Blame view

kernel/exit.c 43.7 KB
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1
2
3
4
5
  /*
   *  linux/kernel/exit.c
   *
   *  Copyright (C) 1991, 1992  Linus Torvalds
   */
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
6
7
8
  #include <linux/mm.h>
  #include <linux/slab.h>
  #include <linux/interrupt.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
9
  #include <linux/module.h>
c59ede7b7   Randy.Dunlap   [PATCH] move capa...
10
  #include <linux/capability.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
11
12
13
  #include <linux/completion.h>
  #include <linux/personality.h>
  #include <linux/tty.h>
da9cbc873   Jens Axboe   block: blkdev.h c...
14
  #include <linux/iocontext.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
15
16
17
18
  #include <linux/key.h>
  #include <linux/security.h>
  #include <linux/cpu.h>
  #include <linux/acct.h>
8f0ab5147   Jay Lan   [PATCH] csa: conv...
19
  #include <linux/tsacct_kern.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
20
  #include <linux/file.h>
9f3acc314   Al Viro   [PATCH] split lin...
21
  #include <linux/fdtable.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
22
  #include <linux/binfmts.h>
ab516013a   Serge E. Hallyn   [PATCH] namespace...
23
  #include <linux/nsproxy.h>
84d737866   Sukadev Bhattiprolu   [PATCH] add child...
24
  #include <linux/pid_namespace.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
25
26
27
28
  #include <linux/ptrace.h>
  #include <linux/profile.h>
  #include <linux/mount.h>
  #include <linux/proc_fs.h>
49d769d52   Eric W. Biederman   Change reparent_t...
29
  #include <linux/kthread.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
30
  #include <linux/mempolicy.h>
c757249af   Shailabh Nagar   [PATCH] per-task-...
31
  #include <linux/taskstats_kern.h>
ca74e92b4   Shailabh Nagar   [PATCH] per-task-...
32
  #include <linux/delayacct.h>
831441862   Rafael J. Wysocki   Freezer: make ker...
33
  #include <linux/freezer.h>
b4f48b636   Paul Menage   Task Control Grou...
34
  #include <linux/cgroup.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
35
  #include <linux/syscalls.h>
7ed20e1ad   Jesper Juhl   [PATCH] convert t...
36
  #include <linux/signal.h>
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
37
  #include <linux/posix-timers.h>
9f46080c4   Matt Helsley   [PATCH] Process E...
38
  #include <linux/cn_proc.h>
de5097c2e   Ingo Molnar   [PATCH] mutex sub...
39
  #include <linux/mutex.h>
0771dfefc   Ingo Molnar   [PATCH] lightweig...
40
  #include <linux/futex.h>
b92ce5589   Jens Axboe   [PATCH] splice: a...
41
  #include <linux/pipe_fs_i.h>
fa84cb935   Al Viro   [PATCH] move call...
42
  #include <linux/audit.h> /* for audit_free() */
83cc5ed3c   Adrian Bunk   [PATCH] kernel/sy...
43
  #include <linux/resource.h>
0d67a46df   David Howells   [PATCH] BLOCK: Re...
44
  #include <linux/blkdev.h>
6eaeeaba3   Eric Dumazet   getrusage(): fill...
45
  #include <linux/task_io_accounting_ops.h>
30199f5a4   Roland McGrath   tracehook: exit
46
  #include <linux/tracehook.h>
5ad4e53bd   Al Viro   Get rid of indire...
47
  #include <linux/fs_struct.h>
d84f4f992   David Howells   CRED: Inaugurate ...
48
  #include <linux/init_task.h>
cdd6c482c   Ingo Molnar   perf: Do the big ...
49
  #include <linux/perf_event.h>
ad8d75fff   Steven Rostedt   tracing/events: m...
50
  #include <trace/events/sched.h>
24f1e32c6   Frederic Weisbecker   hw-breakpoints: R...
51
  #include <linux/hw_breakpoint.h>
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
52
53
54
55
56
  
  #include <asm/uaccess.h>
  #include <asm/unistd.h>
  #include <asm/pgtable.h>
  #include <asm/mmu_context.h>
408b664a7   Adrian Bunk   [PATCH] make lots...
57
  static void exit_mm(struct task_struct * tsk);
d40e48e02   Oleg Nesterov   exit: __exit_sign...
58
  static void __unhash_process(struct task_struct *p, bool group_dead)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
59
60
61
  {
  	nr_threads--;
  	detach_pid(p, PIDTYPE_PID);
d40e48e02   Oleg Nesterov   exit: __exit_sign...
62
  	if (group_dead) {
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
63
64
  		detach_pid(p, PIDTYPE_PGID);
  		detach_pid(p, PIDTYPE_SID);
c97d98931   Oleg Nesterov   [PATCH] kill SET_...
65

5e85d4abe   Eric W. Biederman   [PATCH] task: Mak...
66
  		list_del_rcu(&p->tasks);
9cd80bbb0   Oleg Nesterov   do_wait() optimiz...
67
  		list_del_init(&p->sibling);
73b9ebfe1   Oleg Nesterov   [PATCH] pidhash: ...
68
  		__get_cpu_var(process_counts)--;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
69
  	}
47e65328a   Oleg Nesterov   [PATCH] pids: kil...
70
  	list_del_rcu(&p->thread_group);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
71
  }
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
72
73
74
75
76
77
  /*
   * This function expects the tasklist_lock write-locked.
   */
  static void __exit_signal(struct task_struct *tsk)
  {
  	struct signal_struct *sig = tsk->signal;
d40e48e02   Oleg Nesterov   exit: __exit_sign...
78
  	bool group_dead = thread_group_leader(tsk);
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
79
  	struct sighand_struct *sighand;
4ada856fb   Oleg Nesterov   signals: clear si...
80
  	struct tty_struct *uninitialized_var(tty);
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
81

d11c563dd   Paul E. McKenney   sched: Use lockde...
82
83
  	sighand = rcu_dereference_check(tsk->sighand,
  					rcu_read_lock_held() ||
db1466b3e   Paul E. McKenney   rcu: Use wrapper ...
84
  					lockdep_tasklist_lock_is_held());
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
85
86
87
  	spin_lock(&sighand->siglock);
  
  	posix_cpu_timers_exit(tsk);
d40e48e02   Oleg Nesterov   exit: __exit_sign...
88
  	if (group_dead) {
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
89
  		posix_cpu_timers_exit_group(tsk);
4ada856fb   Oleg Nesterov   signals: clear si...
90
91
  		tty = sig->tty;
  		sig->tty = NULL;
4a5999429   Oleg Nesterov   exit: avoid sig->...
92
  	} else {
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
93
94
95
96
  		/*
  		 * If there is any task waiting for the group exit
  		 * then notify it:
  		 */
d344193a0   Oleg Nesterov   exit: avoid sig->...
97
  		if (sig->notify_count > 0 && !--sig->notify_count)
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
98
  			wake_up_process(sig->group_exit_task);
6db840fa7   Oleg Nesterov   exec: RT sub-thre...
99

6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
100
101
102
103
104
105
106
107
108
109
110
111
  		if (tsk == sig->curr_target)
  			sig->curr_target = next_thread(tsk);
  		/*
  		 * Accumulate here the counters for all threads but the
  		 * group leader as they die, so they can be added into
  		 * the process-wide totals when those are taken.
  		 * The group leader stays around as a zombie as long
  		 * as there are other threads.  When it gets reaped,
  		 * the exit.c code will add its counts into these totals.
  		 * We won't ever get here for the group leader, since it
  		 * will have been the last reference on the signal_struct.
  		 */
0cf55e1ec   Hidetoshi Seto   sched, cputime: I...
112
113
  		sig->utime = cputime_add(sig->utime, tsk->utime);
  		sig->stime = cputime_add(sig->stime, tsk->stime);
d5b7c78e9   Hidetoshi Seto   sched: Remove tas...
114
  		sig->gtime = cputime_add(sig->gtime, tsk->gtime);
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
115
116
117
118
  		sig->min_flt += tsk->min_flt;
  		sig->maj_flt += tsk->maj_flt;
  		sig->nvcsw += tsk->nvcsw;
  		sig->nivcsw += tsk->nivcsw;
6eaeeaba3   Eric Dumazet   getrusage(): fill...
119
120
  		sig->inblock += task_io_get_inblock(tsk);
  		sig->oublock += task_io_get_oublock(tsk);
5995477ab   Andrea Righi   task IO accountin...
121
  		task_io_accounting_add(&sig->ioac, &tsk->ioac);
32bd671d6   Peter Zijlstra   signal: re-add de...
122
  		sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
123
  	}
b3ac022cb   Oleg Nesterov   proc: turn signal...
124
  	sig->nr_threads--;
d40e48e02   Oleg Nesterov   exit: __exit_sign...
125
  	__unhash_process(tsk, group_dead);
5876700cd   Oleg Nesterov   [PATCH] do __unha...
126

da7978b03   Oleg Nesterov   signals: fix sigq...
127
128
129
130
131
  	/*
  	 * Do this under ->siglock, we can race with another thread
  	 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
  	 */
  	flush_sigqueue(&tsk->pending);
a7e5328a0   Oleg Nesterov   [PATCH] cleanup _...
132
  	tsk->sighand = NULL;
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
133
  	spin_unlock(&sighand->siglock);
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
134

a7e5328a0   Oleg Nesterov   [PATCH] cleanup _...
135
  	__cleanup_sighand(sighand);
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
136
  	clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
d40e48e02   Oleg Nesterov   exit: __exit_sign...
137
  	if (group_dead) {
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
138
  		flush_sigqueue(&sig->shared_pending);
4ada856fb   Oleg Nesterov   signals: clear si...
139
  		tty_kref_put(tty);
6a14c5c9d   Oleg Nesterov   [PATCH] move __ex...
140
141
  	}
  }
8c7904a00   Eric W. Biederman   [PATCH] task: RCU...
142
143
  static void delayed_put_task_struct(struct rcu_head *rhp)
  {
0a16b6075   Mathieu Desnoyers   tracing, sched: L...
144
  	struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
4e231c796   Peter Zijlstra   perf: Fix up dela...
145
  	perf_event_delayed_put(tsk);
0a16b6075   Mathieu Desnoyers   tracing, sched: L...
146
147
  	trace_sched_process_free(tsk);
  	put_task_struct(tsk);
8c7904a00   Eric W. Biederman   [PATCH] task: RCU...
148
  }
f470021ad   Roland McGrath   ptrace children r...
149

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
150
151
  void release_task(struct task_struct * p)
  {
36c8b5868   Ingo Molnar   [PATCH] sched: cl...
152
  	struct task_struct *leader;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
153
  	int zap_leader;
1f09f9749   Oleg Nesterov   [PATCH] release_t...
154
  repeat:
dae33574d   Roland McGrath   tracehook: releas...
155
  	tracehook_prepare_release_task(p);
c69e8d9c0   David Howells   CRED: Use RCU to ...
156
  	/* don't need to get the RCU readlock here - the process is dead and
d11c563dd   Paul E. McKenney   sched: Use lockde...
157
158
  	 * can't be modifying its own credentials. But shut RCU-lockdep up */
  	rcu_read_lock();
c69e8d9c0   David Howells   CRED: Use RCU to ...
159
  	atomic_dec(&__task_cred(p)->user->processes);
d11c563dd   Paul E. McKenney   sched: Use lockde...
160
  	rcu_read_unlock();
c69e8d9c0   David Howells   CRED: Use RCU to ...
161

60347f671   Pavel Emelyanov   pid namespaces: p...
162
  	proc_flush_task(p);
0203026b5   Ingo Molnar   perf_counter: fix...
163

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
164
  	write_lock_irq(&tasklist_lock);
dae33574d   Roland McGrath   tracehook: releas...
165
  	tracehook_finish_release_task(p);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
166
  	__exit_signal(p);
35f5cad8c   Oleg Nesterov   [PATCH] revert "O...
167

