Root/
1 | /* |
2 | * linux/kernel/exit.c |
3 | * |
4 | * Copyright (C) 1991, 1992 Linus Torvalds |
5 | */ |
6 | |
7 | #include <linux/mm.h> |
8 | #include <linux/slab.h> |
9 | #include <linux/interrupt.h> |
10 | #include <linux/module.h> |
11 | #include <linux/capability.h> |
12 | #include <linux/completion.h> |
13 | #include <linux/personality.h> |
14 | #include <linux/tty.h> |
15 | #include <linux/iocontext.h> |
16 | #include <linux/key.h> |
17 | #include <linux/security.h> |
18 | #include <linux/cpu.h> |
19 | #include <linux/acct.h> |
20 | #include <linux/tsacct_kern.h> |
21 | #include <linux/file.h> |
22 | #include <linux/fdtable.h> |
23 | #include <linux/binfmts.h> |
24 | #include <linux/nsproxy.h> |
25 | #include <linux/pid_namespace.h> |
26 | #include <linux/ptrace.h> |
27 | #include <linux/profile.h> |
28 | #include <linux/mount.h> |
29 | #include <linux/proc_fs.h> |
30 | #include <linux/kthread.h> |
31 | #include <linux/mempolicy.h> |
32 | #include <linux/taskstats_kern.h> |
33 | #include <linux/delayacct.h> |
34 | #include <linux/freezer.h> |
35 | #include <linux/cgroup.h> |
36 | #include <linux/syscalls.h> |
37 | #include <linux/signal.h> |
38 | #include <linux/posix-timers.h> |
39 | #include <linux/cn_proc.h> |
40 | #include <linux/mutex.h> |
41 | #include <linux/futex.h> |
42 | #include <linux/pipe_fs_i.h> |
43 | #include <linux/audit.h> /* for audit_free() */ |
44 | #include <linux/resource.h> |
45 | #include <linux/blkdev.h> |
46 | #include <linux/task_io_accounting_ops.h> |
47 | #include <linux/tracehook.h> |
48 | #include <linux/fs_struct.h> |
49 | #include <linux/init_task.h> |
50 | #include <linux/perf_counter.h> |
51 | #include <trace/events/sched.h> |
52 | |
53 | #include <asm/uaccess.h> |
54 | #include <asm/unistd.h> |
55 | #include <asm/pgtable.h> |
56 | #include <asm/mmu_context.h> |
57 | #include "cred-internals.h" |
58 | |
59 | static void exit_mm(struct task_struct * tsk); |
60 | |
61 | static void __unhash_process(struct task_struct *p) |
62 | { |
63 | nr_threads--; |
64 | detach_pid(p, PIDTYPE_PID); |
65 | if (thread_group_leader(p)) { |
66 | detach_pid(p, PIDTYPE_PGID); |
67 | detach_pid(p, PIDTYPE_SID); |
68 | |
69 | list_del_rcu(&p->tasks); |
70 | __get_cpu_var(process_counts)--; |
71 | } |
72 | list_del_rcu(&p->thread_group); |
73 | list_del_init(&p->sibling); |
74 | } |
75 | |
76 | /* |
77 | * This function expects the tasklist_lock write-locked. |
78 | */ |
79 | static void __exit_signal(struct task_struct *tsk) |
80 | { |
81 | struct signal_struct *sig = tsk->signal; |
82 | struct sighand_struct *sighand; |
83 | |
84 | BUG_ON(!sig); |
85 | BUG_ON(!atomic_read(&sig->count)); |
86 | |
87 | sighand = rcu_dereference(tsk->sighand); |
88 | spin_lock(&sighand->siglock); |
89 | |
90 | posix_cpu_timers_exit(tsk); |
91 | if (atomic_dec_and_test(&sig->count)) |
92 | posix_cpu_timers_exit_group(tsk); |
93 | else { |
94 | /* |
95 | * If there is any task waiting for the group exit |
96 | * then notify it: |
97 | */ |
98 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) |
99 | wake_up_process(sig->group_exit_task); |
100 | |
101 | if (tsk == sig->curr_target) |
102 | sig->curr_target = next_thread(tsk); |
103 | /* |
104 | * Accumulate here the counters for all threads but the |
105 | * group leader as they die, so they can be added into |
106 | * the process-wide totals when those are taken. |
107 | * The group leader stays around as a zombie as long |
108 | * as there are other threads. When it gets reaped, |
109 | * the exit.c code will add its counts into these totals. |
110 | * We won't ever get here for the group leader, since it |
111 | * will have been the last reference on the signal_struct. |
112 | */ |
113 | sig->utime = cputime_add(sig->utime, task_utime(tsk)); |
114 | sig->stime = cputime_add(sig->stime, task_stime(tsk)); |
115 | sig->gtime = cputime_add(sig->gtime, task_gtime(tsk)); |
116 | sig->min_flt += tsk->min_flt; |
117 | sig->maj_flt += tsk->maj_flt; |
118 | sig->nvcsw += tsk->nvcsw; |
119 | sig->nivcsw += tsk->nivcsw; |
120 | sig->inblock += task_io_get_inblock(tsk); |
121 | sig->oublock += task_io_get_oublock(tsk); |
122 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
123 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
124 | sig = NULL; /* Marker for below. */ |
125 | } |
126 | |
127 | __unhash_process(tsk); |
128 | |
129 | /* |
130 | * Do this under ->siglock, we can race with another thread |
131 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. |
132 | */ |
133 | flush_sigqueue(&tsk->pending); |
134 | |
135 | tsk->signal = NULL; |
136 | tsk->sighand = NULL; |
137 | spin_unlock(&sighand->siglock); |
138 | |
139 | __cleanup_sighand(sighand); |
140 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
141 | if (sig) { |
142 | flush_sigqueue(&sig->shared_pending); |
143 | taskstats_tgid_free(sig); |
144 | /* |
145 | * Make sure ->signal can't go away under rq->lock, |
146 | * see account_group_exec_runtime(). |
147 | */ |
148 | task_rq_unlock_wait(tsk); |
149 | __cleanup_signal(sig); |
150 | } |
151 | } |
152 | |
153 | static void delayed_put_task_struct(struct rcu_head *rhp) |
154 | { |
155 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
156 | |
157 | #ifdef CONFIG_PERF_COUNTERS |
158 | WARN_ON_ONCE(tsk->perf_counter_ctxp); |
159 | #endif |
160 | trace_sched_process_free(tsk); |
161 | put_task_struct(tsk); |
162 | } |
163 | |
164 | |
165 | void release_task(struct task_struct * p) |
166 | { |
167 | struct task_struct *leader; |
168 | int zap_leader; |
169 | repeat: |
170 | tracehook_prepare_release_task(p); |
171 | /* don't need to get the RCU readlock here - the process is dead and |
172 | * can't be modifying its own credentials */ |
173 | atomic_dec(&__task_cred(p)->user->processes); |
174 | |
175 | proc_flush_task(p); |
176 | |
177 | write_lock_irq(&tasklist_lock); |
178 | tracehook_finish_release_task(p); |
179 | __exit_signal(p); |
180 | |
181 | /* |
182 | * If we are the last non-leader member of the thread |
183 | * group, and the leader is zombie, then notify the |
184 | * group leader's parent process. (if it wants notification.) |
185 | */ |
186 | zap_leader = 0; |
187 | leader = p->group_leader; |
188 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { |
189 | BUG_ON(task_detached(leader)); |
190 | do_notify_parent(leader, leader->exit_signal); |
191 | /* |
192 | * If we were the last child thread and the leader has |
193 | * exited already, and the leader's parent ignores SIGCHLD, |
194 | * then we are the one who should release the leader. |
195 | * |
196 | * do_notify_parent() will have marked it self-reaping in |
197 | * that case. |
198 | */ |
199 | zap_leader = task_detached(leader); |
200 | |
201 | /* |
202 | * This maintains the invariant that release_task() |
203 | * only runs on a task in EXIT_DEAD, just for sanity. |
204 | */ |
205 | if (zap_leader) |
