blob: 506a957b665a69bbf84794e233b8032f2d98b280 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/key.h>
17#include <linux/security.h>
18#include <linux/cpu.h>
19#include <linux/acct.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070020#include <linux/tsacct_kern.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/file.h>
22#include <linux/binfmts.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070023#include <linux/nsproxy.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080024#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/ptrace.h>
26#include <linux/profile.h>
27#include <linux/mount.h>
28#include <linux/proc_fs.h>
Eric W. Biederman49d769d2007-05-09 02:34:33 -070029#include <linux/kthread.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/mempolicy.h>
Shailabh Nagarc7572492006-07-14 00:24:40 -070031#include <linux/taskstats_kern.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070032#include <linux/delayacct.h>
Rafael J. Wysocki83144182007-07-17 04:03:35 -070033#include <linux/freezer.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070034#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035#include <linux/syscalls.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070036#include <linux/signal.h>
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080037#include <linux/posix-timers.h>
Matt Helsley9f460802005-11-07 00:59:16 -080038#include <linux/cn_proc.h>
Ingo Molnarde5097c2006-01-09 15:59:21 -080039#include <linux/mutex.h>
Ingo Molnar0771dfe2006-03-27 01:16:22 -080040#include <linux/futex.h>
Ingo Molnar34f192c2006-03-27 01:16:24 -080041#include <linux/compat.h>
Jens Axboeb92ce552006-04-11 13:52:07 +020042#include <linux/pipe_fs_i.h>
Al Virofa84cb92006-03-29 20:30:19 -050043#include <linux/audit.h> /* for audit_free() */
Adrian Bunk83cc5ed2006-06-25 05:47:41 -070044#include <linux/resource.h>
David Howells0d67a462006-08-29 19:05:56 +010045#include <linux/blkdev.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070046#include <linux/task_io_accounting_ops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
48#include <asm/uaccess.h>
49#include <asm/unistd.h>
50#include <asm/pgtable.h>
51#include <asm/mmu_context.h>
52
Adrian Bunk408b6642005-05-01 08:59:29 -070053static void exit_mm(struct task_struct * tsk);
54
Linus Torvalds1da177e2005-04-16 15:20:36 -070055static void __unhash_process(struct task_struct *p)
56{
57 nr_threads--;
58 detach_pid(p, PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 if (thread_group_leader(p)) {
60 detach_pid(p, PIDTYPE_PGID);
61 detach_pid(p, PIDTYPE_SID);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080062
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -070063 list_del_rcu(&p->tasks);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -080064 __get_cpu_var(process_counts)--;
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 }
Oleg Nesterov47e65322006-03-28 16:11:25 -080066 list_del_rcu(&p->thread_group);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080067 remove_parent(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -070068}
69
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080070/*
71 * This function expects the tasklist_lock write-locked.
72 */
73static void __exit_signal(struct task_struct *tsk)
74{
75 struct signal_struct *sig = tsk->signal;
76 struct sighand_struct *sighand;
77
78 BUG_ON(!sig);
79 BUG_ON(!atomic_read(&sig->count));
80
81 rcu_read_lock();
82 sighand = rcu_dereference(tsk->sighand);
83 spin_lock(&sighand->siglock);
84
85 posix_cpu_timers_exit(tsk);
86 if (atomic_dec_and_test(&sig->count))
87 posix_cpu_timers_exit_group(tsk);
88 else {
89 /*
90 * If there is any task waiting for the group exit
91 * then notify it:
92 */
Oleg Nesterov6db840f2007-10-16 23:27:23 -070093 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count)
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080094 wake_up_process(sig->group_exit_task);
Oleg Nesterov6db840f2007-10-16 23:27:23 -070095
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080096 if (tsk == sig->curr_target)
97 sig->curr_target = next_thread(tsk);
98 /*
99 * Accumulate here the counters for all threads but the
100 * group leader as they die, so they can be added into
101 * the process-wide totals when those are taken.
102 * The group leader stays around as a zombie as long
103 * as there are other threads. When it gets reaped,
104 * the exit.c code will add its counts into these totals.
105 * We won't ever get here for the group leader, since it
106 * will have been the last reference on the signal_struct.
107 */
108 sig->utime = cputime_add(sig->utime, tsk->utime);
109 sig->stime = cputime_add(sig->stime, tsk->stime);
Laurent Vivier9ac52312007-10-15 17:00:19 +0200110 sig->gtime = cputime_add(sig->gtime, tsk->gtime);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800111 sig->min_flt += tsk->min_flt;
112 sig->maj_flt += tsk->maj_flt;
113 sig->nvcsw += tsk->nvcsw;
114 sig->nivcsw += tsk->nivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700115 sig->inblock += task_io_get_inblock(tsk);
116 sig->oublock += task_io_get_oublock(tsk);
Balbir Singh172ba842007-07-09 18:52:00 +0200117 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800118 sig = NULL; /* Marker for below. */
119 }
120
Oleg Nesterov58767002006-03-28 16:11:20 -0800121 __unhash_process(tsk);
122
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800123 tsk->signal = NULL;
Oleg Nesterova7e53282006-03-28 16:11:27 -0800124 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800125 spin_unlock(&sighand->siglock);
126 rcu_read_unlock();
127
Oleg Nesterova7e53282006-03-28 16:11:27 -0800128 __cleanup_sighand(sighand);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800129 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
130 flush_sigqueue(&tsk->pending);
131 if (sig) {
132 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov093a8e82006-10-28 10:38:51 -0700133 taskstats_tgid_free(sig);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800134 __cleanup_signal(sig);
135 }
136}
137
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800138static void delayed_put_task_struct(struct rcu_head *rhp)
139{
140 put_task_struct(container_of(rhp, struct task_struct, rcu));
141}
142
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143void release_task(struct task_struct * p)
144{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700145 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800147repeat:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 atomic_dec(&p->user->processes);
Pavel Emelyanov60347f62007-10-18 23:40:03 -0700149 proc_flush_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 write_lock_irq(&tasklist_lock);
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800151 ptrace_unlink(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
153 __exit_signal(p);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -0800154
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 /*
156 * If we are the last non-leader member of the thread
157 * group, and the leader is zombie, then notify the
158 * group leader's parent process. (if it wants notification.)
159 */
160 zap_leader = 0;
161 leader = p->group_leader;
162 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
163 BUG_ON(leader->exit_signal == -1);
164 do_notify_parent(leader, leader->exit_signal);
165 /*
166 * If we were the last child thread and the leader has
167 * exited already, and the leader's parent ignores SIGCHLD,
168 * then we are the one who should release the leader.
169 *
170 * do_notify_parent() will have marked it self-reaping in
171 * that case.
172 */
173 zap_leader = (leader->exit_signal == -1);
174 }
175
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 release_thread(p);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800178 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179
180 p = leader;
181 if (unlikely(zap_leader))
182 goto repeat;
183}
184
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185/*
186 * This checks not only the pgrp, but falls back on the pid if no
187 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
188 * without this...
