blob: 3b25b182d2be35cfe8ddc69c53a908da96ce5aec [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/kernel/exit.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7#include <linux/config.h>
8#include <linux/mm.h>
9#include <linux/slab.h>
10#include <linux/interrupt.h>
11#include <linux/smp_lock.h>
12#include <linux/module.h>
13#include <linux/completion.h>
14#include <linux/personality.h>
15#include <linux/tty.h>
16#include <linux/namespace.h>
17#include <linux/key.h>
18#include <linux/security.h>
19#include <linux/cpu.h>
20#include <linux/acct.h>
21#include <linux/file.h>
22#include <linux/binfmts.h>
23#include <linux/ptrace.h>
24#include <linux/profile.h>
25#include <linux/mount.h>
26#include <linux/proc_fs.h>
27#include <linux/mempolicy.h>
28#include <linux/cpuset.h>
29#include <linux/syscalls.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070030#include <linux/signal.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
32#include <asm/uaccess.h>
33#include <asm/unistd.h>
34#include <asm/pgtable.h>
35#include <asm/mmu_context.h>
36
37extern void sem_exit (void);
38extern struct task_struct *child_reaper;
39
40int getrusage(struct task_struct *, int, struct rusage __user *);
41
Adrian Bunk408b6642005-05-01 08:59:29 -070042static void exit_mm(struct task_struct * tsk);
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044static void __unhash_process(struct task_struct *p)
45{
46 nr_threads--;
47 detach_pid(p, PIDTYPE_PID);
48 detach_pid(p, PIDTYPE_TGID);
49 if (thread_group_leader(p)) {
50 detach_pid(p, PIDTYPE_PGID);
51 detach_pid(p, PIDTYPE_SID);
52 if (p->pid)
53 __get_cpu_var(process_counts)--;
54 }
55
56 REMOVE_LINKS(p);
57}
58
59void release_task(struct task_struct * p)
60{
61 int zap_leader;
62 task_t *leader;
63 struct dentry *proc_dentry;
64
65repeat:
66 atomic_dec(&p->user->processes);
67 spin_lock(&p->proc_lock);
68 proc_dentry = proc_pid_unhash(p);
69 write_lock_irq(&tasklist_lock);
70 if (unlikely(p->ptrace))
71 __ptrace_unlink(p);
72 BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
73 __exit_signal(p);
74 __exit_sighand(p);
Christoph Lameter71a22242005-06-23 00:10:05 -070075 /*
76 * Note that the fastpath in sys_times depends on __exit_signal having
77 * updated the counters before a task is removed from the tasklist of
78 * the process by __unhash_process.
79 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 __unhash_process(p);
81
82 /*
83 * If we are the last non-leader member of the thread
84 * group, and the leader is zombie, then notify the
85 * group leader's parent process. (if it wants notification.)
86 */
87 zap_leader = 0;
88 leader = p->group_leader;
89 if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
90 BUG_ON(leader->exit_signal == -1);
91 do_notify_parent(leader, leader->exit_signal);
92 /*
93 * If we were the last child thread and the leader has
94 * exited already, and the leader's parent ignores SIGCHLD,
95 * then we are the one who should release the leader.
96 *
97 * do_notify_parent() will have marked it self-reaping in
98 * that case.
99 */
100 zap_leader = (leader->exit_signal == -1);
101 }
102
103 sched_exit(p);
104 write_unlock_irq(&tasklist_lock);
105 spin_unlock(&p->proc_lock);
106 proc_pid_flush(proc_dentry);
107 release_thread(p);
108 put_task_struct(p);
109
110 p = leader;
111 if (unlikely(zap_leader))
112 goto repeat;
113}
114
115/* we are using it only for SMP init */
116
117void unhash_process(struct task_struct *p)
118{
119 struct dentry *proc_dentry;
120
121 spin_lock(&p->proc_lock);
122 proc_dentry = proc_pid_unhash(p);
123 write_lock_irq(&tasklist_lock);
124 __unhash_process(p);
125 write_unlock_irq(&tasklist_lock);
126 spin_unlock(&p->proc_lock);
127 proc_pid_flush(proc_dentry);
128}
129
130/*
131 * This checks not only the pgrp, but falls back on the pid if no
132 * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
133 * without this...
134 */
135int session_of_pgrp(int pgrp)
136{
137 struct task_struct *p;
138 int sid = -1;
139
140 read_lock(&tasklist_lock);
141 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
142 if (p->signal->session > 0) {
143 sid = p->signal->session;
144 goto out;
145 }
146 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
147 p = find_task_by_pid(pgrp);
148 if (p)
149 sid = p->signal->session;
150out:
151 read_unlock(&tasklist_lock);
152
153 return sid;
154}
155
156/*
157 * Determine if a process group is "orphaned", according to the POSIX
158 * definition in 2.2.2.52. Orphaned process groups are not to be affected
159 * by terminal-generated stop signals. Newly orphaned process groups are
160 * to receive a SIGHUP and a SIGCONT.
161 *
162 * "I ask you, have you ever known what it is to be an orphan?"
163 */
164static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
165{
166 struct task_struct *p;
167 int ret = 1;
168
169 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
170 if (p == ignored_task
171 || p->exit_state
172 || p->real_parent->pid == 1)
173 continue;
174 if (process_group(p->real_parent) != pgrp
175 && p->real_parent->signal->session == p->signal->session) {
176 ret = 0;
177 break;
178 }
179 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
180 return ret; /* (sighing) "Often!" */
181}
182
183int is_orphaned_pgrp(int pgrp)
184{
185 int retval;
186
187 read_lock(&tasklist_lock);
188 retval = will_become_orphaned_pgrp(pgrp, NULL);
189 read_unlock(&tasklist_lock);
190
191 return retval;
192}
193
194static inline int has_stopped_jobs(int pgrp)
195{
196 int retval = 0;
197 struct task_struct *p;
198
199 do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
200 if (p->state != TASK_STOPPED)
201 continue;
202
203 /* If p is stopped by a debugger on a signal that won't
204 stop it, then don't count p as stopped. This isn't
205 perfect but it's a good approximation. */
206 if (unlikely (p->ptrace)
207 && p->exit_code != SIGSTOP
208 && p->exit_code != SIGTSTP
209 && p->exit_code != SIGTTOU
210 && p->exit_code != SIGTTIN)
211 continue;
212
213 retval = 1;
214 break;
215 } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
216 return retval;
217}
218
219/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700220 * reparent_to_init - Reparent the calling kernel thread to the init task.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 *
222 * If a kernel thread is launched as a result of a system call, or if
223 * it ever exits, it should generally reparent itself to init so that
224 * it is correctly cleaned up on exit.
