blob: 2d8be7ebb0f73499f894a1828fd827f0217290f1 [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>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080015#include <linux/mnt_namespace.h>
Jens Axboeda9cbc82008-06-30 20:42:08 +020016#include <linux/iocontext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/key.h>
18#include <linux/security.h>
19#include <linux/cpu.h>
20#include <linux/acct.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070021#include <linux/tsacct_kern.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040023#include <linux/fdtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/binfmts.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070025#include <linux/nsproxy.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080026#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/ptrace.h>
28#include <linux/profile.h>
29#include <linux/mount.h>
30#include <linux/proc_fs.h>
Eric W. Biederman49d769d2007-05-09 02:34:33 -070031#include <linux/kthread.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <linux/mempolicy.h>
Shailabh Nagarc7572492006-07-14 00:24:40 -070033#include <linux/taskstats_kern.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070034#include <linux/delayacct.h>
Rafael J. Wysocki83144182007-07-17 04:03:35 -070035#include <linux/freezer.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070036#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <linux/syscalls.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070038#include <linux/signal.h>
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080039#include <linux/posix-timers.h>
Matt Helsley9f460802005-11-07 00:59:16 -080040#include <linux/cn_proc.h>
Ingo Molnarde5097c2006-01-09 15:59:21 -080041#include <linux/mutex.h>
Ingo Molnar0771dfe2006-03-27 01:16:22 -080042#include <linux/futex.h>
Jens Axboeb92ce552006-04-11 13:52:07 +020043#include <linux/pipe_fs_i.h>
Al Virofa84cb92006-03-29 20:30:19 -050044#include <linux/audit.h> /* for audit_free() */
Adrian Bunk83cc5ed2006-06-25 05:47:41 -070045#include <linux/resource.h>
David Howells0d67a462006-08-29 19:05:56 +010046#include <linux/blkdev.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070047#include <linux/task_io_accounting_ops.h>
Roland McGrath30199f52008-07-25 19:45:46 -070048#include <linux/tracehook.h>
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -040049#include <trace/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
51#include <asm/uaccess.h>
52#include <asm/unistd.h>
53#include <asm/pgtable.h>
54#include <asm/mmu_context.h>
55
Adrian Bunk408b6642005-05-01 08:59:29 -070056static void exit_mm(struct task_struct * tsk);
57
Oleg Nesterovd839fd42008-04-30 00:53:11 -070058static inline int task_detached(struct task_struct *p)
59{
60 return p->exit_signal == -1;
61}
62
Linus Torvalds1da177e2005-04-16 15:20:36 -070063static void __unhash_process(struct task_struct *p)
64{
65 nr_threads--;
66 detach_pid(p, PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 if (thread_group_leader(p)) {
68 detach_pid(p, PIDTYPE_PGID);
69 detach_pid(p, PIDTYPE_SID);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080070
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -070071 list_del_rcu(&p->tasks);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -080072 __get_cpu_var(process_counts)--;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 }
Oleg Nesterov47e65322006-03-28 16:11:25 -080074 list_del_rcu(&p->thread_group);
Roland McGrathf4700212008-03-24 18:36:23 -070075 list_del_init(&p->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -070076}
77
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080078/*
79 * This function expects the tasklist_lock write-locked.
80 */
81static void __exit_signal(struct task_struct *tsk)
82{
83 struct signal_struct *sig = tsk->signal;
84 struct sighand_struct *sighand;
85
86 BUG_ON(!sig);
87 BUG_ON(!atomic_read(&sig->count));
88
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080089 sighand = rcu_dereference(tsk->sighand);
90 spin_lock(&sighand->siglock);
91
92 posix_cpu_timers_exit(tsk);
93 if (atomic_dec_and_test(&sig->count))
94 posix_cpu_timers_exit_group(tsk);
95 else {
96 /*
97 * If there is any task waiting for the group exit
98 * then notify it:
99 */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700100 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count)
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800101 wake_up_process(sig->group_exit_task);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700102
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800103 if (tsk == sig->curr_target)
104 sig->curr_target = next_thread(tsk);
105 /*
106 * Accumulate here the counters for all threads but the
107 * group leader as they die, so they can be added into
108 * the process-wide totals when those are taken.
109 * The group leader stays around as a zombie as long
110 * as there are other threads. When it gets reaped,
111 * the exit.c code will add its counts into these totals.
112 * We won't ever get here for the group leader, since it
113 * will have been the last reference on the signal_struct.
114 */
Balbir Singh49048622008-09-05 18:12:23 +0200115 sig->gtime = cputime_add(sig->gtime, task_gtime(tsk));
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800116 sig->min_flt += tsk->min_flt;
117 sig->maj_flt += tsk->maj_flt;
118 sig->nvcsw += tsk->nvcsw;
119 sig->nivcsw += tsk->nivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700120 sig->inblock += task_io_get_inblock(tsk);
121 sig->oublock += task_io_get_oublock(tsk);
Andrea Righi59954772008-07-27 17:29:15 +0200122 task_io_accounting_add(&sig->ioac, &tsk->ioac);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800123 sig = NULL; /* Marker for below. */
124 }
125
Oleg Nesterov58767002006-03-28 16:11:20 -0800126 __unhash_process(tsk);
127
Oleg Nesterovda7978b2008-05-23 13:04:41 -0700128 /*
129 * Do this under ->siglock, we can race with another thread
130 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
131 */
132 flush_sigqueue(&tsk->pending);
133
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800134 tsk->signal = NULL;
Oleg Nesterova7e53282006-03-28 16:11:27 -0800135 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800136 spin_unlock(&sighand->siglock);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800137
Oleg Nesterova7e53282006-03-28 16:11:27 -0800138 __cleanup_sighand(sighand);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800139 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800140 if (sig) {
141 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov093a8e82006-10-28 10:38:51 -0700142 taskstats_tgid_free(sig);
Oleg Nesterovad474ca2008-11-10 15:39:30 +0100143 /*
144 * Make sure ->signal can't go away under rq->lock,
145 * see account_group_exec_runtime().
146 */
147 task_rq_unlock_wait(tsk);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800148 __cleanup_signal(sig);
149 }
150}
151
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800152static void delayed_put_task_struct(struct rcu_head *rhp)
153{
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400154 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
155
156 trace_sched_process_free(tsk);
157 put_task_struct(tsk);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800158}
159
Roland McGrathf4700212008-03-24 18:36:23 -0700160
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161void release_task(struct task_struct * p)
162{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700163 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800165repeat:
Roland McGrathdae33572008-07-25 19:45:48 -0700166 tracehook_prepare_release_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 atomic_dec(&p->user->processes);
Pavel Emelyanov60347f62007-10-18 23:40:03 -0700168 proc_flush_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 write_lock_irq(&tasklist_lock);
Roland McGrathdae33572008-07-25 19:45:48 -0700170 tracehook_finish_release_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 __exit_signal(p);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -0800172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 /*
174 * If we are the last non-leader member of the thread
175 * group, and the leader is zombie, then notify the
176 * group leader's parent process. (if it wants notification.)
177 */
178 zap_leader = 0;
179 leader = p->group_leader;
180 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700181 BUG_ON(task_detached(leader));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 do_notify_parent(leader, leader->exit_signal);
183 /*
184 * If we were the last child thread and the leader has
185 * exited already, and the leader's parent ignores SIGCHLD,
186 * then we are the one who should release the leader.
187 *
188 * do_notify_parent() will have marked it self-reaping in
189 * that case.
190 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700191 zap_leader = task_detached(leader);
Roland McGrathdae33572008-07-25 19:45:48 -0700192
193 /*
194 * This maintains the invariant that release_task()
195 * only runs on a task in EXIT_DEAD, just for sanity.