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800189 *
190 * The caller must hold rcu lock or the tasklist lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 */
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800192struct pid *session_of_pgrp(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193{
194 struct task_struct *p;
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800195 struct pid *sid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800197 p = pid_task(pgrp, PIDTYPE_PGID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800198 if (p == NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800199 p = pid_task(pgrp, PIDTYPE_PID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800200 if (p != NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800201 sid = task_session(p);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800202
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 return sid;
204}
205
206/*
207 * Determine if a process group is "orphaned", according to the POSIX
208 * definition in 2.2.2.52. Orphaned process groups are not to be affected
209 * by terminal-generated stop signals. Newly orphaned process groups are
210 * to receive a SIGHUP and a SIGCONT.
211 *
212 * "I ask you, have you ever known what it is to be an orphan?"
213 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800214static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
216 struct task_struct *p;
217 int ret = 1;
218
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800219 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 if (p == ignored_task
221 || p->exit_state
Serge E. Hallynb460cbc2007-10-18 23:39:52 -0700222 || is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 continue;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800224 if (task_pgrp(p->real_parent) != pgrp &&
225 task_session(p->real_parent) == task_session(p)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 ret = 0;
227 break;
228 }
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800229 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 return ret; /* (sighing) "Often!" */
231}
232
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800233int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234{
235 int retval;
236
237 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800238 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 read_unlock(&tasklist_lock);
240
241 return retval;
242}
243
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800244static int has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245{
246 int retval = 0;
247 struct task_struct *p;
248
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800249 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Matthew Wilcox338077e2007-12-06 11:09:35 -0500250 if (!task_is_stopped(p))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 retval = 1;
253 break;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800254 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 return retval;
256}
257
258/**
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700259 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 *
261 * If a kernel thread is launched as a result of a system call, or if
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700262 * it ever exits, it should generally reparent itself to kthreadd so it
263 * isn't in the way of other processes and is correctly cleaned up on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 *
265 * The various task state such as scheduling policy and priority may have
266 * been inherited from a user process, so we reset them to sane values here.
267 *
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700268 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700270static void reparent_to_kthreadd(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271{
272 write_lock_irq(&tasklist_lock);
273
274 ptrace_unlink(current);
275 /* Reparent to init */
Oleg Nesterov9b678ec2006-03-28 16:11:05 -0800276 remove_parent(current);
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700277 current->real_parent = current->parent = kthreadd_task;
Oleg Nesterov9b678ec2006-03-28 16:11:05 -0800278 add_parent(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279
280 /* Set the exit signal to SIGCHLD so we signal init on exit */
281 current->exit_signal = SIGCHLD;
282
Ingo Molnare05606d2007-07-09 18:51:59 +0200283 if (task_nice(current) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 set_user_nice(current, 0);
285 /* cpus_allowed? */
286 /* rt_priority? */
287 /* signals? */
288 security_task_reparent_to_init(current);
289 memcpy(current->signal->rlim, init_task.signal->rlim,
290 sizeof(current->signal->rlim));
291 atomic_inc(&(INIT_USER->__count));
292 write_unlock_irq(&tasklist_lock);
293 switch_uid(INIT_USER);
294}
295
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800296void __set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297{
Oren Laadane19f2472006-01-08 01:03:58 -0800298 struct task_struct *curr = current->group_leader;
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800299 pid_t nr = pid_nr(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800301 if (task_session(curr) != pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 detach_pid(curr, PIDTYPE_SID);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800303 attach_pid(curr, PIDTYPE_SID, pid);
304 set_task_session(curr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 }
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800306 if (task_pgrp(curr) != pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 detach_pid(curr, PIDTYPE_PGID);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800308 attach_pid(curr, PIDTYPE_PGID, pid);
309 set_task_pgrp(curr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 }
311}
312
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800313static void set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314{
315 write_lock_irq(&tasklist_lock);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800316 __set_special_pids(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 write_unlock_irq(&tasklist_lock);
318}
319
320/*
321 * Let kernel threads use this to say that they
322 * allow a certain signal (since daemonize() will
323 * have disabled all of them by default).
324 */
325int allow_signal(int sig)
326{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700327 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 return -EINVAL;
329
330 spin_lock_irq(&current->sighand->siglock);
331 sigdelset(&current->blocked, sig);
332 if (!current->mm) {
333 /* Kernel threads handle their own signals.
334 Let the signal code know it'll be handled, so
335 that they don't get converted to SIGKILL or
336 just silently dropped */
337 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
338 }
339 recalc_sigpending();
340 spin_unlock_irq(&current->sighand->siglock);
341 return 0;
342}
343
344EXPORT_SYMBOL(allow_signal);
345
346int disallow_signal(int sig)
347{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700348 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 return -EINVAL;
350
351 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov10ab8252007-05-09 02:34:37 -0700352 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 recalc_sigpending();
354 spin_unlock_irq(&current->sighand->siglock);
355 return 0;
356}
357
358EXPORT_SYMBOL(disallow_signal);
359
360/*
361 * Put all the gunge required to become a kernel thread without
362 * attached user resources in one place where it belongs.
363 */
364
365void daemonize(const char *name, ...)
366{
367 va_list args;
368 struct fs_struct *fs;
369 sigset_t blocked;
370
371 va_start(args, name);
372 vsnprintf(current->comm, sizeof(current->comm), name, args);
373 va_end(args);
374
375 /*
376 * If we were started as result of loading a module, close all of the
377 * user space pages. We don't need them, and if we didn't close them
378 * they would be locked into memory.
379 */
380 exit_mm(current);
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700381 /*
382 * We don't want to have TIF_FREEZE set if the system-wide hibernation
383 * or suspend transition begins right now.
384 */
385 current->flags |= PF_NOFREEZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800387 if (current->nsproxy != &init_nsproxy) {
388 get_nsproxy(&init_nsproxy);
389 switch_task_namespaces(current, &init_nsproxy);
390 }
Oleg Nesterov297bd422008-02-08 04:19:10 -0800391 set_special_pids(&init_struct_pid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -0800392 proc_clear_tty(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393
394 /* Block and flush all signals */
395 sigfillset(&blocked);
396 sigprocmask(SIG_BLOCK, &blocked, NULL);
397 flush_signals(current);
398
399 /* Become as one with the init task */
400
401 exit_fs(current); /* current->fs->count--; */
402 fs = init_task.fs;
403 current->fs = fs;
404 atomic_inc(&fs->count);
Serge E. Hallynab516012006-10-02 02:18:06 -0700405
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700406 exit_files(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 current->files = init_task.files;
408 atomic_inc(&current->files->count);
409
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700410 reparent_to_kthreadd();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411}
412
413EXPORT_SYMBOL(daemonize);
414
Arjan van de Ven858119e2006-01-14 13:20:43 -0800415static void close_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416{
417 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700418 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419
420 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700421
422 /*
423 * It is safe to dereference the fd table without RCU or
424 * ->file_lock because this is the last reference to the
425 * files structure.