206 | leader->exit_state = EXIT_DEAD; |
207 | } |
208 | |
209 | write_unlock_irq(&tasklist_lock); |
210 | release_thread(p); |
211 | call_rcu(&p->rcu, delayed_put_task_struct); |
212 | |
213 | p = leader; |
214 | if (unlikely(zap_leader)) |
215 | goto repeat; |
216 | } |
217 | |
218 | /* |
219 | * This checks not only the pgrp, but falls back on the pid if no |
220 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly |
221 | * without this... |
222 | * |
223 | * The caller must hold rcu lock or the tasklist lock. |
224 | */ |
225 | struct pid *session_of_pgrp(struct pid *pgrp) |
226 | { |
227 | struct task_struct *p; |
228 | struct pid *sid = NULL; |
229 | |
230 | p = pid_task(pgrp, PIDTYPE_PGID); |
231 | if (p == NULL) |
232 | p = pid_task(pgrp, PIDTYPE_PID); |
233 | if (p != NULL) |
234 | sid = task_session(p); |
235 | |
236 | return sid; |
237 | } |
238 | |
239 | /* |
240 | * Determine if a process group is "orphaned", according to the POSIX |
241 | * definition in 2.2.2.52. Orphaned process groups are not to be affected |
242 | * by terminal-generated stop signals. Newly orphaned process groups are |
243 | * to receive a SIGHUP and a SIGCONT. |
244 | * |
245 | * "I ask you, have you ever known what it is to be an orphan?" |
246 | */ |
247 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
248 | { |
249 | struct task_struct *p; |
250 | |
251 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
252 | if ((p == ignored_task) || |
253 | (p->exit_state && thread_group_empty(p)) || |
254 | is_global_init(p->real_parent)) |
255 | continue; |
256 | |
257 | if (task_pgrp(p->real_parent) != pgrp && |
258 | task_session(p->real_parent) == task_session(p)) |
259 | return 0; |
260 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
261 | |
262 | return 1; |
263 | } |
264 | |
265 | int is_current_pgrp_orphaned(void) |
266 | { |
267 | int retval; |
268 | |
269 | read_lock(&tasklist_lock); |
270 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
271 | read_unlock(&tasklist_lock); |
272 | |
273 | return retval; |
274 | } |
275 | |
276 | static int has_stopped_jobs(struct pid *pgrp) |
277 | { |
278 | int retval = 0; |
279 | struct task_struct *p; |
280 | |
281 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
282 | if (!task_is_stopped(p)) |
283 | continue; |
284 | retval = 1; |
285 | break; |
286 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
287 | return retval; |
288 | } |
289 | |
290 | /* |
291 | * Check to see if any process groups have become orphaned as |
292 | * a result of our exiting, and if they have any stopped jobs, |
293 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
294 | */ |
295 | static void |
296 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) |
297 | { |
298 | struct pid *pgrp = task_pgrp(tsk); |
299 | struct task_struct *ignored_task = tsk; |
300 | |
301 | if (!parent) |
302 | /* exit: our father is in a different pgrp than |
303 | * we are and we were the only connection outside. |
304 | */ |
305 | parent = tsk->real_parent; |
306 | else |
307 | /* reparent: our child is in a different pgrp than |
308 | * we are, and it was the only connection outside. |
309 | */ |
310 | ignored_task = NULL; |
311 | |
312 | if (task_pgrp(parent) != pgrp && |
313 | task_session(parent) == task_session(tsk) && |
314 | will_become_orphaned_pgrp(pgrp, ignored_task) && |
315 | has_stopped_jobs(pgrp)) { |
316 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); |
317 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); |
318 | } |
319 | } |
320 | |
321 | /** |
322 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
323 | * |
324 | * If a kernel thread is launched as a result of a system call, or if |
325 | * it ever exits, it should generally reparent itself to kthreadd so it |
326 | * isn't in the way of other processes and is correctly cleaned up on exit. |
327 | * |
328 | * The various task state such as scheduling policy and priority may have |
329 | * been inherited from a user process, so we reset them to sane values here. |
330 | * |
331 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
332 | */ |
333 | static void reparent_to_kthreadd(void) |
334 | { |
335 | write_lock_irq(&tasklist_lock); |
336 | |
337 | ptrace_unlink(current); |
338 | /* Reparent to init */ |
339 | current->real_parent = current->parent = kthreadd_task; |
340 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
341 | |
342 | /* Set the exit signal to SIGCHLD so we signal init on exit */ |
343 | current->exit_signal = SIGCHLD; |
344 | |
345 | if (task_nice(current) < 0) |
346 | set_user_nice(current, 0); |
347 | /* cpus_allowed? */ |
348 | /* rt_priority? */ |
349 | /* signals? */ |
350 | memcpy(current->signal->rlim, init_task.signal->rlim, |
351 | sizeof(current->signal->rlim)); |
352 | |
353 | atomic_inc(&init_cred.usage); |
354 | commit_creds(&init_cred); |
355 | write_unlock_irq(&tasklist_lock); |
356 | } |
357 | |
358 | void __set_special_pids(struct pid *pid) |
359 | { |
360 | struct task_struct *curr = current->group_leader; |
361 | |
362 | if (task_session(curr) != pid) |
363 | change_pid(curr, PIDTYPE_SID, pid); |
364 | |
365 | if (task_pgrp(curr) != pid) |
366 | change_pid(curr, PIDTYPE_PGID, pid); |
367 | } |
368 | |
369 | static void set_special_pids(struct pid *pid) |
370 | { |
371 | write_lock_irq(&tasklist_lock); |
372 | __set_special_pids(pid); |
373 | write_unlock_irq(&tasklist_lock); |
374 | } |
375 | |
376 | /* |
377 | * Let kernel threads use this to say that they allow a certain signal. |
378 | * Must not be used if kthread was cloned with CLONE_SIGHAND. |
379 | */ |
380 | int allow_signal(int sig) |
381 | { |
382 | if (!valid_signal(sig) || sig < 1) |
383 | return -EINVAL; |
384 | |
385 | spin_lock_irq(¤t->sighand->siglock); |
386 | /* This is only needed for daemonize()'ed kthreads */ |
387 | sigdelset(¤t->blocked, sig); |
388 | /* |
389 | * Kernel threads handle their own signals. Let the signal code |
390 | * know it'll be handled, so that they don't get converted to |
391 | * SIGKILL or just silently dropped. |
392 | */ |
393 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; |
394 | recalc_sigpending(); |
395 | spin_unlock_irq(¤t->sighand->siglock); |
396 | return 0; |
397 | } |
398 | |
399 | EXPORT_SYMBOL(allow_signal); |
400 | |
401 | int disallow_signal(int sig) |
402 | { |
403 | if (!valid_signal(sig) || sig < 1) |
404 | return -EINVAL; |
405 | |
406 | spin_lock_irq(¤t->sighand->siglock); |
407 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
408 | recalc_sigpending(); |
409 | spin_unlock_irq(¤t->sighand->siglock); |
410 | return 0; |
411 | } |
412 | |
413 | EXPORT_SYMBOL(disallow_signal); |
414 | |
415 | /* |
416 | * Put all the gunge required to become a kernel thread without |
417 | * attached user resources in one place where it belongs. |
418 | */ |
419 | |
420 | void daemonize(const char *name, ...) |
421 | { |
422 | va_list args; |
