blob: 546774a31a66752a698faa02c4e54f7c39fbef9e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/kernel/exit.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#include <linux/mm.h>
8#include <linux/slab.h>
9#include <linux/interrupt.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/module.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080011#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/completion.h>
13#include <linux/personality.h>
14#include <linux/tty.h>
Jens Axboeda9cbc82008-06-30 20:42:08 +020015#include <linux/iocontext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/key.h>
17#include <linux/security.h>
18#include <linux/cpu.h>
19#include <linux/acct.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070020#include <linux/tsacct_kern.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040022#include <linux/fdtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/binfmts.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070024#include <linux/nsproxy.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080025#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/ptrace.h>
27#include <linux/profile.h>
28#include <linux/mount.h>
29#include <linux/proc_fs.h>
Eric W. Biederman49d769d2007-05-09 02:34:33 -070030#include <linux/kthread.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/mempolicy.h>
Shailabh Nagarc7572492006-07-14 00:24:40 -070032#include <linux/taskstats_kern.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070033#include <linux/delayacct.h>
Rafael J. Wysocki83144182007-07-17 04:03:35 -070034#include <linux/freezer.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070035#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/syscalls.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070037#include <linux/signal.h>
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080038#include <linux/posix-timers.h>
Matt Helsley9f460802005-11-07 00:59:16 -080039#include <linux/cn_proc.h>
Ingo Molnarde5097c2006-01-09 15:59:21 -080040#include <linux/mutex.h>
Ingo Molnar0771dfe2006-03-27 01:16:22 -080041#include <linux/futex.h>
Jens Axboeb92ce552006-04-11 13:52:07 +020042#include <linux/pipe_fs_i.h>
Al Virofa84cb92006-03-29 20:30:19 -050043#include <linux/audit.h> /* for audit_free() */
Adrian Bunk83cc5ed2006-06-25 05:47:41 -070044#include <linux/resource.h>
David Howells0d67a462006-08-29 19:05:56 +010045#include <linux/blkdev.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070046#include <linux/task_io_accounting_ops.h>
Roland McGrath30199f52008-07-25 19:45:46 -070047#include <linux/tracehook.h>
Al Viro5ad4e532009-03-29 19:50:06 -040048#include <linux/fs_struct.h>
David Howellsd84f4f92008-11-14 10:39:23 +110049#include <linux/init_task.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020050#include <linux/perf_event.h>
Steven Rostedtad8d75f2009-04-14 19:39:12 -040051#include <trace/events/sched.h>
Frederic Weisbecker24f1e32c2009-09-09 19:22:48 +020052#include <linux/hw_breakpoint.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
54#include <asm/uaccess.h>
55#include <asm/unistd.h>
56#include <asm/pgtable.h>
57#include <asm/mmu_context.h>
David Howellsd84f4f92008-11-14 10:39:23 +110058#include "cred-internals.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
Adrian Bunk408b6642005-05-01 08:59:29 -070060static void exit_mm(struct task_struct * tsk);
61
Linus Torvalds1da177e2005-04-16 15:20:36 -070062static void __unhash_process(struct task_struct *p)
63{
64 nr_threads--;
65 detach_pid(p, PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 if (thread_group_leader(p)) {
67 detach_pid(p, PIDTYPE_PGID);
68 detach_pid(p, PIDTYPE_SID);
Oleg Nesterovc97d9892006-03-28 16:11:06 -080069
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -070070 list_del_rcu(&p->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -080071 list_del_init(&p->sibling);
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);
Linus Torvalds1da177e2005-04-16 15:20:36 -070075}
76
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080077/*
78 * This function expects the tasklist_lock write-locked.
79 */
80static void __exit_signal(struct task_struct *tsk)
81{
82 struct signal_struct *sig = tsk->signal;
83 struct sighand_struct *sighand;
84
85 BUG_ON(!sig);
86 BUG_ON(!atomic_read(&sig->count));
87
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -080088 sighand = rcu_dereference(tsk->sighand);
89 spin_lock(&sighand->siglock);
90
91 posix_cpu_timers_exit(tsk);
92 if (atomic_dec_and_test(&sig->count))
93 posix_cpu_timers_exit_group(tsk);
94 else {
95 /*
96 * If there is any task waiting for the group exit
97 * then notify it:
98 */
Oleg Nesterov6db840f2007-10-16 23:27:23 -070099 if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count)
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800100 wake_up_process(sig->group_exit_task);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700101
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800102 if (tsk == sig->curr_target)
103 sig->curr_target = next_thread(tsk);
104 /*
105 * Accumulate here the counters for all threads but the
106 * group leader as they die, so they can be added into
107 * the process-wide totals when those are taken.
108 * The group leader stays around as a zombie as long
109 * as there are other threads. When it gets reaped,
110 * the exit.c code will add its counts into these totals.
111 * We won't ever get here for the group leader, since it
112 * will have been the last reference on the signal_struct.
113 */
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +0900114 sig->utime = cputime_add(sig->utime, tsk->utime);
115 sig->stime = cputime_add(sig->stime, tsk->stime);
Hidetoshi Setod5b7c782009-11-26 14:49:05 +0900116 sig->gtime = cputime_add(sig->gtime, tsk->gtime);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800117 sig->min_flt += tsk->min_flt;
118 sig->maj_flt += tsk->maj_flt;
119 sig->nvcsw += tsk->nvcsw;
120 sig->nivcsw += tsk->nivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700121 sig->inblock += task_io_get_inblock(tsk);
122 sig->oublock += task_io_get_oublock(tsk);
Andrea Righi59954772008-07-27 17:29:15 +0200123 task_io_accounting_add(&sig->ioac, &tsk->ioac);
Peter Zijlstra32bd6712009-02-05 12:24:15 +0100124 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800125 sig = NULL; /* Marker for below. */
126 }
127
Oleg Nesterov58767002006-03-28 16:11:20 -0800128 __unhash_process(tsk);
129
Oleg Nesterovda7978b2008-05-23 13:04:41 -0700130 /*
131 * Do this under ->siglock, we can race with another thread
132 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
133 */
134 flush_sigqueue(&tsk->pending);
135
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800136 tsk->signal = NULL;
Oleg Nesterova7e53282006-03-28 16:11:27 -0800137 tsk->sighand = NULL;
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800138 spin_unlock(&sighand->siglock);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800139
Oleg Nesterova7e53282006-03-28 16:11:27 -0800140 __cleanup_sighand(sighand);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800141 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800142 if (sig) {
143 flush_sigqueue(&sig->shared_pending);
Oleg Nesterov093a8e82006-10-28 10:38:51 -0700144 taskstats_tgid_free(sig);
Oleg Nesterovad474ca2008-11-10 15:39:30 +0100145 /*
146 * Make sure ->signal can't go away under rq->lock,
147 * see account_group_exec_runtime().
148 */
149 task_rq_unlock_wait(tsk);
Oleg Nesterov6a14c5c2006-03-28 16:11:18 -0800150 __cleanup_signal(sig);
151 }
152}
153
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800154static void delayed_put_task_struct(struct rcu_head *rhp)
155{
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400156 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
157
Ingo Molnarcdd6c482009-09-21 12:02:48 +0200158#ifdef CONFIG_PERF_EVENTS
159 WARN_ON_ONCE(tsk->perf_event_ctxp);
Ingo Molnareef6cbf2008-12-19 10:20:42 +0100160#endif
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400161 trace_sched_process_free(tsk);
162 put_task_struct(tsk);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800163}
164
Roland McGrathf4700212008-03-24 18:36:23 -0700165
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166void release_task(struct task_struct * p)
167{
Ingo Molnar36c8b582006-07-03 00:25:41 -0700168 struct task_struct *leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 int zap_leader;
Oleg Nesterov1f09f972006-03-28 16:11:11 -0800170repeat:
Roland McGrathdae33572008-07-25 19:45:48 -0700171 tracehook_prepare_release_task(p);
David Howellsc69e8d92008-11-14 10:39:19 +1100172 /* don't need to get the RCU readlock here - the process is dead and
173 * can't be modifying its own credentials */
174 atomic_dec(&__task_cred(p)->user->processes);
175
Pavel Emelyanov60347f62007-10-18 23:40:03 -0700176 proc_flush_task(p);
Ingo Molnar02030262009-05-17 11:24:08 +0200177
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 write_lock_irq(&tasklist_lock);
Roland McGrathdae33572008-07-25 19:45:48 -0700179 tracehook_finish_release_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 __exit_signal(p);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -0800181
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 /*
183 * If we are the last non-leader member of the thread
184 * group, and the leader is zombie, then notify the
185 * group leader's parent process. (if it wants notification.)
