Fix common misspellings
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
1da177e4 23#include <linux/binfmts.h>
ab516013 24#include <linux/nsproxy.h>
84d73786 25#include <linux/pid_namespace.h>
1da177e4
LT
26#include <linux/ptrace.h>
27#include <linux/profile.h>
28#include <linux/mount.h>
29#include <linux/proc_fs.h>
49d769d5 30#include <linux/kthread.h>
1da177e4 31#include <linux/mempolicy.h>
c757249a 32#include <linux/taskstats_kern.h>
ca74e92b 33#include <linux/delayacct.h>
83144186 34#include <linux/freezer.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>
1da177e4
LT
54
55#include <asm/uaccess.h>
56#include <asm/unistd.h>
57#include <asm/pgtable.h>
58#include <asm/mmu_context.h>
59
408b664a
AB
60static void exit_mm(struct task_struct * tsk);
61
d40e48e0 62static void __unhash_process(struct task_struct *p, bool group_dead)
1da177e4
LT
63{
64 nr_threads--;
65 detach_pid(p, PIDTYPE_PID);
d40e48e0 66 if (group_dead) {
1da177e4
LT
67 detach_pid(p, PIDTYPE_PGID);
68 detach_pid(p, PIDTYPE_SID);
c97d9893 69
5e85d4ab 70 list_del_rcu(&p->tasks);
9cd80bbb 71 list_del_init(&p->sibling);
909ea964 72 __this_cpu_dec(process_counts);
1da177e4 73 }
47e65328 74 list_del_rcu(&p->thread_group);
1da177e4
LT
75}
76
6a14c5c9
ON
77/*
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;
d40e48e0 83 bool group_dead = thread_group_leader(tsk);
6a14c5c9 84 struct sighand_struct *sighand;
4ada856f 85 struct tty_struct *uninitialized_var(tty);
6a14c5c9 86
d11c563d
PM
87 sighand = rcu_dereference_check(tsk->sighand,
88 rcu_read_lock_held() ||
db1466b3 89 lockdep_tasklist_lock_is_held());
6a14c5c9
ON
90 spin_lock(&sighand->siglock);
91
92 posix_cpu_timers_exit(tsk);
d40e48e0 93 if (group_dead) {
6a14c5c9 94 posix_cpu_timers_exit_group(tsk);
4ada856f
ON
95 tty = sig->tty;
96 sig->tty = NULL;
4a599942 97 } else {
e0a70217
ON
98 /*
99 * This can only happen if the caller is de_thread().
100 * FIXME: this is the temporary hack, we should teach
101 * posix-cpu-timers to handle this case correctly.
102 */
103 if (unlikely(has_group_leader_pid(tsk)))
104 posix_cpu_timers_exit_group(tsk);
105
6a14c5c9
ON
106 /*
107 * If there is any task waiting for the group exit
108 * then notify it:
109 */
d344193a 110 if (sig->notify_count > 0 && !--sig->notify_count)
6a14c5c9 111 wake_up_process(sig->group_exit_task);
6db840fa 112
6a14c5c9
ON
113 if (tsk == sig->curr_target)
114 sig->curr_target = next_thread(tsk);
115 /*
116 * Accumulate here the counters for all threads but the
117 * group leader as they die, so they can be added into
118 * the process-wide totals when those are taken.
119 * The group leader stays around as a zombie as long
120 * as there are other threads. When it gets reaped,
121 * the exit.c code will add its counts into these totals.
122 * We won't ever get here for the group leader, since it
123 * will have been the last reference on the signal_struct.
124 */
0cf55e1e
HS
125 sig->utime = cputime_add(sig->utime, tsk->utime);
126 sig->stime = cputime_add(sig->stime, tsk->stime);
d5b7c78e 127 sig->gtime = cputime_add(sig->gtime, tsk->gtime);
6a14c5c9
ON
128 sig->min_flt += tsk->min_flt;
129 sig->maj_flt += tsk->maj_flt;
130 sig->nvcsw += tsk->nvcsw;
131 sig->nivcsw += tsk->nivcsw;
6eaeeaba
ED
132 sig->inblock += task_io_get_inblock(tsk);
133 sig->oublock += task_io_get_oublock(tsk);
5995477a 134 task_io_accounting_add(&sig->ioac, &tsk->ioac);
32bd671d 135 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
6a14c5c9
ON
136 }
137
b3ac022c 138 sig->nr_threads--;
d40e48e0 139 __unhash_process(tsk, group_dead);
5876700c 140
da7978b0
ON
141 /*
142 * Do this under ->siglock, we can race with another thread
143 * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
144 */
145 flush_sigqueue(&tsk->pending);
a7e5328a 146 tsk->sighand = NULL;
6a14c5c9 147 spin_unlock(&sighand->siglock);
6a14c5c9 148
a7e5328a 149 __cleanup_sighand(sighand);
6a14c5c9 150 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
d40e48e0 151 if (group_dead) {
6a14c5c9 152 flush_sigqueue(&sig->shared_pending);
4ada856f 153 tty_kref_put(tty);
6a14c5c9
ON
154 }
155}
156
8c7904a0
EB
157static void delayed_put_task_struct(struct rcu_head *rhp)
158{
0a16b607
MD
159 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
160
4e231c79 161 perf_event_delayed_put(tsk);
0a16b607
MD
162 trace_sched_process_free(tsk);
163 put_task_struct(tsk);
8c7904a0
EB
164}
165
f470021a 166
1da177e4
LT
167void release_task(struct task_struct * p)
168{
36c8b586 169 struct task_struct *leader;
1da177e4 170 int zap_leader;
1f09f974 171repeat:
dae33574 172 tracehook_prepare_release_task(p);
c69e8d9c 173 /* don't need to get the RCU readlock here - the process is dead and
d11c563d
PM
174 * can't be modifying its own credentials. But shut RCU-lockdep up */
175 rcu_read_lock();
c69e8d9c 176 atomic_dec(&__task_cred(p)->user->processes);
d11c563d 177 rcu_read_unlock();
c69e8d9c 178
60347f67 179 proc_flush_task(p);
0203026b 180
1da177e4 181 write_lock_irq(&tasklist_lock);
dae33574 182 tracehook_finish_release_task(p);
1da177e4 183 __exit_signal(p);
35f5cad8 184
1da177e4
LT
185 /*
186 * If we are the last non-leader member of the thread
187 * group, and the leader is zombie, then notify the
188 * group leader's parent process. (if it wants notification.)
189 */
190 zap_leader = 0;
191 leader = p->group_leader;
192 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
d839fd4d 193 BUG_ON(task_detached(leader));
1da177e4
LT
194 do_notify_parent(leader, leader->exit_signal);
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.
199 *
200 * do_notify_parent() will have marked it self-reaping in
201 * that case.
202 */
d839fd4d 203 zap_leader = task_detached(leader);
dae33574
RM
204
205 /*
206 * This maintains the invariant that release_task()
207 * only runs on a task in EXIT_DEAD, just for sanity.
208 */
209 if (zap_leader)
210 leader->exit_state = EXIT_DEAD;
1da177e4
LT
211 }
212
1da177e4 213 write_unlock_irq(&tasklist_lock);
1da177e4 214 release_thread(p);
8c7904a0 215 call_rcu(&p->rcu, delayed_put_task_struct);
1da177e4
LT
216
217 p = leader;
218 if (unlikely(zap_leader))
219 goto repeat;
220}
221
1da177e4
LT
222/*
223 * This checks not only the pgrp, but falls back on the pid if no
224 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
225 * without this...
04a2e6a5
EB
226 *
227 * The caller must hold rcu lock or the tasklist lock.
1da177e4 228 */
04a2e6a5 229struct pid *session_of_pgrp(struct pid *pgrp)
1da177e4
LT
230{
231 struct task_struct *p;
04a2e6a5 232 struct pid *sid = NULL;
62dfb554 233
04a2e6a5 234 p = pid_task(pgrp, PIDTYPE_PGID);
62dfb554 235 if (p == NULL)
04a2e6a5 236 p = pid_task(pgrp, PIDTYPE_PID);
62dfb554 237 if (p != NULL)
04a2e6a5 238 sid = task_session(p);
62dfb554 239
1da177e4
LT
240 return sid;
241}
242
243/*
244 * Determine if a process group is "orphaned", according to the POSIX
245 * definition in 2.2.2.52. Orphaned process groups are not to be affected
246 * by terminal-generated stop signals. Newly orphaned process groups are
247 * to receive a SIGHUP and a SIGCONT.
