all our children.
708 * Try to give them to another thread in our thread
709 * group, and if no such member exists, give it to
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800710 * the child reaper process (ie "init") in our pid
711 * space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 */
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700713static struct task_struct *find_new_reaper(struct task_struct *father)
Namhyung Kimd16e15f2010-10-27 15:34:10 -0700714 __releases(&tasklist_lock)
715 __acquires(&tasklist_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700717 struct pid_namespace *pid_ns = task_active_pid_ns(father);
718 struct task_struct *thread;
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700719
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700720 thread = father;
721 while_each_thread(father, thread) {
722 if (thread->flags & PF_EXITING)
723 continue;
724 if (unlikely(pid_ns->child_reaper == father))
725 pid_ns->child_reaper = thread;
726 return thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700728
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700729 if (unlikely(pid_ns->child_reaper == father)) {
730 write_unlock_irq(&tasklist_lock);
731 if (unlikely(pid_ns == &init_pid_ns))
732 panic("Attempted to kill init!");
733
734 zap_pid_ns_processes(pid_ns);
735 write_lock_irq(&tasklist_lock);
736 /*
737 * We can not clear ->child_reaper or leave it alone.
738 * There may by stealth EXIT_DEAD tasks on ->children,
739 * forget_original_parent() must move them somewhere.
740 */
741 pid_ns->child_reaper = init_pid_ns.child_reaper;
742 }
743
744 return pid_ns->child_reaper;
745}
746
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700747/*
748* Any that need to be release_task'd are put on the @dead list.
749 */
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800750static void reparent_leader(struct task_struct *father, struct task_struct *p,
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700751 struct list_head *dead)
752{
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700753 list_move_tail(&p->sibling, &p->real_parent->children);
754
755 if (task_detached(p))
756 return;
757 /*
758 * If this is a threaded reparent there is no need to
759 * notify anyone anything has happened.
760 */
761 if (same_thread_group(p->real_parent, father))
762 return;
763
764 /* We don't want people slaying init. */
765 p->exit_signal = SIGCHLD;
766
767 /* If it has exited notify the new parent about this child's death. */
Oleg Nesterov5cb11442009-06-17 16:27:30 -0700768 if (!task_ptrace(p) &&
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700769 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
770 do_notify_parent(p, p->exit_signal);
771 if (task_detached(p)) {
772 p->exit_state = EXIT_DEAD;
773 list_move_tail(&p->sibling, dead);
774 }
775 }
776
777 kill_orphaned_pgrp(p, father);
778}
779
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780static void forget_original_parent(struct task_struct *father)
781{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700782 struct task_struct *p, *n, *reaper;
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700783 LIST_HEAD(dead_children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
785 write_lock_irq(&tasklist_lock);
Oleg Nesterovc7e49c12010-08-10 18:03:07 -0700786 /*
787 * Note that exit_ptrace() and find_new_reaper() might
788 * drop tasklist_lock and reacquire it.
789 */
790 exit_ptrace(father);
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700791 reaper = find_new_reaper(father);
Roland McGrathf4700212008-03-24 18:36:23 -0700792
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 list_for_each_entry_safe(p, n, &father->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800794 struct task_struct *t = p;
795 do {
796 t->real_parent = reaper;
797 if (t->parent == father) {
798 BUG_ON(task_ptrace(t));
799 t->parent = t->real_parent;
800 }
801 if (t->pdeath_signal)
802 group_send_sig_info(t->pdeath_signal,
803 SEND_SIG_NOINFO, t);
804 } while_each_thread(p, t);
805 reparent_leader(father, p, &dead_children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700807 write_unlock_irq(&tasklist_lock);
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700808
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700809 BUG_ON(!list_empty(&father->children));
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700810
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700811 list_for_each_entry_safe(p, n, &dead_children, sibling) {
812 list_del_init(&p->sibling);
Oleg Nesterov39c626a2009-04-02 16:58:18 -0700813 release_task(p);
814 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815}
816
817/*
818 * Send signals to all our closest relatives so that they know
819 * to properly mourn us..
820 */
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300821static void exit_notify(struct task_struct *tsk, int group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822{
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700823 int signal;
824 void *cookie;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 /*
827 * This does two things:
828 *
829 * A. Make init inherit all the child processes
830 * B. Check to see if any process groups have become orphaned
831 * as a result of our exiting, and if they have any stopped
832 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
833 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700834 forget_original_parent(tsk);
Pavel Emelyanov2e4a7072007-10-18 23:40:01 -0700835 exit_task_namespaces(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700837 write_lock_irq(&tasklist_lock);
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300838 if (group_dead)
839 kill_orphaned_pgrp(tsk->group_leader, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840
Oleg Nesterov24728442007-08-04 01:04:41 +0400841 /* Let father know we died
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 *
843 * Thread signals are configurable, but you aren't going to use
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700844 * that to send signals to arbitary processes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 * That stops right now.