426 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700427 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 for (;;) {
429 unsigned long set;
430 i = j * __NFDBITS;
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800431 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700433 set = fdt->open_fds->fds_bits[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 while (set) {
435 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700436 struct file * file = xchg(&fdt->fd[i], NULL);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800437 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 filp_close(file, files);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800439 cond_resched();
440 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 }
442 i++;
443 set >>= 1;
444 }
445 }
446}
447
448struct files_struct *get_files_struct(struct task_struct *task)
449{
450 struct files_struct *files;
451
452 task_lock(task);
453 files = task->files;
454 if (files)
455 atomic_inc(&files->count);
456 task_unlock(task);
457
458 return files;
459}
460
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800461void put_files_struct(struct files_struct *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700463 struct fdtable *fdt;
464
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 if (atomic_dec_and_test(&files->count)) {
466 close_files(files);
467 /*
468 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700469 * If the fdtable was embedded, pass files for freeing
470 * at the end of the RCU grace period. Otherwise,
471 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700473 fdt = files_fdtable(files);
Vadim Lobanov4fd45812006-12-10 02:21:17 -0800474 if (fdt != &files->fdtab)
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700475 kmem_cache_free(files_cachep, files);
Vadim Lobanov01b2d932006-12-22 01:10:43 -0800476 free_fdtable(fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 }
478}
479
480EXPORT_SYMBOL(put_files_struct);
481
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700482void reset_files_struct(struct task_struct *tsk, struct files_struct *files)
483{
484 struct files_struct *old;
485
486 old = tsk->files;
487 task_lock(tsk);
488 tsk->files = files;
489 task_unlock(tsk);
490 put_files_struct(old);
491}
492EXPORT_SYMBOL(reset_files_struct);
493
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700494static void __exit_files(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495{
496 struct files_struct * files = tsk->files;
497
498 if (files) {
499 task_lock(tsk);
500 tsk->files = NULL;
501 task_unlock(tsk);
502 put_files_struct(files);
503 }
504}
505
506void exit_files(struct task_struct *tsk)
507{
508 __exit_files(tsk);
509}
510
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700511static void __put_fs_struct(struct fs_struct *fs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512{
513 /* No need to hold fs->lock if we are killing it */
514 if (atomic_dec_and_test(&fs->count)) {
Jan Blunck6ac08c32008-02-14 19:34:38 -0800515 path_put(&fs->root);
516 path_put(&fs->pwd);
517 if (fs->altroot.dentry)
518 path_put(&fs->altroot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 kmem_cache_free(fs_cachep, fs);
520 }
521}
522
523void put_fs_struct(struct fs_struct *fs)
524{
525 __put_fs_struct(fs);
526}
527
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700528static void __exit_fs(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529{
530 struct fs_struct * fs = tsk->fs;
531
532 if (fs) {
533 task_lock(tsk);
534 tsk->fs = NULL;
535 task_unlock(tsk);
536 __put_fs_struct(fs);
537 }
538}
539
540void exit_fs(struct task_struct *tsk)
541{
542 __exit_fs(tsk);
543}
544
545EXPORT_SYMBOL_GPL(exit_fs);
546
547/*
548 * Turn us into a lazy TLB process if we
549 * aren't already..
550 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700551static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552{
553 struct mm_struct *mm = tsk->mm;
554
555 mm_release(tsk, mm);
556 if (!mm)
557 return;
558 /*
559 * Serialize with any possible pending coredump.
560 * We must hold mmap_sem around checking core_waiters
561 * and clearing tsk->mm. The core-inducing thread
562 * will increment core_waiters for each thread in the
563 * group with ->mm != NULL.
564 */
565 down_read(&mm->mmap_sem);
566 if (mm->core_waiters) {
567 up_read(&mm->mmap_sem);
568 down_write(&mm->mmap_sem);
569 if (!--mm->core_waiters)
570 complete(mm->core_startup_done);
571 up_write(&mm->mmap_sem);
572
573 wait_for_completion(&mm->core_done);
574 down_read(&mm->mmap_sem);
575 }
576 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700577 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 /* more a memory barrier than a real lock */
579 task_lock(tsk);
580 tsk->mm = NULL;
581 up_read(&mm->mmap_sem);
582 enter_lazy_tlb(mm, current);
Rafael J. Wysocki0c1eecf2007-07-19 01:47:33 -0700583 /* We don't want this task to be frozen prematurely */
584 clear_freeze_flag(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 task_unlock(tsk);
586 mmput(mm);
587}
588
Ingo Molnar36c8b582006-07-03 00:25:41 -0700589static void
590reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591{
Oleg Nesterov241ceee2006-12-24 23:30:44 +0300592 if (p->pdeath_signal)
593 /* We already hold the tasklist_lock here. */
594 group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
595
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 /* Move the child from its dying parent to the new one. */
597 if (unlikely(traced)) {
598 /* Preserve ptrace links if someone else is tracing this child. */
599 list_del_init(&p->ptrace_list);
600 if (p->parent != p->real_parent)
601 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
602 } else {
603 /* If this child is being traced, then we're the one tracing it
604 * anyway, so let go of it.
605 */
606 p->ptrace = 0;
Oleg Nesterov6ac781b2006-03-28 16:11:09 -0800607 remove_parent(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 p->parent = p->real_parent;
Oleg Nesterov6ac781b2006-03-28 16:11:09 -0800609 add_parent(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610
Matthew Wilcox338077e2007-12-06 11:09:35 -0500611 if (task_is_traced(p)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 /*
613 * If it was at a trace stop, turn it into
614 * a normal stop since it's no longer being
615 * traced.
616 */
617 ptrace_untrace(p);
618 }
619 }
620
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700621 /* If this is a threaded reparent there is no need to
622 * notify anyone anything has happened.
623 */
624 if (p->real_parent->group_leader == father->group_leader)
625 return;
626
627 /* We don't want people slaying init. */
628 if (p->exit_signal != -1)
629 p->exit_signal = SIGCHLD;
Eric W. Biedermanb2b2cbc2006-12-21 21:28:40 -0700630
631 /* If we'd notified the old parent about this child's death,
632 * also notify the new parent.
633 */
634 if (!traced && p->exit_state == EXIT_ZOMBIE &&
635 p->exit_signal != -1 && thread_group_empty(p))
636 do_notify_parent(p, p->exit_signal);
637
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 /*
639 * process group orphan check
640 * Case ii: Our child is in a different pgrp
641 * than we are, and it was the only connection
642 * outside, so the child pgrp is now orphaned.
643 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800644 if ((task_pgrp(p) != task_pgrp(father)) &&
645 (task_session(p) == task_session(father))) {
646 struct pid *pgrp = task_pgrp(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647
Cedric Le Goater937949d2006-12-08 02:37:54 -0800648 if (will_become_orphaned_pgrp(pgrp, NULL) &&
649 has_stopped_jobs(pgrp)) {
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800650 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
651 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 }
653 }
654}
655
656/*
657 * When we die, we re-parent all our children.
