blob: 85a83c831856c193570e40a3b7d3e03ef8862d1c [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>
Ingo Molnar34f192c2006-03-27 01:16:24 -080043#include <linux/compat.h>
Jens Axboeb92ce552006-04-11 13:52:07 +020044#include <linux/pipe_fs_i.h>
Al Virofa84cb92006-03-29 20:30:19 -050045#include <linux/audit.h> /* for audit_free() */
Adrian Bunk83cc5ed2006-06-25 05:47:41 -070046#include <linux/resource.h>
David Howells0d67a462006-08-29 19:05:56 +010047#include <linux/blkdev.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070048#include <linux/task_io_accounting_ops.h>
Roland McGrath30199f52008-07-25 19:45:46 -070049#include <linux/tracehook.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->utime = cputime_add(sig->utime, task_utime(tsk));
116 sig->stime = cputime_add(sig->stime, task_stime(tsk));
117 sig->gtime = cputime_add(sig->gtime, task_gtime(tsk));
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800118 sig->min_flt += tsk->min_flt;
119 sig->maj_flt += tsk->maj_flt;
120 sig->nvcsw += tsk->nvcsw;
121 sig->nivcsw += tsk->nivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700122 sig->inblock += task_io_get_inblock(tsk);
123 sig->oublock += task_io_get_oublock(tsk);
Andrea Righi59954772008-07-27 17:29:15 +0200124 task_io_accounting_add(&sig->ioac, &tsk->ioac);
Balbir Singh172ba842007-07-09 18:52:00 +0200125 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800126 sig = NULL; /* Marker for below. */
127 }
128
Oleg Nesterov58767002006-03-28 16:11:20 -0800129 __unhash_process(tsk);
130
Oleg Nesterovda7978b2008-05-23 13:04:41 -0700131 /*
132 * Do this under ->siglock, we can race with another thread
133 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
134 */
135 flush_sigqueue(&tsk->pending);
136
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800137 tsk->signal = NULL;
Oleg Nesterova7e53282006-03-28 16:11:27 -0800138 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800139 spin_unlock(&sighand->siglock);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800140
Oleg Nesterova7e53282006-03-28 16:11:27 -0800141 __cleanup_sighand(sighand);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800142 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800143 if (sig) {
144 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov093a8e82006-10-28 10:38:51 -0700145 taskstats_tgid_free(sig);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800146 __cleanup_signal(sig);
147 }
148}
149
Eric W. Biederman8c7904a02006-03-31 02:31:37 -0800150static void delayed_put_task_struct(struct rcu_head *rhp)
151{
152 put_task_struct(container_of(rhp, struct task_struct, rcu));
153}
154
Roland McGrathf4700212008-03-24 18:36:23 -0700155
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156void release_task(struct task_struct * p)
157{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700158 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800160repeat:
Roland McGrathdae33572008-07-25 19:45:48 -0700161 tracehook_prepare_release_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 atomic_dec(&p->user->processes);
Pavel Emelyanov60347f62007-10-18 23:40:03 -0700163 proc_flush_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 write_lock_irq(&tasklist_lock);
Roland McGrathdae33572008-07-25 19:45:48 -0700165 tracehook_finish_release_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 __exit_signal(p);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -0800167
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 /*
169 * If we are the last non-leader member of the thread
170 * group, and the leader is zombie, then notify the
171 * group leader's parent process. (if it wants notification.)
172 */
173 zap_leader = 0;
174 leader = p->group_leader;
175 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700176 BUG_ON(task_detached(leader));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 do_notify_parent(leader, leader->exit_signal);
178 /*
179 * If we were the last child thread and the leader has
180 * exited already, and the leader's parent ignores SIGCHLD,
181 * then we are the one who should release the leader.
182 *
183 * do_notify_parent() will have marked it self-reaping in
184 * that case.
185 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700186 zap_leader = task_detached(leader);
Roland McGrathdae33572008-07-25 19:45:48 -0700187
188 /*
189 * This maintains the invariant that release_task()
190 * only runs on a task in EXIT_DEAD, just for sanity.
191 */
192 if (zap_leader)
193 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 }
195
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 release_thread(p);
Eric W. Biederman8c7904a02006-03-31 02:31:37 -0800198 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
200 p = leader;
201 if (unlikely(zap_leader))
202 goto repeat;
203}
204
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205/*
206 * This checks not only the pgrp, but falls back on the pid if no
207 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
208 * without this...
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800209 *
210 * The caller must hold rcu lock or the tasklist lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 */
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800212struct pid *session_of_pgrp(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213{
214 struct task_struct *p;
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800215 struct pid *sid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800217 p = pid_task(pgrp, PIDTYPE_PGID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800218 if (p == NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800219 p = pid_task(pgrp, PIDTYPE_PID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800220 if (p != NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800221 sid = task_session(p);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800222
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 return sid;
224}
225
226/*
227 * Determine if a process group is "orphaned", according to the POSIX
228 * definition in 2.2.2.52. Orphaned process groups are not to be affected
229 * by terminal-generated stop signals. Newly orphaned process groups are
230 * to receive a SIGHUP and a SIGCONT.
231 *
232 * "I ask you, have you ever known what it is to be an orphan?"
233 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800234static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235{
236 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800238 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300239 if ((p == ignored_task) ||
240 (p->exit_state && thread_group_empty(p)) ||
241 is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 continue;
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300243
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800244 if (task_pgrp(p->real_parent) != pgrp &&
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300245 task_session(p->real_parent) == task_session(p))
246 return 0;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800247 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300248
249 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250}
251
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800252int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253{
254 int retval;
255
256 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800257 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258 read_unlock(&tasklist_lock);
259
260 return retval;
261}
262
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800263static int has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264{
265 int retval = 0;
266 struct task_struct *p;
267
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800268 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Matthew Wilcox338077e2007-12-06 11:09:35 -0500269 if (!task_is_stopped(p))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 retval = 1;
272 break;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800273 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 return retval;
275}
276
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300277/*
278 * Check to see if any process groups have become orphaned as
279 * a result of our exiting, and if they have any stopped jobs,
280 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
281 */
282static void
283kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
284{
285 struct pid *pgrp = task_pgrp(tsk);
286 struct task_struct *ignored_task = tsk;
287
288 if (!parent)
289 /* exit: our father is in a different pgrp than
290 * we are and we were the only connection outside.
291 */
292 parent = tsk->real_parent;
293 else
294 /* reparent: our child is in a different pgrp than
295 * we are, and it was the only connection outside.
296 */
297 ignored_task = NULL;
298
299 if (task_pgrp(parent) != pgrp &&
300 task_session(parent) == task_session(tsk) &&
301 will_become_orphaned_pgrp(pgrp, ignored_task) &&
302 has_stopped_jobs(pgrp)) {
303 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
304 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
305 }
306}
307
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308/**
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700309 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 *
311 * If a kernel thread is launched as a result of a system call, or if
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700312 * it ever exits, it should generally reparent itself to kthreadd so it
313 * isn't in the way of other processes and is correctly cleaned up on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 *
315 * The various task state such as scheduling policy and priority may have
316 * been inherited from a user process, so we reset them to sane values here.