248 *
249 * "I ask you, have you ever known what it is to be an orphan?"
250 */
0475ac08 251static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
1da177e4
LT
252{
253 struct task_struct *p;
1da177e4 254
0475ac08 255 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
05e83df6
ON
256 if ((p == ignored_task) ||
257 (p->exit_state && thread_group_empty(p)) ||
258 is_global_init(p->real_parent))
1da177e4 259 continue;
05e83df6 260
0475ac08 261 if (task_pgrp(p->real_parent) != pgrp &&
05e83df6
ON
262 task_session(p->real_parent) == task_session(p))
263 return 0;
0475ac08 264 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
05e83df6
ON
265
266 return 1;
1da177e4
LT
267}
268
3e7cd6c4 269int is_current_pgrp_orphaned(void)
1da177e4
LT
270{
271 int retval;
272
273 read_lock(&tasklist_lock);
3e7cd6c4 274 retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
1da177e4
LT
275 read_unlock(&tasklist_lock);
276
277 return retval;
278}
279
0475ac08 280static int has_stopped_jobs(struct pid *pgrp)
1da177e4
LT
281{
282 int retval = 0;
283 struct task_struct *p;
284
0475ac08 285 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
338077e5 286 if (!task_is_stopped(p))
1da177e4 287 continue;
1da177e4
LT
288 retval = 1;
289 break;
0475ac08 290 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1da177e4
LT
291 return retval;
292}
293
f49ee505
ON
294/*
295 * Check to see if any process groups have become orphaned as
296 * a result of our exiting, and if they have any stopped jobs,
297 * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
298 */
299static void
300kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
301{
302 struct pid *pgrp = task_pgrp(tsk);
303 struct task_struct *ignored_task = tsk;
304
305 if (!parent)
306 /* exit: our father is in a different pgrp than
307 * we are and we were the only connection outside.
308 */
309 parent = tsk->real_parent;
310 else
311 /* reparent: our child is in a different pgrp than
312 * we are, and it was the only connection outside.
313 */
314 ignored_task = NULL;
315
316 if (task_pgrp(parent) != pgrp &&
317 task_session(parent) == task_session(tsk) &&
318 will_become_orphaned_pgrp(pgrp, ignored_task) &&
319 has_stopped_jobs(pgrp)) {
320 __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
321 __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
322 }
323}
324
1da177e4 325/**
49d769d5 326 * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
1da177e4
LT
327 *
328 * If a kernel thread is launched as a result of a system call, or if
49d769d5
EB
329 * it ever exits, it should generally reparent itself to kthreadd so it
330 * isn't in the way of other processes and is correctly cleaned up on exit.
1da177e4
LT
331 *
332 * The various task state such as scheduling policy and priority may have
333 * been inherited from a user process, so we reset them to sane values here.
334 *
49d769d5 335 * NOTE that reparent_to_kthreadd() gives the caller full capabilities.
1da177e4 336 */
49d769d5 337static void reparent_to_kthreadd(void)
1da177e4
LT
338{
339 write_lock_irq(&tasklist_lock);
340
341 ptrace_unlink(current);
342 /* Reparent to init */
49d769d5 343 current->real_parent = current->parent = kthreadd_task;
f470021a 344 list_move_tail(&current->sibling, &current->real_parent->children);
1da177e4
LT
345
346 /* Set the exit signal to SIGCHLD so we signal init on exit */
347 current->exit_signal = SIGCHLD;
348
e05606d3 349 if (task_nice(current) < 0)
1da177e4
LT
350 set_user_nice(current, 0);
351 /* cpus_allowed? */
352 /* rt_priority? */
353 /* signals? */
1da177e4
LT
354 memcpy(current->signal->rlim, init_task.signal->rlim,
355 sizeof(current->signal->rlim));
d84f4f99
DH
356
357 atomic_inc(&init_cred.usage);
358 commit_creds(&init_cred);
1da177e4 359 write_unlock_irq(&tasklist_lock);
1da177e4
LT
360}
361
8520d7c7 362void __set_special_pids(struct pid *pid)
1da177e4 363{
e19f247a 364 struct task_struct *curr = current->group_leader;
1da177e4 365
0d0df599 366 if (task_session(curr) != pid)
7d8da096 367 change_pid(curr, PIDTYPE_SID, pid);
1b0f7ffd
ON
368
369 if (task_pgrp(curr) != pid)
7d8da096 370 change_pid(curr, PIDTYPE_PGID, pid);
1da177e4
LT
371}
372
8520d7c7 373static void set_special_pids(struct pid *pid)
1da177e4
LT
374{
375 write_lock_irq(&tasklist_lock);
8520d7c7 376 __set_special_pids(pid);
1da177e4
LT
377 write_unlock_irq(&tasklist_lock);
378}
379
380/*
87245135
ON
381 * Let kernel threads use this to say that they allow a certain signal.
382 * Must not be used if kthread was cloned with CLONE_SIGHAND.
1da177e4
LT
383 */
384int allow_signal(int sig)
385{
7ed20e1a 386 if (!valid_signal(sig) || sig < 1)
1da177e4
LT
387 return -EINVAL;
388
389 spin_lock_irq(&current->sighand->siglock);
87245135 390 /* This is only needed for daemonize()'ed kthreads */
1da177e4 391 sigdelset(&current->blocked, sig);
87245135
ON
392 /*
393 * Kernel threads handle their own signals. Let the signal code
394 * know it'll be handled, so that they don't get converted to
395 * SIGKILL or just silently dropped.
396 */
397 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
1da177e4
LT
398 recalc_sigpending();
399 spin_unlock_irq(&current->sighand->siglock);
400 return 0;
401}
402
403EXPORT_SYMBOL(allow_signal);
404
405int disallow_signal(int sig)
406{
7ed20e1a 407 if (!valid_signal(sig) || sig < 1)
1da177e4
LT
408 return -EINVAL;
409
410 spin_lock_irq(&current->sighand->siglock);
10ab825b 411 current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
1da177e4
LT
412 recalc_sigpending();
413 spin_unlock_irq(&current->sighand->siglock);
414 return 0;
415}
416
417EXPORT_SYMBOL(disallow_signal);
418
419/*
420 * Put all the gunge required to become a kernel thread without
421 * attached user resources in one place where it belongs.
422 */
423
424void daemonize(const char *name, ...)
425{
426 va_list args;
1da177e4
LT
427 sigset_t blocked;
428
429 va_start(args, name);
430 vsnprintf(current->comm, sizeof(current->comm), name, args);
431 va_end(args);
432
433 /*
434 * If we were started as result of loading a module, close all of the
435 * user space pages. We don't need them, and if we didn't close them
436 * they would be locked into memory.
437 */
438 exit_mm(current);
83144186
RW
439 /*
440 * We don't want to have TIF_FREEZE set if the system-wide hibernation
441 * or suspend transition begins right now.
442 */
7b34e428 443 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
1da177e4 444
8520d7c7
ON
445 if (current->nsproxy != &init_nsproxy) {
446 get_nsproxy(&init_nsproxy);
447 switch_task_namespaces(current, &init_nsproxy);
448 }
297bd42b 449 set_special_pids(&init_struct_pid);
24ec839c 450 proc_clear_tty(current);
1da177e4
LT
451
452 /* Block and flush all signals */
453 sigfillset(&blocked);
454 sigprocmask(SIG_BLOCK, &blocked, NULL);
455 flush_signals(current);
456
457 /* Become as one with the init task */
458
3e93cd67 459 daemonize_fs_struct();
d4c5e41f 460 exit_files(current);
1da177e4
LT
461 current->files = init_task.files;
462 atomic_inc(&current->files->count);
463
49d769d5 464 reparent_to_kthreadd();
1da177e4
LT
465}
466
467EXPORT_SYMBOL(daemonize);
468
858119e1 469static void close_files(struct files_struct * files)
1da177e4
LT
470{
471 int i, j;
badf1662 472 struct fdtable *fdt;
1da177e4
LT
473
474 j = 0;
4fb3a538
DS
475
476 /*
477 * It is safe to dereference the fd table without RCU or
478 * ->file_lock because this is the last reference to the
d11c563d 479 * files structure. But use RCU to shut RCU-lockdep up.