658 * Try to give them to another thread in our thread
659 * group, and if no such member exists, give it to
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800660 * the child reaper process (ie "init") in our pid
661 * space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700663static void forget_original_parent(struct task_struct *father)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664{
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -0700665 struct task_struct *p, *n, *reaper = father;
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700666 struct list_head ptrace_dead;
667
668 INIT_LIST_HEAD(&ptrace_dead);
669
670 write_lock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
672 do {
673 reaper = next_thread(reaper);
674 if (reaper == father) {
Sukadev Bhattiprolu88f21d82007-10-18 23:39:50 -0700675 reaper = task_child_reaper(father);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 break;
677 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700678 } while (reaper->flags & PF_EXITING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679
680 /*
681 * There are only two places where our children can be:
682 *
683 * - in our child list
684 * - in our ptraced child list
685 *
686 * Search them and reparent children.
687 */
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -0700688 list_for_each_entry_safe(p, n, &father->children, sibling) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 int ptrace;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
691 ptrace = p->ptrace;
692
693 /* if father isn't the real parent, then ptrace must be enabled */
694 BUG_ON(father != p->real_parent && !ptrace);
695
696 if (father == p->real_parent) {
697 /* reparent with a reaper, real father it's us */
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700698 p->real_parent = reaper;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 reparent_thread(p, father, 0);
700 } else {
701 /* reparent ptraced task to its real parent */
702 __ptrace_unlink (p);
703 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
704 thread_group_empty(p))
705 do_notify_parent(p, p->exit_signal);
706 }
707
708 /*
709 * if the ptraced child is a zombie with exit_signal == -1
710 * we must collect it before we exit, or it will remain
711 * zombie forever since we prevented it from self-reap itself
712 * while it was being traced by us, to be able to see it in wait4.
713 */
714 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700715 list_add(&p->ptrace_list, &ptrace_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700717
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -0700718 list_for_each_entry_safe(p, n, &father->ptrace_children, ptrace_list) {
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700719 p->real_parent = reaper;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 reparent_thread(p, father, 1);
721 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700722
723 write_unlock_irq(&tasklist_lock);
724 BUG_ON(!list_empty(&father->children));
725 BUG_ON(!list_empty(&father->ptrace_children));
726
727 list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_list) {
728 list_del_init(&p->ptrace_list);
729 release_task(p);
730 }
731
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732}
733
734/*
735 * Send signals to all our closest relatives so that they know
736 * to properly mourn us..
737 */
738static void exit_notify(struct task_struct *tsk)
739{
740 int state;
741 struct task_struct *t;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800742 struct pid *pgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 /*
745 * This does two things:
746 *
747 * A. Make init inherit all the child processes
748 * B. Check to see if any process groups have become orphaned
749 * as a result of our exiting, and if they have any stopped
750 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
751 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700752 forget_original_parent(tsk);
Pavel Emelyanov2e4a7072007-10-18 23:40:01 -0700753 exit_task_namespaces(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700755 write_lock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 /*
757 * Check to see if any process groups have become orphaned
758 * as a result of our exiting, and if they have any stopped
759 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
760 *
761 * Case i: Our father is in a different pgrp than we are
762 * and we were the only connection outside, so our pgrp
763 * is about to become orphaned.
764 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 t = tsk->real_parent;
Oleg Nesterov3ae4cba2007-10-16 23:26:58 -0700766
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800767 pgrp = task_pgrp(tsk);
768 if ((task_pgrp(t) != pgrp) &&
Eric W. Biederman14e9d572007-03-28 23:38:16 -0600769 (task_session(t) == task_session(tsk)) &&
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800770 will_become_orphaned_pgrp(pgrp, tsk) &&
771 has_stopped_jobs(pgrp)) {
772 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
773 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 }
775
Oleg Nesterov24728442007-08-04 01:04:41 +0400776 /* Let father know we died
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 *
778 * Thread signals are configurable, but you aren't going to use
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700779 * that to send signals to arbitary processes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 * That stops right now.
781 *
782 * If the parent exec id doesn't match the exec id we saved
783 * when we started then we know the parent has changed security
784 * domain.
785 *
786 * If our self_exec id doesn't match our parent_exec_id then
787 * we have changed execution domain as these two values started
788 * the same after a fork.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
791 ( tsk->parent_exec_id != t->self_exec_id ||
792 tsk->self_exec_id != tsk->parent_exec_id)
793 && !capable(CAP_KILL))
794 tsk->exit_signal = SIGCHLD;
795
796
797 /* If something other than our normal parent is ptracing us, then
798 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
799 * only has special meaning to our real parent.
800 */
801 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
802 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
803 do_notify_parent(tsk, signal);
804 } else if (tsk->ptrace) {
805 do_notify_parent(tsk, SIGCHLD);
806 }
807
808 state = EXIT_ZOMBIE;
Oleg Nesterov24728442007-08-04 01:04:41 +0400809 if (tsk->exit_signal == -1 && likely(!tsk->ptrace))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 state = EXIT_DEAD;
811 tsk->exit_state = state;
812
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700813 if (thread_group_leader(tsk) &&
814 tsk->signal->notify_count < 0 &&
815 tsk->signal->group_exit_task)
816 wake_up_process(tsk->signal->group_exit_task);
817
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 write_unlock_irq(&tasklist_lock);
819
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 /* If the process is dead, release it - nobody will wait for it */
821 if (state == EXIT_DEAD)
822 release_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823}
824
Jeff Dikee18eecb2007-07-15 23:38:48 -0700825#ifdef CONFIG_DEBUG_STACK_USAGE
826static void check_stack_usage(void)
827{
828 static DEFINE_SPINLOCK(low_water_lock);
829 static int lowest_to_date = THREAD_SIZE;
830 unsigned long *n = end_of_stack(current);
831 unsigned long free;
832
833 while (*n == 0)
834 n++;
835 free = (unsigned long)n - (unsigned long)end_of_stack(current);
836
837 if (free >= lowest_to_date)
838 return;
839
840 spin_lock(&low_water_lock);
841 if (free < lowest_to_date) {
842 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
843 "left\n",
844 current->comm, free);
845 lowest_to_date = free;
846 }
847 spin_unlock(&low_water_lock);
848}
849#else
850static inline void check_stack_usage(void) {}
851#endif
852
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700853static inline void exit_child_reaper(struct task_struct *tsk)
854{
Sukadev Bhattiprolu88f21d82007-10-18 23:39:50 -0700855 if (likely(tsk->group_leader != task_child_reaper(tsk)))
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700856 return;
857
Sukadev Bhattiprolu3eb07c82007-10-18 23:40:13 -0700858 if (tsk->nsproxy->pid_ns == &init_pid_ns)
859 panic("Attempted to kill init!");
860
861 /*
862 * @tsk is the last thread in the 'cgroup-init' and is exiting.
863 * Terminate all remaining processes in the namespace and reap them
864 * before exiting @tsk.