317 *
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700318 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700320static void reparent_to_kthreadd(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321{
322 write_lock_irq(&tasklist_lock);
323
324 ptrace_unlink(current);
325 /* Reparent to init */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700326 current->real_parent = current->parent = kthreadd_task;
Roland McGrathf4700212008-03-24 18:36:23 -0700327 list_move_tail(&current->sibling, &current->real_parent->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
329 /* Set the exit signal to SIGCHLD so we signal init on exit */
330 current->exit_signal = SIGCHLD;
331
Ingo Molnare05606d2007-07-09 18:51:59 +0200332 if (task_nice(current) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 set_user_nice(current, 0);
334 /* cpus_allowed? */
335 /* rt_priority? */
336 /* signals? */
337 security_task_reparent_to_init(current);
338 memcpy(current->signal->rlim, init_task.signal->rlim,
339 sizeof(current->signal->rlim));
340 atomic_inc(&(INIT_USER->__count));
341 write_unlock_irq(&tasklist_lock);
342 switch_uid(INIT_USER);
343}
344
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800345void __set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346{
Oren Laadane19f2472006-01-08 01:03:58 -0800347 struct task_struct *curr = current->group_leader;
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800348 pid_t nr = pid_nr(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800350 if (task_session(curr) != pid) {
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700351 change_pid(curr, PIDTYPE_SID, pid);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800352 set_task_session(curr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 }
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800354 if (task_pgrp(curr) != pid) {
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700355 change_pid(curr, PIDTYPE_PGID, pid);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800356 set_task_pgrp(curr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 }
358}
359
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800360static void set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361{
362 write_lock_irq(&tasklist_lock);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800363 __set_special_pids(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 write_unlock_irq(&tasklist_lock);
365}
366
367/*
368 * Let kernel threads use this to say that they
369 * allow a certain signal (since daemonize() will
370 * have disabled all of them by default).
371 */
372int allow_signal(int sig)
373{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700374 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 return -EINVAL;
376
377 spin_lock_irq(&current->sighand->siglock);
378 sigdelset(&current->blocked, sig);
379 if (!current->mm) {
380 /* Kernel threads handle their own signals.
381 Let the signal code know it'll be handled, so
382 that they don't get converted to SIGKILL or
383 just silently dropped */
384 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
385 }
386 recalc_sigpending();
387 spin_unlock_irq(&current->sighand->siglock);
388 return 0;
389}
390
391EXPORT_SYMBOL(allow_signal);
392
393int disallow_signal(int sig)
394{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700395 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 return -EINVAL;
397
398 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov10ab8252007-05-09 02:34:37 -0700399 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 recalc_sigpending();
401 spin_unlock_irq(&current->sighand->siglock);
402 return 0;
403}
404
405EXPORT_SYMBOL(disallow_signal);
406
407/*
408 * Put all the gunge required to become a kernel thread without
409 * attached user resources in one place where it belongs.
410 */
411
412void daemonize(const char *name, ...)
413{
414 va_list args;
415 struct fs_struct *fs;
416 sigset_t blocked;
417
418 va_start(args, name);
419 vsnprintf(current->comm, sizeof(current->comm), name, args);
420 va_end(args);
421
422 /*
423 * If we were started as result of loading a module, close all of the
424 * user space pages. We don't need them, and if we didn't close them
425 * they would be locked into memory.
426 */
427 exit_mm(current);
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700428 /*
429 * We don't want to have TIF_FREEZE set if the system-wide hibernation
430 * or suspend transition begins right now.
431 */
Oleg Nesterov7b34e422008-07-25 01:47:37 -0700432 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800434 if (current->nsproxy != &init_nsproxy) {
435 get_nsproxy(&init_nsproxy);
436 switch_task_namespaces(current, &init_nsproxy);
437 }
Oleg Nesterov297bd422008-02-08 04:19:10 -0800438 set_special_pids(&init_struct_pid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -0800439 proc_clear_tty(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
441 /* Block and flush all signals */
442 sigfillset(&blocked);
443 sigprocmask(SIG_BLOCK, &blocked, NULL);
444 flush_signals(current);
445
446 /* Become as one with the init task */
447
448 exit_fs(current); /* current->fs->count--; */
449 fs = init_task.fs;
450 current->fs = fs;
451 atomic_inc(&fs->count);
Serge E. Hallynab516012006-10-02 02:18:06 -0700452
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700453 exit_files(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 current->files = init_task.files;
455 atomic_inc(&current->files->count);
456
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700457 reparent_to_kthreadd();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458}
459
460EXPORT_SYMBOL(daemonize);
461
Arjan van de Ven858119e2006-01-14 13:20:43 -0800462static void close_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463{
464 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700465 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
467 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700468
469 /*
470 * It is safe to dereference the fd table without RCU or
471 * ->file_lock because this is the last reference to the
472 * files structure.
473 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700474 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 for (;;) {
476 unsigned long set;
477 i = j * __NFDBITS;
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800478 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700480 set = fdt->open_fds->fds_bits[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 while (set) {
482 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700483 struct file * file = xchg(&fdt->fd[i], NULL);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800484 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 filp_close(file, files);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800486 cond_resched();
487 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 }
489 i++;
490 set >>= 1;
491 }
492 }
493}
494
495struct files_struct *get_files_struct(struct task_struct *task)
496{
497 struct files_struct *files;
498
499 task_lock(task);
500 files = task->files;
501 if (files)
502 atomic_inc(&files->count);
503 task_unlock(task);
504
505 return files;
506}
507
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800508void put_files_struct(struct files_struct *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700510 struct fdtable *fdt;
511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 if (atomic_dec_and_test(&files->count)) {
513 close_files(files);
514 /*
515 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700516 * If the fdtable was embedded, pass files for freeing
517 * at the end of the RCU grace period. Otherwise,
518 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700520 fdt = files_fdtable(files);
Vadim Lobanov4fd45812006-12-10 02:21:17 -0800521 if (fdt != &files->fdtab)
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700522 kmem_cache_free(files_cachep, files);
Vadim Lobanov01b2d932006-12-22 01:10:43 -0800523 free_fdtable(fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 }
525}
526
Al Viro3b125382008-04-22 05:31:30 -0400527void reset_files_struct(struct files_struct *files)
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700528{
Al Viro3b125382008-04-22 05:31:30 -0400529 struct task_struct *tsk = current;
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700530 struct files_struct *old;
531
532 old = tsk->files;
533 task_lock(tsk);
534 tsk->files = files;
535 task_unlock(tsk);
536 put_files_struct(old);
537}
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700538
Al Viro1ec7f1d2008-04-22 05:35:42 -0400539void exit_files(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540{
541 struct files_struct * files = tsk->files;
542
543 if (files) {
544 task_lock(tsk);
545 tsk->files = NULL;
546 task_unlock(tsk);
547 put_files_struct(files);
548 }
549}
550
Al Viro1ec7f1d2008-04-22 05:35:42 -0400551void put_fs_struct(struct fs_struct *fs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552{
553 /* No need to hold fs->lock if we are killing it */
554 if (atomic_dec_and_test(&fs->count)) {
Jan Blunck6ac08c32008-02-14 19:34:38 -0800555 path_put(&fs->root);
556 path_put(&fs->pwd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 kmem_cache_free(fs_cachep, fs);
558 }
559}
560
Al Viro1ec7f1d2008-04-22 05:35:42 -0400561void exit_fs(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562{
563 struct fs_struct * fs = tsk->fs;
564
565 if (fs) {
566 task_lock(tsk);
567 tsk->fs = NULL;
568 task_unlock(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400569 put_fs_struct(fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 }
571}
572
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573EXPORT_SYMBOL_GPL(exit_fs);
574
Balbir Singhcf475ad2008-04-29 01:00:16 -0700575#ifdef CONFIG_MM_OWNER
576/*
577 * Task p is exiting and it owned mm, lets find a new owner for it
578 */
579static inline int
580mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
581{
582 /*
583 * If there are other users of the mm and the owner (us) is exiting
584 * we need to find a new owner to take on the responsibility.