196 */
197 if (zap_leader)
198 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 }
200
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 release_thread(p);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800203 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
205 p = leader;
206 if (unlikely(zap_leader))
207 goto repeat;
208}
209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210/*
211 * This checks not only the pgrp, but falls back on the pid if no
212 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
213 * without this...
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800214 *
215 * The caller must hold rcu lock or the tasklist lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 */
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800217struct pid *session_of_pgrp(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218{
219 struct task_struct *p;
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800220 struct pid *sid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800222 p = pid_task(pgrp, PIDTYPE_PGID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800223 if (p == NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800224 p = pid_task(pgrp, PIDTYPE_PID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800225 if (p != NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800226 sid = task_session(p);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800227
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 return sid;
229}
230
231/*
232 * Determine if a process group is "orphaned", according to the POSIX
233 * definition in 2.2.2.52. Orphaned process groups are not to be affected
234 * by terminal-generated stop signals. Newly orphaned process groups are
235 * to receive a SIGHUP and a SIGCONT.
236 *
237 * "I ask you, have you ever known what it is to be an orphan?"
238 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800239static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240{
241 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800243 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300244 if ((p == ignored_task) ||
245 (p->exit_state && thread_group_empty(p)) ||
246 is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 continue;
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300248
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800249 if (task_pgrp(p->real_parent) != pgrp &&
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300250 task_session(p->real_parent) == task_session(p))
251 return 0;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800252 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300253
254 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255}
256
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800257int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258{
259 int retval;
260
261 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800262 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 read_unlock(&tasklist_lock);
264
265 return retval;
266}
267
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800268static int has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269{
270 int retval = 0;
271 struct task_struct *p;
272
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800273 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Matthew Wilcox338077e2007-12-06 11:09:35 -0500274 if (!task_is_stopped(p))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 retval = 1;
277 break;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800278 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 return retval;
280}
281
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300282/*
283 * Check to see if any process groups have become orphaned as
284 * a result of our exiting, and if they have any stopped jobs,
285 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
286 */
287static void
288kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
289{
290 struct pid *pgrp = task_pgrp(tsk);
291 struct task_struct *ignored_task = tsk;
292
293 if (!parent)
294 /* exit: our father is in a different pgrp than
295 * we are and we were the only connection outside.
296 */
297 parent = tsk->real_parent;
298 else
299 /* reparent: our child is in a different pgrp than
300 * we are, and it was the only connection outside.
301 */
302 ignored_task = NULL;
303
304 if (task_pgrp(parent) != pgrp &&
305 task_session(parent) == task_session(tsk) &&
306 will_become_orphaned_pgrp(pgrp, ignored_task) &&
307 has_stopped_jobs(pgrp)) {
308 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
309 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
310 }
311}
312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313/**
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700314 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 *
316 * If a kernel thread is launched as a result of a system call, or if
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700317 * it ever exits, it should generally reparent itself to kthreadd so it
318 * isn't in the way of other processes and is correctly cleaned up on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 *
320 * The various task state such as scheduling policy and priority may have
321 * been inherited from a user process, so we reset them to sane values here.
322 *
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700323 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700325static void reparent_to_kthreadd(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326{
327 write_lock_irq(&tasklist_lock);
328
329 ptrace_unlink(current);
330 /* Reparent to init */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700331 current->real_parent = current->parent = kthreadd_task;
Roland McGrathf4700212008-03-24 18:36:23 -0700332 list_move_tail(&current->sibling, &current->real_parent->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
334 /* Set the exit signal to SIGCHLD so we signal init on exit */
335 current->exit_signal = SIGCHLD;
336
Ingo Molnare05606d2007-07-09 18:51:59 +0200337 if (task_nice(current) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 set_user_nice(current, 0);
339 /* cpus_allowed? */
340 /* rt_priority? */
341 /* signals? */
342 security_task_reparent_to_init(current);
343 memcpy(current->signal->rlim, init_task.signal->rlim,
344 sizeof(current->signal->rlim));
345 atomic_inc(&(INIT_USER->__count));
346 write_unlock_irq(&tasklist_lock);
347 switch_uid(INIT_USER);
348}
349
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800350void __set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351{
Oren Laadane19f2472006-01-08 01:03:58 -0800352 struct task_struct *curr = current->group_leader;
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800353 pid_t nr = pid_nr(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800355 if (task_session(curr) != pid) {
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700356 change_pid(curr, PIDTYPE_SID, pid);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800357 set_task_session(curr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 }
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800359 if (task_pgrp(curr) != pid) {
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700360 change_pid(curr, PIDTYPE_PGID, pid);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800361 set_task_pgrp(curr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 }
363}
364
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800365static void set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366{
367 write_lock_irq(&tasklist_lock);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800368 __set_special_pids(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 write_unlock_irq(&tasklist_lock);
370}
371
372/*
373 * Let kernel threads use this to say that they
374 * allow a certain signal (since daemonize() will
375 * have disabled all of them by default).
376 */
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);
383 sigdelset(&current->blocked, sig);
384 if (!current->mm) {
385 /* Kernel threads handle their own signals.
386 Let the signal code know it'll be handled, so
387 that they don't get converted to SIGKILL or
388 just silently dropped */
389 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
390 }
391 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;
420 struct fs_struct *fs;
421 sigset_t blocked;
422
423 va_start(args, name);
424 vsnprintf(current->comm, sizeof(current->comm), name, args);
425 va_end(args);
426
427 /*
428 * If we were started as result of loading a module, close all of the
429 * user space pages. We don't need them, and if we didn't close them
430 * they would be locked into memory.
431 */
432 exit_mm(current);
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700433 /*
434 * We don't want to have TIF_FREEZE set if the system-wide hibernation
435 * or suspend transition begins right now.
436 */
Oleg Nesterov7b34e422008-07-25 01:47:37 -0700437 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800439 if (current->nsproxy != &init_nsproxy) {
440 get_nsproxy(&init_nsproxy);
441 switch_task_namespaces(current, &init_nsproxy);
442 }
Oleg Nesterov297bd422008-02-08 04:19:10 -0800443 set_special_pids(&init_struct_pid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -0800444 proc_clear_tty(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
446 /* Block and flush all signals */
447 sigfillset(&blocked);
448 sigprocmask(SIG_BLOCK, &blocked, NULL);
449 flush_signals(current);
450
451 /* Become as one with the init task */
452
453 exit_fs(current); /* current->fs->count--; */
454 fs = init_task.fs;
455 current->fs = fs;
456 atomic_inc(&fs->count);
Serge E. Hallynab516012006-10-02 02:18:06 -0700457
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700458 exit_files(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 current->files = init_task.files;
460 atomic_inc(&current->files->count);
461
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700462 reparent_to_kthreadd();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463}
464
465EXPORT_SYMBOL(daemonize);
466
Arjan van de Ven858119e2006-01-14 13:20:43 -0800467static void close_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468{
469 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700470 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
472 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700473
474 /*
475 * It is safe to dereference the fd table without RCU or
476 * ->file_lock because this is the last reference to the
477 * files structure.