186 */
187 zap_leader = 0;
188 leader = p->group_leader;
189 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700190 BUG_ON(task_detached(leader));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 do_notify_parent(leader, leader->exit_signal);
192 /*
193 * If we were the last child thread and the leader has
194 * exited already, and the leader's parent ignores SIGCHLD,
195 * then we are the one who should release the leader.
196 *
197 * do_notify_parent() will have marked it self-reaping in
198 * that case.
199 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700200 zap_leader = task_detached(leader);
Roland McGrathdae33572008-07-25 19:45:48 -0700201
202 /*
203 * This maintains the invariant that release_task()
204 * only runs on a task in EXIT_DEAD, just for sanity.
205 */
206 if (zap_leader)
207 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 }
209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 release_thread(p);
Eric W. Biederman8c7904a2006-03-31 02:31:37 -0800212 call_rcu(&p->rcu, delayed_put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213
214 p = leader;
215 if (unlikely(zap_leader))
216 goto repeat;
217}
218
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219/*
220 * This checks not only the pgrp, but falls back on the pid if no
221 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
222 * without this...
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800223 *
224 * The caller must hold rcu lock or the tasklist lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 */
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800226struct pid *session_of_pgrp(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227{
228 struct task_struct *p;
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800229 struct pid *sid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800231 p = pid_task(pgrp, PIDTYPE_PGID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800232 if (p == NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800233 p = pid_task(pgrp, PIDTYPE_PID);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800234 if (p != NULL)
Eric W. Biederman04a2e6a2007-02-12 00:52:56 -0800235 sid = task_session(p);
Oleg Nesterov62dfb552006-12-08 02:38:03 -0800236
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 return sid;
238}
239
240/*
241 * Determine if a process group is "orphaned", according to the POSIX
242 * definition in 2.2.2.52. Orphaned process groups are not to be affected
243 * by terminal-generated stop signals. Newly orphaned process groups are
244 * to receive a SIGHUP and a SIGCONT.
245 *
246 * "I ask you, have you ever known what it is to be an orphan?"
247 */
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800248static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249{
250 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800252 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300253 if ((p == ignored_task) ||
254 (p->exit_state && thread_group_empty(p)) ||
255 is_global_init(p->real_parent))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 continue;
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300257
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800258 if (task_pgrp(p->real_parent) != pgrp &&
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300259 task_session(p->real_parent) == task_session(p))
260 return 0;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800261 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Oleg Nesterov05e83df2008-03-02 21:44:42 +0300262
263 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264}
265
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800266int is_current_pgrp_orphaned(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267{
268 int retval;
269
270 read_lock(&tasklist_lock);
Eric W. Biederman3e7cd6c2007-02-12 00:52:58 -0800271 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 read_unlock(&tasklist_lock);
273
274 return retval;
275}
276
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800277static int has_stopped_jobs(struct pid *pgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
279 int retval = 0;
280 struct task_struct *p;
281
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800282 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Matthew Wilcox338077e2007-12-06 11:09:35 -0500283 if (!task_is_stopped(p))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 retval = 1;
286 break;
Eric W. Biederman0475ac02007-02-12 00:52:57 -0800287 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 return retval;
289}
290
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300291/*
292 * Check to see if any process groups have become orphaned as
293 * a result of our exiting, and if they have any stopped jobs,
294 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
295 */
296static void
297kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
298{
299 struct pid *pgrp = task_pgrp(tsk);
300 struct task_struct *ignored_task = tsk;
301
302 if (!parent)
303 /* exit: our father is in a different pgrp than
304 * we are and we were the only connection outside.
305 */
306 parent = tsk->real_parent;
307 else
308 /* reparent: our child is in a different pgrp than
309 * we are, and it was the only connection outside.
310 */
311 ignored_task = NULL;
312
313 if (task_pgrp(parent) != pgrp &&
314 task_session(parent) == task_session(tsk) &&
315 will_become_orphaned_pgrp(pgrp, ignored_task) &&
316 has_stopped_jobs(pgrp)) {
317 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
318 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
319 }
320}
321
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322/**
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700323 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 *
325 * If a kernel thread is launched as a result of a system call, or if
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700326 * it ever exits, it should generally reparent itself to kthreadd so it
327 * isn't in the way of other processes and is correctly cleaned up on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 *
329 * The various task state such as scheduling policy and priority may have
330 * been inherited from a user process, so we reset them to sane values here.
331 *
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700332 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700334static void reparent_to_kthreadd(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335{
336 write_lock_irq(&tasklist_lock);
337
338 ptrace_unlink(current);
339 /* Reparent to init */
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700340 current->real_parent = current->parent = kthreadd_task;
Roland McGrathf4700212008-03-24 18:36:23 -0700341 list_move_tail(&current->sibling, &current->real_parent->children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
343 /* Set the exit signal to SIGCHLD so we signal init on exit */
344 current->exit_signal = SIGCHLD;
345
Ingo Molnare05606d2007-07-09 18:51:59 +0200346 if (task_nice(current) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 set_user_nice(current, 0);
348 /* cpus_allowed? */
349 /* rt_priority? */
350 /* signals? */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 memcpy(current->signal->rlim, init_task.signal->rlim,
352 sizeof(current->signal->rlim));
David Howellsd84f4f92008-11-14 10:39:23 +1100353
354 atomic_inc(&init_cred.usage);
355 commit_creds(&init_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357}
358
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800359void __set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360{
Oren Laadane19f2472006-01-08 01:03:58 -0800361 struct task_struct *curr = current->group_leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
Christian Borntraeger0d0df592009-10-26 16:49:34 -0700363 if (task_session(curr) != pid)
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700364 change_pid(curr, PIDTYPE_SID, pid);
Oleg Nesterov1b0f7ff2009-04-02 16:58:39 -0700365
366 if (task_pgrp(curr) != pid)
Oleg Nesterov7d8da092008-04-30 00:54:27 -0700367 change_pid(curr, PIDTYPE_PGID, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368}
369
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800370static void set_special_pids(struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371{
372 write_lock_irq(&tasklist_lock);
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800373 __set_special_pids(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 write_unlock_irq(&tasklist_lock);
375}
376
377/*
Oleg Nesterov87245132009-06-17 16:27:23 -0700378 * Let kernel threads use this to say that they allow a certain signal.
379 * Must not be used if kthread was cloned with CLONE_SIGHAND.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 */
381int allow_signal(int sig)
382{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700383 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 return -EINVAL;
385
386 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov87245132009-06-17 16:27:23 -0700387 /* This is only needed for daemonize()'ed kthreads */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 sigdelset(&current->blocked, sig);
Oleg Nesterov87245132009-06-17 16:27:23 -0700389 /*
390 * Kernel threads handle their own signals. Let the signal code
391 * know it'll be handled, so that they don't get converted to
392 * SIGKILL or just silently dropped.
393 */
394 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 recalc_sigpending();
396 spin_unlock_irq(&current->sighand->siglock);
397 return 0;
398}
399
400EXPORT_SYMBOL(allow_signal);
401
402int disallow_signal(int sig)
403{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700404 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 return -EINVAL;
406
407 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov10ab8252007-05-09 02:34:37 -0700408 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 recalc_sigpending();
410 spin_unlock_irq(&current->sighand->siglock);
411 return 0;
412}
413
414EXPORT_SYMBOL(disallow_signal);
415
416/*
417 * Put all the gunge required to become a kernel thread without
418 * attached user resources in one place where it belongs.
419 */
420
421void daemonize(const char *name, ...)
422{
423 va_list args;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 sigset_t blocked;
425
426 va_start(args, name);
427 vsnprintf(current->comm, sizeof(current->comm), name, args);
428 va_end(args);
429
430 /*
431 * If we were started as result of loading a module, close all of the
432 * user space pages. We don't need them, and if we didn't close them
433 * they would be locked into memory.
434 */
435 exit_mm(current);
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700436 /*
437 * We don't want to have TIF_FREEZE set if the system-wide hibernation
438 * or suspend transition begins right now.