846 *
847 * If the parent exec id doesn't match the exec id we saved
848 * when we started then we know the parent has changed security
849 * domain.
850 *
851 * If our self_exec id doesn't match our parent_exec_id then
852 * we have changed execution domain as these two values started
853 * the same after a fork.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700855 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300856 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
Oleg Nesterov432870d2009-04-06 16:16:02 +0200857 tsk->self_exec_id != tsk->parent_exec_id))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 tsk->exit_signal = SIGCHLD;
859
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700860 signal = tracehook_notify_death(tsk, &cookie, group_dead);
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700861 if (signal >= 0)
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700862 signal = do_notify_parent(tsk, signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700864 tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
Oleg Nesterov9c339162010-05-26 14:43:10 -0700866 /* mt-exec, de_thread() is waiting for group leader */
867 if (unlikely(tsk->signal->notify_count < 0))
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700868 wake_up_process(tsk->signal->group_exit_task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 write_unlock_irq(&tasklist_lock);
870
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700871 tracehook_report_death(tsk, signal, cookie, group_dead);
872
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 /* If the process is dead, release it - nobody will wait for it */
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700874 if (signal == DEATH_REAP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 release_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876}
877
Jeff Dikee18eecb2007-07-15 23:38:48 -0700878#ifdef CONFIG_DEBUG_STACK_USAGE
879static void check_stack_usage(void)
880{
881 static DEFINE_SPINLOCK(low_water_lock);
882 static int lowest_to_date = THREAD_SIZE;
Jeff Dikee18eecb2007-07-15 23:38:48 -0700883 unsigned long free;
884
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500885 free = stack_not_used(current);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700886
887 if (free >= lowest_to_date)
888 return;
889
890 spin_lock(&low_water_lock);
891 if (free < lowest_to_date) {
892 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
893 "left\n",
894 current->comm, free);
895 lowest_to_date = free;
896 }
897 spin_unlock(&low_water_lock);
898}
899#else
900static inline void check_stack_usage(void) {}
901#endif
902
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800903NORET_TYPE void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904{
905 struct task_struct *tsk = current;
906 int group_dead;
907
908 profile_task_exit(tsk);
909
Jens Axboe22e2c502005-06-27 10:55:12 +0200910 WARN_ON(atomic_read(&tsk->fs_excl));
911
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 if (unlikely(in_interrupt()))
913 panic("Aiee, killing interrupt handler!");
914 if (unlikely(!tsk->pid))
915 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
Roland McGrath30199f52008-07-25 19:45:46 -0700917 tracehook_report_exit(&code);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918
David Howellse0e81732009-09-02 09:13:40 +0100919 validate_creds_for_do_exit(tsk);
920
Alexander Nybergdf164db2005-06-23 00:09:13 -0700921 /*
922 * We're taking recursive faults here in do_exit. Safest is to just
923 * leave this task alone and wait for reboot.
924 */
925 if (unlikely(tsk->flags & PF_EXITING)) {
926 printk(KERN_ALERT
927 "Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700928 /*
929 * We can do this unlocked here. The futex code uses
930 * this flag just to verify whether the pi state
931 * cleanup has been done or not. In the worst case it
932 * loops once more. We pretend that the cleanup was
933 * done as there is no way to return. Either the
934 * OWNER_DIED bit is set by now or we push the blocked
935 * task into the wait for ever nirwana as well.
936 */
937 tsk->flags |= PF_EXITPIDONE;
Alexander Nybergdf164db2005-06-23 00:09:13 -0700938 set_current_state(TASK_UNINTERRUPTIBLE);
939 schedule();
940 }
941
Thomas Gleixner3aa551c2009-03-23 18:28:15 +0100942 exit_irq_thread();
943
Oleg Nesterovd12619b2008-02-08 04:19:12 -0800944 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700945 /*
946 * tsk->flags are checked in the futex code to protect against
947 * an exiting task cleaning up the robust pi futexes.
