]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/exit.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
1da177e4 LT |
7 | #include <linux/mm.h> |
8 | #include <linux/slab.h> | |
9 | #include <linux/interrupt.h> | |
1da177e4 | 10 | #include <linux/module.h> |
c59ede7b | 11 | #include <linux/capability.h> |
1da177e4 LT |
12 | #include <linux/completion.h> |
13 | #include <linux/personality.h> | |
14 | #include <linux/tty.h> | |
da9cbc87 | 15 | #include <linux/iocontext.h> |
1da177e4 | 16 | #include <linux/key.h> |
1da177e4 LT |
17 | #include <linux/cpu.h> |
18 | #include <linux/acct.h> | |
8f0ab514 | 19 | #include <linux/tsacct_kern.h> |
1da177e4 | 20 | #include <linux/file.h> |
9f3acc31 | 21 | #include <linux/fdtable.h> |
80d26af8 | 22 | #include <linux/freezer.h> |
1da177e4 | 23 | #include <linux/binfmts.h> |
ab516013 | 24 | #include <linux/nsproxy.h> |
84d73786 | 25 | #include <linux/pid_namespace.h> |
1da177e4 LT |
26 | #include <linux/ptrace.h> |
27 | #include <linux/profile.h> | |
28 | #include <linux/mount.h> | |
29 | #include <linux/proc_fs.h> | |
49d769d5 | 30 | #include <linux/kthread.h> |
1da177e4 | 31 | #include <linux/mempolicy.h> |
c757249a | 32 | #include <linux/taskstats_kern.h> |
ca74e92b | 33 | #include <linux/delayacct.h> |
b4f48b63 | 34 | #include <linux/cgroup.h> |
1da177e4 | 35 | #include <linux/syscalls.h> |
7ed20e1a | 36 | #include <linux/signal.h> |
6a14c5c9 | 37 | #include <linux/posix-timers.h> |
9f46080c | 38 | #include <linux/cn_proc.h> |
de5097c2 | 39 | #include <linux/mutex.h> |
0771dfef | 40 | #include <linux/futex.h> |
b92ce558 | 41 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 42 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 43 | #include <linux/resource.h> |
0d67a46d | 44 | #include <linux/blkdev.h> |
6eaeeaba | 45 | #include <linux/task_io_accounting_ops.h> |
30199f5a | 46 | #include <linux/tracehook.h> |
5ad4e53b | 47 | #include <linux/fs_struct.h> |
d84f4f99 | 48 | #include <linux/init_task.h> |
cdd6c482 | 49 | #include <linux/perf_event.h> |
ad8d75ff | 50 | #include <trace/events/sched.h> |
24f1e32c | 51 | #include <linux/hw_breakpoint.h> |
3d5992d2 | 52 | #include <linux/oom.h> |
54848d73 | 53 | #include <linux/writeback.h> |
40401530 | 54 | #include <linux/shm.h> |
5c9a8750 | 55 | #include <linux/kcov.h> |
53d3eaa3 | 56 | #include <linux/random.h> |
8f95c90c | 57 | #include <linux/rcuwait.h> |
1da177e4 | 58 | |
7c0f6ba6 | 59 | #include <linux/uaccess.h> |
1da177e4 LT |
60 | #include <asm/unistd.h> |
61 | #include <asm/pgtable.h> | |
62 | #include <asm/mmu_context.h> | |
63 | ||
d40e48e0 | 64 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
65 | { |
66 | nr_threads--; | |
50d75f8d | 67 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 68 | if (group_dead) { |
1da177e4 LT |
69 | detach_pid(p, PIDTYPE_PGID); |
70 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 71 | |
5e85d4ab | 72 | list_del_rcu(&p->tasks); |
9cd80bbb | 73 | list_del_init(&p->sibling); |
909ea964 | 74 | __this_cpu_dec(process_counts); |
1da177e4 | 75 | } |
47e65328 | 76 | list_del_rcu(&p->thread_group); |
0c740d0a | 77 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
78 | } |
79 | ||
6a14c5c9 ON |
80 | /* |
81 | * This function expects the tasklist_lock write-locked. | |
82 | */ | |
83 | static void __exit_signal(struct task_struct *tsk) | |
84 | { | |
85 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 86 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 87 | struct sighand_struct *sighand; |
4ada856f | 88 | struct tty_struct *uninitialized_var(tty); |
5613fda9 | 89 | u64 utime, stime; |
6a14c5c9 | 90 | |
d11c563d | 91 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 92 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
93 | spin_lock(&sighand->siglock); |
94 | ||
baa73d9e | 95 | #ifdef CONFIG_POSIX_TIMERS |
6a14c5c9 | 96 | posix_cpu_timers_exit(tsk); |
d40e48e0 | 97 | if (group_dead) { |
6a14c5c9 | 98 | posix_cpu_timers_exit_group(tsk); |
4a599942 | 99 | } else { |
e0a70217 ON |
100 | /* |
101 | * This can only happen if the caller is de_thread(). | |
102 | * FIXME: this is the temporary hack, we should teach | |
103 | * posix-cpu-timers to handle this case correctly. | |
104 | */ | |
105 | if (unlikely(has_group_leader_pid(tsk))) | |
106 | posix_cpu_timers_exit_group(tsk); | |
baa73d9e NP |
107 | } |
108 | #endif | |
e0a70217 | 109 | |
baa73d9e NP |
110 | if (group_dead) { |
111 | tty = sig->tty; | |
112 | sig->tty = NULL; | |
113 | } else { | |
6a14c5c9 ON |
114 | /* |
115 | * If there is any task waiting for the group exit | |
116 | * then notify it: | |
117 | */ | |
d344193a | 118 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 119 | wake_up_process(sig->group_exit_task); |
6db840fa | 120 | |
6a14c5c9 ON |
121 | if (tsk == sig->curr_target) |
122 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
123 | } |
124 | ||
53d3eaa3 NP |
125 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
126 | sizeof(unsigned long long)); | |
127 | ||
90ed9cbe | 128 | /* |
26e75b5c ON |
129 | * Accumulate here the counters for all threads as they die. We could |
130 | * skip the group leader because it is the last user of signal_struct, | |
131 | * but we want to avoid the race with thread_group_cputime() which can | |
132 | * see the empty ->thread_head list. | |
90ed9cbe RR |
133 | */ |
134 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 135 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
136 | sig->utime += utime; |
137 | sig->stime += stime; | |
138 | sig->gtime += task_gtime(tsk); | |
139 | sig->min_flt += tsk->min_flt; | |
140 | sig->maj_flt += tsk->maj_flt; | |
141 | sig->nvcsw += tsk->nvcsw; | |
142 | sig->nivcsw += tsk->nivcsw; | |
143 | sig->inblock += task_io_get_inblock(tsk); | |
144 | sig->oublock += task_io_get_oublock(tsk); | |
145 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
146 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 147 | sig->nr_threads--; |
d40e48e0 | 148 | __unhash_process(tsk, group_dead); |
e78c3496 | 149 | write_sequnlock(&sig->stats_lock); |
5876700c | 150 | |
da7978b0 ON |
151 | /* |
152 | * Do this under ->siglock, we can race with another thread | |
153 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
154 | */ | |
155 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 156 | tsk->sighand = NULL; |
6a14c5c9 | 157 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 158 | |
a7e5328a | 159 | __cleanup_sighand(sighand); |
a0be55de | 160 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 161 | if (group_dead) { |
6a14c5c9 | 162 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 163 | tty_kref_put(tty); |
6a14c5c9 ON |
164 | } |
165 | } | |
166 | ||
8c7904a0 EB |
167 | static void delayed_put_task_struct(struct rcu_head *rhp) |
168 | { | |
0a16b607 MD |
169 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
170 | ||
4e231c79 | 171 | perf_event_delayed_put(tsk); |
0a16b607 MD |
172 | trace_sched_process_free(tsk); |
173 | put_task_struct(tsk); | |
8c7904a0 EB |
174 | } |
175 | ||
f470021a | 176 | |
a0be55de | 177 | void release_task(struct task_struct *p) |
1da177e4 | 178 | { |
36c8b586 | 179 | struct task_struct *leader; |
1da177e4 | 180 | int zap_leader; |
1f09f974 | 181 | repeat: |
c69e8d9c | 182 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
183 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
184 | rcu_read_lock(); | |
c69e8d9c | 185 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 186 | rcu_read_unlock(); |
c69e8d9c | 187 | |
60347f67 | 188 | proc_flush_task(p); |
0203026b | 189 | |
1da177e4 | 190 | write_lock_irq(&tasklist_lock); |
a288eecc | 191 | ptrace_release_task(p); |
1da177e4 | 192 | __exit_signal(p); |
35f5cad8 | 193 | |
1da177e4 LT |
194 | /* |
195 | * If we are the last non-leader member of the thread | |
196 | * group, and the leader is zombie, then notify the | |
197 | * group leader's parent process. (if it wants notification.) | |
198 | */ | |
199 | zap_leader = 0; | |
200 | leader = p->group_leader; | |
a0be55de IA |
201 | if (leader != p && thread_group_empty(leader) |
202 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
203 | /* |
