]>
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> | |
4eb5aaa3 | 9 | #include <linux/sched/autogroup.h> |
6e84f315 | 10 | #include <linux/sched/mm.h> |
03441a34 | 11 | #include <linux/sched/stat.h> |
29930025 | 12 | #include <linux/sched/task.h> |
68db0cf1 | 13 | #include <linux/sched/task_stack.h> |
32ef5517 | 14 | #include <linux/sched/cputime.h> |
1da177e4 | 15 | #include <linux/interrupt.h> |
1da177e4 | 16 | #include <linux/module.h> |
c59ede7b | 17 | #include <linux/capability.h> |
1da177e4 LT |
18 | #include <linux/completion.h> |
19 | #include <linux/personality.h> | |
20 | #include <linux/tty.h> | |
da9cbc87 | 21 | #include <linux/iocontext.h> |
1da177e4 | 22 | #include <linux/key.h> |
1da177e4 LT |
23 | #include <linux/cpu.h> |
24 | #include <linux/acct.h> | |
8f0ab514 | 25 | #include <linux/tsacct_kern.h> |
1da177e4 | 26 | #include <linux/file.h> |
9f3acc31 | 27 | #include <linux/fdtable.h> |
80d26af8 | 28 | #include <linux/freezer.h> |
1da177e4 | 29 | #include <linux/binfmts.h> |
ab516013 | 30 | #include <linux/nsproxy.h> |
84d73786 | 31 | #include <linux/pid_namespace.h> |
1da177e4 LT |
32 | #include <linux/ptrace.h> |
33 | #include <linux/profile.h> | |
34 | #include <linux/mount.h> | |
35 | #include <linux/proc_fs.h> | |
49d769d5 | 36 | #include <linux/kthread.h> |
1da177e4 | 37 | #include <linux/mempolicy.h> |
c757249a | 38 | #include <linux/taskstats_kern.h> |
ca74e92b | 39 | #include <linux/delayacct.h> |
b4f48b63 | 40 | #include <linux/cgroup.h> |
1da177e4 | 41 | #include <linux/syscalls.h> |
7ed20e1a | 42 | #include <linux/signal.h> |
6a14c5c9 | 43 | #include <linux/posix-timers.h> |
9f46080c | 44 | #include <linux/cn_proc.h> |
de5097c2 | 45 | #include <linux/mutex.h> |
0771dfef | 46 | #include <linux/futex.h> |
b92ce558 | 47 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 48 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 49 | #include <linux/resource.h> |
0d67a46d | 50 | #include <linux/blkdev.h> |
6eaeeaba | 51 | #include <linux/task_io_accounting_ops.h> |
30199f5a | 52 | #include <linux/tracehook.h> |
5ad4e53b | 53 | #include <linux/fs_struct.h> |
d84f4f99 | 54 | #include <linux/init_task.h> |
cdd6c482 | 55 | #include <linux/perf_event.h> |
ad8d75ff | 56 | #include <trace/events/sched.h> |
24f1e32c | 57 | #include <linux/hw_breakpoint.h> |
3d5992d2 | 58 | #include <linux/oom.h> |
54848d73 | 59 | #include <linux/writeback.h> |
40401530 | 60 | #include <linux/shm.h> |
5c9a8750 | 61 | #include <linux/kcov.h> |
53d3eaa3 | 62 | #include <linux/random.h> |
8f95c90c | 63 | #include <linux/rcuwait.h> |
7e95a225 | 64 | #include <linux/compat.h> |
1da177e4 | 65 | |
7c0f6ba6 | 66 | #include <linux/uaccess.h> |
1da177e4 LT |
67 | #include <asm/unistd.h> |
68 | #include <asm/pgtable.h> | |
69 | #include <asm/mmu_context.h> | |
70 | ||
d40e48e0 | 71 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
72 | { |
73 | nr_threads--; | |
50d75f8d | 74 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 75 | if (group_dead) { |
1da177e4 LT |
76 | detach_pid(p, PIDTYPE_PGID); |
77 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 78 | |
5e85d4ab | 79 | list_del_rcu(&p->tasks); |
9cd80bbb | 80 | list_del_init(&p->sibling); |
909ea964 | 81 | __this_cpu_dec(process_counts); |
1da177e4 | 82 | } |
47e65328 | 83 | list_del_rcu(&p->thread_group); |
0c740d0a | 84 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
85 | } |
86 | ||
6a14c5c9 ON |
87 | /* |
88 | * This function expects the tasklist_lock write-locked. | |
89 | */ | |
90 | static void __exit_signal(struct task_struct *tsk) | |
91 | { | |
92 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 93 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 94 | struct sighand_struct *sighand; |
