]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - kernel/exit.c
notifier-error-inject: fix module names in Kconfig
[mirror_ubuntu-artful-kernel.git] / kernel / exit.c
CommitLineData
1da177e4
LT
1/*
2 * linux/kernel/exit.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
1da177e4
LT
7#include <linux/mm.h>
8#include <linux/slab.h>
9#include <linux/interrupt.h>
1da177e4 10#include <linux/module.h>
c59ede7b 11#include <linux/capability.h>
1da177e4
LT
12#include <linux/completion.h>
13#include <linux/personality.h>
14#include <linux/tty.h>
da9cbc87 15#include <linux/iocontext.h>
1da177e4
LT
16#include <linux/key.h>
17#include <linux/security.h>
18#include <linux/cpu.h>
19#include <linux/acct.h>
8f0ab514 20#include <linux/tsacct_kern.h>
1da177e4 21#include <linux/file.h>
9f3acc31 22#include <linux/fdtable.h>
80d26af8 23#include <linux/freezer.h>
1da177e4 24#include <linux/binfmts.h>
ab516013 25#include <linux/nsproxy.h>
84d73786 26#include <linux/pid_namespace.h>
1da177e4
LT
27#include <linux/ptrace.h>
28#include <linux/profile.h>
29#include <linux/mount.h>
30#include <linux/proc_fs.h>
49d769d5 31#include <linux/kthread.h>
1da177e4 32#include <linux/mempolicy.h>
c757249a 33#include <linux/taskstats_kern.h>
ca74e92b 34#include <linux/delayacct.h>
b4f48b63 35#include <linux/cgroup.h>
1da177e4 36#include <linux/syscalls.h>
7ed20e1a 37#include <linux/signal.h>
6a14c5c9 38#include <linux/posix-timers.h>
9f46080c 39#include <linux/cn_proc.h>
de5097c2 40#include <linux/mutex.h>
0771dfef 41#include <linux/futex.h>
b92ce558 42#include <linux/pipe_fs_i.h>
fa84cb93 43#include <linux/audit.h> /* for audit_free() */
83cc5ed3 44#include <linux/resource.h>
0d67a46d 45#include <linux/blkdev.h>
6eaeeaba 46#include <linux/task_io_accounting_ops.h>
30199f5a 47#include <linux/tracehook.h>
5ad4e53b 48#include <linux/fs_struct.h>
d84f4f99 49#include <linux/init_task.h>
cdd6c482 50#include <linux/perf_event.h>
ad8d75ff 51#include <trace/events/sched.h>
24f1e32c 52#include <linux/hw_breakpoint.h>
3d5992d2 53#include <linux/oom.h>
54848d73 54#include <linux/writeback.h>
40401530 55#include <linux/shm.h>
1da177e4
LT
56
57#include <asm/uaccess.h>
58#include <asm/unistd.h>
59#include <asm/pgtable.h>
60#include <asm/mmu_context.h>
61
408b664a
AB
62static void exit_mm(struct task_struct * tsk);
63
d40e48e0 64static void __unhash_process(struct task_struct *p, bool group_dead)
1da177e4
LT
65{
66 nr_threads--;
50d75f8d 67 detach_pid(p, PIDTYPE_PID);
d40e48e0 68 if (group_dead) {
1da177e4
LT
69 detach_pid(p, PIDTYPE_PGID);
70 detach_pid(p, PIDTYPE_SID);
c97d9893 71
5e85d4ab 72 list_del_rcu(&p->tasks);
9cd80bbb 73 list_del_init(&p->sibling);
909ea964 74 __this_cpu_dec(process_counts);
1da177e4 75 }
47e65328 76 list_del_rcu(&p->thread_group);
1da177e4
LT
77}
78
6a14c5c9
ON
79/*
80 * This function expects the tasklist_lock write-locked.
81 */
82static void __exit_signal(struct task_struct *tsk)
83{
84 struct signal_struct *sig = tsk->signal;
d40e48e0 85 bool group_dead = thread_group_leader(tsk);
6a14c5c9 86 struct sighand_struct *sighand;
4ada856f 87 struct tty_struct *uninitialized_var(tty);
6fac4829 88 cputime_t utime, stime;
6a14c5c9 89
d11c563d 90 sighand = rcu_dereference_check(tsk->sighand,
db1466b3 91 lockdep_tasklist_lock_is_held());
6a14c5c9
ON
92 spin_lock(&sighand->siglock);
93
94 posix_cpu_timers_exit(tsk);
d40e48e0 95 if (group_dead) {
6a14c5c9 96 posix_cpu_timers_exit_group(tsk);
4ada856f
ON
97 tty = sig->tty;
98 sig->tty = NULL;
4a599942 99 } else {
e0a70217
ON
100 /*
101 * This can only happen if the caller is de_thread().
102 * FIXME: this is the temporary hack, we should teach
103 * posix-cpu-timers to handle this case correctly.
104 */
105 if (unlikely(has_group_leader_pid(tsk)))
106 posix_cpu_timers_exit_group(tsk);
107
6a14c5c9
ON
108 /*
109 * If there is any task waiting for the group exit
110 * then notify it:
111 */
d344193a 112 if (sig->notify_count > 0 && !--sig->notify_count)
6a14c5c9 113 wake_up_process(sig->group_exit_task);
6db840fa 114
6a14c5c9
ON
115 if (tsk == sig->curr_target)
116 sig->curr_target = next_thread(tsk);
117 /*
118 * Accumulate here the counters for all threads but the
119 * group leader as they die, so they can be added into
120 * the process-wide totals when those are taken.
121 * The group leader stays around as a zombie as long
122 * as there are other threads. When it gets reaped,
123 * the exit.c code will add its counts into these totals.
124 * We won't ever get here for the group leader, since it
125 * will have been the last reference on the signal_struct.
126 */
6fac4829
FW
127 task_cputime(tsk, &utime, &stime);
128 sig->utime += utime;
129 sig->stime += stime;
130 sig->gtime += task_gtime(tsk);
6a14c5c9
ON
131 sig->min_flt += tsk->min_flt;
132 sig->maj_flt += tsk->maj_flt;
133 sig->nvcsw += tsk->nvcsw;
134 sig->nivcsw += tsk->nivcsw;
6eaeeaba
ED
135 sig->inblock += task_io_get_inblock(tsk);
136 sig->oublock += task_io_get_oublock(tsk);
5995477a 137 task_io_accounting_add(&sig->ioac, &tsk->ioac);
32bd671d 138 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
6a14c5c9
ON
139 }
140
b3ac022c 141 sig->nr_threads--;
d40e48e0 142 __unhash_process(tsk, group_dead);
5876700c 143
da7978b0
ON
144 /*
145 * Do this under ->siglock, we can race with another thread
146 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
147 */
148 flush_sigqueue(&tsk->pending);
a7e5328a 149 tsk->sighand = NULL;
6a14c5c9 150 spin_unlock(&sighand->siglock);
6a14c5c9 151
a7e5328a 152 __cleanup_sighand(sighand);
6a14c5c9 153 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
d40e48e0 154 if (group_dead) {
6a14c5c9 155 flush_sigqueue(&sig->shared_pending);
4ada856f 156 tty_kref_put(tty);
6a14c5c9
ON
157 }
158}
159
8c7904a0
EB
160static void delayed_put_task_struct(struct rcu_head *rhp)
161{
0a16b607
MD
162 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
163
4e231c79 164 perf_event_delayed_put(tsk);
0a16b607
MD
165 trace_sched_process_free(tsk);
166 put_task_struct(tsk);
8c7904a0
EB
167}
168
f470021a 169
1da177e4
LT
170void release_task(struct task_struct * p)
171{
36c8b586 172 struct task_struct *leader;
1da177e4 173 int zap_leader;
1f09f974 174repeat:
c69e8d9c 175 /* don't need to get the RCU readlock here - the process is dead and
d11c563d
PM
176 * can't be modifying its own credentials. But shut RCU-lockdep up */
177 rcu_read_lock();
c69e8d9c 178 atomic_dec(&__task_cred(p)->user->processes);
d11c563d 179 rcu_read_unlock();
c69e8d9c 180
60347f67 181 proc_flush_task(p);
0203026b 182
1da177e4 183 write_lock_irq(&tasklist_lock);
a288eecc 184 ptrace_release_task(p);
1da177e4 185 __exit_signal(p);
35f5cad8 186
1da177e4
LT
187 /*
188 * If we are the last non-leader member of the thread
189 * group, and the leader is zombie, then notify the
190 * group leader's parent process. (if it wants notification.)