948 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -0700949 smp_mb();
Thomas Gleixner1d615482009-11-17 14:54:03 +0100950 raw_spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 if (unlikely(in_atomic()))
953 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -0700954 current->comm, task_pid_nr(current),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 preempt_count());
956
957 acct_update_integrals(tsk);
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -0800958 /* sync mm's RSS info before statistics gathering */
KAMEZAWA Hiroyukia3a2e762010-04-06 14:34:42 -0700959 if (tsk->mm)
960 sync_mm_rss(tsk, tsk->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -0700962 if (group_dead) {
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700963 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -0700964 exit_itimers(tsk->signal);
Jiri Pirko1f102062009-09-22 16:44:10 -0700965 if (tsk->mm)
966 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
Andrew Mortonc3068952005-08-04 16:49:32 -0700967 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700968 acct_collect(code, group_dead);
Miloslav Trmac522ed772007-07-15 23:40:56 -0700969 if (group_dead)
970 tty_audit_exit();
Al Virofa84cb92006-03-29 20:30:19 -0500971 if (unlikely(tsk->audit_context))
972 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -0800973
Jonathan Limf2ab6d82007-08-30 23:56:23 -0700974 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -0800975 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -0700976
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 exit_mm(tsk);
978
KaiGai Kohei0e464812006-06-25 05:49:24 -0700979 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700980 acct_process();
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400981 trace_sched_process_exit(tsk);
982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 exit_sem(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400984 exit_files(tsk);
985 exit_fs(tsk);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700986 check_stack_usage();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 exit_thread();
Paul Menageb4f48b62007-10-18 23:39:33 -0700988 cgroup_exit(tsk, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989
Alan Cox5ec93d12009-11-30 13:18:45 +0000990 if (group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 disassociate_ctty(1);
992
Al Viroa1261f52005-11-13 16:06:55 -0800993 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
Matt Helsley9f460802005-11-07 00:59:16 -0800995 proc_exit_connector(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200996
997 /*
Frederic Weisbecker24f1e32c2009-09-09 19:22:48 +0200998 * FIXME: do that only when needed, using sched_exit tracepoint
999 */
1000 flush_ptrace_hw_breakpoint(tsk);
1001 /*
Ingo Molnar33b2fb32009-05-17 11:08:41 +02001002 * Flush inherited counters to the parent - before the parent
1003 * gets woken up by child-exit notifications.
1004 */
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001005 perf_event_exit_task(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +02001006
Oleg Nesterov821c7de2008-03-02 21:44:44 +03001007 exit_notify(tsk, group_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008#ifdef CONFIG_NUMA
Miao Xiec0ff7452010-05-24 14:32:08 -07001009 task_lock(tsk);
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001010 mpol_put(tsk->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 tsk->mempolicy = NULL;
Miao Xiec0ff7452010-05-24 14:32:08 -07001012 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001014#ifdef CONFIG_FUTEX
Ingo Molnarc87e2832006-06-27 02:54:58 -07001015 if (unlikely(current->pi_state_cache))
1016 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001017#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001018 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001019 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -08001020 */
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001021 debug_check_no_locks_held(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001022 /*
1023 * We can do this unlocked here. The futex code uses this flag
1024 * just to verify whether the pi state cleanup has been done
1025 * or not. In the worst case it loops once more.
1026 */
1027 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028
Al Viroafc847b2006-02-28 12:51:55 -05001029 if (tsk->io_context)
Louis Rillingb69f2292009-12-04 14:52:42 +01001030 exit_io_context(tsk);
Al Viroafc847b2006-02-28 12:51:55 -05001031
Jens Axboeb92ce552006-04-11 13:52:07 +02001032 if (tsk->splice_pipe)
1033 __free_pipe_info(tsk->splice_pipe);
1034
David Howellse0e81732009-09-02 09:13:40 +01001035 validate_creds_for_do_exit(tsk);
1036
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001037 preempt_disable();
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001038 exit_rcu();
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001039 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -07001040 tsk->state = TASK_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 schedule();
1042 BUG();
1043 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -07001044 for (;;)
1045 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046}
1047
Russ Anderson012914d2005-04-23 00:08:00 -07001048EXPORT_SYMBOL_GPL(do_exit);
1049
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1051{
1052 if (comp)
1053 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001054
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 do_exit(code);
1056}
1057
1058EXPORT_SYMBOL(complete_and_exit);
1059
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001060SYSCALL_DEFINE1(exit, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061{
1062 do_exit((error_code&0xff)<<8);
1063}
1064
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065/*
1066 * Take down every thread in the group. This is called by fatal signals
1067 * as well as by sys_exit_group (below).
1068 */
1069NORET_TYPE void
1070do_group_exit(int exit_code)
1071{
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001072 struct signal_struct *sig = current->signal;
1073
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1075
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001076 if (signal_group_exit(sig))
1077 exit_code = sig->group_exit_code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 else if (!thread_group_empty(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 struct sighand_struct *const sighand = current->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001081 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 /* Another thread got here before we took the lock. */
1083 exit_code = sig->group_exit_code;
1084 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001086 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 zap_other_threads(current);
1088 }
1089 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 }
1091
1092 do_exit(exit_code);
1093 /* NOTREACHED */
1094}
1095
1096/*
1097 * this kills every thread in the thread group. Note that any externally
1098 * wait4()-ing process will get the correct exit code - even if this
1099 * thread is not the thread group leader.