225 *
226 * The various task state such as scheduling policy and priority may have
227 * been inherited from a user process, so we reset them to sane values here.
228 *
229 * NOTE that reparent_to_init() gives the caller full capabilities.
230 */
Coywolf Qi Hunt6c46ada2005-04-16 15:26:01 -0700231static inline void reparent_to_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232{
233 write_lock_irq(&tasklist_lock);
234
235 ptrace_unlink(current);
236 /* Reparent to init */
237 REMOVE_LINKS(current);
238 current->parent = child_reaper;
239 current->real_parent = child_reaper;
240 SET_LINKS(current);
241
242 /* Set the exit signal to SIGCHLD so we signal init on exit */
243 current->exit_signal = SIGCHLD;
244
245 if ((current->policy == SCHED_NORMAL) && (task_nice(current) < 0))
246 set_user_nice(current, 0);
247 /* cpus_allowed? */
248 /* rt_priority? */
249 /* signals? */
250 security_task_reparent_to_init(current);
251 memcpy(current->signal->rlim, init_task.signal->rlim,
252 sizeof(current->signal->rlim));
253 atomic_inc(&(INIT_USER->__count));
254 write_unlock_irq(&tasklist_lock);
255 switch_uid(INIT_USER);
256}
257
258void __set_special_pids(pid_t session, pid_t pgrp)
259{
260 struct task_struct *curr = current;
261
262 if (curr->signal->session != session) {
263 detach_pid(curr, PIDTYPE_SID);
264 curr->signal->session = session;
265 attach_pid(curr, PIDTYPE_SID, session);
266 }
267 if (process_group(curr) != pgrp) {
268 detach_pid(curr, PIDTYPE_PGID);
269 curr->signal->pgrp = pgrp;
270 attach_pid(curr, PIDTYPE_PGID, pgrp);
271 }
272}
273
274void set_special_pids(pid_t session, pid_t pgrp)
275{
276 write_lock_irq(&tasklist_lock);
277 __set_special_pids(session, pgrp);
278 write_unlock_irq(&tasklist_lock);
279}
280
281/*
282 * Let kernel threads use this to say that they
283 * allow a certain signal (since daemonize() will
284 * have disabled all of them by default).
285 */
286int allow_signal(int sig)
287{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700288 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 return -EINVAL;
290
291 spin_lock_irq(&current->sighand->siglock);
292 sigdelset(&current->blocked, sig);
293 if (!current->mm) {
294 /* Kernel threads handle their own signals.
295 Let the signal code know it'll be handled, so
296 that they don't get converted to SIGKILL or
297 just silently dropped */
298 current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
299 }
300 recalc_sigpending();
301 spin_unlock_irq(&current->sighand->siglock);
302 return 0;
303}
304
305EXPORT_SYMBOL(allow_signal);
306
307int disallow_signal(int sig)
308{
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700309 if (!valid_signal(sig) || sig < 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 return -EINVAL;
311
312 spin_lock_irq(&current->sighand->siglock);
313 sigaddset(&current->blocked, sig);
314 recalc_sigpending();
315 spin_unlock_irq(&current->sighand->siglock);
316 return 0;
317}
318
319EXPORT_SYMBOL(disallow_signal);
320
321/*
322 * Put all the gunge required to become a kernel thread without
323 * attached user resources in one place where it belongs.
324 */
325
326void daemonize(const char *name, ...)
327{
328 va_list args;
329 struct fs_struct *fs;
330 sigset_t blocked;
331
332 va_start(args, name);
333 vsnprintf(current->comm, sizeof(current->comm), name, args);
334 va_end(args);
335
336 /*
337 * If we were started as result of loading a module, close all of the
338 * user space pages. We don't need them, and if we didn't close them
339 * they would be locked into memory.
340 */
341 exit_mm(current);
342
343 set_special_pids(1, 1);
344 down(&tty_sem);
345 current->signal->tty = NULL;
346 up(&tty_sem);
347
348 /* Block and flush all signals */
349 sigfillset(&blocked);
350 sigprocmask(SIG_BLOCK, &blocked, NULL);
351 flush_signals(current);
352
353 /* Become as one with the init task */
354
355 exit_fs(current); /* current->fs->count--; */
356 fs = init_task.fs;
357 current->fs = fs;
358 atomic_inc(&fs->count);
359 exit_files(current);
360 current->files = init_task.files;
361 atomic_inc(&current->files->count);
362
363 reparent_to_init();
364}
365
366EXPORT_SYMBOL(daemonize);
367
368static inline void close_files(struct files_struct * files)
369{
370 int i, j;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700371 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
373 j = 0;
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700374
375 /*
376 * It is safe to dereference the fd table without RCU or
377 * ->file_lock because this is the last reference to the
378 * files structure.
379 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700380 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 for (;;) {
382 unsigned long set;
383 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700384 if (i >= fdt->max_fdset || i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700386 set = fdt->open_fds->fds_bits[j++];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 while (set) {
388 if (set & 1) {
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700389 struct file * file = xchg(&fdt->fd[i], NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 if (file)
391 filp_close(file, files);
392 }
393 i++;
394 set >>= 1;
395 }
396 }
397}
398
399struct files_struct *get_files_struct(struct task_struct *task)
400{
401 struct files_struct *files;
402
403 task_lock(task);
404 files = task->files;
405 if (files)
406 atomic_inc(&files->count);
407 task_unlock(task);
408
409 return files;
410}
411
412void fastcall put_files_struct(struct files_struct *files)
413{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700414 struct fdtable *fdt;
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 if (atomic_dec_and_test(&files->count)) {
417 close_files(files);
418 /*
419 * Free the fd and fdset arrays if we expanded them.