4fb3a538 480 */
d11c563d 481 rcu_read_lock();
badf1662 482 fdt = files_fdtable(files);
d11c563d 483 rcu_read_unlock();
1da177e4
LT
484 for (;;) {
485 unsigned long set;
486 i = j * __NFDBITS;
bbea9f69 487 if (i >= fdt->max_fds)
1da177e4 488 break;
badf1662 489 set = fdt->open_fds->fds_bits[j++];
1da177e4
LT
490 while (set) {
491 if (set & 1) {
badf1662 492 struct file * file = xchg(&fdt->fd[i], NULL);
944be0b2 493 if (file) {
1da177e4 494 filp_close(file, files);
944be0b2
IM
495 cond_resched();
496 }
1da177e4
LT
497 }
498 i++;
499 set >>= 1;
500 }
501 }
502}
503
504struct files_struct *get_files_struct(struct task_struct *task)
505{
506 struct files_struct *files;
507
508 task_lock(task);
509 files = task->files;
510 if (files)
511 atomic_inc(&files->count);
512 task_unlock(task);
513
514 return files;
515}
516
7ad5b3a5 517void put_files_struct(struct files_struct *files)
1da177e4 518{
badf1662
DS
519 struct fdtable *fdt;
520
1da177e4
LT
521 if (atomic_dec_and_test(&files->count)) {
522 close_files(files);
523 /*
524 * Free the fd and fdset arrays if we expanded them.
ab2af1f5
DS
525 * If the fdtable was embedded, pass files for freeing
526 * at the end of the RCU grace period. Otherwise,
527 * you can free files immediately.
1da177e4 528 */
d11c563d 529 rcu_read_lock();
badf1662 530 fdt = files_fdtable(files);
4fd45812 531 if (fdt != &files->fdtab)
ab2af1f5 532 kmem_cache_free(files_cachep, files);
01b2d93c 533 free_fdtable(fdt);
d11c563d 534 rcu_read_unlock();
1da177e4
LT
535 }
536}
537
3b125388 538void reset_files_struct(struct files_struct *files)
3b9b8ab6 539{
3b125388 540 struct task_struct *tsk = current;
3b9b8ab6
KK
541 struct files_struct *old;
542
543 old = tsk->files;
544 task_lock(tsk);
545 tsk->files = files;
546 task_unlock(tsk);
547 put_files_struct(old);
548}
3b9b8ab6 549
1ec7f1dd 550void exit_files(struct task_struct *tsk)
1da177e4
LT
551{
552 struct files_struct * files = tsk->files;
553
554 if (files) {
555 task_lock(tsk);
556 tsk->files = NULL;
557 task_unlock(tsk);
558 put_files_struct(files);
559 }
560}
561
cf475ad2
BS
562#ifdef CONFIG_MM_OWNER
563/*
564 * Task p is exiting and it owned mm, lets find a new owner for it
565 */
566static inline int
567mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
568{
569 /*
570 * If there are other users of the mm and the owner (us) is exiting
571 * we need to find a new owner to take on the responsibility.
572 */
cf475ad2
BS
573 if (atomic_read(&mm->mm_users) <= 1)
574 return 0;
575 if (mm->owner != p)
576 return 0;
577 return 1;
578}
579
580void mm_update_next_owner(struct mm_struct *mm)
581{
582 struct task_struct *c, *g, *p = current;
583
584retry:
585 if (!mm_need_new_owner(mm, p))
586 return;
587
588 read_lock(&tasklist_lock);
589 /*
590 * Search in the children
591 */
592 list_for_each_entry(c, &p->children, sibling) {
593 if (c->mm == mm)
594 goto assign_new_owner;
595 }
596
597 /*
598 * Search in the siblings
599 */
dea33cfd 600 list_for_each_entry(c, &p->real_parent->children, sibling) {
cf475ad2
BS
601 if (c->mm == mm)
602 goto assign_new_owner;
603 }
604
605 /*
606 * Search through everything else. We should not get
607 * here often
608 */
609 do_each_thread(g, c) {
610 if (c->mm == mm)
611 goto assign_new_owner;
612 } while_each_thread(g, c);
613
614 read_unlock(&tasklist_lock);
31a78f23
BS
615 /*
616 * We found no owner yet mm_users > 1: this implies that we are
617 * most likely racing with swapoff (try_to_unuse()) or /proc or
e5991371 618 * ptrace or page migration (get_task_mm()). Mark owner as NULL.
31a78f23 619 */
31a78f23 620 mm->owner = NULL;
cf475ad2
BS
621 return;
622
623assign_new_owner:
624 BUG_ON(c == p);
625 get_task_struct(c);
626 /*
627 * The task_lock protects c->mm from changing.
628 * We always want mm->owner->mm == mm
629 */
630 task_lock(c);
e5991371
HD
631 /*
632 * Delay read_unlock() till we have the task_lock()
633 * to ensure that c does not slip away underneath us
634 */
635 read_unlock(&tasklist_lock);
cf475ad2
BS
636 if (c->mm != mm) {
637 task_unlock(c);
638 put_task_struct(c);
639 goto retry;
640 }
cf475ad2
BS
641 mm->owner = c;
642 task_unlock(c);
643 put_task_struct(c);
644}
645#endif /* CONFIG_MM_OWNER */
646
1da177e4
LT
647/*
648 * Turn us into a lazy TLB process if we
649 * aren't already..
650 */
408b664a 651static void exit_mm(struct task_struct * tsk)
1da177e4
LT
652{
653 struct mm_struct *mm = tsk->mm;
b564daf8 654 struct core_state *core_state;
1da177e4
LT
655
656 mm_release(tsk, mm);
657 if (!mm)
658 return;
659 /*
660 * Serialize with any possible pending coredump.
999d9fc1 661 * We must hold mmap_sem around checking core_state
1da177e4 662 * and clearing tsk->mm. The core-inducing thread
999d9fc1 663 * will increment ->nr_threads for each thread in the
1da177e4
LT
664 * group with ->mm != NULL.
665 */
666 down_read(&mm->mmap_sem);
b564daf8
ON
667 core_state = mm->core_state;
668 if (core_state) {
669 struct core_thread self;
1da177e4 670 up_read(&mm->mmap_sem);
1da177e4 671
b564daf8
ON
672 self.task = tsk;
673 self.next = xchg(&core_state->dumper.next, &self);
674 /*
675 * Implies mb(), the result of xchg() must be visible
676 * to core_state->dumper.
677 */
678 if (atomic_dec_and_test(&core_state->nr_threads))
679 complete(&core_state->startup);
1da177e4 680
a94e2d40
ON
681 for (;;) {
682 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
683 if (!self.task) /* see coredump_finish() */
684 break;
685 schedule();
686 }
687 __set_task_state(tsk, TASK_RUNNING);
1da177e4
LT
688 down_read(&mm->mmap_sem);
689 }
690 atomic_inc(&mm->mm_count);
125e1874 691 BUG_ON(mm != tsk->active_mm);
1da177e4
LT
692 /* more a memory barrier than a real lock */
693 task_lock(tsk);
694 tsk->mm = NULL;
695 up_read(&mm->mmap_sem);
696 enter_lazy_tlb(mm, current);
0c1eecfb
RW
697 /* We don't want this task to be frozen prematurely */
698 clear_freeze_flag(tsk);
3d5992d2
YH
699 if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
700 atomic_dec(&mm->oom_disable_count);
1da177e4 701 task_unlock(tsk);
cf475ad2 702 mm_update_next_owner(mm);
1da177e4
LT
703 mmput(mm);
704}
705
1da177e4
LT
706/*
707 * When we die, we re-parent all our children.
708 * Try to give them to another thread in our thread
709 * group, and if no such member exists, give it to
84d73786
SB
710 * the child reaper process (ie "init") in our pid
711 * space.
1da177e4 712 */
950bbabb 713static struct task_struct *find_new_reaper(struct task_struct *father)
d16e15f5
NK
714 __releases(&tasklist_lock)
715 __acquires(&tasklist_lock)
1da177e4 716{
950bbabb
ON
717 struct pid_namespace *pid_ns = task_active_pid_ns(father);
718 struct task_struct *thread;
1da177e4 719
950bbabb
ON
720 thread = father;
721 while_each_thread(father, thread) {
722 if (thread->flags & PF_EXITING)
723 continue;
724 if (unlikely(pid_ns->child_reaper == father))
725 pid_ns->child_reaper = thread;
726 return thread;
727 }
1da177e4 728
950bbabb
ON
729 if (unlikely(pid_ns->child_reaper == father)) {
730 write_unlock_irq(&tasklist_lock);
731 if (unlikely(pid_ns == &init_pid_ns))
732 panic("Attempted to kill init!");
1da177e4 733
950bbabb
ON
734 zap_pid_ns_processes(pid_ns);
735 write_lock_irq(&tasklist_lock);
1da177e4 736 /*
950bbabb
ON
737 * We can not clear ->child_reaper or leave it alone.