1100 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001101SYSCALL_DEFINE1(exit_group, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102{
1103 do_group_exit((error_code & 0xff) << 8);
Heiko Carstens2ed7c032009-01-14 14:13:54 +01001104 /* NOTREACHED */
1105 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106}
1107
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001108struct wait_opts {
1109 enum pid_type wo_type;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001110 int wo_flags;
Richard Kennedye1eb1eb2009-06-17 16:27:42 -07001111 struct pid *wo_pid;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001112
1113 struct siginfo __user *wo_info;
1114 int __user *wo_stat;
1115 struct rusage __user *wo_rusage;
1116
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001117 wait_queue_t child_wait;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001118 int notask_error;
1119};
1120
Oleg Nesterov989264f2009-09-23 15:56:49 -07001121static inline
1122struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
Eric W. Biederman161550d2008-02-08 04:19:14 -08001123{
Oleg Nesterov989264f2009-09-23 15:56:49 -07001124 if (type != PIDTYPE_PID)
1125 task = task->group_leader;
1126 return task->pids[type].pid;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001127}
1128
Oleg Nesterov989264f2009-09-23 15:56:49 -07001129static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001130{
1131 return wo->wo_type == PIDTYPE_MAX ||
1132 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1133}
1134
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001135static int eligible_child(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136{
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001137 if (!eligible_pid(wo, p))
1138 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 /* Wait for all children (clone and not) if __WALL is set;
1140 * otherwise, wait for clone children *only* if __WCLONE is
1141 * set; otherwise, wait for non-clone children *only*. (Note:
1142 * A "clone" child here is one that reports to its parent
1143 * using a signal other than SIGCHLD.) */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001144 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1145 && !(wo->wo_flags & __WALL))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
Roland McGrath14dd0b82008-03-30 18:41:25 -07001148 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149}
1150
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001151static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1152 pid_t pid, uid_t uid, int why, int status)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001154 struct siginfo __user *infop;
1155 int retval = wo->wo_rusage
1156 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -07001157
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001159 infop = wo->wo_info;
Vitaly Mayatskikhb6fe2d12009-09-23 15:56:52 -07001160 if (infop) {
1161 if (!retval)
1162 retval = put_user(SIGCHLD, &infop->si_signo);
1163 if (!retval)
1164 retval = put_user(0, &infop->si_errno);
1165 if (!retval)
1166 retval = put_user((short)why, &infop->si_code);
1167 if (!retval)
1168 retval = put_user(pid, &infop->si_pid);
1169 if (!retval)
1170 retval = put_user(uid, &infop->si_uid);
1171 if (!retval)
1172 retval = put_user(status, &infop->si_status);
1173 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 if (!retval)
1175 retval = pid;
1176 return retval;
1177}
1178
1179/*
1180 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1181 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1182 * the lock and this task is uninteresting. If we return nonzero, we have
1183 * released the lock and the system call should return.
1184 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001185static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186{
1187 unsigned long state;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001188 int retval, status, traced;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001189 pid_t pid = task_pid_vnr(p);
David Howellsc69e8d92008-11-14 10:39:19 +11001190 uid_t uid = __task_cred(p)->uid;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001191 struct siginfo __user *infop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001193 if (!likely(wo->wo_flags & WEXITED))
Roland McGrath98abed02008-03-19 19:24:59 -07001194 return 0;
1195
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001196 if (unlikely(wo->wo_flags & WNOWAIT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 int exit_code = p->exit_code;
Thiago Farinaf3abd4f2010-03-05 13:42:52 -08001198 int why;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 get_task_struct(p);
1201 read_unlock(&tasklist_lock);
1202 if ((exit_code & 0x7f) == 0) {
1203 why = CLD_EXITED;
1204 status = exit_code >> 8;
1205 } else {
1206 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1207 status = exit_code & 0x7f;
1208 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001209 return wait_noreap_copyout(wo, p, pid, uid, why, status);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 }
1211
1212 /*
1213 * Try to move the task's state to DEAD
1214 * only one thread is allowed to do this:
1215 */
1216 state = xchg(&p->exit_state, EXIT_DEAD);
1217 if (state != EXIT_ZOMBIE) {
1218 BUG_ON(state != EXIT_DEAD);
1219 return 0;
1220 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221
Oleg Nesterov53b6f9f2008-04-30 00:53:13 -07001222 traced = ptrace_reparented(p);
Oleg Nesterovbefca962009-06-18 16:49:11 -07001223 /*
1224 * It can be ptraced but not reparented, check
1225 * !task_detached() to filter out sub-threads.