439 */
Oleg Nesterov7b34e422008-07-25 01:47:37 -0700440 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
Oleg Nesterov8520d7c2008-02-08 04:19:09 -0800442 if (current->nsproxy != &init_nsproxy) {
443 get_nsproxy(&init_nsproxy);
444 switch_task_namespaces(current, &init_nsproxy);
445 }
Oleg Nesterov297bd422008-02-08 04:19:10 -0800446 set_special_pids(&init_struct_pid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -0800447 proc_clear_tty(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448
449 /* Block and flush all signals */
450 sigfillset(&blocked);
451 sigprocmask(SIG_BLOCK, &blocked, NULL);
452 flush_signals(current);
453
454 /* Become as one with the init task */
455
Al Viro3e93cd62009-03-29 19:00:13 -0400456 daemonize_fs_struct();
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700457 exit_files(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 current->files = init_task.files;
459 atomic_inc(&current->files->count);
460
Eric W. Biederman49d769d2007-05-09 02:34:33 -0700461 reparent_to_kthreadd();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462}
463
464EXPORT_SYMBOL(daemonize);
465
Arjan van de Ven858119e2006-01-14 13:20:43 -0800466static void close_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467{
468 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700469 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
471 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700472
473 /*
474 * It is safe to dereference the fd table without RCU or
475 * ->file_lock because this is the last reference to the
476 * files structure.
477 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700478 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 for (;;) {
480 unsigned long set;
481 i = j * __NFDBITS;
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800482 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700484 set = fdt->open_fds->fds_bits[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 while (set) {
486 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700487 struct file * file = xchg(&fdt->fd[i], NULL);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800488 if (file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 filp_close(file, files);
Ingo Molnar944be0b2007-02-12 00:52:26 -0800490 cond_resched();
491 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 }
493 i++;
494 set >>= 1;
495 }
496 }
497}
498
499struct files_struct *get_files_struct(struct task_struct *task)
500{
501 struct files_struct *files;
502
503 task_lock(task);
504 files = task->files;
505 if (files)
506 atomic_inc(&files->count);
507 task_unlock(task);
508
509 return files;
510}
511
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800512void put_files_struct(struct files_struct *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700514 struct fdtable *fdt;
515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 if (atomic_dec_and_test(&files->count)) {
517 close_files(files);
518 /*
519 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700520 * If the fdtable was embedded, pass files for freeing
521 * at the end of the RCU grace period. Otherwise,
522 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700524 fdt = files_fdtable(files);
Vadim Lobanov4fd45812006-12-10 02:21:17 -0800525 if (fdt != &files->fdtab)
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700526 kmem_cache_free(files_cachep, files);
Vadim Lobanov01b2d932006-12-22 01:10:43 -0800527 free_fdtable(fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 }
529}
530
Al Viro3b125382008-04-22 05:31:30 -0400531void reset_files_struct(struct files_struct *files)
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700532{
Al Viro3b125382008-04-22 05:31:30 -0400533 struct task_struct *tsk = current;
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700534 struct files_struct *old;
535
536 old = tsk->files;
537 task_lock(tsk);
538 tsk->files = files;
539 task_unlock(tsk);
540 put_files_struct(old);
541}
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700542
Al Viro1ec7f1d2008-04-22 05:35:42 -0400543void exit_files(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544{
545 struct files_struct * files = tsk->files;
546
547 if (files) {
548 task_lock(tsk);
549 tsk->files = NULL;
550 task_unlock(tsk);
551 put_files_struct(files);
552 }
553}
554
Balbir Singhcf475ad2008-04-29 01:00:16 -0700555#ifdef CONFIG_MM_OWNER
556/*
557 * Task p is exiting and it owned mm, lets find a new owner for it
558 */
559static inline int
560mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
561{
562 /*
563 * If there are other users of the mm and the owner (us) is exiting
564 * we need to find a new owner to take on the responsibility.
565 */
Balbir Singhcf475ad2008-04-29 01:00:16 -0700566 if (atomic_read(&mm->mm_users) <= 1)
567 return 0;
568 if (mm->owner != p)
569 return 0;
570 return 1;
571}
572
573void mm_update_next_owner(struct mm_struct *mm)
574{
575 struct task_struct *c, *g, *p = current;
576
577retry:
578 if (!mm_need_new_owner(mm, p))
579 return;
580
581 read_lock(&tasklist_lock);
582 /*
583 * Search in the children
584 */
585 list_for_each_entry(c, &p->children, sibling) {
586 if (c->mm == mm)
587 goto assign_new_owner;
588 }
589
590 /*
591 * Search in the siblings
592 */
Oleg Nesterovdea33cf2009-06-17 16:27:29 -0700593 list_for_each_entry(c, &p->real_parent->children, sibling) {
Balbir Singhcf475ad2008-04-29 01:00:16 -0700594 if (c->mm == mm)
595 goto assign_new_owner;
596 }
597
598 /*
599 * Search through everything else. We should not get
600 * here often
601 */
602 do_each_thread(g, c) {
603 if (c->mm == mm)
604 goto assign_new_owner;
605 } while_each_thread(g, c);
606
607 read_unlock(&tasklist_lock);
Balbir Singh31a78f22008-09-28 23:09:31 +0100608 /*
609 * We found no owner yet mm_users > 1: this implies that we are
610 * most likely racing with swapoff (try_to_unuse()) or /proc or
Hugh Dickinse5991372009-01-06 14:39:22 -0800611 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
Balbir Singh31a78f22008-09-28 23:09:31 +0100612 */
Balbir Singh31a78f22008-09-28 23:09:31 +0100613 mm->owner = NULL;
Balbir Singhcf475ad2008-04-29 01:00:16 -0700614 return;
615
616assign_new_owner:
617 BUG_ON(c == p);
618 get_task_struct(c);
619 /*
620 * The task_lock protects c->mm from changing.
621 * We always want mm->owner->mm == mm
622 */
623 task_lock(c);
Hugh Dickinse5991372009-01-06 14:39:22 -0800624 /*
625 * Delay read_unlock() till we have the task_lock()
626 * to ensure that c does not slip away underneath us
627 */
628 read_unlock(&tasklist_lock);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700629 if (c->mm != mm) {
630 task_unlock(c);
631 put_task_struct(c);
632 goto retry;
633 }
Balbir Singhcf475ad2008-04-29 01:00:16 -0700634 mm->owner = c;
635 task_unlock(c);
636 put_task_struct(c);
637}
638#endif /* CONFIG_MM_OWNER */
639
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640/*
641 * Turn us into a lazy TLB process if we
642 * aren't already..
643 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700644static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645{
646 struct mm_struct *mm = tsk->mm;
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700647 struct core_state *core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
649 mm_release(tsk, mm);
650 if (!mm)
651 return;
652 /*
653 * Serialize with any possible pending coredump.
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700654 * We must hold mmap_sem around checking core_state
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 * and clearing tsk->mm. The core-inducing thread
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700656 * will increment ->nr_threads for each thread in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 * group with ->mm != NULL.
658 */
659 down_read(&mm->mmap_sem);
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700660 core_state = mm->core_state;
661 if (core_state) {
662 struct core_thread self;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 up_read(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
Oleg Nesterovb564daf2008-07-25 01:47:44 -0700665 self.task = tsk;
666 self.next = xchg(&core_state->dumper.next, &self);
667 /*
668 * Implies mb(), the result of xchg() must be visible
669 * to core_state->dumper.
670 */
671 if (atomic_dec_and_test(&core_state->nr_threads))
672 complete(&core_state->startup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673
Oleg Nesterova94e2d42008-07-25 01:47:46 -0700674 for (;;) {
675 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
676 if (!self.task) /* see coredump_finish() */
677 break;
678 schedule();
679 }
680 __set_task_state(tsk, TASK_RUNNING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 down_read(&mm->mmap_sem);
682 }
683 atomic_inc(&mm->mm_count);
Eric Sesterhenn125e1872006-06-23 02:06:06 -0700684 BUG_ON(mm != tsk->active_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 /* more a memory barrier than a real lock */
686 task_lock(tsk);
687 tsk->mm = NULL;
688 up_read(&mm->mmap_sem);
689 enter_lazy_tlb(mm, current);
Rafael J. Wysocki0c1eecf2007-07-19 01:47:33 -0700690 /* We don't want this task to be frozen prematurely */
691 clear_freeze_flag(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 task_unlock(tsk);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700693 mm_update_next_owner(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 mmput(mm);
695}
696
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697/*
698 * When we die, we re-parent all our children.