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
168
169
170
171
172
173
174
175
  	/*
  	 * If we are the last non-leader member of the thread
  	 * group, and the leader is zombie, then notify the
  	 * group leader's parent process. (if it wants notification.)
  	 */
  	zap_leader = 0;
  	leader = p->group_leader;
  	if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
d839fd4d2   Oleg Nesterov   ptrace: introduce...
176
  		BUG_ON(task_detached(leader));
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
177
178
179
180
181
182
183
184
185
  		do_notify_parent(leader, leader->exit_signal);
  		/*
  		 * If we were the last child thread and the leader has
  		 * exited already, and the leader's parent ignores SIGCHLD,
  		 * then we are the one who should release the leader.
  		 *
  		 * do_notify_parent() will have marked it self-reaping in
  		 * that case.
  		 */
d839fd4d2   Oleg Nesterov   ptrace: introduce...
186
  		zap_leader = task_detached(leader);
dae33574d   Roland McGrath   tracehook: releas...
187
188
189
190
191
192
193
  
  		/*
  		 * This maintains the invariant that release_task()
  		 * only runs on a task in EXIT_DEAD, just for sanity.
  		 */
  		if (zap_leader)
  			leader->exit_state = EXIT_DEAD;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
194
  	}
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
195
  	write_unlock_irq(&tasklist_lock);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
196
  	release_thread(p);
8c7904a00   Eric W. Biederman   [PATCH] task: RCU...
197
  	call_rcu(&p->rcu, delayed_put_task_struct);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
198
199
200
201
202
  
  	p = leader;
  	if (unlikely(zap_leader))
  		goto repeat;
  }
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
203
204
205
206
  /*
   * This checks not only the pgrp, but falls back on the pid if no
   * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
   * without this...
04a2e6a5c   Eric W. Biederman   [PATCH] pid: make...
207
208
   *
   * The caller must hold rcu lock or the tasklist lock.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
209
   */
04a2e6a5c   Eric W. Biederman   [PATCH] pid: make...
210
  struct pid *session_of_pgrp(struct pid *pgrp)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
211
212
  {
  	struct task_struct *p;
04a2e6a5c   Eric W. Biederman   [PATCH] pid: make...
213
  	struct pid *sid = NULL;
62dfb5541   Oleg Nesterov   [PATCH] session_o...
214

04a2e6a5c   Eric W. Biederman   [PATCH] pid: make...
215
  	p = pid_task(pgrp, PIDTYPE_PGID);
62dfb5541   Oleg Nesterov   [PATCH] session_o...
216
  	if (p == NULL)
04a2e6a5c   Eric W. Biederman   [PATCH] pid: make...
217
  		p = pid_task(pgrp, PIDTYPE_PID);
62dfb5541   Oleg Nesterov   [PATCH] session_o...
218
  	if (p != NULL)
04a2e6a5c   Eric W. Biederman   [PATCH] pid: make...
219
  		sid = task_session(p);
62dfb5541   Oleg Nesterov   [PATCH] session_o...
220

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
221
222
223
224
225
226
227
228
229
230
231
  	return sid;
  }
  
  /*
   * Determine if a process group is "orphaned", according to the POSIX
   * definition in 2.2.2.52.  Orphaned process groups are not to be affected
   * by terminal-generated stop signals.  Newly orphaned process groups are
   * to receive a SIGHUP and a SIGCONT.
   *
   * "I ask you, have you ever known what it is to be an orphan?"
   */
0475ac084   Eric W. Biederman   [PATCH] pid: use ...
232
  static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
233
234
  {
  	struct task_struct *p;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
235

0475ac084   Eric W. Biederman   [PATCH] pid: use ...
236
  	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
05e83df62   Oleg Nesterov   will_become_orpha...
237
238
239
  		if ((p == ignored_task) ||
  		    (p->exit_state && thread_group_empty(p)) ||
  		    is_global_init(p->real_parent))
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
240
  			continue;
05e83df62   Oleg Nesterov   will_become_orpha...
241

0475ac084   Eric W. Biederman   [PATCH] pid: use ...
242
  		if (task_pgrp(p->real_parent) != pgrp &&
05e83df62   Oleg Nesterov   will_become_orpha...
243
244
  		    task_session(p->real_parent) == task_session(p))
  			return 0;
0475ac084   Eric W. Biederman   [PATCH] pid: use ...
245
  	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
05e83df62   Oleg Nesterov   will_become_orpha...
246
247
  
  	return 1;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
248
  }
3e7cd6c41   Eric W. Biederman   [PATCH] pid: repl...
249
  int is_current_pgrp_orphaned(void)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
250
251
252
253
  {
  	int retval;
  
  	read_lock(&tasklist_lock);
3e7cd6c41   Eric W. Biederman   [PATCH] pid: repl...
254
  	retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
255
256
257
258
  	read_unlock(&tasklist_lock);
  
  	return retval;
  }
0475ac084   Eric W. Biederman   [PATCH] pid: use ...
259
  static int has_stopped_jobs(struct pid *pgrp)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
260
261
262
  {
  	int retval = 0;
  	struct task_struct *p;
0475ac084   Eric W. Biederman   [PATCH] pid: use ...
263
  	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
338077e54   Matthew Wilcox   exit: Use task_is_*
264
  		if (!task_is_stopped(p))
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
265
  			continue;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
266
267
  		retval = 1;
  		break;
0475ac084   Eric W. Biederman   [PATCH] pid: use ...
268
  	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
269
270
  	return retval;
  }
f49ee505b   Oleg Nesterov   introduce kill_or...
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
  /*
   * Check to see if any process groups have become orphaned as
   * a result of our exiting, and if they have any stopped jobs,
   * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
   */
  static void
  kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
  {
  	struct pid *pgrp = task_pgrp(tsk);
  	struct task_struct *ignored_task = tsk;
  
  	if (!parent)
  		 /* exit: our father is in a different pgrp than
  		  * we are and we were the only connection outside.
  		  */
  		parent = tsk->real_parent;
  	else
  		/* reparent: our child is in a different pgrp than
  		 * we are, and it was the only connection outside.
  		 */
  		ignored_task = NULL;
  
  	if (task_pgrp(parent) != pgrp &&
  	    task_session(parent) == task_session(tsk) &&
  	    will_become_orphaned_pgrp(pgrp, ignored_task) &&
  	    has_stopped_jobs(pgrp)) {
  		__kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
  		__kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
  	}
  }
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
301
  /**
49d769d52   Eric W. Biederman   Change reparent_t...
302
   * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
303
304
   *
   * If a kernel thread is launched as a result of a system call, or if
49d769d52   Eric W. Biederman   Change reparent_t...
305
306
   * it ever exits, it should generally reparent itself to kthreadd so it
   * isn't in the way of other processes and is correctly cleaned up on exit.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
307
308
309
310
   *
   * The various task state such as scheduling policy and priority may have
   * been inherited from a user process, so we reset them to sane values here.
   *
49d769d52   Eric W. Biederman   Change reparent_t...
311
   * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
312
   */
49d769d52   Eric W. Biederman   Change reparent_t...
313
  static void reparent_to_kthreadd(void)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
314
315
316
317
318
  {
  	write_lock_irq(&tasklist_lock);
  
  	ptrace_unlink(current);
  	/* Reparent to init */
49d769d52   Eric W. Biederman   Change reparent_t...
319
  	current->real_parent = current->parent = kthreadd_task;
f470021ad   Roland McGrath   ptrace children r...
320
  	list_move_tail(&current->sibling, &current->real_parent->children);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
321
322
323
  
  	/* Set the exit signal to SIGCHLD so we signal init on exit */
  	current->exit_signal = SIGCHLD;
e05606d33   Ingo Molnar   sched: clean up t...
324
  	if (task_nice(current) < 0)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
325
326
327
328
  		set_user_nice(current, 0);
  	/* cpus_allowed? */
  	/* rt_priority? */
  	/* signals? */
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
329
330
  	memcpy(current->signal->rlim, init_task.signal->rlim,
  	       sizeof(current->signal->rlim));
d84f4f992   David Howells   CRED: Inaugurate ...
331
332
333
  
  	atomic_inc(&init_cred.usage);
  	commit_creds(&init_cred);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
334
  	write_unlock_irq(&tasklist_lock);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
335
  }
8520d7c7f   Oleg Nesterov   teach set_special...
336
  void __set_special_pids(struct pid *pid)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
337
  {
e19f247a3   Oren Laadan   [PATCH] setpgid: ...
338
  	struct task_struct *curr = current->group_leader;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
339

0d0df599f   Christian Borntraeger   connector: fix re...
340
  	if (task_session(curr) != pid)
7d8da0962   Oleg Nesterov   pids: __set_speci...
341
  		change_pid(curr, PIDTYPE_SID, pid);
1b0f7ffd0   Oleg Nesterov   pids: kill signal...
342
343
  
  	if (task_pgrp(curr) != pid)
7d8da0962   Oleg Nesterov   pids: __set_speci...
344
  		change_pid(curr, PIDTYPE_PGID, pid);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
345
  }
8520d7c7f   Oleg Nesterov   teach set_special...
346
  static void set_special_pids(struct pid *pid)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
347
348
  {
  	write_lock_irq(&tasklist_lock);
8520d7c7f   Oleg Nesterov   teach set_special...
349
  	__set_special_pids(pid);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
350
351
352
353
  	write_unlock_irq(&tasklist_lock);
  }
  
  /*
87245135d   Oleg Nesterov   allow_signal: kil...
354
355
   * Let kernel threads use this to say that they allow a certain signal.
   * Must not be used if kthread was cloned with CLONE_SIGHAND.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
356
357
358
   */
  int allow_signal(int sig)
  {
7ed20e1ad   Jesper Juhl   [PATCH] convert t...
359
  	if (!valid_signal(sig) || sig < 1)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
360
361
362
  		return -EINVAL;
  
  	spin_lock_irq(&current->sighand->siglock);
87245135d   Oleg Nesterov   allow_signal: kil...
363
  	/* This is only needed for daemonize()'ed kthreads */
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
364
  	sigdelset(&current->blocked, sig);
87245135d   Oleg Nesterov   allow_signal: kil...
365
366
367
368
369
370
  	/*
  	 * Kernel threads handle their own signals. Let the signal code
  	 * know it'll be handled, so that they don't get converted to
  	 * SIGKILL or just silently dropped.
  	 */
  	current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
371
372
373
374
375
376
377
378
379
  	recalc_sigpending();
  	spin_unlock_irq(&current->sighand->siglock);
  	return 0;
  }
  
  EXPORT_SYMBOL(allow_signal);
  
  int disallow_signal(int sig)
  {
7ed20e1ad   Jesper Juhl   [PATCH] convert t...
380
  	if (!valid_signal(sig) || sig < 1)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
381
382
383
  		return -EINVAL;
  
  	spin_lock_irq(&current->sighand->siglock);
10ab825bd   Oleg Nesterov   change kernel thr...
384
  	current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
  	recalc_sigpending();
  	spin_unlock_irq(&current->sighand->siglock);
  	return 0;
  }
  
  EXPORT_SYMBOL(disallow_signal);
  
  /*
   *	Put all the gunge required to become a kernel thread without
   *	attached user resources in one place where it belongs.
   */
  
  void daemonize(const char *name, ...)
  {
  	va_list args;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
400
401
402
403
404
405
406
407
408
409
410
411
  	sigset_t blocked;
  
  	va_start(args, name);
  	vsnprintf(current->comm, sizeof(current->comm), name, args);
  	va_end(args);
  