865 *
866 * Note that @tsk (last thread of cgroup-init) may not necessarily
867 * be the child-reaper (i.e main thread of cgroup-init) of the
868 * namespace i.e the child_reaper may have already exited.
869 *
870 * Even after a child_reaper exits, we let it inherit orphaned children,
871 * because, pid_ns->child_reaper remains valid as long as there is
872 * at least one living sub-thread in the cgroup init.
873
874 * This living sub-thread of the cgroup-init will be notified when
875 * a child inherited by the 'child-reaper' exits (do_notify_parent()
876 * uses __group_send_sig_info()). Further, when reaping child processes,
877 * do_wait() iterates over children of all living sub threads.
878
879 * i.e even though 'child_reaper' thread is listed as the parent of the
880 * orphaned children, any living sub-thread in the cgroup-init can
881 * perform the role of the child_reaper.
882 */
883 zap_pid_ns_processes(tsk->nsproxy->pid_ns);
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700884}
885
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800886NORET_TYPE void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887{
888 struct task_struct *tsk = current;
889 int group_dead;
890
891 profile_task_exit(tsk);
892
Jens Axboe22e2c502005-06-27 10:55:12 +0200893 WARN_ON(atomic_read(&tsk->fs_excl));
894
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 if (unlikely(in_interrupt()))
896 panic("Aiee, killing interrupt handler!");
897 if (unlikely(!tsk->pid))
898 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899
900 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
901 current->ptrace_message = code;
902 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
903 }
904
Alexander Nybergdf164db2005-06-23 00:09:13 -0700905 /*
906 * We're taking recursive faults here in do_exit. Safest is to just
907 * leave this task alone and wait for reboot.
908 */
909 if (unlikely(tsk->flags & PF_EXITING)) {
910 printk(KERN_ALERT
911 "Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700912 /*
913 * We can do this unlocked here. The futex code uses
914 * this flag just to verify whether the pi state
915 * cleanup has been done or not. In the worst case it
916 * loops once more. We pretend that the cleanup was
917 * done as there is no way to return. Either the
918 * OWNER_DIED bit is set by now or we push the blocked
919 * task into the wait for ever nirwana as well.
920 */
921 tsk->flags |= PF_EXITPIDONE;
Al Viroafc847b2006-02-28 12:51:55 -0500922 if (tsk->io_context)
923 exit_io_context();
Alexander Nybergdf164db2005-06-23 00:09:13 -0700924 set_current_state(TASK_UNINTERRUPTIBLE);
925 schedule();
926 }
927
Oleg Nesterovd12619b2008-02-08 04:19:12 -0800928 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700929 /*
930 * tsk->flags are checked in the futex code to protect against
931 * an exiting task cleaning up the robust pi futexes.
932 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -0700933 smp_mb();
934 spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 if (unlikely(in_atomic()))
937 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -0700938 current->comm, task_pid_nr(current),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 preempt_count());
940
941 acct_update_integrals(tsk);
Hugh Dickins365e9c872005-10-29 18:16:18 -0700942 if (tsk->mm) {
943 update_hiwater_rss(tsk->mm);
944 update_hiwater_vm(tsk->mm);
945 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -0700947 if (group_dead) {
Oleg Nesterov84eb6462007-10-16 23:26:49 -0700948 exit_child_reaper(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700949 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -0700950 exit_itimers(tsk->signal);
Andrew Mortonc3068952005-08-04 16:49:32 -0700951 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700952 acct_collect(code, group_dead);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700953#ifdef CONFIG_FUTEX
Ingo Molnar0771dfe2006-03-27 01:16:22 -0800954 if (unlikely(tsk->robust_list))
955 exit_robust_list(tsk);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700956#ifdef CONFIG_COMPAT
Ingo Molnar34f192c2006-03-27 01:16:24 -0800957 if (unlikely(tsk->compat_robust_list))
958 compat_exit_robust_list(tsk);
959#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700960#endif
Miloslav Trmac522ed772007-07-15 23:40:56 -0700961 if (group_dead)
962 tty_audit_exit();
Al Virofa84cb92006-03-29 20:30:19 -0500963 if (unlikely(tsk->audit_context))
964 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -0800965
Jonathan Limf2ab6d82007-08-30 23:56:23 -0700966 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -0800967 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -0700968
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 exit_mm(tsk);
970
KaiGai Kohei0e464812006-06-25 05:49:24 -0700971 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700972 acct_process();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 exit_sem(tsk);
974 __exit_files(tsk);
975 __exit_fs(tsk);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700976 check_stack_usage();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 exit_thread();
Paul Menageb4f48b62007-10-18 23:39:33 -0700978 cgroup_exit(tsk, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 exit_keys(tsk);
980
981 if (group_dead && tsk->signal->leader)
982 disassociate_ctty(1);
983
Al Viroa1261f52005-11-13 16:06:55 -0800984 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 if (tsk->binfmt)
986 module_put(tsk->binfmt->module);
987
Matt Helsley9f460802005-11-07 00:59:16 -0800988 proc_exit_connector(tsk);
Serge E. Hallyn0f245282007-01-30 15:28:23 -0600989 exit_notify(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990#ifdef CONFIG_NUMA
991 mpol_free(tsk->mempolicy);
992 tsk->mempolicy = NULL;
993#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700994#ifdef CONFIG_FUTEX
Ingo Molnarde5097c2006-01-09 15:59:21 -0800995 /*
Ingo Molnarc87e2832006-06-27 02:54:58 -0700996 * This must happen late, after the PID is not
997 * hashed anymore:
998 */
999 if (unlikely(!list_empty(&tsk->pi_state_list)))
1000 exit_pi_state_list(tsk);
1001 if (unlikely(current->pi_state_cache))
1002 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001003#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001004 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001005 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -08001006 */
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001007 debug_check_no_locks_held(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001008 /*
1009 * We can do this unlocked here. The futex code uses this flag
1010 * just to verify whether the pi state cleanup has been done
1011 * or not. In the worst case it loops once more.
1012 */
1013 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
Al Viroafc847b2006-02-28 12:51:55 -05001015 if (tsk->io_context)
1016 exit_io_context();
1017
Jens Axboeb92ce552006-04-11 13:52:07 +02001018 if (tsk->splice_pipe)
1019 __free_pipe_info(tsk->splice_pipe);
1020
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001021 preempt_disable();
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001022 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -07001023 tsk->state = TASK_DEAD;
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001024
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 schedule();
1026 BUG();
1027 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -07001028 for (;;)
1029 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030}
1031
Russ Anderson012914d2005-04-23 00:08:00 -07001032EXPORT_SYMBOL_GPL(do_exit);
1033
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1035{
1036 if (comp)
1037 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001038
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 do_exit(code);
1040}
1041
1042EXPORT_SYMBOL(complete_and_exit);
1043
1044asmlinkage long sys_exit(int error_code)
1045{
1046 do_exit((error_code&0xff)<<8);
1047}
1048
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049/*
1050 * Take down every thread in the group. This is called by fatal signals
1051 * as well as by sys_exit_group (below).