478 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700479 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 for (;;) {
481 unsigned long set;
482 i = j * __NFDBITS;
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800483 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700485 set = fdt->open_fds->fds_bits[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 while (set) {
487 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700488 struct file * file = xchg(&fdt->fd[i], NULL);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800489 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 filp_close(file, files);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800491 cond_resched();
492 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 }
494 i++;
495 set >>= 1;
496 }
497 }
498}
499
500struct files_struct *get_files_struct(struct task_struct *task)
501{
502 struct files_struct *files;
503
504 task_lock(task);
505 files = task->files;
506 if (files)
507 atomic_inc(&files->count);
508 task_unlock(task);
509
510 return files;
511}
512
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800513void put_files_struct(struct files_struct *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700515 struct fdtable *fdt;
516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 if (atomic_dec_and_test(&files->count)) {
518 close_files(files);
519 /*
520 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700521 * If the fdtable was embedded, pass files for freeing
522 * at the end of the RCU grace period. Otherwise,
523 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700525 fdt = files_fdtable(files);
Vadim Lobanov4fd45812006-12-10 02:21:17 -0800526 if (fdt != &files->fdtab)
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700527 kmem_cache_free(files_cachep, files);
Vadim Lobanov01b2d932006-12-22 01:10:43 -0800528 free_fdtable(fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 }
530}
531
Al Viro3b125382008-04-22 05:31:30 -0400532void reset_files_struct(struct files_struct *files)
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700533{
Al Viro3b125382008-04-22 05:31:30 -0400534 struct task_struct *tsk = current;
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700535 struct files_struct *old;
536
537 old = tsk->files;
538 task_lock(tsk);
539 tsk->files = files;
540 task_unlock(tsk);
541 put_files_struct(old);
542}
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700543
Al Viro1ec7f1d2008-04-22 05:35:42 -0400544void exit_files(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545{
546 struct files_struct * files = tsk->files;
547
548 if (files) {
549 task_lock(tsk);
550 tsk->files = NULL;
551 task_unlock(tsk);
552 put_files_struct(files);
553 }
554}
555
Al Viro1ec7f1d2008-04-22 05:35:42 -0400556void put_fs_struct(struct fs_struct *fs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557{
558 /* No need to hold fs->lock if we are killing it */
559 if (atomic_dec_and_test(&fs->count)) {
Jan Blunck6ac08c32008-02-14 19:34:38 -0800560 path_put(&fs->root);
561 path_put(&fs->pwd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 kmem_cache_free(fs_cachep, fs);
563 }
564}
565
Al Viro1ec7f1d2008-04-22 05:35:42 -0400566void exit_fs(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567{
568 struct fs_struct * fs = tsk->fs;
569
570 if (fs) {
571 task_lock(tsk);
572 tsk->fs = NULL;
573 task_unlock(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400574 put_fs_struct(fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 }
576}
577
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578EXPORT_SYMBOL_GPL(exit_fs);
579
Balbir Singhcf475ad2008-04-29 01:00:16 -0700580#ifdef CONFIG_MM_OWNER
581/*
582 * Task p is exiting and it owned mm, lets find a new owner for it
583 */
584static inline int
585mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
586{
587 /*
588 * If there are other users of the mm and the owner (us) is exiting
589 * we need to find a new owner to take on the responsibility.
590 */
Balbir Singhcf475ad2008-04-29 01:00:16 -0700591 if (atomic_read(&mm->mm_users) <= 1)
592 return 0;
593 if (mm->owner != p)
594 return 0;
595 return 1;
596}
597
598void mm_update_next_owner(struct mm_struct *mm)
599{
600 struct task_struct *c, *g, *p = current;
601
602retry:
603 if (!mm_need_new_owner(mm, p))
604 return;
605
606 read_lock(&tasklist_lock);
607 /*
608 * Search in the children
609 */
610 list_for_each_entry(c, &p->children, sibling) {
611 if (c->mm == mm)
612 goto assign_new_owner;
613 }
614
615 /*
616 * Search in the siblings
617 */
618 list_for_each_entry(c, &p->parent->children, sibling) {
619 if (c->mm == mm)
620 goto assign_new_owner;
621 }
622
623 /*
624 * Search through everything else. We should not get
625 * here often
626 */
627 do_each_thread(g, c) {
628 if (c->mm == mm)
629 goto assign_new_owner;
630 } while_each_thread(g, c);
631
632 read_unlock(&tasklist_lock);
Balbir Singh31a78f22008-09-28 23:09:31 +0100633 /*
634 * We found no owner yet mm_users > 1: this implies that we are
635 * most likely racing with swapoff (try_to_unuse()) or /proc or
636 * ptrace or page migration (get_task_mm()). Mark owner as NULL,
637 * so that subsystems can understand the callback and take action.
638 */
639 down_write(&mm->mmap_sem);
640 cgroup_mm_owner_callbacks(mm->owner, NULL);
641 mm->owner = NULL;
642 up_write(&mm->mmap_sem);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700643 return;
644
645assign_new_owner:
646 BUG_ON(c == p);
647 get_task_struct(c);
Balbir Singh9363b9f2008-10-15 22:01:05 -0700648 read_unlock(&tasklist_lock);
649 down_write(&mm->mmap_sem);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700650 /*
651 * The task_lock protects c->mm from changing.
652 * We always want mm->owner->mm == mm
653 */
654 task_lock(c);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700655 if (c->mm != mm) {
656 task_unlock(c);
Balbir Singh9363b9f2008-10-15 22:01:05 -0700657 up_write(&mm->mmap_sem);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700658 put_task_struct(c);
659 goto retry;
660 }
661 cgroup_mm_owner_callbacks(mm->owner, c);
662 mm->owner = c;
663 task_unlock(c);
Balbir Singh9363b9f2008-10-15 22:01:05 -0700664 up_write(&mm->mmap_sem);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700665 put_task_struct(c);
666}
667#endif /* CONFIG_MM_OWNER */
668
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669/*
670 * Turn us into a lazy TLB process if we
671 * aren't already..
672 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700673static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674{
675 struct mm_struct *mm = tsk->mm;
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700676 struct core_state *core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
678 mm_release(tsk, mm);
679 if (!mm)
680 return;
681 /*
682 * Serialize with any possible pending coredump.
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700683 * We must hold mmap_sem around checking core_state
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 * and clearing tsk->mm. The core-inducing thread
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700685 * will increment ->nr_threads for each thread in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 * group with ->mm != NULL.
687 */
688 down_read(&mm->mmap_sem);
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700689 core_state = mm->core_state;
690 if (core_state) {
691 struct core_thread self;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 up_read(&mm->mmap_sem);
Oleg Nesterovc5f1cc82008-07-25 01:47:42 -0700693
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700694 self.task = tsk;
695 self.next = xchg(&core_state->dumper.next, &self);
696 /*
697 * Implies mb(), the result of xchg() must be visible
698 * to core_state->dumper.
699 */
700 if (atomic_dec_and_test(&core_state->nr_threads))
701 complete(&core_state->startup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
Oleg Nesterova94e2d42008-07-25 01:47:46 -0700703 for (;;) {
704 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
705 if (!self.task) /* see coredump_finish() */
706 break;
707 schedule();
708 }
709 __set_task_state(tsk, TASK_RUNNING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 down_read(&mm->mmap_sem);
711 }
712 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700713 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 /* more a memory barrier than a real lock */
715 task_lock(tsk);
716 tsk->mm = NULL;
717 up_read(&mm->mmap_sem);
718 enter_lazy_tlb(mm, current);
Rafael J. Wysocki0c1eecf2007-07-19 01:47:33 -0700719 /* We don't want this task to be frozen prematurely */
720 clear_freeze_flag(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 task_unlock(tsk);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700722 mm_update_next_owner(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 mmput(mm);
724}
725
Roland McGrathf4700212008-03-24 18:36:23 -0700726/*
Roland McGrath666f1642008-04-08 23:12:30 -0700727 * Return nonzero if @parent's children should reap themselves.
728 *
729 * Called with write_lock_irq(&tasklist_lock) held.
730 */
731static int ignoring_children(struct task_struct *parent)
732{
733 int ret;
734 struct sighand_struct *psig = parent->sighand;
735 unsigned long flags;
736 spin_lock_irqsave(&psig->siglock, flags);
737 ret = (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
738 (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT));
739 spin_unlock_irqrestore(&psig->siglock, flags);
740 return ret;
741}
742
743/*
Roland McGrathf4700212008-03-24 18:36:23 -0700744 * Detach all tasks we were using ptrace on.
745 * Any that need to be release_task'd are put on the @dead list.
746 *
747 * Called with write_lock(&tasklist_lock) held.