204 | * If we were the last child thread and the leader has | |
205 | * exited already, and the leader's parent ignores SIGCHLD, | |
206 | * then we are the one who should release the leader. | |
dae33574 | 207 | */ |
86773473 | 208 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
209 | if (zap_leader) |
210 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
211 | } |
212 | ||
1da177e4 | 213 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 214 | release_thread(p); |
8c7904a0 | 215 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
216 | |
217 | p = leader; | |
218 | if (unlikely(zap_leader)) | |
219 | goto repeat; | |
220 | } | |
221 | ||
150593bf ON |
222 | /* |
223 | * Note that if this function returns a valid task_struct pointer (!NULL) | |
224 | * task->usage must remain >0 for the duration of the RCU critical section. | |
225 | */ | |
226 | struct task_struct *task_rcu_dereference(struct task_struct **ptask) | |
227 | { | |
228 | struct sighand_struct *sighand; | |
229 | struct task_struct *task; | |
230 | ||
231 | /* | |
232 | * We need to verify that release_task() was not called and thus | |
233 | * delayed_put_task_struct() can't run and drop the last reference | |
234 | * before rcu_read_unlock(). We check task->sighand != NULL, | |
235 | * but we can read the already freed and reused memory. | |
236 | */ | |
237 | retry: | |
238 | task = rcu_dereference(*ptask); | |
239 | if (!task) | |
240 | return NULL; | |
241 | ||
242 | probe_kernel_address(&task->sighand, sighand); | |
243 | ||
244 | /* | |
245 | * Pairs with atomic_dec_and_test() in put_task_struct(). If this task | |
246 | * was already freed we can not miss the preceding update of this | |
247 | * pointer. | |
248 | */ | |
249 | smp_rmb(); | |
250 | if (unlikely(task != READ_ONCE(*ptask))) | |
251 | goto retry; | |
252 | ||
253 | /* | |
254 | * We've re-checked that "task == *ptask", now we have two different | |
255 | * cases: | |
256 | * | |
257 | * 1. This is actually the same task/task_struct. In this case | |
258 | * sighand != NULL tells us it is still alive. | |
259 | * | |
260 | * 2. This is another task which got the same memory for task_struct. | |
261 | * We can't know this of course, and we can not trust | |
262 | * sighand != NULL. | |
263 | * | |
264 | * In this case we actually return a random value, but this is | |
265 | * correct. | |
266 | * | |
267 | * If we return NULL - we can pretend that we actually noticed that | |
268 | * *ptask was updated when the previous task has exited. Or pretend | |
269 | * that probe_slab_address(&sighand) reads NULL. | |
270 | * | |
271 | * If we return the new task (because sighand is not NULL for any | |
272 | * reason) - this is fine too. This (new) task can't go away before | |
273 | * another gp pass. | |
274 | * | |
275 | * And note: We could even eliminate the false positive if re-read | |
276 | * task->sighand once again to avoid the falsely NULL. But this case | |
277 | * is very unlikely so we don't care. | |
278 | */ | |
279 | if (!sighand) | |
280 | return NULL; | |
281 | ||
282 | return task; | |
283 | } | |
284 | ||
8f95c90c DB |
285 | void rcuwait_wake_up(struct rcuwait *w) |
286 | { | |
287 | struct task_struct *task; | |
288 | ||
289 | rcu_read_lock(); | |
290 | ||
291 | /* | |
292 | * Order condition vs @task, such that everything prior to the load | |
293 | * of @task is visible. This is the condition as to why the user called | |
294 | * rcuwait_trywake() in the first place. Pairs with set_current_state() | |
295 | * barrier (A) in rcuwait_wait_event(). | |
296 | * | |
297 | * WAIT WAKE | |
298 | * [S] tsk = current [S] cond = true | |
299 | * MB (A) MB (B) | |
300 | * [L] cond [L] tsk | |
301 | */ | |
302 | smp_rmb(); /* (B) */ | |
303 | ||
304 | /* | |
305 | * Avoid using task_rcu_dereference() magic as long as we are careful, | |
306 | * see comment in rcuwait_wait_event() regarding ->exit_state. | |
307 | */ | |
308 | task = rcu_dereference(w->task); | |
309 | if (task) | |
310 | wake_up_process(task); | |
311 | rcu_read_unlock(); | |
312 | } | |
313 | ||
150593bf ON |
314 | struct task_struct *try_get_task_struct(struct task_struct **ptask) |
315 | { | |
316 | struct task_struct *task; | |
317 | ||
318 | rcu_read_lock(); | |
319 | task = task_rcu_dereference(ptask); | |
320 | if (task) | |
321 | get_task_struct(task); | |
322 | rcu_read_unlock(); | |
323 | ||
324 | return task; | |
325 | } | |
326 | ||
1da177e4 LT |
327 | /* |
328 | * Determine if a process group is "orphaned", according to the POSIX | |
329 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
330 | * by terminal-generated stop signals. Newly orphaned process groups are | |
331 | * to receive a SIGHUP and a SIGCONT. | |
332 | * | |
333 | * "I ask you, have you ever known what it is to be an orphan?" | |
334 | */ | |
a0be55de IA |
335 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
336 | struct task_struct *ignored_task) | |
1da177e4 LT |
337 | { |
338 | struct task_struct *p; | |
1da177e4 | 339 | |
0475ac08 | 340 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
341 | if ((p == ignored_task) || |
342 | (p->exit_state && thread_group_empty(p)) || | |
343 | is_global_init(p->real_parent)) | |
1da177e4 | 344 | continue; |
05e83df6 | 345 | |
0475ac08 | 346 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
347 | task_session(p->real_parent) == task_session(p)) |
348 | return 0; | |
0475ac08 | 349 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
350 | |
351 | return 1; | |
1da177e4 LT |
352 | } |
353 | ||
3e7cd6c4 | 354 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
355 | { |
356 | int retval; | |
357 | ||
358 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 359 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
360 | read_unlock(&tasklist_lock); |
361 | ||
362 | return retval; | |
363 | } | |
364 | ||
961c4675 | 365 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 366 | { |
1da177e4 LT |
367 | struct task_struct *p; |
368 | ||
0475ac08 | 369 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
370 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
371 | return true; | |
0475ac08 | 372 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
373 | |
374 | return false; | |
1da177e4 LT |
375 | } |
376 | ||
f49ee505 ON |
377 | /* |
378 | * Check to see if any process groups have become orphaned as | |
379 | * a result of our exiting, and if they have any stopped jobs, | |
380 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
381 | */ | |
382 | static void | |
383 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
384 | { | |
385 | struct pid *pgrp = task_pgrp(tsk); | |
386 | struct task_struct *ignored_task = tsk; | |
387 | ||
388 | if (!parent) | |
a0be55de IA |
389 | /* exit: our father is in a different pgrp than |
390 | * we are and we were the only connection outside. | |
391 | */ | |
f49ee505 ON |
392 | parent = tsk->real_parent; |
393 | else | |
394 | /* reparent: our child is in a different pgrp than | |
395 | * we are, and it was the only connection outside. | |
396 | */ | |
397 | ignored_task = NULL; | |
398 | ||
399 | if (task_pgrp(parent) != pgrp && | |
400 | task_session(parent) == task_session(tsk) && | |
401 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
402 | has_stopped_jobs(pgrp)) { | |
403 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
404 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
405 | } | |
406 | } | |
407 | ||
f98bafa0 | 408 | #ifdef CONFIG_MEMCG |
cf475ad2 | 409 | /* |
733eda7a | 410 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 411 | */ |
cf475ad2 BS |
412 | void mm_update_next_owner(struct mm_struct *mm) |
413 | { | |
414 | struct task_struct *c, *g, *p = current; | |
415 | ||
416 | retry: | |
733eda7a KH |
417 | /* |
418 | * If the exiting or execing task is not the owner, it's | |
419 | * someone else's problem. | |
420 | */ | |
421 | if (mm->owner != p) | |
cf475ad2 | 422 | return; |
733eda7a KH |
423 | /* |
424 | * The current owner is exiting/execing and there are no other | |
425 | * candidates. Do not leave the mm pointing to a possibly | |
426 | * freed task structure. | |
427 | */ | |