4ada856f | 95 | struct tty_struct *uninitialized_var(tty); |
5613fda9 | 96 | u64 utime, stime; |
6a14c5c9 | 97 | |
d11c563d | 98 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 99 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
100 | spin_lock(&sighand->siglock); |
101 | ||
baa73d9e | 102 | #ifdef CONFIG_POSIX_TIMERS |
6a14c5c9 | 103 | posix_cpu_timers_exit(tsk); |
d40e48e0 | 104 | if (group_dead) { |
6a14c5c9 | 105 | posix_cpu_timers_exit_group(tsk); |
4a599942 | 106 | } else { |
e0a70217 ON |
107 | /* |
108 | * This can only happen if the caller is de_thread(). | |
109 | * FIXME: this is the temporary hack, we should teach | |
110 | * posix-cpu-timers to handle this case correctly. | |
111 | */ | |
112 | if (unlikely(has_group_leader_pid(tsk))) | |
113 | posix_cpu_timers_exit_group(tsk); | |
baa73d9e NP |
114 | } |
115 | #endif | |
e0a70217 | 116 | |
baa73d9e NP |
117 | if (group_dead) { |
118 | tty = sig->tty; | |
119 | sig->tty = NULL; | |
120 | } else { | |
6a14c5c9 ON |
121 | /* |
122 | * If there is any task waiting for the group exit | |
123 | * then notify it: | |
124 | */ | |
d344193a | 125 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 126 | wake_up_process(sig->group_exit_task); |
6db840fa | 127 | |
6a14c5c9 ON |
128 | if (tsk == sig->curr_target) |
129 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
130 | } |
131 | ||
53d3eaa3 NP |
132 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
133 | sizeof(unsigned long long)); | |
134 | ||
90ed9cbe | 135 | /* |
26e75b5c ON |
136 | * Accumulate here the counters for all threads as they die. We could |
137 | * skip the group leader because it is the last user of signal_struct, | |
138 | * but we want to avoid the race with thread_group_cputime() which can | |
139 | * see the empty ->thread_head list. | |
90ed9cbe RR |
140 | */ |
141 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 142 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
143 | sig->utime += utime; |
144 | sig->stime += stime; | |
145 | sig->gtime += task_gtime(tsk); | |
146 | sig->min_flt += tsk->min_flt; | |
147 | sig->maj_flt += tsk->maj_flt; | |
148 | sig->nvcsw += tsk->nvcsw; | |
149 | sig->nivcsw += tsk->nivcsw; | |
150 | sig->inblock += task_io_get_inblock(tsk); | |
151 | sig->oublock += task_io_get_oublock(tsk); | |
152 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
153 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 154 | sig->nr_threads--; |
d40e48e0 | 155 | __unhash_process(tsk, group_dead); |
e78c3496 | 156 | write_sequnlock(&sig->stats_lock); |
5876700c | 157 | |
da7978b0 ON |
158 | /* |
159 | * Do this under ->siglock, we can race with another thread | |
160 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
161 | */ | |
162 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 163 | tsk->sighand = NULL; |
6a14c5c9 | 164 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 165 | |
a7e5328a | 166 | __cleanup_sighand(sighand); |
a0be55de | 167 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 168 | if (group_dead) { |
6a14c5c9 | 169 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 170 | tty_kref_put(tty); |
6a14c5c9 ON |
171 | } |
172 | } | |
173 | ||
8c7904a0 EB |
174 | static void delayed_put_task_struct(struct rcu_head *rhp) |
175 | { | |
0a16b607 MD |
176 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
177 | ||
4e231c79 | 178 | perf_event_delayed_put(tsk); |
0a16b607 MD |
179 | trace_sched_process_free(tsk); |
180 | put_task_struct(tsk); | |
8c7904a0 EB |
181 | } |
182 | ||
f470021a | 183 | |
a0be55de | 184 | void release_task(struct task_struct *p) |
1da177e4 | 185 | { |
36c8b586 | 186 | struct task_struct *leader; |
1da177e4 | 187 | int zap_leader; |
1f09f974 | 188 | repeat: |