1052 */
1053NORET_TYPE void
1054do_group_exit(int exit_code)
1055{
1056 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1057
1058 if (current->signal->flags & SIGNAL_GROUP_EXIT)
1059 exit_code = current->signal->group_exit_code;
1060 else if (!thread_group_empty(current)) {
1061 struct signal_struct *const sig = current->signal;
1062 struct sighand_struct *const sighand = current->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001064 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 /* Another thread got here before we took the lock. */
1066 exit_code = sig->group_exit_code;
1067 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001069 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 zap_other_threads(current);
1071 }
1072 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 }
1074
1075 do_exit(exit_code);
1076 /* NOTREACHED */
1077}
1078
1079/*
1080 * this kills every thread in the thread group. Note that any externally
1081 * wait4()-ing process will get the correct exit code - even if this
1082 * thread is not the thread group leader.
1083 */
1084asmlinkage void sys_exit_group(int error_code)
1085{
1086 do_group_exit((error_code & 0xff) << 8);
1087}
1088
Eric W. Biederman161550d2008-02-08 04:19:14 -08001089static struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1090{
1091 struct pid *pid = NULL;
1092 if (type == PIDTYPE_PID)
1093 pid = task->pids[type].pid;
1094 else if (type < PIDTYPE_MAX)
1095 pid = task->group_leader->pids[type].pid;
1096 return pid;
1097}
1098
1099static int eligible_child(enum pid_type type, struct pid *pid, int options,
1100 struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101{
Roland McGrath73243282007-05-06 14:50:20 -07001102 int err;
1103
Eric W. Biederman161550d2008-02-08 04:19:14 -08001104 if (type < PIDTYPE_MAX) {
1105 if (task_pid_type(p, type) != pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 return 0;
1107 }
1108
1109 /*
1110 * Do not consider detached threads that are
1111 * not ptraced:
1112 */
1113 if (p->exit_signal == -1 && !p->ptrace)
1114 return 0;
1115
1116 /* Wait for all children (clone and not) if __WALL is set;
1117 * otherwise, wait for clone children *only* if __WCLONE is
1118 * set; otherwise, wait for non-clone children *only*. (Note:
1119 * A "clone" child here is one that reports to its parent
1120 * using a signal other than SIGCHLD.) */
1121 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
1122 && !(options & __WALL))
1123 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
Roland McGrath73243282007-05-06 14:50:20 -07001125 err = security_task_wait(p);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001126 if (likely(!err))
1127 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128
Eric W. Biederman161550d2008-02-08 04:19:14 -08001129 if (type != PIDTYPE_PID)
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001130 return 0;
1131 /* This child was explicitly requested, abort */
1132 read_unlock(&tasklist_lock);
1133 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134}
1135
Ingo Molnar36c8b582006-07-03 00:25:41 -07001136static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 int why, int status,
1138 struct siginfo __user *infop,
1139 struct rusage __user *rusagep)
1140{
1141 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -07001142
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 put_task_struct(p);
1144 if (!retval)
1145 retval = put_user(SIGCHLD, &infop->si_signo);
1146 if (!retval)
1147 retval = put_user(0, &infop->si_errno);
1148 if (!retval)
1149 retval = put_user((short)why, &infop->si_code);
1150 if (!retval)
1151 retval = put_user(pid, &infop->si_pid);
1152 if (!retval)
1153 retval = put_user(uid, &infop->si_uid);
1154 if (!retval)
1155 retval = put_user(status, &infop->si_status);
1156 if (!retval)
1157 retval = pid;
1158 return retval;
1159}
1160
1161/*
1162 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1163 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1164 * the lock and this task is uninteresting. If we return nonzero, we have
1165 * released the lock and the system call should return.
1166 */
Ingo Molnar36c8b582006-07-03 00:25:41 -07001167static int wait_task_zombie(struct task_struct *p, int noreap,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 struct siginfo __user *infop,
1169 int __user *stat_addr, struct rusage __user *ru)
1170{
1171 unsigned long state;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001172 int retval, status, traced;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001173 pid_t pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174
1175 if (unlikely(noreap)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 uid_t uid = p->uid;
1177 int exit_code = p->exit_code;
1178 int why, status;
1179
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 get_task_struct(p);
1181 read_unlock(&tasklist_lock);
1182 if ((exit_code & 0x7f) == 0) {
1183 why = CLD_EXITED;
1184 status = exit_code >> 8;
1185 } else {
1186 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1187 status = exit_code & 0x7f;
1188 }
1189 return wait_noreap_copyout(p, pid, uid, why,
1190 status, infop, ru);
1191 }
1192
1193 /*
1194 * Try to move the task's state to DEAD
1195 * only one thread is allowed to do this:
1196 */
1197 state = xchg(&p->exit_state, EXIT_DEAD);
1198 if (state != EXIT_ZOMBIE) {
1199 BUG_ON(state != EXIT_DEAD);
1200 return 0;
1201 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001203 /* traced means p->ptrace, but not vice versa */
1204 traced = (p->real_parent != p->parent);
1205
1206 if (likely(!traced)) {
Jesper Juhl3795e162006-01-09 20:54:39 -08001207 struct signal_struct *psig;
1208 struct signal_struct *sig;
1209
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 /*
1211 * The resource counters for the group leader are in its
1212 * own task_struct. Those for dead threads in the group
1213 * are in its signal_struct, as are those for the child
1214 * processes it has previously reaped. All these
1215 * accumulate in the parent's signal_struct c* fields.
1216 *
1217 * We don't bother to take a lock here to protect these
1218 * p->signal fields, because they are only touched by
1219 * __exit_signal, which runs with tasklist_lock
1220 * write-locked anyway, and so is excluded here. We do
1221 * need to protect the access to p->parent->signal fields,
1222 * as other threads in the parent group can be right
1223 * here reaping other children at the same time.
1224 */
1225 spin_lock_irq(&p->parent->sighand->siglock);
Jesper Juhl3795e162006-01-09 20:54:39 -08001226 psig = p->parent->signal;
1227 sig = p->signal;
1228 psig->cutime =
1229 cputime_add(psig->cutime,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 cputime_add(p->utime,
Jesper Juhl3795e162006-01-09 20:54:39 -08001231 cputime_add(sig->utime,
1232 sig->cutime)));
1233 psig->cstime =
1234 cputime_add(psig->cstime,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 cputime_add(p->stime,
Jesper Juhl3795e162006-01-09 20:54:39 -08001236 cputime_add(sig->stime,
1237 sig->cstime)));
Laurent Vivier9ac52312007-10-15 17:00:19 +02001238 psig->cgtime =
1239 cputime_add(psig->cgtime,
1240 cputime_add(p->gtime,
1241 cputime_add(sig->gtime,
1242 sig->cgtime)));
Jesper Juhl3795e162006-01-09 20:54:39 -08001243 psig->cmin_flt +=
1244 p->min_flt + sig->min_flt + sig->cmin_flt;
1245 psig->cmaj_flt +=
1246 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1247 psig->cnvcsw +=
1248 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1249 psig->cnivcsw +=
1250 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001251 psig->cinblock +=
1252 task_io_get_inblock(p) +
1253 sig->inblock + sig->cinblock;
1254 psig->coublock +=
1255 task_io_get_oublock(p) +
1256 sig->oublock + sig->coublock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 spin_unlock_irq(&p->parent->sighand->siglock);
1258 }
1259
1260 /*
1261 * Now we are sure this task is interesting, and no other
1262 * thread can reap it because we set its state to EXIT_DEAD.