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700420 * If the fdtable was embedded, pass files for freeing
421 * at the end of the RCU grace period. Otherwise,
422 * you can free files immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700424 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700425 if (fdt == &files->fdtab)
426 fdt->free_files = files;
427 else
428 kmem_cache_free(files_cachep, files);
429 free_fdtable(fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 }
431}
432
433EXPORT_SYMBOL(put_files_struct);
434
435static inline void __exit_files(struct task_struct *tsk)
436{
437 struct files_struct * files = tsk->files;
438
439 if (files) {
440 task_lock(tsk);
441 tsk->files = NULL;
442 task_unlock(tsk);
443 put_files_struct(files);
444 }
445}
446
447void exit_files(struct task_struct *tsk)
448{
449 __exit_files(tsk);
450}
451
452static inline void __put_fs_struct(struct fs_struct *fs)
453{
454 /* No need to hold fs->lock if we are killing it */
455 if (atomic_dec_and_test(&fs->count)) {
456 dput(fs->root);
457 mntput(fs->rootmnt);
458 dput(fs->pwd);
459 mntput(fs->pwdmnt);
460 if (fs->altroot) {
461 dput(fs->altroot);
462 mntput(fs->altrootmnt);
463 }
464 kmem_cache_free(fs_cachep, fs);
465 }
466}
467
468void put_fs_struct(struct fs_struct *fs)
469{
470 __put_fs_struct(fs);
471}
472
473static inline void __exit_fs(struct task_struct *tsk)
474{
475 struct fs_struct * fs = tsk->fs;
476
477 if (fs) {
478 task_lock(tsk);
479 tsk->fs = NULL;
480 task_unlock(tsk);
481 __put_fs_struct(fs);
482 }
483}
484
485void exit_fs(struct task_struct *tsk)
486{
487 __exit_fs(tsk);
488}
489
490EXPORT_SYMBOL_GPL(exit_fs);
491
492/*
493 * Turn us into a lazy TLB process if we
494 * aren't already..
495 */
Adrian Bunk408b6642005-05-01 08:59:29 -0700496static void exit_mm(struct task_struct * tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497{
498 struct mm_struct *mm = tsk->mm;
499
500 mm_release(tsk, mm);
501 if (!mm)
502 return;
503 /*
504 * Serialize with any possible pending coredump.
505 * We must hold mmap_sem around checking core_waiters
506 * and clearing tsk->mm. The core-inducing thread
507 * will increment core_waiters for each thread in the
508 * group with ->mm != NULL.
509 */
510 down_read(&mm->mmap_sem);
511 if (mm->core_waiters) {
512 up_read(&mm->mmap_sem);
513 down_write(&mm->mmap_sem);
514 if (!--mm->core_waiters)
515 complete(mm->core_startup_done);
516 up_write(&mm->mmap_sem);
517
518 wait_for_completion(&mm->core_done);
519 down_read(&mm->mmap_sem);
520 }
521 atomic_inc(&mm->mm_count);
522 if (mm != tsk->active_mm) BUG();
523 /* more a memory barrier than a real lock */
524 task_lock(tsk);
525 tsk->mm = NULL;
526 up_read(&mm->mmap_sem);
527 enter_lazy_tlb(mm, current);
528 task_unlock(tsk);
529 mmput(mm);
530}
531
532static inline void choose_new_parent(task_t *p, task_t *reaper, task_t *child_reaper)
533{
534 /*
535 * Make sure we're not reparenting to ourselves and that
536 * the parent is not a zombie.
537 */
538 BUG_ON(p == reaper || reaper->exit_state >= EXIT_ZOMBIE);
539 p->real_parent = reaper;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540}
541
542static inline void reparent_thread(task_t *p, task_t *father, int traced)
543{
544 /* We don't want people slaying init. */
545 if (p->exit_signal != -1)
546 p->exit_signal = SIGCHLD;
547
548 if (p->pdeath_signal)
549 /* We already hold the tasklist_lock here. */
550 group_send_sig_info(p->pdeath_signal, (void *) 0, p);
551
552 /* Move the child from its dying parent to the new one. */
553 if (unlikely(traced)) {
554 /* Preserve ptrace links if someone else is tracing this child. */
555 list_del_init(&p->ptrace_list);
556 if (p->parent != p->real_parent)
557 list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
558 } else {
559 /* If this child is being traced, then we're the one tracing it
560 * anyway, so let go of it.
561 */
562 p->ptrace = 0;
563 list_del_init(&p->sibling);
564 p->parent = p->real_parent;
565 list_add_tail(&p->sibling, &p->parent->children);
566
567 /* If we'd notified the old parent about this child's death,
568 * also notify the new parent.
569 */
570 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
571 thread_group_empty(p))
572 do_notify_parent(p, p->exit_signal);
573 else if (p->state == TASK_TRACED) {
574 /*
575 * If it was at a trace stop, turn it into
576 * a normal stop since it's no longer being
577 * traced.
578 */
579 ptrace_untrace(p);
580 }
581 }
582
583 /*
584 * process group orphan check
585 * Case ii: Our child is in a different pgrp
586 * than we are, and it was the only connection
587 * outside, so the child pgrp is now orphaned.
588 */
589 if ((process_group(p) != process_group(father)) &&
590 (p->signal->session == father->signal->session)) {
591 int pgrp = process_group(p);
592
593 if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
594 __kill_pg_info(SIGHUP, (void *)1, pgrp);
595 __kill_pg_info(SIGCONT, (void *)1, pgrp);
596 }
597 }
598}
599
600/*
601 * When we die, we re-parent all our children.
602 * Try to give them to another thread in our thread
603 * group, and if no such member exists, give it to
604 * the global child reaper process (ie "init")
605 */
606static inline void forget_original_parent(struct task_struct * father,
607 struct list_head *to_release)
608{
609 struct task_struct *p, *reaper = father;
610 struct list_head *_p, *_n;
611
612 do {
613 reaper = next_thread(reaper);
614 if (reaper == father) {
615 reaper = child_reaper;
616 break;
617 }
618 } while (reaper->exit_state);
619
620 /*
621 * There are only two places where our children can be:
622 *
623 * - in our child list
624 * - in our ptraced child list
625 *
626 * Search them and reparent children.