1226 */
1227 if (likely(!traced) && likely(!task_detached(p))) {
Jesper Juhl3795e162006-01-09 20:54:39 -08001228 struct signal_struct *psig;
1229 struct signal_struct *sig;
Jiri Pirko1f102062009-09-22 16:44:10 -07001230 unsigned long maxrss;
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001231 cputime_t tgutime, tgstime;
Jesper Juhl3795e162006-01-09 20:54:39 -08001232
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 /*
1234 * The resource counters for the group leader are in its
1235 * own task_struct. Those for dead threads in the group
1236 * are in its signal_struct, as are those for the child
1237 * processes it has previously reaped. All these
1238 * accumulate in the parent's signal_struct c* fields.
1239 *
1240 * We don't bother to take a lock here to protect these
1241 * p->signal fields, because they are only touched by
1242 * __exit_signal, which runs with tasklist_lock
1243 * write-locked anyway, and so is excluded here. We do
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001244 * need to protect the access to parent->signal fields,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 * as other threads in the parent group can be right
1246 * here reaping other children at the same time.
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001247 *
1248 * We use thread_group_times() to get times for the thread
1249 * group, which consolidates times for all threads in the
1250 * group including the group leader.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 */
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001252 thread_group_times(p, &tgutime, &tgstime);
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001253 spin_lock_irq(&p->real_parent->sighand->siglock);
1254 psig = p->real_parent->signal;
Jesper Juhl3795e162006-01-09 20:54:39 -08001255 sig = p->signal;
1256 psig->cutime =
1257 cputime_add(psig->cutime,
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001258 cputime_add(tgutime,
1259 sig->cutime));
Jesper Juhl3795e162006-01-09 20:54:39 -08001260 psig->cstime =
1261 cputime_add(psig->cstime,
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001262 cputime_add(tgstime,
1263 sig->cstime));
Laurent Vivier9ac52312007-10-15 17:00:19 +02001264 psig->cgtime =
1265 cputime_add(psig->cgtime,
1266 cputime_add(p->gtime,
1267 cputime_add(sig->gtime,
1268 sig->cgtime)));
Jesper Juhl3795e162006-01-09 20:54:39 -08001269 psig->cmin_flt +=
1270 p->min_flt + sig->min_flt + sig->cmin_flt;
1271 psig->cmaj_flt +=
1272 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1273 psig->cnvcsw +=
1274 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1275 psig->cnivcsw +=
1276 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001277 psig->cinblock +=
1278 task_io_get_inblock(p) +
1279 sig->inblock + sig->cinblock;
1280 psig->coublock +=
1281 task_io_get_oublock(p) +
1282 sig->oublock + sig->coublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001283 maxrss = max(sig->maxrss, sig->cmaxrss);
1284 if (psig->cmaxrss < maxrss)
1285 psig->cmaxrss = maxrss;
Andrea Righi59954772008-07-27 17:29:15 +02001286 task_io_accounting_add(&psig->ioac, &p->ioac);
1287 task_io_accounting_add(&psig->ioac, &sig->ioac);
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001288 spin_unlock_irq(&p->real_parent->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 }
1290
1291 /*
1292 * Now we are sure this task is interesting, and no other
1293 * thread can reap it because we set its state to EXIT_DEAD.
1294 */
1295 read_unlock(&tasklist_lock);
1296
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001297 retval = wo->wo_rusage
1298 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1300 ? p->signal->group_exit_code : p->exit_code;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001301 if (!retval && wo->wo_stat)
1302 retval = put_user(status, wo->wo_stat);
1303
1304 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 if (!retval && infop)
1306 retval = put_user(SIGCHLD, &infop->si_signo);
1307 if (!retval && infop)
1308 retval = put_user(0, &infop->si_errno);
1309 if (!retval && infop) {
1310 int why;
1311
1312 if ((status & 0x7f) == 0) {
1313 why = CLD_EXITED;
1314 status >>= 8;
1315 } else {
1316 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1317 status &= 0x7f;
1318 }
1319 retval = put_user((short)why, &infop->si_code);
1320 if (!retval)
1321 retval = put_user(status, &infop->si_status);
1322 }
1323 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001324 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 if (!retval && infop)
David Howellsc69e8d92008-11-14 10:39:19 +11001326 retval = put_user(uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001327 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001328 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001329
1330 if (traced) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001332 /* We dropped tasklist, ptracer could die and untrace */
1333 ptrace_unlink(p);
1334 /*
1335 * If this is not a detached task, notify the parent.
1336 * If it's still not detached after that, don't release
1337 * it now.