585 */
Balbir Singhcf475ad2008-04-29 01:00:16 -0700586 if (atomic_read(&mm->mm_users) <= 1)
587 return 0;
588 if (mm->owner != p)
589 return 0;
590 return 1;
591}
592
593void mm_update_next_owner(struct mm_struct *mm)
594{
595 struct task_struct *c, *g, *p = current;
596
597retry:
598 if (!mm_need_new_owner(mm, p))
599 return;
600
601 read_lock(&tasklist_lock);
602 /*
603 * Search in the children
604 */
605 list_for_each_entry(c, &p->children, sibling) {
606 if (c->mm == mm)
607 goto assign_new_owner;
608 }
609
610 /*
611 * Search in the siblings
612 */
613 list_for_each_entry(c, &p->parent->children, sibling) {
614 if (c->mm == mm)
615 goto assign_new_owner;
616 }
617
618 /*
619 * Search through everything else. We should not get
620 * here often
621 */
622 do_each_thread(g, c) {
623 if (c->mm == mm)
624 goto assign_new_owner;
625 } while_each_thread(g, c);
626
627 read_unlock(&tasklist_lock);
Balbir Singh31a78f22008-09-28 23:09:31 +0100628 /*
629 * We found no owner yet mm_users > 1: this implies that we are
630 * most likely racing with swapoff (try_to_unuse()) or /proc or
631 * ptrace or page migration (get_task_mm()). Mark owner as NULL,
632 * so that subsystems can understand the callback and take action.
633 */
634 down_write(&mm->mmap_sem);
635 cgroup_mm_owner_callbacks(mm->owner, NULL);
636 mm->owner = NULL;
637 up_write(&mm->mmap_sem);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700638 return;
639
640assign_new_owner:
641 BUG_ON(c == p);
642 get_task_struct(c);
643 /*
644 * The task_lock protects c->mm from changing.
645 * We always want mm->owner->mm == mm
646 */
647 task_lock(c);
648 /*
649 * Delay read_unlock() till we have the task_lock()
650 * to ensure that c does not slip away underneath us
651 */
652 read_unlock(&tasklist_lock);
653 if (c->mm != mm) {
654 task_unlock(c);
655 put_task_struct(c);
656 goto retry;
657 }
658 cgroup_mm_owner_callbacks(mm->owner, c);
659 mm->owner = c;
660 task_unlock(c);
661 put_task_struct(c);
662}
663#endif /* CONFIG_MM_OWNER */
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665/*
666 * Turn us into a lazy TLB process if we
667 * aren't already..
668 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700669static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670{
671 struct mm_struct *mm = tsk->mm;
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700672 struct core_state *core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673
674 mm_release(tsk, mm);
675 if (!mm)
676 return;
677 /*
678 * Serialize with any possible pending coredump.
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700679 * We must hold mmap_sem around checking core_state
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 * and clearing tsk->mm. The core-inducing thread
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700681 * will increment ->nr_threads for each thread in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 * group with ->mm != NULL.
683 */
684 down_read(&mm->mmap_sem);
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700685 core_state = mm->core_state;
686 if (core_state) {
687 struct core_thread self;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 up_read(&mm->mmap_sem);
Oleg Nesterovc5f1cc82008-07-25 01:47:42 -0700689
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700690 self.task = tsk;
691 self.next = xchg(&core_state->dumper.next, &self);
692 /*
693 * Implies mb(), the result of xchg() must be visible
694 * to core_state->dumper.
695 */
696 if (atomic_dec_and_test(&core_state->nr_threads))
697 complete(&core_state->startup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
Oleg Nesterova94e2d42008-07-25 01:47:46 -0700699 for (;;) {
700 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
701 if (!self.task) /* see coredump_finish() */
702 break;
703 schedule();
704 }
705 __set_task_state(tsk, TASK_RUNNING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 down_read(&mm->mmap_sem);
707 }
708 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700709 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 /* more a memory barrier than a real lock */
711 task_lock(tsk);
712 tsk->mm = NULL;
713 up_read(&mm->mmap_sem);
714 enter_lazy_tlb(mm, current);
Rafael J. Wysocki0c1eecf2007-07-19 01:47:33 -0700715 /* We don't want this task to be frozen prematurely */
716 clear_freeze_flag(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 task_unlock(tsk);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700718 mm_update_next_owner(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 mmput(mm);
720}
721
Roland McGrathf4700212008-03-24 18:36:23 -0700722/*
Roland McGrath666f1642008-04-08 23:12:30 -0700723 * Return nonzero if @parent's children should reap themselves.
724 *
725 * Called with write_lock_irq(&tasklist_lock) held.
726 */
727static int ignoring_children(struct task_struct *parent)
728{
729 int ret;
730 struct sighand_struct *psig = parent->sighand;
731 unsigned long flags;
732 spin_lock_irqsave(&psig->siglock, flags);
733 ret = (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
734 (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT));
735 spin_unlock_irqrestore(&psig->siglock, flags);
736 return ret;
737}
738
739/*
Roland McGrathf4700212008-03-24 18:36:23 -0700740 * Detach all tasks we were using ptrace on.
741 * Any that need to be release_task'd are put on the @dead list.
742 *
743 * Called with write_lock(&tasklist_lock) held.
744 */
745static void ptrace_exit(struct task_struct *parent, struct list_head *dead)
746{
747 struct task_struct *p, *n;
Roland McGrath666f1642008-04-08 23:12:30 -0700748 int ign = -1;
Roland McGrathf4700212008-03-24 18:36:23 -0700749
750 list_for_each_entry_safe(p, n, &parent->ptraced, ptrace_entry) {
751 __ptrace_unlink(p);
752
753 if (p->exit_state != EXIT_ZOMBIE)
754 continue;
755
756 /*
757 * If it's a zombie, our attachedness prevented normal
758 * parent notification or self-reaping. Do notification
759 * now if it would have happened earlier. If it should
760 * reap itself, add it to the @dead list. We can't call
761 * release_task() here because we already hold tasklist_lock.
762 *
763 * If it's our own child, there is no notification to do.
Roland McGrath666f1642008-04-08 23:12:30 -0700764 * But if our normal children self-reap, then this child
765 * was prevented by ptrace and we must reap it now.
Roland McGrathf4700212008-03-24 18:36:23 -0700766 */
767 if (!task_detached(p) && thread_group_empty(p)) {
768 if (!same_thread_group(p->real_parent, parent))
769 do_notify_parent(p, p->exit_signal);
Roland McGrath666f1642008-04-08 23:12:30 -0700770 else {
771 if (ign < 0)
772 ign = ignoring_children(parent);
773 if (ign)
774 p->exit_signal = -1;
775 }
Roland McGrathf4700212008-03-24 18:36:23 -0700776 }
777
778 if (task_detached(p)) {
779 /*
780 * Mark it as in the process of being reaped.
781 */
782 p->exit_state = EXIT_DEAD;
783 list_add(&p->ptrace_entry, dead);
784 }
785 }
786}
787
788/*
789 * Finish up exit-time ptrace cleanup.
790 *
791 * Called without locks.