748 */
749static void ptrace_exit(struct task_struct *parent, struct list_head *dead)
750{
751 struct task_struct *p, *n;
Roland McGrath666f1642008-04-08 23:12:30 -0700752 int ign = -1;
Roland McGrathf4700212008-03-24 18:36:23 -0700753
754 list_for_each_entry_safe(p, n, &parent->ptraced, ptrace_entry) {
755 __ptrace_unlink(p);
756
757 if (p->exit_state != EXIT_ZOMBIE)
758 continue;
759
760 /*
761 * If it's a zombie, our attachedness prevented normal
762 * parent notification or self-reaping. Do notification
763 * now if it would have happened earlier. If it should
764 * reap itself, add it to the @dead list. We can't call
765 * release_task() here because we already hold tasklist_lock.
766 *
767 * If it's our own child, there is no notification to do.
Roland McGrath666f1642008-04-08 23:12:30 -0700768 * But if our normal children self-reap, then this child
769 * was prevented by ptrace and we must reap it now.
Roland McGrathf4700212008-03-24 18:36:23 -0700770 */
771 if (!task_detached(p) && thread_group_empty(p)) {
772 if (!same_thread_group(p->real_parent, parent))
773 do_notify_parent(p, p->exit_signal);
Roland McGrath666f1642008-04-08 23:12:30 -0700774 else {
775 if (ign < 0)
776 ign = ignoring_children(parent);
777 if (ign)
778 p->exit_signal = -1;
779 }
Roland McGrathf4700212008-03-24 18:36:23 -0700780 }
781
782 if (task_detached(p)) {
783 /*
784 * Mark it as in the process of being reaped.
785 */
786 p->exit_state = EXIT_DEAD;
787 list_add(&p->ptrace_entry, dead);
788 }
789 }
790}
791
792/*
793 * Finish up exit-time ptrace cleanup.
794 *
795 * Called without locks.
796 */
797static void ptrace_exit_finish(struct task_struct *parent,
798 struct list_head *dead)
799{
800 struct task_struct *p, *n;
801
802 BUG_ON(!list_empty(&parent->ptraced));
803
804 list_for_each_entry_safe(p, n, dead, ptrace_entry) {
805 list_del_init(&p->ptrace_entry);
806 release_task(p);
807 }
808}
809
810static void reparent_thread(struct task_struct *p, struct task_struct *father)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811{
Oleg Nesterov241ceee2006-12-24 23:30:44 +0300812 if (p->pdeath_signal)
813 /* We already hold the tasklist_lock here. */
814 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
815
Roland McGrathf4700212008-03-24 18:36:23 -0700816 list_move_tail(&p->sibling, &p->real_parent->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700818 /* If this is a threaded reparent there is no need to
819 * notify anyone anything has happened.
820 */
Oleg Nesterov376e1d22008-04-30 00:53:12 -0700821 if (same_thread_group(p->real_parent, father))
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700822 return;
823
824 /* We don't want people slaying init. */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700825 if (!task_detached(p))
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700826 p->exit_signal = SIGCHLD;
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700827
828 /* If we'd notified the old parent about this child's death,
829 * also notify the new parent.
830 */
Roland McGrathf4700212008-03-24 18:36:23 -0700831 if (!ptrace_reparented(p) &&
832 p->exit_state == EXIT_ZOMBIE &&
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700833 !task_detached(p) && thread_group_empty(p))
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700834 do_notify_parent(p, p->exit_signal);
835
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300836 kill_orphaned_pgrp(p, father);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837}
838
839/*
840 * When we die, we re-parent all our children.
841 * Try to give them to another thread in our thread
842 * group, and if no such member exists, give it to
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800843 * the child reaper process (ie "init") in our pid
844 * space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 */
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700846static struct task_struct *find_new_reaper(struct task_struct *father)
847{
848 struct pid_namespace *pid_ns = task_active_pid_ns(father);
849 struct task_struct *thread;
850
851 thread = father;
852 while_each_thread(father, thread) {
853 if (thread->flags & PF_EXITING)
854 continue;
855 if (unlikely(pid_ns->child_reaper == father))
856 pid_ns->child_reaper = thread;
857 return thread;
858 }
859
860 if (unlikely(pid_ns->child_reaper == father)) {
861 write_unlock_irq(&tasklist_lock);
862 if (unlikely(pid_ns == &init_pid_ns))
863 panic("Attempted to kill init!");
864
865 zap_pid_ns_processes(pid_ns);
866 write_lock_irq(&tasklist_lock);
867 /*
868 * We can not clear ->child_reaper or leave it alone.
869 * There may by stealth EXIT_DEAD tasks on ->children,
870 * forget_original_parent() must move them somewhere.
871 */
872 pid_ns->child_reaper = init_pid_ns.child_reaper;
873 }
874
875 return pid_ns->child_reaper;
876}
877
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700878static void forget_original_parent(struct task_struct *father)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700880 struct task_struct *p, *n, *reaper;
Roland McGrathf4700212008-03-24 18:36:23 -0700881 LIST_HEAD(ptrace_dead);
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700882
883 write_lock_irq(&tasklist_lock);
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700884 reaper = find_new_reaper(father);
Roland McGrathf4700212008-03-24 18:36:23 -0700885 /*
886 * First clean up ptrace if we were using it.
887 */
888 ptrace_exit(father, &ptrace_dead);
889
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -0700890 list_for_each_entry_safe(p, n, &father->children, sibling) {
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700891 p->real_parent = reaper;
Roland McGrathf4700212008-03-24 18:36:23 -0700892 if (p->parent == father) {
893 BUG_ON(p->ptrace);
894 p->parent = p->real_parent;
895 }
896 reparent_thread(p, father);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700898
899 write_unlock_irq(&tasklist_lock);
900 BUG_ON(!list_empty(&father->children));
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700901
Roland McGrathf4700212008-03-24 18:36:23 -0700902 ptrace_exit_finish(father, &ptrace_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903}
904
905/*
906 * Send signals to all our closest relatives so that they know
907 * to properly mourn us..
908 */
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300909static void exit_notify(struct task_struct *tsk, int group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910{
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700911 int signal;
912 void *cookie;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 /*
915 * This does two things:
916 *
917 * A. Make init inherit all the child processes
918 * B. Check to see if any process groups have become orphaned
919 * as a result of our exiting, and if they have any stopped
920 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
921 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700922 forget_original_parent(tsk);
Pavel Emelyanov2e4a7072007-10-18 23:40:01 -0700923 exit_task_namespaces(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700925 write_lock_irq(&tasklist_lock);
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300926 if (group_dead)
927 kill_orphaned_pgrp(tsk->group_leader, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
Oleg Nesterov24728442007-08-04 01:04:41 +0400929 /* Let father know we died
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 *
931 * Thread signals are configurable, but you aren't going to use
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700932 * that to send signals to arbitary processes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 * That stops right now.
934 *
935 * If the parent exec id doesn't match the exec id we saved
936 * when we started then we know the parent has changed security
937 * domain.