c69e8d9c | 189 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
190 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
191 | rcu_read_lock(); | |
c69e8d9c | 192 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 193 | rcu_read_unlock(); |
c69e8d9c | 194 | |
60347f67 | 195 | proc_flush_task(p); |
9f698e21 | 196 | cgroup_release(p); |
0203026b | 197 | |
1da177e4 | 198 | write_lock_irq(&tasklist_lock); |
a288eecc | 199 | ptrace_release_task(p); |
1da177e4 | 200 | __exit_signal(p); |
35f5cad8 | 201 | |
1da177e4 LT |
202 | /* |
203 | * If we are the last non-leader member of the thread | |
204 | * group, and the leader is zombie, then notify the | |
205 | * group leader's parent process. (if it wants notification.) | |
206 | */ | |
207 | zap_leader = 0; | |
208 | leader = p->group_leader; | |
a0be55de IA |
209 | if (leader != p && thread_group_empty(leader) |
210 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
211 | /* |
212 | * If we were the last child thread and the leader has | |
213 | * exited already, and the leader's parent ignores SIGCHLD, | |
214 | * then we are the one who should release the leader. | |
dae33574 | 215 | */ |
86773473 | 216 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
217 | if (zap_leader) |
218 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
219 | } |
220 | ||
1da177e4 | 221 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 222 | release_thread(p); |
8c7904a0 | 223 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
224 | |
225 | p = leader; | |
226 | if (unlikely(zap_leader)) | |
227 | goto repeat; | |
228 | } | |
229 | ||
150593bf ON |
230 | /* |
231 | * Note that if this function returns a valid task_struct pointer (!NULL) | |
232 | * task->usage must remain >0 for the duration of the RCU critical section. | |
233 | */ | |
234 | struct task_struct *task_rcu_dereference(struct task_struct **ptask) | |
235 | { | |
236 | struct sighand_struct *sighand; | |
237 | struct task_struct *task; | |
238 | ||
239 | /* | |
240 | * We need to verify that release_task() was not called and thus | |
241 | * delayed_put_task_struct() can't run and drop the last reference | |
242 | * before rcu_read_unlock(). We check task->sighand != NULL, | |
243 | * but we can read the already freed and reused memory. | |
244 | */ | |
245 | retry: | |
246 | task = rcu_dereference(*ptask); | |
247 | if (!task) | |
248 | return NULL; | |
249 | ||
250 | probe_kernel_address(&task->sighand, sighand); | |
251 | ||
252 | /* | |
253 | * Pairs with atomic_dec_and_test() in put_task_struct(). If this task | |
254 | * was already freed we can not miss the preceding update of this | |
255 | * pointer. | |
256 | */ | |
257 | smp_rmb(); | |
258 | if (unlikely(task != READ_ONCE(*ptask))) | |
259 | goto retry; | |
260 | ||
261 | /* | |
262 | * We've re-checked that "task == *ptask", now we have two different | |
263 | * cases: | |
264 | * | |
265 | * 1. This is actually the same task/task_struct. In this case | |
266 | * sighand != NULL tells us it is still alive. | |
267 | * | |
268 | * 2. This is another task which got the same memory for task_struct. | |
269 | * We can't know this of course, and we can not trust | |
270 | * sighand != NULL. | |
271 | * | |
272 | * In this case we actually return a random value, but this is | |
273 | * correct. | |
274 | * | |
275 | * If we return NULL - we can pretend that we actually noticed that | |
276 | * *ptask was updated when the previous task has exited. Or pretend | |
277 | * that probe_slab_address(&sighand) reads NULL. | |
278 | * | |
279 | * If we return the new task (because sighand is not NULL for any | |
280 | * reason) - this is fine too. This (new) task can't go away before | |
281 | * another gp pass. | |
282 | * | |