792 */
793static void ptrace_exit_finish(struct task_struct *parent,
794 struct list_head *dead)
795{
796 struct task_struct *p, *n;
797
798 BUG_ON(!list_empty(&parent->ptraced));
799
800 list_for_each_entry_safe(p, n, dead, ptrace_entry) {
801 list_del_init(&p->ptrace_entry);
802 release_task(p);
803 }
804}
805
806static void reparent_thread(struct task_struct *p, struct task_struct *father)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807{
Oleg Nesterov241ceee2006-12-24 23:30:44 +0300808 if (p->pdeath_signal)
809 /* We already hold the tasklist_lock here. */
810 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
811
Roland McGrathf4700212008-03-24 18:36:23 -0700812 list_move_tail(&p->sibling, &p->real_parent->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700814 /* If this is a threaded reparent there is no need to
815 * notify anyone anything has happened.
816 */
Oleg Nesterov376e1d22008-04-30 00:53:12 -0700817 if (same_thread_group(p->real_parent, father))
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700818 return;
819
820 /* We don't want people slaying init. */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700821 if (!task_detached(p))
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700822 p->exit_signal = SIGCHLD;
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700823
824 /* If we'd notified the old parent about this child's death,
825 * also notify the new parent.
826 */
Roland McGrathf4700212008-03-24 18:36:23 -0700827 if (!ptrace_reparented(p) &&
828 p->exit_state == EXIT_ZOMBIE &&
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700829 !task_detached(p) && thread_group_empty(p))
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700830 do_notify_parent(p, p->exit_signal);
831
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300832 kill_orphaned_pgrp(p, father);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833}
834
835/*
836 * When we die, we re-parent all our children.
837 * Try to give them to another thread in our thread
838 * group, and if no such member exists, give it to
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800839 * the child reaper process (ie "init") in our pid
840 * space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 */
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700842static struct task_struct *find_new_reaper(struct task_struct *father)
843{
844 struct pid_namespace *pid_ns = task_active_pid_ns(father);
845 struct task_struct *thread;
846
847 thread = father;
848 while_each_thread(father, thread) {
849 if (thread->flags & PF_EXITING)
850 continue;
851 if (unlikely(pid_ns->child_reaper == father))
852 pid_ns->child_reaper = thread;
853 return thread;
854 }
855
856 if (unlikely(pid_ns->child_reaper == father)) {
857 write_unlock_irq(&tasklist_lock);
858 if (unlikely(pid_ns == &init_pid_ns))
859 panic("Attempted to kill init!");
860
861 zap_pid_ns_processes(pid_ns);
862 write_lock_irq(&tasklist_lock);
863 /*
864 * We can not clear ->child_reaper or leave it alone.
865 * There may by stealth EXIT_DEAD tasks on ->children,
866 * forget_original_parent() must move them somewhere.
867 */
868 pid_ns->child_reaper = init_pid_ns.child_reaper;
869 }
870
871 return pid_ns->child_reaper;
872}
873
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700874static void forget_original_parent(struct task_struct *father)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700876 struct task_struct *p, *n, *reaper;
Roland McGrathf4700212008-03-24 18:36:23 -0700877 LIST_HEAD(ptrace_dead);
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700878
879 write_lock_irq(&tasklist_lock);
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700880 reaper = find_new_reaper(father);
Roland McGrathf4700212008-03-24 18:36:23 -0700881 /*
882 * First clean up ptrace if we were using it.
883 */
884 ptrace_exit(father, &ptrace_dead);
885
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -0700886 list_for_each_entry_safe(p, n, &father->children, sibling) {
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700887 p->real_parent = reaper;
Roland McGrathf4700212008-03-24 18:36:23 -0700888 if (p->parent == father) {
889 BUG_ON(p->ptrace);
890 p->parent = p->real_parent;
891 }
892 reparent_thread(p, father);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700894
895 write_unlock_irq(&tasklist_lock);
896 BUG_ON(!list_empty(&father->children));
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700897
Roland McGrathf4700212008-03-24 18:36:23 -0700898 ptrace_exit_finish(father, &ptrace_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899}
900
901/*
902 * Send signals to all our closest relatives so that they know
903 * to properly mourn us..
904 */
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300905static void exit_notify(struct task_struct *tsk, int group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906{
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700907 int signal;
908 void *cookie;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 /*
911 * This does two things:
912 *
913 * A. Make init inherit all the child processes
914 * B. Check to see if any process groups have become orphaned
915 * as a result of our exiting, and if they have any stopped
916 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
917 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700918 forget_original_parent(tsk);
Pavel Emelyanov2e4a7072007-10-18 23:40:01 -0700919 exit_task_namespaces(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700921 write_lock_irq(&tasklist_lock);
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300922 if (group_dead)
923 kill_orphaned_pgrp(tsk->group_leader, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924
Oleg Nesterov24728442007-08-04 01:04:41 +0400925 /* Let father know we died
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 *
927 * Thread signals are configurable, but you aren't going to use
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700928 * that to send signals to arbitary processes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 * That stops right now.
930 *
931 * If the parent exec id doesn't match the exec id we saved
932 * when we started then we know the parent has changed security
933 * domain.
934 *
935 * If our self_exec id doesn't match our parent_exec_id then
936 * we have changed execution domain as these two values started
937 * the same after a fork.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700939 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300940 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700941 tsk->self_exec_id != tsk->parent_exec_id) &&
942 !capable(CAP_KILL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 tsk->exit_signal = SIGCHLD;
944
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700945 signal = tracehook_notify_death(tsk, &cookie, group_dead);
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700946 if (signal >= 0)
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700947 signal = do_notify_parent(tsk, signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700949 tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950
Oleg Nesterov2800d8d2008-04-30 00:53:12 -0700951 /* mt-exec, de_thread() is waiting for us */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700952 if (thread_group_leader(tsk) &&
Steve VanDeBogart2633f0e2008-08-26 15:14:36 -0700953 tsk->signal->group_exit_task &&
954 tsk->signal->notify_count < 0)
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700955 wake_up_process(tsk->signal->group_exit_task);
956
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 write_unlock_irq(&tasklist_lock);
958
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700959 tracehook_report_death(tsk, signal, cookie, group_dead);
960
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 /* If the process is dead, release it - nobody will wait for it */
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700962 if (signal == DEATH_REAP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 release_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964}
965
Jeff Dikee18eecb2007-07-15 23:38:48 -0700966#ifdef CONFIG_DEBUG_STACK_USAGE
967static void check_stack_usage(void)
968{
969 static DEFINE_SPINLOCK(low_water_lock);
970 static int lowest_to_date = THREAD_SIZE;
971 unsigned long *n = end_of_stack(current);
972 unsigned long free;
973
974 while (*n == 0)
975 n++;
976 free = (unsigned long)n - (unsigned long)end_of_stack(current);
977
978 if (free >= lowest_to_date)
979 return;
980
981 spin_lock(&low_water_lock);
982 if (free < lowest_to_date) {
983 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
984 "left\n",
985 current->comm, free);
986 lowest_to_date = free;
987 }
988 spin_unlock(&low_water_lock);
989}
990#else
991static inline void check_stack_usage(void) {}
992#endif
993
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800994NORET_TYPE void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995{
996 struct task_struct *tsk = current;
997 int group_dead;
998
999 profile_task_exit(tsk);
1000
Jens Axboe22e2c502005-06-27 10:55:12 +02001001 WARN_ON(atomic_read(&tsk->fs_excl));
1002
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 if (unlikely(in_interrupt()))
1004 panic("Aiee, killing interrupt handler!");
1005 if (unlikely(!tsk->pid))
1006 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007
Roland McGrath30199f52008-07-25 19:45:46 -07001008 tracehook_report_exit(&code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
Alexander Nybergdf164db2005-06-23 00:09:13 -07001010 /*
1011 * We're taking recursive faults here in do_exit. Safest is to just
1012 * leave this task alone and wait for reboot.