  	/*
  	 * If we were started as result of loading a module, close all of the
  	 * user space pages.  We don't need them, and if we didn't close them
  	 * they would be locked into memory.
  	 */
  	exit_mm(current);
831441862   Rafael J. Wysocki   Freezer: make ker...
412
413
414
415
  	/*
  	 * We don't want to have TIF_FREEZE set if the system-wide hibernation
  	 * or suspend transition begins right now.
  	 */
7b34e4283   Oleg Nesterov   introduce PF_KTHR...
416
  	current->flags |= (PF_NOFREEZE | PF_KTHREAD);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
417

8520d7c7f   Oleg Nesterov   teach set_special...
418
419
420
421
  	if (current->nsproxy != &init_nsproxy) {
  		get_nsproxy(&init_nsproxy);
  		switch_task_namespaces(current, &init_nsproxy);
  	}
297bd42b1   Oleg Nesterov   move daemonized k...
422
  	set_special_pids(&init_struct_pid);
24ec839c4   Peter Zijlstra   [PATCH] tty: ->si...
423
  	proc_clear_tty(current);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
424
425
426
427
428
429
430
  
  	/* Block and flush all signals */
  	sigfillset(&blocked);
  	sigprocmask(SIG_BLOCK, &blocked, NULL);
  	flush_signals(current);
  
  	/* Become as one with the init task */
3e93cd671   Al Viro   Take fs_struct ha...
431
  	daemonize_fs_struct();
d4c5e41f3   Daniel Walker   whitespace fixes:...
432
  	exit_files(current);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
433
434
  	current->files = init_task.files;
  	atomic_inc(&current->files->count);
49d769d52   Eric W. Biederman   Change reparent_t...
435
  	reparent_to_kthreadd();
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
436
437
438
  }
  
  EXPORT_SYMBOL(daemonize);
858119e15   Arjan van de Ven   [PATCH] Unlinline...
439
  static void close_files(struct files_struct * files)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
440
441
  {
  	int i, j;
badf16621   Dipankar Sarma   [PATCH] files: br...
442
  	struct fdtable *fdt;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
443
444
  
  	j = 0;
4fb3a5386   Dipankar Sarma   [PATCH] files: fi...
445
446
447
448
  
  	/*
  	 * It is safe to dereference the fd table without RCU or
  	 * ->file_lock because this is the last reference to the
d11c563dd   Paul E. McKenney   sched: Use lockde...
449
  	 * files structure.  But use RCU to shut RCU-lockdep up.
4fb3a5386   Dipankar Sarma   [PATCH] files: fi...
450
  	 */
d11c563dd   Paul E. McKenney   sched: Use lockde...
451
  	rcu_read_lock();
badf16621   Dipankar Sarma   [PATCH] files: br...
452
  	fdt = files_fdtable(files);
d11c563dd   Paul E. McKenney   sched: Use lockde...
453
  	rcu_read_unlock();
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
454
455
456
  	for (;;) {
  		unsigned long set;
  		i = j * __NFDBITS;
bbea9f696   Vadim Lobanov   [PATCH] fdtable: ...
457
  		if (i >= fdt->max_fds)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
458
  			break;
badf16621   Dipankar Sarma   [PATCH] files: br...
459
  		set = fdt->open_fds->fds_bits[j++];
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
460
461
  		while (set) {
  			if (set & 1) {
badf16621   Dipankar Sarma   [PATCH] files: br...
462
  				struct file * file = xchg(&fdt->fd[i], NULL);
944be0b22   Ingo Molnar   [PATCH] close_fil...
463
  				if (file) {
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
464
  					filp_close(file, files);
944be0b22   Ingo Molnar   [PATCH] close_fil...
465
466
  					cond_resched();
  				}
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
  			}
  			i++;
  			set >>= 1;
  		}
  	}
  }
  
  struct files_struct *get_files_struct(struct task_struct *task)
  {
  	struct files_struct *files;
  
  	task_lock(task);
  	files = task->files;
  	if (files)
  		atomic_inc(&files->count);
  	task_unlock(task);
  
  	return files;
  }
7ad5b3a50   Harvey Harrison   kernel: remove fa...
486
  void put_files_struct(struct files_struct *files)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
487
  {
badf16621   Dipankar Sarma   [PATCH] files: br...
488
  	struct fdtable *fdt;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
489
490
491
492
  	if (atomic_dec_and_test(&files->count)) {
  		close_files(files);
  		/*
  		 * Free the fd and fdset arrays if we expanded them.
ab2af1f50   Dipankar Sarma   [PATCH] files: fi...
493
494
495
  		 * If the fdtable was embedded, pass files for freeing
  		 * at the end of the RCU grace period. Otherwise,
  		 * you can free files immediately.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
496
  		 */
d11c563dd   Paul E. McKenney   sched: Use lockde...
497
  		rcu_read_lock();
badf16621   Dipankar Sarma   [PATCH] files: br...
498
  		fdt = files_fdtable(files);
4fd45812c   Vadim Lobanov   [PATCH] fdtable: ...
499
  		if (fdt != &files->fdtab)
ab2af1f50   Dipankar Sarma   [PATCH] files: fi...
500
  			kmem_cache_free(files_cachep, files);
01b2d93ca   Vadim Lobanov   [PATCH] fdtable: ...
501
  		free_fdtable(fdt);
d11c563dd   Paul E. McKenney   sched: Use lockde...
502
  		rcu_read_unlock();
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
503
504
  	}
  }
3b1253880   Al Viro   [PATCH] sanitize ...
505
  void reset_files_struct(struct files_struct *files)
3b9b8ab65   Kirill Korotaev   [PATCH] Fix unser...
506
  {
3b1253880   Al Viro   [PATCH] sanitize ...
507
  	struct task_struct *tsk = current;
3b9b8ab65   Kirill Korotaev   [PATCH] Fix unser...
508
509
510
511
512
513
514
515
  	struct files_struct *old;
  
  	old = tsk->files;
  	task_lock(tsk);
  	tsk->files = files;
  	task_unlock(tsk);
  	put_files_struct(old);
  }
3b9b8ab65   Kirill Korotaev   [PATCH] Fix unser...
516

1ec7f1ddb   Al Viro   [PATCH] get rid o...
517
  void exit_files(struct task_struct *tsk)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
518
519
520
521
522
523
524
525
526
527
  {
  	struct files_struct * files = tsk->files;
  
  	if (files) {
  		task_lock(tsk);
  		tsk->files = NULL;
  		task_unlock(tsk);
  		put_files_struct(files);
  	}
  }
cf475ad28   Balbir Singh   cgroups: add an o...
528
529
530
531
532
533
534
535
536
537
538
  #ifdef CONFIG_MM_OWNER
  /*
   * Task p is exiting and it owned mm, lets find a new owner for it
   */
  static inline int
  mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
  {
  	/*
  	 * If there are other users of the mm and the owner (us) is exiting
  	 * we need to find a new owner to take on the responsibility.
  	 */
cf475ad28   Balbir Singh   cgroups: add an o...
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
  	if (atomic_read(&mm->mm_users) <= 1)
  		return 0;
  	if (mm->owner != p)
  		return 0;
  	return 1;
  }
  
  void mm_update_next_owner(struct mm_struct *mm)
  {
  	struct task_struct *c, *g, *p = current;
  
  retry:
  	if (!mm_need_new_owner(mm, p))
  		return;
  
  	read_lock(&tasklist_lock);
  	/*
  	 * Search in the children
  	 */
  	list_for_each_entry(c, &p->children, sibling) {
  		if (c->mm == mm)
  			goto assign_new_owner;
  	}
  
  	/*
  	 * Search in the siblings
  	 */
dea33cfd9   Oleg Nesterov   ptrace: mm_need_n...
566
  	list_for_each_entry(c, &p->real_parent->children, sibling) {
cf475ad28   Balbir Singh   cgroups: add an o...
567
568
569
570
571
572
573
574
575
576
577
578
579
580
  		if (c->mm == mm)
  			goto assign_new_owner;
  	}
  
  	/*
  	 * Search through everything else. We should not get
  	 * here often
  	 */
  	do_each_thread(g, c) {
  		if (c->mm == mm)
  			goto assign_new_owner;
  	} while_each_thread(g, c);
  
  	read_unlock(&tasklist_lock);
31a78f23b   Balbir Singh   mm owner: fix rac...
581
582
583
  	/*
  	 * We found no owner yet mm_users > 1: this implies that we are
  	 * most likely racing with swapoff (try_to_unuse()) or /proc or
e5991371e   Hugh Dickins   mm: remove cgroup...
584
  	 * ptrace or page migration (get_task_mm()).  Mark owner as NULL.
31a78f23b   Balbir Singh   mm owner: fix rac...
585
  	 */
31a78f23b   Balbir Singh   mm owner: fix rac...
586
  	mm->owner = NULL;
cf475ad28   Balbir Singh   cgroups: add an o...
587
588
589
590
591
592
593
594
595
596
  	return;
  
  assign_new_owner:
  	BUG_ON(c == p);
  	get_task_struct(c);
  	/*
  	 * The task_lock protects c->mm from changing.
  	 * We always want mm->owner->mm == mm
  	 */
  	task_lock(c);
e5991371e   Hugh Dickins   mm: remove cgroup...
597
598
599
600
601
  	/*
  	 * Delay read_unlock() till we have the task_lock()
  	 * to ensure that c does not slip away underneath us
  	 */
  	read_unlock(&tasklist_lock);
cf475ad28   Balbir Singh   cgroups: add an o...
602
603
604
605
606
  	if (c->mm != mm) {
  		task_unlock(c);
  		put_task_struct(c);
  		goto retry;
  	}
cf475ad28   Balbir Singh   cgroups: add an o...
607
608
609
610
611
  	mm->owner = c;
  	task_unlock(c);
  	put_task_struct(c);
  }
  #endif /* CONFIG_MM_OWNER */
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
612
613
614
615
  /*
   * Turn us into a lazy TLB process if we
   * aren't already..
   */
408b664a7   Adrian Bunk   [PATCH] make lots...
616
  static void exit_mm(struct task_struct * tsk)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
617
618
  {
  	struct mm_struct *mm = tsk->mm;
b564daf80   Oleg Nesterov   coredump: constru...
619
  	struct core_state *core_state;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
620
621
622
623
624
625
  
  	mm_release(tsk, mm);
  	if (!mm)
  		return;
  	/*
  	 * Serialize with any possible pending coredump.
999d9fc16   Oleg Nesterov   coredump: move mm...
626
  	 * We must hold mmap_sem around checking core_state
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
627
  	 * and clearing tsk->mm.  The core-inducing thread
999d9fc16   Oleg Nesterov   coredump: move mm...
628
  	 * will increment ->nr_threads for each thread in the
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
629
630
631
  	 * group with ->mm != NULL.
  	 */
  	down_read(&mm->mmap_sem);
b564daf80   Oleg Nesterov   coredump: constru...
632
633
634
  	core_state = mm->core_state;
  	if (core_state) {
  		struct core_thread self;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
635
  		up_read(&mm->mmap_sem);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
636

b564daf80   Oleg Nesterov   coredump: constru...
637
638
639
640
641
642
643
644
  		self.task = tsk;
  		self.next = xchg(&core_state->dumper.next, &self);
  		/*
  		 * Implies mb(), the result of xchg() must be visible
  		 * to core_state->dumper.
  		 */
  		if (atomic_dec_and_test(&core_state->nr_threads))
  			complete(&core_state->startup);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
645

a94e2d408   Oleg Nesterov   coredump: kill mm...
646
647
648
649
650
651
652
  		for (;;) {
  			set_task_state(tsk, TASK_UNINTERRUPTIBLE);
  			if (!self.task) /* see coredump_finish() */
  				break;
  			schedule();
  		}
  		__set_task_state(tsk, TASK_RUNNING);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
653
654
655
  		down_read(&mm->mmap_sem);
  	}
  	atomic_inc(&mm->mm_count);
125e18745   Eric Sesterhenn   [PATCH] More BUG_...
656
  	BUG_ON(mm != tsk->active_mm);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
657
658
659
660
661
  	/* more a memory barrier than a real lock */
  	task_lock(tsk);
  	tsk->mm = NULL;
  	up_read(&mm->mmap_sem);
  	enter_lazy_tlb(mm, current);
0c1eecfb3   Rafael J. Wysocki   Freezer: avoid fr...
662
663
  	/* We don't want this task to be frozen prematurely */
  	clear_freeze_flag(tsk);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
664
  	task_unlock(tsk);
cf475ad28   Balbir Singh   cgroups: add an o...
665
  	mm_update_next_owner(mm);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
666
667
  	mmput(mm);
  }
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
668
669
670
671
  /*
   * When we die, we re-parent all our children.
   * Try to give them to another thread in our thread
   * group, and if no such member exists, give it to
84d737866   Sukadev Bhattiprolu   [PATCH] add child...
672
673
   * the child reaper process (ie "init") in our pid
   * space.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
674
   */
950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
675
  static struct task_struct *find_new_reaper(struct task_struct *father)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
676
  {
950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
677
678
  	struct pid_namespace *pid_ns = task_active_pid_ns(father);
  	struct task_struct *thread;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
679

