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> |
ca49ca71 | 48 | #include <linux/userfaultfd_k.h> |
d84f4f99 | 49 | #include <linux/init_task.h> |
cdd6c482 | 50 | #include <linux/perf_event.h> |
ad8d75ff | 51 | #include <trace/events/sched.h> |
24f1e32c | 52 | #include <linux/hw_breakpoint.h> |
3d5992d2 | 53 | #include <linux/oom.h> |
54848d73 | 54 | #include <linux/writeback.h> |
40401530 | 55 | #include <linux/shm.h> |
5c9a8750 | 56 | #include <linux/kcov.h> |
53d3eaa3 | 57 | #include <linux/random.h> |
8f95c90c | 58 | #include <linux/rcuwait.h> |
1da177e4 | 59 | |
7c0f6ba6 | 60 | #include <linux/uaccess.h> |
1da177e4 LT |
61 | #include <asm/unistd.h> |
62 | #include <asm/pgtable.h> | |
63 | #include <asm/mmu_context.h> | |
64 | ||
d40e48e0 | 65 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
66 | { |
67 | nr_threads--; | |
50d75f8d | 68 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 69 | if (group_dead) { |
1da177e4 LT |
70 | detach_pid(p, PIDTYPE_PGID); |
71 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 72 | |
5e85d4ab | 73 | list_del_rcu(&p->tasks); |
9cd80bbb | 74 | list_del_init(&p->sibling); |
909ea964 | 75 | __this_cpu_dec(process_counts); |
1da177e4 | 76 | } |
47e65328 | 77 | list_del_rcu(&p->thread_group); |
0c740d0a | 78 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
79 | } |
80 | ||
6a14c5c9 ON |
81 | /* |
82 | * This function expects the tasklist_lock write-locked. | |
83 | */ | |
84 | static void __exit_signal(struct task_struct *tsk) | |
85 | { | |
86 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 87 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 88 | struct sighand_struct *sighand; |
4ada856f | 89 | struct tty_struct *uninitialized_var(tty); |
5613fda9 | 90 | u64 utime, stime; |
6a14c5c9 | 91 | |
d11c563d | 92 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 93 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
94 | spin_lock(&sighand->siglock); |
95 | ||
baa73d9e | 96 | #ifdef CONFIG_POSIX_TIMERS |
6a14c5c9 | 97 | posix_cpu_timers_exit(tsk); |
d40e48e0 | 98 | if (group_dead) { |
6a14c5c9 | 99 | posix_cpu_timers_exit_group(tsk); |
4a599942 | 100 | } else { |
e0a70217 ON |
101 | /* |
102 | * This can only happen if the caller is de_thread(). | |
103 | * FIXME: this is the temporary hack, we should teach | |
104 | * posix-cpu-timers to handle this case correctly. | |
105 | */ | |
106 | if (unlikely(has_group_leader_pid(tsk))) | |
107 | posix_cpu_timers_exit_group(tsk); | |
baa73d9e NP |
108 | } |
109 | #endif | |
e0a70217 | 110 | |
baa73d9e NP |
111 | if (group_dead) { |
112 | tty = sig->tty; | |
113 | sig->tty = NULL; | |
114 | } else { | |
6a14c5c9 ON |
115 | /* |
116 | * If there is any task waiting for the group exit | |
117 | * then notify it: | |
118 | */ | |
d344193a | 119 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 120 | wake_up_process(sig->group_exit_task); |
6db840fa | 121 | |
6a14c5c9 ON |
122 | if (tsk == sig->curr_target) |
123 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
124 | } |
125 | ||
53d3eaa3 NP |
126 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
127 | sizeof(unsigned long long)); | |
128 | ||
90ed9cbe | 129 | /* |
26e75b5c ON |
130 | * Accumulate here the counters for all threads as they die. We could |
131 | * skip the group leader because it is the last user of signal_struct, | |
132 | * but we want to avoid the race with thread_group_cputime() which can | |
133 | * see the empty ->thread_head list. | |
90ed9cbe RR |
134 | */ |
135 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 136 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
137 | sig->utime += utime; |
138 | sig->stime += stime; | |
139 | sig->gtime += task_gtime(tsk); | |
140 | sig->min_flt += tsk->min_flt; | |
141 | sig->maj_flt += tsk->maj_flt; | |
142 | sig->nvcsw += tsk->nvcsw; | |
143 | sig->nivcsw += tsk->nivcsw; | |
144 | sig->inblock += task_io_get_inblock(tsk); | |
145 | sig->oublock += task_io_get_oublock(tsk); | |
146 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
147 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 148 | sig->nr_threads--; |
d40e48e0 | 149 | __unhash_process(tsk, group_dead); |
e78c3496 | 150 | write_sequnlock(&sig->stats_lock); |
5876700c | 151 | |
da7978b0 ON |
152 | /* |
153 | * Do this under ->siglock, we can race with another thread | |
154 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
155 | */ | |
156 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 157 | tsk->sighand = NULL; |
6a14c5c9 | 158 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 159 | |
a7e5328a | 160 | __cleanup_sighand(sighand); |
a0be55de | 161 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 162 | if (group_dead) { |
6a14c5c9 | 163 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 164 | tty_kref_put(tty); |
6a14c5c9 ON |
165 | } |
166 | } | |
167 | ||
8c7904a0 EB |
168 | static void delayed_put_task_struct(struct rcu_head *rhp) |
169 | { | |
0a16b607 MD |
170 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
171 | ||
4e231c79 | 172 | perf_event_delayed_put(tsk); |
0a16b607 MD |
173 | trace_sched_process_free(tsk); |
174 | put_task_struct(tsk); | |
8c7904a0 EB |
175 | } |
176 | ||
f470021a | 177 | |
a0be55de | 178 | void release_task(struct task_struct *p) |
1da177e4 | 179 | { |
36c8b586 | 180 | struct task_struct *leader; |
1da177e4 | 181 | int zap_leader; |
1f09f974 | 182 | repeat: |
c69e8d9c | 183 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
184 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
185 | rcu_read_lock(); | |
c69e8d9c | 186 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 187 | rcu_read_unlock(); |
c69e8d9c | 188 | |
60347f67 | 189 | proc_flush_task(p); |
0203026b | 190 | |
1da177e4 | 191 | write_lock_irq(&tasklist_lock); |
a288eecc | 192 | ptrace_release_task(p); |
1da177e4 | 193 | __exit_signal(p); |
35f5cad8 | 194 | |
1da177e4 LT |
195 | /* |
196 | * If we are the last non-leader member of the thread | |
197 | * group, and the leader is zombie, then notify the | |
198 | * group leader's parent process. (if it wants notification.) | |
199 | */ | |
200 | zap_leader = 0; | |
201 | leader = p->group_leader; | |
a0be55de IA |
202 | if (leader != p && thread_group_empty(leader) |
203 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
204 | /* |
205 | * If we were the last child thread and the leader has | |
206 | * exited already, and the leader's parent ignores SIGCHLD, | |
207 | * then we are the one who should release the leader. | |
dae33574 | 208 | */ |
86773473 | 209 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
210 | if (zap_leader) |
211 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
212 | } |
213 | ||
1da177e4 | 214 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 215 | release_thread(p); |
8c7904a0 | 216 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
217 | |
218 | p = leader; | |
219 | if (unlikely(zap_leader)) | |
220 | goto repeat; | |
221 | } | |
222 | ||
150593bf ON |
223 | /* |
224 | * Note that if this function returns a valid task_struct pointer (!NULL) | |
225 | * task->usage must remain >0 for the duration of the RCU critical section. | |
226 | */ | |
227 | struct task_struct *task_rcu_dereference(struct task_struct **ptask) | |
228 | { | |
229 | struct sighand_struct *sighand; | |
230 | struct task_struct *task; | |
231 | ||
232 | /* | |
233 | * We need to verify that release_task() was not called and thus | |
234 | * delayed_put_task_struct() can't run and drop the last reference | |
235 | * before rcu_read_unlock(). We check task->sighand != NULL, | |
236 | * but we can read the already freed and reused memory. | |
237 | */ | |
238 | retry: | |
239 | task = rcu_dereference(*ptask); | |
240 | if (!task) | |
241 | return NULL; | |
242 | ||
243 | probe_kernel_address(&task->sighand, sighand); | |
244 | ||
245 | /* | |
246 | * Pairs with atomic_dec_and_test() in put_task_struct(). If this task | |