423 | sigset_t blocked; |
424 | |
425 | va_start(args, name); |
426 | vsnprintf(current->comm, sizeof(current->comm), name, args); |
427 | va_end(args); |
428 | |
429 | /* |
430 | * If we were started as result of loading a module, close all of the |
431 | * user space pages. We don't need them, and if we didn't close them |
432 | * they would be locked into memory. |
433 | */ |
434 | exit_mm(current); |
435 | /* |
436 | * We don't want to have TIF_FREEZE set if the system-wide hibernation |
437 | * or suspend transition begins right now. |
438 | */ |
439 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
440 | |
441 | if (current->nsproxy != &init_nsproxy) { |
442 | get_nsproxy(&init_nsproxy); |
443 | switch_task_namespaces(current, &init_nsproxy); |
444 | } |
445 | set_special_pids(&init_struct_pid); |
446 | proc_clear_tty(current); |
447 | |
448 | /* Block and flush all signals */ |
449 | sigfillset(&blocked); |
450 | sigprocmask(SIG_BLOCK, &blocked, NULL); |
451 | flush_signals(current); |
452 | |
453 | /* Become as one with the init task */ |
454 | |
455 | daemonize_fs_struct(); |
456 | exit_files(current); |
457 | current->files = init_task.files; |
458 | atomic_inc(¤t->files->count); |
459 | |
460 | reparent_to_kthreadd(); |
461 | } |
462 | |
463 | EXPORT_SYMBOL(daemonize); |
464 | |
465 | static void close_files(struct files_struct * files) |
466 | { |
467 | int i, j; |
468 | struct fdtable *fdt; |
469 | |
470 | j = 0; |
471 | |
472 | /* |
473 | * It is safe to dereference the fd table without RCU or |
474 | * ->file_lock because this is the last reference to the |
475 | * files structure. |
476 | */ |
477 | fdt = files_fdtable(files); |
478 | for (;;) { |
479 | unsigned long set; |
480 | i = j * __NFDBITS; |
481 | if (i >= fdt->max_fds) |
482 | break; |
483 | set = fdt->open_fds->fds_bits[j++]; |
484 | while (set) { |
485 | if (set & 1) { |
486 | struct file * file = xchg(&fdt->fd[i], NULL); |
487 | if (file) { |
488 | filp_close(file, files); |
489 | cond_resched(); |
490 | } |
491 | } |
492 | i++; |
493 | set >>= 1; |
494 | } |
495 | } |
496 | } |
497 | |
498 | struct files_struct *get_files_struct(struct task_struct *task) |
499 | { |
500 | struct files_struct *files; |
501 | |
502 | task_lock(task); |
503 | files = task->files; |
504 | if (files) |
505 | atomic_inc(&files->count); |
506 | task_unlock(task); |
507 | |
508 | return files; |
509 | } |
510 | EXPORT_SYMBOL_GPL(get_files_struct); |
511 | |
512 | void put_files_struct(struct files_struct *files) |
513 | { |
514 | struct fdtable *fdt; |
515 | |
516 | if (atomic_dec_and_test(&files->count)) { |
517 | close_files(files); |
518 | /* |
519 | * Free the fd and fdset arrays if we expanded them. |
520 | * If the fdtable was embedded, pass files for freeing |
521 | * at the end of the RCU grace period. Otherwise, |
522 | * you can free files immediately. |
523 | */ |
524 | fdt = files_fdtable(files); |
525 | if (fdt != &files->fdtab) |
526 | kmem_cache_free(files_cachep, files); |
527 | free_fdtable(fdt); |
528 | } |
529 | } |
530 | EXPORT_SYMBOL_GPL(put_files_struct); |
531 | |
532 | void reset_files_struct(struct files_struct *files) |
533 | { |
534 | struct task_struct *tsk = current; |
535 | struct files_struct *old; |
536 | |
537 | old = tsk->files; |
538 | task_lock(tsk); |
539 | tsk->files = files; |
540 | task_unlock(tsk); |
541 | put_files_struct(old); |
542 | } |
543 | |
544 | void exit_files(struct task_struct *tsk) |
545 | { |
546 | struct files_struct * files = tsk->files; |
547 | |
548 | if (files) { |
549 | task_lock(tsk); |
550 | tsk->files = NULL; |
551 | task_unlock(tsk); |
552 | put_files_struct(files); |
553 | } |
554 | } |
555 | |
556 | #ifdef CONFIG_MM_OWNER |
557 | /* |
558 | * Task p is exiting and it owned mm, lets find a new owner for it |
559 | */ |
560 | static inline int |
561 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) |
562 | { |
563 | /* |
564 | * If there are other users of the mm and the owner (us) is exiting |
565 | * we need to find a new owner to take on the responsibility. |
566 | */ |
567 | if (atomic_read(&mm->mm_users) <= 1) |
568 | return 0; |
569 | if (mm->owner != p) |
570 | return 0; |
571 | return 1; |
572 | } |
573 | |
574 | void mm_update_next_owner(struct mm_struct *mm) |
575 | { |
576 | struct task_struct *c, *g, *p = current; |
577 | |
578 | retry: |
579 | if (!mm_need_new_owner(mm, p)) |
580 | return; |
581 | |
582 | read_lock(&tasklist_lock); |
583 | /* |
584 | * Search in the children |
585 | */ |
586 | list_for_each_entry(c, &p->children, sibling) { |
587 | if (c->mm == mm) |
588 | goto assign_new_owner; |
589 | } |
590 | |
591 | /* |
592 | * Search in the siblings |
593 | */ |
594 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
595 | if (c->mm == mm) |
596 | goto assign_new_owner; |
597 | } |
598 | |
599 | /* |
600 | * Search through everything else. We should not get |
601 | * here often |
602 | */ |
603 | do_each_thread(g, c) { |
604 | if (c->mm == mm) |
605 | goto assign_new_owner; |
606 | } while_each_thread(g, c); |
607 | |
608 | read_unlock(&tasklist_lock); |
609 | /* |
610 | * We found no owner yet mm_users > 1: this implies that we are |
611 | * most likely racing with swapoff (try_to_unuse()) or /proc or |
612 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
613 | */ |
614 | mm->owner = NULL; |
615 | return; |
616 | |
617 | assign_new_owner: |
618 | BUG_ON(c == p); |
619 | get_task_struct(c); |
620 | /* |
621 | * The task_lock protects c->mm from changing. |
622 | * We always want mm->owner->mm == mm |
623 | */ |
624 | task_lock(c); |
625 | /* |
626 | * Delay read_unlock() till we have the task_lock() |
627 | * to ensure that c does not slip away underneath us |
628 | */ |
629 | read_unlock(&tasklist_lock); |
630 | if (c->mm != mm) { |
631 | task_unlock(c); |
632 | put_task_struct(c); |
633 | goto retry; |
634 | } |
635 | mm->owner = c; |
636 | task_unlock(c); |
637 | put_task_struct(c); |
638 | } |
639 | #endif /* CONFIG_MM_OWNER */ |
640 | |
641 | /* |
642 | * Turn us into a lazy TLB process if we |
643 | * aren't already.. |
644 | */ |
645 | static void exit_mm(struct task_struct * tsk) |
646 | { |
647 | struct mm_struct *mm = tsk->mm; |
648 | struct core_state *core_state; |
649 | |
650 | mm_release(tsk, mm); |
651 | if (!mm) |
652 | return; |
653 | /* |
654 | * Serialize with any possible pending coredump. |
655 | * We must hold mmap_sem around checking core_state |
656 | * and clearing tsk->mm. The core-inducing thread |
657 | * will increment ->nr_threads for each thread in the |
658 | * group with ->mm != NULL. |
659 | */ |
660 | down_read(&mm->mmap_sem); |
661 | core_state = mm->core_state; |
662 | if (core_state) { |
663 | struct core_thread self; |
664 | up_read(&mm->mmap_sem); |
665 | |
666 | self.task = tsk; |
667 | self.next = xchg(&core_state->dumper.next, &self); |
668 | /* |