950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
680
681
682
683
684
685
686
687
  	thread = father;
  	while_each_thread(father, thread) {
  		if (thread->flags & PF_EXITING)
  			continue;
  		if (unlikely(pid_ns->child_reaper == father))
  			pid_ns->child_reaper = thread;
  		return thread;
  	}
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
688

950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
689
690
691
692
  	if (unlikely(pid_ns->child_reaper == father)) {
  		write_unlock_irq(&tasklist_lock);
  		if (unlikely(pid_ns == &init_pid_ns))
  			panic("Attempted to kill init!");
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
693

950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
694
695
  		zap_pid_ns_processes(pid_ns);
  		write_lock_irq(&tasklist_lock);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
696
  		/*
950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
697
698
699
  		 * We can not clear ->child_reaper or leave it alone.
  		 * There may by stealth EXIT_DEAD tasks on ->children,
  		 * forget_original_parent() must move them somewhere.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
700
  		 */
950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
701
  		pid_ns->child_reaper = init_pid_ns.child_reaper;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
702
  	}
762a24bee   Oleg Nesterov   pid namespaces: r...
703

950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
704
705
  	return pid_ns->child_reaper;
  }
5dfc80be7   Oleg Nesterov   forget_original_p...
706
707
708
  /*
  * Any that need to be release_task'd are put on the @dead list.
   */
9cd80bbb0   Oleg Nesterov   do_wait() optimiz...
709
  static void reparent_leader(struct task_struct *father, struct task_struct *p,
5dfc80be7   Oleg Nesterov   forget_original_p...
710
711
  				struct list_head *dead)
  {
5dfc80be7   Oleg Nesterov   forget_original_p...
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
  	list_move_tail(&p->sibling, &p->real_parent->children);
  
  	if (task_detached(p))
  		return;
  	/*
  	 * If this is a threaded reparent there is no need to
  	 * notify anyone anything has happened.
  	 */
  	if (same_thread_group(p->real_parent, father))
  		return;
  
  	/* We don't want people slaying init.  */
  	p->exit_signal = SIGCHLD;
  
  	/* If it has exited notify the new parent about this child's death. */
5cb114468   Oleg Nesterov   ptrace: do not us...
727
  	if (!task_ptrace(p) &&
5dfc80be7   Oleg Nesterov   forget_original_p...
728
729
730
731
732
733
734
735
736
737
  	    p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
  		do_notify_parent(p, p->exit_signal);
  		if (task_detached(p)) {
  			p->exit_state = EXIT_DEAD;
  			list_move_tail(&p->sibling, dead);
  		}
  	}
  
  	kill_orphaned_pgrp(p, father);
  }
762a24bee   Oleg Nesterov   pid namespaces: r...
738
  static void forget_original_parent(struct task_struct *father)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
739
  {
950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
740
  	struct task_struct *p, *n, *reaper;
5dfc80be7   Oleg Nesterov   forget_original_p...
741
  	LIST_HEAD(dead_children);
762a24bee   Oleg Nesterov   pid namespaces: r...
742
743
  
  	write_lock_irq(&tasklist_lock);
c7e49c148   Oleg Nesterov   ptrace: optimize ...
744
745
746
747
748
  	/*
  	 * Note that exit_ptrace() and find_new_reaper() might
  	 * drop tasklist_lock and reacquire it.
  	 */
  	exit_ptrace(father);
950bbabb5   Oleg Nesterov   pid_ns: (BUG 1139...
749
  	reaper = find_new_reaper(father);
f470021ad   Roland McGrath   ptrace children r...
750

03ff17979   Matthias Kaehlcke   kernel/exit.c: Us...
751
  	list_for_each_entry_safe(p, n, &father->children, sibling) {
9cd80bbb0   Oleg Nesterov   do_wait() optimiz...
752
753
754
755
756
757
758
759
760
761
762
763
  		struct task_struct *t = p;
  		do {
  			t->real_parent = reaper;
  			if (t->parent == father) {
  				BUG_ON(task_ptrace(t));
  				t->parent = t->real_parent;
  			}
  			if (t->pdeath_signal)
  				group_send_sig_info(t->pdeath_signal,
  						    SEND_SIG_NOINFO, t);
  		} while_each_thread(p, t);
  		reparent_leader(father, p, &dead_children);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
764
  	}
762a24bee   Oleg Nesterov   pid namespaces: r...
765
  	write_unlock_irq(&tasklist_lock);
5dfc80be7   Oleg Nesterov   forget_original_p...
766

762a24bee   Oleg Nesterov   pid namespaces: r...
767
  	BUG_ON(!list_empty(&father->children));
762a24bee   Oleg Nesterov   pid namespaces: r...
768

5dfc80be7   Oleg Nesterov   forget_original_p...
769
770
  	list_for_each_entry_safe(p, n, &dead_children, sibling) {
  		list_del_init(&p->sibling);
39c626ae4   Oleg Nesterov   forget_original_p...
771
772
  		release_task(p);
  	}
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
773
774
775
776
777
778
  }
  
  /*
   * Send signals to all our closest relatives so that they know
   * to properly mourn us..
   */
821c7de71   Oleg Nesterov   exit_notify: fix ...
779
  static void exit_notify(struct task_struct *tsk, int group_dead)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
780
  {
2b2a1ff64   Roland McGrath   tracehook: death
781
782
  	int signal;
  	void *cookie;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
783

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
784
785
786
787
788
789
790
791
  	/*
  	 * This does two things:
  	 *
    	 * A.  Make init inherit all the child processes
  	 * B.  Check to see if any process groups have become orphaned
  	 *	as a result of our exiting, and if they have any stopped
  	 *	jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
  	 */
762a24bee   Oleg Nesterov   pid namespaces: r...
792
  	forget_original_parent(tsk);
2e4a70726   Pavel Emelyanov   pid namespaces: m...
793
  	exit_task_namespaces(tsk);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
794

762a24bee   Oleg Nesterov   pid namespaces: r...
795
  	write_lock_irq(&tasklist_lock);
821c7de71   Oleg Nesterov   exit_notify: fix ...
796
797
  	if (group_dead)
  		kill_orphaned_pgrp(tsk->group_leader, NULL);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
798

247284481   Oleg Nesterov   Kill some obsolet...
799
  	/* Let father know we died
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
800
801
  	 *
  	 * Thread signals are configurable, but you aren't going to use
d4c5e41f3   Daniel Walker   whitespace fixes:...
802
  	 * that to send signals to arbitary processes.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
803
804
805
806
807
808
809
810
811
  	 * That stops right now.
  	 *
  	 * If the parent exec id doesn't match the exec id we saved
  	 * when we started then we know the parent has changed security
  	 * domain.
  	 *
  	 * If our self_exec id doesn't match our parent_exec_id then
  	 * we have changed execution domain as these two values started
  	 * the same after a fork.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
812
  	 */
d839fd4d2   Oleg Nesterov   ptrace: introduce...
813
  	if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
f49ee505b   Oleg Nesterov   introduce kill_or...
814
  	    (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
432870dab   Oleg Nesterov   exit_notify: kill...
815
  	     tsk->self_exec_id != tsk->parent_exec_id))
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
816
  		tsk->exit_signal = SIGCHLD;
2b2a1ff64   Roland McGrath   tracehook: death
817
  	signal = tracehook_notify_death(tsk, &cookie, group_dead);
5c7edcd7e   Roland McGrath   tracehook: fix ex...
818
  	if (signal >= 0)
2b2a1ff64   Roland McGrath   tracehook: death
819
  		signal = do_notify_parent(tsk, signal);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
820

5c7edcd7e   Roland McGrath   tracehook: fix ex...
821
  	tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
822

9c3391684   Oleg Nesterov   exit: exit_notify...
823
824
  	/* mt-exec, de_thread() is waiting for group leader */
  	if (unlikely(tsk->signal->notify_count < 0))
6db840fa7   Oleg Nesterov   exec: RT sub-thre...
825
  		wake_up_process(tsk->signal->group_exit_task);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
826
  	write_unlock_irq(&tasklist_lock);
2b2a1ff64   Roland McGrath   tracehook: death
827
  	tracehook_report_death(tsk, signal, cookie, group_dead);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
828
  	/* If the process is dead, release it - nobody will wait for it */
5c7edcd7e   Roland McGrath   tracehook: fix ex...
829
  	if (signal == DEATH_REAP)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
830
  		release_task(tsk);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
831
  }
e18eecb8b   Jeff Dike   Add generic exit-...
832
833
834
835
836
  #ifdef CONFIG_DEBUG_STACK_USAGE
  static void check_stack_usage(void)
  {
  	static DEFINE_SPINLOCK(low_water_lock);
  	static int lowest_to_date = THREAD_SIZE;
e18eecb8b   Jeff Dike   Add generic exit-...
837
  	unsigned long free;
7c9f8861e   Eric Sandeen   stackprotector: u...
838
  	free = stack_not_used(current);
e18eecb8b   Jeff Dike   Add generic exit-...
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
  
  	if (free >= lowest_to_date)
  		return;
  
  	spin_lock(&low_water_lock);
  	if (free < lowest_to_date) {
  		printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
  				"left
  ",
  				current->comm, free);
  		lowest_to_date = free;
  	}
  	spin_unlock(&low_water_lock);
  }
  #else
  static inline void check_stack_usage(void) {}
  #endif
7ad5b3a50   Harvey Harrison   kernel: remove fa...
856
  NORET_TYPE void do_exit(long code)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
857
858
859
860
861
  {
  	struct task_struct *tsk = current;
  	int group_dead;
  
  	profile_task_exit(tsk);
22e2c507c   Jens Axboe   [PATCH] Update cf...
862
  	WARN_ON(atomic_read(&tsk->fs_excl));
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
863
864
865
866
  	if (unlikely(in_interrupt()))
  		panic("Aiee, killing interrupt handler!");
  	if (unlikely(!tsk->pid))
  		panic("Attempted to kill the idle task!");
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
867