938 *
939 * If our self_exec id doesn't match our parent_exec_id then
940 * we have changed execution domain as these two values started
941 * the same after a fork.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700943 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300944 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700945 tsk->self_exec_id != tsk->parent_exec_id) &&
946 !capable(CAP_KILL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 tsk->exit_signal = SIGCHLD;
948
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700949 signal = tracehook_notify_death(tsk, &cookie, group_dead);
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700950 if (signal >= 0)
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700951 signal = do_notify_parent(tsk, signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700953 tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954
Oleg Nesterov2800d8d2008-04-30 00:53:12 -0700955 /* mt-exec, de_thread() is waiting for us */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700956 if (thread_group_leader(tsk) &&
Steve VanDeBogart2633f0e2008-08-26 15:14:36 -0700957 tsk->signal->group_exit_task &&
958 tsk->signal->notify_count < 0)
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700959 wake_up_process(tsk->signal->group_exit_task);
960
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 write_unlock_irq(&tasklist_lock);
962
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700963 tracehook_report_death(tsk, signal, cookie, group_dead);
964
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 /* If the process is dead, release it - nobody will wait for it */
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700966 if (signal == DEATH_REAP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 release_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968}
969
Jeff Dikee18eecb2007-07-15 23:38:48 -0700970#ifdef CONFIG_DEBUG_STACK_USAGE
971static void check_stack_usage(void)
972{
973 static DEFINE_SPINLOCK(low_water_lock);
974 static int lowest_to_date = THREAD_SIZE;
975 unsigned long *n = end_of_stack(current);
976 unsigned long free;
977
978 while (*n == 0)
979 n++;
980 free = (unsigned long)n - (unsigned long)end_of_stack(current);
981
982 if (free >= lowest_to_date)
983 return;
984
985 spin_lock(&low_water_lock);
986 if (free < lowest_to_date) {
987 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
988 "left\n",
989 current->comm, free);
990 lowest_to_date = free;
991 }
992 spin_unlock(&low_water_lock);
993}
994#else
995static inline void check_stack_usage(void) {}
996#endif
997
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800998NORET_TYPE void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999{
1000 struct task_struct *tsk = current;
1001 int group_dead;
1002
1003 profile_task_exit(tsk);
1004
Jens Axboe22e2c502005-06-27 10:55:12 +02001005 WARN_ON(atomic_read(&tsk->fs_excl));
1006
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 if (unlikely(in_interrupt()))
1008 panic("Aiee, killing interrupt handler!");
1009 if (unlikely(!tsk->pid))
1010 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
Roland McGrath30199f52008-07-25 19:45:46 -07001012 tracehook_report_exit(&code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013
Alexander Nybergdf164db2005-06-23 00:09:13 -07001014 /*
1015 * We're taking recursive faults here in do_exit. Safest is to just
1016 * leave this task alone and wait for reboot.
1017 */
1018 if (unlikely(tsk->flags & PF_EXITING)) {
1019 printk(KERN_ALERT
1020 "Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001021 /*
1022 * We can do this unlocked here. The futex code uses
1023 * this flag just to verify whether the pi state
1024 * cleanup has been done or not. In the worst case it
1025 * loops once more. We pretend that the cleanup was
1026 * done as there is no way to return. Either the
1027 * OWNER_DIED bit is set by now or we push the blocked
1028 * task into the wait for ever nirwana as well.
1029 */
1030 tsk->flags |= PF_EXITPIDONE;
Al Viroafc847b2006-02-28 12:51:55 -05001031 if (tsk->io_context)
1032 exit_io_context();
Alexander Nybergdf164db2005-06-23 00:09:13 -07001033 set_current_state(TASK_UNINTERRUPTIBLE);
1034 schedule();
1035 }
1036
Oleg Nesterovd12619b2008-02-08 04:19:12 -08001037 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001038 /*
1039 * tsk->flags are checked in the futex code to protect against
1040 * an exiting task cleaning up the robust pi futexes.
1041 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -07001042 smp_mb();
1043 spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 if (unlikely(in_atomic()))
1046 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -07001047 current->comm, task_pid_nr(current),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 preempt_count());
1049
1050 acct_update_integrals(tsk);
Hugh Dickins365e9c872005-10-29 18:16:18 -07001051 if (tsk->mm) {
1052 update_hiwater_rss(tsk->mm);
1053 update_hiwater_vm(tsk->mm);
1054 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -07001056 if (group_dead) {
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001057 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -07001058 exit_itimers(tsk->signal);
Andrew Mortonc3068952005-08-04 16:49:32 -07001059 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -07001060 acct_collect(code, group_dead);
Miloslav Trmac522ed772007-07-15 23:40:56 -07001061 if (group_dead)
1062 tty_audit_exit();
Al Virofa84cb92006-03-29 20:30:19 -05001063 if (unlikely(tsk->audit_context))
1064 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -08001065
Jonathan Limf2ab6d82007-08-30 23:56:23 -07001066 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -08001067 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -07001068
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 exit_mm(tsk);
1070
KaiGai Kohei0e464812006-06-25 05:49:24 -07001071 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -07001072 acct_process();
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001073 trace_sched_process_exit(tsk);
1074
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 exit_sem(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -04001076 exit_files(tsk);
1077 exit_fs(tsk);
Jeff Dikee18eecb2007-07-15 23:38:48 -07001078 check_stack_usage();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 exit_thread();
Paul Menageb4f48b62007-10-18 23:39:33 -07001080 cgroup_exit(tsk, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 exit_keys(tsk);
1082
1083 if (group_dead && tsk->signal->leader)
1084 disassociate_ctty(1);
1085
Al Viroa1261f52005-11-13 16:06:55 -08001086 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 if (tsk->binfmt)
1088 module_put(tsk->binfmt->module);
1089
Matt Helsley9f460802005-11-07 00:59:16 -08001090 proc_exit_connector(tsk);
Oleg Nesterov821c7de2008-03-02 21:44:44 +03001091 exit_notify(tsk, group_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001093 mpol_put(tsk->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 tsk->mempolicy = NULL;
1095#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001096#ifdef CONFIG_FUTEX
Ingo Molnarde5097c2006-01-09 15:59:21 -08001097 /*
Ingo Molnarc87e2832006-06-27 02:54:58 -07001098 * This must happen late, after the PID is not
1099 * hashed anymore:
1100 */
1101 if (unlikely(!list_empty(&tsk->pi_state_list)))
1102 exit_pi_state_list(tsk);
1103 if (unlikely(current->pi_state_cache))
1104 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001105#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001106 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001107 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -08001108 */
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001109 debug_check_no_locks_held(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001110 /*
1111 * We can do this unlocked here. The futex code uses this flag
1112 * just to verify whether the pi state cleanup has been done
1113 * or not. In the worst case it loops once more.
1114 */
1115 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116
Al Viroafc847b2006-02-28 12:51:55 -05001117 if (tsk->io_context)
1118 exit_io_context();
1119
Jens Axboeb92ce552006-04-11 13:52:07 +02001120 if (tsk->splice_pipe)
1121 __free_pipe_info(tsk->splice_pipe);
1122
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001123 preempt_disable();
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001124 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -07001125 tsk->state = TASK_DEAD;
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001126
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 schedule();
1128 BUG();
1129 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -07001130 for (;;)
1131 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132}
1133
Russ Anderson012914d2005-04-23 00:08:00 -07001134EXPORT_SYMBOL_GPL(do_exit);
1135
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1137{
1138 if (comp)
1139 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001140
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 do_exit(code);
1142}
1143
1144EXPORT_SYMBOL(complete_and_exit);
1145
1146asmlinkage long sys_exit(int error_code)
1147{
1148 do_exit((error_code&0xff)<<8);
1149}
1150
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151/*
1152 * Take down every thread in the group. This is called by fatal signals
1153 * as well as by sys_exit_group (below).
1154 */
1155NORET_TYPE void
1156do_group_exit(int exit_code)
1157{
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001158 struct signal_struct *sig = current->signal;
1159
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1161
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001162 if (signal_group_exit(sig))
1163 exit_code = sig->group_exit_code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 else if (!thread_group_empty(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 struct sighand_struct *const sighand = current->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001167 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 /* Another thread got here before we took the lock. */
1169 exit_code = sig->group_exit_code;
1170 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001172 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 zap_other_threads(current);
1174 }
1175 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 }
1177
1178 do_exit(exit_code);
1179 /* NOTREACHED */
1180}
1181
1182/*
1183 * this kills every thread in the thread group. Note that any externally
1184 * wait4()-ing process will get the correct exit code - even if this
1185 * thread is not the thread group leader.