191 */
192 zap_leader = 0;
193 leader = p->group_leader;
194 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
1da177e4
LT
195 /*
196 * If we were the last child thread and the leader has
197 * exited already, and the leader's parent ignores SIGCHLD,
198 * then we are the one who should release the leader.
dae33574 199 */
86773473 200 zap_leader = do_notify_parent(leader, leader->exit_signal);
dae33574
RM
201 if (zap_leader)
202 leader->exit_state = EXIT_DEAD;
1da177e4
LT
203 }
204
1da177e4 205 write_unlock_irq(&tasklist_lock);
1da177e4 206 release_thread(p);
8c7904a0 207 call_rcu(&p->rcu, delayed_put_task_struct);
1da177e4
LT
208
209 p = leader;
210 if (unlikely(zap_leader))
211 goto repeat;
212}
213
1da177e4
LT
214/*
215 * This checks not only the pgrp, but falls back on the pid if no
216 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
217 * without this...
04a2e6a5
EB
218 *
219 * The caller must hold rcu lock or the tasklist lock.
1da177e4 220 */
04a2e6a5 221struct pid *session_of_pgrp(struct pid *pgrp)
1da177e4
LT
222{
223 struct task_struct *p;
04a2e6a5 224 struct pid *sid = NULL;
62dfb554 225
04a2e6a5 226 p = pid_task(pgrp, PIDTYPE_PGID);
62dfb554 227 if (p == NULL)
04a2e6a5 228 p = pid_task(pgrp, PIDTYPE_PID);
62dfb554 229 if (p != NULL)
04a2e6a5 230 sid = task_session(p);
62dfb554 231
1da177e4
LT
232 return sid;
233}
234
235/*
236 * Determine if a process group is "orphaned", according to the POSIX
237 * definition in 2.2.2.52. Orphaned process groups are not to be affected
238 * by terminal-generated stop signals. Newly orphaned process groups are
239 * to receive a SIGHUP and a SIGCONT.
240 *
241 * "I ask you, have you ever known what it is to be an orphan?"
242 */
0475ac08 243static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
1da177e4
LT
244{
245 struct task_struct *p;
1da177e4 246
0475ac08 247 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
05e83df6
ON
248 if ((p == ignored_task) ||
249 (p->exit_state && thread_group_empty(p)) ||
250 is_global_init(p->real_parent))
1da177e4 251 continue;
05e83df6 252
0475ac08 253 if (task_pgrp(p->real_parent) != pgrp &&
05e83df6
ON
254 task_session(p->real_parent) == task_session(p))
255 return 0;
0475ac08 256 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
05e83df6
ON
257
258 return 1;
1da177e4
LT
259}
260
3e7cd6c4 261int is_current_pgrp_orphaned(void)
1da177e4
LT
262{
263 int retval;
264
265 read_lock(&tasklist_lock);
3e7cd6c4 266 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
1da177e4
LT
267 read_unlock(&tasklist_lock);
268
269 return retval;
270}
271
961c4675 272static bool has_stopped_jobs(struct pid *pgrp)
1da177e4 273{
1da177e4
LT
274 struct task_struct *p;
275
0475ac08 276 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
961c4675
ON
277 if (p->signal->flags & SIGNAL_STOP_STOPPED)
278 return true;
0475ac08 279 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
961c4675
ON
280
281 return false;
1da177e4
LT
282}
283
f49ee505
ON
284/*
285 * Check to see if any process groups have become orphaned as
286 * a result of our exiting, and if they have any stopped jobs,
287 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
288 */
289static void
290kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
291{
292 struct pid *pgrp = task_pgrp(tsk);
293 struct task_struct *ignored_task = tsk;
294
295 if (!parent)
296 /* exit: our father is in a different pgrp than
297 * we are and we were the only connection outside.
298 */
299 parent = tsk->real_parent;
300 else
301 /* reparent: our child is in a different pgrp than
302 * we are, and it was the only connection outside.
303 */
304 ignored_task = NULL;
305
306 if (task_pgrp(parent) != pgrp &&
307 task_session(parent) == task_session(tsk) &&
308 will_become_orphaned_pgrp(pgrp, ignored_task) &&
309 has_stopped_jobs(pgrp)) {
310 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
311 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
312 }
313}
314
8520d7c7 315void __set_special_pids(struct pid *pid)
1da177e4 316{
e19f247a 317 struct task_struct *curr = current->group_leader;
1da177e4 318
0d0df599 319 if (task_session(curr) != pid)
7d8da096 320 change_pid(curr, PIDTYPE_SID, pid);
1b0f7ffd
ON
321
322 if (task_pgrp(curr) != pid)
7d8da096 323 change_pid(curr, PIDTYPE_PGID, pid);
1da177e4
LT
324}
325
1da177e4 326/*
87245135
ON
327 * Let kernel threads use this to say that they allow a certain signal.
328 * Must not be used if kthread was cloned with CLONE_SIGHAND.
1da177e4
LT
329 */
330int allow_signal(int sig)
331{
7ed20e1a 332 if (!valid_signal(sig) || sig < 1)
1da177e4
LT
333 return -EINVAL;
334
335 spin_lock_irq(&current->sighand->siglock);
87245135 336 /* This is only needed for daemonize()'ed kthreads */
1da177e4 337 sigdelset(&current->blocked, sig);
87245135
ON
338 /*
339 * Kernel threads handle their own signals. Let the signal code
340 * know it'll be handled, so that they don't get converted to
341 * SIGKILL or just silently dropped.
342 */
343 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
1da177e4
LT
344 recalc_sigpending();
345 spin_unlock_irq(&current->sighand->siglock);
346 return 0;
347}
348
349EXPORT_SYMBOL(allow_signal);
350
351int disallow_signal(int sig)
352{
7ed20e1a 353 if (!valid_signal(sig) || sig < 1)
1da177e4
LT
354 return -EINVAL;
355
356 spin_lock_irq(&current->sighand->siglock);
10ab825b 357 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
1da177e4
LT
358 recalc_sigpending();
359 spin_unlock_irq(&current->sighand->siglock);
360 return 0;
361}
362
363EXPORT_SYMBOL(disallow_signal);
364
cf475ad2
BS
365#ifdef CONFIG_MM_OWNER
366/*
733eda7a 367 * A task is exiting. If it owned this mm, find a new owner for the mm.
cf475ad2 368 */
cf475ad2
BS
369void mm_update_next_owner(struct mm_struct *mm)
370{
371 struct task_struct *c, *g, *p = current;
372
373retry:
733eda7a
KH
374 /*
375 * If the exiting or execing task is not the owner, it's
376 * someone else's problem.
377 */
378 if (mm->owner != p)
cf475ad2 379 return;
733eda7a
KH
380 /*
381 * The current owner is exiting/execing and there are no other
382 * candidates. Do not leave the mm pointing to a possibly
383 * freed task structure.
384 */
385 if (atomic_read(&mm->mm_users) <= 1) {
386 mm->owner = NULL;
387 return;
388 }
cf475ad2
BS
389
390 read_lock(&tasklist_lock);
391 /*
392 * Search in the children
393 */
394 list_for_each_entry(c, &p->children, sibling) {
395 if (c->mm == mm)
396 goto assign_new_owner;
397 }
398
399 /*
400 * Search in the siblings
401 */
dea33cfd 402 list_for_each_entry(c, &p->real_parent->children, sibling) {
cf475ad2
BS
403 if (c->mm == mm)
404 goto assign_new_owner;
405 }
406
407 /*
408 * Search through everything else. We should not get
409 * here often
410 */
411 do_each_thread(g, c) {
412 if (c->mm == mm)
413 goto assign_new_owner;
414 } while_each_thread(g, c);
415
416 read_unlock(&tasklist_lock);
31a78f23
BS
417 /*
418 * We found no owner yet mm_users > 1: this implies that we are
419 * most likely racing with swapoff (try_to_unuse()) or /proc or
e5991371 420 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
31a78f23 421 */
31a78f23 422 mm->owner = NULL;
cf475ad2
BS
423 return;
424
425assign_new_owner:
426 BUG_ON(c == p);
427 get_task_struct(c);
428 /*
429 * The task_lock protects c->mm from changing.