30199f5a4   Roland McGrath   tracehook: exit
868
  	tracehook_report_exit(&code);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
869

e0e817392   David Howells   CRED: Add some co...
870
  	validate_creds_for_do_exit(tsk);
df164db5f   Alexander Nyberg   [PATCH] avoid res...
871
872
873
874
875
876
877
878
  	/*
  	 * We're taking recursive faults here in do_exit. Safest is to just
  	 * leave this task alone and wait for reboot.
  	 */
  	if (unlikely(tsk->flags & PF_EXITING)) {
  		printk(KERN_ALERT
  			"Fixing recursive fault but reboot is needed!
  ");
778e9a9c3   Alexey Kuznetsov   pi-futex: fix exi...
879
880
881
882
883
884
885
886
887
888
  		/*
  		 * We can do this unlocked here. The futex code uses
  		 * this flag just to verify whether the pi state
  		 * cleanup has been done or not. In the worst case it
  		 * loops once more. We pretend that the cleanup was
  		 * done as there is no way to return. Either the
  		 * OWNER_DIED bit is set by now or we push the blocked
  		 * task into the wait for ever nirwana as well.
  		 */
  		tsk->flags |= PF_EXITPIDONE;
df164db5f   Alexander Nyberg   [PATCH] avoid res...
889
890
891
  		set_current_state(TASK_UNINTERRUPTIBLE);
  		schedule();
  	}
3aa551c9b   Thomas Gleixner   genirq: add threa...
892
  	exit_irq_thread();
d12619b5f   Oleg Nesterov   fix group stop wi...
893
  	exit_signals(tsk);  /* sets PF_EXITING */
778e9a9c3   Alexey Kuznetsov   pi-futex: fix exi...
894
895
896
897
  	/*
  	 * tsk->flags are checked in the futex code to protect against
  	 * an exiting task cleaning up the robust pi futexes.
  	 */
d2ee7198c   Oleg Nesterov   pi-futex: set PF_...
898
  	smp_mb();
1d6154825   Thomas Gleixner   sched: Convert pi...
899
  	raw_spin_unlock_wait(&tsk->pi_lock);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
900

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
901
902
903
  	if (unlikely(in_atomic()))
  		printk(KERN_INFO "note: %s[%d] exited with preempt_count %d
  ",
ba25f9dcc   Pavel Emelyanov   Use helpers to ob...
904
  				current->comm, task_pid_nr(current),
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
905
906
907
  				preempt_count());
  
  	acct_update_integrals(tsk);
34e55232e   KAMEZAWA Hiroyuki   mm: avoid false s...
908
  	/* sync mm's RSS info before statistics gathering */
a3a2e76c7   KAMEZAWA Hiroyuki   mm: avoid null-po...
909
910
  	if (tsk->mm)
  		sync_mm_rss(tsk, tsk->mm);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
911
  	group_dead = atomic_dec_and_test(&tsk->signal->live);
c30689516   Andrew Morton   [PATCH] revert "t...
912
  	if (group_dead) {
778e9a9c3   Alexey Kuznetsov   pi-futex: fix exi...
913
  		hrtimer_cancel(&tsk->signal->real_timer);
25f407f0b   Roland McGrath   [PATCH] Call exit...
914
  		exit_itimers(tsk->signal);
1f10206cf   Jiri Pirko   getrusage: fill r...
915
916
  		if (tsk->mm)
  			setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
c30689516   Andrew Morton   [PATCH] revert "t...
917
  	}
f6ec29a42   KaiGai Kohei   [PATCH] pacct: av...
918
  	acct_collect(code, group_dead);
522ed7767   Miloslav Trmac   Audit: add TTY in...
919
920
  	if (group_dead)
  		tty_audit_exit();
fa84cb935   Al Viro   [PATCH] move call...
921
922
  	if (unlikely(tsk->audit_context))
  		audit_free(tsk);
115085ea0   Oleg Nesterov   [PATCH] taskstats...
923

f2ab6d888   Jonathan Lim   Assign task_struc...
924
  	tsk->exit_code = code;
115085ea0   Oleg Nesterov   [PATCH] taskstats...
925
  	taskstats_exit(tsk, group_dead);
c757249af   Shailabh Nagar   [PATCH] per-task-...
926

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
927
  	exit_mm(tsk);
0e4648141   KaiGai Kohei   [PATCH] pacct: ad...
928
  	if (group_dead)
f6ec29a42   KaiGai Kohei   [PATCH] pacct: av...
929
  		acct_process();
0a16b6075   Mathieu Desnoyers   tracing, sched: L...
930
  	trace_sched_process_exit(tsk);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
931
  	exit_sem(tsk);
1ec7f1ddb   Al Viro   [PATCH] get rid o...
932
933
  	exit_files(tsk);
  	exit_fs(tsk);
e18eecb8b   Jeff Dike   Add generic exit-...
934
  	check_stack_usage();
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
935
  	exit_thread();
b4f48b636   Paul Menage   Task Control Grou...
936
  	cgroup_exit(tsk, 1);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
937

5ec93d115   Alan Cox   tty: Move the lea...
938
  	if (group_dead)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
939
  		disassociate_ctty(1);
a1261f546   Al Viro   [PATCH] m68k: int...
940
  	module_put(task_thread_info(tsk)->exec_domain->module);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
941

9f46080c4   Matt Helsley   [PATCH] Process E...
942
  	proc_exit_connector(tsk);
33b2fb303   Ingo Molnar   perf_counter: fix...
943
944
  
  	/*
24f1e32c6   Frederic Weisbecker   hw-breakpoints: R...
945
946
947
948
  	 * FIXME: do that only when needed, using sched_exit tracepoint
  	 */
  	flush_ptrace_hw_breakpoint(tsk);
  	/*
33b2fb303   Ingo Molnar   perf_counter: fix...
949
950
951
  	 * Flush inherited counters to the parent - before the parent
  	 * gets woken up by child-exit notifications.
  	 */
cdd6c482c   Ingo Molnar   perf: Do the big ...
952
  	perf_event_exit_task(tsk);
33b2fb303   Ingo Molnar   perf_counter: fix...
953

821c7de71   Oleg Nesterov   exit_notify: fix ...
954
  	exit_notify(tsk, group_dead);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
955
  #ifdef CONFIG_NUMA
c0ff7453b   Miao Xie   cpuset,mm: fix no...
956
  	task_lock(tsk);
f0be3d32b   Lee Schermerhorn   mempolicy: rename...
957
  	mpol_put(tsk->mempolicy);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
958
  	tsk->mempolicy = NULL;
c0ff7453b   Miao Xie   cpuset,mm: fix no...
959
  	task_unlock(tsk);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
960
  #endif
42b2dd0a0   Alexey Dobriyan   Shrink task_struc...
961
  #ifdef CONFIG_FUTEX
c87e2837b   Ingo Molnar   [PATCH] pi-futex:...
962
963
  	if (unlikely(current->pi_state_cache))
  		kfree(current->pi_state_cache);
42b2dd0a0   Alexey Dobriyan   Shrink task_struc...
964
  #endif
c87e2837b   Ingo Molnar   [PATCH] pi-futex:...
965
  	/*
9a11b49a8   Ingo Molnar   [PATCH] lockdep: ...
966
  	 * Make sure we are holding no locks:
de5097c2e   Ingo Molnar   [PATCH] mutex sub...
967
  	 */
9a11b49a8   Ingo Molnar   [PATCH] lockdep: ...
968
  	debug_check_no_locks_held(tsk);
778e9a9c3   Alexey Kuznetsov   pi-futex: fix exi...
969
970
971
972
973
974
  	/*
  	 * We can do this unlocked here. The futex code uses this flag
  	 * just to verify whether the pi state cleanup has been done
  	 * or not. In the worst case it loops once more.
  	 */
  	tsk->flags |= PF_EXITPIDONE;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
975

afc847b7d   Al Viro   [PATCH] don't do ...
976
  	if (tsk->io_context)
b69f22920   Louis Rilling   block: Fix io_con...
977
  		exit_io_context(tsk);
afc847b7d   Al Viro   [PATCH] don't do ...
978

b92ce5589   Jens Axboe   [PATCH] splice: a...
979
980
  	if (tsk->splice_pipe)
  		__free_pipe_info(tsk->splice_pipe);
e0e817392   David Howells   CRED: Add some co...
981
  	validate_creds_for_do_exit(tsk);
7407251a0   Coywolf Qi Hunt   [PATCH] PF_DEAD c...
982
  	preempt_disable();
f41d911f8   Paul E. McKenney   rcu: Merge preemp...
983
  	exit_rcu();
55a101f8f   Oleg Nesterov   [PATCH] kill PF_D...
984
  	/* causes final put_task_struct in finish_task_switch(). */
c394cc9fb   Oleg Nesterov   [PATCH] introduce...
985
  	tsk->state = TASK_DEAD;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
986
987
988
  	schedule();
  	BUG();
  	/* Avoid "noreturn function does return".  */
54306cf04   Alan Cox   [PATCH] exit: fix...
989
990
  	for (;;)
  		cpu_relax();	/* For when BUG is null */
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
991
  }
012914dad   Russ Anderson   [patch] MCA recov...
992
  EXPORT_SYMBOL_GPL(do_exit);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
993
994
995
996
  NORET_TYPE void complete_and_exit(struct completion *comp, long code)
  {
  	if (comp)
  		complete(comp);
55a101f8f   Oleg Nesterov   [PATCH] kill PF_D...
997

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
998
999
1000
1001
  	do_exit(code);
  }
  
  EXPORT_SYMBOL(complete_and_exit);
754fe8d29   Heiko Carstens   [CVE-2009-0029] S...
1002
  SYSCALL_DEFINE1(exit, int, error_code)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1003
1004
1005
  {
  	do_exit((error_code&0xff)<<8);
  }
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1006
1007
1008
1009
1010
1011
1012
  /*
   * Take down every thread in the group.  This is called by fatal signals
   * as well as by sys_exit_group (below).
   */
  NORET_TYPE void
  do_group_exit(int exit_code)
  {
bfc4b0890   Oleg Nesterov   signals: do_group...
1013
  	struct signal_struct *sig = current->signal;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1014
  	BUG_ON(exit_code & 0x80); /* core dumps don't get here */
bfc4b0890   Oleg Nesterov   signals: do_group...
1015
1016
  	if (signal_group_exit(sig))
  		exit_code = sig->group_exit_code;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1017
  	else if (!thread_group_empty(current)) {
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1018
  		struct sighand_struct *const sighand = current->sighand;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1019
  		spin_lock_irq(&sighand->siglock);
ed5d2cac1   Oleg Nesterov   exec: rework the ...
1020
  		if (signal_group_exit(sig))
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1021
1022
1023
  			/* Another thread got here before we took the lock.  */
  			exit_code = sig->group_exit_code;
  		else {
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1024
  			sig->group_exit_code = exit_code;
ed5d2cac1   Oleg Nesterov   exec: rework the ...
1025
  			sig->flags = SIGNAL_GROUP_EXIT;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1026
1027
1028
  			zap_other_threads(current);
  		}
  		spin_unlock_irq(&sighand->siglock);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
  	}
  
  	do_exit(exit_code);
  	/* NOTREACHED */
  }
  
  /*
   * this kills every thread in the thread group. Note that any externally
   * wait4()-ing process will get the correct exit code - even if this
   * thread is not the thread group leader.
   */
754fe8d29   Heiko Carstens   [CVE-2009-0029] S...
1040
  SYSCALL_DEFINE1(exit_group, int, error_code)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1041
1042
  {
  	do_group_exit((error_code & 0xff) << 8);
2ed7c03ec   Heiko Carstens   [CVE-2009-0029] C...
1043
1044
  	/* NOTREACHED */
  	return 0;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1045
  }
9e8ae01d1   Oleg Nesterov   introduce "struct...
1046
1047
  struct wait_opts {
  	enum pid_type		wo_type;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1048
  	int			wo_flags;
e1eb1ebcc   Richard Kennedy   mm: exit.c reorde...
1049
  	struct pid		*wo_pid;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1050
1051
1052
1053
  
  	struct siginfo __user	*wo_info;
  	int __user		*wo_stat;
  	struct rusage __user	*wo_rusage;
0b7570e77   Oleg Nesterov   do_wait() wakeup ...
1054
  	wait_queue_t		child_wait;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1055
1056
  	int			notask_error;
  };
989264f46   Oleg Nesterov   do_wait-wakeup-op...
1057
1058
  static inline
  struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
161550d74   Eric W. Biederman   pid: sys_wait... ...
1059
  {
989264f46   Oleg Nesterov   do_wait-wakeup-op...
1060
1061
1062
  	if (type != PIDTYPE_PID)
  		task = task->group_leader;
  	return task->pids[type].pid;
161550d74   Eric W. Biederman   pid: sys_wait... ...
1063
  }
989264f46   Oleg Nesterov   do_wait-wakeup-op...
1064
  static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1065
  {
5c01ba49e   Oleg Nesterov   do_wait-wakeup-op...
1066
1067
1068
  	return	wo->wo_type == PIDTYPE_MAX ||
  		task_pid_type(p, wo->wo_type) == wo->wo_pid;
  }
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1069