1013 */
1014 if (unlikely(tsk->flags & PF_EXITING)) {
1015 printk(KERN_ALERT
1016 "Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001017 /*
1018 * We can do this unlocked here. The futex code uses
1019 * this flag just to verify whether the pi state
1020 * cleanup has been done or not. In the worst case it
1021 * loops once more. We pretend that the cleanup was
1022 * done as there is no way to return. Either the
1023 * OWNER_DIED bit is set by now or we push the blocked
1024 * task into the wait for ever nirwana as well.
1025 */
1026 tsk->flags |= PF_EXITPIDONE;
Al Viroafc847b2006-02-28 12:51:55 -05001027 if (tsk->io_context)
1028 exit_io_context();
Alexander Nybergdf164db2005-06-23 00:09:13 -07001029 set_current_state(TASK_UNINTERRUPTIBLE);
1030 schedule();
1031 }
1032
Oleg Nesterovd12619b2008-02-08 04:19:12 -08001033 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001034 /*
1035 * tsk->flags are checked in the futex code to protect against
1036 * an exiting task cleaning up the robust pi futexes.
1037 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -07001038 smp_mb();
1039 spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 if (unlikely(in_atomic()))
1042 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -07001043 current->comm, task_pid_nr(current),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 preempt_count());
1045
1046 acct_update_integrals(tsk);
Hugh Dickins365e9c872005-10-29 18:16:18 -07001047 if (tsk->mm) {
1048 update_hiwater_rss(tsk->mm);
1049 update_hiwater_vm(tsk->mm);
1050 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -07001052 if (group_dead) {
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001053 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -07001054 exit_itimers(tsk->signal);
Andrew Mortonc3068952005-08-04 16:49:32 -07001055 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -07001056 acct_collect(code, group_dead);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001057#ifdef CONFIG_FUTEX
Ingo Molnar0771dfe2006-03-27 01:16:22 -08001058 if (unlikely(tsk->robust_list))
1059 exit_robust_list(tsk);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001060#ifdef CONFIG_COMPAT
Ingo Molnar34f192c2006-03-27 01:16:24 -08001061 if (unlikely(tsk->compat_robust_list))
1062 compat_exit_robust_list(tsk);
1063#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001064#endif
Miloslav Trmac522ed772007-07-15 23:40:56 -07001065 if (group_dead)
1066 tty_audit_exit();
Al Virofa84cb92006-03-29 20:30:19 -05001067 if (unlikely(tsk->audit_context))
1068 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -08001069
Jonathan Limf2ab6d82007-08-30 23:56:23 -07001070 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -08001071 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -07001072
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 exit_mm(tsk);
1074
KaiGai Kohei0e464812006-06-25 05:49:24 -07001075 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -07001076 acct_process();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 exit_sem(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -04001078 exit_files(tsk);
1079 exit_fs(tsk);
Jeff Dikee18eecb2007-07-15 23:38:48 -07001080 check_stack_usage();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 exit_thread();
Paul Menageb4f48b62007-10-18 23:39:33 -07001082 cgroup_exit(tsk, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 exit_keys(tsk);
1084
1085 if (group_dead && tsk->signal->leader)
1086 disassociate_ctty(1);
1087
Al Viroa1261f542005-11-13 16:06:55 -08001088 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 if (tsk->binfmt)
1090 module_put(tsk->binfmt->module);
1091
Matt Helsley9f460802005-11-07 00:59:16 -08001092 proc_exit_connector(tsk);
Oleg Nesterov821c7de2008-03-02 21:44:44 +03001093 exit_notify(tsk, group_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001095 mpol_put(tsk->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 tsk->mempolicy = NULL;
1097#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001098#ifdef CONFIG_FUTEX
Ingo Molnarde5097c2006-01-09 15:59:21 -08001099 /*
Ingo Molnarc87e2832006-06-27 02:54:58 -07001100 * This must happen late, after the PID is not
1101 * hashed anymore:
1102 */
1103 if (unlikely(!list_empty(&tsk->pi_state_list)))
1104 exit_pi_state_list(tsk);
1105 if (unlikely(current->pi_state_cache))
1106 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001107#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001108 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001109 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -08001110 */
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001111 debug_check_no_locks_held(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001112 /*
1113 * We can do this unlocked here. The futex code uses this flag
1114 * just to verify whether the pi state cleanup has been done
1115 * or not. In the worst case it loops once more.
1116 */
1117 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118
Al Viroafc847b2006-02-28 12:51:55 -05001119 if (tsk->io_context)
1120 exit_io_context();
1121
Jens Axboeb92ce552006-04-11 13:52:07 +02001122 if (tsk->splice_pipe)
1123 __free_pipe_info(tsk->splice_pipe);
1124
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001125 preempt_disable();
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001126 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -07001127 tsk->state = TASK_DEAD;
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001128
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 schedule();
1130 BUG();
1131 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -07001132 for (;;)
1133 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134}
1135
Russ Anderson012914d2005-04-23 00:08:00 -07001136EXPORT_SYMBOL_GPL(do_exit);
1137
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1139{
1140 if (comp)
1141 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001142
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 do_exit(code);
1144}
1145
1146EXPORT_SYMBOL(complete_and_exit);
1147
1148asmlinkage long sys_exit(int error_code)
1149{
1150 do_exit((error_code&0xff)<<8);
1151}
1152
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153/*
1154 * Take down every thread in the group. This is called by fatal signals
1155 * as well as by sys_exit_group (below).
1156 */
1157NORET_TYPE void
1158do_group_exit(int exit_code)
1159{
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001160 struct signal_struct *sig = current->signal;
1161
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1163
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001164 if (signal_group_exit(sig))
1165 exit_code = sig->group_exit_code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 else if (!thread_group_empty(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 struct sighand_struct *const sighand = current->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001169 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 /* Another thread got here before we took the lock. */
1171 exit_code = sig->group_exit_code;
1172 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001174 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 zap_other_threads(current);
1176 }
1177 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 }
1179
1180 do_exit(exit_code);
1181 /* NOTREACHED */
1182}
1183
1184/*
1185 * this kills every thread in the thread group. Note that any externally
1186 * wait4()-ing process will get the correct exit code - even if this
1187 * thread is not the thread group leader.