699 * Try to give them to another thread in our thread
700 * group, and if no such member exists, give it to
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800701 * the child reaper process (ie "init") in our pid
702 * space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 */
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700704static struct task_struct *find_new_reaper(struct task_struct *father)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700706 struct pid_namespace *pid_ns = task_active_pid_ns(father);
707 struct task_struct *thread;
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700708
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700709 thread = father;
710 while_each_thread(father, thread) {
711 if (thread->flags & PF_EXITING)
712 continue;
713 if (unlikely(pid_ns->child_reaper == father))
714 pid_ns->child_reaper = thread;
715 return thread;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700717
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700718 if (unlikely(pid_ns->child_reaper == father)) {
719 write_unlock_irq(&tasklist_lock);
720 if (unlikely(pid_ns == &init_pid_ns))
721 panic("Attempted to kill init!");
722
723 zap_pid_ns_processes(pid_ns);
724 write_lock_irq(&tasklist_lock);
725 /*
726 * We can not clear ->child_reaper or leave it alone.
727 * There may by stealth EXIT_DEAD tasks on ->children,
728 * forget_original_parent() must move them somewhere.
729 */
730 pid_ns->child_reaper = init_pid_ns.child_reaper;
731 }
732
733 return pid_ns->child_reaper;
734}
735
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700736/*
737* Any that need to be release_task'd are put on the @dead list.
738 */
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800739static void reparent_leader(struct task_struct *father, struct task_struct *p,
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700740 struct list_head *dead)
741{
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700742 list_move_tail(&p->sibling, &p->real_parent->children);
743
744 if (task_detached(p))
745 return;
746 /*
747 * If this is a threaded reparent there is no need to
748 * notify anyone anything has happened.
749 */
750 if (same_thread_group(p->real_parent, father))
751 return;
752
753 /* We don't want people slaying init. */
754 p->exit_signal = SIGCHLD;
755
756 /* If it has exited notify the new parent about this child's death. */
Oleg Nesterov5cb11442009-06-17 16:27:30 -0700757 if (!task_ptrace(p) &&
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700758 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
759 do_notify_parent(p, p->exit_signal);
760 if (task_detached(p)) {
761 p->exit_state = EXIT_DEAD;
762 list_move_tail(&p->sibling, dead);
763 }
764 }
765
766 kill_orphaned_pgrp(p, father);
767}
768
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769static void forget_original_parent(struct task_struct *father)
770{
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700771 struct task_struct *p, *n, *reaper;
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700772 LIST_HEAD(dead_children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Oleg Nesterov39c626a2009-04-02 16:58:18 -0700774 exit_ptrace(father);
775
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 write_lock_irq(&tasklist_lock);
Oleg Nesterov950bbab2008-09-02 14:35:49 -0700777 reaper = find_new_reaper(father);
Roland McGrathf4700212008-03-24 18:36:23 -0700778
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 list_for_each_entry_safe(p, n, &father->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800780 struct task_struct *t = p;
781 do {
782 t->real_parent = reaper;
783 if (t->parent == father) {
784 BUG_ON(task_ptrace(t));
785 t->parent = t->real_parent;
786 }
787 if (t->pdeath_signal)
788 group_send_sig_info(t->pdeath_signal,
789 SEND_SIG_NOINFO, t);
790 } while_each_thread(p, t);
791 reparent_leader(father, p, &dead_children);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 }
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700793 write_unlock_irq(&tasklist_lock);
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700794
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700795 BUG_ON(!list_empty(&father->children));
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700796
Oleg Nesterov5dfc80b2009-04-02 16:58:19 -0700797 list_for_each_entry_safe(p, n, &dead_children, sibling) {
798 list_del_init(&p->sibling);
Oleg Nesterov39c626a2009-04-02 16:58:18 -0700799 release_task(p);
800 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801}
802
803/*
804 * Send signals to all our closest relatives so that they know
805 * to properly mourn us..
806 */
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300807static void exit_notify(struct task_struct *tsk, int group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700809 int signal;
810 void *cookie;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 /*
813 * This does two things:
814 *
815 * A. Make init inherit all the child processes
816 * B. Check to see if any process groups have become orphaned
817 * as a result of our exiting, and if they have any stopped
818 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
819 */
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700820 forget_original_parent(tsk);
Pavel Emelyanov2e4a7072007-10-18 23:40:01 -0700821 exit_task_namespaces(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
Oleg Nesterov762a24b2007-10-18 23:40:00 -0700823 write_lock_irq(&tasklist_lock);
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300824 if (group_dead)
825 kill_orphaned_pgrp(tsk->group_leader, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Oleg Nesterov24728442007-08-04 01:04:41 +0400827 /* Let father know we died
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 *
829 * Thread signals are configurable, but you aren't going to use
Daniel Walkerd4c5e412007-10-18 23:39:59 -0700830 * that to send signals to arbitary processes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 * That stops right now.
832 *
833 * If the parent exec id doesn't match the exec id we saved
834 * when we started then we know the parent has changed security
835 * domain.
836 *
837 * If our self_exec id doesn't match our parent_exec_id then
838 * we have changed execution domain as these two values started
839 * the same after a fork.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -0700841 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
Oleg Nesterovf49ee502008-03-02 21:44:40 +0300842 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
Oleg Nesterov432870d2009-04-06 16:16:02 +0200843 tsk->self_exec_id != tsk->parent_exec_id))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 tsk->exit_signal = SIGCHLD;
845
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700846 signal = tracehook_notify_death(tsk, &cookie, group_dead);
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700847 if (signal >= 0)
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700848 signal = do_notify_parent(tsk, signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700850 tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
Oleg Nesterov2800d8d2008-04-30 00:53:12 -0700852 /* mt-exec, de_thread() is waiting for us */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700853 if (thread_group_leader(tsk) &&
Steve VanDeBogart2633f0e2008-08-26 15:14:36 -0700854 tsk->signal->group_exit_task &&
855 tsk->signal->notify_count < 0)
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700856 wake_up_process(tsk->signal->group_exit_task);
857
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 write_unlock_irq(&tasklist_lock);
859
Roland McGrath2b2a1ff2008-07-25 19:45:54 -0700860 tracehook_report_death(tsk, signal, cookie, group_dead);
861
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 /* If the process is dead, release it - nobody will wait for it */
Roland McGrath5c7edcd2008-07-31 02:04:09 -0700863 if (signal == DEATH_REAP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 release_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865}
866
Jeff Dikee18eecb2007-07-15 23:38:48 -0700867#ifdef CONFIG_DEBUG_STACK_USAGE
868static void check_stack_usage(void)
869{
870 static DEFINE_SPINLOCK(low_water_lock);
871 static int lowest_to_date = THREAD_SIZE;
Jeff Dikee18eecb2007-07-15 23:38:48 -0700872 unsigned long free;
873
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500874 free = stack_not_used(current);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700875
876 if (free >= lowest_to_date)
877 return;
878
879 spin_lock(&low_water_lock);
880 if (free < lowest_to_date) {
881 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
882 "left\n",
883 current->comm, free);
884 lowest_to_date = free;
885 }
886 spin_unlock(&low_water_lock);
887}
888#else
889static inline void check_stack_usage(void) {}
890#endif
891
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800892NORET_TYPE void do_exit(long code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893{
894 struct task_struct *tsk = current;
895 int group_dead;
896
897 profile_task_exit(tsk);
898
Jens Axboe22e2c502005-06-27 10:55:12 +0200899 WARN_ON(atomic_read(&tsk->fs_excl));
900
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 if (unlikely(in_interrupt()))
902 panic("Aiee, killing interrupt handler!");
903 if (unlikely(!tsk->pid))
904 panic("Attempted to kill the idle task!");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
Roland McGrath30199f52008-07-25 19:45:46 -0700906 tracehook_report_exit(&code);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907
David Howellse0e81732009-09-02 09:13:40 +0100908 validate_creds_for_do_exit(tsk);
909
Alexander Nybergdf164db2005-06-23 00:09:13 -0700910 /*
911 * We're taking recursive faults here in do_exit. Safest is to just
912 * leave this task alone and wait for reboot.