430 * We always want mm->owner->mm == mm
431 */
432 task_lock(c);
e5991371
HD
433 /*
434 * Delay read_unlock() till we have the task_lock()
435 * to ensure that c does not slip away underneath us
436 */
437 read_unlock(&tasklist_lock);
cf475ad2
BS
438 if (c->mm != mm) {
439 task_unlock(c);
440 put_task_struct(c);
441 goto retry;
442 }
cf475ad2
BS
443 mm->owner = c;
444 task_unlock(c);
445 put_task_struct(c);
446}
447#endif /* CONFIG_MM_OWNER */
448
1da177e4
LT
449/*
450 * Turn us into a lazy TLB process if we
451 * aren't already..
452 */
408b664a 453static void exit_mm(struct task_struct * tsk)
1da177e4
LT
454{
455 struct mm_struct *mm = tsk->mm;
b564daf8 456 struct core_state *core_state;
1da177e4 457
48d212a2 458 mm_release(tsk, mm);
1da177e4
LT
459 if (!mm)
460 return;
4fe7efdb 461 sync_mm_rss(mm);
1da177e4
LT
462 /*
463 * Serialize with any possible pending coredump.
999d9fc1 464 * We must hold mmap_sem around checking core_state
1da177e4 465 * and clearing tsk->mm. The core-inducing thread
999d9fc1 466 * will increment ->nr_threads for each thread in the
1da177e4
LT
467 * group with ->mm != NULL.
468 */
469 down_read(&mm->mmap_sem);
b564daf8
ON
470 core_state = mm->core_state;
471 if (core_state) {
472 struct core_thread self;
1da177e4 473 up_read(&mm->mmap_sem);
1da177e4 474
b564daf8
ON
475 self.task = tsk;
476 self.next = xchg(&core_state->dumper.next, &self);
477 /*
478 * Implies mb(), the result of xchg() must be visible
479 * to core_state->dumper.
480 */
481 if (atomic_dec_and_test(&core_state->nr_threads))
482 complete(&core_state->startup);
1da177e4 483
a94e2d40
ON
484 for (;;) {
485 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
486 if (!self.task) /* see coredump_finish() */
487 break;
80d26af8 488 freezable_schedule();
a94e2d40
ON
489 }
490 __set_task_state(tsk, TASK_RUNNING);
1da177e4
LT
491 down_read(&mm->mmap_sem);
492 }
493 atomic_inc(&mm->mm_count);
125e1874 494 BUG_ON(mm != tsk->active_mm);
1da177e4
LT
495 /* more a memory barrier than a real lock */
496 task_lock(tsk);
497 tsk->mm = NULL;
498 up_read(&mm->mmap_sem);
499 enter_lazy_tlb(mm, current);
500 task_unlock(tsk);
cf475ad2 501 mm_update_next_owner(mm);
1da177e4
LT
502 mmput(mm);
503}
504
1da177e4 505/*
ebec18a6
LP
506 * When we die, we re-parent all our children, and try to:
507 * 1. give them to another thread in our thread group, if such a member exists
508 * 2. give it to the first ancestor process which prctl'd itself as a
509 * child_subreaper for its children (like a service manager)
510 * 3. give it to the init process (PID 1) in our pid namespace
1da177e4 511 */
950bbabb 512static struct task_struct *find_new_reaper(struct task_struct *father)
d16e15f5
NK
513 __releases(&tasklist_lock)
514 __acquires(&tasklist_lock)
1da177e4 515{
950bbabb
ON
516 struct pid_namespace *pid_ns = task_active_pid_ns(father);
517 struct task_struct *thread;
1da177e4 518
950bbabb
ON
519 thread = father;
520 while_each_thread(father, thread) {
521 if (thread->flags & PF_EXITING)
522 continue;
523 if (unlikely(pid_ns->child_reaper == father))
524 pid_ns->child_reaper = thread;
525 return thread;
526 }
1da177e4 527
950bbabb
ON
528 if (unlikely(pid_ns->child_reaper == father)) {
529 write_unlock_irq(&tasklist_lock);
397a21f2
DV
530 if (unlikely(pid_ns == &init_pid_ns)) {
531 panic("Attempted to kill init! exitcode=0x%08x\n",
532 father->signal->group_exit_code ?:
533 father->exit_code);
534 }
1da177e4 535
950bbabb
ON
536 zap_pid_ns_processes(pid_ns);
537 write_lock_irq(&tasklist_lock);
ebec18a6
LP
538 } else if (father->signal->has_child_subreaper) {
539 struct task_struct *reaper;
540
541 /*
542 * Find the first ancestor marked as child_subreaper.
543 * Note that the code below checks same_thread_group(reaper,
544 * pid_ns->child_reaper). This is what we need to DTRT in a
545 * PID namespace. However we still need the check above, see
546 * http://marc.info/?l=linux-kernel&m=131385460420380
547 */
548 for (reaper = father->real_parent;
549 reaper != &init_task;
550 reaper = reaper->real_parent) {
551 if (same_thread_group(reaper, pid_ns->child_reaper))
552 break;
553 if (!reaper->signal->is_child_subreaper)
554 continue;
555 thread = reaper;
556 do {
557 if (!(thread->flags & PF_EXITING))
558 return reaper;
559 } while_each_thread(reaper, thread);
560 }
1da177e4 561 }
762a24be 562
950bbabb
ON
563 return pid_ns->child_reaper;
564}
565
5dfc80be
ON
566/*
567* Any that need to be release_task'd are put on the @dead list.
568 */
9cd80bbb 569static void reparent_leader(struct task_struct *father, struct task_struct *p,
5dfc80be
ON
570 struct list_head *dead)
571{
5dfc80be
ON
572 list_move_tail(&p->sibling, &p->real_parent->children);
573
0976a03e 574 if (p->exit_state == EXIT_DEAD)
5dfc80be
ON
575 return;
576 /*
577 * If this is a threaded reparent there is no need to
578 * notify anyone anything has happened.
579 */
580 if (same_thread_group(p->real_parent, father))
581 return;
582
583 /* We don't want people slaying init. */
584 p->exit_signal = SIGCHLD;
585
586 /* If it has exited notify the new parent about this child's death. */
d21142ec 587 if (!p->ptrace &&
5dfc80be 588 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
86773473 589 if (do_notify_parent(p, p->exit_signal)) {
5dfc80be
ON
590 p->exit_state = EXIT_DEAD;
591 list_move_tail(&p->sibling, dead);
592 }
593 }
594
595 kill_orphaned_pgrp(p, father);
596}
597
762a24be 598static void forget_original_parent(struct task_struct *father)
1da177e4 599{
950bbabb 600 struct task_struct *p, *n, *reaper;
5dfc80be 601 LIST_HEAD(dead_children);
762a24be
ON
602
603 write_lock_irq(&tasklist_lock);
c7e49c14
ON
604 /*
605 * Note that exit_ptrace() and find_new_reaper() might
606 * drop tasklist_lock and reacquire it.
607 */
608 exit_ptrace(father);
950bbabb 609 reaper = find_new_reaper(father);
f470021a 610
03ff1797 611 list_for_each_entry_safe(p, n, &father->children, sibling) {
9cd80bbb
ON
612 struct task_struct *t = p;
613 do {
614 t->real_parent = reaper;
615 if (t->parent == father) {
d21142ec 616 BUG_ON(t->ptrace);
9cd80bbb
ON
617 t->parent = t->real_parent;
618 }
619 if (t->pdeath_signal)
620 group_send_sig_info(t->pdeath_signal,
621 SEND_SIG_NOINFO, t);
622 } while_each_thread(p, t);
623 reparent_leader(father, p, &dead_children);
1da177e4 624 }
762a24be 625 write_unlock_irq(&tasklist_lock);
5dfc80be 626
762a24be 627 BUG_ON(!list_empty(&father->children));
762a24be 628
5dfc80be
ON
629 list_for_each_entry_safe(p, n, &dead_children, sibling) {
630 list_del_init(&p->sibling);
39c626ae
ON
631 release_task(p);
632 }
1da177e4
LT
633}
634
635/*
636 * Send signals to all our closest relatives so that they know
637 * to properly mourn us..
