blob: ceffc67b564ae28eb6b46be67bfcf4f0ab7e8c6a [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
54#include <asm/uaccess.h>
55#include <asm/unistd.h>
56#include <asm/pgtable.h>
57#include <asm/mmu_context.h>
58
Adrian Bunk408b6642005-05-01 08:59:29 -070059static void exit_mm(struct task_struct * tsk);
60
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070061static void __unhash_process(struct task_struct *p, bool group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -070062{
63 nr_threads--;
64 detach_pid(p, PIDTYPE_PID);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070065 if (group_dead) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 detach_pid(p, PIDTYPE_PGID);
67 detach_pid(p, PIDTYPE_SID);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080068
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -070069 list_del_rcu(&p->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -080070 list_del_init(&p->sibling);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -080071 __get_cpu_var(process_counts)--;
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 }
Oleg Nesterov47e65322006-03-28 16:11:25 -080073 list_del_rcu(&p->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -070074}
75
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080076/*
77 * This function expects the tasklist_lock write-locked.
78 */
79static void __exit_signal(struct task_struct *tsk)
80{
81 struct signal_struct *sig = tsk->signal;
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070082 bool group_dead = thread_group_leader(tsk);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080083 struct sighand_struct *sighand;
Oleg Nesterov4ada8562010-05-26 14:43:17 -070084 struct tty_struct *uninitialized_var(tty);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080085
Paul E. McKenneyd11c5632010-02-22 17:04:50 -080086 sighand = rcu_dereference_check(tsk->sighand,
87 rcu_read_lock_held() ||
Paul E. McKenneydb1466b2010-03-03 07:46:56 -080088 lockdep_tasklist_lock_is_held());
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080089 spin_lock(&sighand->siglock);
90
91 posix_cpu_timers_exit(tsk);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -070092 if (group_dead) {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080093 posix_cpu_timers_exit_group(tsk);
Oleg Nesterov4ada8562010-05-26 14:43:17 -070094 tty = sig->tty;
95 sig->tty = NULL;
Oleg Nesterov4a599942010-05-26 14:43:12 -070096 } else {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080097 /*
98 * If there is any task waiting for the group exit
99 * then notify it:
100 */
Oleg Nesterovd3441932010-05-26 14:43:11 -0700101 if (sig->notify_count > 0 && !--sig->notify_count)
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800102 wake_up_process(sig->group_exit_task);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700103
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800104 if (tsk == sig->curr_target)
105 sig->curr_target = next_thread(tsk);
106 /*
107 * Accumulate here the counters for all threads but the
108 * group leader as they die, so they can be added into
109 * the process-wide totals when those are taken.
110 * The group leader stays around as a zombie as long
111 * as there are other threads. When it gets reaped,
112 * the exit.c code will add its counts into these totals.
113 * We won't ever get here for the group leader, since it
114 * will have been the last reference on the signal_struct.
115 */
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +0900116 sig->utime = cputime_add(sig->utime, tsk->utime);
117 sig->stime = cputime_add(sig->stime, tsk->stime);
Hidetoshi Setod5b7c782009-11-26 14:49:05 +0900118 sig->gtime = cputime_add(sig->gtime, tsk->gtime);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800119 sig->min_flt += tsk->min_flt;
120 sig->maj_flt += tsk->maj_flt;
121 sig->nvcsw += tsk->nvcsw;
122 sig->nivcsw += tsk->nivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700123 sig->inblock += task_io_get_inblock(tsk);
124 sig->oublock += task_io_get_oublock(tsk);
Andrea Righi59954772008-07-27 17:29:15 +0200125 task_io_accounting_add(&sig->ioac, &tsk->ioac);
Peter Zijlstra32bd6712009-02-05 12:24:15 +0100126 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800127 }
128
Oleg Nesterovb3ac0222010-05-26 14:43:24 -0700129 sig->nr_threads--;
Oleg Nesterovd40e48e2010-05-26 14:43:19 -0700130 __unhash_process(tsk, group_dead);
Oleg Nesterov58767002006-03-28 16:11:20 -0800131
Oleg Nesterovda7978b2008-05-23 13:04:41 -0700132 /*
133 * Do this under ->siglock, we can race with another thread
134 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
135 */
136 flush_sigqueue(&tsk->pending);
Oleg Nesterova7e53282006-03-28 16:11:27 -0800137 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800138 spin_unlock(&sighand->siglock);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800139
Oleg Nesterova7e53282006-03-28 16:11:27 -0800140 __cleanup_sighand(sighand);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800141 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
Oleg Nesterovd40e48e2010-05-26 14:43:19 -0700142 if (group_dead) {
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800143 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov4ada8562010-05-26 14:43:17 -0700144 tty_kref_put(tty);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800145 }
146}
147
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800148static void delayed_put_task_struct(struct rcu_head *rhp)
149{
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400150 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
151
Ingo Molnarcdd6c482009-09-21 12:02:48 +0200152#ifdef CONFIG_PERF_EVENTS
153 WARN_ON_ONCE(tsk->perf_event_ctxp);
Ingo Molnareef6cbf2008-12-19 10:20:42 +0100154#endif
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400155 trace_sched_process_free(tsk);
156 put_task_struct(tsk);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800157}
158
Roland McGrathf4700212008-03-24 18:36:23 -0700159
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160void release_task(struct task_struct * p)
161{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700162 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800164repeat:
Roland McGrathdae33572008-07-25 19:45:48 -0700165 tracehook_prepare_release_task(p);
David Howellsc69e8d92008-11-14 10:39:19 +1100166 /* don't need to get the RCU readlock here - the process is dead and
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800167 * can't be modifying its own credentials. But shut RCU-lockdep up */
168 rcu_read_lock();
David Howellsc69e8d92008-11-14 10:39:19 +1100169 atomic_dec(&__task_cred(p)->user->processes);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800170 rcu_read_unlock();
David Howellsc69e8d92008-11-14 10:39:19 +1100171
Pavel Emelyanov60347f62007-10-18 23:40:03 -0700172 proc_flush_task(p);
Ingo Molnar02030262009-05-17 11:24:08 +0200173
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 write_lock_irq(&tasklist_lock);
Roland McGrathdae33572008-07-25 19:45:48 -0700175 tracehook_finish_release_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 __exit_signal(p);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -0800177
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 /*
179 * If we are the last non-leader member of the thread
180 * group, and the leader is zombie, then notify the
181 * group leader's parent process. (if it wants notification.)
182 */
183 zap_leader = 0;
184 leader = p->group_leader;
185 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700186 BUG_ON(task_detached(leader));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 do_notify_parent(leader, leader->exit_signal);
188 /*
189 * If we were the last child thread and the leader has
190 * exited already, and the leader's parent ignores SIGCHLD,
191 * then we are the one who should release the leader.
192 *
193 * do_notify_parent() will have marked it self-reaping in
194 * that case.
195 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700196 zap_leader = task_detached(leader);
Roland McGrathdae33572008-07-25 19:45:48 -0700197
198 /*
199 * This maintains the invariant that release_task()
200 * only runs on a task in EXIT_DEAD, just for sanity.
201 */
202 if (zap_leader)
203 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 }
205
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 release_thread(p);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800208 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209
210 p = leader;
211 if (unlikely(zap_leader))
212 goto repeat;
213}
214
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215/*
216 * This checks not only the pgrp, but falls back on the pid if no
217 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
218 * without this...