247 | * was already freed we can not miss the preceding update of this | |
248 | * pointer. | |
249 | */ | |
250 | smp_rmb(); | |
251 | if (unlikely(task != READ_ONCE(*ptask))) | |
252 | goto retry; | |
253 | ||
254 | /* | |
255 | * We've re-checked that "task == *ptask", now we have two different | |
256 | * cases: | |
257 | * | |
258 | * 1. This is actually the same task/task_struct. In this case | |
259 | * sighand != NULL tells us it is still alive. | |
260 | * | |
261 | * 2. This is another task which got the same memory for task_struct. | |
262 | * We can't know this of course, and we can not trust | |
263 | * sighand != NULL. | |
264 | * | |
265 | * In this case we actually return a random value, but this is | |
266 | * correct. | |
267 | * | |
268 | * If we return NULL - we can pretend that we actually noticed that | |
269 | * *ptask was updated when the previous task has exited. Or pretend | |
270 | * that probe_slab_address(&sighand) reads NULL. | |
271 | * | |
272 | * If we return the new task (because sighand is not NULL for any | |
273 | * reason) - this is fine too. This (new) task can't go away before | |
274 | * another gp pass. | |
275 | * | |
276 | * And note: We could even eliminate the false positive if re-read | |
277 | * task->sighand once again to avoid the falsely NULL. But this case | |
278 | * is very unlikely so we don't care. | |
279 | */ | |
280 | if (!sighand) | |
281 | return NULL; | |
282 | ||
283 | return task; | |
284 | } | |
285 | ||
8f95c90c DB |
286 | void rcuwait_wake_up(struct rcuwait *w) |
287 | { | |
288 | struct task_struct *task; | |
289 | ||
290 | rcu_read_lock(); | |
291 | ||
292 | /* | |
293 | * Order condition vs @task, such that everything prior to the load | |
294 | * of @task is visible. This is the condition as to why the user called | |
295 | * rcuwait_trywake() in the first place. Pairs with set_current_state() | |
296 | * barrier (A) in rcuwait_wait_event(). | |
297 | * | |
298 | * WAIT WAKE | |
299 | * [S] tsk = current [S] cond = true | |
300 | * MB (A) MB (B) | |
301 | * [L] cond [L] tsk | |
302 | */ | |
303 | smp_rmb(); /* (B) */ | |
304 | ||
305 | /* | |
306 | * Avoid using task_rcu_dereference() magic as long as we are careful, | |
307 | * see comment in rcuwait_wait_event() regarding ->exit_state. | |
308 | */ | |
309 | task = rcu_dereference(w->task); | |
310 | if (task) | |
311 | wake_up_process(task); | |
312 | rcu_read_unlock(); | |
313 | } | |
314 | ||
150593bf ON |
315 | struct task_struct *try_get_task_struct(struct task_struct **ptask) |
316 | { | |
317 | struct task_struct *task; | |
318 | ||
319 | rcu_read_lock(); | |
320 | task = task_rcu_dereference(ptask); | |
321 | if (task) | |
322 | get_task_struct(task); | |
323 | rcu_read_unlock(); | |
324 | ||
325 | return task; | |
326 | } | |
327 | ||
1da177e4 LT |
328 | /* |
329 | * Determine if a process group is "orphaned", according to the POSIX | |
330 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
331 | * by terminal-generated stop signals. Newly orphaned process groups are | |
332 | * to receive a SIGHUP and a SIGCONT. | |
333 | * | |
334 | * "I ask you, have you ever known what it is to be an orphan?" | |
335 | */ | |
a0be55de IA |
336 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
337 | struct task_struct *ignored_task) | |
1da177e4 LT |
338 | { |
339 | struct task_struct *p; | |
1da177e4 | 340 | |
0475ac08 | 341 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
342 | if ((p == ignored_task) || |
343 | (p->exit_state && thread_group_empty(p)) || | |
344 | is_global_init(p->real_parent)) | |
1da177e4 | 345 | continue; |
05e83df6 | 346 | |
0475ac08 | 347 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
348 | task_session(p->real_parent) == task_session(p)) |
349 | return 0; | |
0475ac08 | 350 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
351 | |
352 | return 1; | |
1da177e4 LT |
353 | } |
354 | ||
3e7cd6c4 | 355 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
356 | { |
357 | int retval; | |
358 | ||
359 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 360 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
361 | read_unlock(&tasklist_lock); |
362 | ||
363 | return retval; | |
364 | } | |
365 | ||
961c4675 | 366 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 367 | { |
1da177e4 LT |
368 | struct task_struct *p; |
369 | ||
0475ac08 | 370 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
371 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
372 | return true; | |
0475ac08 | 373 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
374 | |
375 | return false; | |
1da177e4 LT |
376 | } |
377 | ||
f49ee505 ON |
378 | /* |
379 | * Check to see if any process groups have become orphaned as | |
380 | * a result of our exiting, and if they have any stopped jobs, | |
381 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
382 | */ | |
383 | static void | |
384 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
385 | { | |
386 | struct pid *pgrp = task_pgrp(tsk); | |
387 | struct task_struct *ignored_task = tsk; | |
388 | ||
389 | if (!parent) | |
a0be55de IA |
390 | /* exit: our father is in a different pgrp than |
391 | * we are and we were the only connection outside. | |
392 | */ | |
f49ee505 ON |
393 | parent = tsk->real_parent; |
394 | else | |
395 | /* reparent: our child is in a different pgrp than | |
396 | * we are, and it was the only connection outside. | |
397 | */ | |
398 | ignored_task = NULL; | |
399 | ||
400 | if (task_pgrp(parent) != pgrp && | |
401 | task_session(parent) == task_session(tsk) && | |
402 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
403 | has_stopped_jobs(pgrp)) { | |
404 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
405 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
406 | } | |
407 | } | |
408 | ||
f98bafa0 | 409 | #ifdef CONFIG_MEMCG |
cf475ad2 | 410 | /* |
733eda7a | 411 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 412 | */ |
cf475ad2 BS |
413 | void mm_update_next_owner(struct mm_struct *mm) |
414 | { | |
415 | struct task_struct *c, *g, *p = current; | |
416 | ||
417 | retry: | |
733eda7a KH |
418 | /* |
419 | * If the exiting or execing task is not the owner, it's | |
420 | * someone else's problem. | |
421 | */ | |
422 | if (mm->owner != p) | |
cf475ad2 | 423 | return; |
733eda7a KH |
424 | /* |
425 | * The current owner is exiting/execing and there are no other | |
426 | * candidates. Do not leave the mm pointing to a possibly | |
427 | * freed task structure. | |
428 | */ | |
429 | if (atomic_read(&mm->mm_users) <= 1) { | |
430 | mm->owner = NULL; | |
431 | return; | |
432 | } | |
cf475ad2 BS |
433 | |
434 | read_lock(&tasklist_lock); | |
435 | /* | |
436 | * Search in the children | |
437 | */ | |
438 | list_for_each_entry(c, &p->children, sibling) { | |
439 | if (c->mm == mm) | |
440 | goto assign_new_owner; | |
441 | } | |
442 | ||
443 | /* | |
444 | * Search in the siblings | |
445 | */ | |
dea33cfd | 446 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
447 | if (c->mm == mm) |
448 | goto assign_new_owner; | |
449 | } | |
450 | ||
451 | /* | |
f87fb599 | 452 | * Search through everything else, we should not get here often. |
cf475ad2 | 453 | */ |
39af1765 ON |
454 | for_each_process(g) { |
455 | if (g->flags & PF_KTHREAD) | |
456 | continue; | |
457 | for_each_thread(g, c) { | |
458 | if (c->mm == mm) | |
459 | goto assign_new_owner; | |
460 | if (c->mm) | |
461 | break; | |
462 | } | |
f87fb599 | 463 | } |
cf475ad2 | 464 | read_unlock(&tasklist_lock); |
31a78f23 BS |
465 | /* |
466 | * We found no owner yet mm_users > 1: this implies that we are | |
467 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 468 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 469 | */ |
31a78f23 | 470 | mm->owner = NULL; |
cf475ad2 BS |
471 | return; |
472 | ||
473 | assign_new_owner: | |
474 | BUG_ON(c == p); | |
475 | get_task_struct(c); | |