428 | if (atomic_read(&mm->mm_users) <= 1) { | |
429 | mm->owner = NULL; | |
430 | return; | |
431 | } | |
cf475ad2 BS |
432 | |
433 | read_lock(&tasklist_lock); | |
434 | /* | |
435 | * Search in the children | |
436 | */ | |
437 | list_for_each_entry(c, &p->children, sibling) { | |
438 | if (c->mm == mm) | |
439 | goto assign_new_owner; | |
440 | } | |
441 | ||
442 | /* | |
443 | * Search in the siblings | |
444 | */ | |
dea33cfd | 445 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
446 | if (c->mm == mm) |
447 | goto assign_new_owner; | |
448 | } | |
449 | ||
450 | /* | |
f87fb599 | 451 | * Search through everything else, we should not get here often. |
cf475ad2 | 452 | */ |
39af1765 ON |
453 | for_each_process(g) { |
454 | if (g->flags & PF_KTHREAD) | |
455 | continue; | |
456 | for_each_thread(g, c) { | |
457 | if (c->mm == mm) | |
458 | goto assign_new_owner; | |
459 | if (c->mm) | |
460 | break; | |
461 | } | |
f87fb599 | 462 | } |
cf475ad2 | 463 | read_unlock(&tasklist_lock); |
31a78f23 BS |
464 | /* |
465 | * We found no owner yet mm_users > 1: this implies that we are | |
466 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 467 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 468 | */ |
31a78f23 | 469 | mm->owner = NULL; |
cf475ad2 BS |
470 | return; |
471 | ||
472 | assign_new_owner: | |
473 | BUG_ON(c == p); | |
474 | get_task_struct(c); | |
475 | /* | |
476 | * The task_lock protects c->mm from changing. | |
477 | * We always want mm->owner->mm == mm | |
478 | */ | |
479 | task_lock(c); | |
e5991371 HD |
480 | /* |
481 | * Delay read_unlock() till we have the task_lock() | |
482 | * to ensure that c does not slip away underneath us | |
483 | */ | |
484 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
485 | if (c->mm != mm) { |
486 | task_unlock(c); | |
487 | put_task_struct(c); | |
488 | goto retry; | |
489 | } | |
cf475ad2 BS |
490 | mm->owner = c; |
491 | task_unlock(c); | |
492 | put_task_struct(c); | |
493 | } | |
f98bafa0 | 494 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 495 | |
1da177e4 LT |
496 | /* |
497 | * Turn us into a lazy TLB process if we | |
498 | * aren't already.. | |
499 | */ | |
0039962a | 500 | static void exit_mm(void) |
1da177e4 | 501 | { |
0039962a | 502 | struct mm_struct *mm = current->mm; |
b564daf8 | 503 | struct core_state *core_state; |
1da177e4 | 504 | |
0039962a | 505 | mm_release(current, mm); |
1da177e4 LT |
506 | if (!mm) |
507 | return; | |
4fe7efdb | 508 | sync_mm_rss(mm); |
1da177e4 LT |
509 | /* |
510 | * Serialize with any possible pending coredump. | |
999d9fc1 | 511 | * We must hold mmap_sem around checking core_state |
1da177e4 | 512 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 513 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
514 | * group with ->mm != NULL. |
515 | */ | |
516 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
517 | core_state = mm->core_state; |
518 | if (core_state) { | |
519 | struct core_thread self; | |
a0be55de | 520 | |
1da177e4 | 521 | up_read(&mm->mmap_sem); |
1da177e4 | 522 | |
0039962a | 523 | self.task = current; |
b564daf8 ON |
524 | self.next = xchg(&core_state->dumper.next, &self); |
525 | /* | |
526 | * Implies mb(), the result of xchg() must be visible | |
527 | * to core_state->dumper. | |
528 | */ | |
529 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
530 | complete(&core_state->startup); | |
1da177e4 | 531 | |
a94e2d40 | 532 | for (;;) { |
642fa448 | 533 | set_current_state(TASK_UNINTERRUPTIBLE); |
a94e2d40 ON |
534 | if (!self.task) /* see coredump_finish() */ |
535 | break; | |
80d26af8 | 536 | freezable_schedule(); |
a94e2d40 | 537 | } |
642fa448 | 538 | __set_current_state(TASK_RUNNING); |
1da177e4 LT |
539 | down_read(&mm->mmap_sem); |
540 | } | |
541 | atomic_inc(&mm->mm_count); | |
0039962a | 542 | BUG_ON(mm != current->active_mm); |
1da177e4 | 543 | /* more a memory barrier than a real lock */ |
0039962a DB |
544 | task_lock(current); |
545 | current->mm = NULL; | |
1da177e4 LT |
546 | up_read(&mm->mmap_sem); |
547 | enter_lazy_tlb(mm, current); | |
0039962a | 548 | task_unlock(current); |
cf475ad2 | 549 | mm_update_next_owner(mm); |
1da177e4 | 550 | mmput(mm); |
c32b3cbe | 551 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 552 | exit_oom_victim(); |
1da177e4 LT |
553 | } |
554 | ||
c9dc05bf ON |
555 | static struct task_struct *find_alive_thread(struct task_struct *p) |
556 | { | |
557 | struct task_struct *t; | |
558 | ||
559 | for_each_thread(p, t) { | |
560 | if (!(t->flags & PF_EXITING)) | |
561 | return t; | |
562 | } | |
563 | return NULL; | |
564 | } | |
565 | ||
1109909c ON |
566 | static struct task_struct *find_child_reaper(struct task_struct *father) |
567 | __releases(&tasklist_lock) | |
568 | __acquires(&tasklist_lock) | |
569 | { | |
570 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
571 | struct task_struct *reaper = pid_ns->child_reaper; | |
572 | ||
573 | if (likely(reaper != father)) | |
574 | return reaper; | |
575 | ||
c9dc05bf ON |
576 | reaper = find_alive_thread(father); |
577 | if (reaper) { | |
1109909c ON |
578 | pid_ns->child_reaper = reaper; |
579 | return reaper; | |
580 | } | |
581 | ||
582 | write_unlock_irq(&tasklist_lock); | |
583 | if (unlikely(pid_ns == &init_pid_ns)) { | |
584 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
585 | father->signal->group_exit_code ?: father->exit_code); | |
586 | } | |
587 | zap_pid_ns_processes(pid_ns); | |
588 | write_lock_irq(&tasklist_lock); | |
589 | ||
590 | return father; | |
591 | } | |
592 | ||
1da177e4 | 593 | /* |
ebec18a6 LP |
594 | * When we die, we re-parent all our children, and try to: |
595 | * 1. give them to another thread in our thread group, if such a member exists | |
596 | * 2. give it to the first ancestor process which prctl'd itself as a | |
597 | * child_subreaper for its children (like a service manager) | |
598 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 599 | */ |
1109909c ON |
600 | static struct task_struct *find_new_reaper(struct task_struct *father, |
601 | struct task_struct *child_reaper) | |
1da177e4 | 602 | { |
c9dc05bf | 603 | struct task_struct *thread, *reaper; |
1da177e4 | 604 | |
c9dc05bf ON |
605 | thread = find_alive_thread(father); |
606 | if (thread) | |
950bbabb | 607 | return thread; |
1da177e4 | 608 | |
7d24e2df | 609 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 610 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 611 | /* |
175aed3f | 612 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
613 | * We can't check reaper != child_reaper to ensure we do not |
614 | * cross the namespaces, the exiting parent could be injected | |
615 | * by setns() + fork(). | |
616 | * We check pid->level, this is slightly more efficient than | |
617 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 618 | */ |
c6c70f44 ON |
619 | for (reaper = father->real_parent; |
620 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 621 | reaper = reaper->real_parent) { |
175aed3f | 622 | if (reaper == &init_task) |
ebec18a6 LP |
623 | break; |
624 | if (!reaper->signal->is_child_subreaper) | |
625 | continue; | |
c9dc05bf ON |
626 | thread = find_alive_thread(reaper); |
627 | if (thread) | |
628 | return thread; | |
ebec18a6 | 629 | } |
1da177e4 | 630 | } |
762a24be | 631 | |
1109909c | 632 | return child_reaper; |
950bbabb ON |
633 | } |
634 | ||
5dfc80be ON |
635 | /* |
636 | * Any that need to be release_task'd are put on the @dead list. | |
637 | */ | |
9cd80bbb | 638 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
639 | struct list_head *dead) |
640 | { | |
2831096e | 641 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
642 | return; |
643 | ||
abd50b39 | 644 | /* We don't want people slaying init. */ |
5dfc80be ON |
645 | p->exit_signal = SIGCHLD; |
646 | ||
647 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 648 | if (!p->ptrace && |
5dfc80be | 649 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 650 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 651 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 652 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
653 | } |
654 | } | |