283 | * And note: We could even eliminate the false positive if re-read | |
284 | * task->sighand once again to avoid the falsely NULL. But this case | |
285 | * is very unlikely so we don't care. | |
286 | */ | |
287 | if (!sighand) | |
288 | return NULL; | |
289 | ||
290 | return task; | |
291 | } | |
292 | ||
8f95c90c DB |
293 | void rcuwait_wake_up(struct rcuwait *w) |
294 | { | |
295 | struct task_struct *task; | |
296 | ||
297 | rcu_read_lock(); | |
298 | ||
299 | /* | |
300 | * Order condition vs @task, such that everything prior to the load | |
301 | * of @task is visible. This is the condition as to why the user called | |
302 | * rcuwait_trywake() in the first place. Pairs with set_current_state() | |
303 | * barrier (A) in rcuwait_wait_event(). | |
304 | * | |
305 | * WAIT WAKE | |
306 | * [S] tsk = current [S] cond = true | |
307 | * MB (A) MB (B) | |
308 | * [L] cond [L] tsk | |
309 | */ | |
4120dd24 | 310 | smp_mb(); /* (B) */ |
8f95c90c DB |
311 | |
312 | /* | |
313 | * Avoid using task_rcu_dereference() magic as long as we are careful, | |
314 | * see comment in rcuwait_wait_event() regarding ->exit_state. | |
315 | */ | |
316 | task = rcu_dereference(w->task); | |
317 | if (task) | |
318 | wake_up_process(task); | |
319 | rcu_read_unlock(); | |
320 | } | |
321 | ||
1da177e4 LT |
322 | /* |
323 | * Determine if a process group is "orphaned", according to the POSIX | |
324 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
325 | * by terminal-generated stop signals. Newly orphaned process groups are | |
326 | * to receive a SIGHUP and a SIGCONT. | |
327 | * | |
328 | * "I ask you, have you ever known what it is to be an orphan?" | |
329 | */ | |
a0be55de IA |
330 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
331 | struct task_struct *ignored_task) | |
1da177e4 LT |
332 | { |
333 | struct task_struct *p; | |
1da177e4 | 334 | |
0475ac08 | 335 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
336 | if ((p == ignored_task) || |
337 | (p->exit_state && thread_group_empty(p)) || | |
338 | is_global_init(p->real_parent)) | |
1da177e4 | 339 | continue; |
05e83df6 | 340 | |
0475ac08 | 341 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
342 | task_session(p->real_parent) == task_session(p)) |
343 | return 0; | |
0475ac08 | 344 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
345 | |
346 | return 1; | |
1da177e4 LT |
347 | } |
348 | ||
3e7cd6c4 | 349 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
350 | { |
351 | int retval; | |
352 | ||
353 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 354 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
355 | read_unlock(&tasklist_lock); |
356 | ||
357 | return retval; | |
358 | } | |
359 | ||
961c4675 | 360 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 361 | { |
1da177e4 LT |
362 | struct task_struct *p; |
363 | ||
0475ac08 | 364 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
365 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
366 | return true; | |
0475ac08 | 367 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
368 | |
369 | return false; | |
1da177e4 LT |
370 | } |
371 | ||
f49ee505 ON |
372 | /* |
373 | * Check to see if any process groups have become orphaned as | |
374 | * a result of our exiting, and if they have any stopped jobs, | |
375 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
376 | */ | |
377 | static void | |
378 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
379 | { | |
380 | struct pid *pgrp = task_pgrp(tsk); | |
381 | struct task_struct *ignored_task = tsk; | |
382 | ||
383 | if (!parent) | |
a0be55de IA |
384 | /* exit: our father is in a different pgrp than |
385 | * we are and we were the only connection outside. | |
386 | */ | |