476 | /* | |
477 | * The task_lock protects c->mm from changing. | |
478 | * We always want mm->owner->mm == mm | |
479 | */ | |
480 | task_lock(c); | |
e5991371 HD |
481 | /* |
482 | * Delay read_unlock() till we have the task_lock() | |
483 | * to ensure that c does not slip away underneath us | |
484 | */ | |
485 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
486 | if (c->mm != mm) { |
487 | task_unlock(c); | |
488 | put_task_struct(c); | |
489 | goto retry; | |
490 | } | |
cf475ad2 BS |
491 | mm->owner = c; |
492 | task_unlock(c); | |
493 | put_task_struct(c); | |
494 | } | |
f98bafa0 | 495 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 496 | |
1da177e4 LT |
497 | /* |
498 | * Turn us into a lazy TLB process if we | |
499 | * aren't already.. | |
500 | */ | |
0039962a | 501 | static void exit_mm(void) |
1da177e4 | 502 | { |
0039962a | 503 | struct mm_struct *mm = current->mm; |
b564daf8 | 504 | struct core_state *core_state; |
1da177e4 | 505 | |
0039962a | 506 | mm_release(current, mm); |
1da177e4 LT |
507 | if (!mm) |
508 | return; | |
4fe7efdb | 509 | sync_mm_rss(mm); |
1da177e4 LT |
510 | /* |
511 | * Serialize with any possible pending coredump. | |
999d9fc1 | 512 | * We must hold mmap_sem around checking core_state |
1da177e4 | 513 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 514 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
515 | * group with ->mm != NULL. |
516 | */ | |
517 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
518 | core_state = mm->core_state; |
519 | if (core_state) { | |
520 | struct core_thread self; | |
a0be55de | 521 | |
1da177e4 | 522 | up_read(&mm->mmap_sem); |
1da177e4 | 523 | |
0039962a | 524 | self.task = current; |
b564daf8 ON |
525 | self.next = xchg(&core_state->dumper.next, &self); |
526 | /* | |
527 | * Implies mb(), the result of xchg() must be visible | |
528 | * to core_state->dumper. | |
529 | */ | |
530 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
531 | complete(&core_state->startup); | |
1da177e4 | 532 | |
a94e2d40 | 533 | for (;;) { |
642fa448 | 534 | set_current_state(TASK_UNINTERRUPTIBLE); |
a94e2d40 ON |
535 | if (!self.task) /* see coredump_finish() */ |
536 | break; | |
80d26af8 | 537 | freezable_schedule(); |
a94e2d40 | 538 | } |
642fa448 | 539 | __set_current_state(TASK_RUNNING); |
1da177e4 LT |
540 | down_read(&mm->mmap_sem); |
541 | } | |
542 | atomic_inc(&mm->mm_count); | |
0039962a | 543 | BUG_ON(mm != current->active_mm); |
1da177e4 | 544 | /* more a memory barrier than a real lock */ |
0039962a DB |
545 | task_lock(current); |
546 | current->mm = NULL; | |
1da177e4 LT |
547 | up_read(&mm->mmap_sem); |
548 | enter_lazy_tlb(mm, current); | |
0039962a | 549 | task_unlock(current); |
cf475ad2 | 550 | mm_update_next_owner(mm); |
ca49ca71 | 551 | userfaultfd_exit(mm); |
1da177e4 | 552 | mmput(mm); |
c32b3cbe | 553 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 554 | exit_oom_victim(); |
1da177e4 LT |
555 | } |
556 | ||
c9dc05bf ON |
557 | static struct task_struct *find_alive_thread(struct task_struct *p) |
558 | { | |
559 | struct task_struct *t; | |
560 | ||
561 | for_each_thread(p, t) { | |
562 | if (!(t->flags & PF_EXITING)) | |
563 | return t; | |
564 | } | |
565 | return NULL; | |
566 | } | |
567 | ||
1109909c ON |
568 | static struct task_struct *find_child_reaper(struct task_struct *father) |
569 | __releases(&tasklist_lock) | |
570 | __acquires(&tasklist_lock) | |
571 | { | |
572 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
573 | struct task_struct *reaper = pid_ns->child_reaper; | |
574 | ||
575 | if (likely(reaper != father)) | |
576 | return reaper; | |
577 | ||
c9dc05bf ON |
578 | reaper = find_alive_thread(father); |
579 | if (reaper) { | |
1109909c ON |
580 | pid_ns->child_reaper = reaper; |
581 | return reaper; | |
582 | } | |
583 | ||
584 | write_unlock_irq(&tasklist_lock); | |
585 | if (unlikely(pid_ns == &init_pid_ns)) { | |
586 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
587 | father->signal->group_exit_code ?: father->exit_code); | |
588 | } | |
589 | zap_pid_ns_processes(pid_ns); | |
590 | write_lock_irq(&tasklist_lock); | |
591 | ||
592 | return father; | |
593 | } | |
594 | ||
1da177e4 | 595 | /* |
ebec18a6 LP |
596 | * When we die, we re-parent all our children, and try to: |
597 | * 1. give them to another thread in our thread group, if such a member exists | |
598 | * 2. give it to the first ancestor process which prctl'd itself as a | |
599 | * child_subreaper for its children (like a service manager) | |
600 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 601 | */ |
1109909c ON |
602 | static struct task_struct *find_new_reaper(struct task_struct *father, |
603 | struct task_struct *child_reaper) | |
1da177e4 | 604 | { |
c9dc05bf | 605 | struct task_struct *thread, *reaper; |
1da177e4 | 606 | |
c9dc05bf ON |
607 | thread = find_alive_thread(father); |
608 | if (thread) | |
950bbabb | 609 | return thread; |
1da177e4 | 610 | |
7d24e2df | 611 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 612 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 613 | /* |
175aed3f | 614 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
615 | * We can't check reaper != child_reaper to ensure we do not |
616 | * cross the namespaces, the exiting parent could be injected | |
617 | * by setns() + fork(). | |
618 | * We check pid->level, this is slightly more efficient than | |
619 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 620 | */ |
c6c70f44 ON |
621 | for (reaper = father->real_parent; |
622 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 623 | reaper = reaper->real_parent) { |
175aed3f | 624 | if (reaper == &init_task) |
ebec18a6 LP |
625 | break; |
626 | if (!reaper->signal->is_child_subreaper) | |
627 | continue; | |
c9dc05bf ON |
628 | thread = find_alive_thread(reaper); |
629 | if (thread) | |
630 | return thread; | |
ebec18a6 | 631 | } |
1da177e4 | 632 | } |
762a24be | 633 | |
1109909c | 634 | return child_reaper; |
950bbabb ON |
635 | } |
636 | ||
5dfc80be ON |
637 | /* |
638 | * Any that need to be release_task'd are put on the @dead list. | |
639 | */ | |
9cd80bbb | 640 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
641 | struct list_head *dead) |
642 | { | |
2831096e | 643 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
644 | return; |
645 | ||
abd50b39 | 646 | /* We don't want people slaying init. */ |
5dfc80be ON |
647 | p->exit_signal = SIGCHLD; |
648 | ||
649 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 650 | if (!p->ptrace && |
5dfc80be | 651 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 652 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 653 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 654 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
655 | } |
656 | } | |
657 | ||
658 | kill_orphaned_pgrp(p, father); | |
659 | } | |
660 | ||
482a3767 ON |
661 | /* |
662 | * This does two things: | |
663 | * | |
664 | * A. Make init inherit all the child processes | |
665 | * B. Check to see if any process groups have become orphaned | |
666 | * as a result of our exiting, and if they have any stopped | |
667 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
668 | */ | |
669 | static void forget_original_parent(struct task_struct *father, | |
670 | struct list_head *dead) | |
1da177e4 | 671 | { |
482a3767 | 672 | struct task_struct *p, *t, *reaper; |
762a24be | 673 | |
7c8bd232 | 674 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 675 | exit_ptrace(father, dead); |
f470021a | 676 | |
7c8bd232 | 677 | /* Can drop and reacquire tasklist_lock */ |
1109909c | 678 | reaper = find_child_reaper(father); |
ad9e206a | 679 | if (list_empty(&father->children)) |
482a3767 | 680 | return; |
1109909c ON |
681 | |
682 | reaper = find_new_reaper(father, reaper); | |
2831096e | 683 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 684 | for_each_thread(p, t) { |