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800219 *
220 * The caller must hold rcu lock or the tasklist lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 */
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800222struct pid *session_of_pgrp(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223{
224 struct task_struct *p;
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800225 struct pid *sid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800227 p = pid_task(pgrp, PIDTYPE_PGID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800228 if (p == NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800229 p = pid_task(pgrp, PIDTYPE_PID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800230 if (p != NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800231 sid = task_session(p);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800232
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 return sid;
234}
235
236/*
237 * Determine if a process group is "orphaned", according to the POSIX
238 * definition in 2.2.2.52. Orphaned process groups are not to be affected
239 * by terminal-generated stop signals. Newly orphaned process groups are
240 * to receive a SIGHUP and a SIGCONT.
241 *
242 * "I ask you, have you ever known what it is to be an orphan?"
243 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800244static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245{
246 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800248 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300249 if ((p == ignored_task) ||
250 (p->exit_state && thread_group_empty(p)) ||
251 is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 continue;
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300253
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800254 if (task_pgrp(p->real_parent) != pgrp &&
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300255 task_session(p->real_parent) == task_session(p))
256 return 0;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800257 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300258
259 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260}
261
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800262int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263{
264 int retval;
265
266 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800267 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 read_unlock(&tasklist_lock);
269
270 return retval;
271}
272
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800273static int has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274{
275 int retval = 0;
276 struct task_struct *p;
277
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800278 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Matthew Wilcox338077e2007-12-06 11:09:35 -0500279 if (!task_is_stopped(p))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 retval = 1;
282 break;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800283 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 return retval;
285}
286
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300287/*
288 * Check to see if any process groups have become orphaned as
289 * a result of our exiting, and if they have any stopped jobs,
290 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
291 */
292static void
293kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
294{
295 struct pid *pgrp = task_pgrp(tsk);
296 struct task_struct *ignored_task = tsk;
297
298 if (!parent)
299 /* exit: our father is in a different pgrp than
300 * we are and we were the only connection outside.
301 */
302 parent = tsk->real_parent;
303 else
304 /* reparent: our child is in a different pgrp than
305 * we are, and it was the only connection outside.
306 */
307 ignored_task = NULL;
308
309 if (task_pgrp(parent) != pgrp &&
310 task_session(parent) == task_session(tsk) &&
311 will_become_orphaned_pgrp(pgrp, ignored_task) &&
312 has_stopped_jobs(pgrp)) {
313 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
314 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
315 }
316}
317
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318/**
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700319 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 *
321 * If a kernel thread is launched as a result of a system call, or if
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700322 * it ever exits, it should generally reparent itself to kthreadd so it
323 * isn't in the way of other processes and is correctly cleaned up on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 *
325 * The various task state such as scheduling policy and priority may have
326 * been inherited from a user process, so we reset them to sane values here.
327 *
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700328 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700330static void reparent_to_kthreadd(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331{
332 write_lock_irq(&tasklist_lock);
333
334 ptrace_unlink(current);
335 /* Reparent to init */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700336 current->real_parent = current->parent = kthreadd_task;
Roland McGrathf4700212008-03-24 18:36:23 -0700337 list_move_tail(&current->sibling, &current->real_parent->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338
339 /* Set the exit signal to SIGCHLD so we signal init on exit */
340 current->exit_signal = SIGCHLD;
341
Ingo Molnare05606d2007-07-09 18:51:59 +0200342 if (task_nice(current) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 set_user_nice(current, 0);
344 /* cpus_allowed? */
345 /* rt_priority? */
346 /* signals? */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 memcpy(current->signal->rlim, init_task.signal->rlim,
348 sizeof(current->signal->rlim));
David Howellsd84f4f92008-11-14 10:39:23 +1100349
350 atomic_inc(&init_cred.usage);
351 commit_creds(&init_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353}
354
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800355void __set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356{
Oren Laadane19f2472006-01-08 01:03:58 -0800357 struct task_struct *curr = current->group_leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358
Christian Borntraeger0d0df592009-10-26 16:49:34 -0700359 if (task_session(curr) != pid)
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700360 change_pid(curr, PIDTYPE_SID, pid);
Oleg Nesterov1b0f7ff2009-04-02 16:58:39 -0700361
362 if (task_pgrp(curr) != pid)
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700363 change_pid(curr, PIDTYPE_PGID, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364}
365
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800366static void set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367{
368 write_lock_irq(&tasklist_lock);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800369 __set_special_pids(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 write_unlock_irq(&tasklist_lock);
371}
372
373/*
Oleg Nesterov87245132009-06-17 16:27:23 -0700374 * Let kernel threads use this to say that they allow a certain signal.
375 * Must not be used if kthread was cloned with CLONE_SIGHAND.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 */
377int allow_signal(int sig)
378{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700379 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 return -EINVAL;
381
382 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov87245132009-06-17 16:27:23 -0700383 /* This is only needed for daemonize()'ed kthreads */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 sigdelset(&current->blocked, sig);
Oleg Nesterov87245132009-06-17 16:27:23 -0700385 /*
386 * Kernel threads handle their own signals. Let the signal code
387 * know it'll be handled, so that they don't get converted to
388 * SIGKILL or just silently dropped.
389 */
390 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 recalc_sigpending();
392 spin_unlock_irq(&current->sighand->siglock);
393 return 0;
394}
395
396EXPORT_SYMBOL(allow_signal);
397
398int disallow_signal(int sig)
399{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700400 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 return -EINVAL;
402
403 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov10ab8252007-05-09 02:34:37 -0700404 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 recalc_sigpending();
406 spin_unlock_irq(&current->sighand->siglock);
407 return 0;
408}
409
410EXPORT_SYMBOL(disallow_signal);
411
412/*
413 * Put all the gunge required to become a kernel thread without
414 * attached user resources in one place where it belongs.
415 */
416
417void daemonize(const char *name, ...)
418{
419 va_list args;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 sigset_t blocked;
421
422 va_start(args, name);
423 vsnprintf(current->comm, sizeof(current->comm), name, args);
424 va_end(args);
425
426 /*
427 * If we were started as result of loading a module, close all of the
428 * user space pages. We don't need them, and if we didn't close them
429 * they would be locked into memory.
430 */
431 exit_mm(current);
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700432 /*
433 * We don't want to have TIF_FREEZE set if the system-wide hibernation
434 * or suspend transition begins right now.
435 */
Oleg Nesterov7b34e422008-07-25 01:47:37 -0700436 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800438 if (current->nsproxy != &init_nsproxy) {
439 get_nsproxy(&init_nsproxy);
440 switch_task_namespaces(current, &init_nsproxy);
441 }
Oleg Nesterov297bd422008-02-08 04:19:10 -0800442 set_special_pids(&init_struct_pid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -0800443 proc_clear_tty(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
445 /* Block and flush all signals */
446 sigfillset(&blocked);
447 sigprocmask(SIG_BLOCK, &blocked, NULL);
448 flush_signals(current);
449
450 /* Become as one with the init task */
451
Al Viro3e93cd62009-03-29 19:00:13 -0400452 daemonize_fs_struct();
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700453 exit_files(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 current->files = init_task.files;
455 atomic_inc(&current->files->count);
456
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700457 reparent_to_kthreadd();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458}
459
460EXPORT_SYMBOL(daemonize);
461
Arjan van de Ven858119e2006-01-14 13:20:43 -0800462static void close_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463{
464 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700465 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
467 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700468
469 /*
470 * It is safe to dereference the fd table without RCU or
471 * ->file_lock because this is the last reference to the
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800472 * files structure. But use RCU to shut RCU-lockdep up.