1263 */
1264 read_unlock(&tasklist_lock);
1265
1266 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1267 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1268 ? p->signal->group_exit_code : p->exit_code;
1269 if (!retval && stat_addr)
1270 retval = put_user(status, stat_addr);
1271 if (!retval && infop)
1272 retval = put_user(SIGCHLD, &infop->si_signo);
1273 if (!retval && infop)
1274 retval = put_user(0, &infop->si_errno);
1275 if (!retval && infop) {
1276 int why;
1277
1278 if ((status & 0x7f) == 0) {
1279 why = CLD_EXITED;
1280 status >>= 8;
1281 } else {
1282 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1283 status &= 0x7f;
1284 }
1285 retval = put_user((short)why, &infop->si_code);
1286 if (!retval)
1287 retval = put_user(status, &infop->si_status);
1288 }
1289 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001290 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 if (!retval && infop)
1292 retval = put_user(p->uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001293 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001294 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001295
1296 if (traced) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001298 /* We dropped tasklist, ptracer could die and untrace */
1299 ptrace_unlink(p);
1300 /*
1301 * If this is not a detached task, notify the parent.
1302 * If it's still not detached after that, don't release
1303 * it now.
1304 */
1305 if (p->exit_signal != -1) {
1306 do_notify_parent(p, p->exit_signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 if (p->exit_signal != -1) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001308 p->exit_state = EXIT_ZOMBIE;
1309 p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310 }
1311 }
1312 write_unlock_irq(&tasklist_lock);
1313 }
1314 if (p != NULL)
1315 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001316
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 return retval;
1318}
1319
1320/*
1321 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1322 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1323 * the lock and this task is uninteresting. If we return nonzero, we have
1324 * released the lock and the system call should return.
1325 */
Oleg Nesterov1bad95c2008-02-08 04:19:03 -08001326static int wait_task_stopped(struct task_struct *p,
Ingo Molnar36c8b582006-07-03 00:25:41 -07001327 int noreap, struct siginfo __user *infop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 int __user *stat_addr, struct rusage __user *ru)
1329{
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001330 int retval, exit_code, why;
1331 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001332 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001334 exit_code = 0;
1335 spin_lock_irq(&p->sighand->siglock);
1336
1337 if (unlikely(!task_is_stopped_or_traced(p)))
1338 goto unlock_sig;
1339
Oleg Nesterov1bad95c2008-02-08 04:19:03 -08001340 if (!(p->ptrace & PT_PTRACED) && p->signal->group_stop_count > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 /*
1342 * A group stop is in progress and this is the group leader.
1343 * We won't report until all threads have stopped.
1344 */
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001345 goto unlock_sig;
1346
1347 exit_code = p->exit_code;
1348 if (!exit_code)
1349 goto unlock_sig;
1350
1351 if (!noreap)
1352 p->exit_code = 0;
1353
1354 uid = p->uid;
1355unlock_sig:
1356 spin_unlock_irq(&p->sighand->siglock);
1357 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 return 0;
1359
1360 /*
1361 * Now we are pretty sure this task is interesting.
1362 * Make sure it doesn't get reaped out from under us while we
1363 * give up the lock and then examine it below. We don't want to
1364 * keep holding onto the tasklist_lock while we call getrusage and
1365 * possibly take page faults for user memory.
1366 */
1367 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001368 pid = task_pid_vnr(p);
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001369 why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 read_unlock(&tasklist_lock);
1371
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001372 if (unlikely(noreap))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 return wait_noreap_copyout(p, pid, uid,
Scott James Remnante6ceb322007-11-28 16:22:07 -08001374 why, exit_code,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 infop, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376
1377 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1378 if (!retval && stat_addr)
1379 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1380 if (!retval && infop)
1381 retval = put_user(SIGCHLD, &infop->si_signo);
1382 if (!retval && infop)
1383 retval = put_user(0, &infop->si_errno);
1384 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001385 retval = put_user(why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 if (!retval && infop)
1387 retval = put_user(exit_code, &infop->si_status);
1388 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001389 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001391 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001393 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 put_task_struct(p);
1395
1396 BUG_ON(!retval);
1397 return retval;
1398}
1399
1400/*
1401 * Handle do_wait work for one task in a live, non-stopped state.
1402 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1403 * the lock and this task is uninteresting. If we return nonzero, we have
1404 * released the lock and the system call should return.
1405 */
Ingo Molnar36c8b582006-07-03 00:25:41 -07001406static int wait_task_continued(struct task_struct *p, int noreap,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 struct siginfo __user *infop,
1408 int __user *stat_addr, struct rusage __user *ru)
1409{
1410 int retval;
1411 pid_t pid;
1412 uid_t uid;
1413
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1415 return 0;
1416
1417 spin_lock_irq(&p->sighand->siglock);
1418 /* Re-check with the lock held. */
1419 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1420 spin_unlock_irq(&p->sighand->siglock);
1421 return 0;
1422 }
1423 if (!noreap)
1424 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1425 spin_unlock_irq(&p->sighand->siglock);
1426
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001427 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 uid = p->uid;
1429 get_task_struct(p);
1430 read_unlock(&tasklist_lock);
1431
1432 if (!infop) {
1433 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1434 put_task_struct(p);
1435 if (!retval && stat_addr)
1436 retval = put_user(0xffff, stat_addr);
1437 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001438 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 } else {
1440 retval = wait_noreap_copyout(p, pid, uid,
1441 CLD_CONTINUED, SIGCONT,
1442 infop, ru);
1443 BUG_ON(retval == 0);
1444 }
1445
1446 return retval;
1447}
1448
Eric W. Biederman161550d2008-02-08 04:19:14 -08001449static long do_wait(enum pid_type type, struct pid *pid, int options,
1450 struct siginfo __user *infop, int __user *stat_addr,
1451 struct rusage __user *ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452{
1453 DECLARE_WAITQUEUE(wait, current);
1454 struct task_struct *tsk;
1455 int flag, retval;
1456
1457 add_wait_queue(&current->signal->wait_chldexit,&wait);
1458repeat:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001459 /* If there is nothing that can match our critier just get out */
1460 retval = -ECHILD;
1461 if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type])))
1462 goto end;
1463
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 /*
1465 * We will set this flag if we see any child that might later
1466 * match our criteria, even if we are not able to reap it yet.