913 */
914 if (unlikely(tsk->flags & PF_EXITING)) {
915 printk(KERN_ALERT
916 "Fixing recursive fault but reboot is needed!\n");
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700917 /*
918 * We can do this unlocked here. The futex code uses
919 * this flag just to verify whether the pi state
920 * cleanup has been done or not. In the worst case it
921 * loops once more. We pretend that the cleanup was
922 * done as there is no way to return. Either the
923 * OWNER_DIED bit is set by now or we push the blocked
924 * task into the wait for ever nirwana as well.
925 */
926 tsk->flags |= PF_EXITPIDONE;
Alexander Nybergdf164db2005-06-23 00:09:13 -0700927 set_current_state(TASK_UNINTERRUPTIBLE);
928 schedule();
929 }
930
Thomas Gleixner3aa551c2009-03-23 18:28:15 +0100931 exit_irq_thread();
932
Oleg Nesterovd12619b2008-02-08 04:19:12 -0800933 exit_signals(tsk); /* sets PF_EXITING */
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700934 /*
935 * tsk->flags are checked in the futex code to protect against
936 * an exiting task cleaning up the robust pi futexes.
937 */
Oleg Nesterovd2ee7192007-10-16 23:26:47 -0700938 smp_mb();
Thomas Gleixner1d615482009-11-17 14:54:03 +0100939 raw_spin_unlock_wait(&tsk->pi_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 if (unlikely(in_atomic()))
942 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -0700943 current->comm, task_pid_nr(current),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 preempt_count());
945
946 acct_update_integrals(tsk);
Oleg Nesterov901608d2009-01-06 14:40:29 -0800947
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -0700949 if (group_dead) {
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -0700950 hrtimer_cancel(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -0700951 exit_itimers(tsk->signal);
Jiri Pirko1f102062009-09-22 16:44:10 -0700952 if (tsk->mm)
953 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
Andrew Mortonc3068952005-08-04 16:49:32 -0700954 }
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700955 acct_collect(code, group_dead);
Miloslav Trmac522ed772007-07-15 23:40:56 -0700956 if (group_dead)
957 tty_audit_exit();
Al Virofa84cb92006-03-29 20:30:19 -0500958 if (unlikely(tsk->audit_context))
959 audit_free(tsk);
Oleg Nesterov115085e2006-12-06 20:36:51 -0800960
Jonathan Limf2ab6d82007-08-30 23:56:23 -0700961 tsk->exit_code = code;
Oleg Nesterov115085e2006-12-06 20:36:51 -0800962 taskstats_exit(tsk, group_dead);
Shailabh Nagarc7572492006-07-14 00:24:40 -0700963
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 exit_mm(tsk);
965
KaiGai Kohei0e464812006-06-25 05:49:24 -0700966 if (group_dead)
KaiGai Koheif6ec29a2006-06-25 05:49:25 -0700967 acct_process();
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -0400968 trace_sched_process_exit(tsk);
969
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 exit_sem(tsk);
Al Viro1ec7f1d2008-04-22 05:35:42 -0400971 exit_files(tsk);
972 exit_fs(tsk);
Jeff Dikee18eecb2007-07-15 23:38:48 -0700973 check_stack_usage();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 exit_thread();
Paul Menageb4f48b62007-10-18 23:39:33 -0700975 cgroup_exit(tsk, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976
Alan Cox5ec93d12009-11-30 13:18:45 +0000977 if (group_dead)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 disassociate_ctty(1);
979
Al Viroa1261f52005-11-13 16:06:55 -0800980 module_put(task_thread_info(tsk)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981
Matt Helsley9f460802005-11-07 00:59:16 -0800982 proc_exit_connector(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200983
984 /*
Frederic Weisbecker24f1e32c2009-09-09 19:22:48 +0200985 * FIXME: do that only when needed, using sched_exit tracepoint
986 */
987 flush_ptrace_hw_breakpoint(tsk);
988 /*
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200989 * Flush inherited counters to the parent - before the parent
990 * gets woken up by child-exit notifications.
991 */
Ingo Molnarcdd6c482009-09-21 12:02:48 +0200992 perf_event_exit_task(tsk);
Ingo Molnar33b2fb32009-05-17 11:08:41 +0200993
Oleg Nesterov821c7de2008-03-02 21:44:44 +0300994 exit_notify(tsk, group_dead);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -0700996 mpol_put(tsk->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 tsk->mempolicy = NULL;
998#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -0700999#ifdef CONFIG_FUTEX
Ingo Molnarc87e2832006-06-27 02:54:58 -07001000 if (unlikely(current->pi_state_cache))
1001 kfree(current->pi_state_cache);
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001002#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001003 /*
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001004 * Make sure we are holding no locks:
Ingo Molnarde5097c2006-01-09 15:59:21 -08001005 */
Ingo Molnar9a11b49a2006-07-03 00:24:33 -07001006 debug_check_no_locks_held(tsk);
Alexey Kuznetsov778e9a92007-06-08 13:47:00 -07001007 /*
1008 * We can do this unlocked here. The futex code uses this flag
1009 * just to verify whether the pi state cleanup has been done
1010 * or not. In the worst case it loops once more.
1011 */
1012 tsk->flags |= PF_EXITPIDONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013
Al Viroafc847b2006-02-28 12:51:55 -05001014 if (tsk->io_context)
Louis Rillingb69f2292009-12-04 14:52:42 +01001015 exit_io_context(tsk);
Al Viroafc847b2006-02-28 12:51:55 -05001016
Jens Axboeb92ce552006-04-11 13:52:07 +02001017 if (tsk->splice_pipe)
1018 __free_pipe_info(tsk->splice_pipe);
1019
David Howellse0e81732009-09-02 09:13:40 +01001020 validate_creds_for_do_exit(tsk);
1021
Coywolf Qi Hunt74072512005-10-30 15:02:47 -08001022 preempt_disable();
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001023 exit_rcu();
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001024 /* causes final put_task_struct in finish_task_switch(). */
Oleg Nesterovc394cc92006-09-29 02:01:11 -07001025 tsk->state = TASK_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 schedule();
1027 BUG();
1028 /* Avoid "noreturn function does return". */
Alan Cox54306cf2006-09-29 02:00:42 -07001029 for (;;)
1030 cpu_relax(); /* For when BUG is null */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031}
1032
Russ Anderson012914d2005-04-23 00:08:00 -07001033EXPORT_SYMBOL_GPL(do_exit);
1034
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1036{
1037 if (comp)
1038 complete(comp);
Oleg Nesterov55a101f2006-09-29 02:01:10 -07001039
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 do_exit(code);
1041}
1042
1043EXPORT_SYMBOL(complete_and_exit);
1044
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001045SYSCALL_DEFINE1(exit, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046{
1047 do_exit((error_code&0xff)<<8);
1048}
1049
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050/*
1051 * Take down every thread in the group. This is called by fatal signals
1052 * as well as by sys_exit_group (below).
1053 */
1054NORET_TYPE void
1055do_group_exit(int exit_code)
1056{
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001057 struct signal_struct *sig = current->signal;
1058
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1060
Oleg Nesterovbfc4b082008-04-30 00:52:36 -07001061 if (signal_group_exit(sig))
1062 exit_code = sig->group_exit_code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 else if (!thread_group_empty(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 struct sighand_struct *const sighand = current->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 spin_lock_irq(&sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001066 if (signal_group_exit(sig))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 /* Another thread got here before we took the lock. */
1068 exit_code = sig->group_exit_code;
1069 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 sig->group_exit_code = exit_code;
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001071 sig->flags = SIGNAL_GROUP_EXIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 zap_other_threads(current);
1073 }
1074 spin_unlock_irq(&sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 }
1076
1077 do_exit(exit_code);
1078 /* NOTREACHED */
1079}
1080
1081/*
1082 * this kills every thread in the thread group. Note that any externally
1083 * wait4()-ing process will get the correct exit code - even if this
1084 * thread is not the thread group leader.