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700473 */
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800474 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700475 fdt = files_fdtable(files);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800476 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 for (;;) {
478 unsigned long set;
479 i = j * __NFDBITS;
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800480 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700482 set = fdt->open_fds->fds_bits[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 while (set) {
484 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700485 struct file * file = xchg(&fdt->fd[i], NULL);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800486 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 filp_close(file, files);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800488 cond_resched();
489 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 }
491 i++;
492 set >>= 1;
493 }
494 }
495}
496
497struct files_struct *get_files_struct(struct task_struct *task)
498{
499 struct files_struct *files;
500
501 task_lock(task);
502 files = task->files;
503 if (files)
504 atomic_inc(&files->count);
505 task_unlock(task);
506
507 return files;
508}
509
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800510void put_files_struct(struct files_struct *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700512 struct fdtable *fdt;
513
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 if (atomic_dec_and_test(&files->count)) {
515 close_files(files);
516 /*
517 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700518 * If the fdtable was embedded, pass files for freeing
519 * at the end of the RCU grace period. Otherwise,
520 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 */
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800522 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700523 fdt = files_fdtable(files);
Vadim Lobanov4fd45812006-12-10 02:21:17 -0800524 if (fdt != &files->fdtab)
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700525 kmem_cache_free(files_cachep, files);
Vadim Lobanov01b2d932006-12-22 01:10:43 -0800526 free_fdtable(fdt);
Paul E. McKenneyd11c5632010-02-22 17:04:50 -0800527 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 }
529}
530
Al Viro3b125382008-04-22 05:31:30 -0400531void reset_files_struct(struct files_struct *files)
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700532{
Al Viro3b125382008-04-22 05:31:30 -0400533 struct task_struct *tsk = current;
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700534 struct files_struct *old;
535
536 old = tsk->files;
537 task_lock(tsk);
538 tsk->files = files;
539 task_unlock(tsk);
540 put_files_struct(old);
541}
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700542
Al Viro1ec7f1d2008-04-22 05:35:42 -0400543void exit_files(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544{
545 struct files_struct * files = tsk->files;
546
547 if (files) {
548 task_lock(tsk);
549 tsk->files = NULL;
550 task_unlock(tsk);
551 put_files_struct(files);
552 }
553}
554
Balbir Singhcf475ad2008-04-29 01:00:16 -0700555#ifdef CONFIG_MM_OWNER
556/*
557 * Task p is exiting and it owned mm, lets find a new owner for it
558 */
559static inline int
560mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
561{
562 /*
563 * If there are other users of the mm and the owner (us) is exiting
564 * we need to find a new owner to take on the responsibility.
565 */
Balbir Singhcf475ad2008-04-29 01:00:16 -0700566 if (atomic_read(&mm->mm_users) <= 1)
567 return 0;
568 if (mm->owner != p)
569 return 0;
570 return 1;
571}
572
573void mm_update_next_owner(struct mm_struct *mm)
574{
575 struct task_struct *c, *g, *p = current;
576
577retry:
578 if (!mm_need_new_owner(mm, p))
579 return;
580
581 read_lock(&tasklist_lock);
582 /*
583 * Search in the children
584 */
585 list_for_each_entry(c, &p->children, sibling) {
586 if (c->mm == mm)
587 goto assign_new_owner;
588 }
589
590 /*
591 * Search in the siblings
592 */
Oleg Nesterovdea33cf2009-06-17 16:27:29 -0700593 list_for_each_entry(c, &p->real_parent->children, sibling) {
Balbir Singhcf475ad2008-04-29 01:00:16 -0700594 if (c->mm == mm)
595 goto assign_new_owner;
596 }
597
598 /*
599 * Search through everything else. We should not get
600 * here often
601 */
602 do_each_thread(g, c) {
603 if (c->mm == mm)
604 goto assign_new_owner;
605 } while_each_thread(g, c);
606
607 read_unlock(&tasklist_lock);
Balbir Singh31a78f22008-09-28 23:09:31 +0100608 /*
609 * We found no owner yet mm_users > 1: this implies that we are
610 * most likely racing with swapoff (try_to_unuse()) or /proc or
Hugh Dickinse5991372009-01-06 14:39:22 -0800611 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
Balbir Singh31a78f22008-09-28 23:09:31 +0100612 */
Balbir Singh31a78f22008-09-28 23:09:31 +0100613 mm->owner = NULL;
Balbir Singhcf475ad2008-04-29 01:00:16 -0700614 return;
615
616assign_new_owner:
617 BUG_ON(c == p);
618 get_task_struct(c);
619 /*
620 * The task_lock protects c->mm from changing.
621 * We always want mm->owner->mm == mm
622 */
623 task_lock(c);
Hugh Dickinse5991372009-01-06 14:39:22 -0800624 /*
625 * Delay read_unlock() till we have the task_lock()
626 * to ensure that c does not slip away underneath us
627 */
628 read_unlock(&tasklist_lock);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700629 if (c->mm != mm) {
630 task_unlock(c);
631 put_task_struct(c);
632 goto retry;
633 }
Balbir Singhcf475ad2008-04-29 01:00:16 -0700634 mm->owner = c;
635 task_unlock(c);
636 put_task_struct(c);
637}
638#endif /* CONFIG_MM_OWNER */
639
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640/*
641 * Turn us into a lazy TLB process if we
642 * aren't already..
643 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700644static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645{
646 struct mm_struct *mm = tsk->mm;
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700647 struct core_state *core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
649 mm_release(tsk, mm);
650 if (!mm)
651 return;
652 /*
653 * Serialize with any possible pending coredump.
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700654 * We must hold mmap_sem around checking core_state
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 * and clearing tsk->mm. The core-inducing thread
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700656 * will increment ->nr_threads for each thread in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 * group with ->mm != NULL.
658 */
659 down_read(&mm->mmap_sem);
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700660 core_state = mm->core_state;
661 if (core_state) {
662 struct core_thread self;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 up_read(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700665 self.task = tsk;
666 self.next = xchg(&core_state->dumper.next, &self);
667 /*
668 * Implies mb(), the result of xchg() must be visible
669 * to core_state->dumper.
670 */
671 if (atomic_dec_and_test(&core_state->nr_threads))
672 complete(&core_state->startup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673
Oleg Nesterova94e2d42008-07-25 01:47:46 -0700674 for (;;) {
675 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
676 if (!self.task) /* see coredump_finish() */
677 break;
678 schedule();
679 }
680 __set_task_state(tsk, TASK_RUNNING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 down_read(&mm->mmap_sem);
682 }
683 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700684 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 /* more a memory barrier than a real lock */
686 task_lock(tsk);
687 tsk->mm = NULL;
688 up_read(&mm->mmap_sem);
689 enter_lazy_tlb(mm, current);
Rafael J. Wysocki0c1eecf2007-07-19 01:47:33 -0700690 /* We don't want this task to be frozen prematurely */
691 clear_freeze_flag(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 task_unlock(tsk);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700693 mm_update_next_owner(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 mmput(mm);
695}
696
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697/*
698 * When we die, we re-parent all our children.