669 | * Implies mb(), the result of xchg() must be visible |
670 | * to core_state->dumper. |
671 | */ |
672 | if (atomic_dec_and_test(&core_state->nr_threads)) |
673 | complete(&core_state->startup); |
674 | |
675 | for (;;) { |
676 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
677 | if (!self.task) /* see coredump_finish() */ |
678 | break; |
679 | schedule(); |
680 | } |
681 | __set_task_state(tsk, TASK_RUNNING); |
682 | down_read(&mm->mmap_sem); |
683 | } |
684 | atomic_inc(&mm->mm_count); |
685 | BUG_ON(mm != tsk->active_mm); |
686 | /* more a memory barrier than a real lock */ |
687 | task_lock(tsk); |
688 | tsk->mm = NULL; |
689 | up_read(&mm->mmap_sem); |
690 | enter_lazy_tlb(mm, current); |
691 | /* We don't want this task to be frozen prematurely */ |
692 | clear_freeze_flag(tsk); |
693 | task_unlock(tsk); |
694 | mm_update_next_owner(mm); |
695 | mmput(mm); |
696 | } |
697 | |
698 | /* |
699 | * When we die, we re-parent all our children. |
700 | * Try to give them to another thread in our thread |
701 | * group, and if no such member exists, give it to |
702 | * the child reaper process (ie "init") in our pid |
703 | * space. |
704 | */ |
705 | static struct task_struct *find_new_reaper(struct task_struct *father) |
706 | { |
707 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
708 | struct task_struct *thread; |
709 | |
710 | thread = father; |
711 | while_each_thread(father, thread) { |
712 | if (thread->flags & PF_EXITING) |
713 | continue; |
714 | if (unlikely(pid_ns->child_reaper == father)) |
715 | pid_ns->child_reaper = thread; |
716 | return thread; |
717 | } |
718 | |
719 | if (unlikely(pid_ns->child_reaper == father)) { |
720 | write_unlock_irq(&tasklist_lock); |
721 | if (unlikely(pid_ns == &init_pid_ns)) |
722 | panic("Attempted to kill init!"); |
723 | |
724 | zap_pid_ns_processes(pid_ns); |
725 | write_lock_irq(&tasklist_lock); |
726 | /* |
727 | * We can not clear ->child_reaper or leave it alone. |
728 | * There may by stealth EXIT_DEAD tasks on ->children, |
729 | * forget_original_parent() must move them somewhere. |
730 | */ |
731 | pid_ns->child_reaper = init_pid_ns.child_reaper; |
732 | } |
733 | |
734 | return pid_ns->child_reaper; |
735 | } |
736 | |
737 | /* |
738 | * Any that need to be release_task'd are put on the @dead list. |
739 | */ |
740 | static void reparent_thread(struct task_struct *father, struct task_struct *p, |
741 | struct list_head *dead) |
742 | { |
743 | if (p->pdeath_signal) |
744 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); |
745 | |
746 | list_move_tail(&p->sibling, &p->real_parent->children); |
747 | |
748 | if (task_detached(p)) |
749 | return; |
750 | /* |
751 | * If this is a threaded reparent there is no need to |
752 | * notify anyone anything has happened. |
753 | */ |
754 | if (same_thread_group(p->real_parent, father)) |
755 | return; |
756 | |
757 | /* We don't want people slaying init. */ |
758 | p->exit_signal = SIGCHLD; |
759 | |
760 | /* If it has exited notify the new parent about this child's death. */ |
761 | if (!task_ptrace(p) && |
762 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
763 | do_notify_parent(p, p->exit_signal); |
764 | if (task_detached(p)) { |
765 | p->exit_state = EXIT_DEAD; |
766 | list_move_tail(&p->sibling, dead); |
767 | } |
768 | } |
769 | |
770 | kill_orphaned_pgrp(p, father); |
771 | } |
772 | |
773 | static void forget_original_parent(struct task_struct *father) |
774 | { |
775 | struct task_struct *p, *n, *reaper; |
776 | LIST_HEAD(dead_children); |
777 | |
778 | exit_ptrace(father); |
779 | |
780 | write_lock_irq(&tasklist_lock); |
781 | reaper = find_new_reaper(father); |
782 | |
783 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
784 | p->real_parent = reaper; |
785 | if (p->parent == father) { |
786 | BUG_ON(task_ptrace(p)); |
787 | p->parent = p->real_parent; |
788 | } |
789 | reparent_thread(father, p, &dead_children); |
790 | } |
791 | write_unlock_irq(&tasklist_lock); |
792 | |
793 | BUG_ON(!list_empty(&father->children)); |
794 | |
795 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
796 | list_del_init(&p->sibling); |
797 | release_task(p); |
798 | } |
799 | } |
800 | |
801 | /* |
802 | * Send signals to all our closest relatives so that they know |
803 | * to properly mourn us.. |
804 | */ |
805 | static void exit_notify(struct task_struct *tsk, int group_dead) |
806 | { |
807 | int signal; |
808 | void *cookie; |
809 | |
810 | /* |
811 | * This does two things: |
812 | * |
813 | * A. Make init inherit all the child processes |
814 | * B. Check to see if any process groups have become orphaned |
815 | * as a result of our exiting, and if they have any stopped |
816 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
817 | */ |
818 | forget_original_parent(tsk); |
819 | exit_task_namespaces(tsk); |
820 | |
821 | write_lock_irq(&tasklist_lock); |
822 | if (group_dead) |
823 | kill_orphaned_pgrp(tsk->group_leader, NULL); |
824 | |
825 | /* Let father know we died |
826 | * |
827 | * Thread signals are configurable, but you aren't going to use |
828 | * that to send signals to arbitary processes. |
829 | * That stops right now. |
830 | * |
831 | * If the parent exec id doesn't match the exec id we saved |
832 | * when we started then we know the parent has changed security |
833 | * domain. |
834 | * |
835 | * If our self_exec id doesn't match our parent_exec_id then |
836 | * we have changed execution domain as these two values started |
837 | * the same after a fork. |
838 | */ |
839 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
840 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
841 | tsk->self_exec_id != tsk->parent_exec_id)) |
842 | tsk->exit_signal = SIGCHLD; |
843 | |
844 | signal = tracehook_notify_death(tsk, &cookie, group_dead); |
845 | if (signal >= 0) |
846 | signal = do_notify_parent(tsk, signal); |
847 | |
848 | tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE; |
849 | |
850 | /* mt-exec, de_thread() is waiting for us */ |
851 | if (thread_group_leader(tsk) && |
852 | tsk->signal->group_exit_task && |
853 | tsk->signal->notify_count < 0) |
854 | wake_up_process(tsk->signal->group_exit_task); |
855 | |
856 | write_unlock_irq(&tasklist_lock); |
857 | |
858 | tracehook_report_death(tsk, signal, cookie, group_dead); |
859 | |
860 | /* If the process is dead, release it - nobody will wait for it */ |
861 | if (signal == DEATH_REAP) |
862 | release_task(tsk); |
863 | } |
864 | |
865 | #ifdef CONFIG_DEBUG_STACK_USAGE |
866 | static void check_stack_usage(void) |
867 | { |
868 | static DEFINE_SPINLOCK(low_water_lock); |
869 | static int lowest_to_date = THREAD_SIZE; |
870 | unsigned long free; |
871 | |
872 | free = stack_not_used(current); |
873 | |
874 | if (free >= lowest_to_date) |
875 | return; |
876 | |
877 | spin_lock(&low_water_lock); |
878 | if (free < lowest_to_date) { |