1188 */
1189asmlinkage void sys_exit_group(int error_code)
1190{
1191 do_group_exit((error_code & 0xff) << 8);
1192}
1193
Eric W. Biederman161550d2008-02-08 04:19:14 -08001194static struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1195{
1196 struct pid *pid = NULL;
1197 if (type == PIDTYPE_PID)
1198 pid = task->pids[type].pid;
1199 else if (type < PIDTYPE_MAX)
1200 pid = task->group_leader->pids[type].pid;
1201 return pid;
1202}
1203
1204static int eligible_child(enum pid_type type, struct pid *pid, int options,
1205 struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206{
Roland McGrath73243282007-05-06 14:50:20 -07001207 int err;
1208
Eric W. Biederman161550d2008-02-08 04:19:14 -08001209 if (type < PIDTYPE_MAX) {
1210 if (task_pid_type(p, type) != pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 return 0;
1212 }
1213
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 /* Wait for all children (clone and not) if __WALL is set;
1215 * otherwise, wait for clone children *only* if __WCLONE is
1216 * set; otherwise, wait for non-clone children *only*. (Note:
1217 * A "clone" child here is one that reports to its parent
1218 * using a signal other than SIGCHLD.) */
1219 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1220 && !(options & __WALL))
1221 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222
Roland McGrath73243282007-05-06 14:50:20 -07001223 err = security_task_wait(p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001224 if (err)
1225 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226
Roland McGrath14dd0b82008-03-30 18:41:25 -07001227 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228}
1229
Ingo Molnar36c8b582006-07-03 00:25:41 -07001230static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 int why, int status,
1232 struct siginfo __user *infop,
1233 struct rusage __user *rusagep)
1234{
1235 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -07001236
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 put_task_struct(p);
1238 if (!retval)
1239 retval = put_user(SIGCHLD, &infop->si_signo);
1240 if (!retval)
1241 retval = put_user(0, &infop->si_errno);
1242 if (!retval)
1243 retval = put_user((short)why, &infop->si_code);
1244 if (!retval)
1245 retval = put_user(pid, &infop->si_pid);
1246 if (!retval)
1247 retval = put_user(uid, &infop->si_uid);
1248 if (!retval)
1249 retval = put_user(status, &infop->si_status);
1250 if (!retval)
1251 retval = pid;
1252 return retval;
1253}
1254
1255/*
1256 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1257 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1258 * the lock and this task is uninteresting. If we return nonzero, we have
1259 * released the lock and the system call should return.
1260 */
Roland McGrath98abed02008-03-19 19:24:59 -07001261static int wait_task_zombie(struct task_struct *p, int options,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 struct siginfo __user *infop,
1263 int __user *stat_addr, struct rusage __user *ru)
1264{
1265 unsigned long state;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001266 int retval, status, traced;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001267 pid_t pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268
Roland McGrath98abed02008-03-19 19:24:59 -07001269 if (!likely(options & WEXITED))
1270 return 0;
1271
1272 if (unlikely(options & WNOWAIT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 uid_t uid = p->uid;
1274 int exit_code = p->exit_code;
1275 int why, status;
1276
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 get_task_struct(p);
1278 read_unlock(&tasklist_lock);
1279 if ((exit_code & 0x7f) == 0) {
1280 why = CLD_EXITED;
1281 status = exit_code >> 8;
1282 } else {
1283 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1284 status = exit_code & 0x7f;
1285 }
1286 return wait_noreap_copyout(p, pid, uid, why,
1287 status, infop, ru);
1288 }
1289
1290 /*
1291 * Try to move the task's state to DEAD
1292 * only one thread is allowed to do this:
1293 */
1294 state = xchg(&p->exit_state, EXIT_DEAD);
1295 if (state != EXIT_ZOMBIE) {
1296 BUG_ON(state != EXIT_DEAD);
1297 return 0;
1298 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
Oleg Nesterov53b6f9f2008-04-30 00:53:13 -07001300 traced = ptrace_reparented(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001301
1302 if (likely(!traced)) {
Jesper Juhl3795e162006-01-09 20:54:39 -08001303 struct signal_struct *psig;
1304 struct signal_struct *sig;
1305
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 /*
1307 * The resource counters for the group leader are in its
1308 * own task_struct. Those for dead threads in the group
1309 * are in its signal_struct, as are those for the child
1310 * processes it has previously reaped. All these
1311 * accumulate in the parent's signal_struct c* fields.
1312 *
1313 * We don't bother to take a lock here to protect these
1314 * p->signal fields, because they are only touched by
1315 * __exit_signal, which runs with tasklist_lock
1316 * write-locked anyway, and so is excluded here. We do
1317 * need to protect the access to p->parent->signal fields,
1318 * as other threads in the parent group can be right
1319 * here reaping other children at the same time.
1320 */
1321 spin_lock_irq(&p->parent->sighand->siglock);
Jesper Juhl3795e162006-01-09 20:54:39 -08001322 psig = p->parent->signal;
1323 sig = p->signal;
1324 psig->cutime =
1325 cputime_add(psig->cutime,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 cputime_add(p->utime,
Jesper Juhl3795e162006-01-09 20:54:39 -08001327 cputime_add(sig->utime,
1328 sig->cutime)));
1329 psig->cstime =
1330 cputime_add(psig->cstime,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 cputime_add(p->stime,
Jesper Juhl3795e162006-01-09 20:54:39 -08001332 cputime_add(sig->stime,
1333 sig->cstime)));
Laurent Vivier9ac52312007-10-15 17:00:19 +02001334 psig->cgtime =
1335 cputime_add(psig->cgtime,
1336 cputime_add(p->gtime,
1337 cputime_add(sig->gtime,
1338 sig->cgtime)));
Jesper Juhl3795e162006-01-09 20:54:39 -08001339 psig->cmin_flt +=
1340 p->min_flt + sig->min_flt + sig->cmin_flt;
1341 psig->cmaj_flt +=
1342 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1343 psig->cnvcsw +=
1344 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1345 psig->cnivcsw +=
1346 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001347 psig->cinblock +=
1348 task_io_get_inblock(p) +
1349 sig->inblock + sig->cinblock;
1350 psig->coublock +=
1351 task_io_get_oublock(p) +
1352 sig->oublock + sig->coublock;
Andrea Righi59954772008-07-27 17:29:15 +02001353 task_io_accounting_add(&psig->ioac, &p->ioac);
1354 task_io_accounting_add(&psig->ioac, &sig->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 spin_unlock_irq(&p->parent->sighand->siglock);
1356 }
1357
1358 /*
1359 * Now we are sure this task is interesting, and no other
1360 * thread can reap it because we set its state to EXIT_DEAD.
1361 */
1362 read_unlock(&tasklist_lock);
1363
1364 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1365 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1366 ? p->signal->group_exit_code : p->exit_code;
1367 if (!retval && stat_addr)
1368 retval = put_user(status, stat_addr);
1369 if (!retval && infop)
1370 retval = put_user(SIGCHLD, &infop->si_signo);
1371 if (!retval && infop)
1372 retval = put_user(0, &infop->si_errno);
1373 if (!retval && infop) {
1374 int why;
1375
1376 if ((status & 0x7f) == 0) {
1377 why = CLD_EXITED;
1378 status >>= 8;
1379 } else {
1380 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1381 status &= 0x7f;
1382 }
1383 retval = put_user((short)why, &infop->si_code);
1384 if (!retval)
1385 retval = put_user(status, &infop->si_status);
1386 }
1387 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001388 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 if (!retval && infop)
1390 retval = put_user(p->uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001391 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001392 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001393
1394 if (traced) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001396 /* We dropped tasklist, ptracer could die and untrace */
1397 ptrace_unlink(p);
1398 /*
1399 * If this is not a detached task, notify the parent.
1400 * If it's still not detached after that, don't release
1401 * it now.
1402 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001403 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001404 do_notify_parent(p, p->exit_signal);
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001405 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001406 p->exit_state = EXIT_ZOMBIE;
1407 p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 }
1409 }
1410 write_unlock_irq(&tasklist_lock);
1411 }
1412 if (p != NULL)
1413 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001414
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 return retval;
1416}
1417
1418/*
1419 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1420 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1421 * the lock and this task is uninteresting. If we return nonzero, we have
1422 * released the lock and the system call should return.