627 */
628 list_for_each_safe(_p, _n, &father->children) {
629 int ptrace;
630 p = list_entry(_p,struct task_struct,sibling);
631
632 ptrace = p->ptrace;
633
634 /* if father isn't the real parent, then ptrace must be enabled */
635 BUG_ON(father != p->real_parent && !ptrace);
636
637 if (father == p->real_parent) {
638 /* reparent with a reaper, real father it's us */
639 choose_new_parent(p, reaper, child_reaper);
640 reparent_thread(p, father, 0);
641 } else {
642 /* reparent ptraced task to its real parent */
643 __ptrace_unlink (p);
644 if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
645 thread_group_empty(p))
646 do_notify_parent(p, p->exit_signal);
647 }
648
649 /*
650 * if the ptraced child is a zombie with exit_signal == -1
651 * we must collect it before we exit, or it will remain
652 * zombie forever since we prevented it from self-reap itself
653 * while it was being traced by us, to be able to see it in wait4.
654 */
655 if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
656 list_add(&p->ptrace_list, to_release);
657 }
658 list_for_each_safe(_p, _n, &father->ptrace_children) {
659 p = list_entry(_p,struct task_struct,ptrace_list);
660 choose_new_parent(p, reaper, child_reaper);
661 reparent_thread(p, father, 1);
662 }
663}
664
665/*
666 * Send signals to all our closest relatives so that they know
667 * to properly mourn us..
668 */
669static void exit_notify(struct task_struct *tsk)
670{
671 int state;
672 struct task_struct *t;
673 struct list_head ptrace_dead, *_p, *_n;
674
675 if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
676 && !thread_group_empty(tsk)) {
677 /*
678 * This occurs when there was a race between our exit
679 * syscall and a group signal choosing us as the one to
680 * wake up. It could be that we are the only thread
681 * alerted to check for pending signals, but another thread
682 * should be woken now to take the signal since we will not.
683 * Now we'll wake all the threads in the group just to make
684 * sure someone gets all the pending signals.
685 */
686 read_lock(&tasklist_lock);
687 spin_lock_irq(&tsk->sighand->siglock);
688 for (t = next_thread(tsk); t != tsk; t = next_thread(t))
689 if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
690 recalc_sigpending_tsk(t);
691 if (signal_pending(t))
692 signal_wake_up(t, 0);
693 }
694 spin_unlock_irq(&tsk->sighand->siglock);
695 read_unlock(&tasklist_lock);
696 }
697
698 write_lock_irq(&tasklist_lock);
699
700 /*
701 * This does two things:
702 *
703 * A. Make init inherit all the child processes
704 * B. Check to see if any process groups have become orphaned
705 * as a result of our exiting, and if they have any stopped
706 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
707 */
708
709 INIT_LIST_HEAD(&ptrace_dead);
710 forget_original_parent(tsk, &ptrace_dead);
711 BUG_ON(!list_empty(&tsk->children));
712 BUG_ON(!list_empty(&tsk->ptrace_children));
713
714 /*
715 * Check to see if any process groups have become orphaned
716 * as a result of our exiting, and if they have any stopped
717 * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
718 *
719 * Case i: Our father is in a different pgrp than we are
720 * and we were the only connection outside, so our pgrp
721 * is about to become orphaned.
722 */
723
724 t = tsk->real_parent;
725
726 if ((process_group(t) != process_group(tsk)) &&
727 (t->signal->session == tsk->signal->session) &&
728 will_become_orphaned_pgrp(process_group(tsk), tsk) &&
729 has_stopped_jobs(process_group(tsk))) {
730 __kill_pg_info(SIGHUP, (void *)1, process_group(tsk));
731 __kill_pg_info(SIGCONT, (void *)1, process_group(tsk));
732 }
733
734 /* Let father know we died
735 *
736 * Thread signals are configurable, but you aren't going to use
737 * that to send signals to arbitary processes.
738 * That stops right now.
739 *
740 * If the parent exec id doesn't match the exec id we saved
741 * when we started then we know the parent has changed security
742 * domain.
743 *
744 * If our self_exec id doesn't match our parent_exec_id then
745 * we have changed execution domain as these two values started
746 * the same after a fork.
747 *
748 */
749
750 if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
751 ( tsk->parent_exec_id != t->self_exec_id ||
752 tsk->self_exec_id != tsk->parent_exec_id)
753 && !capable(CAP_KILL))
754 tsk->exit_signal = SIGCHLD;
755
756
757 /* If something other than our normal parent is ptracing us, then
758 * send it a SIGCHLD instead of honoring exit_signal. exit_signal
759 * only has special meaning to our real parent.
760 */
761 if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
762 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
763 do_notify_parent(tsk, signal);
764 } else if (tsk->ptrace) {
765 do_notify_parent(tsk, SIGCHLD);
766 }
767
768 state = EXIT_ZOMBIE;
769 if (tsk->exit_signal == -1 &&
770 (likely(tsk->ptrace == 0) ||
771 unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
772 state = EXIT_DEAD;
773 tsk->exit_state = state;
774
775 write_unlock_irq(&tasklist_lock);
776
777 list_for_each_safe(_p, _n, &ptrace_dead) {
778 list_del_init(_p);
779 t = list_entry(_p,struct task_struct,ptrace_list);
780 release_task(t);
781 }
782
783 /* If the process is dead, release it - nobody will wait for it */
784 if (state == EXIT_DEAD)
785 release_task(tsk);
786
787 /* PF_DEAD causes final put_task_struct after we schedule. */
788 preempt_disable();
789 tsk->flags |= PF_DEAD;
790}
791
792fastcall NORET_TYPE void do_exit(long code)
793{
794 struct task_struct *tsk = current;
795 int group_dead;
796
797 profile_task_exit(tsk);
798
Jens Axboe22e2c502005-06-27 10:55:12 +0200799 WARN_ON(atomic_read(&tsk->fs_excl));
800
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 if (unlikely(in_interrupt()))
802 panic("Aiee, killing interrupt handler!");
803 if (unlikely(!tsk->pid))
804 panic("Attempted to kill the idle task!");
805 if (unlikely(tsk->pid == 1))
806 panic("Attempted to kill init!");
807 if (tsk->io_context)
808 exit_io_context();
809
810 if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
811 current->ptrace_message = code;
812 ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
813 }
814
Alexander Nybergdf164db2005-06-23 00:09:13 -0700815 /*
816 * We're taking recursive faults here in do_exit. Safest is to just
817 * leave this task alone and wait for reboot.