655 | ||
656 | kill_orphaned_pgrp(p, father); | |
657 | } | |
658 | ||
482a3767 ON |
659 | /* |
660 | * This does two things: | |
661 | * | |
662 | * A. Make init inherit all the child processes | |
663 | * B. Check to see if any process groups have become orphaned | |
664 | * as a result of our exiting, and if they have any stopped | |
665 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
666 | */ | |
667 | static void forget_original_parent(struct task_struct *father, | |
668 | struct list_head *dead) | |
1da177e4 | 669 | { |
482a3767 | 670 | struct task_struct *p, *t, *reaper; |
762a24be | 671 | |
7c8bd232 | 672 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 673 | exit_ptrace(father, dead); |
f470021a | 674 | |
7c8bd232 | 675 | /* Can drop and reacquire tasklist_lock */ |
1109909c | 676 | reaper = find_child_reaper(father); |
ad9e206a | 677 | if (list_empty(&father->children)) |
482a3767 | 678 | return; |
1109909c ON |
679 | |
680 | reaper = find_new_reaper(father, reaper); | |
2831096e | 681 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 682 | for_each_thread(p, t) { |
9cd80bbb | 683 | t->real_parent = reaper; |
57a05918 ON |
684 | BUG_ON((!t->ptrace) != (t->parent == father)); |
685 | if (likely(!t->ptrace)) | |
9cd80bbb | 686 | t->parent = t->real_parent; |
9cd80bbb ON |
687 | if (t->pdeath_signal) |
688 | group_send_sig_info(t->pdeath_signal, | |
689 | SEND_SIG_NOINFO, t); | |
57a05918 | 690 | } |
2831096e ON |
691 | /* |
692 | * If this is a threaded reparent there is no need to | |
693 | * notify anyone anything has happened. | |
694 | */ | |
695 | if (!same_thread_group(reaper, father)) | |
482a3767 | 696 | reparent_leader(father, p, dead); |
1da177e4 | 697 | } |
2831096e | 698 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
699 | } |
700 | ||
701 | /* | |
702 | * Send signals to all our closest relatives so that they know | |
703 | * to properly mourn us.. | |
704 | */ | |
821c7de7 | 705 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 706 | { |
53c8f9f1 | 707 | bool autoreap; |
482a3767 ON |
708 | struct task_struct *p, *n; |
709 | LIST_HEAD(dead); | |
1da177e4 | 710 | |
762a24be | 711 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
712 | forget_original_parent(tsk, &dead); |
713 | ||
821c7de7 ON |
714 | if (group_dead) |
715 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 716 | |
45cdf5cc ON |
717 | if (unlikely(tsk->ptrace)) { |
718 | int sig = thread_group_leader(tsk) && | |
719 | thread_group_empty(tsk) && | |
720 | !ptrace_reparented(tsk) ? | |
721 | tsk->exit_signal : SIGCHLD; | |
722 | autoreap = do_notify_parent(tsk, sig); | |
723 | } else if (thread_group_leader(tsk)) { | |
724 | autoreap = thread_group_empty(tsk) && | |
725 | do_notify_parent(tsk, tsk->exit_signal); | |
726 | } else { | |
727 | autoreap = true; | |
728 | } | |
1da177e4 | 729 | |
53c8f9f1 | 730 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
6c66e7db ON |
731 | if (tsk->exit_state == EXIT_DEAD) |
732 | list_add(&tsk->ptrace_entry, &dead); | |
1da177e4 | 733 | |
9c339168 ON |
734 | /* mt-exec, de_thread() is waiting for group leader */ |
735 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 736 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
737 | write_unlock_irq(&tasklist_lock); |
738 | ||
482a3767 ON |
739 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
740 | list_del_init(&p->ptrace_entry); | |
741 | release_task(p); | |
742 | } | |
1da177e4 LT |
743 | } |
744 | ||
e18eecb8 JD |
745 | #ifdef CONFIG_DEBUG_STACK_USAGE |
746 | static void check_stack_usage(void) | |
747 | { | |
748 | static DEFINE_SPINLOCK(low_water_lock); | |
749 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
750 | unsigned long free; |
751 | ||
7c9f8861 | 752 | free = stack_not_used(current); |
e18eecb8 JD |
753 | |
754 | if (free >= lowest_to_date) | |
755 | return; | |
756 | ||
757 | spin_lock(&low_water_lock); | |
758 | if (free < lowest_to_date) { | |
627393d4 | 759 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 760 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
761 | lowest_to_date = free; |
762 | } | |
763 | spin_unlock(&low_water_lock); | |
764 | } | |
765 | #else | |
766 | static inline void check_stack_usage(void) {} | |
767 | #endif | |
768 | ||
9af6528e | 769 | void __noreturn do_exit(long code) |
1da177e4 LT |
770 | { |
771 | struct task_struct *tsk = current; | |
772 | int group_dead; | |
3f95aa81 | 773 | TASKS_RCU(int tasks_rcu_i); |
1da177e4 LT |
774 | |
775 | profile_task_exit(tsk); | |
5c9a8750 | 776 | kcov_task_exit(tsk); |
1da177e4 | 777 | |
73c10101 | 778 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 779 | |
1da177e4 LT |
780 | if (unlikely(in_interrupt())) |
781 | panic("Aiee, killing interrupt handler!"); | |
782 | if (unlikely(!tsk->pid)) | |
783 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 784 | |
33dd94ae NE |
785 | /* |
786 | * If do_exit is called because this processes oopsed, it's possible | |
787 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
788 | * continuing. Amongst other possible reasons, this is to prevent | |
789 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
790 | * kernel address. | |
791 | */ | |
792 | set_fs(USER_DS); | |
793 | ||
a288eecc | 794 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 795 | |
e0e81739 DH |
796 | validate_creds_for_do_exit(tsk); |
797 | ||
df164db5 AN |
798 | /* |
799 | * We're taking recursive faults here in do_exit. Safest is to just | |
800 | * leave this task alone and wait for reboot. | |
801 | */ | |
802 | if (unlikely(tsk->flags & PF_EXITING)) { | |
a0be55de | 803 | pr_alert("Fixing recursive fault but reboot is needed!\n"); |
778e9a9c AK |
804 | /* |
805 | * We can do this unlocked here. The futex code uses | |
806 | * this flag just to verify whether the pi state | |
807 | * cleanup has been done or not. In the worst case it | |
808 | * loops once more. We pretend that the cleanup was | |
809 | * done as there is no way to return. Either the | |
810 | * OWNER_DIED bit is set by now or we push the blocked | |
811 | * task into the wait for ever nirwana as well. | |
812 | */ | |
813 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
814 | set_current_state(TASK_UNINTERRUPTIBLE); |
815 | schedule(); | |
816 | } | |
817 | ||
d12619b5 | 818 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c | 819 | /* |
be3e7844 PZ |
820 | * Ensure that all new tsk->pi_lock acquisitions must observe |
821 | * PF_EXITING. Serializes against futex.c:attach_to_pi_owner(). | |
778e9a9c | 822 | */ |
d2ee7198 | 823 | smp_mb(); |
be3e7844 PZ |
824 | /* |
825 | * Ensure that we must observe the pi_state in exit_mm() -> | |
826 | * mm_release() -> exit_pi_state_list(). | |
827 | */ | |
1d615482 | 828 | raw_spin_unlock_wait(&tsk->pi_lock); |
1da177e4 | 829 | |
1dc0fffc | 830 | if (unlikely(in_atomic())) { |
a0be55de IA |
831 | pr_info("note: %s[%d] exited with preempt_count %d\n", |
832 | current->comm, task_pid_nr(current), | |
833 | preempt_count()); | |
1dc0fffc PZ |
834 | preempt_count_set(PREEMPT_ENABLED); |
835 | } | |
1da177e4 | 836 | |
48d212a2 LT |
837 | /* sync mm's RSS info before statistics gathering */ |
838 | if (tsk->mm) | |
839 | sync_mm_rss(tsk->mm); | |
51229b49 | 840 | acct_update_integrals(tsk); |
1da177e4 | 841 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 842 | if (group_dead) { |
baa73d9e | 843 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 844 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 845 | exit_itimers(tsk->signal); |
baa73d9e | 846 | #endif |
1f10206c JP |
847 | if (tsk->mm) |
848 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 849 | } |
f6ec29a4 | 850 | acct_collect(code, group_dead); |
522ed776 MT |
851 | if (group_dead) |
852 | tty_audit_exit(); | |
a4ff8dba | 853 | audit_free(tsk); |
115085ea | 854 | |
48d212a2 | 855 | tsk->exit_code = code; |
115085ea | 856 | taskstats_exit(tsk, group_dead); |
c757249a | 857 | |
0039962a | 858 | exit_mm(); |
1da177e4 | 859 | |
0e464814 | 860 | if (group_dead) |
f6ec29a4 | 861 | acct_process(); |
0a16b607 MD |
862 | trace_sched_process_exit(tsk); |
863 | ||
1da177e4 | 864 | exit_sem(tsk); |