699 * Try to give them to another thread in our thread
700 * group, and if no such member exists, give it to
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800701 * the child reaper process (ie "init") in our pid
702 * space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 */
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700704static struct task_struct *find_new_reaper(struct task_struct *father)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700706 struct pid_namespace *pid_ns = task_active_pid_ns(father);
707 struct task_struct *thread;
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700708
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700709 thread = father;
710 while_each_thread(father, thread) {
711 if (thread->flags & PF_EXITING)
712 continue;
713 if (unlikely(pid_ns->child_reaper == father))
714 pid_ns->child_reaper = thread;
715 return thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700717
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700718 if (unlikely(pid_ns->child_reaper == father)) {
719 write_unlock_irq(&tasklist_lock);
720 if (unlikely(pid_ns == &init_pid_ns))
721 panic("Attempted to kill init!");
722
723 zap_pid_ns_processes(pid_ns);
724 write_lock_irq(&tasklist_lock);
725 /*
726 * We can not clear ->child_reaper or leave it alone.
727 * There may by stealth EXIT_DEAD tasks on ->children,
728 * forget_original_parent() must move them somewhere.
729 */
730 pid_ns->child_reaper = init_pid_ns.child_reaper;
731 }
732
733 return pid_ns->child_reaper;
734}
735
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700736/*
737* Any that need to be release_task'd are put on the @dead list.
738 */
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800739static void reparent_leader(struct task_struct *father, struct task_struct *p,
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700740 struct list_head *dead)
741{
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700742 list_move_tail(&p->sibling, &p->real_parent->children);
743
744 if (task_detached(p))
745 return;
746 /*
747 * If this is a threaded reparent there is no need to
748 * notify anyone anything has happened.
749 */
750 if (same_thread_group(p->real_parent, father))
751 return;
752
753 /* We don't want people slaying init. */
754 p->exit_signal = SIGCHLD;
755
756 /* If it has exited notify the new parent about this child's death. */
Oleg Nesterov5cb11442009-06-17 16:27:30 -0700757 if (!task_ptrace(p) &&
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700758 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
759 do_notify_parent(p, p->exit_signal);
760 if (task_detached(p)) {
761 p->exit_state = EXIT_DEAD;
762 list_move_tail(&p->sibling, dead);
763 }
764 }
765
766 kill_orphaned_pgrp(p, father);
767}
768
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769static void forget_original_parent(struct task_struct *father)
770{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700771 struct task_struct *p, *n, *reaper;
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700772 LIST_HEAD(dead_children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Oleg Nesterov39c626a2009-04-02 16:58:18 -0700774 exit_ptrace(father);
775
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 write_lock_irq(&tasklist_lock);
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700777 reaper = find_new_reaper(father);
Roland McGrathf4700212008-03-24 18:36:23 -0700778
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 list_for_each_entry_safe(p, n, &father->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800780 struct task_struct *t = p;
781 do {
782 t->real_parent = reaper;
783 if (t->parent == father) {
784 BUG_ON(task_ptrace(t));
785 t->parent = t->real_parent;
786 }
787 if (t->pdeath_signal)
788 group_send_sig_info(t->pdeath_signal,
789 SEND_SIG_NOINFO, t);
790 } while_each_thread(p, t);
791 reparent_leader(father, p, &dead_children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700793 write_unlock_irq(&tasklist_lock);
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700794
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700795 BUG_ON(!list_empty(&father->children));
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700796
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700797 list_for_each_entry_safe(p, n, &dead_children, sibling) {
798 list_del_init(&p->sibling);
Oleg Nesterov39c626a2009-04-02 16:58:18 -0700799 release_task(p);
800 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801}
802
803/*
804 * Send signals to all our closest relatives so that they know
805 * to properly mourn us..
806 */
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300807static void exit_notify(struct task_struct *tsk, int group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700809 int signal;
810 void *cookie;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 /*
813 * This does two things:
814 *
815 * A. Make init inherit all the child processes
816 * B. Check to see if any process groups have become orphaned
817 * as a result of our exiting, and if they have any stopped
818 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
819 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700820 forget_original_parent(tsk);
Pavel Emelyanov2e4a7072007-10-18 23:40:01 -0700821 exit_task_namespaces(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700823 write_lock_irq(&tasklist_lock);
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300824 if (group_dead)
825 kill_orphaned_pgrp(tsk->group_leader, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Oleg Nesterov24728442007-08-04 01:04:41 +0400827 /* Let father know we died
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 *
829 * Thread signals are configurable, but you aren't going to use
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700830 * that to send signals to arbitary processes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 * That stops right now.
832 *
833 * If the parent exec id doesn't match the exec id we saved
834 * when we started then we know the parent has changed security
835 * domain.
836 *
837 * If our self_exec id doesn't match our parent_exec_id then
838 * we have changed execution domain as these two values started
839 * the same after a fork.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700841 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300842 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
Oleg Nesterov432870d2009-04-06 16:16:02 +0200843 tsk->self_exec_id != tsk->parent_exec_id))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 tsk->exit_signal = SIGCHLD;
845
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700846 signal = tracehook_notify_death(tsk, &cookie, group_dead);
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700847 if (signal >= 0)
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700848 signal = do_notify_parent(tsk, signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700850 tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
Oleg Nesterov9c339162010-05-26 14:43:10 -0700852 /* mt-exec, de_thread() is waiting for group leader */
853 if (unlikely(tsk->signal->notify_count < 0))
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700854 wake_up_process(tsk->signal->group_exit_task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 write_unlock_irq(&tasklist_lock);
856
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700857 tracehook_report_death(tsk, signal, cookie, group_dead);
858
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 /* If the process is dead, release it - nobody will wait for it */
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700860 if (signal == DEATH_REAP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 release_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862}
863
Jeff Dikee18eecb2007-07-15 23:38:48 -0700864#ifdef CONFIG_DEBUG_STACK_USAGE
865static void check_stack_usage(void)
866{
867 static DEFINE_SPINLOCK(low_water_lock);
868 static int lowest_to_date = THREAD_SIZE;
Jeff Dikee18eecb2007-07-15 23:38:48 -0700869 unsigned long free;
870
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500871 free = stack_not_used(current);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700872
873 if (free >= lowest_to_date)
874 return;
875
876 spin_lock(&low_water_lock);
877 if (free < lowest_to_date) {
878 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
879 "left\n",
880 current->comm, free);
881 lowest_to_date = free;
882 }
883 spin_unlock(&low_water_lock);
884}
885#else
886static inline void check_stack_usage(void) {}
887#endif
888
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800889NORET_TYPE void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890{
891 struct task_struct *tsk = current;
892 int group_dead;
893
894 profile_task_exit(tsk);
895
Jens Axboe22e2c502005-06-27 10:55:12 +0200896 WARN_ON(atomic_read(&tsk->fs_excl));
897
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 if (unlikely(in_interrupt()))
899 panic("Aiee, killing interrupt handler!");
900 if (unlikely(!tsk->pid))
901 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
Roland McGrath30199f52008-07-25 19:45:46 -0700903 tracehook_report_exit(&code);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
David Howellse0e81732009-09-02 09:13:40 +0100905 validate_creds_for_do_exit(tsk);
906
Alexander Nybergdf164db2005-06-23 00:09:13 -0700907 /*
908 * We're taking recursive faults here in do_exit. Safest is to just
909 * leave this task alone and wait for reboot.