9cd80bbb | 685 | t->real_parent = reaper; |
57a05918 ON |
686 | BUG_ON((!t->ptrace) != (t->parent == father)); |
687 | if (likely(!t->ptrace)) | |
9cd80bbb | 688 | t->parent = t->real_parent; |
9cd80bbb ON |
689 | if (t->pdeath_signal) |
690 | group_send_sig_info(t->pdeath_signal, | |
691 | SEND_SIG_NOINFO, t); | |
57a05918 | 692 | } |
2831096e ON |
693 | /* |
694 | * If this is a threaded reparent there is no need to | |
695 | * notify anyone anything has happened. | |
696 | */ | |
697 | if (!same_thread_group(reaper, father)) | |
482a3767 | 698 | reparent_leader(father, p, dead); |
1da177e4 | 699 | } |
2831096e | 700 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
701 | } |
702 | ||
703 | /* | |
704 | * Send signals to all our closest relatives so that they know | |
705 | * to properly mourn us.. | |
706 | */ | |
821c7de7 | 707 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 708 | { |
53c8f9f1 | 709 | bool autoreap; |
482a3767 ON |
710 | struct task_struct *p, *n; |
711 | LIST_HEAD(dead); | |
1da177e4 | 712 | |
762a24be | 713 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
714 | forget_original_parent(tsk, &dead); |
715 | ||
821c7de7 ON |
716 | if (group_dead) |
717 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 718 | |
45cdf5cc ON |
719 | if (unlikely(tsk->ptrace)) { |
720 | int sig = thread_group_leader(tsk) && | |
721 | thread_group_empty(tsk) && | |
722 | !ptrace_reparented(tsk) ? | |
723 | tsk->exit_signal : SIGCHLD; | |
724 | autoreap = do_notify_parent(tsk, sig); | |
725 | } else if (thread_group_leader(tsk)) { | |
726 | autoreap = thread_group_empty(tsk) && | |
727 | do_notify_parent(tsk, tsk->exit_signal); | |
728 | } else { | |
729 | autoreap = true; | |
730 | } | |
1da177e4 | 731 | |
53c8f9f1 | 732 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
6c66e7db ON |
733 | if (tsk->exit_state == EXIT_DEAD) |
734 | list_add(&tsk->ptrace_entry, &dead); | |
1da177e4 | 735 | |
9c339168 ON |
736 | /* mt-exec, de_thread() is waiting for group leader */ |
737 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 738 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
739 | write_unlock_irq(&tasklist_lock); |
740 | ||
482a3767 ON |
741 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
742 | list_del_init(&p->ptrace_entry); | |
743 | release_task(p); | |
744 | } | |
1da177e4 LT |
745 | } |
746 | ||
e18eecb8 JD |
747 | #ifdef CONFIG_DEBUG_STACK_USAGE |
748 | static void check_stack_usage(void) | |
749 | { | |
750 | static DEFINE_SPINLOCK(low_water_lock); | |
751 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
752 | unsigned long free; |
753 | ||
7c9f8861 | 754 | free = stack_not_used(current); |
e18eecb8 JD |
755 | |
756 | if (free >= lowest_to_date) | |
757 | return; | |
758 | ||
759 | spin_lock(&low_water_lock); | |
760 | if (free < lowest_to_date) { | |
627393d4 | 761 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 762 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
763 | lowest_to_date = free; |
764 | } | |
765 | spin_unlock(&low_water_lock); | |
766 | } | |
767 | #else | |
768 | static inline void check_stack_usage(void) {} | |
769 | #endif | |
770 | ||
9af6528e | 771 | void __noreturn do_exit(long code) |
1da177e4 LT |
772 | { |
773 | struct task_struct *tsk = current; | |
774 | int group_dead; | |
3f95aa81 | 775 | TASKS_RCU(int tasks_rcu_i); |
1da177e4 LT |
776 | |
777 | profile_task_exit(tsk); | |
5c9a8750 | 778 | kcov_task_exit(tsk); |
1da177e4 | 779 | |
73c10101 | 780 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 781 | |
1da177e4 LT |
782 | if (unlikely(in_interrupt())) |
783 | panic("Aiee, killing interrupt handler!"); | |
784 | if (unlikely(!tsk->pid)) | |
785 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 786 | |
33dd94ae NE |
787 | /* |
788 | * If do_exit is called because this processes oopsed, it's possible | |
789 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
790 | * continuing. Amongst other possible reasons, this is to prevent | |
791 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
792 | * kernel address. | |
793 | */ | |
794 | set_fs(USER_DS); | |
795 | ||
a288eecc | 796 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 797 | |
e0e81739 DH |
798 | validate_creds_for_do_exit(tsk); |
799 | ||
df164db5 AN |
800 | /* |
801 | * We're taking recursive faults here in do_exit. Safest is to just | |
802 | * leave this task alone and wait for reboot. | |
803 | */ | |
804 | if (unlikely(tsk->flags & PF_EXITING)) { | |
a0be55de | 805 | pr_alert("Fixing recursive fault but reboot is needed!\n"); |
778e9a9c AK |
806 | /* |
807 | * We can do this unlocked here. The futex code uses | |
808 | * this flag just to verify whether the pi state | |
809 | * cleanup has been done or not. In the worst case it | |
810 | * loops once more. We pretend that the cleanup was | |
811 | * done as there is no way to return. Either the | |
812 | * OWNER_DIED bit is set by now or we push the blocked | |
813 | * task into the wait for ever nirwana as well. | |
814 | */ | |
815 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
816 | set_current_state(TASK_UNINTERRUPTIBLE); |
817 | schedule(); | |
818 | } | |
819 | ||
d12619b5 | 820 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c | 821 | /* |
be3e7844 PZ |
822 | * Ensure that all new tsk->pi_lock acquisitions must observe |
823 | * PF_EXITING. Serializes against futex.c:attach_to_pi_owner(). | |
778e9a9c | 824 | */ |
d2ee7198 | 825 | smp_mb(); |
be3e7844 PZ |
826 | /* |
827 | * Ensure that we must observe the pi_state in exit_mm() -> | |
828 | * mm_release() -> exit_pi_state_list(). | |
829 | */ | |
1d615482 | 830 | raw_spin_unlock_wait(&tsk->pi_lock); |
1da177e4 | 831 | |
1dc0fffc | 832 | if (unlikely(in_atomic())) { |
a0be55de IA |
833 | pr_info("note: %s[%d] exited with preempt_count %d\n", |
834 | current->comm, task_pid_nr(current), | |
835 | preempt_count()); | |
1dc0fffc PZ |
836 | preempt_count_set(PREEMPT_ENABLED); |
837 | } | |
1da177e4 | 838 | |
48d212a2 LT |
839 | /* sync mm's RSS info before statistics gathering */ |
840 | if (tsk->mm) | |
841 | sync_mm_rss(tsk->mm); | |
51229b49 | 842 | acct_update_integrals(tsk); |
1da177e4 | 843 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 844 | if (group_dead) { |
baa73d9e | 845 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 846 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 847 | exit_itimers(tsk->signal); |
baa73d9e | 848 | #endif |
1f10206c JP |
849 | if (tsk->mm) |
850 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 851 | } |
f6ec29a4 | 852 | acct_collect(code, group_dead); |
522ed776 MT |
853 | if (group_dead) |
854 | tty_audit_exit(); | |
a4ff8dba | 855 | audit_free(tsk); |
115085ea | 856 | |
48d212a2 | 857 | tsk->exit_code = code; |
115085ea | 858 | taskstats_exit(tsk, group_dead); |
c757249a | 859 | |
0039962a | 860 | exit_mm(); |
1da177e4 | 861 | |
0e464814 | 862 | if (group_dead) |
f6ec29a4 | 863 | acct_process(); |
0a16b607 MD |
864 | trace_sched_process_exit(tsk); |
865 | ||
1da177e4 | 866 | exit_sem(tsk); |
b34a6b1d | 867 | exit_shm(tsk); |
1ec7f1dd AV |
868 | exit_files(tsk); |
869 | exit_fs(tsk); | |
c39df5fa ON |
870 | if (group_dead) |
871 | disassociate_ctty(1); | |
8aac6270 | 872 | exit_task_namespaces(tsk); |
ed3e694d | 873 | exit_task_work(tsk); |
e6464694 | 874 | exit_thread(tsk); |
0b3fcf17 SE |
875 | |
876 | /* | |
877 | * Flush inherited counters to the parent - before the parent | |
878 | * gets woken up by child-exit notifications. | |
879 | * | |
880 | * because of cgroup mode, must be called before cgroup_exit() | |
881 | */ | |
882 | perf_event_exit_task(tsk); | |
883 | ||
8e5bfa8c | 884 | sched_autogroup_exit_task(tsk); |
1ec41830 | 885 | cgroup_exit(tsk); |
1da177e4 | 886 | |
24f1e32c FW |
887 | /* |
888 | * FIXME: do that only when needed, using sched_exit tracepoint | |
889 | */ | |
7c8df286 | 890 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 891 | |