1338 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001339 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001340 do_notify_parent(p, p->exit_signal);
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001341 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001342 p->exit_state = EXIT_ZOMBIE;
1343 p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 }
1345 }
1346 write_unlock_irq(&tasklist_lock);
1347 }
1348 if (p != NULL)
1349 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001350
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 return retval;
1352}
1353
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001354static int *task_stopped_code(struct task_struct *p, bool ptrace)
1355{
1356 if (ptrace) {
1357 if (task_is_stopped_or_traced(p))
1358 return &p->exit_code;
1359 } else {
1360 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1361 return &p->signal->group_exit_code;
1362 }
1363 return NULL;
1364}
1365
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366/*
1367 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1368 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1369 * the lock and this task is uninteresting. If we return nonzero, we have
1370 * released the lock and the system call should return.
1371 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001372static int wait_task_stopped(struct wait_opts *wo,
1373 int ptrace, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001375 struct siginfo __user *infop;
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001376 int retval, exit_code, *p_code, why;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001377 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001378 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379
Oleg Nesterov47918022009-06-17 16:27:39 -07001380 /*
1381 * Traditionally we see ptrace'd stopped tasks regardless of options.
1382 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001383 if (!ptrace && !(wo->wo_flags & WUNTRACED))
Roland McGrath98abed02008-03-19 19:24:59 -07001384 return 0;
1385
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001386 exit_code = 0;
1387 spin_lock_irq(&p->sighand->siglock);
1388
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001389 p_code = task_stopped_code(p, ptrace);
1390 if (unlikely(!p_code))
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001391 goto unlock_sig;
1392
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001393 exit_code = *p_code;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001394 if (!exit_code)
1395 goto unlock_sig;
1396
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001397 if (!unlikely(wo->wo_flags & WNOWAIT))
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001398 *p_code = 0;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001399
Daniel J Bluemanf362b732010-08-17 23:56:55 +01001400 uid = task_uid(p);
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001401unlock_sig:
1402 spin_unlock_irq(&p->sighand->siglock);
1403 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 return 0;
1405
1406 /*
1407 * Now we are pretty sure this task is interesting.
1408 * Make sure it doesn't get reaped out from under us while we
1409 * give up the lock and then examine it below. We don't want to
1410 * keep holding onto the tasklist_lock while we call getrusage and
1411 * possibly take page faults for user memory.
1412 */
1413 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001414 pid = task_pid_vnr(p);
Roland McGrathf4700212008-03-24 18:36:23 -07001415 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 read_unlock(&tasklist_lock);
1417
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001418 if (unlikely(wo->wo_flags & WNOWAIT))
1419 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001421 retval = wo->wo_rusage
1422 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1423 if (!retval && wo->wo_stat)
1424 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1425
1426 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 if (!retval && infop)
1428 retval = put_user(SIGCHLD, &infop->si_signo);
1429 if (!retval && infop)
1430 retval = put_user(0, &infop->si_errno);
1431 if (!retval && infop)
Roland McGrath6efcae42008-03-08 11:41:22 -08001432 retval = put_user((short)why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 if (!retval && infop)
1434 retval = put_user(exit_code, &infop->si_status);
1435 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001436 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001438 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001440 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 put_task_struct(p);
1442
1443 BUG_ON(!retval);
1444 return retval;
1445}
1446
1447/*
1448 * Handle do_wait work for one task in a live, non-stopped state.
1449 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1450 * the lock and this task is uninteresting. If we return nonzero, we have
1451 * released the lock and the system call should return.
1452 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001453static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454{
1455 int retval;
1456 pid_t pid;
1457 uid_t uid;
1458
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001459 if (!unlikely(wo->wo_flags & WCONTINUED))
Roland McGrath98abed02008-03-19 19:24:59 -07001460 return 0;
1461
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1463 return 0;
1464
1465 spin_lock_irq(&p->sighand->siglock);
1466 /* Re-check with the lock held. */
1467 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1468 spin_unlock_irq(&p->sighand->siglock);
1469 return 0;
1470 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001471 if (!unlikely(wo->wo_flags & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
Daniel J Bluemanf362b732010-08-17 23:56:55 +01001473 uid = task_uid(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 spin_unlock_irq(&p->sighand->siglock);
1475
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001476 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 get_task_struct(p);
1478 read_unlock(&tasklist_lock);
1479
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001480 if (!wo->wo_info) {
1481 retval = wo->wo_rusage
1482 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001484 if (!retval && wo->wo_stat)
1485 retval = put_user(0xffff, wo->wo_stat);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001487 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 } else {
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001489 retval = wait_noreap_copyout(wo, p, pid, uid,
1490 CLD_CONTINUED, SIGCONT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 BUG_ON(retval == 0);
1492 }
1493
1494 return retval;
1495}
1496
Roland McGrath98abed02008-03-19 19:24:59 -07001497/*
1498 * Consider @p for a wait by @parent.