818 */
819 if (unlikely(tsk->flags & PF_EXITING)) {
820 printk(KERN_ALERT
821 "Fixing recursive fault but reboot is needed!\n");
822 set_current_state(TASK_UNINTERRUPTIBLE);
823 schedule();
824 }
825
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 tsk->flags |= PF_EXITING;
827
828 /*
829 * Make sure we don't try to process any timer firings
830 * while we are already exiting.
831 */
832 tsk->it_virt_expires = cputime_zero;
833 tsk->it_prof_expires = cputime_zero;
834 tsk->it_sched_expires = 0;
835
836 if (unlikely(in_atomic()))
837 printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
838 current->comm, current->pid,
839 preempt_count());
840
841 acct_update_integrals(tsk);
842 update_mem_hiwater(tsk);
843 group_dead = atomic_dec_and_test(&tsk->signal->live);
Andrew Mortonc3068952005-08-04 16:49:32 -0700844 if (group_dead) {
845 del_timer_sync(&tsk->signal->real_timer);
Roland McGrath25f407f2005-10-21 15:03:29 -0700846 exit_itimers(tsk->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 acct_process(code);
Andrew Mortonc3068952005-08-04 16:49:32 -0700848 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 exit_mm(tsk);
850
851 exit_sem(tsk);
852 __exit_files(tsk);
853 __exit_fs(tsk);
854 exit_namespace(tsk);
855 exit_thread();
856 cpuset_exit(tsk);
857 exit_keys(tsk);
858
859 if (group_dead && tsk->signal->leader)
860 disassociate_ctty(1);
861
862 module_put(tsk->thread_info->exec_domain->module);
863 if (tsk->binfmt)
864 module_put(tsk->binfmt->module);
865
866 tsk->exit_code = code;
867 exit_notify(tsk);
868#ifdef CONFIG_NUMA
869 mpol_free(tsk->mempolicy);
870 tsk->mempolicy = NULL;
871#endif
872
873 BUG_ON(!(current->flags & PF_DEAD));
874 schedule();
875 BUG();
876 /* Avoid "noreturn function does return". */
877 for (;;) ;
878}
879
Russ Anderson012914d2005-04-23 00:08:00 -0700880EXPORT_SYMBOL_GPL(do_exit);
881
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882NORET_TYPE void complete_and_exit(struct completion *comp, long code)
883{
884 if (comp)
885 complete(comp);
886
887 do_exit(code);
888}
889
890EXPORT_SYMBOL(complete_and_exit);
891
892asmlinkage long sys_exit(int error_code)
893{
894 do_exit((error_code&0xff)<<8);
895}
896
897task_t fastcall *next_thread(const task_t *p)
898{
899 return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
900}
901
902EXPORT_SYMBOL(next_thread);
903
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 */
908NORET_TYPE void
909do_group_exit(int exit_code)
910{
911 BUG_ON(exit_code & 0x80); /* core dumps don't get here */
912
913 if (current->signal->flags & SIGNAL_GROUP_EXIT)
914 exit_code = current->signal->group_exit_code;
915 else if (!thread_group_empty(current)) {
916 struct signal_struct *const sig = current->signal;
917 struct sighand_struct *const sighand = current->sighand;
918 read_lock(&tasklist_lock);
919 spin_lock_irq(&sighand->siglock);
920 if (sig->flags & SIGNAL_GROUP_EXIT)
921 /* Another thread got here before we took the lock. */
922 exit_code = sig->group_exit_code;
923 else {
924 sig->flags = SIGNAL_GROUP_EXIT;
925 sig->group_exit_code = exit_code;
926 zap_other_threads(current);
927 }
928 spin_unlock_irq(&sighand->siglock);
929 read_unlock(&tasklist_lock);
930 }
931
932 do_exit(exit_code);
933 /* NOTREACHED */
934}
935
936/*
937 * this kills every thread in the thread group. Note that any externally
938 * wait4()-ing process will get the correct exit code - even if this
939 * thread is not the thread group leader.
940 */
941asmlinkage void sys_exit_group(int error_code)
942{
943 do_group_exit((error_code & 0xff) << 8);
944}
945
946static int eligible_child(pid_t pid, int options, task_t *p)
947{
948 if (pid > 0) {
949 if (p->pid != pid)
950 return 0;
951 } else if (!pid) {
952 if (process_group(p) != process_group(current))
953 return 0;
954 } else if (pid != -1) {
955 if (process_group(p) != -pid)
956 return 0;
957 }
958
959 /*
960 * Do not consider detached threads that are
961 * not ptraced:
962 */
963 if (p->exit_signal == -1 && !p->ptrace)
964 return 0;
965
966 /* Wait for all children (clone and not) if __WALL is set;
967 * otherwise, wait for clone children *only* if __WCLONE is
968 * set; otherwise, wait for non-clone children *only*. (Note:
969 * A "clone" child here is one that reports to its parent
970 * using a signal other than SIGCHLD.) */
971 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
972 && !(options & __WALL))
973 return 0;
974 /*
975 * Do not consider thread group leaders that are
976 * in a non-empty thread group:
977 */
978 if (current->tgid != p->tgid && delay_group_leader(p))
979 return 2;
980
981 if (security_task_wait(p))
982 return 0;
983
984 return 1;
985}
986
987static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
988 int why, int status,
989 struct siginfo __user *infop,
990 struct rusage __user *rusagep)
991{
992 int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
993 put_task_struct(p);
994 if (!retval)
995 retval = put_user(SIGCHLD, &infop->si_signo);
996 if (!retval)
997 retval = put_user(0, &infop->si_errno);
998 if (!retval)
999 retval = put_user((short)why, &infop->si_code);
1000 if (!retval)
1001 retval = put_user(pid, &infop->si_pid);
1002 if (!retval)
1003 retval = put_user(uid, &infop->si_uid);
1004 if (!retval)
1005 retval = put_user(status, &infop->si_status);
1006 if (!retval)
1007 retval = pid;
1008 return retval;
1009}
1010
1011/*
1012 * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
1013 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1014 * the lock and this task is uninteresting. If we return nonzero, we have
1015 * released the lock and the system call should return.