1186 */
1187asmlinkage void sys_exit_group(int error_code)
1188{
1189 do_group_exit((error_code & 0xff) << 8);
1190}
1191
Eric W. Biederman161550d2008-02-08 04:19:14 -08001192static struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1193{
1194 struct pid *pid = NULL;
1195 if (type == PIDTYPE_PID)
1196 pid = task->pids[type].pid;
1197 else if (type < PIDTYPE_MAX)
1198 pid = task->group_leader->pids[type].pid;
1199 return pid;
1200}
1201
1202static int eligible_child(enum pid_type type, struct pid *pid, int options,
1203 struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204{
Roland McGrath73243282007-05-06 14:50:20 -07001205 int err;
1206
Eric W. Biederman161550d2008-02-08 04:19:14 -08001207 if (type < PIDTYPE_MAX) {
1208 if (task_pid_type(p, type) != pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 return 0;
1210 }
1211
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 /* Wait for all children (clone and not) if __WALL is set;
1213 * otherwise, wait for clone children *only* if __WCLONE is
1214 * set; otherwise, wait for non-clone children *only*. (Note:
1215 * A "clone" child here is one that reports to its parent
1216 * using a signal other than SIGCHLD.) */
1217 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1218 && !(options & __WALL))
1219 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220
Roland McGrath73243282007-05-06 14:50:20 -07001221 err = security_task_wait(p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001222 if (err)
1223 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224
Roland McGrath14dd0b82008-03-30 18:41:25 -07001225 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226}
1227
Ingo Molnar36c8b582006-07-03 00:25:41 -07001228static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 int why, int status,
1230 struct siginfo __user *infop,
1231 struct rusage __user *rusagep)
1232{
1233 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -07001234
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 put_task_struct(p);
1236 if (!retval)
1237 retval = put_user(SIGCHLD, &infop->si_signo);
1238 if (!retval)
1239 retval = put_user(0, &infop->si_errno);
1240 if (!retval)
1241 retval = put_user((short)why, &infop->si_code);
1242 if (!retval)
1243 retval = put_user(pid, &infop->si_pid);
1244 if (!retval)
1245 retval = put_user(uid, &infop->si_uid);
1246 if (!retval)
1247 retval = put_user(status, &infop->si_status);
1248 if (!retval)
1249 retval = pid;
1250 return retval;
1251}
1252
1253/*
1254 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1255 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1256 * the lock and this task is uninteresting. If we return nonzero, we have
1257 * released the lock and the system call should return.
1258 */
Roland McGrath98abed02008-03-19 19:24:59 -07001259static int wait_task_zombie(struct task_struct *p, int options,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 struct siginfo __user *infop,
1261 int __user *stat_addr, struct rusage __user *ru)
1262{
1263 unsigned long state;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001264 int retval, status, traced;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001265 pid_t pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
Roland McGrath98abed02008-03-19 19:24:59 -07001267 if (!likely(options & WEXITED))
1268 return 0;
1269
1270 if (unlikely(options & WNOWAIT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 uid_t uid = p->uid;
1272 int exit_code = p->exit_code;
1273 int why, status;
1274
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 get_task_struct(p);
1276 read_unlock(&tasklist_lock);
1277 if ((exit_code & 0x7f) == 0) {
1278 why = CLD_EXITED;
1279 status = exit_code >> 8;
1280 } else {
1281 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1282 status = exit_code & 0x7f;
1283 }
1284 return wait_noreap_copyout(p, pid, uid, why,
1285 status, infop, ru);
1286 }
1287
1288 /*
1289 * Try to move the task's state to DEAD
1290 * only one thread is allowed to do this:
1291 */
1292 state = xchg(&p->exit_state, EXIT_DEAD);
1293 if (state != EXIT_ZOMBIE) {
1294 BUG_ON(state != EXIT_DEAD);
1295 return 0;
1296 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297
Oleg Nesterov53b6f9f2008-04-30 00:53:13 -07001298 traced = ptrace_reparented(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001299
1300 if (likely(!traced)) {
Jesper Juhl3795e162006-01-09 20:54:39 -08001301 struct signal_struct *psig;
1302 struct signal_struct *sig;
Frank Mayharf06febc2008-09-12 09:54:39 -07001303 struct task_cputime cputime;
Jesper Juhl3795e162006-01-09 20:54:39 -08001304
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 /*
1306 * The resource counters for the group leader are in its
1307 * own task_struct. Those for dead threads in the group
1308 * are in its signal_struct, as are those for the child
1309 * processes it has previously reaped. All these
1310 * accumulate in the parent's signal_struct c* fields.
1311 *
1312 * We don't bother to take a lock here to protect these
1313 * p->signal fields, because they are only touched by
1314 * __exit_signal, which runs with tasklist_lock
1315 * write-locked anyway, and so is excluded here. We do
1316 * need to protect the access to p->parent->signal fields,
1317 * as other threads in the parent group can be right
1318 * here reaping other children at the same time.
Frank Mayharf06febc2008-09-12 09:54:39 -07001319 *
1320 * We use thread_group_cputime() to get times for the thread
1321 * group, which consolidates times for all threads in the
1322 * group including the group leader.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 */
1324 spin_lock_irq(&p->parent->sighand->siglock);
Jesper Juhl3795e162006-01-09 20:54:39 -08001325 psig = p->parent->signal;
1326 sig = p->signal;
Frank Mayharf06febc2008-09-12 09:54:39 -07001327 thread_group_cputime(p, &cputime);
Jesper Juhl3795e162006-01-09 20:54:39 -08001328 psig->cutime =
1329 cputime_add(psig->cutime,
Frank Mayharf06febc2008-09-12 09:54:39 -07001330 cputime_add(cputime.utime,
1331 sig->cutime));
Jesper Juhl3795e162006-01-09 20:54:39 -08001332 psig->cstime =
1333 cputime_add(psig->cstime,
Frank Mayharf06febc2008-09-12 09:54:39 -07001334 cputime_add(cputime.stime,
1335 sig->cstime));
Laurent Vivier9ac52312007-10-15 17:00:19 +02001336 psig->cgtime =
1337 cputime_add(psig->cgtime,
1338 cputime_add(p->gtime,
1339 cputime_add(sig->gtime,
1340 sig->cgtime)));
Jesper Juhl3795e162006-01-09 20:54:39 -08001341 psig->cmin_flt +=
1342 p->min_flt + sig->min_flt + sig->cmin_flt;
1343 psig->cmaj_flt +=
1344 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1345 psig->cnvcsw +=
1346 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1347 psig->cnivcsw +=
1348 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001349 psig->cinblock +=
1350 task_io_get_inblock(p) +
1351 sig->inblock + sig->cinblock;
1352 psig->coublock +=
1353 task_io_get_oublock(p) +
1354 sig->oublock + sig->coublock;
Andrea Righi59954772008-07-27 17:29:15 +02001355 task_io_accounting_add(&psig->ioac, &p->ioac);
1356 task_io_accounting_add(&psig->ioac, &sig->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357 spin_unlock_irq(&p->parent->sighand->siglock);
1358 }
1359
1360 /*
1361 * Now we are sure this task is interesting, and no other
1362 * thread can reap it because we set its state to EXIT_DEAD.
1363 */
1364 read_unlock(&tasklist_lock);
1365
1366 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1367 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1368 ? p->signal->group_exit_code : p->exit_code;
1369 if (!retval && stat_addr)
1370 retval = put_user(status, stat_addr);
1371 if (!retval && infop)
1372 retval = put_user(SIGCHLD, &infop->si_signo);
1373 if (!retval && infop)
1374 retval = put_user(0, &infop->si_errno);
1375 if (!retval && infop) {
1376 int why;
1377
1378 if ((status & 0x7f) == 0) {
1379 why = CLD_EXITED;
1380 status >>= 8;
1381 } else {
1382 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1383 status &= 0x7f;
1384 }
1385 retval = put_user((short)why, &infop->si_code);
1386 if (!retval)
1387 retval = put_user(status, &infop->si_status);
1388 }
1389 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001390 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 if (!retval && infop)
1392 retval = put_user(p->uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001393 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001394 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001395
1396 if (traced) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001398 /* We dropped tasklist, ptracer could die and untrace */
1399 ptrace_unlink(p);
1400 /*
1401 * If this is not a detached task, notify the parent.