1499 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001500 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001501 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1502 * Returns zero if the search for a child should continue;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001503 * then ->notask_error is 0 if @p is an eligible child,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001504 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001505 */
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001506static int wait_consider_task(struct wait_opts *wo, int ptrace,
1507 struct task_struct *p)
Roland McGrath98abed02008-03-19 19:24:59 -07001508{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001509 int ret = eligible_child(wo, p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001510 if (!ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001511 return ret;
1512
Oleg Nesterova2322e12009-09-23 15:56:45 -07001513 ret = security_task_wait(p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001514 if (unlikely(ret < 0)) {
1515 /*
1516 * If we have not yet seen any eligible child,
1517 * then let this error code replace -ECHILD.
1518 * A permission error will give the user a clue
1519 * to look for security policy problems, rather
1520 * than for mysterious wait bugs.
1521 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001522 if (wo->notask_error)
1523 wo->notask_error = ret;
Oleg Nesterov78a3d9d2009-04-29 18:01:23 +02001524 return 0;
Roland McGrath14dd0b82008-03-30 18:41:25 -07001525 }
1526
Oleg Nesterov5cb11442009-06-17 16:27:30 -07001527 if (likely(!ptrace) && unlikely(task_ptrace(p))) {
Roland McGrathf4700212008-03-24 18:36:23 -07001528 /*
1529 * This child is hidden by ptrace.
1530 * We aren't allowed to see it now, but eventually we will.
1531 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001532 wo->notask_error = 0;
Roland McGrathf4700212008-03-24 18:36:23 -07001533 return 0;
1534 }
1535
Roland McGrath98abed02008-03-19 19:24:59 -07001536 if (p->exit_state == EXIT_DEAD)
1537 return 0;
1538
1539 /*
1540 * We don't reap group leaders with subthreads.
1541 */
1542 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001543 return wait_task_zombie(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001544
1545 /*
1546 * It's stopped or running now, so it might
1547 * later continue, exit, or stop again.
1548 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001549 wo->notask_error = 0;
Roland McGrath98abed02008-03-19 19:24:59 -07001550
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001551 if (task_stopped_code(p, ptrace))
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001552 return wait_task_stopped(wo, ptrace, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001553
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001554 return wait_task_continued(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001555}
1556
1557/*
1558 * Do the work of do_wait() for one thread in the group, @tsk.
1559 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001560 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001561 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1562 * Returns zero if the search for a child should continue; then
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001563 * ->notask_error is 0 if there were any eligible children,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001564 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001565 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001566static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001567{
1568 struct task_struct *p;
1569
1570 list_for_each_entry(p, &tsk->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001571 int ret = wait_consider_task(wo, 0, p);
1572 if (ret)
1573 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001574 }
1575
1576 return 0;
1577}
1578
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001579static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001580{
1581 struct task_struct *p;
1582
Roland McGrathf4700212008-03-24 18:36:23 -07001583 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001584 int ret = wait_consider_task(wo, 1, p);
Roland McGrathf4700212008-03-24 18:36:23 -07001585 if (ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001586 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001587 }
1588
1589 return 0;
1590}
1591
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001592static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1593 int sync, void *key)
1594{
1595 struct wait_opts *wo = container_of(wait, struct wait_opts,
1596 child_wait);
1597 struct task_struct *p = key;
1598
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001599 if (!eligible_pid(wo, p))
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001600 return 0;
1601
Oleg Nesterovb4fe5182009-09-23 15:56:47 -07001602 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1603 return 0;
1604
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001605 return default_wake_function(wait, mode, sync, key);
1606}
1607
Oleg Nesterova7f07652009-09-23 15:56:44 -07001608void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1609{
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001610 __wake_up_sync_key(&parent->signal->wait_chldexit,
1611 TASK_INTERRUPTIBLE, 1, p);
Oleg Nesterova7f07652009-09-23 15:56:44 -07001612}
1613
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001614static long do_wait(struct wait_opts *wo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 struct task_struct *tsk;
Roland McGrath98abed02008-03-19 19:24:59 -07001617 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001619 trace_sched_process_wait(wo->wo_pid);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001620
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001621 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1622 wo->child_wait.private = current;
1623 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624repeat:
Roland McGrath98abed02008-03-19 19:24:59 -07001625 /*
1626 * If there is nothing that can match our critiera just get out.
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001627 * We will clear ->notask_error to zero if we see any child that
1628 * might later match our criteria, even if we are not able to reap
1629 * it yet.