49f5903b | 892 | TASKS_RCU(preempt_disable()); |
3f95aa81 | 893 | TASKS_RCU(tasks_rcu_i = __srcu_read_lock(&tasks_rcu_exit_srcu)); |
49f5903b | 894 | TASKS_RCU(preempt_enable()); |
821c7de7 | 895 | exit_notify(tsk, group_dead); |
ef982393 | 896 | proc_exit_connector(tsk); |
c11600e4 | 897 | mpol_put_task_policy(tsk); |
42b2dd0a | 898 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
899 | if (unlikely(current->pi_state_cache)) |
900 | kfree(current->pi_state_cache); | |
42b2dd0a | 901 | #endif |
de5097c2 | 902 | /* |
9a11b49a | 903 | * Make sure we are holding no locks: |
de5097c2 | 904 | */ |
1b1d2fb4 | 905 | debug_check_no_locks_held(); |
778e9a9c AK |
906 | /* |
907 | * We can do this unlocked here. The futex code uses this flag | |
908 | * just to verify whether the pi state cleanup has been done | |
909 | * or not. In the worst case it loops once more. | |
910 | */ | |
911 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 912 | |
afc847b7 | 913 | if (tsk->io_context) |
b69f2292 | 914 | exit_io_context(tsk); |
afc847b7 | 915 | |
b92ce558 | 916 | if (tsk->splice_pipe) |
4b8a8f1e | 917 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 918 | |
5640f768 ED |
919 | if (tsk->task_frag.page) |
920 | put_page(tsk->task_frag.page); | |
921 | ||
e0e81739 DH |
922 | validate_creds_for_do_exit(tsk); |
923 | ||
4bcb8232 | 924 | check_stack_usage(); |
7407251a | 925 | preempt_disable(); |
54848d73 WF |
926 | if (tsk->nr_dirtied) |
927 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 928 | exit_rcu(); |
3f95aa81 | 929 | TASKS_RCU(__srcu_read_unlock(&tasks_rcu_exit_srcu, tasks_rcu_i)); |
b5740f4b | 930 | |
9af6528e | 931 | do_task_dead(); |
1da177e4 | 932 | } |
012914da RA |
933 | EXPORT_SYMBOL_GPL(do_exit); |
934 | ||
9402c95f | 935 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
936 | { |
937 | if (comp) | |
938 | complete(comp); | |
55a101f8 | 939 | |
1da177e4 LT |
940 | do_exit(code); |
941 | } | |
1da177e4 LT |
942 | EXPORT_SYMBOL(complete_and_exit); |
943 | ||
754fe8d2 | 944 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
945 | { |
946 | do_exit((error_code&0xff)<<8); | |
947 | } | |
948 | ||
1da177e4 LT |
949 | /* |
950 | * Take down every thread in the group. This is called by fatal signals | |
951 | * as well as by sys_exit_group (below). | |
952 | */ | |
9402c95f | 953 | void |
1da177e4 LT |
954 | do_group_exit(int exit_code) |
955 | { | |
bfc4b089 ON |
956 | struct signal_struct *sig = current->signal; |
957 | ||
1da177e4 LT |
958 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
959 | ||
bfc4b089 ON |
960 | if (signal_group_exit(sig)) |
961 | exit_code = sig->group_exit_code; | |
1da177e4 | 962 | else if (!thread_group_empty(current)) { |
1da177e4 | 963 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 964 | |
1da177e4 | 965 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 966 | if (signal_group_exit(sig)) |
1da177e4 LT |
967 | /* Another thread got here before we took the lock. */ |
968 | exit_code = sig->group_exit_code; | |
969 | else { | |
1da177e4 | 970 | sig->group_exit_code = exit_code; |
ed5d2cac | 971 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
972 | zap_other_threads(current); |
973 | } | |
974 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
975 | } |
976 | ||
977 | do_exit(exit_code); | |
978 | /* NOTREACHED */ | |
979 | } | |
980 | ||
981 | /* | |
982 | * this kills every thread in the thread group. Note that any externally | |
983 | * wait4()-ing process will get the correct exit code - even if this | |
984 | * thread is not the thread group leader. | |
985 | */ | |
754fe8d2 | 986 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
987 | { |
988 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
989 | /* NOTREACHED */ |
990 | return 0; | |
1da177e4 LT |
991 | } |
992 | ||
9e8ae01d ON |
993 | struct wait_opts { |
994 | enum pid_type wo_type; | |
9e8ae01d | 995 | int wo_flags; |
e1eb1ebc | 996 | struct pid *wo_pid; |
9e8ae01d ON |
997 | |
998 | struct siginfo __user *wo_info; | |
999 | int __user *wo_stat; | |
1000 | struct rusage __user *wo_rusage; | |
1001 | ||
0b7570e7 | 1002 | wait_queue_t child_wait; |
9e8ae01d ON |
1003 | int notask_error; |
1004 | }; | |
1005 | ||
989264f4 ON |
1006 | static inline |
1007 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 1008 | { |
989264f4 ON |
1009 | if (type != PIDTYPE_PID) |
1010 | task = task->group_leader; | |
1011 | return task->pids[type].pid; | |
161550d7 EB |
1012 | } |
1013 | ||
989264f4 | 1014 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1015 | { |
5c01ba49 ON |
1016 | return wo->wo_type == PIDTYPE_MAX || |
1017 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1018 | } | |
1da177e4 | 1019 | |
bf959931 ON |
1020 | static int |
1021 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
1022 | { |
1023 | if (!eligible_pid(wo, p)) | |
1024 | return 0; | |
bf959931 ON |
1025 | |
1026 | /* | |
1027 | * Wait for all children (clone and not) if __WALL is set or | |
1028 | * if it is traced by us. | |
1029 | */ | |
1030 | if (ptrace || (wo->wo_flags & __WALL)) | |
1031 | return 1; | |
1032 | ||
1033 | /* | |
1034 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
1035 | * otherwise, wait for non-clone children *only*. | |
1036 | * | |
1037 | * Note: a "clone" child here is one that reports to its parent | |
1038 | * using a signal other than SIGCHLD, or a non-leader thread which | |
1039 | * we can only see if it is traced by us. | |
1040 | */ | |
1041 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 1042 | return 0; |
1da177e4 | 1043 | |
14dd0b81 | 1044 | return 1; |
1da177e4 LT |
1045 | } |
1046 | ||
9e8ae01d ON |
1047 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
1048 | pid_t pid, uid_t uid, int why, int status) | |
1da177e4 | 1049 | { |
9e8ae01d ON |
1050 | struct siginfo __user *infop; |
1051 | int retval = wo->wo_rusage | |
1052 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
36c8b586 | 1053 | |
1da177e4 | 1054 | put_task_struct(p); |
9e8ae01d | 1055 | infop = wo->wo_info; |
b6fe2d11 VM |
1056 | if (infop) { |
1057 | if (!retval) | |
1058 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1059 | if (!retval) | |
1060 | retval = put_user(0, &infop->si_errno); | |
1061 | if (!retval) | |
1062 | retval = put_user((short)why, &infop->si_code); | |
1063 | if (!retval) | |
1064 | retval = put_user(pid, &infop->si_pid); | |
1065 | if (!retval) | |
1066 | retval = put_user(uid, &infop->si_uid); | |
1067 | if (!retval) | |
1068 | retval = put_user(status, &infop->si_status); | |
1069 | } | |
1da177e4 LT |
1070 | if (!retval) |
1071 | retval = pid; | |
1072 | return retval; | |
1073 | } | |
1074 | ||
1075 | /* | |
1076 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1077 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1078 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1079 | * released the lock and the system call should return. | |
1080 | */ | |
9e8ae01d | 1081 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1082 | { |
f6507f83 | 1083 | int state, retval, status; |
6c5f3e7b | 1084 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1085 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
9e8ae01d | 1086 | struct siginfo __user *infop; |
1da177e4 | 1087 | |
9e8ae01d | 1088 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1089 | return 0; |
1090 | ||
9e8ae01d | 1091 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1da177e4 | 1092 | int exit_code = p->exit_code; |
f3abd4f9 | 1093 | int why; |
1da177e4 | 1094 | |
1da177e4 LT |
1095 | get_task_struct(p); |
1096 | read_unlock(&tasklist_lock); | |
1029a2b5 PZ |
1097 | sched_annotate_sleep(); |
1098 | ||
1da177e4 LT |
1099 | if ((exit_code & 0x7f) == 0) { |
1100 | why = CLD_EXITED; | |
1101 | status = exit_code >> 8; | |
1102 | } else { | |
1103 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1104 | status = exit_code & 0x7f; | |
1105 | } | |