f49ee505 ON |
387 | parent = tsk->real_parent; |
388 | else | |
389 | /* reparent: our child is in a different pgrp than | |
390 | * we are, and it was the only connection outside. | |
391 | */ | |
392 | ignored_task = NULL; | |
393 | ||
394 | if (task_pgrp(parent) != pgrp && | |
395 | task_session(parent) == task_session(tsk) && | |
396 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
397 | has_stopped_jobs(pgrp)) { | |
398 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
399 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
400 | } | |
401 | } | |
402 | ||
f98bafa0 | 403 | #ifdef CONFIG_MEMCG |
cf475ad2 | 404 | /* |
733eda7a | 405 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 406 | */ |
cf475ad2 BS |
407 | void mm_update_next_owner(struct mm_struct *mm) |
408 | { | |
409 | struct task_struct *c, *g, *p = current; | |
410 | ||
411 | retry: | |
733eda7a KH |
412 | /* |
413 | * If the exiting or execing task is not the owner, it's | |
414 | * someone else's problem. | |
415 | */ | |
416 | if (mm->owner != p) | |
cf475ad2 | 417 | return; |
733eda7a KH |
418 | /* |
419 | * The current owner is exiting/execing and there are no other | |
420 | * candidates. Do not leave the mm pointing to a possibly | |
421 | * freed task structure. | |
422 | */ | |
423 | if (atomic_read(&mm->mm_users) <= 1) { | |
424 | mm->owner = NULL; | |
425 | return; | |
426 | } | |
cf475ad2 BS |
427 | |
428 | read_lock(&tasklist_lock); | |
429 | /* | |
430 | * Search in the children | |
431 | */ | |
432 | list_for_each_entry(c, &p->children, sibling) { | |
433 | if (c->mm == mm) | |
434 | goto assign_new_owner; | |
435 | } | |
436 | ||
437 | /* | |
438 | * Search in the siblings | |
439 | */ | |
dea33cfd | 440 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
441 | if (c->mm == mm) |
442 | goto assign_new_owner; | |
443 | } | |
444 | ||
445 | /* | |
f87fb599 | 446 | * Search through everything else, we should not get here often. |
cf475ad2 | 447 | */ |
39af1765 ON |
448 | for_each_process(g) { |
449 | if (g->flags & PF_KTHREAD) | |
450 | continue; | |
451 | for_each_thread(g, c) { | |
452 | if (c->mm == mm) | |
453 | goto assign_new_owner; | |
454 | if (c->mm) | |
455 | break; | |
456 | } | |
f87fb599 | 457 | } |
cf475ad2 | 458 | read_unlock(&tasklist_lock); |
31a78f23 BS |
459 | /* |
460 | * We found no owner yet mm_users > 1: this implies that we are | |
461 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 462 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 463 | */ |
31a78f23 | 464 | mm->owner = NULL; |
cf475ad2 BS |
465 | return; |
466 | ||
467 | assign_new_owner: | |
468 | BUG_ON(c == p); | |
469 | get_task_struct(c); | |
470 | /* | |
471 | * The task_lock protects c->mm from changing. | |
472 | * We always want mm->owner->mm == mm | |
473 | */ | |
474 | task_lock(c); | |
e5991371 HD |
475 | /* |
476 | * Delay read_unlock() till we have the task_lock() | |
477 | * to ensure that c does not slip away underneath us | |
478 | */ | |
479 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
480 | if (c->mm != mm) { |
481 | task_unlock(c); | |
482 | put_task_struct(c); | |
483 | goto retry; | |
484 | } | |
cf475ad2 BS |
485 | mm->owner = c; |
486 | task_unlock(c); | |
487 | put_task_struct(c); | |
488 | } | |
f98bafa0 | 489 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 490 | |
1da177e4 LT |
491 | /* |
492 | * Turn us into a lazy TLB process if we | |
493 | * aren't already.. | |
494 | */ | |
0039962a | 495 | static void exit_mm(void) |
1da177e4 | 496 | { |
0039962a | 497 | struct mm_struct *mm = current->mm; |
b564daf8 | 498 | struct core_state *core_state; |
1da177e4 | 499 | |
0c10e4b6 | 500 | exit_mm_release(current, mm); |
1da177e4 LT |
501 | if (!mm) |
502 | return; | |