1423 */
Roland McGrathf4700212008-03-24 18:36:23 -07001424static int wait_task_stopped(int ptrace, struct task_struct *p,
Roland McGrath98abed02008-03-19 19:24:59 -07001425 int options, struct siginfo __user *infop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 int __user *stat_addr, struct rusage __user *ru)
1427{
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001428 int retval, exit_code, why;
1429 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001430 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431
Roland McGrathf4700212008-03-24 18:36:23 -07001432 if (!(options & WUNTRACED))
Roland McGrath98abed02008-03-19 19:24:59 -07001433 return 0;
1434
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001435 exit_code = 0;
1436 spin_lock_irq(&p->sighand->siglock);
1437
1438 if (unlikely(!task_is_stopped_or_traced(p)))
1439 goto unlock_sig;
1440
Roland McGrathf4700212008-03-24 18:36:23 -07001441 if (!ptrace && p->signal->group_stop_count > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 /*
1443 * A group stop is in progress and this is the group leader.
1444 * We won't report until all threads have stopped.
1445 */
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001446 goto unlock_sig;
1447
1448 exit_code = p->exit_code;
1449 if (!exit_code)
1450 goto unlock_sig;
1451
Roland McGrath98abed02008-03-19 19:24:59 -07001452 if (!unlikely(options & WNOWAIT))
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001453 p->exit_code = 0;
1454
1455 uid = p->uid;
1456unlock_sig:
1457 spin_unlock_irq(&p->sighand->siglock);
1458 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 return 0;
1460
1461 /*
1462 * Now we are pretty sure this task is interesting.
1463 * Make sure it doesn't get reaped out from under us while we
1464 * give up the lock and then examine it below. We don't want to
1465 * keep holding onto the tasklist_lock while we call getrusage and
1466 * possibly take page faults for user memory.
1467 */
1468 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001469 pid = task_pid_vnr(p);
Roland McGrathf4700212008-03-24 18:36:23 -07001470 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 read_unlock(&tasklist_lock);
1472
Roland McGrath98abed02008-03-19 19:24:59 -07001473 if (unlikely(options & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 return wait_noreap_copyout(p, pid, uid,
Scott James Remnante6ceb322007-11-28 16:22:07 -08001475 why, exit_code,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 infop, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477
1478 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1479 if (!retval && stat_addr)
1480 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1481 if (!retval && infop)
1482 retval = put_user(SIGCHLD, &infop->si_signo);
1483 if (!retval && infop)
1484 retval = put_user(0, &infop->si_errno);
1485 if (!retval && infop)
Roland McGrath6efcae462008-03-08 11:41:22 -08001486 retval = put_user((short)why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 if (!retval && infop)
1488 retval = put_user(exit_code, &infop->si_status);
1489 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001490 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001492 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001494 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 put_task_struct(p);
1496
1497 BUG_ON(!retval);
1498 return retval;
1499}
1500
1501/*
1502 * Handle do_wait work for one task in a live, non-stopped state.
1503 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1504 * the lock and this task is uninteresting. If we return nonzero, we have
1505 * released the lock and the system call should return.
1506 */
Roland McGrath98abed02008-03-19 19:24:59 -07001507static int wait_task_continued(struct task_struct *p, int options,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 struct siginfo __user *infop,
1509 int __user *stat_addr, struct rusage __user *ru)
1510{
1511 int retval;
1512 pid_t pid;
1513 uid_t uid;
1514
Roland McGrath98abed02008-03-19 19:24:59 -07001515 if (!unlikely(options & WCONTINUED))
1516 return 0;
1517
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1519 return 0;
1520
1521 spin_lock_irq(&p->sighand->siglock);
1522 /* Re-check with the lock held. */
1523 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1524 spin_unlock_irq(&p->sighand->siglock);
1525 return 0;
1526 }
Roland McGrath98abed02008-03-19 19:24:59 -07001527 if (!unlikely(options & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1529 spin_unlock_irq(&p->sighand->siglock);
1530
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001531 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 uid = p->uid;
1533 get_task_struct(p);
1534 read_unlock(&tasklist_lock);
1535
1536 if (!infop) {
1537 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1538 put_task_struct(p);
1539 if (!retval && stat_addr)
1540 retval = put_user(0xffff, stat_addr);
1541 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001542 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 } else {
1544 retval = wait_noreap_copyout(p, pid, uid,
1545 CLD_CONTINUED, SIGCONT,
1546 infop, ru);
1547 BUG_ON(retval == 0);
1548 }
1549
1550 return retval;
1551}
1552
Roland McGrath98abed02008-03-19 19:24:59 -07001553/*
1554 * Consider @p for a wait by @parent.
1555 *
1556 * -ECHILD should be in *@notask_error before the first call.
1557 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1558 * Returns zero if the search for a child should continue;
Roland McGrath14dd0b82008-03-30 18:41:25 -07001559 * then *@notask_error is 0 if @p is an eligible child,
1560 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001561 */
Roland McGrathf4700212008-03-24 18:36:23 -07001562static int wait_consider_task(struct task_struct *parent, int ptrace,
Roland McGrath98abed02008-03-19 19:24:59 -07001563 struct task_struct *p, int *notask_error,
1564 enum pid_type type, struct pid *pid, int options,
1565 struct siginfo __user *infop,
1566 int __user *stat_addr, struct rusage __user *ru)
1567{
1568 int ret = eligible_child(type, pid, options, p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001569 if (!ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001570 return ret;
1571
Roland McGrath14dd0b82008-03-30 18:41:25 -07001572 if (unlikely(ret < 0)) {
1573 /*
1574 * If we have not yet seen any eligible child,
1575 * then let this error code replace -ECHILD.
1576 * A permission error will give the user a clue
1577 * to look for security policy problems, rather
1578 * than for mysterious wait bugs.
1579 */
1580 if (*notask_error)
1581 *notask_error = ret;
1582 }
1583
Roland McGrathf4700212008-03-24 18:36:23 -07001584 if (likely(!ptrace) && unlikely(p->ptrace)) {
1585 /*
1586 * This child is hidden by ptrace.
1587 * We aren't allowed to see it now, but eventually we will.
1588 */
1589 *notask_error = 0;
1590 return 0;
1591 }
1592
Roland McGrath98abed02008-03-19 19:24:59 -07001593 if (p->exit_state == EXIT_DEAD)
1594 return 0;
1595
1596 /*
1597 * We don't reap group leaders with subthreads.
1598 */
1599 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
1600 return wait_task_zombie(p, options, infop, stat_addr, ru);
1601
1602 /*
1603 * It's stopped or running now, so it might
1604 * later continue, exit, or stop again.
1605 */
1606 *notask_error = 0;
1607
1608 if (task_is_stopped_or_traced(p))
Roland McGrathf4700212008-03-24 18:36:23 -07001609 return wait_task_stopped(ptrace, p, options,
1610 infop, stat_addr, ru);
Roland McGrath98abed02008-03-19 19:24:59 -07001611
1612 return wait_task_continued(p, options, infop, stat_addr, ru);
1613}
1614
1615/*
1616 * Do the work of do_wait() for one thread in the group, @tsk.
1617 *
1618 * -ECHILD should be in *@notask_error before the first call.