638 */
821c7de7 639static void exit_notify(struct task_struct *tsk, int group_dead)
1da177e4 640{
53c8f9f1 641 bool autoreap;
1da177e4 642
1da177e4
LT
643 /*
644 * This does two things:
645 *
646 * A. Make init inherit all the child processes
647 * B. Check to see if any process groups have become orphaned
648 * as a result of our exiting, and if they have any stopped
649 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
650 */
762a24be 651 forget_original_parent(tsk);
2e4a7072 652 exit_task_namespaces(tsk);
1da177e4 653
762a24be 654 write_lock_irq(&tasklist_lock);
821c7de7
ON
655 if (group_dead)
656 kill_orphaned_pgrp(tsk->group_leader, NULL);
1da177e4 657
45cdf5cc
ON
658 if (unlikely(tsk->ptrace)) {
659 int sig = thread_group_leader(tsk) &&
660 thread_group_empty(tsk) &&
661 !ptrace_reparented(tsk) ?
662 tsk->exit_signal : SIGCHLD;
663 autoreap = do_notify_parent(tsk, sig);
664 } else if (thread_group_leader(tsk)) {
665 autoreap = thread_group_empty(tsk) &&
666 do_notify_parent(tsk, tsk->exit_signal);
667 } else {
668 autoreap = true;
669 }
1da177e4 670
53c8f9f1 671 tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE;
1da177e4 672
9c339168
ON
673 /* mt-exec, de_thread() is waiting for group leader */
674 if (unlikely(tsk->signal->notify_count < 0))
6db840fa 675 wake_up_process(tsk->signal->group_exit_task);
1da177e4
LT
676 write_unlock_irq(&tasklist_lock);
677
1da177e4 678 /* If the process is dead, release it - nobody will wait for it */
53c8f9f1 679 if (autoreap)
1da177e4 680 release_task(tsk);
1da177e4
LT
681}
682
e18eecb8
JD
683#ifdef CONFIG_DEBUG_STACK_USAGE
684static void check_stack_usage(void)
685{
686 static DEFINE_SPINLOCK(low_water_lock);
687 static int lowest_to_date = THREAD_SIZE;
e18eecb8
JD
688 unsigned long free;
689
7c9f8861 690 free = stack_not_used(current);
e18eecb8
JD
691
692 if (free >= lowest_to_date)
693 return;
694
695 spin_lock(&low_water_lock);
696 if (free < lowest_to_date) {
168eeccb
TB
697 printk(KERN_WARNING "%s (%d) used greatest stack depth: "
698 "%lu bytes left\n",
699 current->comm, task_pid_nr(current), free);
e18eecb8
JD
700 lowest_to_date = free;
701 }
702 spin_unlock(&low_water_lock);
703}
704#else
705static inline void check_stack_usage(void) {}
706#endif
707
9402c95f 708void do_exit(long code)
1da177e4
LT
709{
710 struct task_struct *tsk = current;
711 int group_dead;
712
713 profile_task_exit(tsk);
714
73c10101 715 WARN_ON(blk_needs_flush_plug(tsk));
22e2c507 716
1da177e4
LT
717 if (unlikely(in_interrupt()))
718 panic("Aiee, killing interrupt handler!");
719 if (unlikely(!tsk->pid))
720 panic("Attempted to kill the idle task!");
1da177e4 721
33dd94ae
NE
722 /*
723 * If do_exit is called because this processes oopsed, it's possible
724 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
725 * continuing. Amongst other possible reasons, this is to prevent
726 * mm_release()->clear_child_tid() from writing to a user-controlled
727 * kernel address.
728 */
729 set_fs(USER_DS);
730
a288eecc 731 ptrace_event(PTRACE_EVENT_EXIT, code);
1da177e4 732
e0e81739
DH
733 validate_creds_for_do_exit(tsk);
734
df164db5
AN
735 /*
736 * We're taking recursive faults here in do_exit. Safest is to just
737 * leave this task alone and wait for reboot.
738 */
739 if (unlikely(tsk->flags & PF_EXITING)) {
740 printk(KERN_ALERT
741 "Fixing recursive fault but reboot is needed!\n");
778e9a9c
AK
742 /*
743 * We can do this unlocked here. The futex code uses
744 * this flag just to verify whether the pi state
745 * cleanup has been done or not. In the worst case it
746 * loops once more. We pretend that the cleanup was
747 * done as there is no way to return. Either the
748 * OWNER_DIED bit is set by now or we push the blocked
749 * task into the wait for ever nirwana as well.
750 */
751 tsk->flags |= PF_EXITPIDONE;
df164db5
AN
752 set_current_state(TASK_UNINTERRUPTIBLE);
753 schedule();
754 }
755
d12619b5 756 exit_signals(tsk); /* sets PF_EXITING */
778e9a9c
AK
757 /*
758 * tsk->flags are checked in the futex code to protect against
ed3e694d 759 * an exiting task cleaning up the robust pi futexes.
778e9a9c 760 */
d2ee7198 761 smp_mb();
1d615482 762 raw_spin_unlock_wait(&tsk->pi_lock);
1da177e4 763
1da177e4
LT
764 if (unlikely(in_atomic()))
765 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
ba25f9dc 766 current->comm, task_pid_nr(current),
1da177e4
LT
767 preempt_count());
768
769 acct_update_integrals(tsk);
48d212a2
LT
770 /* sync mm's RSS info before statistics gathering */
771 if (tsk->mm)
772 sync_mm_rss(tsk->mm);
1da177e4 773 group_dead = atomic_dec_and_test(&tsk->signal->live);
c3068951 774 if (group_dead) {
778e9a9c 775 hrtimer_cancel(&tsk->signal->real_timer);
25f407f0 776 exit_itimers(tsk->signal);
1f10206c
JP
777 if (tsk->mm)
778 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
c3068951 779 }
f6ec29a4 780 acct_collect(code, group_dead);
522ed776
MT
781 if (group_dead)
782 tty_audit_exit();
a4ff8dba 783 audit_free(tsk);
115085ea 784
48d212a2 785 tsk->exit_code = code;
115085ea 786 taskstats_exit(tsk, group_dead);
c757249a 787
1da177e4
LT
788 exit_mm(tsk);
789
0e464814 790 if (group_dead)
f6ec29a4 791 acct_process();
0a16b607
MD
792 trace_sched_process_exit(tsk);
793
1da177e4 794 exit_sem(tsk);
b34a6b1d 795 exit_shm(tsk);
1ec7f1dd
AV
796 exit_files(tsk);
797 exit_fs(tsk);
ed3e694d 798 exit_task_work(tsk);
e18eecb8 799 check_stack_usage();
1da177e4 800 exit_thread();
0b3fcf17
SE
801
802 /*
803 * Flush inherited counters to the parent - before the parent
804 * gets woken up by child-exit notifications.
805 *
806 * because of cgroup mode, must be called before cgroup_exit()
807 */
808 perf_event_exit_task(tsk);
809
b4f48b63 810 cgroup_exit(tsk, 1);
1da177e4 811
5ec93d11 812 if (group_dead)
1da177e4
LT
813 disassociate_ctty(1);
814
a1261f54 815 module_put(task_thread_info(tsk)->exec_domain->module);
1da177e4 816
9f46080c 817 proc_exit_connector(tsk);
33b2fb30 818
24f1e32c
FW
819 /*
820 * FIXME: do that only when needed, using sched_exit tracepoint
821 */
bf26c018 822 ptrace_put_breakpoints(tsk);
33b2fb30 823
821c7de7 824 exit_notify(tsk, group_dead);
1da177e4 825#ifdef CONFIG_NUMA
c0ff7453 826 task_lock(tsk);
f0be3d32 827 mpol_put(tsk->mempolicy);
1da177e4 828 tsk->mempolicy = NULL;
c0ff7453 829 task_unlock(tsk);
1da177e4 830#endif
42b2dd0a 831#ifdef CONFIG_FUTEX
c87e2837
IM
832 if (unlikely(current->pi_state_cache))
833 kfree(current->pi_state_cache);
42b2dd0a 834#endif
de5097c2 835 /*
9a11b49a 836 * Make sure we are holding no locks:
de5097c2 837 */
dbf520a9 838 debug_check_no_locks_held(tsk);
778e9a9c
AK
839 /*
840 * We can do this unlocked here. The futex code uses this flag
841 * just to verify whether the pi state cleanup has been done
842 * or not. In the worst case it loops once more.