5c01ba49e   Oleg Nesterov   do_wait-wakeup-op...
1070
1071
1072
1073
  static int eligible_child(struct wait_opts *wo, struct task_struct *p)
  {
  	if (!eligible_pid(wo, p))
  		return 0;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1074
1075
1076
1077
1078
  	/* Wait for all children (clone and not) if __WALL is set;
  	 * otherwise, wait for clone children *only* if __WCLONE is
  	 * set; otherwise, wait for non-clone children *only*.  (Note:
  	 * A "clone" child here is one that reports to its parent
  	 * using a signal other than SIGCHLD.) */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1079
1080
  	if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
  	    && !(wo->wo_flags & __WALL))
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1081
  		return 0;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1082

14dd0b814   Roland McGrath   do_wait: return s...
1083
  	return 1;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1084
  }
9e8ae01d1   Oleg Nesterov   introduce "struct...
1085
1086
  static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
  				pid_t pid, uid_t uid, int why, int status)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1087
  {
9e8ae01d1   Oleg Nesterov   introduce "struct...
1088
1089
1090
  	struct siginfo __user *infop;
  	int retval = wo->wo_rusage
  		? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
36c8b5868   Ingo Molnar   [PATCH] sched: cl...
1091

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1092
  	put_task_struct(p);
9e8ae01d1   Oleg Nesterov   introduce "struct...
1093
  	infop = wo->wo_info;
b6fe2d117   Vitaly Mayatskikh   wait_noreap_copyo...
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
  	if (infop) {
  		if (!retval)
  			retval = put_user(SIGCHLD, &infop->si_signo);
  		if (!retval)
  			retval = put_user(0, &infop->si_errno);
  		if (!retval)
  			retval = put_user((short)why, &infop->si_code);
  		if (!retval)
  			retval = put_user(pid, &infop->si_pid);
  		if (!retval)
  			retval = put_user(uid, &infop->si_uid);
  		if (!retval)
  			retval = put_user(status, &infop->si_status);
  	}
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
  	if (!retval)
  		retval = pid;
  	return retval;
  }
  
  /*
   * Handle sys_wait4 work for one task in state EXIT_ZOMBIE.  We hold
   * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
   * the lock and this task is uninteresting.  If we return nonzero, we have
   * released the lock and the system call should return.
   */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1119
  static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1120
1121
  {
  	unsigned long state;
2f4e6e2a8   Oleg Nesterov   wait_task_zombie:...
1122
  	int retval, status, traced;
6c5f3e7b4   Pavel Emelyanov   Pidns: make full ...
1123
  	pid_t pid = task_pid_vnr(p);
c69e8d9c0   David Howells   CRED: Use RCU to ...
1124
  	uid_t uid = __task_cred(p)->uid;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1125
  	struct siginfo __user *infop;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1126

9e8ae01d1   Oleg Nesterov   introduce "struct...
1127
  	if (!likely(wo->wo_flags & WEXITED))
98abed020   Roland McGrath   do_wait reorganiz...
1128
  		return 0;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1129
  	if (unlikely(wo->wo_flags & WNOWAIT)) {
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1130
  		int exit_code = p->exit_code;
f3abd4f95   Thiago Farina   kernel/exit.c: fi...
1131
  		int why;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1132

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1133
1134
1135
1136
1137
1138
1139
1140
1141
  		get_task_struct(p);
  		read_unlock(&tasklist_lock);
  		if ((exit_code & 0x7f) == 0) {
  			why = CLD_EXITED;
  			status = exit_code >> 8;
  		} else {
  			why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
  			status = exit_code & 0x7f;
  		}
9e8ae01d1   Oleg Nesterov   introduce "struct...
1142
  		return wait_noreap_copyout(wo, p, pid, uid, why, status);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
  	}
  
  	/*
  	 * Try to move the task's state to DEAD
  	 * only one thread is allowed to do this:
  	 */
  	state = xchg(&p->exit_state, EXIT_DEAD);
  	if (state != EXIT_ZOMBIE) {
  		BUG_ON(state != EXIT_DEAD);
  		return 0;
  	}
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1154

53b6f9fbd   Oleg Nesterov   ptrace: introduce...
1155
  	traced = ptrace_reparented(p);
befca9677   Oleg Nesterov   ptrace: wait_task...
1156
1157
1158
1159
1160
  	/*
  	 * It can be ptraced but not reparented, check
  	 * !task_detached() to filter out sub-threads.
  	 */
  	if (likely(!traced) && likely(!task_detached(p))) {
3795e1616   Jesper Juhl   [PATCH] Decrease ...
1161
1162
  		struct signal_struct *psig;
  		struct signal_struct *sig;
1f10206cf   Jiri Pirko   getrusage: fill r...
1163
  		unsigned long maxrss;
0cf55e1ec   Hidetoshi Seto   sched, cputime: I...
1164
  		cputime_t tgutime, tgstime;
3795e1616   Jesper Juhl   [PATCH] Decrease ...
1165

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
  		/*
  		 * The resource counters for the group leader are in its
  		 * own task_struct.  Those for dead threads in the group
  		 * are in its signal_struct, as are those for the child
  		 * processes it has previously reaped.  All these
  		 * accumulate in the parent's signal_struct c* fields.
  		 *
  		 * We don't bother to take a lock here to protect these
  		 * p->signal fields, because they are only touched by
  		 * __exit_signal, which runs with tasklist_lock
  		 * write-locked anyway, and so is excluded here.  We do
d1e98f429   Oleg Nesterov   ptrace: wait_task...
1177
  		 * need to protect the access to parent->signal fields,
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1178
1179
  		 * as other threads in the parent group can be right
  		 * here reaping other children at the same time.
0cf55e1ec   Hidetoshi Seto   sched, cputime: I...
1180
1181
1182
1183
  		 *
  		 * We use thread_group_times() to get times for the thread
  		 * group, which consolidates times for all threads in the
  		 * group including the group leader.
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1184
  		 */
0cf55e1ec   Hidetoshi Seto   sched, cputime: I...
1185
  		thread_group_times(p, &tgutime, &tgstime);
d1e98f429   Oleg Nesterov   ptrace: wait_task...
1186
1187
  		spin_lock_irq(&p->real_parent->sighand->siglock);
  		psig = p->real_parent->signal;
3795e1616   Jesper Juhl   [PATCH] Decrease ...
1188
1189
1190
  		sig = p->signal;
  		psig->cutime =
  			cputime_add(psig->cutime,
0cf55e1ec   Hidetoshi Seto   sched, cputime: I...
1191
1192
  			cputime_add(tgutime,
  				    sig->cutime));
3795e1616   Jesper Juhl   [PATCH] Decrease ...
1193
1194
  		psig->cstime =
  			cputime_add(psig->cstime,
0cf55e1ec   Hidetoshi Seto   sched, cputime: I...
1195
1196
  			cputime_add(tgstime,
  				    sig->cstime));
9ac52315d   Laurent Vivier   sched: guest CPU ...
1197
1198
1199
1200
1201
  		psig->cgtime =
  			cputime_add(psig->cgtime,
  			cputime_add(p->gtime,
  			cputime_add(sig->gtime,
  				    sig->cgtime)));
3795e1616   Jesper Juhl   [PATCH] Decrease ...
1202
1203
1204
1205
1206
1207
1208
1209
  		psig->cmin_flt +=
  			p->min_flt + sig->min_flt + sig->cmin_flt;
  		psig->cmaj_flt +=
  			p->maj_flt + sig->maj_flt + sig->cmaj_flt;
  		psig->cnvcsw +=
  			p->nvcsw + sig->nvcsw + sig->cnvcsw;
  		psig->cnivcsw +=
  			p->nivcsw + sig->nivcsw + sig->cnivcsw;
6eaeeaba3   Eric Dumazet   getrusage(): fill...
1210
1211
1212
1213
1214
1215
  		psig->cinblock +=
  			task_io_get_inblock(p) +
  			sig->inblock + sig->cinblock;
  		psig->coublock +=
  			task_io_get_oublock(p) +
  			sig->oublock + sig->coublock;
1f10206cf   Jiri Pirko   getrusage: fill r...
1216
1217
1218
  		maxrss = max(sig->maxrss, sig->cmaxrss);
  		if (psig->cmaxrss < maxrss)
  			psig->cmaxrss = maxrss;
5995477ab   Andrea Righi   task IO accountin...
1219
1220
  		task_io_accounting_add(&psig->ioac, &p->ioac);
  		task_io_accounting_add(&psig->ioac, &sig->ioac);
d1e98f429   Oleg Nesterov   ptrace: wait_task...
1221
  		spin_unlock_irq(&p->real_parent->sighand->siglock);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1222
1223
1224
1225
1226
1227
1228
  	}
  
  	/*
  	 * Now we are sure this task is interesting, and no other
  	 * thread can reap it because we set its state to EXIT_DEAD.
  	 */
  	read_unlock(&tasklist_lock);
9e8ae01d1   Oleg Nesterov   introduce "struct...
1229
1230
  	retval = wo->wo_rusage
  		? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1231
1232
  	status = (p->signal->flags & SIGNAL_GROUP_EXIT)
  		? p->signal->group_exit_code : p->exit_code;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1233
1234
1235
1236
  	if (!retval && wo->wo_stat)
  		retval = put_user(status, wo->wo_stat);
  
  	infop = wo->wo_info;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
  	if (!retval && infop)
  		retval = put_user(SIGCHLD, &infop->si_signo);
  	if (!retval && infop)
  		retval = put_user(0, &infop->si_errno);
  	if (!retval && infop) {
  		int why;
  
  		if ((status & 0x7f) == 0) {
  			why = CLD_EXITED;
  			status >>= 8;
  		} else {
  			why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
  			status &= 0x7f;
  		}
  		retval = put_user((short)why, &infop->si_code);
  		if (!retval)
  			retval = put_user(status, &infop->si_status);
  	}
  	if (!retval && infop)
3a515e4a6   Oleg Nesterov   wait_task_continu...
1256
  		retval = put_user(pid, &infop->si_pid);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1257
  	if (!retval && infop)
c69e8d9c0   David Howells   CRED: Use RCU to ...
1258
  		retval = put_user(uid, &infop->si_uid);
2f4e6e2a8   Oleg Nesterov   wait_task_zombie:...
1259
  	if (!retval)
3a515e4a6   Oleg Nesterov   wait_task_continu...
1260
  		retval = pid;
2f4e6e2a8   Oleg Nesterov   wait_task_zombie:...
1261
1262
  
  	if (traced) {
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1263
  		write_lock_irq(&tasklist_lock);
2f4e6e2a8   Oleg Nesterov   wait_task_zombie:...
1264
1265
1266
1267
1268
1269
1270
  		/* We dropped tasklist, ptracer could die and untrace */
  		ptrace_unlink(p);
  		/*
  		 * If this is not a detached task, notify the parent.
  		 * If it's still not detached after that, don't release
  		 * it now.
  		 */
d839fd4d2   Oleg Nesterov   ptrace: introduce...
1271
  		if (!task_detached(p)) {
2f4e6e2a8   Oleg Nesterov   wait_task_zombie:...
1272
  			do_notify_parent(p, p->exit_signal);
d839fd4d2   Oleg Nesterov   ptrace: introduce...
1273
  			if (!task_detached(p)) {
2f4e6e2a8   Oleg Nesterov   wait_task_zombie:...
1274
1275
  				p->exit_state = EXIT_ZOMBIE;
  				p = NULL;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1276
1277
1278
1279
1280
1281
  			}
  		}
  		write_unlock_irq(&tasklist_lock);
  	}
  	if (p != NULL)
  		release_task(p);
2f4e6e2a8   Oleg Nesterov   wait_task_zombie:...
1282