1085 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001086SYSCALL_DEFINE1(exit_group, int, error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087{
1088 do_group_exit((error_code & 0xff) << 8);
Heiko Carstens2ed7c032009-01-14 14:13:54 +01001089 /* NOTREACHED */
1090 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091}
1092
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001093struct wait_opts {
1094 enum pid_type wo_type;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001095 int wo_flags;
Richard Kennedye1eb1eb2009-06-17 16:27:42 -07001096 struct pid *wo_pid;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001097
1098 struct siginfo __user *wo_info;
1099 int __user *wo_stat;
1100 struct rusage __user *wo_rusage;
1101
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001102 wait_queue_t child_wait;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001103 int notask_error;
1104};
1105
Oleg Nesterov989264f2009-09-23 15:56:49 -07001106static inline
1107struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
Eric W. Biederman161550d2008-02-08 04:19:14 -08001108{
Oleg Nesterov989264f2009-09-23 15:56:49 -07001109 if (type != PIDTYPE_PID)
1110 task = task->group_leader;
1111 return task->pids[type].pid;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001112}
1113
Oleg Nesterov989264f2009-09-23 15:56:49 -07001114static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001115{
1116 return wo->wo_type == PIDTYPE_MAX ||
1117 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1118}
1119
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001120static int eligible_child(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121{
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001122 if (!eligible_pid(wo, p))
1123 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 /* Wait for all children (clone and not) if __WALL is set;
1125 * otherwise, wait for clone children *only* if __WCLONE is
1126 * set; otherwise, wait for non-clone children *only*. (Note:
1127 * A "clone" child here is one that reports to its parent
1128 * using a signal other than SIGCHLD.) */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001129 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1130 && !(wo->wo_flags & __WALL))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132
Roland McGrath14dd0b82008-03-30 18:41:25 -07001133 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134}
1135
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001136static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1137 pid_t pid, uid_t uid, int why, int status)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001139 struct siginfo __user *infop;
1140 int retval = wo->wo_rusage
1141 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Ingo Molnar36c8b582006-07-03 00:25:41 -07001142
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001144 infop = wo->wo_info;
Vitaly Mayatskikhb6fe2d12009-09-23 15:56:52 -07001145 if (infop) {
1146 if (!retval)
1147 retval = put_user(SIGCHLD, &infop->si_signo);
1148 if (!retval)
1149 retval = put_user(0, &infop->si_errno);
1150 if (!retval)
1151 retval = put_user((short)why, &infop->si_code);
1152 if (!retval)
1153 retval = put_user(pid, &infop->si_pid);
1154 if (!retval)
1155 retval = put_user(uid, &infop->si_uid);
1156 if (!retval)
1157 retval = put_user(status, &infop->si_status);
1158 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 if (!retval)
1160 retval = pid;
1161 return retval;
1162}
1163
1164/*
1165 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1166 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1167 * the lock and this task is uninteresting. If we return nonzero, we have
1168 * released the lock and the system call should return.
1169 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001170static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171{
1172 unsigned long state;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001173 int retval, status, traced;
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001174 pid_t pid = task_pid_vnr(p);
David Howellsc69e8d92008-11-14 10:39:19 +11001175 uid_t uid = __task_cred(p)->uid;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001176 struct siginfo __user *infop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001178 if (!likely(wo->wo_flags & WEXITED))
Roland McGrath98abed02008-03-19 19:24:59 -07001179 return 0;
1180
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001181 if (unlikely(wo->wo_flags & WNOWAIT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 int exit_code = p->exit_code;
1183 int why, status;
1184
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 get_task_struct(p);
1186 read_unlock(&tasklist_lock);
1187 if ((exit_code & 0x7f) == 0) {
1188 why = CLD_EXITED;
1189 status = exit_code >> 8;
1190 } else {
1191 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1192 status = exit_code & 0x7f;
1193 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001194 return wait_noreap_copyout(wo, p, pid, uid, why, status);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 }
1196
1197 /*
1198 * Try to move the task's state to DEAD
1199 * only one thread is allowed to do this:
1200 */
1201 state = xchg(&p->exit_state, EXIT_DEAD);
1202 if (state != EXIT_ZOMBIE) {
1203 BUG_ON(state != EXIT_DEAD);
1204 return 0;
1205 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206
Oleg Nesterov53b6f9f2008-04-30 00:53:13 -07001207 traced = ptrace_reparented(p);
Oleg Nesterovbefca962009-06-18 16:49:11 -07001208 /*
1209 * It can be ptraced but not reparented, check
1210 * !task_detached() to filter out sub-threads.
1211 */
1212 if (likely(!traced) && likely(!task_detached(p))) {
Jesper Juhl3795e162006-01-09 20:54:39 -08001213 struct signal_struct *psig;
1214 struct signal_struct *sig;
Jiri Pirko1f102062009-09-22 16:44:10 -07001215 unsigned long maxrss;
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001216 cputime_t tgutime, tgstime;
Jesper Juhl3795e162006-01-09 20:54:39 -08001217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 /*
1219 * The resource counters for the group leader are in its
1220 * own task_struct. Those for dead threads in the group
1221 * are in its signal_struct, as are those for the child
1222 * processes it has previously reaped. All these
1223 * accumulate in the parent's signal_struct c* fields.
1224 *
1225 * We don't bother to take a lock here to protect these
1226 * p->signal fields, because they are only touched by
1227 * __exit_signal, which runs with tasklist_lock
1228 * write-locked anyway, and so is excluded here. We do
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001229 * need to protect the access to parent->signal fields,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 * as other threads in the parent group can be right
1231 * here reaping other children at the same time.
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001232 *
1233 * We use thread_group_times() to get times for the thread
1234 * group, which consolidates times for all threads in the
1235 * group including the group leader.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 */
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001237 thread_group_times(p, &tgutime, &tgstime);
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001238 spin_lock_irq(&p->real_parent->sighand->siglock);
1239 psig = p->real_parent->signal;
Jesper Juhl3795e162006-01-09 20:54:39 -08001240 sig = p->signal;
1241 psig->cutime =
1242 cputime_add(psig->cutime,
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001243 cputime_add(tgutime,
1244 sig->cutime));
Jesper Juhl3795e162006-01-09 20:54:39 -08001245 psig->cstime =
1246 cputime_add(psig->cstime,
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001247 cputime_add(tgstime,
1248 sig->cstime));
Laurent Vivier9ac52312007-10-15 17:00:19 +02001249 psig->cgtime =
1250 cputime_add(psig->cgtime,
1251 cputime_add(p->gtime,
1252 cputime_add(sig->gtime,
1253 sig->cgtime)));
Jesper Juhl3795e162006-01-09 20:54:39 -08001254 psig->cmin_flt +=
1255 p->min_flt + sig->min_flt + sig->cmin_flt;
1256 psig->cmaj_flt +=
1257 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1258 psig->cnvcsw +=
1259 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1260 psig->cnivcsw +=
1261 p->nivcsw + sig->nivcsw + sig->cnivcsw;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001262 psig->cinblock +=
1263 task_io_get_inblock(p) +
1264 sig->inblock + sig->cinblock;
1265 psig->coublock +=
1266 task_io_get_oublock(p) +
1267 sig->oublock + sig->coublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001268 maxrss = max(sig->maxrss, sig->cmaxrss);
1269 if (psig->cmaxrss < maxrss)
1270 psig->cmaxrss = maxrss;
Andrea Righi59954772008-07-27 17:29:15 +02001271 task_io_accounting_add(&psig->ioac, &p->ioac);
1272 task_io_accounting_add(&psig->ioac, &sig->ioac);
Oleg Nesterovd1e98f42009-06-17 16:27:34 -07001273 spin_unlock_irq(&p->real_parent->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 }
1275
1276 /*
1277 * Now we are sure this task is interesting, and no other
1278 * thread can reap it because we set its state to EXIT_DEAD.
1279 */
1280 read_unlock(&tasklist_lock);
1281
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001282 retval = wo->wo_rusage
1283 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1285 ? p->signal->group_exit_code : p->exit_code;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001286 if (!retval && wo->wo_stat)
1287 retval = put_user(status, wo->wo_stat);
1288
1289 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 if (!retval && infop)
1291 retval = put_user(SIGCHLD, &infop->si_signo);
1292 if (!retval && infop)
1293 retval = put_user(0, &infop->si_errno);
1294 if (!retval && infop) {
1295 int why;
1296
1297 if ((status & 0x7f) == 0) {
1298 why = CLD_EXITED;
1299 status >>= 8;
1300 } else {
1301 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1302 status &= 0x7f;
1303 }
1304 retval = put_user((short)why, &infop->si_code);
1305 if (!retval)
1306 retval = put_user(status, &infop->si_status);
1307 }
1308 if (!retval && infop)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001309 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310 if (!retval && infop)
David Howellsc69e8d92008-11-14 10:39:19 +11001311 retval = put_user(uid, &infop->si_uid);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001312 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001313 retval = pid;
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001314
1315 if (traced) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 write_lock_irq(&tasklist_lock);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001317 /* We dropped tasklist, ptracer could die and untrace */
1318 ptrace_unlink(p);
1319 /*
1320 * If this is not a detached task, notify the parent.