879 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " |
880 | "left\n", |
881 | current->comm, free); |
882 | lowest_to_date = free; |
883 | } |
884 | spin_unlock(&low_water_lock); |
885 | } |
886 | #else |
887 | static inline void check_stack_usage(void) {} |
888 | #endif |
889 | |
890 | NORET_TYPE void do_exit(long code) |
891 | { |
892 | struct task_struct *tsk = current; |
893 | int group_dead; |
894 | |
895 | profile_task_exit(tsk); |
896 | |
897 | WARN_ON(atomic_read(&tsk->fs_excl)); |
898 | |
899 | if (unlikely(in_interrupt())) |
900 | panic("Aiee, killing interrupt handler!"); |
901 | if (unlikely(!tsk->pid)) |
902 | panic("Attempted to kill the idle task!"); |
903 | |
904 | tracehook_report_exit(&code); |
905 | |
906 | /* |
907 | * We're taking recursive faults here in do_exit. Safest is to just |
908 | * leave this task alone and wait for reboot. |
909 | */ |
910 | if (unlikely(tsk->flags & PF_EXITING)) { |
911 | printk(KERN_ALERT |
912 | "Fixing recursive fault but reboot is needed!\n"); |
913 | /* |
914 | * We can do this unlocked here. The futex code uses |
915 | * this flag just to verify whether the pi state |
916 | * cleanup has been done or not. In the worst case it |
917 | * loops once more. We pretend that the cleanup was |
918 | * done as there is no way to return. Either the |
919 | * OWNER_DIED bit is set by now or we push the blocked |
920 | * task into the wait for ever nirwana as well. |
921 | */ |
922 | tsk->flags |= PF_EXITPIDONE; |
923 | set_current_state(TASK_UNINTERRUPTIBLE); |
924 | schedule(); |
925 | } |
926 | |
927 | exit_irq_thread(); |
928 | |
929 | exit_signals(tsk); /* sets PF_EXITING */ |
930 | /* |
931 | * tsk->flags are checked in the futex code to protect against |
932 | * an exiting task cleaning up the robust pi futexes. |
933 | */ |
934 | smp_mb(); |
935 | spin_unlock_wait(&tsk->pi_lock); |
936 | |
937 | if (unlikely(in_atomic())) |
938 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", |
939 | current->comm, task_pid_nr(current), |
940 | preempt_count()); |
941 | |
942 | acct_update_integrals(tsk); |
943 | |
944 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
945 | if (group_dead) { |
946 | hrtimer_cancel(&tsk->signal->real_timer); |
947 | exit_itimers(tsk->signal); |
948 | } |
949 | acct_collect(code, group_dead); |
950 | if (group_dead) |
951 | tty_audit_exit(); |
952 | if (unlikely(tsk->audit_context)) |
953 | audit_free(tsk); |
954 | |
955 | tsk->exit_code = code; |
956 | taskstats_exit(tsk, group_dead); |
957 | |
958 | exit_mm(tsk); |
959 | |
960 | if (group_dead) |
961 | acct_process(); |
962 | trace_sched_process_exit(tsk); |
963 | |
964 | exit_sem(tsk); |
965 | exit_files(tsk); |
966 | exit_fs(tsk); |
967 | check_stack_usage(); |
968 | exit_thread(); |
969 | cgroup_exit(tsk, 1); |
970 | |
971 | if (group_dead && tsk->signal->leader) |
972 | disassociate_ctty(1); |
973 | |
974 | module_put(task_thread_info(tsk)->exec_domain->module); |
975 | if (tsk->binfmt) |
976 | module_put(tsk->binfmt->module); |
977 | |
978 | proc_exit_connector(tsk); |
979 | |
980 | /* |
981 | * Flush inherited counters to the parent - before the parent |
982 | * gets woken up by child-exit notifications. |
983 | */ |
984 | perf_counter_exit_task(tsk); |
985 | |
986 | exit_notify(tsk, group_dead); |
987 | #ifdef CONFIG_NUMA |
988 | mpol_put(tsk->mempolicy); |
989 | tsk->mempolicy = NULL; |
990 | #endif |
991 | #ifdef CONFIG_FUTEX |
992 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
993 | exit_pi_state_list(tsk); |
994 | if (unlikely(current->pi_state_cache)) |
995 | kfree(current->pi_state_cache); |
996 | #endif |
997 | /* |
998 | * Make sure we are holding no locks: |
999 | */ |
1000 | debug_check_no_locks_held(tsk); |
1001 | /* |
1002 | * We can do this unlocked here. The futex code uses this flag |
1003 | * just to verify whether the pi state cleanup has been done |
1004 | * or not. In the worst case it loops once more. |
1005 | */ |
1006 | tsk->flags |= PF_EXITPIDONE; |
1007 | |
1008 | if (tsk->io_context) |
1009 | exit_io_context(); |
1010 | |
1011 | if (tsk->splice_pipe) |
1012 | __free_pipe_info(tsk->splice_pipe); |
1013 | |
1014 | preempt_disable(); |
1015 | /* causes final put_task_struct in finish_task_switch(). */ |
1016 | tsk->state = TASK_DEAD; |
1017 | schedule(); |
1018 | BUG(); |
1019 | /* Avoid "noreturn function does return". */ |
1020 | for (;;) |
1021 | cpu_relax(); /* For when BUG is null */ |
1022 | } |
1023 | |
1024 | EXPORT_SYMBOL_GPL(do_exit); |
1025 | |
1026 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1027 | { |
1028 | if (comp) |
1029 | complete(comp); |
1030 | |
1031 | do_exit(code); |
1032 | } |
1033 | |
1034 | EXPORT_SYMBOL(complete_and_exit); |
1035 | |
1036 | SYSCALL_DEFINE1(exit, int, error_code) |
1037 | { |
1038 | do_exit((error_code&0xff)<<8); |
1039 | } |
1040 | |
1041 | /* |
1042 | * Take down every thread in the group. This is called by fatal signals |
1043 | * as well as by sys_exit_group (below). |
1044 | */ |
1045 | NORET_TYPE void |
1046 | do_group_exit(int exit_code) |
1047 | { |
1048 | struct signal_struct *sig = current->signal; |
1049 | |
1050 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1051 | |
1052 | if (signal_group_exit(sig)) |
1053 | exit_code = sig->group_exit_code; |
1054 | else if (!thread_group_empty(current)) { |
1055 | struct sighand_struct *const sighand = current->sighand; |
1056 | spin_lock_irq(&sighand->siglock); |
1057 | if (signal_group_exit(sig)) |
1058 | /* Another thread got here before we took the lock. */ |
1059 | exit_code = sig->group_exit_code; |
1060 | else { |
1061 | sig->group_exit_code = exit_code; |
1062 | sig->flags = SIGNAL_GROUP_EXIT; |
1063 | zap_other_threads(current); |
1064 | } |
1065 | spin_unlock_irq(&sighand->siglock); |
1066 | } |
1067 | |
1068 | do_exit(exit_code); |
1069 | /* NOTREACHED */ |
1070 | } |
1071 | |
1072 | /* |
1073 | * this kills every thread in the thread group. Note that any externally |
1074 | * wait4()-ing process will get the correct exit code - even if this |
1075 | * thread is not the thread group leader. |
1076 | */ |
1077 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1078 | { |
1079 | do_group_exit((error_code & 0xff) << 8); |
1080 | /* NOTREACHED */ |
1081 | return 0; |
1082 | } |
1083 | |
1084 | struct wait_opts { |
1085 | enum pid_type wo_type; |
1086 | int wo_flags; |
1087 | struct pid *wo_pid; |
1088 | |
1089 | struct siginfo __user *wo_info; |
1090 | int __user *wo_stat; |
1091 | struct rusage __user *wo_rusage; |
1092 | |
1093 | int notask_error; |
1094 | }; |
1095 | |
1096 | static struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
1097 | { |
1098 | struct pid *pid = NULL; |
1099 | if (type == PIDTYPE_PID) |
1100 | pid = task->pids[type].pid; |
1101 | else if (type < PIDTYPE_MAX) |
1102 | pid = task->group_leader->pids[type].pid; |
1103 | return pid; |
1104 | } |
1105 | |