1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1283
1284
  	return retval;
  }
90bc8d8b1   Oleg Nesterov   do_wait: fix wait...
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
  static int *task_stopped_code(struct task_struct *p, bool ptrace)
  {
  	if (ptrace) {
  		if (task_is_stopped_or_traced(p))
  			return &p->exit_code;
  	} else {
  		if (p->signal->flags & SIGNAL_STOP_STOPPED)
  			return &p->signal->group_exit_code;
  	}
  	return NULL;
  }
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1296
1297
1298
1299
1300
1301
  /*
   * Handle sys_wait4 work for one task in state TASK_STOPPED.  We hold
   * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
   * the lock and this task is uninteresting.  If we return nonzero, we have
   * released the lock and the system call should return.
   */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1302
1303
  static int wait_task_stopped(struct wait_opts *wo,
  				int ptrace, struct task_struct *p)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1304
  {
9e8ae01d1   Oleg Nesterov   introduce "struct...
1305
  	struct siginfo __user *infop;
90bc8d8b1   Oleg Nesterov   do_wait: fix wait...
1306
  	int retval, exit_code, *p_code, why;
ee7c82da8   Oleg Nesterov   wait_task_stopped...
1307
  	uid_t uid = 0; /* unneeded, required by compiler */
c89507835   Oleg Nesterov   wait_task_stopped...
1308
  	pid_t pid;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1309

47918025e   Oleg Nesterov   shift "ptrace imp...
1310
1311
1312
  	/*
  	 * Traditionally we see ptrace'd stopped tasks regardless of options.
  	 */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1313
  	if (!ptrace && !(wo->wo_flags & WUNTRACED))
98abed020   Roland McGrath   do_wait reorganiz...
1314
  		return 0;
ee7c82da8   Oleg Nesterov   wait_task_stopped...
1315
1316
  	exit_code = 0;
  	spin_lock_irq(&p->sighand->siglock);
90bc8d8b1   Oleg Nesterov   do_wait: fix wait...
1317
1318
  	p_code = task_stopped_code(p, ptrace);
  	if (unlikely(!p_code))
ee7c82da8   Oleg Nesterov   wait_task_stopped...
1319
  		goto unlock_sig;
90bc8d8b1   Oleg Nesterov   do_wait: fix wait...
1320
  	exit_code = *p_code;
ee7c82da8   Oleg Nesterov   wait_task_stopped...
1321
1322
  	if (!exit_code)
  		goto unlock_sig;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1323
  	if (!unlikely(wo->wo_flags & WNOWAIT))
90bc8d8b1   Oleg Nesterov   do_wait: fix wait...
1324
  		*p_code = 0;
ee7c82da8   Oleg Nesterov   wait_task_stopped...
1325

f362b7324   Daniel J Blueman   Fix unprotected a...
1326
  	uid = task_uid(p);
ee7c82da8   Oleg Nesterov   wait_task_stopped...
1327
1328
1329
  unlock_sig:
  	spin_unlock_irq(&p->sighand->siglock);
  	if (!exit_code)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
  		return 0;
  
  	/*
  	 * Now we are pretty sure this task is interesting.
  	 * Make sure it doesn't get reaped out from under us while we
  	 * give up the lock and then examine it below.  We don't want to
  	 * keep holding onto the tasklist_lock while we call getrusage and
  	 * possibly take page faults for user memory.
  	 */
  	get_task_struct(p);
6c5f3e7b4   Pavel Emelyanov   Pidns: make full ...
1340
  	pid = task_pid_vnr(p);
f470021ad   Roland McGrath   ptrace children r...
1341
  	why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1342
  	read_unlock(&tasklist_lock);
9e8ae01d1   Oleg Nesterov   introduce "struct...
1343
1344
1345
1346
1347
1348
1349
  	if (unlikely(wo->wo_flags & WNOWAIT))
  		return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
  
  	retval = wo->wo_rusage
  		? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
  	if (!retval && wo->wo_stat)
  		retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1350

9e8ae01d1   Oleg Nesterov   introduce "struct...
1351
  	infop = wo->wo_info;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1352
1353
1354
1355
1356
  	if (!retval && infop)
  		retval = put_user(SIGCHLD, &infop->si_signo);
  	if (!retval && infop)
  		retval = put_user(0, &infop->si_errno);
  	if (!retval && infop)
6efcae460   Roland McGrath   Fix waitid si_cod...
1357
  		retval = put_user((short)why, &infop->si_code);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1358
1359
1360
  	if (!retval && infop)
  		retval = put_user(exit_code, &infop->si_status);
  	if (!retval && infop)
c89507835   Oleg Nesterov   wait_task_stopped...
1361
  		retval = put_user(pid, &infop->si_pid);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1362
  	if (!retval && infop)
ee7c82da8   Oleg Nesterov   wait_task_stopped...
1363
  		retval = put_user(uid, &infop->si_uid);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1364
  	if (!retval)
c89507835   Oleg Nesterov   wait_task_stopped...
1365
  		retval = pid;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
  	put_task_struct(p);
  
  	BUG_ON(!retval);
  	return retval;
  }
  
  /*
   * Handle do_wait work for one task in a live, non-stopped state.
   * read_lock(&tasklist_lock) on entry.  If we return zero, we still hold
   * the lock and this task is uninteresting.  If we return nonzero, we have
   * released the lock and the system call should return.
   */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1378
  static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1379
1380
1381
1382
  {
  	int retval;
  	pid_t pid;
  	uid_t uid;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1383
  	if (!unlikely(wo->wo_flags & WCONTINUED))
98abed020   Roland McGrath   do_wait reorganiz...
1384
  		return 0;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1385
1386
1387
1388
1389
1390
1391
1392
1393
  	if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
  		return 0;
  
  	spin_lock_irq(&p->sighand->siglock);
  	/* Re-check with the lock held.  */
  	if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
  		spin_unlock_irq(&p->sighand->siglock);
  		return 0;
  	}
9e8ae01d1   Oleg Nesterov   introduce "struct...
1394
  	if (!unlikely(wo->wo_flags & WNOWAIT))
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1395
  		p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
f362b7324   Daniel J Blueman   Fix unprotected a...
1396
  	uid = task_uid(p);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1397
  	spin_unlock_irq(&p->sighand->siglock);
6c5f3e7b4   Pavel Emelyanov   Pidns: make full ...
1398
  	pid = task_pid_vnr(p);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1399
1400
  	get_task_struct(p);
  	read_unlock(&tasklist_lock);
9e8ae01d1   Oleg Nesterov   introduce "struct...
1401
1402
1403
  	if (!wo->wo_info) {
  		retval = wo->wo_rusage
  			? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1404
  		put_task_struct(p);
9e8ae01d1   Oleg Nesterov   introduce "struct...
1405
1406
  		if (!retval && wo->wo_stat)
  			retval = put_user(0xffff, wo->wo_stat);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1407
  		if (!retval)
3a515e4a6   Oleg Nesterov   wait_task_continu...
1408
  			retval = pid;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1409
  	} else {
9e8ae01d1   Oleg Nesterov   introduce "struct...
1410
1411
  		retval = wait_noreap_copyout(wo, p, pid, uid,
  					     CLD_CONTINUED, SIGCONT);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1412
1413
1414
1415
1416
  		BUG_ON(retval == 0);
  	}
  
  	return retval;
  }
98abed020   Roland McGrath   do_wait reorganiz...
1417
1418
1419
  /*
   * Consider @p for a wait by @parent.
   *
9e8ae01d1   Oleg Nesterov   introduce "struct...
1420
   * -ECHILD should be in ->notask_error before the first call.
98abed020   Roland McGrath   do_wait reorganiz...
1421
1422
   * Returns nonzero for a final return, when we have unlocked tasklist_lock.
   * Returns zero if the search for a child should continue;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1423
   * then ->notask_error is 0 if @p is an eligible child,
14dd0b814   Roland McGrath   do_wait: return s...
1424
   * or another error from security_task_wait(), or still -ECHILD.
98abed020   Roland McGrath   do_wait reorganiz...
1425
   */
b6e763f07   Oleg Nesterov   wait_consider_tas...
1426
1427
  static int wait_consider_task(struct wait_opts *wo, int ptrace,
  				struct task_struct *p)
98abed020   Roland McGrath   do_wait reorganiz...
1428
  {
9e8ae01d1   Oleg Nesterov   introduce "struct...
1429
  	int ret = eligible_child(wo, p);
14dd0b814   Roland McGrath   do_wait: return s...
1430
  	if (!ret)
98abed020   Roland McGrath   do_wait reorganiz...
1431
  		return ret;
a2322e1d2   Oleg Nesterov   do_wait() wakeup ...
1432
  	ret = security_task_wait(p);
14dd0b814   Roland McGrath   do_wait: return s...
1433
1434
1435
1436
1437
1438
1439
1440
  	if (unlikely(ret < 0)) {
  		/*
  		 * If we have not yet seen any eligible child,
  		 * then let this error code replace -ECHILD.
  		 * A permission error will give the user a clue
  		 * to look for security policy problems, rather
  		 * than for mysterious wait bugs.
  		 */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1441
1442
  		if (wo->notask_error)
  			wo->notask_error = ret;
78a3d9d56   Oleg Nesterov   do_wait: do take ...
1443
  		return 0;
14dd0b814   Roland McGrath   do_wait: return s...
1444
  	}
5cb114468   Oleg Nesterov   ptrace: do not us...
1445
  	if (likely(!ptrace) && unlikely(task_ptrace(p))) {
f470021ad   Roland McGrath   ptrace children r...
1446
1447
1448
1449
  		/*
  		 * This child is hidden by ptrace.
  		 * We aren't allowed to see it now, but eventually we will.
  		 */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1450
  		wo->notask_error = 0;
f470021ad   Roland McGrath   ptrace children r...
1451
1452
  		return 0;
  	}
98abed020   Roland McGrath   do_wait reorganiz...
1453
1454
1455
1456
1457
1458
1459
  	if (p->exit_state == EXIT_DEAD)
  		return 0;
  
  	/*
  	 * We don't reap group leaders with subthreads.
  	 */
  	if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
9e8ae01d1   Oleg Nesterov   introduce "struct...
1460
  		return wait_task_zombie(wo, p);
98abed020   Roland McGrath   do_wait reorganiz...
1461
1462
1463
1464
1465
  
  	/*
  	 * It's stopped or running now, so it might
  	 * later continue, exit, or stop again.
  	 */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1466
  	wo->notask_error = 0;
98abed020   Roland McGrath   do_wait reorganiz...
1467

90bc8d8b1   Oleg Nesterov   do_wait: fix wait...
1468
  	if (task_stopped_code(p, ptrace))
9e8ae01d1   Oleg Nesterov   introduce "struct...
1469
  		return wait_task_stopped(wo, ptrace, p);
98abed020   Roland McGrath   do_wait reorganiz...
1470

9e8ae01d1   Oleg Nesterov   introduce "struct...
1471
  	return wait_task_continued(wo, p);
98abed020   Roland McGrath   do_wait reorganiz...
1472
1473
1474
1475
1476
  }
  