b34a6b1d | 865 | exit_shm(tsk); |
1ec7f1dd AV |
866 | exit_files(tsk); |
867 | exit_fs(tsk); | |
c39df5fa ON |
868 | if (group_dead) |
869 | disassociate_ctty(1); | |
8aac6270 | 870 | exit_task_namespaces(tsk); |
ed3e694d | 871 | exit_task_work(tsk); |
e6464694 | 872 | exit_thread(tsk); |
0b3fcf17 SE |
873 | |
874 | /* | |
875 | * Flush inherited counters to the parent - before the parent | |
876 | * gets woken up by child-exit notifications. | |
877 | * | |
878 | * because of cgroup mode, must be called before cgroup_exit() | |
879 | */ | |
880 | perf_event_exit_task(tsk); | |
881 | ||
8e5bfa8c | 882 | sched_autogroup_exit_task(tsk); |
1ec41830 | 883 | cgroup_exit(tsk); |
1da177e4 | 884 | |
24f1e32c FW |
885 | /* |
886 | * FIXME: do that only when needed, using sched_exit tracepoint | |
887 | */ | |
7c8df286 | 888 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 889 | |
49f5903b | 890 | TASKS_RCU(preempt_disable()); |
3f95aa81 | 891 | TASKS_RCU(tasks_rcu_i = __srcu_read_lock(&tasks_rcu_exit_srcu)); |
49f5903b | 892 | TASKS_RCU(preempt_enable()); |
821c7de7 | 893 | exit_notify(tsk, group_dead); |
ef982393 | 894 | proc_exit_connector(tsk); |
c11600e4 | 895 | mpol_put_task_policy(tsk); |
42b2dd0a | 896 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
897 | if (unlikely(current->pi_state_cache)) |
898 | kfree(current->pi_state_cache); | |
42b2dd0a | 899 | #endif |
de5097c2 | 900 | /* |
9a11b49a | 901 | * Make sure we are holding no locks: |
de5097c2 | 902 | */ |
1b1d2fb4 | 903 | debug_check_no_locks_held(); |
778e9a9c AK |
904 | /* |
905 | * We can do this unlocked here. The futex code uses this flag | |
906 | * just to verify whether the pi state cleanup has been done | |
907 | * or not. In the worst case it loops once more. | |
908 | */ | |
909 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 910 | |
afc847b7 | 911 | if (tsk->io_context) |
b69f2292 | 912 | exit_io_context(tsk); |
afc847b7 | 913 | |
b92ce558 | 914 | if (tsk->splice_pipe) |
4b8a8f1e | 915 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 916 | |
5640f768 ED |
917 | if (tsk->task_frag.page) |
918 | put_page(tsk->task_frag.page); | |
919 | ||
e0e81739 DH |
920 | validate_creds_for_do_exit(tsk); |
921 | ||
4bcb8232 | 922 | check_stack_usage(); |
7407251a | 923 | preempt_disable(); |
54848d73 WF |
924 | if (tsk->nr_dirtied) |
925 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 926 | exit_rcu(); |
3f95aa81 | 927 | TASKS_RCU(__srcu_read_unlock(&tasks_rcu_exit_srcu, tasks_rcu_i)); |
b5740f4b | 928 | |
9af6528e | 929 | do_task_dead(); |
1da177e4 | 930 | } |
012914da RA |
931 | EXPORT_SYMBOL_GPL(do_exit); |
932 | ||
9402c95f | 933 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
934 | { |
935 | if (comp) | |
936 | complete(comp); | |
55a101f8 | 937 | |
1da177e4 LT |
938 | do_exit(code); |
939 | } | |
1da177e4 LT |
940 | EXPORT_SYMBOL(complete_and_exit); |
941 | ||
754fe8d2 | 942 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
943 | { |
944 | do_exit((error_code&0xff)<<8); | |
945 | } | |
946 | ||
1da177e4 LT |
947 | /* |
948 | * Take down every thread in the group. This is called by fatal signals | |
949 | * as well as by sys_exit_group (below). | |
950 | */ | |
9402c95f | 951 | void |
1da177e4 LT |
952 | do_group_exit(int exit_code) |
953 | { | |
bfc4b089 ON |
954 | struct signal_struct *sig = current->signal; |
955 | ||
1da177e4 LT |
956 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
957 | ||
bfc4b089 ON |
958 | if (signal_group_exit(sig)) |
959 | exit_code = sig->group_exit_code; | |
1da177e4 | 960 | else if (!thread_group_empty(current)) { |
1da177e4 | 961 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 962 | |
1da177e4 | 963 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 964 | if (signal_group_exit(sig)) |
1da177e4 LT |
965 | /* Another thread got here before we took the lock. */ |
966 | exit_code = sig->group_exit_code; | |
967 | else { | |
1da177e4 | 968 | sig->group_exit_code = exit_code; |
ed5d2cac | 969 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
970 | zap_other_threads(current); |
971 | } | |
972 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
973 | } |
974 | ||
975 | do_exit(exit_code); | |
976 | /* NOTREACHED */ | |
977 | } | |
978 | ||
979 | /* | |
980 | * this kills every thread in the thread group. Note that any externally | |
981 | * wait4()-ing process will get the correct exit code - even if this | |
982 | * thread is not the thread group leader. | |
983 | */ | |
754fe8d2 | 984 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
985 | { |
986 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
987 | /* NOTREACHED */ |
988 | return 0; | |
1da177e4 LT |
989 | } |
990 | ||
9e8ae01d ON |
991 | struct wait_opts { |
992 | enum pid_type wo_type; | |
9e8ae01d | 993 | int wo_flags; |
e1eb1ebc | 994 | struct pid *wo_pid; |
9e8ae01d ON |
995 | |
996 | struct siginfo __user *wo_info; | |
997 | int __user *wo_stat; | |
998 | struct rusage __user *wo_rusage; | |
999 | ||
0b7570e7 | 1000 | wait_queue_t child_wait; |
9e8ae01d ON |
1001 | int notask_error; |
1002 | }; | |
1003 | ||
989264f4 ON |
1004 | static inline |
1005 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 1006 | { |
989264f4 ON |
1007 | if (type != PIDTYPE_PID) |
1008 | task = task->group_leader; | |
1009 | return task->pids[type].pid; | |
161550d7 EB |
1010 | } |
1011 | ||
989264f4 | 1012 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1013 | { |
5c01ba49 ON |
1014 | return wo->wo_type == PIDTYPE_MAX || |
1015 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1016 | } | |
1da177e4 | 1017 | |
bf959931 ON |
1018 | static int |
1019 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
1020 | { |
1021 | if (!eligible_pid(wo, p)) | |
1022 | return 0; | |
bf959931 ON |
1023 | |
1024 | /* | |
1025 | * Wait for all children (clone and not) if __WALL is set or | |
1026 | * if it is traced by us. | |
1027 | */ | |
1028 | if (ptrace || (wo->wo_flags & __WALL)) | |
1029 | return 1; | |
1030 | ||
1031 | /* | |
1032 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
1033 | * otherwise, wait for non-clone children *only*. | |
1034 | * | |
1035 | * Note: a "clone" child here is one that reports to its parent | |
1036 | * using a signal other than SIGCHLD, or a non-leader thread which | |
1037 | * we can only see if it is traced by us. | |
1038 | */ | |
1039 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 1040 | return 0; |
1da177e4 | 1041 | |
14dd0b81 | 1042 | return 1; |
1da177e4 LT |
1043 | } |
1044 | ||
9e8ae01d ON |
1045 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
1046 | pid_t pid, uid_t uid, int why, int status) | |
1da177e4 | 1047 | { |
9e8ae01d ON |
1048 | struct siginfo __user *infop; |
1049 | int retval = wo->wo_rusage | |
1050 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
36c8b586 | 1051 | |
1da177e4 | 1052 | put_task_struct(p); |
9e8ae01d | 1053 | infop = wo->wo_info; |
b6fe2d11 VM |
1054 | if (infop) { |
1055 | if (!retval) | |
1056 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1057 | if (!retval) | |
1058 | retval = put_user(0, &infop->si_errno); | |
1059 | if (!retval) | |
1060 | retval = put_user((short)why, &infop->si_code); | |
1061 | if (!retval) | |
1062 | retval = put_user(pid, &infop->si_pid); | |
1063 | if (!retval) | |
1064 | retval = put_user(uid, &infop->si_uid); | |
1065 | if (!retval) | |
1066 | retval = put_user(status, &infop->si_status); | |
1067 | } | |
1da177e4 LT |
1068 | if (!retval) |
1069 | retval = pid; | |
1070 | return retval; | |
1071 | } | |
1072 | ||
1073 | /* | |
1074 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1075 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1076 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1077 | * released the lock and the system call should return. | |
1078 | */ | |
9e8ae01d | 1079 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1080 | { |
f6507f83 | 1081 | int state, retval, status; |
6c5f3e7b | 1082 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1083 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
9e8ae01d | 1084 | struct siginfo __user *infop; |
1da177e4 | 1085 | |