843 */
844 tsk->flags |= PF_EXITPIDONE;
1da177e4 845
afc847b7 846 if (tsk->io_context)
b69f2292 847 exit_io_context(tsk);
afc847b7 848
b92ce558
JA
849 if (tsk->splice_pipe)
850 __free_pipe_info(tsk->splice_pipe);
851
5640f768
ED
852 if (tsk->task_frag.page)
853 put_page(tsk->task_frag.page);
854
e0e81739
DH
855 validate_creds_for_do_exit(tsk);
856
7407251a 857 preempt_disable();
54848d73
WF
858 if (tsk->nr_dirtied)
859 __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied);
f41d911f 860 exit_rcu();
b5740f4b
YG
861
862 /*
863 * The setting of TASK_RUNNING by try_to_wake_up() may be delayed
864 * when the following two conditions become true.
865 * - There is race condition of mmap_sem (It is acquired by
866 * exit_mm()), and
867 * - SMI occurs before setting TASK_RUNINNG.
868 * (or hypervisor of virtual machine switches to other guest)
869 * As a result, we may become TASK_RUNNING after becoming TASK_DEAD
870 *
871 * To avoid it, we have to wait for releasing tsk->pi_lock which
872 * is held by try_to_wake_up()
873 */
874 smp_mb();
875 raw_spin_unlock_wait(&tsk->pi_lock);
876
55a101f8 877 /* causes final put_task_struct in finish_task_switch(). */
c394cc9f 878 tsk->state = TASK_DEAD;
a585042f 879 tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */
1da177e4
LT
880 schedule();
881 BUG();
882 /* Avoid "noreturn function does return". */
54306cf0
AC
883 for (;;)
884 cpu_relax(); /* For when BUG is null */
1da177e4
LT
885}
886
012914da
RA
887EXPORT_SYMBOL_GPL(do_exit);
888
9402c95f 889void complete_and_exit(struct completion *comp, long code)
1da177e4
LT
890{
891 if (comp)
892 complete(comp);
55a101f8 893
1da177e4
LT
894 do_exit(code);
895}
896
897EXPORT_SYMBOL(complete_and_exit);
898
754fe8d2 899SYSCALL_DEFINE1(exit, int, error_code)
1da177e4
LT
900{
901 do_exit((error_code&0xff)<<8);
902}
903
1da177e4
LT
904/*
905 * Take down every thread in the group. This is called by fatal signals
906 * as well as by sys_exit_group (below).
907 */
9402c95f 908void
1da177e4
LT
909do_group_exit(int exit_code)
910{
bfc4b089
ON
911 struct signal_struct *sig = current->signal;
912
1da177e4
LT
913 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
914
bfc4b089
ON
915 if (signal_group_exit(sig))
916 exit_code = sig->group_exit_code;
1da177e4 917 else if (!thread_group_empty(current)) {
1da177e4 918 struct sighand_struct *const sighand = current->sighand;
1da177e4 919 spin_lock_irq(&sighand->siglock);
ed5d2cac 920 if (signal_group_exit(sig))
1da177e4
LT
921 /* Another thread got here before we took the lock. */
922 exit_code = sig->group_exit_code;
923 else {
1da177e4 924 sig->group_exit_code = exit_code;
ed5d2cac 925 sig->flags = SIGNAL_GROUP_EXIT;
1da177e4
LT
926 zap_other_threads(current);
927 }
928 spin_unlock_irq(&sighand->siglock);
1da177e4
LT
929 }
930
931 do_exit(exit_code);
932 /* NOTREACHED */
933}
934
935/*
936 * this kills every thread in the thread group. Note that any externally
937 * wait4()-ing process will get the correct exit code - even if this
938 * thread is not the thread group leader.
939 */
754fe8d2 940SYSCALL_DEFINE1(exit_group, int, error_code)
1da177e4
LT
941{
942 do_group_exit((error_code & 0xff) << 8);
2ed7c03e
HC
943 /* NOTREACHED */
944 return 0;
1da177e4
LT
945}
946
9e8ae01d
ON
947struct wait_opts {
948 enum pid_type wo_type;
9e8ae01d 949 int wo_flags;
e1eb1ebc 950 struct pid *wo_pid;
9e8ae01d
ON
951
952 struct siginfo __user *wo_info;
953 int __user *wo_stat;
954 struct rusage __user *wo_rusage;
955
0b7570e7 956 wait_queue_t child_wait;
9e8ae01d
ON
957 int notask_error;
958};
959
989264f4
ON
960static inline
961struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
161550d7 962{
989264f4
ON
963 if (type != PIDTYPE_PID)
964 task = task->group_leader;
965 return task->pids[type].pid;
161550d7
EB
966}
967
989264f4 968static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1da177e4 969{
5c01ba49
ON
970 return wo->wo_type == PIDTYPE_MAX ||
971 task_pid_type(p, wo->wo_type) == wo->wo_pid;
972}
1da177e4 973
5c01ba49
ON
974static int eligible_child(struct wait_opts *wo, struct task_struct *p)
975{
976 if (!eligible_pid(wo, p))
977 return 0;
1da177e4
LT
978 /* Wait for all children (clone and not) if __WALL is set;
979 * otherwise, wait for clone children *only* if __WCLONE is
980 * set; otherwise, wait for non-clone children *only*. (Note:
981 * A "clone" child here is one that reports to its parent
982 * using a signal other than SIGCHLD.) */
9e8ae01d
ON
983 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
984 && !(wo->wo_flags & __WALL))
1da177e4 985 return 0;
1da177e4 986
14dd0b81 987 return 1;
1da177e4
LT
988}
989
9e8ae01d
ON
990static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
991 pid_t pid, uid_t uid, int why, int status)
1da177e4 992{
9e8ae01d
ON
993 struct siginfo __user *infop;
994 int retval = wo->wo_rusage
995 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
36c8b586 996
1da177e4 997 put_task_struct(p);
9e8ae01d 998 infop = wo->wo_info;
b6fe2d11
VM
999 if (infop) {
1000 if (!retval)
1001 retval = put_user(SIGCHLD, &infop->si_signo);
1002 if (!retval)
1003 retval = put_user(0, &infop->si_errno);
1004 if (!retval)
1005 retval = put_user((short)why, &infop->si_code);
1006 if (!retval)
1007 retval = put_user(pid, &infop->si_pid);
1008 if (!retval)
1009 retval = put_user(uid, &infop->si_uid);
1010 if (!retval)
1011 retval = put_user(status, &infop->si_status);
1012 }
1da177e4
LT
1013 if (!retval)
1014 retval = pid;
1015 return retval;
1016}
1017
1018/*
1019 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1020 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1021 * the lock and this task is uninteresting. If we return nonzero, we have
1022 * released the lock and the system call should return.
1023 */
9e8ae01d 1024static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1da177e4
LT
1025{
1026 unsigned long state;
2f4e6e2a 1027 int retval, status, traced;
6c5f3e7b 1028 pid_t pid = task_pid_vnr(p);
43e13cc1 1029 uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p));
9e8ae01d 1030 struct siginfo __user *infop;
1da177e4 1031
9e8ae01d 1032 if (!likely(wo->wo_flags & WEXITED))
98abed02
RM
1033 return 0;
1034
9e8ae01d 1035 if (unlikely(wo->wo_flags & WNOWAIT)) {
1da177e4 1036 int exit_code = p->exit_code;
f3abd4f9 1037 int why;
1da177e4 1038
1da177e4
LT
1039 get_task_struct(p);
1040 read_unlock(&tasklist_lock);
1041 if ((exit_code & 0x7f) == 0) {
1042 why = CLD_EXITED;
1043 status = exit_code >> 8;
1044 } else {
1045 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1046 status = exit_code & 0x7f;
1047 }
9e8ae01d 1048 return wait_noreap_copyout(wo, p, pid, uid, why, status);
1da177e4
LT
1049 }
1050
1051 /*
1052 * Try to move the task's state to DEAD
1053 * only one thread is allowed to do this:
1054 */
1055 state = xchg(&p->exit_state, EXIT_DEAD);
1056 if (state != EXIT_ZOMBIE) {
1057 BUG_ON(state != EXIT_DEAD);
1058 return 0;
1059 }
1da177e4 1060
53b6f9fb 1061 traced = ptrace_reparented(p);
befca967
ON
1062 /*
1063 * It can be ptraced but not reparented, check
e550f14d 1064 * thread_group_leader() to filter out sub-threads.
befca967 1065 */
e550f14d 1066 if (likely(!traced) && thread_group_leader(p)) {
3795e161
JJ
1067 struct signal_struct *psig;
1068 struct signal_struct *sig;
1f10206c 1069 unsigned long maxrss;
0cf55e1e 1070 cputime_t tgutime, tgstime;
3795e161 1071
1da177e4
LT
1072 /*
1073 * The resource counters for the group leader are in its
1074 * own task_struct. Those for dead threads in the group
1075 * are in its signal_struct, as are those for the child
1076 * processes it has previously reaped. All these
1077 * accumulate in the parent's signal_struct c* fields.