  /*
   * Do the work of do_wait() for one thread in the group, @tsk.
   *
9e8ae01d1   Oleg Nesterov   introduce "struct...
1477
   * -ECHILD should be in ->notask_error before the first call.
98abed020   Roland McGrath   do_wait reorganiz...
1478
1479
   * Returns nonzero for a final return, when we have unlocked tasklist_lock.
   * Returns zero if the search for a child should continue; then
9e8ae01d1   Oleg Nesterov   introduce "struct...
1480
   * ->notask_error is 0 if there were any eligible children,
14dd0b814   Roland McGrath   do_wait: return s...
1481
   * or another error from security_task_wait(), or still -ECHILD.
98abed020   Roland McGrath   do_wait reorganiz...
1482
   */
9e8ae01d1   Oleg Nesterov   introduce "struct...
1483
  static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
98abed020   Roland McGrath   do_wait reorganiz...
1484
1485
1486
1487
  {
  	struct task_struct *p;
  
  	list_for_each_entry(p, &tsk->children, sibling) {
9cd80bbb0   Oleg Nesterov   do_wait() optimiz...
1488
1489
1490
  		int ret = wait_consider_task(wo, 0, p);
  		if (ret)
  			return ret;
98abed020   Roland McGrath   do_wait reorganiz...
1491
1492
1493
1494
  	}
  
  	return 0;
  }
9e8ae01d1   Oleg Nesterov   introduce "struct...
1495
  static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
98abed020   Roland McGrath   do_wait reorganiz...
1496
1497
  {
  	struct task_struct *p;
f470021ad   Roland McGrath   ptrace children r...
1498
  	list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
b6e763f07   Oleg Nesterov   wait_consider_tas...
1499
  		int ret = wait_consider_task(wo, 1, p);
f470021ad   Roland McGrath   ptrace children r...
1500
  		if (ret)
98abed020   Roland McGrath   do_wait reorganiz...
1501
  			return ret;
98abed020   Roland McGrath   do_wait reorganiz...
1502
1503
1504
1505
  	}
  
  	return 0;
  }
0b7570e77   Oleg Nesterov   do_wait() wakeup ...
1506
1507
1508
1509
1510
1511
  static int child_wait_callback(wait_queue_t *wait, unsigned mode,
  				int sync, void *key)
  {
  	struct wait_opts *wo = container_of(wait, struct wait_opts,
  						child_wait);
  	struct task_struct *p = key;
5c01ba49e   Oleg Nesterov   do_wait-wakeup-op...
1512
  	if (!eligible_pid(wo, p))
0b7570e77   Oleg Nesterov   do_wait() wakeup ...
1513
  		return 0;
b4fe51823   Oleg Nesterov   do_wait() wakeup ...
1514
1515
  	if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
  		return 0;
0b7570e77   Oleg Nesterov   do_wait() wakeup ...
1516
1517
  	return default_wake_function(wait, mode, sync, key);
  }
a7f0765ed   Oleg Nesterov   ptrace: __ptrace_...
1518
1519
  void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
  {
0b7570e77   Oleg Nesterov   do_wait() wakeup ...
1520
1521
  	__wake_up_sync_key(&parent->signal->wait_chldexit,
  				TASK_INTERRUPTIBLE, 1, p);
a7f0765ed   Oleg Nesterov   ptrace: __ptrace_...
1522
  }
9e8ae01d1   Oleg Nesterov   introduce "struct...
1523
  static long do_wait(struct wait_opts *wo)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1524
  {
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1525
  	struct task_struct *tsk;
98abed020   Roland McGrath   do_wait reorganiz...
1526
  	int retval;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1527

9e8ae01d1   Oleg Nesterov   introduce "struct...
1528
  	trace_sched_process_wait(wo->wo_pid);
0a16b6075   Mathieu Desnoyers   tracing, sched: L...
1529

0b7570e77   Oleg Nesterov   do_wait() wakeup ...
1530
1531
1532
  	init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
  	wo->child_wait.private = current;
  	add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1533
  repeat:
98abed020   Roland McGrath   do_wait reorganiz...
1534
1535
  	/*
  	 * If there is nothing that can match our critiera just get out.
9e8ae01d1   Oleg Nesterov   introduce "struct...
1536
1537
1538
  	 * We will clear ->notask_error to zero if we see any child that
  	 * might later match our criteria, even if we are not able to reap
  	 * it yet.
98abed020   Roland McGrath   do_wait reorganiz...
1539
  	 */
64a16caf5   Oleg Nesterov   do_wait: simplify...
1540
  	wo->notask_error = -ECHILD;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1541
1542
  	if ((wo->wo_type < PIDTYPE_MAX) &&
  	   (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
64a16caf5   Oleg Nesterov   do_wait: simplify...
1543
  		goto notask;
161550d74   Eric W. Biederman   pid: sys_wait... ...
1544

f95d39d10   Oleg Nesterov   do_wait: fix the ...
1545
  	set_current_state(TASK_INTERRUPTIBLE);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1546
1547
1548
  	read_lock(&tasklist_lock);
  	tsk = current;
  	do {
64a16caf5   Oleg Nesterov   do_wait: simplify...
1549
1550
1551
  		retval = do_wait_thread(wo, tsk);
  		if (retval)
  			goto end;
9e8ae01d1   Oleg Nesterov   introduce "struct...
1552

64a16caf5   Oleg Nesterov   do_wait: simplify...
1553
1554
  		retval = ptrace_do_wait(wo, tsk);
  		if (retval)
98abed020   Roland McGrath   do_wait reorganiz...
1555
  			goto end;
98abed020   Roland McGrath   do_wait reorganiz...
1556

9e8ae01d1   Oleg Nesterov   introduce "struct...
1557
  		if (wo->wo_flags & __WNOTHREAD)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1558
  			break;
a3f6dfb72   Oleg Nesterov   do_wait: kill the...
1559
  	} while_each_thread(current, tsk);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1560
  	read_unlock(&tasklist_lock);
f2cc3eb13   Oleg Nesterov   do_wait: fix secu...
1561

64a16caf5   Oleg Nesterov   do_wait: simplify...
1562
  notask:
9e8ae01d1   Oleg Nesterov   introduce "struct...
1563
1564
  	retval = wo->notask_error;
  	if (!retval && !(wo->wo_flags & WNOHANG)) {
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1565
  		retval = -ERESTARTSYS;
98abed020   Roland McGrath   do_wait reorganiz...
1566
1567
1568
1569
  		if (!signal_pending(current)) {
  			schedule();
  			goto repeat;
  		}
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1570
  	}
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1571
  end:
f95d39d10   Oleg Nesterov   do_wait: fix the ...
1572
  	__set_current_state(TASK_RUNNING);
0b7570e77   Oleg Nesterov   do_wait() wakeup ...
1573
  	remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1574
1575
  	return retval;
  }
17da2bd90   Heiko Carstens   [CVE-2009-0029] S...
1576
1577
  SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
  		infop, int, options, struct rusage __user *, ru)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1578
  {
9e8ae01d1   Oleg Nesterov   introduce "struct...
1579
  	struct wait_opts wo;
161550d74   Eric W. Biederman   pid: sys_wait... ...
1580
1581
  	struct pid *pid = NULL;
  	enum pid_type type;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1582
1583
1584
1585
1586
1587
1588
1589
1590
  	long ret;
  
  	if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
  		return -EINVAL;
  	if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
  		return -EINVAL;
  
  	switch (which) {
  	case P_ALL:
161550d74   Eric W. Biederman   pid: sys_wait... ...
1591
  		type = PIDTYPE_MAX;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1592
1593
  		break;
  	case P_PID:
161550d74   Eric W. Biederman   pid: sys_wait... ...
1594
1595
  		type = PIDTYPE_PID;
  		if (upid <= 0)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1596
1597
1598
  			return -EINVAL;
  		break;
  	case P_PGID:
161550d74   Eric W. Biederman   pid: sys_wait... ...
1599
1600
  		type = PIDTYPE_PGID;
  		if (upid <= 0)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1601
  			return -EINVAL;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1602
1603
1604
1605
  		break;
  	default:
  		return -EINVAL;
  	}
161550d74   Eric W. Biederman   pid: sys_wait... ...
1606
1607
  	if (type < PIDTYPE_MAX)
  		pid = find_get_pid(upid);
9e8ae01d1   Oleg Nesterov   introduce "struct...
1608
1609
1610
1611
1612
1613
1614
1615
  
  	wo.wo_type	= type;
  	wo.wo_pid	= pid;
  	wo.wo_flags	= options;
  	wo.wo_info	= infop;
  	wo.wo_stat	= NULL;
  	wo.wo_rusage	= ru;
  	ret = do_wait(&wo);
dfe16dfa4   Vitaly Mayatskikh   do_wait: fix sys_...
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
  
  	if (ret > 0) {
  		ret = 0;
  	} else if (infop) {
  		/*
  		 * For a WNOHANG return, clear out all the fields
  		 * we would set so the user can easily tell the
  		 * difference.
  		 */
  		if (!ret)
  			ret = put_user(0, &infop->si_signo);
  		if (!ret)
  			ret = put_user(0, &infop->si_errno);
  		if (!ret)
  			ret = put_user(0, &infop->si_code);
  		if (!ret)
  			ret = put_user(0, &infop->si_pid);
  		if (!ret)
  			ret = put_user(0, &infop->si_uid);
  		if (!ret)
  			ret = put_user(0, &infop->si_status);
  	}
161550d74   Eric W. Biederman   pid: sys_wait... ...
1638
  	put_pid(pid);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1639
1640
  
  	/* avoid REGPARM breakage on x86: */
54a015104   Roland McGrath   asmlinkage_protec...
1641
  	asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1642
1643
  	return ret;
  }
754fe8d29   Heiko Carstens   [CVE-2009-0029] S...
1644
1645
  SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
  		int, options, struct rusage __user *, ru)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1646
  {
9e8ae01d1   Oleg Nesterov   introduce "struct...
1647
  	struct wait_opts wo;
161550d74   Eric W. Biederman   pid: sys_wait... ...
1648
1649
  	struct pid *pid = NULL;
  	enum pid_type type;
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1650
1651
1652
1653
1654
  	long ret;
  
  	if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
  			__WNOTHREAD|__WCLONE|__WALL))
  		return -EINVAL;
161550d74   Eric W. Biederman   pid: sys_wait... ...
1655
1656
1657
1658
1659
1660
1661
1662
  
  	if (upid == -1)
  		type = PIDTYPE_MAX;
  	else if (upid < 0) {
  		type = PIDTYPE_PGID;
  		pid = find_get_pid(-upid);
  	} else if (upid == 0) {
  		type = PIDTYPE_PGID;
2ae448efc   Oleg Nesterov   pids: improve get...
1663
  		pid = get_task_pid(current, PIDTYPE_PGID);
161550d74   Eric W. Biederman   pid: sys_wait... ...
1664
1665
1666
1667
  	} else /* upid > 0 */ {
  		type = PIDTYPE_PID;
  		pid = find_get_pid(upid);
  	}
9e8ae01d1   Oleg Nesterov   introduce "struct...
1668
1669
1670
1671
1672
1673
1674
  	wo.wo_type	= type;
  	wo.wo_pid	= pid;
  	wo.wo_flags	= options | WEXITED;
  	wo.wo_info	= NULL;
  	wo.wo_stat	= stat_addr;
  	wo.wo_rusage	= ru;
  	ret = do_wait(&wo);
161550d74   Eric W. Biederman   pid: sys_wait... ...
1675
  	put_pid(pid);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1676
1677
  
  	/* avoid REGPARM breakage on x86: */
54a015104   Roland McGrath   asmlinkage_protec...
1678
  	asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1679
1680
1681
1682
1683
1684
1685
1686
1687
  	return ret;
  }
  
  #ifdef __ARCH_WANT_SYS_WAITPID
  
  /*
   * sys_waitpid() remains for compatibility. waitpid() should be
   * implemented by calling sys_wait4() from libc.a.
   */
17da2bd90   Heiko Carstens   [CVE-2009-0029] S...
1688
  SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1da177e4c   Linus Torvalds   Linux-2.6.12-rc2
1689
1690
1691
1692
1693
  {
  	return sys_wait4(pid, stat_addr, options, NULL);
  }
  
  #endif