9e8ae01d | 1086 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1087 | return 0; |
1088 | ||
9e8ae01d | 1089 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1da177e4 | 1090 | int exit_code = p->exit_code; |
f3abd4f9 | 1091 | int why; |
1da177e4 | 1092 | |
1da177e4 LT |
1093 | get_task_struct(p); |
1094 | read_unlock(&tasklist_lock); | |
1029a2b5 PZ |
1095 | sched_annotate_sleep(); |
1096 | ||
1da177e4 LT |
1097 | if ((exit_code & 0x7f) == 0) { |
1098 | why = CLD_EXITED; | |
1099 | status = exit_code >> 8; | |
1100 | } else { | |
1101 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1102 | status = exit_code & 0x7f; | |
1103 | } | |
9e8ae01d | 1104 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1da177e4 | 1105 | } |
1da177e4 | 1106 | /* |
abd50b39 | 1107 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1108 | */ |
f6507f83 ON |
1109 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1110 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1111 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1112 | return 0; |
986094df ON |
1113 | /* |
1114 | * We own this thread, nobody else can reap it. | |
1115 | */ | |
1116 | read_unlock(&tasklist_lock); | |
1117 | sched_annotate_sleep(); | |
f6507f83 | 1118 | |
befca967 | 1119 | /* |
f6507f83 | 1120 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1121 | */ |
f6507f83 | 1122 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1123 | struct signal_struct *sig = p->signal; |
1124 | struct signal_struct *psig = current->signal; | |
1f10206c | 1125 | unsigned long maxrss; |
5613fda9 | 1126 | u64 tgutime, tgstime; |
3795e161 | 1127 | |
1da177e4 LT |
1128 | /* |
1129 | * The resource counters for the group leader are in its | |
1130 | * own task_struct. Those for dead threads in the group | |
1131 | * are in its signal_struct, as are those for the child | |
1132 | * processes it has previously reaped. All these | |
1133 | * accumulate in the parent's signal_struct c* fields. | |
1134 | * | |
1135 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1136 | * p->signal fields because the whole thread group is dead |
1137 | * and nobody can change them. | |
1138 | * | |
1139 | * psig->stats_lock also protects us from our sub-theads | |
1140 | * which can reap other children at the same time. Until | |
1141 | * we change k_getrusage()-like users to rely on this lock | |
1142 | * we have to take ->siglock as well. | |
0cf55e1e | 1143 | * |
a0be55de IA |
1144 | * We use thread_group_cputime_adjusted() to get times for |
1145 | * the thread group, which consolidates times for all threads | |
1146 | * in the group including the group leader. | |
1da177e4 | 1147 | */ |
e80d0a1a | 1148 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1149 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1150 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1151 | psig->cutime += tgutime + sig->cutime; |
1152 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1153 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1154 | psig->cmin_flt += |
1155 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1156 | psig->cmaj_flt += | |
1157 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1158 | psig->cnvcsw += | |
1159 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1160 | psig->cnivcsw += | |
1161 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1162 | psig->cinblock += |
1163 | task_io_get_inblock(p) + | |
1164 | sig->inblock + sig->cinblock; | |
1165 | psig->coublock += | |
1166 | task_io_get_oublock(p) + | |
1167 | sig->oublock + sig->coublock; | |
1f10206c JP |
1168 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1169 | if (psig->cmaxrss < maxrss) | |
1170 | psig->cmaxrss = maxrss; | |
5995477a AR |
1171 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1172 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1173 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1174 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1175 | } |
1176 | ||
9e8ae01d ON |
1177 | retval = wo->wo_rusage |
1178 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 LT |
1179 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1180 | ? p->signal->group_exit_code : p->exit_code; | |
9e8ae01d ON |
1181 | if (!retval && wo->wo_stat) |
1182 | retval = put_user(status, wo->wo_stat); | |
1183 | ||
1184 | infop = wo->wo_info; | |
1da177e4 LT |
1185 | if (!retval && infop) |
1186 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1187 | if (!retval && infop) | |
1188 | retval = put_user(0, &infop->si_errno); | |
1189 | if (!retval && infop) { | |
1190 | int why; | |
1191 | ||
1192 | if ((status & 0x7f) == 0) { | |
1193 | why = CLD_EXITED; | |
1194 | status >>= 8; | |
1195 | } else { | |
1196 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1197 | status &= 0x7f; | |
1198 | } | |
1199 | retval = put_user((short)why, &infop->si_code); | |
1200 | if (!retval) | |
1201 | retval = put_user(status, &infop->si_status); | |
1202 | } | |
1203 | if (!retval && infop) | |
3a515e4a | 1204 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1205 | if (!retval && infop) |
c69e8d9c | 1206 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1207 | if (!retval) |
3a515e4a | 1208 | retval = pid; |
2f4e6e2a | 1209 | |
b4360690 | 1210 | if (state == EXIT_TRACE) { |
1da177e4 | 1211 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1212 | /* We dropped tasklist, ptracer could die and untrace */ |
1213 | ptrace_unlink(p); | |
b4360690 ON |
1214 | |
1215 | /* If parent wants a zombie, don't release it now */ | |
1216 | state = EXIT_ZOMBIE; | |
1217 | if (do_notify_parent(p, p->exit_signal)) | |
1218 | state = EXIT_DEAD; | |
abd50b39 | 1219 | p->exit_state = state; |
1da177e4 LT |
1220 | write_unlock_irq(&tasklist_lock); |
1221 | } | |
abd50b39 | 1222 | if (state == EXIT_DEAD) |
1da177e4 | 1223 | release_task(p); |
2f4e6e2a | 1224 | |
1da177e4 LT |
1225 | return retval; |
1226 | } | |
1227 | ||
90bc8d8b ON |
1228 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1229 | { | |
1230 | if (ptrace) { | |
570ac933 | 1231 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1232 | return &p->exit_code; |
1233 | } else { | |
1234 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1235 | return &p->signal->group_exit_code; | |
1236 | } | |
1237 | return NULL; | |
1238 | } | |
1239 | ||
19e27463 TH |
1240 | /** |
1241 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1242 | * @wo: wait options | |
1243 | * @ptrace: is the wait for ptrace | |
1244 | * @p: task to wait for | |
1245 | * | |
1246 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1247 | * | |
1248 | * CONTEXT: | |
1249 | * read_lock(&tasklist_lock), which is released if return value is | |
1250 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1251 | * | |
1252 | * RETURNS: | |
1253 | * 0 if wait condition didn't exist and search for other wait conditions | |
1254 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1255 | * success, implies that tasklist_lock is released and wait condition | |
1256 | * search should terminate. | |
1da177e4 | 1257 | */ |
9e8ae01d ON |
1258 | static int wait_task_stopped(struct wait_opts *wo, |
1259 | int ptrace, struct task_struct *p) | |
1da177e4 | 1260 | { |
9e8ae01d | 1261 | struct siginfo __user *infop; |
90bc8d8b | 1262 | int retval, exit_code, *p_code, why; |
ee7c82da | 1263 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1264 | pid_t pid; |
1da177e4 | 1265 | |
47918025 ON |
1266 | /* |
1267 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1268 | */ | |
9e8ae01d | 1269 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1270 | return 0; |
1271 | ||
19e27463 TH |
1272 | if (!task_stopped_code(p, ptrace)) |
1273 | return 0; | |
1274 | ||
ee7c82da ON |
1275 | exit_code = 0; |
1276 | spin_lock_irq(&p->sighand->siglock); | |
1277 | ||
90bc8d8b ON |
1278 | p_code = task_stopped_code(p, ptrace); |
1279 | if (unlikely(!p_code)) | |
ee7c82da ON |
1280 | goto unlock_sig; |
1281 | ||
90bc8d8b | 1282 | exit_code = *p_code; |
ee7c82da ON |
1283 | if (!exit_code) |
1284 | goto unlock_sig; | |
1285 | ||
9e8ae01d | 1286 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1287 | *p_code = 0; |