738 * There may by stealth EXIT_DEAD tasks on ->children,
739 * forget_original_parent() must move them somewhere.
1da177e4 740 */
950bbabb 741 pid_ns->child_reaper = init_pid_ns.child_reaper;
1da177e4 742 }
762a24be 743
950bbabb
ON
744 return pid_ns->child_reaper;
745}
746
5dfc80be
ON
747/*
748* Any that need to be release_task'd are put on the @dead list.
749 */
9cd80bbb 750static void reparent_leader(struct task_struct *father, struct task_struct *p,
5dfc80be
ON
751 struct list_head *dead)
752{
5dfc80be
ON
753 list_move_tail(&p->sibling, &p->real_parent->children);
754
755 if (task_detached(p))
756 return;
757 /*
758 * If this is a threaded reparent there is no need to
759 * notify anyone anything has happened.
760 */
761 if (same_thread_group(p->real_parent, father))
762 return;
763
764 /* We don't want people slaying init. */
765 p->exit_signal = SIGCHLD;
766
767 /* If it has exited notify the new parent about this child's death. */
5cb11446 768 if (!task_ptrace(p) &&
5dfc80be
ON
769 p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) {
770 do_notify_parent(p, p->exit_signal);
771 if (task_detached(p)) {
772 p->exit_state = EXIT_DEAD;
773 list_move_tail(&p->sibling, dead);
774 }
775 }
776
777 kill_orphaned_pgrp(p, father);
778}
779
762a24be 780static void forget_original_parent(struct task_struct *father)
1da177e4 781{
950bbabb 782 struct task_struct *p, *n, *reaper;
5dfc80be 783 LIST_HEAD(dead_children);
762a24be
ON
784
785 write_lock_irq(&tasklist_lock);
c7e49c14
ON
786 /*
787 * Note that exit_ptrace() and find_new_reaper() might
788 * drop tasklist_lock and reacquire it.
789 */
790 exit_ptrace(father);
950bbabb 791 reaper = find_new_reaper(father);
f470021a 792
03ff1797 793 list_for_each_entry_safe(p, n, &father->children, sibling) {
9cd80bbb
ON
794 struct task_struct *t = p;
795 do {
796 t->real_parent = reaper;
797 if (t->parent == father) {
798 BUG_ON(task_ptrace(t));
799 t->parent = t->real_parent;
800 }
801 if (t->pdeath_signal)
802 group_send_sig_info(t->pdeath_signal,
803 SEND_SIG_NOINFO, t);
804 } while_each_thread(p, t);
805 reparent_leader(father, p, &dead_children);
1da177e4 806 }
762a24be 807 write_unlock_irq(&tasklist_lock);
5dfc80be 808
762a24be 809 BUG_ON(!list_empty(&father->children));
762a24be 810
5dfc80be
ON
811 list_for_each_entry_safe(p, n, &dead_children, sibling) {
812 list_del_init(&p->sibling);
39c626ae
ON
813 release_task(p);
814 }
1da177e4
LT
815}
816
817/*
818 * Send signals to all our closest relatives so that they know
819 * to properly mourn us..
820 */
821c7de7 821static void exit_notify(struct task_struct *tsk, int group_dead)
1da177e4 822{
2b2a1ff6
RM
823 int signal;
824 void *cookie;
1da177e4 825
1da177e4
LT
826 /*
827 * This does two things:
828 *
829 * A. Make init inherit all the child processes
830 * B. Check to see if any process groups have become orphaned
831 * as a result of our exiting, and if they have any stopped
832 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
833 */
762a24be 834 forget_original_parent(tsk);
2e4a7072 835 exit_task_namespaces(tsk);
1da177e4 836
762a24be 837 write_lock_irq(&tasklist_lock);
821c7de7
ON
838 if (group_dead)
839 kill_orphaned_pgrp(tsk->group_leader, NULL);
1da177e4 840
24728448 841 /* Let father know we died
1da177e4
LT
842 *
843 * Thread signals are configurable, but you aren't going to use
25985edc 844 * that to send signals to arbitrary processes.
1da177e4
LT
845 * That stops right now.
846 *
847 * If the parent exec id doesn't match the exec id we saved
848 * when we started then we know the parent has changed security
849 * domain.
850 *
851 * If our self_exec id doesn't match our parent_exec_id then
852 * we have changed execution domain as these two values started
853 * the same after a fork.
1da177e4 854 */
d839fd4d 855 if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
f49ee505 856 (tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
432870da 857 tsk->self_exec_id != tsk->parent_exec_id))
1da177e4
LT
858 tsk->exit_signal = SIGCHLD;
859
2b2a1ff6 860 signal = tracehook_notify_death(tsk, &cookie, group_dead);
5c7edcd7 861 if (signal >= 0)
2b2a1ff6 862 signal = do_notify_parent(tsk, signal);
1da177e4 863
5c7edcd7 864 tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
1da177e4 865
9c339168
ON
866 /* mt-exec, de_thread() is waiting for group leader */
867 if (unlikely(tsk->signal->notify_count < 0))
6db840fa 868 wake_up_process(tsk->signal->group_exit_task);
1da177e4
LT
869 write_unlock_irq(&tasklist_lock);
870
2b2a1ff6
RM
871 tracehook_report_death(tsk, signal, cookie, group_dead);
872
1da177e4 873 /* If the process is dead, release it - nobody will wait for it */
5c7edcd7 874 if (signal == DEATH_REAP)
1da177e4 875 release_task(tsk);
1da177e4
LT
876}
877
e18eecb8
JD
878#ifdef CONFIG_DEBUG_STACK_USAGE
879static void check_stack_usage(void)
880{
881 static DEFINE_SPINLOCK(low_water_lock);
882 static int lowest_to_date = THREAD_SIZE;
e18eecb8
JD
883 unsigned long free;
884
7c9f8861 885 free = stack_not_used(current);
e18eecb8
JD
886
887 if (free >= lowest_to_date)
888 return;
889
890 spin_lock(&low_water_lock);
891 if (free < lowest_to_date) {
892 printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
893 "left\n",
894 current->comm, free);
895 lowest_to_date = free;
896 }
897 spin_unlock(&low_water_lock);
898}
899#else
900static inline void check_stack_usage(void) {}
901#endif
902
7ad5b3a5 903NORET_TYPE void do_exit(long code)
1da177e4
LT
904{
905 struct task_struct *tsk = current;
906 int group_dead;
907
908 profile_task_exit(tsk);
909
22e2c507 910 WARN_ON(atomic_read(&tsk->fs_excl));
73c10101 911 WARN_ON(blk_needs_flush_plug(tsk));
22e2c507 912
1da177e4
LT
913 if (unlikely(in_interrupt()))
914 panic("Aiee, killing interrupt handler!");
915 if (unlikely(!tsk->pid))
916 panic("Attempted to kill the idle task!");
1da177e4 917
33dd94ae
NE
918 /*
919 * If do_exit is called because this processes oopsed, it's possible
920 * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before
921 * continuing. Amongst other possible reasons, this is to prevent
922 * mm_release()->clear_child_tid() from writing to a user-controlled
923 * kernel address.
924 */
925 set_fs(USER_DS);
926
30199f5a 927 tracehook_report_exit(&code);
1da177e4 928
e0e81739
DH
929 validate_creds_for_do_exit(tsk);
930
df164db5
AN
931 /*
932 * We're taking recursive faults here in do_exit. Safest is to just
933 * leave this task alone and wait for reboot.
934 */
935 if (unlikely(tsk->flags & PF_EXITING)) {
936 printk(KERN_ALERT
937 "Fixing recursive fault but reboot is needed!\n");
778e9a9c
AK
938 /*
939 * We can do this unlocked here. The futex code uses
940 * this flag just to verify whether the pi state
941 * cleanup has been done or not. In the worst case it
942 * loops once more. We pretend that the cleanup was
943 * done as there is no way to return. Either the
944 * OWNER_DIED bit is set by now or we push the blocked
945 * task into the wait for ever nirwana as well.
946 */
947 tsk->flags |= PF_EXITPIDONE;
df164db5
AN
948 set_current_state(TASK_UNINTERRUPTIBLE);
949 schedule();
950 }
951
3aa551c9
TG
952 exit_irq_thread();
953
d12619b5 954 exit_signals(tsk); /* sets PF_EXITING */
778e9a9c
AK
955 /*
956 * tsk->flags are checked in the futex code to protect against
957 * an exiting task cleaning up the robust pi futexes.