Roland McGrath98abed02008-03-19 19:24:59 -07001630 */
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001631 wo->notask_error = -ECHILD;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001632 if ((wo->wo_type < PIDTYPE_MAX) &&
1633 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001634 goto notask;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001635
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001636 set_current_state(TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 read_lock(&tasklist_lock);
1638 tsk = current;
1639 do {
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001640 retval = do_wait_thread(wo, tsk);
1641 if (retval)
Roland McGrath98abed02008-03-19 19:24:59 -07001642 goto end;
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001643
1644 retval = ptrace_do_wait(wo, tsk);
1645 if (retval)
1646 goto end;
Roland McGrath98abed02008-03-19 19:24:59 -07001647
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001648 if (wo->wo_flags & __WNOTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 break;
Oleg Nesterova3f6dfb2009-06-17 16:27:41 -07001650 } while_each_thread(current, tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001652
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001653notask:
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001654 retval = wo->notask_error;
1655 if (!retval && !(wo->wo_flags & WNOHANG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 retval = -ERESTARTSYS;
Roland McGrath98abed02008-03-19 19:24:59 -07001657 if (!signal_pending(current)) {
1658 schedule();
1659 goto repeat;
1660 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662end:
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001663 __set_current_state(TASK_RUNNING);
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001664 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 return retval;
1666}
1667
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001668SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1669 infop, int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001671 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001672 struct pid *pid = NULL;
1673 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 long ret;
1675
1676 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1677 return -EINVAL;
1678 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1679 return -EINVAL;
1680
1681 switch (which) {
1682 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001683 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 break;
1685 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001686 type = PIDTYPE_PID;
1687 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 return -EINVAL;
1689 break;
1690 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001691 type = PIDTYPE_PGID;
1692 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 break;
1695 default:
1696 return -EINVAL;
1697 }
1698
Eric W. Biederman161550d2008-02-08 04:19:14 -08001699 if (type < PIDTYPE_MAX)
1700 pid = find_get_pid(upid);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001701
1702 wo.wo_type = type;
1703 wo.wo_pid = pid;
1704 wo.wo_flags = options;
1705 wo.wo_info = infop;
1706 wo.wo_stat = NULL;
1707 wo.wo_rusage = ru;
1708 ret = do_wait(&wo);
Vitaly Mayatskikhdfe16df2009-09-23 15:56:51 -07001709
1710 if (ret > 0) {
1711 ret = 0;
1712 } else if (infop) {
1713 /*
1714 * For a WNOHANG return, clear out all the fields
1715 * we would set so the user can easily tell the
1716 * difference.
1717 */
1718 if (!ret)
1719 ret = put_user(0, &infop->si_signo);
1720 if (!ret)
1721 ret = put_user(0, &infop->si_errno);
1722 if (!ret)
1723 ret = put_user(0, &infop->si_code);
1724 if (!ret)
1725 ret = put_user(0, &infop->si_pid);
1726 if (!ret)
1727 ret = put_user(0, &infop->si_uid);
1728 if (!ret)
1729 ret = put_user(0, &infop->si_status);
1730 }
1731
Eric W. Biederman161550d2008-02-08 04:19:14 -08001732 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733
1734 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001735 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 return ret;
1737}
1738
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001739SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1740 int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001742 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001743 struct pid *pid = NULL;
1744 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 long ret;
1746
1747 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1748 __WNOTHREAD|__WCLONE|__WALL))
1749 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001750
1751 if (upid == -1)
1752 type = PIDTYPE_MAX;
1753 else if (upid < 0) {
1754 type = PIDTYPE_PGID;
1755 pid = find_get_pid(-upid);
1756 } else if (upid == 0) {
1757 type = PIDTYPE_PGID;
Oleg Nesterov2ae448e2009-04-02 16:58:36 -07001758 pid = get_task_pid(current, PIDTYPE_PGID);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001759 } else /* upid > 0 */ {
1760 type = PIDTYPE_PID;
1761 pid = find_get_pid(upid);
1762 }
1763
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001764 wo.wo_type = type;
1765 wo.wo_pid = pid;
1766 wo.wo_flags = options | WEXITED;
1767 wo.wo_info = NULL;
1768 wo.wo_stat = stat_addr;
1769 wo.wo_rusage = ru;
1770 ret = do_wait(&wo);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001771 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772
1773 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001774 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 return ret;
1776}
1777
1778#ifdef __ARCH_WANT_SYS_WAITPID
1779
1780/*
1781 * sys_waitpid() remains for compatibility. waitpid() should be
1782 * implemented by calling sys_wait4() from libc.a.
1783 */
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001784SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785{
1786 return sys_wait4(pid, stat_addr, options, NULL);
1787}
1788
1789#endif