ee7c82da | 1288 | |
8ca937a6 | 1289 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1290 | unlock_sig: |
1291 | spin_unlock_irq(&p->sighand->siglock); | |
1292 | if (!exit_code) | |
1da177e4 LT |
1293 | return 0; |
1294 | ||
1295 | /* | |
1296 | * Now we are pretty sure this task is interesting. | |
1297 | * Make sure it doesn't get reaped out from under us while we | |
1298 | * give up the lock and then examine it below. We don't want to | |
1299 | * keep holding onto the tasklist_lock while we call getrusage and | |
1300 | * possibly take page faults for user memory. | |
1301 | */ | |
1302 | get_task_struct(p); | |
6c5f3e7b | 1303 | pid = task_pid_vnr(p); |
f470021a | 1304 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1305 | read_unlock(&tasklist_lock); |
1029a2b5 | 1306 | sched_annotate_sleep(); |
1da177e4 | 1307 | |
9e8ae01d ON |
1308 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1309 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | |
1310 | ||
1311 | retval = wo->wo_rusage | |
1312 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1313 | if (!retval && wo->wo_stat) | |
1314 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | |
1da177e4 | 1315 | |
9e8ae01d | 1316 | infop = wo->wo_info; |
1da177e4 LT |
1317 | if (!retval && infop) |
1318 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1319 | if (!retval && infop) | |
1320 | retval = put_user(0, &infop->si_errno); | |
1321 | if (!retval && infop) | |
6efcae46 | 1322 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1323 | if (!retval && infop) |
1324 | retval = put_user(exit_code, &infop->si_status); | |
1325 | if (!retval && infop) | |
c8950783 | 1326 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1327 | if (!retval && infop) |
ee7c82da | 1328 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1329 | if (!retval) |
c8950783 | 1330 | retval = pid; |
1da177e4 LT |
1331 | put_task_struct(p); |
1332 | ||
1333 | BUG_ON(!retval); | |
1334 | return retval; | |
1335 | } | |
1336 | ||
1337 | /* | |
1338 | * Handle do_wait work for one task in a live, non-stopped state. | |
1339 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1340 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1341 | * released the lock and the system call should return. | |
1342 | */ | |
9e8ae01d | 1343 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1344 | { |
1345 | int retval; | |
1346 | pid_t pid; | |
1347 | uid_t uid; | |
1348 | ||
9e8ae01d | 1349 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1350 | return 0; |
1351 | ||
1da177e4 LT |
1352 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1353 | return 0; | |
1354 | ||
1355 | spin_lock_irq(&p->sighand->siglock); | |
1356 | /* Re-check with the lock held. */ | |
1357 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1358 | spin_unlock_irq(&p->sighand->siglock); | |
1359 | return 0; | |
1360 | } | |
9e8ae01d | 1361 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1362 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1363 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1364 | spin_unlock_irq(&p->sighand->siglock); |
1365 | ||
6c5f3e7b | 1366 | pid = task_pid_vnr(p); |
1da177e4 LT |
1367 | get_task_struct(p); |
1368 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1369 | sched_annotate_sleep(); |
1da177e4 | 1370 | |
9e8ae01d ON |
1371 | if (!wo->wo_info) { |
1372 | retval = wo->wo_rusage | |
1373 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 | 1374 | put_task_struct(p); |
9e8ae01d ON |
1375 | if (!retval && wo->wo_stat) |
1376 | retval = put_user(0xffff, wo->wo_stat); | |
1da177e4 | 1377 | if (!retval) |
3a515e4a | 1378 | retval = pid; |
1da177e4 | 1379 | } else { |
9e8ae01d ON |
1380 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1381 | CLD_CONTINUED, SIGCONT); | |
1da177e4 LT |
1382 | BUG_ON(retval == 0); |
1383 | } | |
1384 | ||
1385 | return retval; | |
1386 | } | |
1387 | ||
98abed02 RM |
1388 | /* |
1389 | * Consider @p for a wait by @parent. | |
1390 | * | |
9e8ae01d | 1391 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1392 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1393 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1394 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1395 | * or still -ECHILD. |
98abed02 | 1396 | */ |
b6e763f0 ON |
1397 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1398 | struct task_struct *p) | |
98abed02 | 1399 | { |
3245d6ac ON |
1400 | /* |
1401 | * We can race with wait_task_zombie() from another thread. | |
1402 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1403 | * can't confuse the checks below. | |
1404 | */ | |
1405 | int exit_state = ACCESS_ONCE(p->exit_state); | |
b3ab0316 ON |
1406 | int ret; |
1407 | ||
3245d6ac | 1408 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1409 | return 0; |
1410 | ||
bf959931 | 1411 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1412 | if (!ret) |
98abed02 RM |
1413 | return ret; |
1414 | ||
3245d6ac | 1415 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1416 | /* |
abd50b39 ON |
1417 | * ptrace == 0 means we are the natural parent. In this case |
1418 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1419 | */ |
abd50b39 | 1420 | if (likely(!ptrace)) |
50b8d257 | 1421 | wo->notask_error = 0; |
823b018e | 1422 | return 0; |
50b8d257 | 1423 | } |
823b018e | 1424 | |
377d75da ON |
1425 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1426 | /* | |
1427 | * If it is traced by its real parent's group, just pretend | |
1428 | * the caller is ptrace_do_wait() and reap this child if it | |
1429 | * is zombie. | |
1430 | * | |
1431 | * This also hides group stop state from real parent; otherwise | |
1432 | * a single stop can be reported twice as group and ptrace stop. | |
1433 | * If a ptracer wants to distinguish these two events for its | |
1434 | * own children it should create a separate process which takes | |
1435 | * the role of real parent. | |
1436 | */ | |
1437 | if (!ptrace_reparented(p)) | |
1438 | ptrace = 1; | |
1439 | } | |
1440 | ||
45cb24a1 | 1441 | /* slay zombie? */ |
3245d6ac | 1442 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1443 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1444 | if (!delay_group_leader(p)) { |
1445 | /* | |
1446 | * A zombie ptracee is only visible to its ptracer. | |
1447 | * Notification and reaping will be cascaded to the | |
1448 | * real parent when the ptracer detaches. | |
1449 | */ | |
1450 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1451 | return wait_task_zombie(wo, p); | |
1452 | } | |
98abed02 | 1453 | |
f470021a | 1454 | /* |
9b84cca2 TH |
1455 | * Allow access to stopped/continued state via zombie by |
1456 | * falling through. Clearing of notask_error is complex. | |
1457 | * | |
1458 | * When !@ptrace: | |
1459 | * | |
1460 | * If WEXITED is set, notask_error should naturally be | |
1461 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1462 | * so, if there are live subthreads, there are events to | |
1463 | * wait for. If all subthreads are dead, it's still safe | |
1464 | * to clear - this function will be called again in finite | |
1465 | * amount time once all the subthreads are released and | |
1466 | * will then return without clearing. | |
1467 | * | |
1468 | * When @ptrace: | |
1469 | * | |
1470 | * Stopped state is per-task and thus can't change once the | |
1471 | * target task dies. Only continued and exited can happen. | |
1472 | * Clear notask_error if WCONTINUED | WEXITED. | |
1473 | */ | |
1474 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1475 | wo->notask_error = 0; | |
1476 | } else { | |
1477 | /* | |
1478 | * @p is alive and it's gonna stop, continue or exit, so | |
1479 | * there always is something to wait for. | |
f470021a | 1480 | */ |
9e8ae01d | 1481 | wo->notask_error = 0; |
f470021a RM |
1482 | } |
1483 | ||
98abed02 | 1484 | /* |
45cb24a1 TH |
1485 | * Wait for stopped. Depending on @ptrace, different stopped state |
1486 | * is used and the two don't interact with each other. | |
98abed02 | 1487 | */ |
19e27463 TH |
1488 | ret = wait_task_stopped(wo, ptrace, p); |
1489 | if (ret) | |
1490 | return ret; | |