958 */
d2ee7198 959 smp_mb();
1d615482 960 raw_spin_unlock_wait(&tsk->pi_lock);
1da177e4 961
1da177e4
LT
962 if (unlikely(in_atomic()))
963 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
ba25f9dc 964 current->comm, task_pid_nr(current),
1da177e4
LT
965 preempt_count());
966
967 acct_update_integrals(tsk);
34e55232 968 /* sync mm's RSS info before statistics gathering */
a3a2e76c
KH
969 if (tsk->mm)
970 sync_mm_rss(tsk, tsk->mm);
1da177e4 971 group_dead = atomic_dec_and_test(&tsk->signal->live);
c3068951 972 if (group_dead) {
778e9a9c 973 hrtimer_cancel(&tsk->signal->real_timer);
25f407f0 974 exit_itimers(tsk->signal);
1f10206c
JP
975 if (tsk->mm)
976 setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm);
c3068951 977 }
f6ec29a4 978 acct_collect(code, group_dead);
522ed776
MT
979 if (group_dead)
980 tty_audit_exit();
fa84cb93
AV
981 if (unlikely(tsk->audit_context))
982 audit_free(tsk);
115085ea 983
f2ab6d88 984 tsk->exit_code = code;
115085ea 985 taskstats_exit(tsk, group_dead);
c757249a 986
1da177e4
LT
987 exit_mm(tsk);
988
0e464814 989 if (group_dead)
f6ec29a4 990 acct_process();
0a16b607
MD
991 trace_sched_process_exit(tsk);
992
1da177e4 993 exit_sem(tsk);
1ec7f1dd
AV
994 exit_files(tsk);
995 exit_fs(tsk);
e18eecb8 996 check_stack_usage();
1da177e4 997 exit_thread();
0b3fcf17
SE
998
999 /*
1000 * Flush inherited counters to the parent - before the parent
1001 * gets woken up by child-exit notifications.
1002 *
1003 * because of cgroup mode, must be called before cgroup_exit()
1004 */
1005 perf_event_exit_task(tsk);
1006
b4f48b63 1007 cgroup_exit(tsk, 1);
1da177e4 1008
5ec93d11 1009 if (group_dead)
1da177e4
LT
1010 disassociate_ctty(1);
1011
a1261f54 1012 module_put(task_thread_info(tsk)->exec_domain->module);
1da177e4 1013
9f46080c 1014 proc_exit_connector(tsk);
33b2fb30 1015
24f1e32c
FW
1016 /*
1017 * FIXME: do that only when needed, using sched_exit tracepoint
1018 */
1019 flush_ptrace_hw_breakpoint(tsk);
33b2fb30 1020
821c7de7 1021 exit_notify(tsk, group_dead);
1da177e4 1022#ifdef CONFIG_NUMA
c0ff7453 1023 task_lock(tsk);
f0be3d32 1024 mpol_put(tsk->mempolicy);
1da177e4 1025 tsk->mempolicy = NULL;
c0ff7453 1026 task_unlock(tsk);
1da177e4 1027#endif
42b2dd0a 1028#ifdef CONFIG_FUTEX
c87e2837
IM
1029 if (unlikely(current->pi_state_cache))
1030 kfree(current->pi_state_cache);
42b2dd0a 1031#endif
de5097c2 1032 /*
9a11b49a 1033 * Make sure we are holding no locks:
de5097c2 1034 */
9a11b49a 1035 debug_check_no_locks_held(tsk);
778e9a9c
AK
1036 /*
1037 * We can do this unlocked here. The futex code uses this flag
1038 * just to verify whether the pi state cleanup has been done
1039 * or not. In the worst case it loops once more.
1040 */
1041 tsk->flags |= PF_EXITPIDONE;
1da177e4 1042
afc847b7 1043 if (tsk->io_context)
b69f2292 1044 exit_io_context(tsk);
afc847b7 1045
b92ce558
JA
1046 if (tsk->splice_pipe)
1047 __free_pipe_info(tsk->splice_pipe);
1048
e0e81739
DH
1049 validate_creds_for_do_exit(tsk);
1050
7407251a 1051 preempt_disable();
f41d911f 1052 exit_rcu();
55a101f8 1053 /* causes final put_task_struct in finish_task_switch(). */
c394cc9f 1054 tsk->state = TASK_DEAD;
1da177e4
LT
1055 schedule();
1056 BUG();
1057 /* Avoid "noreturn function does return". */
54306cf0
AC
1058 for (;;)
1059 cpu_relax(); /* For when BUG is null */
1da177e4
LT
1060}
1061
012914da
RA
1062EXPORT_SYMBOL_GPL(do_exit);
1063
1da177e4
LT
1064NORET_TYPE void complete_and_exit(struct completion *comp, long code)
1065{
1066 if (comp)
1067 complete(comp);
55a101f8 1068
1da177e4
LT
1069 do_exit(code);
1070}
1071
1072EXPORT_SYMBOL(complete_and_exit);
1073
754fe8d2 1074SYSCALL_DEFINE1(exit, int, error_code)
1da177e4
LT
1075{
1076 do_exit((error_code&0xff)<<8);
1077}
1078
1da177e4
LT
1079/*
1080 * Take down every thread in the group. This is called by fatal signals
1081 * as well as by sys_exit_group (below).
1082 */
1083NORET_TYPE void
1084do_group_exit(int exit_code)
1085{
bfc4b089
ON
1086 struct signal_struct *sig = current->signal;
1087
1da177e4
LT
1088 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
1089
bfc4b089
ON
1090 if (signal_group_exit(sig))
1091 exit_code = sig->group_exit_code;
1da177e4 1092 else if (!thread_group_empty(current)) {
1da177e4 1093 struct sighand_struct *const sighand = current->sighand;
1da177e4 1094 spin_lock_irq(&sighand->siglock);
ed5d2cac 1095 if (signal_group_exit(sig))
1da177e4
LT
1096 /* Another thread got here before we took the lock. */
1097 exit_code = sig->group_exit_code;
1098 else {
1da177e4 1099 sig->group_exit_code = exit_code;
ed5d2cac 1100 sig->flags = SIGNAL_GROUP_EXIT;
1da177e4
LT
1101 zap_other_threads(current);
1102 }
1103 spin_unlock_irq(&sighand->siglock);
1da177e4
LT
1104 }
1105
1106 do_exit(exit_code);
1107 /* NOTREACHED */
1108}
1109
1110/*
1111 * this kills every thread in the thread group. Note that any externally
1112 * wait4()-ing process will get the correct exit code - even if this
1113 * thread is not the thread group leader.
1114 */
754fe8d2 1115SYSCALL_DEFINE1(exit_group, int, error_code)
1da177e4
LT
1116{
1117 do_group_exit((error_code & 0xff) << 8);
2ed7c03e
HC
1118 /* NOTREACHED */
1119 return 0;
1da177e4
LT
1120}
1121
9e8ae01d
ON
1122struct wait_opts {
1123 enum pid_type wo_type;
9e8ae01d 1124 int wo_flags;
e1eb1ebc 1125 struct pid *wo_pid;
9e8ae01d
ON
1126
1127 struct siginfo __user *wo_info;
1128 int __user *wo_stat;
1129 struct rusage __user *wo_rusage;
1130
0b7570e7 1131 wait_queue_t child_wait;
9e8ae01d
ON
1132 int notask_error;
1133};
1134
989264f4
ON
1135static inline
1136struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
161550d7 1137{
989264f4
ON
1138 if (type != PIDTYPE_PID)
1139 task = task->group_leader;
1140 return task->pids[type].pid;
161550d7
EB
1141}
1142
989264f4 1143static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1da177e4 1144{
5c01ba49
ON
1145 return wo->wo_type == PIDTYPE_MAX ||
1146 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1147}
1da177e4 1148
5c01ba49
ON
1149static int eligible_child(struct wait_opts *wo, struct task_struct *p)
1150{
1151 if (!eligible_pid(wo, p))
1152 return 0;
1da177e4
LT
1153 /* Wait for all children (clone and not) if __WALL is set;
1154 * otherwise, wait for clone children *only* if __WCLONE is
1155 * set; otherwise, wait for non-clone children *only*. (Note:
1156 * A "clone" child here is one that reports to its parent
1157 * using a signal other than SIGCHLD.) */
9e8ae01d
ON
1158 if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE))
1159 && !(wo->wo_flags & __WALL))
1da177e4 1160 return 0;
1da177e4 1161
14dd0b81 1162 return 1;
1da177e4
LT
1163}
1164
9e8ae01d
ON
1165static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1166 pid_t pid, uid_t uid, int why, int status)
1da177e4 1167{
9e8ae01d
ON
1168 struct siginfo __user *infop;
1169 int retval = wo->wo_rusage
1170 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
36c8b586 1171
1da177e4 1172 put_task_struct(p);
9e8ae01d 1173 infop = wo->wo_info;
b6fe2d11
VM
1174 if (infop) {
1175 if (!retval)
1176 retval = put_user(SIGCHLD, &infop->si_signo);
1177 if (!retval)
1178 retval = put_user(0, &infop->si_errno);
1179 if (!retval)
1180 retval = put_user((short)why, &infop->si_code);
1181 if (!retval)
1182 retval = put_user(pid, &infop->si_pid);
1183 if (!retval)
1184 retval = put_user(uid, &infop->si_uid);
1185 if (!retval)
1186 retval = put_user(status, &infop->si_status);
1187 }
1da177e4
LT
1188 if (!retval)
1189 retval = pid;
1190 return retval;
1191}
1192
1193/*
1194 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1195 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1196 * the lock and this task is uninteresting. If we return nonzero, we have
1197 * released the lock and the system call should return.