1402 * If it's still not detached after that, don't release
1403 * it now.
1404 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001405 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001406 do_notify_parent(p, p->exit_signal);
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001407 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001408 p->exit_state = EXIT_ZOMBIE;
1409 p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 }
1411 }
1412 write_unlock_irq(&tasklist_lock);
1413 }
1414 if (p != NULL)
1415 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001416
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 return retval;
1418}
1419
1420/*
1421 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1422 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1423 * the lock and this task is uninteresting. If we return nonzero, we have
1424 * released the lock and the system call should return.
1425 */
Roland McGrathf4700212008-03-24 18:36:23 -07001426static int wait_task_stopped(int ptrace, struct task_struct *p,
Roland McGrath98abed02008-03-19 19:24:59 -07001427 int options, struct siginfo __user *infop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 int __user *stat_addr, struct rusage __user *ru)
1429{
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001430 int retval, exit_code, why;
1431 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001432 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
Roland McGrathf4700212008-03-24 18:36:23 -07001434 if (!(options & WUNTRACED))
Roland McGrath98abed02008-03-19 19:24:59 -07001435 return 0;
1436
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001437 exit_code = 0;
1438 spin_lock_irq(&p->sighand->siglock);
1439
1440 if (unlikely(!task_is_stopped_or_traced(p)))
1441 goto unlock_sig;
1442
Roland McGrathf4700212008-03-24 18:36:23 -07001443 if (!ptrace && p->signal->group_stop_count > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 /*
1445 * A group stop is in progress and this is the group leader.
1446 * We won't report until all threads have stopped.
1447 */
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001448 goto unlock_sig;
1449
1450 exit_code = p->exit_code;
1451 if (!exit_code)
1452 goto unlock_sig;
1453
Roland McGrath98abed02008-03-19 19:24:59 -07001454 if (!unlikely(options & WNOWAIT))
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001455 p->exit_code = 0;
1456
1457 uid = p->uid;
1458unlock_sig:
1459 spin_unlock_irq(&p->sighand->siglock);
1460 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 return 0;
1462
1463 /*
1464 * Now we are pretty sure this task is interesting.
1465 * Make sure it doesn't get reaped out from under us while we
1466 * give up the lock and then examine it below. We don't want to
1467 * keep holding onto the tasklist_lock while we call getrusage and
1468 * possibly take page faults for user memory.
1469 */
1470 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001471 pid = task_pid_vnr(p);
Roland McGrathf4700212008-03-24 18:36:23 -07001472 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 read_unlock(&tasklist_lock);
1474
Roland McGrath98abed02008-03-19 19:24:59 -07001475 if (unlikely(options & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 return wait_noreap_copyout(p, pid, uid,
Scott James Remnante6ceb322007-11-28 16:22:07 -08001477 why, exit_code,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 infop, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479
1480 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1481 if (!retval && stat_addr)
1482 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1483 if (!retval && infop)
1484 retval = put_user(SIGCHLD, &infop->si_signo);
1485 if (!retval && infop)
1486 retval = put_user(0, &infop->si_errno);
1487 if (!retval && infop)
Roland McGrath6efcae42008-03-08 11:41:22 -08001488 retval = put_user((short)why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 if (!retval && infop)
1490 retval = put_user(exit_code, &infop->si_status);
1491 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001492 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001494 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001496 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 put_task_struct(p);
1498
1499 BUG_ON(!retval);
1500 return retval;
1501}
1502
1503/*
1504 * Handle do_wait work for one task in a live, non-stopped state.
1505 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1506 * the lock and this task is uninteresting. If we return nonzero, we have
1507 * released the lock and the system call should return.
1508 */
Roland McGrath98abed02008-03-19 19:24:59 -07001509static int wait_task_continued(struct task_struct *p, int options,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 struct siginfo __user *infop,
1511 int __user *stat_addr, struct rusage __user *ru)
1512{
1513 int retval;
1514 pid_t pid;
1515 uid_t uid;
1516
Roland McGrath98abed02008-03-19 19:24:59 -07001517 if (!unlikely(options & WCONTINUED))
1518 return 0;
1519
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1521 return 0;
1522
1523 spin_lock_irq(&p->sighand->siglock);
1524 /* Re-check with the lock held. */
1525 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1526 spin_unlock_irq(&p->sighand->siglock);
1527 return 0;
1528 }
Roland McGrath98abed02008-03-19 19:24:59 -07001529 if (!unlikely(options & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1531 spin_unlock_irq(&p->sighand->siglock);
1532
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001533 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 uid = p->uid;
1535 get_task_struct(p);
1536 read_unlock(&tasklist_lock);
1537
1538 if (!infop) {
1539 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1540 put_task_struct(p);
1541 if (!retval && stat_addr)
1542 retval = put_user(0xffff, stat_addr);
1543 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001544 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 } else {
1546 retval = wait_noreap_copyout(p, pid, uid,
1547 CLD_CONTINUED, SIGCONT,
1548 infop, ru);
1549 BUG_ON(retval == 0);
1550 }
1551
1552 return retval;
1553}
1554
Roland McGrath98abed02008-03-19 19:24:59 -07001555/*
1556 * Consider @p for a wait by @parent.
1557 *
1558 * -ECHILD should be in *@notask_error before the first call.
1559 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1560 * Returns zero if the search for a child should continue;
Roland McGrath14dd0b82008-03-30 18:41:25 -07001561 * then *@notask_error is 0 if @p is an eligible child,
1562 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001563 */
Roland McGrathf4700212008-03-24 18:36:23 -07001564static int wait_consider_task(struct task_struct *parent, int ptrace,
Roland McGrath98abed02008-03-19 19:24:59 -07001565 struct task_struct *p, int *notask_error,
1566 enum pid_type type, struct pid *pid, int options,
1567 struct siginfo __user *infop,
1568 int __user *stat_addr, struct rusage __user *ru)
1569{
1570 int ret = eligible_child(type, pid, options, p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001571 if (!ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001572 return ret;
1573
Roland McGrath14dd0b82008-03-30 18:41:25 -07001574 if (unlikely(ret < 0)) {
1575 /*
1576 * If we have not yet seen any eligible child,
1577 * then let this error code replace -ECHILD.
1578 * A permission error will give the user a clue
1579 * to look for security policy problems, rather
1580 * than for mysterious wait bugs.
1581 */
1582 if (*notask_error)
1583 *notask_error = ret;
1584 }
1585
Roland McGrathf4700212008-03-24 18:36:23 -07001586 if (likely(!ptrace) && unlikely(p->ptrace)) {
1587 /*
1588 * This child is hidden by ptrace.
1589 * We aren't allowed to see it now, but eventually we will.
1590 */
1591 *notask_error = 0;
1592 return 0;
1593 }
1594
Roland McGrath98abed02008-03-19 19:24:59 -07001595 if (p->exit_state == EXIT_DEAD)
1596 return 0;
1597
1598 /*
1599 * We don't reap group leaders with subthreads.
1600 */
1601 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
1602 return wait_task_zombie(p, options, infop, stat_addr, ru);
1603
1604 /*
1605 * It's stopped or running now, so it might
1606 * later continue, exit, or stop again.
1607 */
1608 *notask_error = 0;
1609
1610 if (task_is_stopped_or_traced(p))
Roland McGrathf4700212008-03-24 18:36:23 -07001611 return wait_task_stopped(ptrace, p, options,
1612 infop, stat_addr, ru);
Roland McGrath98abed02008-03-19 19:24:59 -07001613
1614 return wait_task_continued(p, options, infop, stat_addr, ru);
1615}
1616
1617/*
1618 * Do the work of do_wait() for one thread in the group, @tsk.