1619 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1620 * Returns zero if the search for a child should continue; then
Roland McGrath14dd0b82008-03-30 18:41:25 -07001621 * *@notask_error is 0 if there were any eligible children,
1622 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001623 */
1624static int do_wait_thread(struct task_struct *tsk, int *notask_error,
1625 enum pid_type type, struct pid *pid, int options,
1626 struct siginfo __user *infop, int __user *stat_addr,
1627 struct rusage __user *ru)
1628{
1629 struct task_struct *p;
1630
1631 list_for_each_entry(p, &tsk->children, sibling) {
Roland McGrathf4700212008-03-24 18:36:23 -07001632 /*
1633 * Do not consider detached threads.
1634 */
1635 if (!task_detached(p)) {
1636 int ret = wait_consider_task(tsk, 0, p, notask_error,
1637 type, pid, options,
1638 infop, stat_addr, ru);
1639 if (ret)
1640 return ret;
1641 }
Roland McGrath98abed02008-03-19 19:24:59 -07001642 }
1643
1644 return 0;
1645}
1646
1647static int ptrace_do_wait(struct task_struct *tsk, int *notask_error,
1648 enum pid_type type, struct pid *pid, int options,
1649 struct siginfo __user *infop, int __user *stat_addr,
1650 struct rusage __user *ru)
1651{
1652 struct task_struct *p;
1653
1654 /*
Roland McGrathf4700212008-03-24 18:36:23 -07001655 * Traditionally we see ptrace'd stopped tasks regardless of options.
Roland McGrath98abed02008-03-19 19:24:59 -07001656 */
Roland McGrathf4700212008-03-24 18:36:23 -07001657 options |= WUNTRACED;
Roland McGrath98abed02008-03-19 19:24:59 -07001658
Roland McGrathf4700212008-03-24 18:36:23 -07001659 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1660 int ret = wait_consider_task(tsk, 1, p, notask_error,
1661 type, pid, options,
1662 infop, stat_addr, ru);
1663 if (ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001664 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001665 }
1666
1667 return 0;
1668}
1669
Eric W. Biederman161550d2008-02-08 04:19:14 -08001670static long do_wait(enum pid_type type, struct pid *pid, int options,
1671 struct siginfo __user *infop, int __user *stat_addr,
1672 struct rusage __user *ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673{
1674 DECLARE_WAITQUEUE(wait, current);
1675 struct task_struct *tsk;
Roland McGrath98abed02008-03-19 19:24:59 -07001676 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
1678 add_wait_queue(&current->signal->wait_chldexit,&wait);
1679repeat:
Roland McGrath98abed02008-03-19 19:24:59 -07001680 /*
1681 * If there is nothing that can match our critiera just get out.
1682 * We will clear @retval to zero if we see any child that might later
1683 * match our criteria, even if we are not able to reap it yet.
1684 */
Eric W. Biederman161550d2008-02-08 04:19:14 -08001685 retval = -ECHILD;
1686 if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type])))
1687 goto end;
1688
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 current->state = TASK_INTERRUPTIBLE;
1690 read_lock(&tasklist_lock);
1691 tsk = current;
1692 do {
Roland McGrath98abed02008-03-19 19:24:59 -07001693 int tsk_result = do_wait_thread(tsk, &retval,
1694 type, pid, options,
1695 infop, stat_addr, ru);
1696 if (!tsk_result)
1697 tsk_result = ptrace_do_wait(tsk, &retval,
1698 type, pid, options,
1699 infop, stat_addr, ru);
1700 if (tsk_result) {
1701 /*
1702 * tasklist_lock is unlocked and we have a final result.
1703 */
1704 retval = tsk_result;
1705 goto end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 }
Roland McGrath98abed02008-03-19 19:24:59 -07001707
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 if (options & __WNOTHREAD)
1709 break;
1710 tsk = next_thread(tsk);
Eric Sesterhenn125e1872006-06-23 02:06:06 -07001711 BUG_ON(tsk->signal != current->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 } while (tsk != current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001714
Roland McGrath98abed02008-03-19 19:24:59 -07001715 if (!retval && !(options & WNOHANG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 retval = -ERESTARTSYS;
Roland McGrath98abed02008-03-19 19:24:59 -07001717 if (!signal_pending(current)) {
1718 schedule();
1719 goto repeat;
1720 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 }
Roland McGrath98abed02008-03-19 19:24:59 -07001722
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723end:
1724 current->state = TASK_RUNNING;
1725 remove_wait_queue(&current->signal->wait_chldexit,&wait);
1726 if (infop) {
1727 if (retval > 0)
Oleg Nesterov9cbab812008-02-08 04:19:02 -08001728 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 else {
1730 /*
1731 * For a WNOHANG return, clear out all the fields
1732 * we would set so the user can easily tell the
1733 * difference.
1734 */
1735 if (!retval)
1736 retval = put_user(0, &infop->si_signo);
1737 if (!retval)
1738 retval = put_user(0, &infop->si_errno);
1739 if (!retval)
1740 retval = put_user(0, &infop->si_code);
1741 if (!retval)
1742 retval = put_user(0, &infop->si_pid);
1743 if (!retval)
1744 retval = put_user(0, &infop->si_uid);
1745 if (!retval)
1746 retval = put_user(0, &infop->si_status);
1747 }
1748 }
1749 return retval;
1750}
1751
Eric W. Biederman161550d2008-02-08 04:19:14 -08001752asmlinkage long sys_waitid(int which, pid_t upid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 struct siginfo __user *infop, int options,
1754 struct rusage __user *ru)
1755{
Eric W. Biederman161550d2008-02-08 04:19:14 -08001756 struct pid *pid = NULL;
1757 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 long ret;
1759
1760 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1761 return -EINVAL;
1762 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1763 return -EINVAL;
1764
1765 switch (which) {
1766 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001767 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 break;
1769 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001770 type = PIDTYPE_PID;
1771 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 return -EINVAL;
1773 break;
1774 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001775 type = PIDTYPE_PGID;
1776 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 break;
1779 default:
1780 return -EINVAL;
1781 }
1782
Eric W. Biederman161550d2008-02-08 04:19:14 -08001783 if (type < PIDTYPE_MAX)
1784 pid = find_get_pid(upid);
1785 ret = do_wait(type, pid, options, infop, NULL, ru);
1786 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787
1788 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001789 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 return ret;
1791}
1792
Eric W. Biederman161550d2008-02-08 04:19:14 -08001793asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 int options, struct rusage __user *ru)
1795{
Eric W. Biederman161550d2008-02-08 04:19:14 -08001796 struct pid *pid = NULL;
1797 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 long ret;
1799
1800 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1801 __WNOTHREAD|__WCLONE|__WALL))
1802 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001803
1804 if (upid == -1)
1805 type = PIDTYPE_MAX;
1806 else if (upid < 0) {
1807 type = PIDTYPE_PGID;
1808 pid = find_get_pid(-upid);
1809 } else if (upid == 0) {
1810 type = PIDTYPE_PGID;
1811 pid = get_pid(task_pgrp(current));
1812 } else /* upid > 0 */ {
1813 type = PIDTYPE_PID;
1814 pid = find_get_pid(upid);
1815 }
1816
1817 ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru);
1818 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819
1820 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001821 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822 return ret;
1823}
1824
1825#ifdef __ARCH_WANT_SYS_WAITPID
1826
1827/*
1828 * sys_waitpid() remains for compatibility. waitpid() should be
1829 * implemented by calling sys_wait4() from libc.a.
1830 */
1831asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1832{
1833 return sys_wait4(pid, stat_addr, options, NULL);
1834}
1835
1836#endif