1078 *
1079 * We don't bother to take a lock here to protect these
1080 * p->signal fields, because they are only touched by
1081 * __exit_signal, which runs with tasklist_lock
1082 * write-locked anyway, and so is excluded here. We do
d1e98f42 1083 * need to protect the access to parent->signal fields,
1da177e4
LT
1084 * as other threads in the parent group can be right
1085 * here reaping other children at the same time.
0cf55e1e 1086 *
e80d0a1a 1087 * We use thread_group_cputime_adjusted() to get times for the thread
0cf55e1e
HS
1088 * group, which consolidates times for all threads in the
1089 * group including the group leader.
1da177e4 1090 */
e80d0a1a 1091 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
d1e98f42
ON
1092 spin_lock_irq(&p->real_parent->sighand->siglock);
1093 psig = p->real_parent->signal;
3795e161 1094 sig = p->signal;
64861634
MS
1095 psig->cutime += tgutime + sig->cutime;
1096 psig->cstime += tgstime + sig->cstime;
6fac4829 1097 psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime;
3795e161
JJ
1098 psig->cmin_flt +=
1099 p->min_flt + sig->min_flt + sig->cmin_flt;
1100 psig->cmaj_flt +=
1101 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1102 psig->cnvcsw +=
1103 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1104 psig->cnivcsw +=
1105 p->nivcsw + sig->nivcsw + sig->cnivcsw;
6eaeeaba
ED
1106 psig->cinblock +=
1107 task_io_get_inblock(p) +
1108 sig->inblock + sig->cinblock;
1109 psig->coublock +=
1110 task_io_get_oublock(p) +
1111 sig->oublock + sig->coublock;
1f10206c
JP
1112 maxrss = max(sig->maxrss, sig->cmaxrss);
1113 if (psig->cmaxrss < maxrss)
1114 psig->cmaxrss = maxrss;
5995477a
AR
1115 task_io_accounting_add(&psig->ioac, &p->ioac);
1116 task_io_accounting_add(&psig->ioac, &sig->ioac);
d1e98f42 1117 spin_unlock_irq(&p->real_parent->sighand->siglock);
1da177e4
LT
1118 }
1119
1120 /*
1121 * Now we are sure this task is interesting, and no other
1122 * thread can reap it because we set its state to EXIT_DEAD.
1123 */
1124 read_unlock(&tasklist_lock);
1125
9e8ae01d
ON
1126 retval = wo->wo_rusage
1127 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1da177e4
LT
1128 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1129 ? p->signal->group_exit_code : p->exit_code;
9e8ae01d
ON
1130 if (!retval && wo->wo_stat)
1131 retval = put_user(status, wo->wo_stat);
1132
1133 infop = wo->wo_info;
1da177e4
LT
1134 if (!retval && infop)
1135 retval = put_user(SIGCHLD, &infop->si_signo);
1136 if (!retval && infop)
1137 retval = put_user(0, &infop->si_errno);
1138 if (!retval && infop) {
1139 int why;
1140
1141 if ((status & 0x7f) == 0) {
1142 why = CLD_EXITED;
1143 status >>= 8;
1144 } else {
1145 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1146 status &= 0x7f;
1147 }
1148 retval = put_user((short)why, &infop->si_code);
1149 if (!retval)
1150 retval = put_user(status, &infop->si_status);
1151 }
1152 if (!retval && infop)
3a515e4a 1153 retval = put_user(pid, &infop->si_pid);
1da177e4 1154 if (!retval && infop)
c69e8d9c 1155 retval = put_user(uid, &infop->si_uid);
2f4e6e2a 1156 if (!retval)
3a515e4a 1157 retval = pid;
2f4e6e2a
ON
1158
1159 if (traced) {
1da177e4 1160 write_lock_irq(&tasklist_lock);
2f4e6e2a
ON
1161 /* We dropped tasklist, ptracer could die and untrace */
1162 ptrace_unlink(p);
1163 /*
86773473
ON
1164 * If this is not a sub-thread, notify the parent.
1165 * If parent wants a zombie, don't release it now.
2f4e6e2a 1166 */
86773473
ON
1167 if (thread_group_leader(p) &&
1168 !do_notify_parent(p, p->exit_signal)) {
1169 p->exit_state = EXIT_ZOMBIE;
1170 p = NULL;
1da177e4
LT
1171 }
1172 write_unlock_irq(&tasklist_lock);
1173 }
1174 if (p != NULL)
1175 release_task(p);
2f4e6e2a 1176
1da177e4
LT
1177 return retval;
1178}
1179
90bc8d8b
ON
1180static int *task_stopped_code(struct task_struct *p, bool ptrace)
1181{
1182 if (ptrace) {
544b2c91
TH
1183 if (task_is_stopped_or_traced(p) &&
1184 !(p->jobctl & JOBCTL_LISTENING))
90bc8d8b
ON
1185 return &p->exit_code;
1186 } else {
1187 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1188 return &p->signal->group_exit_code;
1189 }
1190 return NULL;
1191}
1192
19e27463
TH
1193/**
1194 * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED
1195 * @wo: wait options
1196 * @ptrace: is the wait for ptrace
1197 * @p: task to wait for
1198 *
1199 * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED.
1200 *
1201 * CONTEXT:
1202 * read_lock(&tasklist_lock), which is released if return value is
1203 * non-zero. Also, grabs and releases @p->sighand->siglock.
1204 *
1205 * RETURNS:
1206 * 0 if wait condition didn't exist and search for other wait conditions
1207 * should continue. Non-zero return, -errno on failure and @p's pid on
1208 * success, implies that tasklist_lock is released and wait condition
1209 * search should terminate.
1da177e4 1210 */
9e8ae01d
ON
1211static int wait_task_stopped(struct wait_opts *wo,
1212 int ptrace, struct task_struct *p)
1da177e4 1213{
9e8ae01d 1214 struct siginfo __user *infop;
90bc8d8b 1215 int retval, exit_code, *p_code, why;
ee7c82da 1216 uid_t uid = 0; /* unneeded, required by compiler */
c8950783 1217 pid_t pid;
1da177e4 1218
47918025
ON
1219 /*
1220 * Traditionally we see ptrace'd stopped tasks regardless of options.
1221 */
9e8ae01d 1222 if (!ptrace && !(wo->wo_flags & WUNTRACED))
98abed02
RM
1223 return 0;
1224
19e27463
TH
1225 if (!task_stopped_code(p, ptrace))
1226 return 0;
1227
ee7c82da
ON
1228 exit_code = 0;
1229 spin_lock_irq(&p->sighand->siglock);
1230
90bc8d8b
ON
1231 p_code = task_stopped_code(p, ptrace);
1232 if (unlikely(!p_code))
ee7c82da
ON
1233 goto unlock_sig;
1234
90bc8d8b 1235 exit_code = *p_code;
ee7c82da
ON
1236 if (!exit_code)
1237 goto unlock_sig;
1238
9e8ae01d 1239 if (!unlikely(wo->wo_flags & WNOWAIT))
90bc8d8b 1240 *p_code = 0;
ee7c82da 1241
8ca937a6 1242 uid = from_kuid_munged(current_user_ns(), task_uid(p));
ee7c82da
ON
1243unlock_sig:
1244 spin_unlock_irq(&p->sighand->siglock);
1245 if (!exit_code)
1da177e4
LT
1246 return 0;
1247
1248 /*
1249 * Now we are pretty sure this task is interesting.
1250 * Make sure it doesn't get reaped out from under us while we
1251 * give up the lock and then examine it below. We don't want to
1252 * keep holding onto the tasklist_lock while we call getrusage and
1253 * possibly take page faults for user memory.