1016 */
1017static int wait_task_zombie(task_t *p, int noreap,
1018 struct siginfo __user *infop,
1019 int __user *stat_addr, struct rusage __user *ru)
1020{
1021 unsigned long state;
1022 int retval;
1023 int status;
1024
1025 if (unlikely(noreap)) {
1026 pid_t pid = p->pid;
1027 uid_t uid = p->uid;
1028 int exit_code = p->exit_code;
1029 int why, status;
1030
1031 if (unlikely(p->exit_state != EXIT_ZOMBIE))
1032 return 0;
1033 if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
1034 return 0;
1035 get_task_struct(p);
1036 read_unlock(&tasklist_lock);
1037 if ((exit_code & 0x7f) == 0) {
1038 why = CLD_EXITED;
1039 status = exit_code >> 8;
1040 } else {
1041 why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
1042 status = exit_code & 0x7f;
1043 }
1044 return wait_noreap_copyout(p, pid, uid, why,
1045 status, infop, ru);
1046 }
1047
1048 /*
1049 * Try to move the task's state to DEAD
1050 * only one thread is allowed to do this:
1051 */
1052 state = xchg(&p->exit_state, EXIT_DEAD);
1053 if (state != EXIT_ZOMBIE) {
1054 BUG_ON(state != EXIT_DEAD);
1055 return 0;
1056 }
1057 if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
1058 /*
1059 * This can only happen in a race with a ptraced thread
1060 * dying on another processor.
1061 */
1062 return 0;
1063 }
1064
1065 if (likely(p->real_parent == p->parent) && likely(p->signal)) {
1066 /*
1067 * The resource counters for the group leader are in its
1068 * own task_struct. Those for dead threads in the group
1069 * are in its signal_struct, as are those for the child
1070 * processes it has previously reaped. All these
1071 * accumulate in the parent's signal_struct c* fields.
1072 *
1073 * We don't bother to take a lock here to protect these
1074 * p->signal fields, because they are only touched by
1075 * __exit_signal, which runs with tasklist_lock
1076 * write-locked anyway, and so is excluded here. We do
1077 * need to protect the access to p->parent->signal fields,
1078 * as other threads in the parent group can be right
1079 * here reaping other children at the same time.
1080 */
1081 spin_lock_irq(&p->parent->sighand->siglock);
1082 p->parent->signal->cutime =
1083 cputime_add(p->parent->signal->cutime,
1084 cputime_add(p->utime,
1085 cputime_add(p->signal->utime,
1086 p->signal->cutime)));
1087 p->parent->signal->cstime =
1088 cputime_add(p->parent->signal->cstime,
1089 cputime_add(p->stime,
1090 cputime_add(p->signal->stime,
1091 p->signal->cstime)));
1092 p->parent->signal->cmin_flt +=
1093 p->min_flt + p->signal->min_flt + p->signal->cmin_flt;
1094 p->parent->signal->cmaj_flt +=
1095 p->maj_flt + p->signal->maj_flt + p->signal->cmaj_flt;
1096 p->parent->signal->cnvcsw +=
1097 p->nvcsw + p->signal->nvcsw + p->signal->cnvcsw;
1098 p->parent->signal->cnivcsw +=
1099 p->nivcsw + p->signal->nivcsw + p->signal->cnivcsw;
1100 spin_unlock_irq(&p->parent->sighand->siglock);
1101 }
1102
1103 /*
1104 * Now we are sure this task is interesting, and no other
1105 * thread can reap it because we set its state to EXIT_DEAD.
1106 */
1107 read_unlock(&tasklist_lock);
1108
1109 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1110 status = (p->signal->flags & SIGNAL_GROUP_EXIT)
1111 ? p->signal->group_exit_code : p->exit_code;
1112 if (!retval && stat_addr)
1113 retval = put_user(status, stat_addr);
1114 if (!retval && infop)
1115 retval = put_user(SIGCHLD, &infop->si_signo);
1116 if (!retval && infop)
1117 retval = put_user(0, &infop->si_errno);
1118 if (!retval && infop) {
1119 int why;
1120
1121 if ((status & 0x7f) == 0) {
1122 why = CLD_EXITED;
1123 status >>= 8;
1124 } else {
1125 why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
1126 status &= 0x7f;
1127 }
1128 retval = put_user((short)why, &infop->si_code);
1129 if (!retval)
1130 retval = put_user(status, &infop->si_status);
1131 }
1132 if (!retval && infop)
1133 retval = put_user(p->pid, &infop->si_pid);
1134 if (!retval && infop)
1135 retval = put_user(p->uid, &infop->si_uid);
1136 if (retval) {
1137 // TODO: is this safe?
1138 p->exit_state = EXIT_ZOMBIE;
1139 return retval;
1140 }
1141 retval = p->pid;
1142 if (p->real_parent != p->parent) {
1143 write_lock_irq(&tasklist_lock);
1144 /* Double-check with lock held. */
1145 if (p->real_parent != p->parent) {
1146 __ptrace_unlink(p);
1147 // TODO: is this safe?
1148 p->exit_state = EXIT_ZOMBIE;
1149 /*
1150 * If this is not a detached task, notify the parent.
1151 * If it's still not detached after that, don't release
1152 * it now.
1153 */
1154 if (p->exit_signal != -1) {
1155 do_notify_parent(p, p->exit_signal);
1156 if (p->exit_signal != -1)
1157 p = NULL;
1158 }
1159 }
1160 write_unlock_irq(&tasklist_lock);
1161 }
1162 if (p != NULL)
1163 release_task(p);
1164 BUG_ON(!retval);
1165 return retval;
1166}
1167
1168/*
1169 * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
1170 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1171 * the lock and this task is uninteresting. If we return nonzero, we have
1172 * released the lock and the system call should return.