1106 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
1107 | { |
1108 | int err; |
1109 | |
1110 | if (wo->wo_type < PIDTYPE_MAX) { |
1111 | if (task_pid_type(p, wo->wo_type) != wo->wo_pid) |
1112 | return 0; |
1113 | } |
1114 | |
1115 | /* Wait for all children (clone and not) if __WALL is set; |
1116 | * otherwise, wait for clone children *only* if __WCLONE is |
1117 | * set; otherwise, wait for non-clone children *only*. (Note: |
1118 | * A "clone" child here is one that reports to its parent |
1119 | * using a signal other than SIGCHLD.) */ |
1120 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
1121 | && !(wo->wo_flags & __WALL)) |
1122 | return 0; |
1123 | |
1124 | err = security_task_wait(p); |
1125 | if (err) |
1126 | return err; |
1127 | |
1128 | return 1; |
1129 | } |
1130 | |
1131 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
1132 | pid_t pid, uid_t uid, int why, int status) |
1133 | { |
1134 | struct siginfo __user *infop; |
1135 | int retval = wo->wo_rusage |
1136 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
1137 | |
1138 | put_task_struct(p); |
1139 | infop = wo->wo_info; |
1140 | if (!retval) |
1141 | retval = put_user(SIGCHLD, &infop->si_signo); |
1142 | if (!retval) |
1143 | retval = put_user(0, &infop->si_errno); |
1144 | if (!retval) |
1145 | retval = put_user((short)why, &infop->si_code); |
1146 | if (!retval) |
1147 | retval = put_user(pid, &infop->si_pid); |
1148 | if (!retval) |
1149 | retval = put_user(uid, &infop->si_uid); |
1150 | if (!retval) |
1151 | retval = put_user(status, &infop->si_status); |
1152 | if (!retval) |
1153 | retval = pid; |
1154 | return retval; |
1155 | } |
1156 | |
1157 | /* |
1158 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold |
1159 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
1160 | * the lock and this task is uninteresting. If we return nonzero, we have |
1161 | * released the lock and the system call should return. |
1162 | */ |
1163 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1164 | { |
1165 | unsigned long state; |
1166 | int retval, status, traced; |
1167 | pid_t pid = task_pid_vnr(p); |
1168 | uid_t uid = __task_cred(p)->uid; |
1169 | struct siginfo __user *infop; |
1170 | |
1171 | if (!likely(wo->wo_flags & WEXITED)) |
1172 | return 0; |
1173 | |
1174 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1175 | int exit_code = p->exit_code; |
1176 | int why, status; |
1177 | |
1178 | get_task_struct(p); |
1179 | read_unlock(&tasklist_lock); |
1180 | if ((exit_code & 0x7f) == 0) { |
1181 | why = CLD_EXITED; |
1182 | status = exit_code >> 8; |
1183 | } else { |
1184 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; |
1185 | status = exit_code & 0x7f; |
1186 | } |
1187 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1188 | } |
1189 | |
1190 | /* |
1191 | * Try to move the task's state to DEAD |
1192 | * only one thread is allowed to do this: |
1193 | */ |
1194 | state = xchg(&p->exit_state, EXIT_DEAD); |
1195 | if (state != EXIT_ZOMBIE) { |
1196 | BUG_ON(state != EXIT_DEAD); |
1197 | return 0; |
1198 | } |
1199 | |
1200 | traced = ptrace_reparented(p); |
1201 | /* |
1202 | * It can be ptraced but not reparented, check |
1203 | * !task_detached() to filter out sub-threads. |
1204 | */ |
1205 | if (likely(!traced) && likely(!task_detached(p))) { |
1206 | struct signal_struct *psig; |
1207 | struct signal_struct *sig; |
1208 | |
1209 | /* |
1210 | * The resource counters for the group leader are in its |
1211 | * own task_struct. Those for dead threads in the group |
1212 | * are in its signal_struct, as are those for the child |
1213 | * processes it has previously reaped. All these |
1214 | * accumulate in the parent's signal_struct c* fields. |
1215 | * |
1216 | * We don't bother to take a lock here to protect these |
1217 | * p->signal fields, because they are only touched by |
1218 | * __exit_signal, which runs with tasklist_lock |
1219 | * write-locked anyway, and so is excluded here. We do |
1220 | * need to protect the access to parent->signal fields, |
1221 | * as other threads in the parent group can be right |
1222 | * here reaping other children at the same time. |
1223 | */ |
1224 | spin_lock_irq(&p->real_parent->sighand->siglock); |
1225 | psig = p->real_parent->signal; |
1226 | sig = p->signal; |
1227 | psig->cutime = |
1228 | cputime_add(psig->cutime, |
1229 | cputime_add(p->utime, |
1230 | cputime_add(sig->utime, |
1231 | sig->cutime))); |
1232 | psig->cstime = |
1233 | cputime_add(psig->cstime, |
1234 | cputime_add(p->stime, |
1235 | cputime_add(sig->stime, |
1236 | sig->cstime))); |
1237 | psig->cgtime = |
1238 | cputime_add(psig->cgtime, |
1239 | cputime_add(p->gtime, |
1240 | cputime_add(sig->gtime, |
1241 | sig->cgtime))); |
1242 | psig->cmin_flt += |
1243 | p->min_flt + sig->min_flt + sig->cmin_flt; |
1244 | psig->cmaj_flt += |
1245 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; |
1246 | psig->cnvcsw += |
1247 | p->nvcsw + sig->nvcsw + sig->cnvcsw; |
1248 | psig->cnivcsw += |
1249 | p->nivcsw + sig->nivcsw + sig->cnivcsw; |
1250 | psig->cinblock += |
1251 | task_io_get_inblock(p) + |
1252 | sig->inblock + sig->cinblock; |
1253 | psig->coublock += |
1254 | task_io_get_oublock(p) + |
1255 | sig->oublock + sig->coublock; |
1256 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1257 | task_io_accounting_add(&psig->ioac, &sig->ioac); |
1258 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
1259 | } |
1260 | |
1261 | /* |
1262 | * Now we are sure this task is interesting, and no other |
1263 | * thread can reap it because we set its state to EXIT_DEAD. |
1264 | */ |
1265 | read_unlock(&tasklist_lock); |
1266 | |
1267 | retval = wo->wo_rusage |
1268 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
1269 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1270 | ? p->signal->group_exit_code : p->exit_code; |
1271 | if (!retval && wo->wo_stat) |
1272 | retval = put_user(status, wo->wo_stat); |
1273 | |
1274 | infop = wo->wo_info; |
1275 | if (!retval && infop) |
1276 | retval = put_user(SIGCHLD, &infop->si_signo); |
1277 | if (!retval && infop) |
1278 | retval = put_user(0, &infop->si_errno); |
1279 | if (!retval && infop) { |
1280 | int why; |
1281 | |
1282 | if ((status & 0x7f) == 0) { |
1283 | why = CLD_EXITED; |
1284 | status >>= 8; |
1285 | } else { |
1286 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; |
1287 | status &= 0x7f; |
1288 | } |
1289 | retval = put_user((short)why, &infop->si_code); |
1290 | if (!retval) |
1291 | retval = put_user(status, &infop->si_status); |
1292 | } |
1293 | if (!retval && infop) |
1294 | retval = put_user(pid, &infop->si_pid); |
1295 | if (!retval && infop) |
1296 | retval = put_user(uid, &infop->si_uid); |
1297 | if (!retval) |
1298 | retval = pid; |
1299 | |
1300 | if (traced) { |
1301 | write_lock_irq(&tasklist_lock); |
1302 | /* We dropped tasklist, ptracer could die and untrace */ |