1254 */
1255 get_task_struct(p);
6c5f3e7b 1256 pid = task_pid_vnr(p);
f470021a 1257 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1da177e4
LT
1258 read_unlock(&tasklist_lock);
1259
9e8ae01d
ON
1260 if (unlikely(wo->wo_flags & WNOWAIT))
1261 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
1262
1263 retval = wo->wo_rusage
1264 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1265 if (!retval && wo->wo_stat)
1266 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1da177e4 1267
9e8ae01d 1268 infop = wo->wo_info;
1da177e4
LT
1269 if (!retval && infop)
1270 retval = put_user(SIGCHLD, &infop->si_signo);
1271 if (!retval && infop)
1272 retval = put_user(0, &infop->si_errno);
1273 if (!retval && infop)
6efcae46 1274 retval = put_user((short)why, &infop->si_code);
1da177e4
LT
1275 if (!retval && infop)
1276 retval = put_user(exit_code, &infop->si_status);
1277 if (!retval && infop)
c8950783 1278 retval = put_user(pid, &infop->si_pid);
1da177e4 1279 if (!retval && infop)
ee7c82da 1280 retval = put_user(uid, &infop->si_uid);
1da177e4 1281 if (!retval)
c8950783 1282 retval = pid;
1da177e4
LT
1283 put_task_struct(p);
1284
1285 BUG_ON(!retval);
1286 return retval;
1287}
1288
1289/*
1290 * Handle do_wait work for one task in a live, non-stopped state.
1291 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1292 * the lock and this task is uninteresting. If we return nonzero, we have
1293 * released the lock and the system call should return.
1294 */
9e8ae01d 1295static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1da177e4
LT
1296{
1297 int retval;
1298 pid_t pid;
1299 uid_t uid;
1300
9e8ae01d 1301 if (!unlikely(wo->wo_flags & WCONTINUED))
98abed02
RM
1302 return 0;
1303
1da177e4
LT
1304 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1305 return 0;
1306
1307 spin_lock_irq(&p->sighand->siglock);
1308 /* Re-check with the lock held. */
1309 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1310 spin_unlock_irq(&p->sighand->siglock);
1311 return 0;
1312 }
9e8ae01d 1313 if (!unlikely(wo->wo_flags & WNOWAIT))
1da177e4 1314 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
8ca937a6 1315 uid = from_kuid_munged(current_user_ns(), task_uid(p));
1da177e4
LT
1316 spin_unlock_irq(&p->sighand->siglock);
1317
6c5f3e7b 1318 pid = task_pid_vnr(p);
1da177e4
LT
1319 get_task_struct(p);
1320 read_unlock(&tasklist_lock);
1321
9e8ae01d
ON
1322 if (!wo->wo_info) {
1323 retval = wo->wo_rusage
1324 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1da177e4 1325 put_task_struct(p);
9e8ae01d
ON
1326 if (!retval && wo->wo_stat)
1327 retval = put_user(0xffff, wo->wo_stat);
1da177e4 1328 if (!retval)
3a515e4a 1329 retval = pid;
1da177e4 1330 } else {
9e8ae01d
ON
1331 retval = wait_noreap_copyout(wo, p, pid, uid,
1332 CLD_CONTINUED, SIGCONT);
1da177e4
LT
1333 BUG_ON(retval == 0);
1334 }
1335
1336 return retval;
1337}
1338
98abed02
RM
1339/*
1340 * Consider @p for a wait by @parent.
1341 *
9e8ae01d 1342 * -ECHILD should be in ->notask_error before the first call.
98abed02
RM
1343 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1344 * Returns zero if the search for a child should continue;
9e8ae01d 1345 * then ->notask_error is 0 if @p is an eligible child,
14dd0b81 1346 * or another error from security_task_wait(), or still -ECHILD.
98abed02 1347 */
b6e763f0
ON
1348static int wait_consider_task(struct wait_opts *wo, int ptrace,
1349 struct task_struct *p)
98abed02 1350{
9e8ae01d 1351 int ret = eligible_child(wo, p);
14dd0b81 1352 if (!ret)
98abed02
RM
1353 return ret;
1354
a2322e1d 1355 ret = security_task_wait(p);
14dd0b81
RM
1356 if (unlikely(ret < 0)) {
1357 /*
1358 * If we have not yet seen any eligible child,
1359 * then let this error code replace -ECHILD.
1360 * A permission error will give the user a clue
1361 * to look for security policy problems, rather
1362 * than for mysterious wait bugs.
1363 */
9e8ae01d
ON
1364 if (wo->notask_error)
1365 wo->notask_error = ret;
78a3d9d5 1366 return 0;
14dd0b81
RM
1367 }
1368
823b018e 1369 /* dead body doesn't have much to contribute */
50b8d257
ON
1370 if (unlikely(p->exit_state == EXIT_DEAD)) {
1371 /*
1372 * But do not ignore this task until the tracer does
1373 * wait_task_zombie()->do_notify_parent().
1374 */
1375 if (likely(!ptrace) && unlikely(ptrace_reparented(p)))
1376 wo->notask_error = 0;
823b018e 1377 return 0;
50b8d257 1378 }
823b018e 1379
45cb24a1
TH
1380 /* slay zombie? */
1381 if (p->exit_state == EXIT_ZOMBIE) {
f470021a 1382 /*
45cb24a1
TH
1383 * A zombie ptracee is only visible to its ptracer.
1384 * Notification and reaping will be cascaded to the real
1385 * parent when the ptracer detaches.
f470021a 1386 */
d21142ec 1387 if (likely(!ptrace) && unlikely(p->ptrace)) {
45cb24a1
TH
1388 /* it will become visible, clear notask_error */
1389 wo->notask_error = 0;
1390 return 0;
1391 }
f470021a 1392
9b84cca2
TH
1393 /* we don't reap group leaders with subthreads */
1394 if (!delay_group_leader(p))
1395 return wait_task_zombie(wo, p);
98abed02 1396
f470021a 1397 /*
9b84cca2
TH
1398 * Allow access to stopped/continued state via zombie by
1399 * falling through. Clearing of notask_error is complex.
1400 *
1401 * When !@ptrace:
1402 *
1403 * If WEXITED is set, notask_error should naturally be
1404 * cleared. If not, subset of WSTOPPED|WCONTINUED is set,
1405 * so, if there are live subthreads, there are events to
1406 * wait for. If all subthreads are dead, it's still safe
1407 * to clear - this function will be called again in finite
1408 * amount time once all the subthreads are released and
1409 * will then return without clearing.
1410 *
1411 * When @ptrace:
1412 *
1413 * Stopped state is per-task and thus can't change once the
1414 * target task dies. Only continued and exited can happen.
1415 * Clear notask_error if WCONTINUED | WEXITED.
1416 */
1417 if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
1418 wo->notask_error = 0;
1419 } else {
45cb24a1
TH
1420 /*
1421 * If @p is ptraced by a task in its real parent's group,
1422 * hide group stop/continued state when looking at @p as
1423 * the real parent; otherwise, a single stop can be
1424 * reported twice as group and ptrace stops.
1425 *
1426 * If a ptracer wants to distinguish the two events for its
1427 * own children, it should create a separate process which
1428 * takes the role of real parent.
1429 */
479bf98c 1430 if (likely(!ptrace) && p->ptrace && !ptrace_reparented(p))
45cb24a1
TH
1431 return 0;
1432
9b84cca2
TH
1433 /*
1434 * @p is alive and it's gonna stop, continue or exit, so
1435 * there always is something to wait for.
f470021a 1436 */
9e8ae01d 1437 wo->notask_error = 0;
f470021a
RM
1438 }
1439
98abed02 1440 /*
45cb24a1
TH
1441 * Wait for stopped. Depending on @ptrace, different stopped state
1442 * is used and the two don't interact with each other.
98abed02 1443 */
19e27463
TH
1444 ret = wait_task_stopped(wo, ptrace, p);
1445 if (ret)
1446 return ret;
98abed02
RM
1447
1448 /*
45cb24a1
TH
1449 * Wait for continued. There's only one continued state and the
1450 * ptracer can consume it which can confuse the real parent. Don't
1451 * use WCONTINUED from ptracer. You don't need or want it.
98abed02 1452 */
9e8ae01d 1453 return wait_task_continued(wo, p);
98abed02
RM
1454}
1455
1456/*
1457 * Do the work of do_wait() for one thread in the group, @tsk.