1173 */
1174static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
1175 struct siginfo __user *infop,
1176 int __user *stat_addr, struct rusage __user *ru)
1177{
1178 int retval, exit_code;
1179
1180 if (!p->exit_code)
1181 return 0;
1182 if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
1183 p->signal && p->signal->group_stop_count > 0)
1184 /*
1185 * A group stop is in progress and this is the group leader.
1186 * We won't report until all threads have stopped.
1187 */
1188 return 0;
1189
1190 /*
1191 * Now we are pretty sure this task is interesting.
1192 * Make sure it doesn't get reaped out from under us while we
1193 * give up the lock and then examine it below. We don't want to
1194 * keep holding onto the tasklist_lock while we call getrusage and
1195 * possibly take page faults for user memory.
1196 */
1197 get_task_struct(p);
1198 read_unlock(&tasklist_lock);
1199
1200 if (unlikely(noreap)) {
1201 pid_t pid = p->pid;
1202 uid_t uid = p->uid;
1203 int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
1204
1205 exit_code = p->exit_code;
1206 if (unlikely(!exit_code) ||
Linus Torvalds14bf01b2005-10-01 11:04:18 -07001207 unlikely(p->state & TASK_TRACED))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208 goto bail_ref;
1209 return wait_noreap_copyout(p, pid, uid,
1210 why, (exit_code << 8) | 0x7f,
1211 infop, ru);
1212 }
1213
1214 write_lock_irq(&tasklist_lock);
1215
1216 /*
1217 * This uses xchg to be atomic with the thread resuming and setting
1218 * it. It must also be done with the write lock held to prevent a
1219 * race with the EXIT_ZOMBIE case.
1220 */
1221 exit_code = xchg(&p->exit_code, 0);
1222 if (unlikely(p->exit_state)) {
1223 /*
1224 * The task resumed and then died. Let the next iteration
1225 * catch it in EXIT_ZOMBIE. Note that exit_code might
1226 * already be zero here if it resumed and did _exit(0).
1227 * The task itself is dead and won't touch exit_code again;
1228 * other processors in this function are locked out.
1229 */
1230 p->exit_code = exit_code;
1231 exit_code = 0;
1232 }
1233 if (unlikely(exit_code == 0)) {
1234 /*
1235 * Another thread in this function got to it first, or it
1236 * resumed, or it resumed and then died.
1237 */
1238 write_unlock_irq(&tasklist_lock);
1239bail_ref:
1240 put_task_struct(p);
1241 /*
1242 * We are returning to the wait loop without having successfully
1243 * removed the process and having released the lock. We cannot
1244 * continue, since the "p" task pointer is potentially stale.
1245 *
1246 * Return -EAGAIN, and do_wait() will restart the loop from the
1247 * beginning. Do _not_ re-acquire the lock.
1248 */
1249 return -EAGAIN;
1250 }
1251
1252 /* move to end of parent's list to avoid starvation */
1253 remove_parent(p);
1254 add_parent(p, p->parent);
1255
1256 write_unlock_irq(&tasklist_lock);
1257
1258 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1259 if (!retval && stat_addr)
1260 retval = put_user((exit_code << 8) | 0x7f, stat_addr);
1261 if (!retval && infop)
1262 retval = put_user(SIGCHLD, &infop->si_signo);
1263 if (!retval && infop)
1264 retval = put_user(0, &infop->si_errno);
1265 if (!retval && infop)
1266 retval = put_user((short)((p->ptrace & PT_PTRACED)
1267 ? CLD_TRAPPED : CLD_STOPPED),
1268 &infop->si_code);
1269 if (!retval && infop)
1270 retval = put_user(exit_code, &infop->si_status);
1271 if (!retval && infop)
1272 retval = put_user(p->pid, &infop->si_pid);
1273 if (!retval && infop)
1274 retval = put_user(p->uid, &infop->si_uid);
1275 if (!retval)
1276 retval = p->pid;
1277 put_task_struct(p);
1278
1279 BUG_ON(!retval);
1280 return retval;
1281}
1282
1283/*
1284 * Handle do_wait work for one task in a live, non-stopped state.
1285 * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
1286 * the lock and this task is uninteresting. If we return nonzero, we have
1287 * released the lock and the system call should return.
1288 */
1289static int wait_task_continued(task_t *p, int noreap,
1290 struct siginfo __user *infop,
1291 int __user *stat_addr, struct rusage __user *ru)
1292{
1293 int retval;
1294 pid_t pid;
1295 uid_t uid;
1296
1297 if (unlikely(!p->signal))
1298 return 0;
1299
1300 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
1301 return 0;
1302
1303 spin_lock_irq(&p->sighand->siglock);
1304 /* Re-check with the lock held. */
1305 if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
1306 spin_unlock_irq(&p->sighand->siglock);
1307 return 0;
1308 }
1309 if (!noreap)
1310 p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
1311 spin_unlock_irq(&p->sighand->siglock);
1312
1313 pid = p->pid;
1314 uid = p->uid;
1315 get_task_struct(p);
1316 read_unlock(&tasklist_lock);
1317
1318 if (!infop) {
1319 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
1320 put_task_struct(p);
1321 if (!retval && stat_addr)
1322 retval = put_user(0xffff, stat_addr);
1323 if (!retval)
1324 retval = p->pid;
1325 } else {
1326 retval = wait_noreap_copyout(p, pid, uid,
1327 CLD_CONTINUED, SIGCONT,
1328 infop, ru);
1329 BUG_ON(retval == 0);
1330 }
1331
1332 return retval;
1333}
1334
1335
1336static inline int my_ptrace_child(struct task_struct *p)
1337{
1338 if (!(p->ptrace & PT_PTRACED))
1339 return 0;
1340 if (!(p->ptrace & PT_ATTACHED))
1341 return 1;
1342 /*
1343 * This child was PTRACE_ATTACH'd. We should be seeing it only if
1344 * we are the attacher. If we are the real parent, this is a race
1345 * inside ptrace_attach. It is waiting for the tasklist_lock,
1346 * which we have to switch the parent links, but has already set
1347 * the flags in p->ptrace.