1303 | ptrace_unlink(p); |
1304 | /* |
1305 | * If this is not a detached task, notify the parent. |
1306 | * If it's still not detached after that, don't release |
1307 | * it now. |
1308 | */ |
1309 | if (!task_detached(p)) { |
1310 | do_notify_parent(p, p->exit_signal); |
1311 | if (!task_detached(p)) { |
1312 | p->exit_state = EXIT_ZOMBIE; |
1313 | p = NULL; |
1314 | } |
1315 | } |
1316 | write_unlock_irq(&tasklist_lock); |
1317 | } |
1318 | if (p != NULL) |
1319 | release_task(p); |
1320 | |
1321 | return retval; |
1322 | } |
1323 | |
1324 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1325 | { |
1326 | if (ptrace) { |
1327 | if (task_is_stopped_or_traced(p)) |
1328 | return &p->exit_code; |
1329 | } else { |
1330 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
1331 | return &p->signal->group_exit_code; |
1332 | } |
1333 | return NULL; |
1334 | } |
1335 | |
1336 | /* |
1337 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold |
1338 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
1339 | * the lock and this task is uninteresting. If we return nonzero, we have |
1340 | * released the lock and the system call should return. |
1341 | */ |
1342 | static int wait_task_stopped(struct wait_opts *wo, |
1343 | int ptrace, struct task_struct *p) |
1344 | { |
1345 | struct siginfo __user *infop; |
1346 | int retval, exit_code, *p_code, why; |
1347 | uid_t uid = 0; /* unneeded, required by compiler */ |
1348 | pid_t pid; |
1349 | |
1350 | /* |
1351 | * Traditionally we see ptrace'd stopped tasks regardless of options. |
1352 | */ |
1353 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
1354 | return 0; |
1355 | |
1356 | exit_code = 0; |
1357 | spin_lock_irq(&p->sighand->siglock); |
1358 | |
1359 | p_code = task_stopped_code(p, ptrace); |
1360 | if (unlikely(!p_code)) |
1361 | goto unlock_sig; |
1362 | |
1363 | exit_code = *p_code; |
1364 | if (!exit_code) |
1365 | goto unlock_sig; |
1366 | |
1367 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1368 | *p_code = 0; |
1369 | |
1370 | /* don't need the RCU readlock here as we're holding a spinlock */ |
1371 | uid = __task_cred(p)->uid; |
1372 | unlock_sig: |
1373 | spin_unlock_irq(&p->sighand->siglock); |
1374 | if (!exit_code) |
1375 | return 0; |
1376 | |
1377 | /* |
1378 | * Now we are pretty sure this task is interesting. |
1379 | * Make sure it doesn't get reaped out from under us while we |
1380 | * give up the lock and then examine it below. We don't want to |
1381 | * keep holding onto the tasklist_lock while we call getrusage and |
1382 | * possibly take page faults for user memory. |
1383 | */ |
1384 | get_task_struct(p); |
1385 | pid = task_pid_vnr(p); |
1386 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1387 | read_unlock(&tasklist_lock); |
1388 | |
1389 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1390 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); |
1391 | |
1392 | retval = wo->wo_rusage |
1393 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
1394 | if (!retval && wo->wo_stat) |
1395 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); |
1396 | |
1397 | infop = wo->wo_info; |
1398 | if (!retval && infop) |
1399 | retval = put_user(SIGCHLD, &infop->si_signo); |
1400 | if (!retval && infop) |
1401 | retval = put_user(0, &infop->si_errno); |
1402 | if (!retval && infop) |
1403 | retval = put_user((short)why, &infop->si_code); |
1404 | if (!retval && infop) |
1405 | retval = put_user(exit_code, &infop->si_status); |
1406 | if (!retval && infop) |
1407 | retval = put_user(pid, &infop->si_pid); |
1408 | if (!retval && infop) |
1409 | retval = put_user(uid, &infop->si_uid); |
1410 | if (!retval) |
1411 | retval = pid; |
1412 | put_task_struct(p); |
1413 | |
1414 | BUG_ON(!retval); |
1415 | return retval; |
1416 | } |
1417 | |
1418 | /* |
1419 | * Handle do_wait work for one task in a live, non-stopped state. |
1420 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
1421 | * the lock and this task is uninteresting. If we return nonzero, we have |
1422 | * released the lock and the system call should return. |
1423 | */ |
1424 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1425 | { |
1426 | int retval; |
1427 | pid_t pid; |
1428 | uid_t uid; |
1429 | |
1430 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
1431 | return 0; |
1432 | |
1433 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1434 | return 0; |
1435 | |
1436 | spin_lock_irq(&p->sighand->siglock); |
1437 | /* Re-check with the lock held. */ |
1438 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { |
1439 | spin_unlock_irq(&p->sighand->siglock); |
1440 | return 0; |
1441 | } |
1442 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1443 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
1444 | uid = __task_cred(p)->uid; |
1445 | spin_unlock_irq(&p->sighand->siglock); |
1446 | |
1447 | pid = task_pid_vnr(p); |
1448 | get_task_struct(p); |
1449 | read_unlock(&tasklist_lock); |
1450 | |
1451 | if (!wo->wo_info) { |
1452 | retval = wo->wo_rusage |
1453 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
1454 | put_task_struct(p); |
1455 | if (!retval && wo->wo_stat) |
1456 | retval = put_user(0xffff, wo->wo_stat); |
1457 | if (!retval) |
1458 | retval = pid; |
1459 | } else { |
1460 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1461 | CLD_CONTINUED, SIGCONT); |
1462 | BUG_ON(retval == 0); |
1463 | } |
1464 | |
1465 | return retval; |
1466 | } |
1467 | |
1468 | /* |
1469 | * Consider @p for a wait by @parent. |
1470 | * |
1471 | * -ECHILD should be in ->notask_error before the first call. |
1472 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1473 | * Returns zero if the search for a child should continue; |
1474 | * then ->notask_error is 0 if @p is an eligible child, |
1475 | * or another error from security_task_wait(), or still -ECHILD. |
1476 | */ |
1477 | static int wait_consider_task(struct wait_opts *wo, struct task_struct *parent, |
1478 | int ptrace, struct task_struct *p) |
1479 | { |
1480 | int ret = eligible_child(wo, p); |
1481 | if (!ret) |
1482 | return ret; |
1483 | |
1484 | if (unlikely(ret < 0)) { |
1485 | /* |
1486 | * If we have not yet seen any eligible child, |
1487 | * then let this error code replace -ECHILD. |
1488 | * A permission error will give the user a clue |
1489 | * to look for security policy problems, rather |
1490 | * than for mysterious wait bugs. |
1491 | */ |
1492 | if (wo->notask_error) |
1493 | wo->notask_error = ret; |
1494 | return 0; |
1495 | } |
1496 | |
1497 | if (likely(!ptrace) && unlikely(task_ptrace(p))) { |
1498 | /* |
1499 | * This child is hidden by ptrace. |
1500 | * We aren't allowed to see it now, but eventually we will. |
1501 | */ |
1502 | wo->notask_error = 0; |
1503 | return 0; |
1504 | } |
1505 | |
1506 | if (p->exit_state == EXIT_DEAD) |
1507 | return 0; |
1508 | |
1509 | /* |