910 */
911 if (unlikely(tsk->flags & PF_EXITING)) {
912 printk(KERN_ALERT
913 "Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700914 /*
915 * We can do this unlocked here. The futex code uses
916 * this flag just to verify whether the pi state
917 * cleanup has been done or not. In the worst case it
918 * loops once more. We pretend that the cleanup was
919 * done as there is no way to return. Either the
920 * OWNER_DIED bit is set by now or we push the blocked
921 * task into the wait for ever nirwana as well.
922 */
923 tsk->flags |= PF_EXITPIDONE;
Alexander Nybergdf164db2005-06-23 00:09:13 -0700924 set_current_state(TASK_UNINTERRUPTIBLE);
925 schedule();
926 }
927
Thomas Gleixner3aa551c2009-03-23 18:28:15 +0100928 exit_irq_thread();
929
Oleg Nesterovd12619b2008-02-08 04:19:12 -0800930 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700931 /*
932 * tsk->flags are checked in the futex code to protect against
933 * an exiting task cleaning up the robust pi futexes.
934 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -0700935 smp_mb();
Thomas Gleixner1d615482009-11-17 14:54:03 +0100936 raw_spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 if (unlikely(in_atomic()))
939 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -0700940 current->comm, task_pid_nr(current),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 preempt_count());
942
943 acct_update_integrals(tsk);
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -0800944 /* sync mm's RSS info before statistics gathering */
KAMEZAWA Hiroyukia3a2e762010-04-06 14:34:42 -0700945 if (tsk->mm)
946 sync_mm_rss(tsk, tsk->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -0700948 if (group_dead) {
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700949 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -0700950 exit_itimers(tsk->signal);
Jiri Pirko1f102062009-09-22 16:44:10 -0700951 if (tsk->mm)
952 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
Andrew Mortonc3068952005-08-04 16:49:32 -0700953 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700954 acct_collect(code, group_dead);
Miloslav Trmac522ed772007-07-15 23:40:56 -0700955 if (group_dead)
956 tty_audit_exit();
Al Virofa84cb92006-03-29 20:30:19 -0500957 if (unlikely(tsk->audit_context))
958 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -0800959
Jonathan Limf2ab6d82007-08-30 23:56:23 -0700960 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -0800961 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -0700962
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 exit_mm(tsk);
964
KaiGai Kohei0e464812006-06-25 05:49:24 -0700965 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700966 acct_process();
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400967 trace_sched_process_exit(tsk);
968
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 exit_sem(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400970 exit_files(tsk);
971 exit_fs(tsk);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700972 check_stack_usage();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 exit_thread();
Paul Menageb4f48b62007-10-18 23:39:33 -0700974 cgroup_exit(tsk, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975
Alan Cox5ec93d12009-11-30 13:18:45 +0000976 if (group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 disassociate_ctty(1);
978
Al Viroa1261f52005-11-13 16:06:55 -0800979 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
Matt Helsley9f460802005-11-07 00:59:16 -0800981 proc_exit_connector(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200982
983 /*
Frederic Weisbecker24f1e32c2009-09-09 19:22:48 +0200984 * FIXME: do that only when needed, using sched_exit tracepoint
985 */
986 flush_ptrace_hw_breakpoint(tsk);
987 /*
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200988 * Flush inherited counters to the parent - before the parent
989 * gets woken up by child-exit notifications.
990 */
Ingo Molnarcdd6c482009-09-21 12:02:48 +0200991 perf_event_exit_task(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200992
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300993 exit_notify(tsk, group_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994#ifdef CONFIG_NUMA
Miao Xiec0ff7452010-05-24 14:32:08 -0700995 task_lock(tsk);
Lee Schermerhornf0be3d32008-04-28 02:13:08 -0700996 mpol_put(tsk->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 tsk->mempolicy = NULL;
Miao Xiec0ff7452010-05-24 14:32:08 -0700998 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001000#ifdef CONFIG_FUTEX
Ingo Molnarc87e2832006-06-27 02:54:58 -07001001 if (unlikely(current->pi_state_cache))
1002 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001003#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001004 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001005 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -08001006 */
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001007 debug_check_no_locks_held(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001008 /*
1009 * We can do this unlocked here. The futex code uses this flag
1010 * just to verify whether the pi state cleanup has been done
1011 * or not. In the worst case it loops once more.
1012 */
1013 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
Al Viroafc847b2006-02-28 12:51:55 -05001015 if (tsk->io_context)
Louis Rillingb69f2292009-12-04 14:52:42 +01001016 exit_io_context(tsk);
Al Viroafc847b2006-02-28 12:51:55 -05001017
Jens Axboeb92ce552006-04-11 13:52:07 +02001018 if (tsk->splice_pipe)
1019 __free_pipe_info(tsk->splice_pipe);
1020
David Howellse0e81732009-09-02 09:13:40 +01001021 validate_creds_for_do_exit(tsk);
1022
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001023 preempt_disable();
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001024 exit_rcu();
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001025 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -07001026 tsk->state = TASK_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 schedule();
1028 BUG();
1029 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -07001030 for (;;)
1031 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032}
1033
Russ Anderson012914d2005-04-23 00:08:00 -07001034EXPORT_SYMBOL_GPL(do_exit);
1035
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1037{
1038 if (comp)
1039 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001040
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 do_exit(code);
1042}
1043
1044EXPORT_SYMBOL(complete_and_exit);
1045
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001046SYSCALL_DEFINE1(exit, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047{
1048 do_exit((error_code&0xff)<<8);
1049}
1050
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051/*
1052 * Take down every thread in the group. This is called by fatal signals
1053 * as well as by sys_exit_group (below).
1054 */
1055NORET_TYPE void
1056do_group_exit(int exit_code)
1057{
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001058 struct signal_struct *sig = current->signal;
1059
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1061
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001062 if (signal_group_exit(sig))
1063 exit_code = sig->group_exit_code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 else if (!thread_group_empty(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 struct sighand_struct *const sighand = current->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001067 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 /* Another thread got here before we took the lock. */
1069 exit_code = sig->group_exit_code;
1070 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001072 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 zap_other_threads(current);
1074 }
1075 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 }
1077
1078 do_exit(exit_code);
1079 /* NOTREACHED */
1080}
1081
1082/*
1083 * this kills every thread in the thread group. Note that any externally
1084 * wait4()-ing process will get the correct exit code - even if this
1085 * thread is not the thread group leader.