1348 */
1349 return (p->parent != p->real_parent);
1350}
1351
1352static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
1353 int __user *stat_addr, struct rusage __user *ru)
1354{
1355 DECLARE_WAITQUEUE(wait, current);
1356 struct task_struct *tsk;
1357 int flag, retval;
1358
1359 add_wait_queue(&current->signal->wait_chldexit,&wait);
1360repeat:
1361 /*
1362 * We will set this flag if we see any child that might later
1363 * match our criteria, even if we are not able to reap it yet.
1364 */
1365 flag = 0;
1366 current->state = TASK_INTERRUPTIBLE;
1367 read_lock(&tasklist_lock);
1368 tsk = current;
1369 do {
1370 struct task_struct *p;
1371 struct list_head *_p;
1372 int ret;
1373
1374 list_for_each(_p,&tsk->children) {
1375 p = list_entry(_p,struct task_struct,sibling);
1376
1377 ret = eligible_child(pid, options, p);
1378 if (!ret)
1379 continue;
1380
1381 switch (p->state) {
1382 case TASK_TRACED:
1383 if (!my_ptrace_child(p))
1384 continue;
1385 /*FALLTHROUGH*/
1386 case TASK_STOPPED:
1387 /*
1388 * It's stopped now, so it might later
1389 * continue, exit, or stop again.
1390 */
1391 flag = 1;
1392 if (!(options & WUNTRACED) &&
1393 !my_ptrace_child(p))
1394 continue;
1395 retval = wait_task_stopped(p, ret == 2,
1396 (options & WNOWAIT),
1397 infop,
1398 stat_addr, ru);
1399 if (retval == -EAGAIN)
1400 goto repeat;
1401 if (retval != 0) /* He released the lock. */
1402 goto end;
1403 break;
1404 default:
1405 // case EXIT_DEAD:
1406 if (p->exit_state == EXIT_DEAD)
1407 continue;
1408 // case EXIT_ZOMBIE:
1409 if (p->exit_state == EXIT_ZOMBIE) {
1410 /*
1411 * Eligible but we cannot release
1412 * it yet:
1413 */
1414 if (ret == 2)
1415 goto check_continued;
1416 if (!likely(options & WEXITED))
1417 continue;
1418 retval = wait_task_zombie(
1419 p, (options & WNOWAIT),
1420 infop, stat_addr, ru);
1421 /* He released the lock. */
1422 if (retval != 0)
1423 goto end;
1424 break;
1425 }
1426check_continued:
1427 /*
1428 * It's running now, so it might later
1429 * exit, stop, or stop and then continue.
1430 */
1431 flag = 1;
1432 if (!unlikely(options & WCONTINUED))
1433 continue;
1434 retval = wait_task_continued(
1435 p, (options & WNOWAIT),
1436 infop, stat_addr, ru);
1437 if (retval != 0) /* He released the lock. */
1438 goto end;
1439 break;
1440 }
1441 }
1442 if (!flag) {
1443 list_for_each(_p, &tsk->ptrace_children) {
1444 p = list_entry(_p, struct task_struct,
1445 ptrace_list);
1446 if (!eligible_child(pid, options, p))
1447 continue;
1448 flag = 1;
1449 break;
1450 }
1451 }
1452 if (options & __WNOTHREAD)
1453 break;
1454 tsk = next_thread(tsk);
1455 if (tsk->signal != current->signal)
1456 BUG();
1457 } while (tsk != current);
1458
1459 read_unlock(&tasklist_lock);
1460 if (flag) {
1461 retval = 0;
1462 if (options & WNOHANG)
1463 goto end;
1464 retval = -ERESTARTSYS;
1465 if (signal_pending(current))
1466 goto end;
1467 schedule();
1468 goto repeat;
1469 }
1470 retval = -ECHILD;
1471end:
1472 current->state = TASK_RUNNING;
1473 remove_wait_queue(&current->signal->wait_chldexit,&wait);
1474 if (infop) {
1475 if (retval > 0)
1476 retval = 0;
1477 else {
1478 /*
1479 * For a WNOHANG return, clear out all the fields
1480 * we would set so the user can easily tell the
1481 * difference.
1482 */
1483 if (!retval)
1484 retval = put_user(0, &infop->si_signo);
1485 if (!retval)
1486 retval = put_user(0, &infop->si_errno);
1487 if (!retval)
1488 retval = put_user(0, &infop->si_code);
1489 if (!retval)
1490 retval = put_user(0, &infop->si_pid);
1491 if (!retval)
1492 retval = put_user(0, &infop->si_uid);
1493 if (!retval)
1494 retval = put_user(0, &infop->si_status);
1495 }
1496 }
1497 return retval;
1498}
1499
1500asmlinkage long sys_waitid(int which, pid_t pid,
1501 struct siginfo __user *infop, int options,
1502 struct rusage __user *ru)
1503{
1504 long ret;
1505
1506 if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
1507 return -EINVAL;
1508 if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
1509 return -EINVAL;
1510
1511 switch (which) {
1512 case P_ALL:
1513 pid = -1;
1514 break;
1515 case P_PID:
1516 if (pid <= 0)
1517 return -EINVAL;
1518 break;
1519 case P_PGID:
1520 if (pid <= 0)
1521 return -EINVAL;
1522 pid = -pid;
1523 break;
1524 default:
1525 return -EINVAL;
1526 }
1527
1528 ret = do_wait(pid, options, infop, NULL, ru);
1529
1530 /* avoid REGPARM breakage on x86: */
1531 prevent_tail_call(ret);
1532 return ret;
1533}
1534
1535asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
1536 int options, struct rusage __user *ru)
1537{
1538 long ret;
1539
1540 if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
1541 __WNOTHREAD|__WCLONE|__WALL))
1542 return -EINVAL;
1543 ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
1544
1545 /* avoid REGPARM breakage on x86: */
1546 prevent_tail_call(ret);
1547 return ret;
1548}
1549
1550#ifdef __ARCH_WANT_SYS_WAITPID
1551
1552/*
1553 * sys_waitpid() remains for compatibility. waitpid() should be
1554 * implemented by calling sys_wait4() from libc.a.
1555 */
1556asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
1557{
1558 return sys_wait4(pid, stat_addr, options, NULL);
1559}
1560
1561#endif