9e8ae01d | 1106 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1da177e4 | 1107 | } |
1da177e4 | 1108 | /* |
abd50b39 | 1109 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1110 | */ |
f6507f83 ON |
1111 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1112 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1113 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1114 | return 0; |
986094df ON |
1115 | /* |
1116 | * We own this thread, nobody else can reap it. | |
1117 | */ | |
1118 | read_unlock(&tasklist_lock); | |
1119 | sched_annotate_sleep(); | |
f6507f83 | 1120 | |
befca967 | 1121 | /* |
f6507f83 | 1122 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1123 | */ |
f6507f83 | 1124 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1125 | struct signal_struct *sig = p->signal; |
1126 | struct signal_struct *psig = current->signal; | |
1f10206c | 1127 | unsigned long maxrss; |
5613fda9 | 1128 | u64 tgutime, tgstime; |
3795e161 | 1129 | |
1da177e4 LT |
1130 | /* |
1131 | * The resource counters for the group leader are in its | |
1132 | * own task_struct. Those for dead threads in the group | |
1133 | * are in its signal_struct, as are those for the child | |
1134 | * processes it has previously reaped. All these | |
1135 | * accumulate in the parent's signal_struct c* fields. | |
1136 | * | |
1137 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1138 | * p->signal fields because the whole thread group is dead |
1139 | * and nobody can change them. | |
1140 | * | |
1141 | * psig->stats_lock also protects us from our sub-theads | |
1142 | * which can reap other children at the same time. Until | |
1143 | * we change k_getrusage()-like users to rely on this lock | |
1144 | * we have to take ->siglock as well. | |
0cf55e1e | 1145 | * |
a0be55de IA |
1146 | * We use thread_group_cputime_adjusted() to get times for |
1147 | * the thread group, which consolidates times for all threads | |
1148 | * in the group including the group leader. | |
1da177e4 | 1149 | */ |
e80d0a1a | 1150 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1151 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1152 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1153 | psig->cutime += tgutime + sig->cutime; |
1154 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1155 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1156 | psig->cmin_flt += |
1157 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1158 | psig->cmaj_flt += | |
1159 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1160 | psig->cnvcsw += | |
1161 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1162 | psig->cnivcsw += | |
1163 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1164 | psig->cinblock += |
1165 | task_io_get_inblock(p) + | |
1166 | sig->inblock + sig->cinblock; | |
1167 | psig->coublock += | |
1168 | task_io_get_oublock(p) + | |
1169 | sig->oublock + sig->coublock; | |
1f10206c JP |
1170 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1171 | if (psig->cmaxrss < maxrss) | |
1172 | psig->cmaxrss = maxrss; | |
5995477a AR |
1173 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1174 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1175 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1176 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1177 | } |
1178 | ||
9e8ae01d ON |
1179 | retval = wo->wo_rusage |
1180 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 LT |
1181 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1182 | ? p->signal->group_exit_code : p->exit_code; | |
9e8ae01d ON |
1183 | if (!retval && wo->wo_stat) |
1184 | retval = put_user(status, wo->wo_stat); | |
1185 | ||
1186 | infop = wo->wo_info; | |
1da177e4 LT |
1187 | if (!retval && infop) |
1188 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1189 | if (!retval && infop) | |
1190 | retval = put_user(0, &infop->si_errno); | |
1191 | if (!retval && infop) { | |
1192 | int why; | |
1193 | ||
1194 | if ((status & 0x7f) == 0) { | |
1195 | why = CLD_EXITED; | |
1196 | status >>= 8; | |
1197 | } else { | |
1198 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1199 | status &= 0x7f; | |
1200 | } | |
1201 | retval = put_user((short)why, &infop->si_code); | |
1202 | if (!retval) | |
1203 | retval = put_user(status, &infop->si_status); | |
1204 | } | |
1205 | if (!retval && infop) | |
3a515e4a | 1206 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1207 | if (!retval && infop) |
c69e8d9c | 1208 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1209 | if (!retval) |
3a515e4a | 1210 | retval = pid; |
2f4e6e2a | 1211 | |
b4360690 | 1212 | if (state == EXIT_TRACE) { |
1da177e4 | 1213 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1214 | /* We dropped tasklist, ptracer could die and untrace */ |
1215 | ptrace_unlink(p); | |
b4360690 ON |
1216 | |
1217 | /* If parent wants a zombie, don't release it now */ | |
1218 | state = EXIT_ZOMBIE; | |
1219 | if (do_notify_parent(p, p->exit_signal)) | |
1220 | state = EXIT_DEAD; | |
abd50b39 | 1221 | p->exit_state = state; |
1da177e4 LT |
1222 | write_unlock_irq(&tasklist_lock); |
1223 | } | |
abd50b39 | 1224 | if (state == EXIT_DEAD) |
1da177e4 | 1225 | release_task(p); |
2f4e6e2a | 1226 | |
1da177e4 LT |
1227 | return retval; |
1228 | } | |
1229 | ||
90bc8d8b ON |
1230 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1231 | { | |
1232 | if (ptrace) { | |
570ac933 | 1233 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1234 | return &p->exit_code; |
1235 | } else { | |
1236 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1237 | return &p->signal->group_exit_code; | |
1238 | } | |
1239 | return NULL; | |
1240 | } | |
1241 | ||
19e27463 TH |
1242 | /** |
1243 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1244 | * @wo: wait options | |
1245 | * @ptrace: is the wait for ptrace | |
1246 | * @p: task to wait for | |
1247 | * | |
1248 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1249 | * | |
1250 | * CONTEXT: | |
1251 | * read_lock(&tasklist_lock), which is released if return value is | |
1252 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1253 | * | |
1254 | * RETURNS: | |
1255 | * 0 if wait condition didn't exist and search for other wait conditions | |
1256 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1257 | * success, implies that tasklist_lock is released and wait condition | |
1258 | * search should terminate. | |
1da177e4 | 1259 | */ |
9e8ae01d ON |
1260 | static int wait_task_stopped(struct wait_opts *wo, |
1261 | int ptrace, struct task_struct *p) | |
1da177e4 | 1262 | { |
9e8ae01d | 1263 | struct siginfo __user *infop; |
90bc8d8b | 1264 | int retval, exit_code, *p_code, why; |
ee7c82da | 1265 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1266 | pid_t pid; |
1da177e4 | 1267 | |
47918025 ON |
1268 | /* |
1269 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1270 | */ | |
9e8ae01d | 1271 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1272 | return 0; |
1273 | ||
19e27463 TH |
1274 | if (!task_stopped_code(p, ptrace)) |
1275 | return 0; | |
1276 | ||
ee7c82da ON |
1277 | exit_code = 0; |
1278 | spin_lock_irq(&p->sighand->siglock); | |
1279 | ||
90bc8d8b ON |
1280 | p_code = task_stopped_code(p, ptrace); |
1281 | if (unlikely(!p_code)) | |
ee7c82da ON |
1282 | goto unlock_sig; |
1283 | ||
90bc8d8b | 1284 | exit_code = *p_code; |
ee7c82da ON |
1285 | if (!exit_code) |
1286 | goto unlock_sig; | |
1287 | ||
9e8ae01d | 1288 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1289 | *p_code = 0; |
ee7c82da | 1290 | |
8ca937a6 | 1291 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1292 | unlock_sig: |
1293 | spin_unlock_irq(&p->sighand->siglock); | |
1294 | if (!exit_code) | |
1da177e4 LT |
1295 | return 0; |
1296 | ||
1297 | /* | |
1298 | * Now we are pretty sure this task is interesting. | |
1299 | * Make sure it doesn't get reaped out from under us while we | |
1300 | * give up the lock and then examine it below. We don't want to | |