98abed02 RM |
1491 | |
1492 | /* | |
45cb24a1 TH |
1493 | * Wait for continued. There's only one continued state and the |
1494 | * ptracer can consume it which can confuse the real parent. Don't | |
1495 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1496 | */ |
9e8ae01d | 1497 | return wait_task_continued(wo, p); |
98abed02 RM |
1498 | } |
1499 | ||
1500 | /* | |
1501 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1502 | * | |
9e8ae01d | 1503 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1504 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1505 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1506 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1507 | * or still -ECHILD. |
98abed02 | 1508 | */ |
9e8ae01d | 1509 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1510 | { |
1511 | struct task_struct *p; | |
1512 | ||
1513 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1514 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1515 | |
9cd80bbb ON |
1516 | if (ret) |
1517 | return ret; | |
98abed02 RM |
1518 | } |
1519 | ||
1520 | return 0; | |
1521 | } | |
1522 | ||
9e8ae01d | 1523 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1524 | { |
1525 | struct task_struct *p; | |
1526 | ||
f470021a | 1527 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1528 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1529 | |
f470021a | 1530 | if (ret) |
98abed02 | 1531 | return ret; |
98abed02 RM |
1532 | } |
1533 | ||
1534 | return 0; | |
1535 | } | |
1536 | ||
0b7570e7 ON |
1537 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1538 | int sync, void *key) | |
1539 | { | |
1540 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1541 | child_wait); | |
1542 | struct task_struct *p = key; | |
1543 | ||
5c01ba49 | 1544 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1545 | return 0; |
1546 | ||
b4fe5182 ON |
1547 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1548 | return 0; | |
1549 | ||
0b7570e7 ON |
1550 | return default_wake_function(wait, mode, sync, key); |
1551 | } | |
1552 | ||
a7f0765e ON |
1553 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1554 | { | |
0b7570e7 ON |
1555 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1556 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1557 | } |
1558 | ||
9e8ae01d | 1559 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1560 | { |
1da177e4 | 1561 | struct task_struct *tsk; |
98abed02 | 1562 | int retval; |
1da177e4 | 1563 | |
9e8ae01d | 1564 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1565 | |
0b7570e7 ON |
1566 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1567 | wo->child_wait.private = current; | |
1568 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1569 | repeat: |
98abed02 | 1570 | /* |
3da56d16 | 1571 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1572 | * We will clear ->notask_error to zero if we see any child that |
1573 | * might later match our criteria, even if we are not able to reap | |
1574 | * it yet. | |
98abed02 | 1575 | */ |
64a16caf | 1576 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1577 | if ((wo->wo_type < PIDTYPE_MAX) && |
1578 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1579 | goto notask; |
161550d7 | 1580 | |
f95d39d1 | 1581 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1582 | read_lock(&tasklist_lock); |
1583 | tsk = current; | |
1584 | do { | |
64a16caf ON |
1585 | retval = do_wait_thread(wo, tsk); |
1586 | if (retval) | |
1587 | goto end; | |
9e8ae01d | 1588 | |
64a16caf ON |
1589 | retval = ptrace_do_wait(wo, tsk); |
1590 | if (retval) | |
98abed02 | 1591 | goto end; |
98abed02 | 1592 | |
9e8ae01d | 1593 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1594 | break; |
a3f6dfb7 | 1595 | } while_each_thread(current, tsk); |
1da177e4 | 1596 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1597 | |
64a16caf | 1598 | notask: |
9e8ae01d ON |
1599 | retval = wo->notask_error; |
1600 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1601 | retval = -ERESTARTSYS; |
98abed02 RM |
1602 | if (!signal_pending(current)) { |
1603 | schedule(); | |
1604 | goto repeat; | |
1605 | } | |
1da177e4 | 1606 | } |
1da177e4 | 1607 | end: |
f95d39d1 | 1608 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1609 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1610 | return retval; |
1611 | } | |
1612 | ||
17da2bd9 HC |
1613 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1614 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1615 | { |
9e8ae01d | 1616 | struct wait_opts wo; |
161550d7 EB |
1617 | struct pid *pid = NULL; |
1618 | enum pid_type type; | |
1da177e4 LT |
1619 | long ret; |
1620 | ||
91c4e8ea ON |
1621 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1622 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1623 | return -EINVAL; |
1624 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1625 | return -EINVAL; | |
1626 | ||
1627 | switch (which) { | |
1628 | case P_ALL: | |
161550d7 | 1629 | type = PIDTYPE_MAX; |
1da177e4 LT |
1630 | break; |
1631 | case P_PID: | |
161550d7 EB |
1632 | type = PIDTYPE_PID; |
1633 | if (upid <= 0) | |
1da177e4 LT |
1634 | return -EINVAL; |
1635 | break; | |
1636 | case P_PGID: | |
161550d7 EB |
1637 | type = PIDTYPE_PGID; |
1638 | if (upid <= 0) | |
1da177e4 | 1639 | return -EINVAL; |
1da177e4 LT |
1640 | break; |
1641 | default: | |
1642 | return -EINVAL; | |
1643 | } | |
1644 | ||
161550d7 EB |
1645 | if (type < PIDTYPE_MAX) |
1646 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1647 | |
1648 | wo.wo_type = type; | |
1649 | wo.wo_pid = pid; | |
1650 | wo.wo_flags = options; | |
1651 | wo.wo_info = infop; | |
1652 | wo.wo_stat = NULL; | |
1653 | wo.wo_rusage = ru; | |
1654 | ret = do_wait(&wo); | |
dfe16dfa VM |
1655 | |
1656 | if (ret > 0) { | |
1657 | ret = 0; | |
1658 | } else if (infop) { | |
1659 | /* | |
1660 | * For a WNOHANG return, clear out all the fields | |
1661 | * we would set so the user can easily tell the | |
1662 | * difference. | |
1663 | */ | |
1664 | if (!ret) | |
1665 | ret = put_user(0, &infop->si_signo); | |
1666 | if (!ret) | |
1667 | ret = put_user(0, &infop->si_errno); | |
1668 | if (!ret) | |
1669 | ret = put_user(0, &infop->si_code); | |
1670 | if (!ret) | |
1671 | ret = put_user(0, &infop->si_pid); | |
1672 | if (!ret) | |
1673 | ret = put_user(0, &infop->si_uid); | |
1674 | if (!ret) | |
1675 | ret = put_user(0, &infop->si_status); | |
1676 | } | |
1677 | ||
161550d7 | 1678 | put_pid(pid); |
1da177e4 LT |
1679 | return ret; |
1680 | } | |
1681 | ||
754fe8d2 HC |
1682 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1683 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1684 | { |
9e8ae01d | 1685 | struct wait_opts wo; |
161550d7 EB |
1686 | struct pid *pid = NULL; |
1687 | enum pid_type type; | |
1da177e4 LT |
1688 | long ret; |
1689 | ||
1690 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1691 | __WNOTHREAD|__WCLONE|__WALL)) | |
1692 | return -EINVAL; | |
161550d7 EB |
1693 | |
1694 | if (upid == -1) | |
1695 | type = PIDTYPE_MAX; | |
1696 | else if (upid < 0) { | |
1697 | type = PIDTYPE_PGID; | |
1698 | pid = find_get_pid(-upid); | |
1699 | } else if (upid == 0) { | |
1700 | type = PIDTYPE_PGID; | |
2ae448ef | 1701 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1702 | } else /* upid > 0 */ { |
1703 | type = PIDTYPE_PID; | |
1704 | pid = find_get_pid(upid); | |
1705 | } | |
1706 | ||
9e8ae01d ON |
1707 | wo.wo_type = type; |
1708 | wo.wo_pid = pid; | |
1709 | wo.wo_flags = options | WEXITED; | |
1710 | wo.wo_info = NULL; | |
1711 | wo.wo_stat = stat_addr; | |
1712 | wo.wo_rusage = ru; | |
1713 | ret = do_wait(&wo); | |
161550d7 | 1714 | put_pid(pid); |
1da177e4 | 1715 | |
1da177e4 LT |
1716 | return ret; |
1717 | } | |
1718 | ||
1719 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1720 | ||
1721 | /* | |
1722 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1723 | * implemented by calling sys_wait4() from libc.a. | |
1724 | */ | |
17da2bd9 | 1725 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1726 | { |
1727 | return sys_wait4(pid, stat_addr, options, NULL); | |
1728 | } | |
1729 | ||
1730 | #endif |