1619 *
1620 * -ECHILD should be in *@notask_error before the first call.
1621 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1622 * Returns zero if the search for a child should continue; then
Roland McGrath14dd0b82008-03-30 18:41:25 -07001623 * *@notask_error is 0 if there were any eligible children,
1624 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001625 */
1626static int do_wait_thread(struct task_struct *tsk, int *notask_error,
1627 enum pid_type type, struct pid *pid, int options,
1628 struct siginfo __user *infop, int __user *stat_addr,
1629 struct rusage __user *ru)
1630{
1631 struct task_struct *p;
1632
1633 list_for_each_entry(p, &tsk->children, sibling) {
Roland McGrathf4700212008-03-24 18:36:23 -07001634 /*
1635 * Do not consider detached threads.
1636 */
1637 if (!task_detached(p)) {
1638 int ret = wait_consider_task(tsk, 0, p, notask_error,
1639 type, pid, options,
1640 infop, stat_addr, ru);
1641 if (ret)
1642 return ret;
1643 }
Roland McGrath98abed02008-03-19 19:24:59 -07001644 }
1645
1646 return 0;
1647}
1648
1649static int ptrace_do_wait(struct task_struct *tsk, int *notask_error,
1650 enum pid_type type, struct pid *pid, int options,
1651 struct siginfo __user *infop, int __user *stat_addr,
1652 struct rusage __user *ru)
1653{
1654 struct task_struct *p;
1655
1656 /*
Roland McGrathf4700212008-03-24 18:36:23 -07001657 * Traditionally we see ptrace'd stopped tasks regardless of options.
Roland McGrath98abed02008-03-19 19:24:59 -07001658 */
Roland McGrathf4700212008-03-24 18:36:23 -07001659 options |= WUNTRACED;
Roland McGrath98abed02008-03-19 19:24:59 -07001660
Roland McGrathf4700212008-03-24 18:36:23 -07001661 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1662 int ret = wait_consider_task(tsk, 1, p, notask_error,
1663 type, pid, options,
1664 infop, stat_addr, ru);
1665 if (ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001666 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001667 }
1668
1669 return 0;
1670}
1671
Eric W. Biederman161550d2008-02-08 04:19:14 -08001672static long do_wait(enum pid_type type, struct pid *pid, int options,
1673 struct siginfo __user *infop, int __user *stat_addr,
1674 struct rusage __user *ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675{
1676 DECLARE_WAITQUEUE(wait, current);
1677 struct task_struct *tsk;
Roland McGrath98abed02008-03-19 19:24:59 -07001678 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001680 trace_sched_process_wait(pid);
1681
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 add_wait_queue(&current->signal->wait_chldexit,&wait);
1683repeat:
Roland McGrath98abed02008-03-19 19:24:59 -07001684 /*
1685 * If there is nothing that can match our critiera just get out.
1686 * We will clear @retval to zero if we see any child that might later
1687 * match our criteria, even if we are not able to reap it yet.
1688 */
Eric W. Biederman161550d2008-02-08 04:19:14 -08001689 retval = -ECHILD;
1690 if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type])))
1691 goto end;
1692
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 current->state = TASK_INTERRUPTIBLE;
1694 read_lock(&tasklist_lock);
1695 tsk = current;
1696 do {
Roland McGrath98abed02008-03-19 19:24:59 -07001697 int tsk_result = do_wait_thread(tsk, &retval,
1698 type, pid, options,
1699 infop, stat_addr, ru);
1700 if (!tsk_result)
1701 tsk_result = ptrace_do_wait(tsk, &retval,
1702 type, pid, options,
1703 infop, stat_addr, ru);
1704 if (tsk_result) {
1705 /*
1706 * tasklist_lock is unlocked and we have a final result.
1707 */
1708 retval = tsk_result;
1709 goto end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 }
Roland McGrath98abed02008-03-19 19:24:59 -07001711
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 if (options & __WNOTHREAD)
1713 break;
1714 tsk = next_thread(tsk);
Eric Sesterhenn125e1872006-06-23 02:06:06 -07001715 BUG_ON(tsk->signal != current->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 } while (tsk != current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001718
Roland McGrath98abed02008-03-19 19:24:59 -07001719 if (!retval && !(options & WNOHANG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 retval = -ERESTARTSYS;
Roland McGrath98abed02008-03-19 19:24:59 -07001721 if (!signal_pending(current)) {
1722 schedule();
1723 goto repeat;
1724 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 }
Roland McGrath98abed02008-03-19 19:24:59 -07001726
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727end:
1728 current->state = TASK_RUNNING;
1729 remove_wait_queue(&current->signal->wait_chldexit,&wait);
1730 if (infop) {
1731 if (retval > 0)
Oleg Nesterov9cbab812008-02-08 04:19:02 -08001732 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 else {
1734 /*
1735 * For a WNOHANG return, clear out all the fields
1736 * we would set so the user can easily tell the
1737 * difference.
1738 */
1739 if (!retval)
1740 retval = put_user(0, &infop->si_signo);
1741 if (!retval)
1742 retval = put_user(0, &infop->si_errno);
1743 if (!retval)
1744 retval = put_user(0, &infop->si_code);
1745 if (!retval)
1746 retval = put_user(0, &infop->si_pid);
1747 if (!retval)
1748 retval = put_user(0, &infop->si_uid);
1749 if (!retval)
1750 retval = put_user(0, &infop->si_status);
1751 }
1752 }
1753 return retval;
1754}
1755
Eric W. Biederman161550d2008-02-08 04:19:14 -08001756asmlinkage long sys_waitid(int which, pid_t upid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 struct siginfo __user *infop, int options,
1758 struct rusage __user *ru)
1759{
Eric W. Biederman161550d2008-02-08 04:19:14 -08001760 struct pid *pid = NULL;
1761 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 long ret;
1763
1764 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1765 return -EINVAL;
1766 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1767 return -EINVAL;
1768
1769 switch (which) {
1770 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001771 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 break;
1773 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001774 type = PIDTYPE_PID;
1775 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 return -EINVAL;
1777 break;
1778 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001779 type = PIDTYPE_PGID;
1780 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 break;
1783 default:
1784 return -EINVAL;
1785 }
1786
Eric W. Biederman161550d2008-02-08 04:19:14 -08001787 if (type < PIDTYPE_MAX)
1788 pid = find_get_pid(upid);
1789 ret = do_wait(type, pid, options, infop, NULL, ru);
1790 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791
1792 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001793 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 return ret;
1795}
1796
Eric W. Biederman161550d2008-02-08 04:19:14 -08001797asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 int options, struct rusage __user *ru)
1799{
Eric W. Biederman161550d2008-02-08 04:19:14 -08001800 struct pid *pid = NULL;
1801 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 long ret;
1803
1804 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1805 __WNOTHREAD|__WCLONE|__WALL))
1806 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001807
1808 if (upid == -1)
1809 type = PIDTYPE_MAX;
1810 else if (upid < 0) {
1811 type = PIDTYPE_PGID;
1812 pid = find_get_pid(-upid);
1813 } else if (upid == 0) {
1814 type = PIDTYPE_PGID;
1815 pid = get_pid(task_pgrp(current));
1816 } else /* upid > 0 */ {
1817 type = PIDTYPE_PID;
1818 pid = find_get_pid(upid);
1819 }
1820
1821 ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru);
1822 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823
1824 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001825 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 return ret;
1827}
1828
1829#ifdef __ARCH_WANT_SYS_WAITPID
1830
1831/*
1832 * sys_waitpid() remains for compatibility. waitpid() should be
1833 * implemented by calling sys_wait4() from libc.a.
1834 */
1835asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1836{
1837 return sys_wait4(pid, stat_addr, options, NULL);
1838}
1839
1840#endif