1510 | * We don't reap group leaders with subthreads. |
1511 | */ |
1512 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) |
1513 | return wait_task_zombie(wo, p); |
1514 | |
1515 | /* |
1516 | * It's stopped or running now, so it might |
1517 | * later continue, exit, or stop again. |
1518 | */ |
1519 | wo->notask_error = 0; |
1520 | |
1521 | if (task_stopped_code(p, ptrace)) |
1522 | return wait_task_stopped(wo, ptrace, p); |
1523 | |
1524 | return wait_task_continued(wo, p); |
1525 | } |
1526 | |
1527 | /* |
1528 | * Do the work of do_wait() for one thread in the group, @tsk. |
1529 | * |
1530 | * -ECHILD should be in ->notask_error before the first call. |
1531 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1532 | * Returns zero if the search for a child should continue; then |
1533 | * ->notask_error is 0 if there were any eligible children, |
1534 | * or another error from security_task_wait(), or still -ECHILD. |
1535 | */ |
1536 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
1537 | { |
1538 | struct task_struct *p; |
1539 | |
1540 | list_for_each_entry(p, &tsk->children, sibling) { |
1541 | /* |
1542 | * Do not consider detached threads. |
1543 | */ |
1544 | if (!task_detached(p)) { |
1545 | int ret = wait_consider_task(wo, tsk, 0, p); |
1546 | if (ret) |
1547 | return ret; |
1548 | } |
1549 | } |
1550 | |
1551 | return 0; |
1552 | } |
1553 | |
1554 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
1555 | { |
1556 | struct task_struct *p; |
1557 | |
1558 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
1559 | int ret = wait_consider_task(wo, tsk, 1, p); |
1560 | if (ret) |
1561 | return ret; |
1562 | } |
1563 | |
1564 | return 0; |
1565 | } |
1566 | |
1567 | static long do_wait(struct wait_opts *wo) |
1568 | { |
1569 | DECLARE_WAITQUEUE(wait, current); |
1570 | struct task_struct *tsk; |
1571 | int retval; |
1572 | |
1573 | trace_sched_process_wait(wo->wo_pid); |
1574 | |
1575 | add_wait_queue(¤t->signal->wait_chldexit,&wait); |
1576 | repeat: |
1577 | /* |
1578 | * If there is nothing that can match our critiera just get out. |
1579 | * We will clear ->notask_error to zero if we see any child that |
1580 | * might later match our criteria, even if we are not able to reap |
1581 | * it yet. |
1582 | */ |
1583 | wo->notask_error = -ECHILD; |
1584 | if ((wo->wo_type < PIDTYPE_MAX) && |
1585 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) |
1586 | goto notask; |
1587 | |
1588 | set_current_state(TASK_INTERRUPTIBLE); |
1589 | read_lock(&tasklist_lock); |
1590 | tsk = current; |
1591 | do { |
1592 | retval = do_wait_thread(wo, tsk); |
1593 | if (retval) |
1594 | goto end; |
1595 | |
1596 | retval = ptrace_do_wait(wo, tsk); |
1597 | if (retval) |
1598 | goto end; |
1599 | |
1600 | if (wo->wo_flags & __WNOTHREAD) |
1601 | break; |
1602 | } while_each_thread(current, tsk); |
1603 | read_unlock(&tasklist_lock); |
1604 | |
1605 | notask: |
1606 | retval = wo->notask_error; |
1607 | if (!retval && !(wo->wo_flags & WNOHANG)) { |
1608 | retval = -ERESTARTSYS; |
1609 | if (!signal_pending(current)) { |
1610 | schedule(); |
1611 | goto repeat; |
1612 | } |
1613 | } |
1614 | end: |
1615 | __set_current_state(TASK_RUNNING); |
1616 | remove_wait_queue(¤t->signal->wait_chldexit,&wait); |
1617 | if (wo->wo_info) { |
1618 | struct siginfo __user *infop = wo->wo_info; |
1619 | |
1620 | if (retval > 0) |
1621 | retval = 0; |
1622 | else { |
1623 | /* |
1624 | * For a WNOHANG return, clear out all the fields |
1625 | * we would set so the user can easily tell the |
1626 | * difference. |
1627 | */ |
1628 | if (!retval) |
1629 | retval = put_user(0, &infop->si_signo); |
1630 | if (!retval) |
1631 | retval = put_user(0, &infop->si_errno); |
1632 | if (!retval) |
1633 | retval = put_user(0, &infop->si_code); |
1634 | if (!retval) |
1635 | retval = put_user(0, &infop->si_pid); |
1636 | if (!retval) |
1637 | retval = put_user(0, &infop->si_uid); |
1638 | if (!retval) |
1639 | retval = put_user(0, &infop->si_status); |
1640 | } |
1641 | } |
1642 | return retval; |
1643 | } |
1644 | |
1645 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1646 | infop, int, options, struct rusage __user *, ru) |
1647 | { |
1648 | struct wait_opts wo; |
1649 | struct pid *pid = NULL; |
1650 | enum pid_type type; |
1651 | long ret; |
1652 | |
1653 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) |
1654 | return -EINVAL; |
1655 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) |
1656 | return -EINVAL; |
1657 | |
1658 | switch (which) { |
1659 | case P_ALL: |
1660 | type = PIDTYPE_MAX; |
1661 | break; |
1662 | case P_PID: |
1663 | type = PIDTYPE_PID; |
1664 | if (upid <= 0) |
1665 | return -EINVAL; |
1666 | break; |
1667 | case P_PGID: |
1668 | type = PIDTYPE_PGID; |
1669 | if (upid <= 0) |
1670 | return -EINVAL; |
1671 | break; |
1672 | default: |
1673 | return -EINVAL; |
1674 | } |
1675 | |
1676 | if (type < PIDTYPE_MAX) |
1677 | pid = find_get_pid(upid); |
1678 | |
1679 | wo.wo_type = type; |
1680 | wo.wo_pid = pid; |
1681 | wo.wo_flags = options; |
1682 | wo.wo_info = infop; |
1683 | wo.wo_stat = NULL; |
1684 | wo.wo_rusage = ru; |
1685 | ret = do_wait(&wo); |
1686 | put_pid(pid); |
1687 | |
1688 | /* avoid REGPARM breakage on x86: */ |
1689 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1690 | return ret; |
1691 | } |
1692 | |
1693 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1694 | int, options, struct rusage __user *, ru) |
1695 | { |
1696 | struct wait_opts wo; |
1697 | struct pid *pid = NULL; |
1698 | enum pid_type type; |
1699 | long ret; |
1700 | |
1701 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| |
1702 | __WNOTHREAD|__WCLONE|__WALL)) |
1703 | return -EINVAL; |
1704 | |
1705 | if (upid == -1) |
1706 | type = PIDTYPE_MAX; |
1707 | else if (upid < 0) { |
1708 | type = PIDTYPE_PGID; |
1709 | pid = find_get_pid(-upid); |
1710 | } else if (upid == 0) { |
1711 | type = PIDTYPE_PGID; |
1712 | pid = get_task_pid(current, PIDTYPE_PGID); |
1713 | } else /* upid > 0 */ { |
1714 | type = PIDTYPE_PID; |
1715 | pid = find_get_pid(upid); |
1716 | } |
1717 | |
1718 | wo.wo_type = type; |
1719 | wo.wo_pid = pid; |
1720 | wo.wo_flags = options | WEXITED; |
1721 | wo.wo_info = NULL; |
1722 | wo.wo_stat = stat_addr; |
1723 | wo.wo_rusage = ru; |
1724 | ret = do_wait(&wo); |
1725 | put_pid(pid); |
1726 | |
1727 | /* avoid REGPARM breakage on x86: */ |
1728 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1729 | return ret; |
1730 | } |
1731 | |
1732 | #ifdef __ARCH_WANT_SYS_WAITPID |
1733 | |
1734 | /* |
1735 | * sys_waitpid() remains for compatibility. waitpid() should be |
1736 | * implemented by calling sys_wait4() from libc.a. |
1737 | */ |
1738 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1739 | { |
1740 | return sys_wait4(pid, stat_addr, options, NULL); |
1741 | } |
1742 | |
1743 | #endif |
1744 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9