1321 * If it's still not detached after that, don't release
1322 * it now.
1323 */
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001324 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001325 do_notify_parent(p, p->exit_signal);
Oleg Nesterovd839fd42008-04-30 00:53:11 -07001326 if (!task_detached(p)) {
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001327 p->exit_state = EXIT_ZOMBIE;
1328 p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 }
1330 }
1331 write_unlock_irq(&tasklist_lock);
1332 }
1333 if (p != NULL)
1334 release_task(p);
Oleg Nesterov2f4e6e22007-10-16 23:26:58 -07001335
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 return retval;
1337}
1338
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001339static int *task_stopped_code(struct task_struct *p, bool ptrace)
1340{
1341 if (ptrace) {
1342 if (task_is_stopped_or_traced(p))
1343 return &p->exit_code;
1344 } else {
1345 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1346 return &p->signal->group_exit_code;
1347 }
1348 return NULL;
1349}
1350
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351/*
1352 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1353 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1354 * the lock and this task is uninteresting. If we return nonzero, we have
1355 * released the lock and the system call should return.
1356 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001357static int wait_task_stopped(struct wait_opts *wo,
1358 int ptrace, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001360 struct siginfo __user *infop;
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001361 int retval, exit_code, *p_code, why;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001362 uid_t uid = 0; /* unneeded, required by compiler */
Oleg Nesterovc8950782007-11-28 16:21:24 -08001363 pid_t pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364
Oleg Nesterov47918022009-06-17 16:27:39 -07001365 /*
1366 * Traditionally we see ptrace'd stopped tasks regardless of options.
1367 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001368 if (!ptrace && !(wo->wo_flags & WUNTRACED))
Roland McGrath98abed02008-03-19 19:24:59 -07001369 return 0;
1370
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001371 exit_code = 0;
1372 spin_lock_irq(&p->sighand->siglock);
1373
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001374 p_code = task_stopped_code(p, ptrace);
1375 if (unlikely(!p_code))
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001376 goto unlock_sig;
1377
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001378 exit_code = *p_code;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001379 if (!exit_code)
1380 goto unlock_sig;
1381
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001382 if (!unlikely(wo->wo_flags & WNOWAIT))
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001383 *p_code = 0;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001384
David Howellsc69e8d92008-11-14 10:39:19 +11001385 /* don't need the RCU readlock here as we're holding a spinlock */
1386 uid = __task_cred(p)->uid;
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001387unlock_sig:
1388 spin_unlock_irq(&p->sighand->siglock);
1389 if (!exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 return 0;
1391
1392 /*
1393 * Now we are pretty sure this task is interesting.
1394 * Make sure it doesn't get reaped out from under us while we
1395 * give up the lock and then examine it below. We don't want to
1396 * keep holding onto the tasklist_lock while we call getrusage and
1397 * possibly take page faults for user memory.
1398 */
1399 get_task_struct(p);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001400 pid = task_pid_vnr(p);
Roland McGrathf4700212008-03-24 18:36:23 -07001401 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 read_unlock(&tasklist_lock);
1403
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001404 if (unlikely(wo->wo_flags & WNOWAIT))
1405 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001407 retval = wo->wo_rusage
1408 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1409 if (!retval && wo->wo_stat)
1410 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1411
1412 infop = wo->wo_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 if (!retval && infop)
1414 retval = put_user(SIGCHLD, &infop->si_signo);
1415 if (!retval && infop)
1416 retval = put_user(0, &infop->si_errno);
1417 if (!retval && infop)
Roland McGrath6efcae42008-03-08 11:41:22 -08001418 retval = put_user((short)why, &infop->si_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 if (!retval && infop)
1420 retval = put_user(exit_code, &infop->si_status);
1421 if (!retval && infop)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001422 retval = put_user(pid, &infop->si_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 if (!retval && infop)
Oleg Nesterovee7c82d2008-02-08 04:19:01 -08001424 retval = put_user(uid, &infop->si_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 if (!retval)
Oleg Nesterovc8950782007-11-28 16:21:24 -08001426 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 put_task_struct(p);
1428
1429 BUG_ON(!retval);
1430 return retval;
1431}
1432
1433/*
1434 * Handle do_wait work for one task in a live, non-stopped state.
1435 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1436 * the lock and this task is uninteresting. If we return nonzero, we have
1437 * released the lock and the system call should return.
1438 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001439static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440{
1441 int retval;
1442 pid_t pid;
1443 uid_t uid;
1444
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001445 if (!unlikely(wo->wo_flags & WCONTINUED))
Roland McGrath98abed02008-03-19 19:24:59 -07001446 return 0;
1447
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1449 return 0;
1450
1451 spin_lock_irq(&p->sighand->siglock);
1452 /* Re-check with the lock held. */
1453 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1454 spin_unlock_irq(&p->sighand->siglock);
1455 return 0;
1456 }
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001457 if (!unlikely(wo->wo_flags & WNOWAIT))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
David Howellsc69e8d92008-11-14 10:39:19 +11001459 uid = __task_cred(p)->uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 spin_unlock_irq(&p->sighand->siglock);
1461
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001462 pid = task_pid_vnr(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 get_task_struct(p);
1464 read_unlock(&tasklist_lock);
1465
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001466 if (!wo->wo_info) {
1467 retval = wo->wo_rusage
1468 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 put_task_struct(p);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001470 if (!retval && wo->wo_stat)
1471 retval = put_user(0xffff, wo->wo_stat);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 if (!retval)
Oleg Nesterov3a515e42008-02-08 04:19:07 -08001473 retval = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 } else {
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001475 retval = wait_noreap_copyout(wo, p, pid, uid,
1476 CLD_CONTINUED, SIGCONT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 BUG_ON(retval == 0);
1478 }
1479
1480 return retval;
1481}
1482
Roland McGrath98abed02008-03-19 19:24:59 -07001483/*
1484 * Consider @p for a wait by @parent.
1485 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001486 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001487 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1488 * Returns zero if the search for a child should continue;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001489 * then ->notask_error is 0 if @p is an eligible child,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001490 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001491 */
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001492static int wait_consider_task(struct wait_opts *wo, int ptrace,
1493 struct task_struct *p)
Roland McGrath98abed02008-03-19 19:24:59 -07001494{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001495 int ret = eligible_child(wo, p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001496 if (!ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001497 return ret;
1498
Oleg Nesterova2322e12009-09-23 15:56:45 -07001499 ret = security_task_wait(p);
Roland McGrath14dd0b82008-03-30 18:41:25 -07001500 if (unlikely(ret < 0)) {
1501 /*
1502 * If we have not yet seen any eligible child,
1503 * then let this error code replace -ECHILD.
1504 * A permission error will give the user a clue
1505 * to look for security policy problems, rather
1506 * than for mysterious wait bugs.
1507 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001508 if (wo->notask_error)
1509 wo->notask_error = ret;
Oleg Nesterov78a3d9d2009-04-29 18:01:23 +02001510 return 0;
Roland McGrath14dd0b82008-03-30 18:41:25 -07001511 }
1512
Oleg Nesterov5cb11442009-06-17 16:27:30 -07001513 if (likely(!ptrace) && unlikely(task_ptrace(p))) {
Roland McGrathf4700212008-03-24 18:36:23 -07001514 /*
1515 * This child is hidden by ptrace.
1516 * We aren't allowed to see it now, but eventually we will.
1517 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001518 wo->notask_error = 0;
Roland McGrathf4700212008-03-24 18:36:23 -07001519 return 0;
1520 }
1521
Roland McGrath98abed02008-03-19 19:24:59 -07001522 if (p->exit_state == EXIT_DEAD)
1523 return 0;
1524
1525 /*
1526 * We don't reap group leaders with subthreads.
1527 */
1528 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001529 return wait_task_zombie(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001530
1531 /*
1532 * It's stopped or running now, so it might
1533 * later continue, exit, or stop again.
1534 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001535 wo->notask_error = 0;
Roland McGrath98abed02008-03-19 19:24:59 -07001536
Oleg Nesterov90bc8d82009-04-02 16:57:58 -07001537 if (task_stopped_code(p, ptrace))
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001538 return wait_task_stopped(wo, ptrace, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001539
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001540 return wait_task_continued(wo, p);
Roland McGrath98abed02008-03-19 19:24:59 -07001541}
1542
1543/*
1544 * Do the work of do_wait() for one thread in the group, @tsk.