1086 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001087SYSCALL_DEFINE1(exit_group, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088{
1089 do_group_exit((error_code & 0xff) << 8);
Heiko Carstens2ed7c032009-01-14 14:13:54 +01001090 /* NOTREACHED */
1091 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092}
1093
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001094struct wait_opts {
1095 enum pid_type wo_type;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001096 int wo_flags;
Richard Kennedye1eb1eb2009-06-17 16:27:42 -07001097 struct pid *wo_pid;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001098
1099 struct siginfo __user *wo_info;
1100 int __user *wo_stat;
1101 struct rusage __user *wo_rusage;
1102
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001103 wait_queue_t child_wait;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001104 int notask_error;
1105};
1106
Oleg Nesterov989264f2009-09-23 15:56:49 -07001107static inline
1108struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
Eric W. Biederman161550d2008-02-08 04:19:14 -08001109{
Oleg Nesterov989264f2009-09-23 15:56:49 -07001110 if (type != PIDTYPE_PID)
1111 task = task->group_leader;
1112 return task->pids[type].pid;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001113}
1114
Oleg Nesterov989264f2009-09-23 15:56:49 -07001115static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001116{
1117 return wo->wo_type == PIDTYPE_MAX ||
1118 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1119}
1120
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001121static int eligible_child(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122{
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001123 if (!eligible_pid(wo, p))
1124 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 /* Wait for all children (clone and not) if __WALL is set;
1126 * otherwise, wait for clone children *only* if __WCLONE is
1127 * set; otherwise, wait for non-clone children *only*. (Note:
1128 * A "clone" child here is one that reports to its parent
1129 * using a signal other than SIGCHLD.) */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001130 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1131 && !(wo->wo_flags & __WALL))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
Roland McGrath14dd0b82008-03-30 18:41:25 -07001134 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135}
1136
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001137static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1138 pid_t pid, uid_t uid, int why, int status)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001140 struct siginfo __user *infop;
1141 int retval = wo->wo_rusage
1142 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -07001143
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001145 infop = wo->wo_info;
Vitaly Mayatskikhb6fe2d12009-09-23 15:56:52 -07001146 if (infop) {
1147 if (!retval)
1148 retval = put_user(SIGCHLD, &infop->si_signo);
1149 if (!retval)
1150 retval = put_user(0, &infop->si_errno);
1151 if (!retval)
1152 retval = put_user((short)why, &infop->si_code);
1153 if (!retval)
1154 retval = put_user(pid, &infop->si_pid);
1155 if (!retval)
1156 retval = put_user(uid, &infop->si_uid);
1157 if (!retval)
1158 retval = put_user(status, &infop->si_status);
1159 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 if (!retval)
1161 retval = pid;
1162 return retval;
1163}
1164
1165/*
1166 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1167 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1168 * the lock and this task is uninteresting. If we return nonzero, we have
1169 * released the lock and the system call should return.
1170 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001171static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172{
1173 unsigned long state;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001174 int retval, status, traced;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001175 pid_t pid = task_pid_vnr(p);
David Howellsc69e8d92008-11-14 10:39:19 +11001176 uid_t uid = __task_cred(p)->uid;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001177 struct siginfo __user *infop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001179 if (!likely(wo->wo_flags & WEXITED))
Roland McGrath98abed02008-03-19 19:24:59 -07001180 return 0;
1181
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001182 if (unlikely(wo->wo_flags & WNOWAIT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 int exit_code = p->exit_code;
Thiago Farinaf3abd4f2010-03-05 13:42:52 -08001184 int why;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 get_task_struct(p);
1187 read_unlock(&tasklist_lock);
1188 if ((exit_code & 0x7f) == 0) {
1189 why = CLD_EXITED;
1190 status = exit_code >> 8;
1191 } else {
1192 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1193 status = exit_code & 0x7f;
1194 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001195 return wait_noreap_copyout(wo, p, pid, uid, why, status);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 }
1197
1198 /*
1199 * Try to move the task's state to DEAD
1200 * only one thread is allowed to do this:
1201 */
1202 state = xchg(&p->exit_state, EXIT_DEAD);
1203 if (state != EXIT_ZOMBIE) {
1204 BUG_ON(state != EXIT_DEAD);
1205 return 0;
1206 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207
Oleg Nesterov53b6f9f2008-04-30 00:53:13 -07001208 traced = ptrace_reparented(p);
Oleg Nesterovbefca962009-06-18 16:49:11 -07001209 /*
1210 * It can be ptraced but not reparented, check
1211 * !task_detached() to filter out sub-threads.
1212 */
1213 if (likely(!traced) && likely(!task_detached(p))) {
Jesper Juhl3795e162006-01-09 20:54:39 -08001214 struct signal_struct *psig;
1215 struct signal_struct *sig;
Jiri Pirko1f102062009-09-22 16:44:10 -07001216 unsigned long maxrss;
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001217 cputime_t tgutime, tgstime;
Jesper Juhl3795e162006-01-09 20:54:39 -08001218
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 /*
1220 * The resource counters for the group leader are in its
1221 * own task_struct. Those for dead threads in the group
1222 * are in its signal_struct, as are those for the child
1223 * processes it has previously reaped. All these
1224 * accumulate in the parent's signal_struct c* fields.
1225 *
1226 * We don't bother to take a lock here to protect these
1227 * p->signal fields, because they are only touched by
1228 * __exit_signal, which runs with tasklist_lock
1229 * write-locked anyway, and so is excluded here. We do
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001230 * need to protect the access to parent->signal fields,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 * as other threads in the parent group can be right
1232 * here reaping other children at the same time.
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001233 *
1234 * We use thread_group_times() to get times for the thread
1235 * group, which consolidates times for all threads in the
1236 * group including the group leader.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 */
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001238 thread_group_times(p, &tgutime, &tgstime);
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001239 spin_lock_irq(&p->real_parent->sighand->siglock);
1240 psig = p->real_parent->signal;
Jesper Juhl3795e162006-01-09 20:54:39 -08001241 sig = p->signal;
1242 psig->cutime =
1243 cputime_add(psig->cutime,
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001244 cputime_add(tgutime,
1245 sig->cutime));
Jesper Juhl3795e162006-01-09 20:54:39 -08001246 psig->cstime =
1247 cputime_add(psig->cstime,
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001248 cputime_add(tgstime,
1249 sig->cstime));
Laurent Vivier9ac52312007-10-15 17:00:19 +02001250 psig->cgtime =
1251 cputime_add(psig->cgtime,
1252 cputime_add(p->gtime,
1253 cputime_add(sig->gtime,
1254 sig->cgtime)));
Jesper Juhl3795e162006-01-09 20:54:39 -08001255 psig->cmin_flt +=
1256 p->min_flt + sig->min_flt + sig->cmin_flt;
1257 psig->cmaj_flt +=
1258 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1259 psig->cnvcsw +=
1260 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1261 psig->cnivcsw +=
1262 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001263 psig->cinblock +=
1264 task_io_get_inblock(p) +
1265 sig->inblock + sig->cinblock;
1266 psig->coublock +=
1267 task_io_get_oublock(p) +
1268 sig->oublock + sig->coublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001269 maxrss = max(sig->maxrss, sig->cmaxrss);
1270 if (psig->cmaxrss < maxrss)
1271 psig->cmaxrss = maxrss;
Andrea Righi59954772008-07-27 17:29:15 +02001272 task_io_accounting_add(&psig->ioac, &p->ioac);
1273 task_io_accounting_add(&psig->ioac, &sig->ioac);
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001274 spin_unlock_irq(&p->real_parent->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 }
1276
1277 /*
1278 * Now we are sure this task is interesting, and no other
1279 * thread can reap it because we set its state to EXIT_DEAD.