1301 | * keep holding onto the tasklist_lock while we call getrusage and | |
1302 | * possibly take page faults for user memory. | |
1303 | */ | |
1304 | get_task_struct(p); | |
6c5f3e7b | 1305 | pid = task_pid_vnr(p); |
f470021a | 1306 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1307 | read_unlock(&tasklist_lock); |
1029a2b5 | 1308 | sched_annotate_sleep(); |
1da177e4 | 1309 | |
9e8ae01d ON |
1310 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1311 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | |
1312 | ||
1313 | retval = wo->wo_rusage | |
1314 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1315 | if (!retval && wo->wo_stat) | |
1316 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | |
1da177e4 | 1317 | |
9e8ae01d | 1318 | infop = wo->wo_info; |
1da177e4 LT |
1319 | if (!retval && infop) |
1320 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1321 | if (!retval && infop) | |
1322 | retval = put_user(0, &infop->si_errno); | |
1323 | if (!retval && infop) | |
6efcae46 | 1324 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1325 | if (!retval && infop) |
1326 | retval = put_user(exit_code, &infop->si_status); | |
1327 | if (!retval && infop) | |
c8950783 | 1328 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1329 | if (!retval && infop) |
ee7c82da | 1330 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1331 | if (!retval) |
c8950783 | 1332 | retval = pid; |
1da177e4 LT |
1333 | put_task_struct(p); |
1334 | ||
1335 | BUG_ON(!retval); | |
1336 | return retval; | |
1337 | } | |
1338 | ||
1339 | /* | |
1340 | * Handle do_wait work for one task in a live, non-stopped state. | |
1341 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1342 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1343 | * released the lock and the system call should return. | |
1344 | */ | |
9e8ae01d | 1345 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1346 | { |
1347 | int retval; | |
1348 | pid_t pid; | |
1349 | uid_t uid; | |
1350 | ||
9e8ae01d | 1351 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1352 | return 0; |
1353 | ||
1da177e4 LT |
1354 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1355 | return 0; | |
1356 | ||
1357 | spin_lock_irq(&p->sighand->siglock); | |
1358 | /* Re-check with the lock held. */ | |
1359 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1360 | spin_unlock_irq(&p->sighand->siglock); | |
1361 | return 0; | |
1362 | } | |
9e8ae01d | 1363 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1364 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1365 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1366 | spin_unlock_irq(&p->sighand->siglock); |
1367 | ||
6c5f3e7b | 1368 | pid = task_pid_vnr(p); |
1da177e4 LT |
1369 | get_task_struct(p); |
1370 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1371 | sched_annotate_sleep(); |
1da177e4 | 1372 | |
9e8ae01d ON |
1373 | if (!wo->wo_info) { |
1374 | retval = wo->wo_rusage | |
1375 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 | 1376 | put_task_struct(p); |
9e8ae01d ON |
1377 | if (!retval && wo->wo_stat) |
1378 | retval = put_user(0xffff, wo->wo_stat); | |
1da177e4 | 1379 | if (!retval) |
3a515e4a | 1380 | retval = pid; |
1da177e4 | 1381 | } else { |
9e8ae01d ON |
1382 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1383 | CLD_CONTINUED, SIGCONT); | |
1da177e4 LT |
1384 | BUG_ON(retval == 0); |
1385 | } | |
1386 | ||
1387 | return retval; | |
1388 | } | |
1389 | ||
98abed02 RM |
1390 | /* |
1391 | * Consider @p for a wait by @parent. | |
1392 | * | |
9e8ae01d | 1393 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1394 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1395 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1396 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1397 | * or still -ECHILD. |
98abed02 | 1398 | */ |
b6e763f0 ON |
1399 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1400 | struct task_struct *p) | |
98abed02 | 1401 | { |
3245d6ac ON |
1402 | /* |
1403 | * We can race with wait_task_zombie() from another thread. | |
1404 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1405 | * can't confuse the checks below. | |
1406 | */ | |
1407 | int exit_state = ACCESS_ONCE(p->exit_state); | |
b3ab0316 ON |
1408 | int ret; |
1409 | ||
3245d6ac | 1410 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1411 | return 0; |
1412 | ||
bf959931 | 1413 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1414 | if (!ret) |
98abed02 RM |
1415 | return ret; |
1416 | ||
3245d6ac | 1417 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1418 | /* |
abd50b39 ON |
1419 | * ptrace == 0 means we are the natural parent. In this case |
1420 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1421 | */ |
abd50b39 | 1422 | if (likely(!ptrace)) |
50b8d257 | 1423 | wo->notask_error = 0; |
823b018e | 1424 | return 0; |
50b8d257 | 1425 | } |
823b018e | 1426 | |
377d75da ON |
1427 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1428 | /* | |
1429 | * If it is traced by its real parent's group, just pretend | |
1430 | * the caller is ptrace_do_wait() and reap this child if it | |
1431 | * is zombie. | |
1432 | * | |
1433 | * This also hides group stop state from real parent; otherwise | |
1434 | * a single stop can be reported twice as group and ptrace stop. | |
1435 | * If a ptracer wants to distinguish these two events for its | |
1436 | * own children it should create a separate process which takes | |
1437 | * the role of real parent. | |
1438 | */ | |
1439 | if (!ptrace_reparented(p)) | |
1440 | ptrace = 1; | |
1441 | } | |
1442 | ||
45cb24a1 | 1443 | /* slay zombie? */ |
3245d6ac | 1444 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1445 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1446 | if (!delay_group_leader(p)) { |
1447 | /* | |
1448 | * A zombie ptracee is only visible to its ptracer. | |
1449 | * Notification and reaping will be cascaded to the | |
1450 | * real parent when the ptracer detaches. | |
1451 | */ | |
1452 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1453 | return wait_task_zombie(wo, p); | |
1454 | } | |
98abed02 | 1455 | |
f470021a | 1456 | /* |
9b84cca2 TH |
1457 | * Allow access to stopped/continued state via zombie by |
1458 | * falling through. Clearing of notask_error is complex. | |
1459 | * | |
1460 | * When !@ptrace: | |
1461 | * | |
1462 | * If WEXITED is set, notask_error should naturally be | |
1463 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1464 | * so, if there are live subthreads, there are events to | |
1465 | * wait for. If all subthreads are dead, it's still safe | |
1466 | * to clear - this function will be called again in finite | |
1467 | * amount time once all the subthreads are released and | |
1468 | * will then return without clearing. | |
1469 | * | |
1470 | * When @ptrace: | |
1471 | * | |
1472 | * Stopped state is per-task and thus can't change once the | |
1473 | * target task dies. Only continued and exited can happen. | |
1474 | * Clear notask_error if WCONTINUED | WEXITED. | |
1475 | */ | |
1476 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1477 | wo->notask_error = 0; | |
1478 | } else { | |
1479 | /* | |
1480 | * @p is alive and it's gonna stop, continue or exit, so | |
1481 | * there always is something to wait for. | |
f470021a | 1482 | */ |
9e8ae01d | 1483 | wo->notask_error = 0; |
f470021a RM |
1484 | } |
1485 | ||
98abed02 | 1486 | /* |
45cb24a1 TH |
1487 | * Wait for stopped. Depending on @ptrace, different stopped state |
1488 | * is used and the two don't interact with each other. | |
98abed02 | 1489 | */ |
19e27463 TH |
1490 | ret = wait_task_stopped(wo, ptrace, p); |
1491 | if (ret) | |
1492 | return ret; | |
98abed02 RM |
1493 | |
1494 | /* | |
45cb24a1 TH |
1495 | * Wait for continued. There's only one continued state and the |
1496 | * ptracer can consume it which can confuse the real parent. Don't | |