1545 *
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001546 * -ECHILD should be in ->notask_error before the first call.
Roland McGrath98abed02008-03-19 19:24:59 -07001547 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1548 * Returns zero if the search for a child should continue; then
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001549 * ->notask_error is 0 if there were any eligible children,
Roland McGrath14dd0b82008-03-30 18:41:25 -07001550 * or another error from security_task_wait(), or still -ECHILD.
Roland McGrath98abed02008-03-19 19:24:59 -07001551 */
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001552static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001553{
1554 struct task_struct *p;
1555
1556 list_for_each_entry(p, &tsk->children, sibling) {
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001557 int ret = wait_consider_task(wo, 0, p);
1558 if (ret)
1559 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001560 }
1561
1562 return 0;
1563}
1564
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001565static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
Roland McGrath98abed02008-03-19 19:24:59 -07001566{
1567 struct task_struct *p;
1568
Roland McGrathf4700212008-03-24 18:36:23 -07001569 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
Oleg Nesterovb6e763f2009-09-23 15:56:50 -07001570 int ret = wait_consider_task(wo, 1, p);
Roland McGrathf4700212008-03-24 18:36:23 -07001571 if (ret)
Roland McGrath98abed02008-03-19 19:24:59 -07001572 return ret;
Roland McGrath98abed02008-03-19 19:24:59 -07001573 }
1574
1575 return 0;
1576}
1577
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001578static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1579 int sync, void *key)
1580{
1581 struct wait_opts *wo = container_of(wait, struct wait_opts,
1582 child_wait);
1583 struct task_struct *p = key;
1584
Oleg Nesterov5c01ba42009-09-23 15:56:48 -07001585 if (!eligible_pid(wo, p))
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001586 return 0;
1587
Oleg Nesterovb4fe5182009-09-23 15:56:47 -07001588 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1589 return 0;
1590
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001591 return default_wake_function(wait, mode, sync, key);
1592}
1593
Oleg Nesterova7f07652009-09-23 15:56:44 -07001594void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1595{
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001596 __wake_up_sync_key(&parent->signal->wait_chldexit,
1597 TASK_INTERRUPTIBLE, 1, p);
Oleg Nesterova7f07652009-09-23 15:56:44 -07001598}
1599
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001600static long do_wait(struct wait_opts *wo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 struct task_struct *tsk;
Roland McGrath98abed02008-03-19 19:24:59 -07001603 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001605 trace_sched_process_wait(wo->wo_pid);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001606
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001607 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1608 wo->child_wait.private = current;
1609 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610repeat:
Roland McGrath98abed02008-03-19 19:24:59 -07001611 /*
1612 * If there is nothing that can match our critiera just get out.
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001613 * We will clear ->notask_error to zero if we see any child that
1614 * might later match our criteria, even if we are not able to reap
1615 * it yet.
Roland McGrath98abed02008-03-19 19:24:59 -07001616 */
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001617 wo->notask_error = -ECHILD;
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001618 if ((wo->wo_type < PIDTYPE_MAX) &&
1619 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001620 goto notask;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001621
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001622 set_current_state(TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 read_lock(&tasklist_lock);
1624 tsk = current;
1625 do {
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001626 retval = do_wait_thread(wo, tsk);
1627 if (retval)
Roland McGrath98abed02008-03-19 19:24:59 -07001628 goto end;
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001629
1630 retval = ptrace_do_wait(wo, tsk);
1631 if (retval)
1632 goto end;
Roland McGrath98abed02008-03-19 19:24:59 -07001633
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001634 if (wo->wo_flags & __WNOTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 break;
Oleg Nesterova3f6dfb2009-06-17 16:27:41 -07001636 } while_each_thread(current, tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 read_unlock(&tasklist_lock);
Oleg Nesterovf2cc3eb2008-02-08 04:19:06 -08001638
Oleg Nesterov64a16ca2009-06-17 16:27:40 -07001639notask:
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001640 retval = wo->notask_error;
1641 if (!retval && !(wo->wo_flags & WNOHANG)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 retval = -ERESTARTSYS;
Roland McGrath98abed02008-03-19 19:24:59 -07001643 if (!signal_pending(current)) {
1644 schedule();
1645 goto repeat;
1646 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648end:
Oleg Nesterovf95d39d2009-06-17 16:27:42 -07001649 __set_current_state(TASK_RUNNING);
Oleg Nesterov0b7570e2009-09-23 15:56:46 -07001650 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 return retval;
1652}
1653
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001654SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1655 infop, int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001657 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001658 struct pid *pid = NULL;
1659 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 long ret;
1661
1662 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1663 return -EINVAL;
1664 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1665 return -EINVAL;
1666
1667 switch (which) {
1668 case P_ALL:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001669 type = PIDTYPE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 break;
1671 case P_PID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001672 type = PIDTYPE_PID;
1673 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 return -EINVAL;
1675 break;
1676 case P_PGID:
Eric W. Biederman161550d2008-02-08 04:19:14 -08001677 type = PIDTYPE_PGID;
1678 if (upid <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 break;
1681 default:
1682 return -EINVAL;
1683 }
1684
Eric W. Biederman161550d2008-02-08 04:19:14 -08001685 if (type < PIDTYPE_MAX)
1686 pid = find_get_pid(upid);
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001687
1688 wo.wo_type = type;
1689 wo.wo_pid = pid;
1690 wo.wo_flags = options;
1691 wo.wo_info = infop;
1692 wo.wo_stat = NULL;
1693 wo.wo_rusage = ru;
1694 ret = do_wait(&wo);
Vitaly Mayatskikhdfe16df2009-09-23 15:56:51 -07001695
1696 if (ret > 0) {
1697 ret = 0;
1698 } else if (infop) {
1699 /*
1700 * For a WNOHANG return, clear out all the fields
1701 * we would set so the user can easily tell the
1702 * difference.
1703 */
1704 if (!ret)
1705 ret = put_user(0, &infop->si_signo);
1706 if (!ret)
1707 ret = put_user(0, &infop->si_errno);
1708 if (!ret)
1709 ret = put_user(0, &infop->si_code);
1710 if (!ret)
1711 ret = put_user(0, &infop->si_pid);
1712 if (!ret)
1713 ret = put_user(0, &infop->si_uid);
1714 if (!ret)
1715 ret = put_user(0, &infop->si_status);
1716 }
1717
Eric W. Biederman161550d2008-02-08 04:19:14 -08001718 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719
1720 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001721 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 return ret;
1723}
1724
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001725SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1726 int, options, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727{
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001728 struct wait_opts wo;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001729 struct pid *pid = NULL;
1730 enum pid_type type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 long ret;
1732
1733 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1734 __WNOTHREAD|__WCLONE|__WALL))
1735 return -EINVAL;
Eric W. Biederman161550d2008-02-08 04:19:14 -08001736
1737 if (upid == -1)
1738 type = PIDTYPE_MAX;
1739 else if (upid < 0) {
1740 type = PIDTYPE_PGID;
1741 pid = find_get_pid(-upid);
1742 } else if (upid == 0) {
1743 type = PIDTYPE_PGID;
Oleg Nesterov2ae448e2009-04-02 16:58:36 -07001744 pid = get_task_pid(current, PIDTYPE_PGID);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001745 } else /* upid > 0 */ {
1746 type = PIDTYPE_PID;
1747 pid = find_get_pid(upid);
1748 }
1749
Oleg Nesterov9e8ae012009-06-17 16:27:39 -07001750 wo.wo_type = type;
1751 wo.wo_pid = pid;
1752 wo.wo_flags = options | WEXITED;
1753 wo.wo_info = NULL;
1754 wo.wo_stat = stat_addr;
1755 wo.wo_rusage = ru;
1756 ret = do_wait(&wo);
Eric W. Biederman161550d2008-02-08 04:19:14 -08001757 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758
1759 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001760 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 return ret;
1762}
1763
1764#ifdef __ARCH_WANT_SYS_WAITPID
1765
1766/*
1767 * sys_waitpid() remains for compatibility. waitpid() should be
1768 * implemented by calling sys_wait4() from libc.a.
1769 */
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001770SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771{
1772 return sys_wait4(pid, stat_addr, options, NULL);
1773}
1774
1775#endif