1198 */
9e8ae01d 1199static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
1da177e4
LT
1200{
1201 unsigned long state;
2f4e6e2a 1202 int retval, status, traced;
6c5f3e7b 1203 pid_t pid = task_pid_vnr(p);
c69e8d9c 1204 uid_t uid = __task_cred(p)->uid;
9e8ae01d 1205 struct siginfo __user *infop;
1da177e4 1206
9e8ae01d 1207 if (!likely(wo->wo_flags & WEXITED))
98abed02
RM
1208 return 0;
1209
9e8ae01d 1210 if (unlikely(wo->wo_flags & WNOWAIT)) {
1da177e4 1211 int exit_code = p->exit_code;
f3abd4f9 1212 int why;
1da177e4 1213
1da177e4
LT
1214 get_task_struct(p);
1215 read_unlock(&tasklist_lock);
1216 if ((exit_code & 0x7f) == 0) {
1217 why = CLD_EXITED;
1218 status = exit_code >> 8;
1219 } else {
1220 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1221 status = exit_code & 0x7f;
1222 }
9e8ae01d 1223 return wait_noreap_copyout(wo, p, pid, uid, why, status);
1da177e4
LT
1224 }
1225
1226 /*
1227 * Try to move the task's state to DEAD
1228 * only one thread is allowed to do this:
1229 */
1230 state = xchg(&p->exit_state, EXIT_DEAD);
1231 if (state != EXIT_ZOMBIE) {
1232 BUG_ON(state != EXIT_DEAD);
1233 return 0;
1234 }
1da177e4 1235
53b6f9fb 1236 traced = ptrace_reparented(p);
befca967
ON
1237 /*
1238 * It can be ptraced but not reparented, check
1239 * !task_detached() to filter out sub-threads.
1240 */
1241 if (likely(!traced) && likely(!task_detached(p))) {
3795e161
JJ
1242 struct signal_struct *psig;
1243 struct signal_struct *sig;
1f10206c 1244 unsigned long maxrss;
0cf55e1e 1245 cputime_t tgutime, tgstime;
3795e161 1246
1da177e4
LT
1247 /*
1248 * The resource counters for the group leader are in its
1249 * own task_struct. Those for dead threads in the group
1250 * are in its signal_struct, as are those for the child
1251 * processes it has previously reaped. All these
1252 * accumulate in the parent's signal_struct c* fields.
1253 *
1254 * We don't bother to take a lock here to protect these
1255 * p->signal fields, because they are only touched by
1256 * __exit_signal, which runs with tasklist_lock
1257 * write-locked anyway, and so is excluded here. We do
d1e98f42 1258 * need to protect the access to parent->signal fields,
1da177e4
LT
1259 * as other threads in the parent group can be right
1260 * here reaping other children at the same time.
0cf55e1e
HS
1261 *
1262 * We use thread_group_times() to get times for the thread
1263 * group, which consolidates times for all threads in the
1264 * group including the group leader.
1da177e4 1265 */
0cf55e1e 1266 thread_group_times(p, &tgutime, &tgstime);
d1e98f42
ON
1267 spin_lock_irq(&p->real_parent->sighand->siglock);
1268 psig = p->real_parent->signal;
3795e161
JJ
1269 sig = p->signal;
1270 psig->cutime =
1271 cputime_add(psig->cutime,
0cf55e1e
HS
1272 cputime_add(tgutime,
1273 sig->cutime));
3795e161
JJ
1274 psig->cstime =
1275 cputime_add(psig->cstime,
0cf55e1e
HS
1276 cputime_add(tgstime,
1277 sig->cstime));
9ac52315
LV
1278 psig->cgtime =
1279 cputime_add(psig->cgtime,
1280 cputime_add(p->gtime,
1281 cputime_add(sig->gtime,
1282 sig->cgtime)));
3795e161
JJ
1283 psig->cmin_flt +=
1284 p->min_flt + sig->min_flt + sig->cmin_flt;
1285 psig->cmaj_flt +=
1286 p->maj_flt + sig->maj_flt + sig->cmaj_flt;
1287 psig->cnvcsw +=
1288 p->nvcsw + sig->nvcsw + sig->cnvcsw;
1289 psig->cnivcsw +=
1290 p->nivcsw + sig->nivcsw + sig->cnivcsw;
6eaeeaba
ED
1291 psig->cinblock +=
1292 task_io_get_inblock(p) +
1293 sig->inblock + sig->cinblock;
1294 psig->coublock +=
1295 task_io_get_oublock(p) +
1296 sig->oublock + sig->coublock;
1f10206c
JP
1297 maxrss = max(sig->maxrss, sig->cmaxrss);
1298 if (psig->cmaxrss < maxrss)
1299 psig->cmaxrss = maxrss;
5995477a
AR
1300 task_io_accounting_add(&psig->ioac, &p->ioac);
1301 task_io_accounting_add(&psig->ioac, &sig->ioac);
d1e98f42 1302 spin_unlock_irq(&p->real_parent->sighand->siglock);
1da177e4
LT
1303 }
1304
1305 /*
1306 * Now we are sure this task is interesting, and no other
1307 * thread can reap it because we set its state to EXIT_DEAD.
1308 */
1309 read_unlock(&tasklist_lock);
1310
9e8ae01d
ON
1311 retval = wo->wo_rusage
1312 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1da177e4
LT
1313 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1314 ? p->signal->group_exit_code : p->exit_code;
9e8ae01d
ON
1315 if (!retval && wo->wo_stat)
1316 retval = put_user(status, wo->wo_stat);
1317
1318 infop = wo->wo_info;
1da177e4
LT
1319 if (!retval && infop)
1320 retval = put_user(SIGCHLD, &infop->si_signo);
1321 if (!retval && infop)
1322 retval = put_user(0, &infop->si_errno);
1323 if (!retval && infop) {
1324 int why;
1325
1326 if ((status & 0x7f) == 0) {
1327 why = CLD_EXITED;
1328 status >>= 8;
1329 } else {
1330 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1331 status &= 0x7f;
1332 }
1333 retval = put_user((short)why, &infop->si_code);
1334 if (!retval)
1335 retval = put_user(status, &infop->si_status);
1336 }
1337 if (!retval && infop)
3a515e4a 1338 retval = put_user(pid, &infop->si_pid);
1da177e4 1339 if (!retval && infop)
c69e8d9c 1340 retval = put_user(uid, &infop->si_uid);
2f4e6e2a 1341 if (!retval)
3a515e4a 1342 retval = pid;
2f4e6e2a
ON
1343
1344 if (traced) {
1da177e4 1345 write_lock_irq(&tasklist_lock);
2f4e6e2a
ON
1346 /* We dropped tasklist, ptracer could die and untrace */
1347 ptrace_unlink(p);
1348 /*
1349 * If this is not a detached task, notify the parent.
1350 * If it's still not detached after that, don't release
1351 * it now.