1280 */
1281 read_unlock(&tasklist_lock);
1282
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001283 retval = wo->wo_rusage
1284 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1286 ? p->signal->group_exit_code : p->exit_code;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001287 if (!retval && wo->wo_stat)
1288 retval = put_user(status, wo->wo_stat);
1289
1290 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 if (!retval && infop)
1292 retval = put_user(SIGCHLD, &infop->si_signo);
1293 if (!retval && infop)
1294 retval = put_user(0, &infop->si_errno);
1295 if (!retval && infop) {
1296 int why;
1297
1298 if ((status & 0x7f) == 0) {
1299 why = CLD_EXITED;
1300 status >>= 8;
1301 } else {
1302 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1303 status &= 0x7f;
1304 }
1305 retval = put_user((short)why, &infop->si_code);
1306 if (!retval)
1307 retval = put_user(status, &infop->si_status);
1308 }
1309 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001310 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 if (!retval && infop)
David Howellsc69e8d92008-11-14 10:39:19 +11001312 retval = put_user(uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001313 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001314 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001315
1316 if (traced) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001318 /* We dropped tasklist, ptracer could die and untrace */
1319 ptrace_unlink(p);
1320 /*
1321 * If this is not a detached task, notify the parent.
1322 * If it's still not detached after that, don't release
1323 * it now.
1324 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001325 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001326 do_notify_parent(p, p->exit_signal);
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001327 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001328 p->exit_state = EXIT_ZOMBIE;
1329 p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 }
1331 }
1332 write_unlock_irq(&tasklist_lock);
1333 }
1334 if (p != NULL)
1335 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001336
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 return retval;
1338}
1339
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001340static int *task_stopped_code(struct task_struct *p, bool ptrace)
1341{
1342 if (ptrace) {
1343 if (task_is_stopped_or_traced(p))
1344 return &p->exit_code;
1345 } else {
1346 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1347 return &p->signal->group_exit_code;
1348 }
1349 return NULL;
1350}
1351
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352/*
1353 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1354 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1355 * the lock and this task is uninteresting. If we return nonzero, we have
1356 * released the lock and the system call should return.
1357 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001358static int wait_task_stopped(struct wait_opts *wo,
1359 int ptrace, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001361 struct siginfo __user *infop;
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001362 int retval, exit_code, *p_code, why;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001363 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001364 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
Oleg Nesterov47918022009-06-17 16:27:39 -07001366 /*
1367 * Traditionally we see ptrace'd stopped tasks regardless of options.
1368 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001369 if (!ptrace && !(wo->wo_flags & WUNTRACED))
Roland McGrath98abed02008-03-19 19:24:59 -07001370 return 0;
1371
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001372 exit_code = 0;
1373 spin_lock_irq(&p->sighand->siglock);
1374
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001375 p_code = task_stopped_code(p, ptrace);
1376 if (unlikely(!p_code))
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001377 goto unlock_sig;
1378
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001379 exit_code = *p_code;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001380 if (!exit_code)
1381 goto unlock_sig;
1382
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001383 if (!unlikely(wo->wo_flags & WNOWAIT))
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001384 *p_code = 0;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001385
David Howellsc69e8d92008-11-14 10:39:19 +11001386 /* don't need the RCU readlock here as we're holding a spinlock */
1387 uid = __task_cred(p)->uid;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001388unlock_sig:
1389 spin_unlock_irq(&p->sighand->siglock);
1390 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 return 0;
1392
1393 /*
1394 * Now we are pretty sure this task is interesting.
1395 * Make sure it doesn't get reaped out from under us while we
1396 * give up the lock and then examine it below. We don't want to
1397 * keep holding onto the tasklist_lock while we call getrusage and
1398 * possibly take page faults for user memory.
1399 */
1400 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001401 pid = task_pid_vnr(p);
Roland McGrathf4700212008-03-24 18:36:23 -07001402 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 read_unlock(&tasklist_lock);
1404
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001405 if (unlikely(wo->wo_flags & WNOWAIT))
1406 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001408 retval = wo->wo_rusage
1409 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1410 if (!retval && wo->wo_stat)
1411 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1412
1413 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 if (!retval && infop)
1415 retval = put_user(SIGCHLD, &infop->si_signo);
1416 if (!retval && infop)
1417 retval = put_user(0, &infop->si_errno);
1418 if (!retval && infop)
Roland McGrath6efcae42008-03-08 11:41:22 -08001419 retval = put_user((short)why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 if (!retval && infop)
1421 retval = put_user(exit_code, &infop->si_status);
1422 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001423 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001425 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001427 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 put_task_struct(p);
1429
1430 BUG_ON(!retval);
1431 return retval;
1432}
1433
1434/*
1435 * Handle do_wait work for one task in a live, non-stopped state.
1436 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1437 * the lock and this task is uninteresting. If we return nonzero, we have
1438 * released the lock and the system call should return.
1439 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001440static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441{
1442 int retval;
1443 pid_t pid;
1444 uid_t uid;
1445
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001446 if (!unlikely(wo->wo_flags & WCONTINUED))
Roland McGrath98abed02008-03-19 19:24:59 -07001447 return 0;
1448
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1450 return 0;
1451
1452 spin_lock_irq(&p->sighand->siglock);
1453 /* Re-check with the lock held. */
1454 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1455 spin_unlock_irq(&p->sighand->siglock);
1456 return 0;
1457 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001458 if (!unlikely(wo->wo_flags & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
David Howellsc69e8d92008-11-14 10:39:19 +11001460 uid = __task_cred(p)->uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 spin_unlock_irq(&p->sighand->siglock);
1462
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001463 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 get_task_struct(p);
1465 read_unlock(&tasklist_lock);
1466
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001467 if (!wo->wo_info) {
1468 retval = wo->wo_rusage
1469 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001471 if (!retval && wo->wo_stat)
1472 retval = put_user(0xffff, wo->wo_stat);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001474 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 } else {
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001476 retval = wait_noreap_copyout(wo, p, pid, uid,
1477 CLD_CONTINUED, SIGCONT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 BUG_ON(retval == 0);
1479 }
1480
1481 return retval;
1482}
1483
Roland McGrath98abed02008-03-19 19:24:59 -07001484/*
1485 * Consider @p for a wait by @parent.
1486 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001487 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001488 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1489 * Returns zero if the search for a child should continue;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001490 * then ->notask_error is 0 if @p is an eligible child,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001491 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001492 */
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001493static int wait_consider_task(struct wait_opts *wo, int ptrace,
1494 struct task_struct *p)
Roland McGrath98abed02008-03-19 19:24:59 -07001495{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001496 int ret = eligible_child(wo, p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001497 if (!ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001498 return ret;
1499
Oleg Nesterova2322e12009-09-23 15:56:45 -07001500 ret = security_task_wait(p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001501 if (unlikely(ret < 0)) {
1502 /*
1503 * If we have not yet seen any eligible child,
1504 * then let this error code replace -ECHILD.
1505 * A permission error will give the user a clue
1506 * to look for security policy problems, rather
1507 * than for mysterious wait bugs.
1508 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001509 if (wo->notask_error)
1510 wo->notask_error = ret;
Oleg Nesterov78a3d9d2009-04-29 18:01:23 +02001511 return 0;
Roland McGrath14dd0b82008-03-30 18:41:25 -07001512 }
1513
Oleg Nesterov5cb11442009-06-17 16:27:30 -07001514 if (likely(!ptrace) && unlikely(task_ptrace(p))) {
Roland McGrathf4700212008-03-24 18:36:23 -07001515 /*
1516 * This child is hidden by ptrace.
1517 * We aren't allowed to see it now, but eventually we will.
1518 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001519 wo->notask_error = 0;
Roland McGrathf4700212008-03-24 18:36:23 -07001520 return 0;
1521 }
1522
Roland McGrath98abed02008-03-19 19:24:59 -07001523 if (p->exit_state == EXIT_DEAD)
1524 return 0;
1525
1526 /*
1527 * We don't reap group leaders with subthreads.