1497 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1498 | */ |
9e8ae01d | 1499 | return wait_task_continued(wo, p); |
98abed02 RM |
1500 | } |
1501 | ||
1502 | /* | |
1503 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1504 | * | |
9e8ae01d | 1505 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1506 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1507 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1508 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1509 | * or still -ECHILD. |
98abed02 | 1510 | */ |
9e8ae01d | 1511 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1512 | { |
1513 | struct task_struct *p; | |
1514 | ||
1515 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1516 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1517 | |
9cd80bbb ON |
1518 | if (ret) |
1519 | return ret; | |
98abed02 RM |
1520 | } |
1521 | ||
1522 | return 0; | |
1523 | } | |
1524 | ||
9e8ae01d | 1525 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1526 | { |
1527 | struct task_struct *p; | |
1528 | ||
f470021a | 1529 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1530 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1531 | |
f470021a | 1532 | if (ret) |
98abed02 | 1533 | return ret; |
98abed02 RM |
1534 | } |
1535 | ||
1536 | return 0; | |
1537 | } | |
1538 | ||
0b7570e7 ON |
1539 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1540 | int sync, void *key) | |
1541 | { | |
1542 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1543 | child_wait); | |
1544 | struct task_struct *p = key; | |
1545 | ||
5c01ba49 | 1546 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1547 | return 0; |
1548 | ||
b4fe5182 ON |
1549 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1550 | return 0; | |
1551 | ||
0b7570e7 ON |
1552 | return default_wake_function(wait, mode, sync, key); |
1553 | } | |
1554 | ||
a7f0765e ON |
1555 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1556 | { | |
0b7570e7 ON |
1557 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1558 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1559 | } |
1560 | ||
9e8ae01d | 1561 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1562 | { |
1da177e4 | 1563 | struct task_struct *tsk; |
98abed02 | 1564 | int retval; |
1da177e4 | 1565 | |
9e8ae01d | 1566 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1567 | |
0b7570e7 ON |
1568 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1569 | wo->child_wait.private = current; | |
1570 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1571 | repeat: |
98abed02 | 1572 | /* |
3da56d16 | 1573 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1574 | * We will clear ->notask_error to zero if we see any child that |
1575 | * might later match our criteria, even if we are not able to reap | |
1576 | * it yet. | |
98abed02 | 1577 | */ |
64a16caf | 1578 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1579 | if ((wo->wo_type < PIDTYPE_MAX) && |
1580 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1581 | goto notask; |
161550d7 | 1582 | |
f95d39d1 | 1583 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1584 | read_lock(&tasklist_lock); |
1585 | tsk = current; | |
1586 | do { | |
64a16caf ON |
1587 | retval = do_wait_thread(wo, tsk); |
1588 | if (retval) | |
1589 | goto end; | |
9e8ae01d | 1590 | |
64a16caf ON |
1591 | retval = ptrace_do_wait(wo, tsk); |
1592 | if (retval) | |
98abed02 | 1593 | goto end; |
98abed02 | 1594 | |
9e8ae01d | 1595 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1596 | break; |
a3f6dfb7 | 1597 | } while_each_thread(current, tsk); |
1da177e4 | 1598 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1599 | |
64a16caf | 1600 | notask: |
9e8ae01d ON |
1601 | retval = wo->notask_error; |
1602 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1603 | retval = -ERESTARTSYS; |
98abed02 RM |
1604 | if (!signal_pending(current)) { |
1605 | schedule(); | |
1606 | goto repeat; | |
1607 | } | |
1da177e4 | 1608 | } |
1da177e4 | 1609 | end: |
f95d39d1 | 1610 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1611 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1612 | return retval; |
1613 | } | |
1614 | ||
17da2bd9 HC |
1615 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1616 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1617 | { |
9e8ae01d | 1618 | struct wait_opts wo; |
161550d7 EB |
1619 | struct pid *pid = NULL; |
1620 | enum pid_type type; | |
1da177e4 LT |
1621 | long ret; |
1622 | ||
91c4e8ea ON |
1623 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1624 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1625 | return -EINVAL; |
1626 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1627 | return -EINVAL; | |
1628 | ||
1629 | switch (which) { | |
1630 | case P_ALL: | |
161550d7 | 1631 | type = PIDTYPE_MAX; |
1da177e4 LT |
1632 | break; |
1633 | case P_PID: | |
161550d7 EB |
1634 | type = PIDTYPE_PID; |
1635 | if (upid <= 0) | |
1da177e4 LT |
1636 | return -EINVAL; |
1637 | break; | |
1638 | case P_PGID: | |
161550d7 EB |
1639 | type = PIDTYPE_PGID; |
1640 | if (upid <= 0) | |
1da177e4 | 1641 | return -EINVAL; |
1da177e4 LT |
1642 | break; |
1643 | default: | |
1644 | return -EINVAL; | |
1645 | } | |
1646 | ||
161550d7 EB |
1647 | if (type < PIDTYPE_MAX) |
1648 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1649 | |
1650 | wo.wo_type = type; | |
1651 | wo.wo_pid = pid; | |
1652 | wo.wo_flags = options; | |
1653 | wo.wo_info = infop; | |
1654 | wo.wo_stat = NULL; | |
1655 | wo.wo_rusage = ru; | |
1656 | ret = do_wait(&wo); | |
dfe16dfa VM |
1657 | |
1658 | if (ret > 0) { | |
1659 | ret = 0; | |
1660 | } else if (infop) { | |
1661 | /* | |
1662 | * For a WNOHANG return, clear out all the fields | |
1663 | * we would set so the user can easily tell the | |
1664 | * difference. | |
1665 | */ | |
1666 | if (!ret) | |
1667 | ret = put_user(0, &infop->si_signo); | |
1668 | if (!ret) | |
1669 | ret = put_user(0, &infop->si_errno); | |
1670 | if (!ret) | |
1671 | ret = put_user(0, &infop->si_code); | |
1672 | if (!ret) | |
1673 | ret = put_user(0, &infop->si_pid); | |
1674 | if (!ret) | |
1675 | ret = put_user(0, &infop->si_uid); | |
1676 | if (!ret) | |
1677 | ret = put_user(0, &infop->si_status); | |
1678 | } | |
1679 | ||
161550d7 | 1680 | put_pid(pid); |
1da177e4 LT |
1681 | return ret; |
1682 | } | |
1683 | ||
754fe8d2 HC |
1684 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1685 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1686 | { |
9e8ae01d | 1687 | struct wait_opts wo; |
161550d7 EB |
1688 | struct pid *pid = NULL; |
1689 | enum pid_type type; | |
1da177e4 LT |
1690 | long ret; |
1691 | ||
1692 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1693 | __WNOTHREAD|__WCLONE|__WALL)) | |
1694 | return -EINVAL; | |
161550d7 EB |
1695 | |
1696 | if (upid == -1) | |
1697 | type = PIDTYPE_MAX; | |
1698 | else if (upid < 0) { | |
1699 | type = PIDTYPE_PGID; | |
1700 | pid = find_get_pid(-upid); | |
1701 | } else if (upid == 0) { | |
1702 | type = PIDTYPE_PGID; | |
2ae448ef | 1703 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1704 | } else /* upid > 0 */ { |
1705 | type = PIDTYPE_PID; | |
1706 | pid = find_get_pid(upid); | |
1707 | } | |
1708 | ||
9e8ae01d ON |
1709 | wo.wo_type = type; |
1710 | wo.wo_pid = pid; | |
1711 | wo.wo_flags = options | WEXITED; | |
1712 | wo.wo_info = NULL; | |
1713 | wo.wo_stat = stat_addr; | |
1714 | wo.wo_rusage = ru; | |
1715 | ret = do_wait(&wo); | |
161550d7 | 1716 | put_pid(pid); |
1da177e4 | 1717 | |
1da177e4 LT |
1718 | return ret; |
1719 | } | |
1720 | ||
1721 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1722 | ||
1723 | /* | |
1724 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1725 | * implemented by calling sys_wait4() from libc.a. | |
1726 | */ | |
17da2bd9 | 1727 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1728 | { |
1729 | return sys_wait4(pid, stat_addr, options, NULL); | |
1730 | } | |
1731 | ||
1732 | #endif |