1352 */
d839fd4d 1353 if (!task_detached(p)) {
2f4e6e2a 1354 do_notify_parent(p, p->exit_signal);
d839fd4d 1355 if (!task_detached(p)) {
2f4e6e2a
ON
1356 p->exit_state = EXIT_ZOMBIE;
1357 p = NULL;
1da177e4
LT
1358 }
1359 }
1360 write_unlock_irq(&tasklist_lock);
1361 }
1362 if (p != NULL)
1363 release_task(p);
2f4e6e2a 1364
1da177e4
LT
1365 return retval;
1366}
1367
90bc8d8b
ON
1368static int *task_stopped_code(struct task_struct *p, bool ptrace)
1369{
1370 if (ptrace) {
1371 if (task_is_stopped_or_traced(p))
1372 return &p->exit_code;
1373 } else {
1374 if (p->signal->flags & SIGNAL_STOP_STOPPED)
1375 return &p->signal->group_exit_code;
1376 }
1377 return NULL;
1378}
1379
1da177e4
LT
1380/*
1381 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1382 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1383 * the lock and this task is uninteresting. If we return nonzero, we have
1384 * released the lock and the system call should return.
1385 */
9e8ae01d
ON
1386static int wait_task_stopped(struct wait_opts *wo,
1387 int ptrace, struct task_struct *p)
1da177e4 1388{
9e8ae01d 1389 struct siginfo __user *infop;
90bc8d8b 1390 int retval, exit_code, *p_code, why;
ee7c82da 1391 uid_t uid = 0; /* unneeded, required by compiler */
c8950783 1392 pid_t pid;
1da177e4 1393
47918025
ON
1394 /*
1395 * Traditionally we see ptrace'd stopped tasks regardless of options.
1396 */
9e8ae01d 1397 if (!ptrace && !(wo->wo_flags & WUNTRACED))
98abed02
RM
1398 return 0;
1399
ee7c82da
ON
1400 exit_code = 0;
1401 spin_lock_irq(&p->sighand->siglock);
1402
90bc8d8b
ON
1403 p_code = task_stopped_code(p, ptrace);
1404 if (unlikely(!p_code))
ee7c82da
ON
1405 goto unlock_sig;
1406
90bc8d8b 1407 exit_code = *p_code;
ee7c82da
ON
1408 if (!exit_code)
1409 goto unlock_sig;
1410
9e8ae01d 1411 if (!unlikely(wo->wo_flags & WNOWAIT))
90bc8d8b 1412 *p_code = 0;
ee7c82da 1413
f362b732 1414 uid = task_uid(p);
ee7c82da
ON
1415unlock_sig:
1416 spin_unlock_irq(&p->sighand->siglock);
1417 if (!exit_code)
1da177e4
LT
1418 return 0;
1419
1420 /*
1421 * Now we are pretty sure this task is interesting.
1422 * Make sure it doesn't get reaped out from under us while we
1423 * give up the lock and then examine it below. We don't want to
1424 * keep holding onto the tasklist_lock while we call getrusage and
1425 * possibly take page faults for user memory.
1426 */
1427 get_task_struct(p);
6c5f3e7b 1428 pid = task_pid_vnr(p);
f470021a 1429 why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
1da177e4
LT
1430 read_unlock(&tasklist_lock);
1431
9e8ae01d
ON
1432 if (unlikely(wo->wo_flags & WNOWAIT))
1433 return wait_noreap_copyout(wo, p, pid, uid, why, exit_code);
1434
1435 retval = wo->wo_rusage
1436 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1437 if (!retval && wo->wo_stat)
1438 retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat);
1da177e4 1439
9e8ae01d 1440 infop = wo->wo_info;
1da177e4
LT
1441 if (!retval && infop)
1442 retval = put_user(SIGCHLD, &infop->si_signo);
1443 if (!retval && infop)
1444 retval = put_user(0, &infop->si_errno);
1445 if (!retval && infop)
6efcae46 1446 retval = put_user((short)why, &infop->si_code);
1da177e4
LT
1447 if (!retval && infop)
1448 retval = put_user(exit_code, &infop->si_status);
1449 if (!retval && infop)
c8950783 1450 retval = put_user(pid, &infop->si_pid);
1da177e4 1451 if (!retval && infop)
ee7c82da 1452 retval = put_user(uid, &infop->si_uid);
1da177e4 1453 if (!retval)
c8950783 1454 retval = pid;
1da177e4
LT
1455 put_task_struct(p);
1456
1457 BUG_ON(!retval);
1458 return retval;
1459}
1460
1461/*
1462 * Handle do_wait work for one task in a live, non-stopped state.
1463 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1464 * the lock and this task is uninteresting. If we return nonzero, we have
1465 * released the lock and the system call should return.
1466 */
9e8ae01d 1467static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1da177e4
LT
1468{
1469 int retval;
1470 pid_t pid;
1471 uid_t uid;
1472
9e8ae01d 1473 if (!unlikely(wo->wo_flags & WCONTINUED))
98abed02
RM
1474 return 0;
1475
1da177e4
LT
1476 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1477 return 0;
1478
1479 spin_lock_irq(&p->sighand->siglock);
1480 /* Re-check with the lock held. */
1481 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1482 spin_unlock_irq(&p->sighand->siglock);
1483 return 0;
1484 }
9e8ae01d 1485 if (!unlikely(wo->wo_flags & WNOWAIT))
1da177e4 1486 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
f362b732 1487 uid = task_uid(p);
1da177e4
LT
1488 spin_unlock_irq(&p->sighand->siglock);
1489
6c5f3e7b 1490 pid = task_pid_vnr(p);
1da177e4
LT
1491 get_task_struct(p);
1492 read_unlock(&tasklist_lock);
1493
9e8ae01d
ON
1494 if (!wo->wo_info) {
1495 retval = wo->wo_rusage
1496 ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0;
1da177e4 1497 put_task_struct(p);
9e8ae01d
ON
1498 if (!retval && wo->wo_stat)
1499 retval = put_user(0xffff, wo->wo_stat);
1da177e4 1500 if (!retval)
3a515e4a 1501 retval = pid;
1da177e4 1502 } else {
9e8ae01d
ON
1503 retval = wait_noreap_copyout(wo, p, pid, uid,
1504 CLD_CONTINUED, SIGCONT);
1da177e4
LT
1505 BUG_ON(retval == 0);
1506 }
1507
1508 return retval;
1509}
1510
98abed02
RM
1511/*
1512 * Consider @p for a wait by @parent.
1513 *
9e8ae01d 1514 * -ECHILD should be in ->notask_error before the first call.
98abed02
RM
1515 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1516 * Returns zero if the search for a child should continue;
9e8ae01d 1517 * then ->notask_error is 0 if @p is an eligible child,
14dd0b81 1518 * or another error from security_task_wait(), or still -ECHILD.
98abed02 1519 */
b6e763f0
ON
1520static int wait_consider_task(struct wait_opts *wo, int ptrace,
1521 struct task_struct *p)
98abed02 1522{
9e8ae01d 1523 int ret = eligible_child(wo, p);
14dd0b81 1524 if (!ret)
98abed02
RM
1525 return ret;
1526
a2322e1d 1527 ret = security_task_wait(p);
14dd0b81
RM
1528 if (unlikely(ret < 0)) {
1529 /*
1530 * If we have not yet seen any eligible child,
1531 * then let this error code replace -ECHILD.
1532 * A permission error will give the user a clue
1533 * to look for security policy problems, rather
1534 * than for mysterious wait bugs.
1535 */
9e8ae01d
ON
1536 if (wo->notask_error)
1537 wo->notask_error = ret;
78a3d9d5 1538 return 0;
14dd0b81
RM
1539 }
1540
5cb11446 1541 if (likely(!ptrace) && unlikely(task_ptrace(p))) {
f470021a
RM
1542 /*
1543 * This child is hidden by ptrace.
1544 * We aren't allowed to see it now, but eventually we will.
1545 */
9e8ae01d 1546 wo->notask_error = 0;
f470021a
RM
1547 return 0;
1548 }
1549
98abed02
RM
1550 if (p->exit_state == EXIT_DEAD)
1551 return 0;
1552
1553 /*
1554 * We don't reap group leaders with subthreads.
1555 */
1556 if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
9e8ae01d 1557 return wait_task_zombie(wo, p);
98abed02
RM
1558
1559 /*
1560 * It's stopped or running now, so it might
1561 * later continue, exit, or stop again.
1562 */
9e8ae01d 1563 wo->notask_error = 0;
98abed02 1564
90bc8d8b 1565 if (task_stopped_code(p, ptrace))
9e8ae01d 1566 return wait_task_stopped(wo, ptrace, p);
98abed02 1567
9e8ae01d 1568 return wait_task_continued(wo, p);
98abed02
RM
1569}
1570
1571/*
1572 * Do the work of do_wait() for one thread in the group, @tsk.