1528 */
1529 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001530 return wait_task_zombie(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001531
1532 /*
1533 * It's stopped or running now, so it might
1534 * later continue, exit, or stop again.
1535 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001536 wo->notask_error = 0;
Roland McGrath98abed02008-03-19 19:24:59 -07001537
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001538 if (task_stopped_code(p, ptrace))
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001539 return wait_task_stopped(wo, ptrace, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001540
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001541 return wait_task_continued(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001542}
1543
1544/*
1545 * Do the work of do_wait() for one thread in the group, @tsk.
1546 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001547 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001548 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1549 * Returns zero if the search for a child should continue; then
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001550 * ->notask_error is 0 if there were any eligible children,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001551 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001552 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001553static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001554{
1555 struct task_struct *p;
1556
1557 list_for_each_entry(p, &tsk->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001558 int ret = wait_consider_task(wo, 0, p);
1559 if (ret)
1560 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001561 }
1562
1563 return 0;
1564}
1565
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001566static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001567{
1568 struct task_struct *p;
1569
Roland McGrathf4700212008-03-24 18:36:23 -07001570 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001571 int ret = wait_consider_task(wo, 1, p);
Roland McGrathf4700212008-03-24 18:36:23 -07001572 if (ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001573 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001574 }
1575
1576 return 0;
1577}
1578
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001579static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1580 int sync, void *key)
1581{
1582 struct wait_opts *wo = container_of(wait, struct wait_opts,
1583 child_wait);
1584 struct task_struct *p = key;
1585
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001586 if (!eligible_pid(wo, p))
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001587 return 0;
1588
Oleg Nesterovb4fe5182009-09-23 15:56:47 -07001589 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1590 return 0;
1591
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001592 return default_wake_function(wait, mode, sync, key);
1593}
1594
Oleg Nesterova7f07652009-09-23 15:56:44 -07001595void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1596{
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001597 __wake_up_sync_key(&parent->signal->wait_chldexit,
1598 TASK_INTERRUPTIBLE, 1, p);
Oleg Nesterova7f07652009-09-23 15:56:44 -07001599}
1600
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001601static long do_wait(struct wait_opts *wo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 struct task_struct *tsk;
Roland McGrath98abed02008-03-19 19:24:59 -07001604 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001606 trace_sched_process_wait(wo->wo_pid);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001607
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001608 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1609 wo->child_wait.private = current;
1610 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611repeat:
Roland McGrath98abed02008-03-19 19:24:59 -07001612 /*
1613 * If there is nothing that can match our critiera just get out.
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001614 * We will clear ->notask_error to zero if we see any child that
1615 * might later match our criteria, even if we are not able to reap
1616 * it yet.
Roland McGrath98abed02008-03-19 19:24:59 -07001617 */
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001618 wo->notask_error = -ECHILD;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001619 if ((wo->wo_type < PIDTYPE_MAX) &&
1620 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001621 goto notask;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001622
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001623 set_current_state(TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 read_lock(&tasklist_lock);
1625 tsk = current;
1626 do {
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001627 retval = do_wait_thread(wo, tsk);
1628 if (retval)
Roland McGrath98abed02008-03-19 19:24:59 -07001629 goto end;
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001630
1631 retval = ptrace_do_wait(wo, tsk);
1632 if (retval)
1633 goto end;
Roland McGrath98abed02008-03-19 19:24:59 -07001634
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001635 if (wo->wo_flags & __WNOTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 break;
Oleg Nesterova3f6dfb2009-06-17 16:27:41 -07001637 } while_each_thread(current, tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001639
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001640notask:
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001641 retval = wo->notask_error;
1642 if (!retval && !(wo->wo_flags & WNOHANG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 retval = -ERESTARTSYS;
Roland McGrath98abed02008-03-19 19:24:59 -07001644 if (!signal_pending(current)) {
1645 schedule();
1646 goto repeat;
1647 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649end:
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001650 __set_current_state(TASK_RUNNING);
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001651 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 return retval;
1653}
1654
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001655SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1656 infop, int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001658 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001659 struct pid *pid = NULL;
1660 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 long ret;
1662
1663 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1664 return -EINVAL;
1665 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1666 return -EINVAL;
1667
1668 switch (which) {
1669 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001670 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 break;
1672 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001673 type = PIDTYPE_PID;
1674 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 return -EINVAL;
1676 break;
1677 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001678 type = PIDTYPE_PGID;
1679 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 break;
1682 default:
1683 return -EINVAL;
1684 }
1685
Eric W. Biederman161550d2008-02-08 04:19:14 -08001686 if (type < PIDTYPE_MAX)
1687 pid = find_get_pid(upid);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001688
1689 wo.wo_type = type;
1690 wo.wo_pid = pid;
1691 wo.wo_flags = options;
1692 wo.wo_info = infop;
1693 wo.wo_stat = NULL;
1694 wo.wo_rusage = ru;
1695 ret = do_wait(&wo);
Vitaly Mayatskikhdfe16df2009-09-23 15:56:51 -07001696
1697 if (ret > 0) {
1698 ret = 0;
1699 } else if (infop) {
1700 /*
1701 * For a WNOHANG return, clear out all the fields
1702 * we would set so the user can easily tell the
1703 * difference.
1704 */
1705 if (!ret)
1706 ret = put_user(0, &infop->si_signo);
1707 if (!ret)
1708 ret = put_user(0, &infop->si_errno);
1709 if (!ret)
1710 ret = put_user(0, &infop->si_code);
1711 if (!ret)
1712 ret = put_user(0, &infop->si_pid);
1713 if (!ret)
1714 ret = put_user(0, &infop->si_uid);
1715 if (!ret)
1716 ret = put_user(0, &infop->si_status);
1717 }
1718
Eric W. Biederman161550d2008-02-08 04:19:14 -08001719 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720
1721 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001722 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 return ret;
1724}
1725
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001726SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1727 int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001729 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001730 struct pid *pid = NULL;
1731 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 long ret;
1733
1734 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1735 __WNOTHREAD|__WCLONE|__WALL))
1736 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001737
1738 if (upid == -1)
1739 type = PIDTYPE_MAX;
1740 else if (upid < 0) {
1741 type = PIDTYPE_PGID;
1742 pid = find_get_pid(-upid);
1743 } else if (upid == 0) {
1744 type = PIDTYPE_PGID;
Oleg Nesterov2ae448e2009-04-02 16:58:36 -07001745 pid = get_task_pid(current, PIDTYPE_PGID);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001746 } else /* upid > 0 */ {
1747 type = PIDTYPE_PID;
1748 pid = find_get_pid(upid);
1749 }
1750
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001751 wo.wo_type = type;
1752 wo.wo_pid = pid;
1753 wo.wo_flags = options | WEXITED;
1754 wo.wo_info = NULL;
1755 wo.wo_stat = stat_addr;
1756 wo.wo_rusage = ru;
1757 ret = do_wait(&wo);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001758 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759
1760 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001761 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 return ret;
1763}
1764
1765#ifdef __ARCH_WANT_SYS_WAITPID
1766
1767/*
1768 * sys_waitpid() remains for compatibility. waitpid() should be
1769 * implemented by calling sys_wait4() from libc.a.
1770 */
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001771SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772{
1773 return sys_wait4(pid, stat_addr, options, NULL);
1774}
1775
1776#endif