1467 */
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001468 flag = retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 current->state = TASK_INTERRUPTIBLE;
1470 read_lock(&tasklist_lock);
1471 tsk = current;
1472 do {
1473 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -07001475 list_for_each_entry(p, &tsk->children, sibling) {
Eric W. Biederman161550d2008-02-08 04:19:14 -08001476 int ret = eligible_child(type, pid, options, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 if (!ret)
1478 continue;
1479
Roland McGrath73243282007-05-06 14:50:20 -07001480 if (unlikely(ret < 0)) {
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001481 retval = ret;
1482 } else if (task_is_stopped_or_traced(p)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 /*
1484 * It's stopped now, so it might later
1485 * continue, exit, or stop again.
1486 */
1487 flag = 1;
Oleg Nesterov34a17382008-02-08 04:18:59 -08001488 if (!(p->ptrace & PT_PTRACED) &&
1489 !(options & WUNTRACED))
1490 continue;
Matthew Wilcox338077e2007-12-06 11:09:35 -05001491
Oleg Nesterov1bad95c2008-02-08 04:19:03 -08001492 retval = wait_task_stopped(p,
Matthew Wilcox338077e2007-12-06 11:09:35 -05001493 (options & WNOWAIT), infop,
1494 stat_addr, ru);
Oleg Nesterov96fabbf2008-02-08 04:19:04 -08001495 } else if (p->exit_state == EXIT_ZOMBIE &&
1496 !delay_group_leader(p)) {
Matthew Wilcox338077e2007-12-06 11:09:35 -05001497 /*
Oleg Nesterov96fabbf2008-02-08 04:19:04 -08001498 * We don't reap group leaders with subthreads.
Matthew Wilcox338077e2007-12-06 11:09:35 -05001499 */
Matthew Wilcox338077e2007-12-06 11:09:35 -05001500 if (!likely(options & WEXITED))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 continue;
Matthew Wilcox338077e2007-12-06 11:09:35 -05001502 retval = wait_task_zombie(p,
1503 (options & WNOWAIT), infop,
1504 stat_addr, ru);
Oleg Nesterov0a76fe82008-02-06 01:37:06 -08001505 } else if (p->exit_state != EXIT_DEAD) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 /*
1507 * It's running now, so it might later
1508 * exit, stop, or stop and then continue.
1509 */
1510 flag = 1;
1511 if (!unlikely(options & WCONTINUED))
1512 continue;
Matthew Wilcox338077e2007-12-06 11:09:35 -05001513 retval = wait_task_continued(p,
1514 (options & WNOWAIT), infop,
1515 stat_addr, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 }
Oleg Nesterov9cbab812008-02-08 04:19:02 -08001517 if (retval != 0) /* tasklist_lock released */
1518 goto end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 }
1520 if (!flag) {
Matthias Kaehlcke03ff1792007-10-18 23:39:57 -07001521 list_for_each_entry(p, &tsk->ptrace_children,
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001522 ptrace_list) {
Eric W. Biederman161550d2008-02-08 04:19:14 -08001523 flag = eligible_child(type, pid, options, p);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001524 if (!flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 continue;
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001526 if (likely(flag > 0))
1527 break;
1528 retval = flag;
1529 goto end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 }
1531 }
1532 if (options & __WNOTHREAD)
1533 break;
1534 tsk = next_thread(tsk);
Eric Sesterhenn125e1872006-06-23 02:06:06 -07001535 BUG_ON(tsk->signal != current->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 } while (tsk != current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001538
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 if (flag) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 if (options & WNOHANG)
1541 goto end;
1542 retval = -ERESTARTSYS;
1543 if (signal_pending(current))
1544 goto end;
1545 schedule();
1546 goto repeat;
1547 }
1548 retval = -ECHILD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549end:
1550 current->state = TASK_RUNNING;
1551 remove_wait_queue(&current->signal->wait_chldexit,&wait);
1552 if (infop) {
1553 if (retval > 0)
Oleg Nesterov9cbab812008-02-08 04:19:02 -08001554 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 else {
1556 /*
1557 * For a WNOHANG return, clear out all the fields
1558 * we would set so the user can easily tell the
1559 * difference.
1560 */
1561 if (!retval)
1562 retval = put_user(0, &infop->si_signo);
1563 if (!retval)
1564 retval = put_user(0, &infop->si_errno);
1565 if (!retval)
1566 retval = put_user(0, &infop->si_code);
1567 if (!retval)
1568 retval = put_user(0, &infop->si_pid);
1569 if (!retval)
1570 retval = put_user(0, &infop->si_uid);
1571 if (!retval)
1572 retval = put_user(0, &infop->si_status);
1573 }
1574 }
1575 return retval;
1576}
1577
Eric W. Biederman161550d2008-02-08 04:19:14 -08001578asmlinkage long sys_waitid(int which, pid_t upid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 struct siginfo __user *infop, int options,
1580 struct rusage __user *ru)
1581{
Eric W. Biederman161550d2008-02-08 04:19:14 -08001582 struct pid *pid = NULL;
1583 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 long ret;
1585
1586 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1587 return -EINVAL;
1588 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1589 return -EINVAL;
1590
1591 switch (which) {
1592 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001593 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 break;
1595 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001596 type = PIDTYPE_PID;
1597 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 return -EINVAL;
1599 break;
1600 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001601 type = PIDTYPE_PGID;
1602 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 break;
1605 default:
1606 return -EINVAL;
1607 }
1608
Eric W. Biederman161550d2008-02-08 04:19:14 -08001609 if (type < PIDTYPE_MAX)
1610 pid = find_get_pid(upid);
1611 ret = do_wait(type, pid, options, infop, NULL, ru);
1612 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613
1614 /* avoid REGPARM breakage on x86: */
1615 prevent_tail_call(ret);
1616 return ret;
1617}
1618
Eric W. Biederman161550d2008-02-08 04:19:14 -08001619asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 int options, struct rusage __user *ru)
1621{
Eric W. Biederman161550d2008-02-08 04:19:14 -08001622 struct pid *pid = NULL;
1623 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 long ret;
1625
1626 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1627 __WNOTHREAD|__WCLONE|__WALL))
1628 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001629
1630 if (upid == -1)
1631 type = PIDTYPE_MAX;
1632 else if (upid < 0) {
1633 type = PIDTYPE_PGID;
1634 pid = find_get_pid(-upid);
1635 } else if (upid == 0) {
1636 type = PIDTYPE_PGID;
1637 pid = get_pid(task_pgrp(current));
1638 } else /* upid > 0 */ {
1639 type = PIDTYPE_PID;
1640 pid = find_get_pid(upid);
1641 }
1642
1643 ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru);
1644 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645
1646 /* avoid REGPARM breakage on x86: */
1647 prevent_tail_call(ret);
1648 return ret;
1649}
1650
1651#ifdef __ARCH_WANT_SYS_WAITPID
1652
1653/*
1654 * sys_waitpid() remains for compatibility. waitpid() should be
1655 * implemented by calling sys_wait4() from libc.a.
1656 */
1657asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1658{
1659 return sys_wait4(pid, stat_addr, options, NULL);
1660}
1661
1662#endif