1573 *
9e8ae01d 1574 * -ECHILD should be in ->notask_error before the first call.
98abed02
RM
1575 * Returns nonzero for a final return, when we have unlocked tasklist_lock.
1576 * Returns zero if the search for a child should continue; then
9e8ae01d 1577 * ->notask_error is 0 if there were any eligible children,
14dd0b81 1578 * or another error from security_task_wait(), or still -ECHILD.
98abed02 1579 */
9e8ae01d 1580static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
98abed02
RM
1581{
1582 struct task_struct *p;
1583
1584 list_for_each_entry(p, &tsk->children, sibling) {
9cd80bbb
ON
1585 int ret = wait_consider_task(wo, 0, p);
1586 if (ret)
1587 return ret;
98abed02
RM
1588 }
1589
1590 return 0;
1591}
1592
9e8ae01d 1593static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
98abed02
RM
1594{
1595 struct task_struct *p;
1596
f470021a 1597 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
b6e763f0 1598 int ret = wait_consider_task(wo, 1, p);
f470021a 1599 if (ret)
98abed02 1600 return ret;
98abed02
RM
1601 }
1602
1603 return 0;
1604}
1605
0b7570e7
ON
1606static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1607 int sync, void *key)
1608{
1609 struct wait_opts *wo = container_of(wait, struct wait_opts,
1610 child_wait);
1611 struct task_struct *p = key;
1612
5c01ba49 1613 if (!eligible_pid(wo, p))
0b7570e7
ON
1614 return 0;
1615
b4fe5182
ON
1616 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1617 return 0;
1618
0b7570e7
ON
1619 return default_wake_function(wait, mode, sync, key);
1620}
1621
a7f0765e
ON
1622void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1623{
0b7570e7
ON
1624 __wake_up_sync_key(&parent->signal->wait_chldexit,
1625 TASK_INTERRUPTIBLE, 1, p);
a7f0765e
ON
1626}
1627
9e8ae01d 1628static long do_wait(struct wait_opts *wo)
1da177e4 1629{
1da177e4 1630 struct task_struct *tsk;
98abed02 1631 int retval;
1da177e4 1632
9e8ae01d 1633 trace_sched_process_wait(wo->wo_pid);
0a16b607 1634
0b7570e7
ON
1635 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1636 wo->child_wait.private = current;
1637 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1da177e4 1638repeat:
98abed02
RM
1639 /*
1640 * If there is nothing that can match our critiera just get out.
9e8ae01d
ON
1641 * We will clear ->notask_error to zero if we see any child that
1642 * might later match our criteria, even if we are not able to reap
1643 * it yet.
98abed02 1644 */
64a16caf 1645 wo->notask_error = -ECHILD;
9e8ae01d
ON
1646 if ((wo->wo_type < PIDTYPE_MAX) &&
1647 (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type])))
64a16caf 1648 goto notask;
161550d7 1649
f95d39d1 1650 set_current_state(TASK_INTERRUPTIBLE);
1da177e4
LT
1651 read_lock(&tasklist_lock);
1652 tsk = current;
1653 do {
64a16caf
ON
1654 retval = do_wait_thread(wo, tsk);
1655 if (retval)
1656 goto end;
9e8ae01d 1657
64a16caf
ON
1658 retval = ptrace_do_wait(wo, tsk);
1659 if (retval)
98abed02 1660 goto end;
98abed02 1661
9e8ae01d 1662 if (wo->wo_flags & __WNOTHREAD)
1da177e4 1663 break;
a3f6dfb7 1664 } while_each_thread(current, tsk);
1da177e4 1665 read_unlock(&tasklist_lock);
f2cc3eb1 1666
64a16caf 1667notask:
9e8ae01d
ON
1668 retval = wo->notask_error;
1669 if (!retval && !(wo->wo_flags & WNOHANG)) {
1da177e4 1670 retval = -ERESTARTSYS;
98abed02
RM
1671 if (!signal_pending(current)) {
1672 schedule();
1673 goto repeat;
1674 }
1da177e4 1675 }
1da177e4 1676end:
f95d39d1 1677 __set_current_state(TASK_RUNNING);
0b7570e7 1678 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1da177e4
LT
1679 return retval;
1680}
1681
17da2bd9
HC
1682SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1683 infop, int, options, struct rusage __user *, ru)
1da177e4 1684{
9e8ae01d 1685 struct wait_opts wo;
161550d7
EB
1686 struct pid *pid = NULL;
1687 enum pid_type type;
1da177e4
LT
1688 long ret;
1689
1690 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1691 return -EINVAL;
1692 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1693 return -EINVAL;
1694
1695 switch (which) {
1696 case P_ALL:
161550d7 1697 type = PIDTYPE_MAX;
1da177e4
LT
1698 break;
1699 case P_PID:
161550d7
EB
1700 type = PIDTYPE_PID;
1701 if (upid <= 0)
1da177e4
LT
1702 return -EINVAL;
1703 break;
1704 case P_PGID:
161550d7
EB
1705 type = PIDTYPE_PGID;
1706 if (upid <= 0)
1da177e4 1707 return -EINVAL;
1da177e4
LT
1708 break;
1709 default:
1710 return -EINVAL;
1711 }
1712
161550d7
EB
1713 if (type < PIDTYPE_MAX)
1714 pid = find_get_pid(upid);
9e8ae01d
ON
1715
1716 wo.wo_type = type;
1717 wo.wo_pid = pid;
1718 wo.wo_flags = options;
1719 wo.wo_info = infop;
1720 wo.wo_stat = NULL;
1721 wo.wo_rusage = ru;
1722 ret = do_wait(&wo);
dfe16dfa
VM
1723
1724 if (ret > 0) {
1725 ret = 0;
1726 } else if (infop) {
1727 /*
1728 * For a WNOHANG return, clear out all the fields
1729 * we would set so the user can easily tell the
1730 * difference.
1731 */
1732 if (!ret)
1733 ret = put_user(0, &infop->si_signo);
1734 if (!ret)
1735 ret = put_user(0, &infop->si_errno);
1736 if (!ret)
1737 ret = put_user(0, &infop->si_code);
1738 if (!ret)
1739 ret = put_user(0, &infop->si_pid);
1740 if (!ret)
1741 ret = put_user(0, &infop->si_uid);
1742 if (!ret)
1743 ret = put_user(0, &infop->si_status);
1744 }
1745
161550d7 1746 put_pid(pid);
1da177e4
LT
1747
1748 /* avoid REGPARM breakage on x86: */
54a01510 1749 asmlinkage_protect(5, ret, which, upid, infop, options, ru);
1da177e4
LT
1750 return ret;
1751}
1752
754fe8d2
HC
1753SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr,
1754 int, options, struct rusage __user *, ru)
1da177e4 1755{
9e8ae01d 1756 struct wait_opts wo;
161550d7
EB
1757 struct pid *pid = NULL;
1758 enum pid_type type;
1da177e4
LT
1759 long ret;
1760
1761 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1762 __WNOTHREAD|__WCLONE|__WALL))
1763 return -EINVAL;
161550d7
EB
1764
1765 if (upid == -1)
1766 type = PIDTYPE_MAX;
1767 else if (upid < 0) {
1768 type = PIDTYPE_PGID;
1769 pid = find_get_pid(-upid);
1770 } else if (upid == 0) {
1771 type = PIDTYPE_PGID;
2ae448ef 1772 pid = get_task_pid(current, PIDTYPE_PGID);
161550d7
EB
1773 } else /* upid > 0 */ {
1774 type = PIDTYPE_PID;
1775 pid = find_get_pid(upid);
1776 }
1777
9e8ae01d
ON
1778 wo.wo_type = type;
1779 wo.wo_pid = pid;
1780 wo.wo_flags = options | WEXITED;
1781 wo.wo_info = NULL;
1782 wo.wo_stat = stat_addr;
1783 wo.wo_rusage = ru;
1784 ret = do_wait(&wo);
161550d7 1785 put_pid(pid);
1da177e4
LT
1786
1787 /* avoid REGPARM breakage on x86: */
54a01510 1788 asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
1da177e4
LT
1789 return ret;
1790}
1791
1792#ifdef __ARCH_WANT_SYS_WAITPID
1793
1794/*
1795 * sys_waitpid() remains for compatibility. waitpid() should be
1796 * implemented by calling sys_wait4() from libc.a.
1797 */
17da2bd9 1798SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options)
1da177e4
LT
1799{
1800 return sys_wait4(pid, stat_addr, options, NULL);
1801}
1802
1803#endif