1458 *
9e8ae01d 1459 * -ECHILD should be in ->notask_error before the first call.
98abed02
RM
1460 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1461 * Returns zero if the search for a child should continue; then
9e8ae01d 1462 * ->notask_error is 0 if there were any eligible children,
14dd0b81 1463 * or another error from security_task_wait(), or still -ECHILD.
98abed02 1464 */
9e8ae01d 1465static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
98abed02
RM
1466{
1467 struct task_struct *p;
1468
1469 list_for_each_entry(p, &tsk->children, sibling) {
9cd80bbb
ON
1470 int ret = wait_consider_task(wo, 0, p);
1471 if (ret)
1472 return ret;
98abed02
RM
1473 }
1474
1475 return 0;
1476}
1477
9e8ae01d 1478static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
98abed02
RM
1479{
1480 struct task_struct *p;
1481
f470021a 1482 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
b6e763f0 1483 int ret = wait_consider_task(wo, 1, p);
f470021a 1484 if (ret)
98abed02 1485 return ret;
98abed02
RM
1486 }
1487
1488 return 0;
1489}
1490
0b7570e7
ON
1491static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1492 int sync, void *key)
1493{
1494 struct wait_opts *wo = container_of(wait, struct wait_opts,
1495 child_wait);
1496 struct task_struct *p = key;
1497
5c01ba49 1498 if (!eligible_pid(wo, p))
0b7570e7
ON
1499 return 0;
1500
b4fe5182
ON
1501 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1502 return 0;
1503
0b7570e7
ON
1504 return default_wake_function(wait, mode, sync, key);
1505}
1506
a7f0765e
ON
1507void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1508{
0b7570e7
ON
1509 __wake_up_sync_key(&parent->signal->wait_chldexit,
1510 TASK_INTERRUPTIBLE, 1, p);
a7f0765e
ON
1511}
1512
9e8ae01d 1513static long do_wait(struct wait_opts *wo)
1da177e4 1514{
1da177e4 1515 struct task_struct *tsk;
98abed02 1516 int retval;
1da177e4 1517
9e8ae01d 1518 trace_sched_process_wait(wo->wo_pid);
0a16b607 1519
0b7570e7
ON
1520 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1521 wo->child_wait.private = current;
1522 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1da177e4 1523repeat:
98abed02
RM
1524 /*
1525 * If there is nothing that can match our critiera just get out.
9e8ae01d
ON
1526 * We will clear ->notask_error to zero if we see any child that
1527 * might later match our criteria, even if we are not able to reap
1528 * it yet.
98abed02 1529 */
64a16caf 1530 wo->notask_error = -ECHILD;
9e8ae01d
ON
1531 if ((wo->wo_type < PIDTYPE_MAX) &&
1532 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
64a16caf 1533 goto notask;
161550d7 1534
f95d39d1 1535 set_current_state(TASK_INTERRUPTIBLE);
1da177e4
LT
1536 read_lock(&tasklist_lock);
1537 tsk = current;
1538 do {
64a16caf
ON
1539 retval = do_wait_thread(wo, tsk);
1540 if (retval)
1541 goto end;
9e8ae01d 1542
64a16caf
ON
1543 retval = ptrace_do_wait(wo, tsk);
1544 if (retval)
98abed02 1545 goto end;
98abed02 1546
9e8ae01d 1547 if (wo->wo_flags & __WNOTHREAD)
1da177e4 1548 break;
a3f6dfb7 1549 } while_each_thread(current, tsk);
1da177e4 1550 read_unlock(&tasklist_lock);
f2cc3eb1 1551
64a16caf 1552notask:
9e8ae01d
ON
1553 retval = wo->notask_error;
1554 if (!retval && !(wo->wo_flags & WNOHANG)) {
1da177e4 1555 retval = -ERESTARTSYS;
98abed02
RM
1556 if (!signal_pending(current)) {
1557 schedule();
1558 goto repeat;
1559 }
1da177e4 1560 }
1da177e4 1561end:
f95d39d1 1562 __set_current_state(TASK_RUNNING);
0b7570e7 1563 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1da177e4
LT
1564 return retval;
1565}
1566
17da2bd9
HC
1567SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1568 infop, int, options, struct rusage __user *, ru)
1da177e4 1569{
9e8ae01d 1570 struct wait_opts wo;
161550d7
EB
1571 struct pid *pid = NULL;
1572 enum pid_type type;
1da177e4
LT
1573 long ret;
1574
1575 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1576 return -EINVAL;
1577 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1578 return -EINVAL;
1579
1580 switch (which) {
1581 case P_ALL:
161550d7 1582 type = PIDTYPE_MAX;
1da177e4
LT
1583 break;
1584 case P_PID:
161550d7
EB
1585 type = PIDTYPE_PID;
1586 if (upid <= 0)
1da177e4
LT
1587 return -EINVAL;
1588 break;
1589 case P_PGID:
161550d7
EB
1590 type = PIDTYPE_PGID;
1591 if (upid <= 0)
1da177e4 1592 return -EINVAL;
1da177e4
LT
1593 break;
1594 default:
1595 return -EINVAL;
1596 }
1597
161550d7
EB
1598 if (type < PIDTYPE_MAX)
1599 pid = find_get_pid(upid);
9e8ae01d
ON
1600
1601 wo.wo_type = type;
1602 wo.wo_pid = pid;
1603 wo.wo_flags = options;
1604 wo.wo_info = infop;
1605 wo.wo_stat = NULL;
1606 wo.wo_rusage = ru;
1607 ret = do_wait(&wo);
dfe16dfa
VM
1608
1609 if (ret > 0) {
1610 ret = 0;
1611 } else if (infop) {
1612 /*
1613 * For a WNOHANG return, clear out all the fields
1614 * we would set so the user can easily tell the
1615 * difference.
1616 */
1617 if (!ret)
1618 ret = put_user(0, &infop->si_signo);
1619 if (!ret)
1620 ret = put_user(0, &infop->si_errno);
1621 if (!ret)
1622 ret = put_user(0, &infop->si_code);
1623 if (!ret)
1624 ret = put_user(0, &infop->si_pid);
1625 if (!ret)
1626 ret = put_user(0, &infop->si_uid);
1627 if (!ret)
1628 ret = put_user(0, &infop->si_status);
1629 }
1630
161550d7 1631 put_pid(pid);
1da177e4
LT
1632
1633 /* avoid REGPARM breakage on x86: */
54a01510 1634 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1da177e4
LT
1635 return ret;
1636}
1637
754fe8d2
HC
1638SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1639 int, options, struct rusage __user *, ru)
1da177e4 1640{
9e8ae01d 1641 struct wait_opts wo;
161550d7
EB
1642 struct pid *pid = NULL;
1643 enum pid_type type;
1da177e4
LT
1644 long ret;
1645
1646 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1647 __WNOTHREAD|__WCLONE|__WALL))
1648 return -EINVAL;
161550d7
EB
1649
1650 if (upid == -1)
1651 type = PIDTYPE_MAX;
1652 else if (upid < 0) {
1653 type = PIDTYPE_PGID;
1654 pid = find_get_pid(-upid);
1655 } else if (upid == 0) {
1656 type = PIDTYPE_PGID;
2ae448ef 1657 pid = get_task_pid(current, PIDTYPE_PGID);
161550d7
EB
1658 } else /* upid > 0 */ {
1659 type = PIDTYPE_PID;
1660 pid = find_get_pid(upid);
1661 }
1662
9e8ae01d
ON
1663 wo.wo_type = type;
1664 wo.wo_pid = pid;
1665 wo.wo_flags = options | WEXITED;
1666 wo.wo_info = NULL;
1667 wo.wo_stat = stat_addr;
1668 wo.wo_rusage = ru;
1669 ret = do_wait(&wo);
161550d7 1670 put_pid(pid);
1da177e4
LT
1671
1672 /* avoid REGPARM breakage on x86: */
54a01510 1673 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1da177e4
LT
1674 return ret;
1675}
1676
1677#ifdef __ARCH_WANT_SYS_WAITPID
1678
1679/*
1680 * sys_waitpid() remains for compatibility. waitpid() should be
1681 * implemented by calling sys_wait4() from libc.a.
1682 */
17da2bd9 1683SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1da177e4
LT
1684{
1685 return sys_wait4(pid, stat_addr, options, NULL);
1686}
1687
1688#endif