4fe7efdb | 503 | sync_mm_rss(mm); |
1da177e4 LT |
504 | /* |
505 | * Serialize with any possible pending coredump. | |
999d9fc1 | 506 | * We must hold mmap_sem around checking core_state |
1da177e4 | 507 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 508 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
509 | * group with ->mm != NULL. |
510 | */ | |
511 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
512 | core_state = mm->core_state; |
513 | if (core_state) { | |
514 | struct core_thread self; | |
a0be55de | 515 | |
1da177e4 | 516 | up_read(&mm->mmap_sem); |
1da177e4 | 517 | |
0039962a | 518 | self.task = current; |
b564daf8 ON |
519 | self.next = xchg(&core_state->dumper.next, &self); |
520 | /* | |
521 | * Implies mb(), the result of xchg() must be visible | |
522 | * to core_state->dumper. | |
523 | */ | |
524 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
525 | complete(&core_state->startup); | |
1da177e4 | 526 | |
a94e2d40 | 527 | for (;;) { |
642fa448 | 528 | set_current_state(TASK_UNINTERRUPTIBLE); |
a94e2d40 ON |
529 | if (!self.task) /* see coredump_finish() */ |
530 | break; | |
80d26af8 | 531 | freezable_schedule(); |
a94e2d40 | 532 | } |
642fa448 | 533 | __set_current_state(TASK_RUNNING); |
1da177e4 LT |
534 | down_read(&mm->mmap_sem); |
535 | } | |
f1f10076 | 536 | mmgrab(mm); |
0039962a | 537 | BUG_ON(mm != current->active_mm); |
1da177e4 | 538 | /* more a memory barrier than a real lock */ |
0039962a DB |
539 | task_lock(current); |
540 | current->mm = NULL; | |
1da177e4 LT |
541 | up_read(&mm->mmap_sem); |
542 | enter_lazy_tlb(mm, current); | |
0039962a | 543 | task_unlock(current); |
cf475ad2 | 544 | mm_update_next_owner(mm); |
1da177e4 | 545 | mmput(mm); |
c32b3cbe | 546 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 547 | exit_oom_victim(); |
1da177e4 LT |
548 | } |
549 | ||
c9dc05bf ON |
550 | static struct task_struct *find_alive_thread(struct task_struct *p) |
551 | { | |
552 | struct task_struct *t; | |
553 | ||
554 | for_each_thread(p, t) { | |
555 | if (!(t->flags & PF_EXITING)) | |
556 | return t; | |
557 | } | |
558 | return NULL; | |
559 | } | |
560 | ||
8bca422a AV |
561 | static struct task_struct *find_child_reaper(struct task_struct *father, |
562 | struct list_head *dead) | |
1109909c ON |
563 | __releases(&tasklist_lock) |
564 | __acquires(&tasklist_lock) | |
565 | { | |
566 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
567 | struct task_struct *reaper = pid_ns->child_reaper; | |
8bca422a | 568 | struct task_struct *p, *n; |
1109909c ON |
569 | |
570 | if (likely(reaper != father)) | |
571 | return reaper; | |
572 | ||
c9dc05bf ON |
573 | reaper = find_alive_thread(father); |
574 | if (reaper) { | |
1109909c ON |
575 | pid_ns->child_reaper = reaper; |
576 | return reaper; | |
577 | } | |
578 | ||
579 | write_unlock_irq(&tasklist_lock); | |
8bca422a AV |
580 | |
581 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
582 | list_del_init(&p->ptrace_entry); | |
583 | release_task(p); | |
584 | } | |
585 | ||
1109909c ON |
586 | zap_pid_ns_processes(pid_ns); |
587 | write_lock_irq(&tasklist_lock); | |
588 | ||
589 | return father; | |
590 | } | |
591 | ||
1da177e4 | 592 | /* |
ebec18a6 LP |
593 | * When we die, we re-parent all our children, and try to: |
594 | * 1. give them to another thread in our thread group, if such a member exists | |
595 | * 2. give it to the first ancestor process which prctl'd itself as a | |
596 | * child_subreaper for its children (like a service manager) | |
597 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 598 | */ |
1109909c ON |
599 | static struct task_struct *find_new_reaper(struct task_struct *father, |
600 | struct task_struct *child_reaper) | |
1da177e4 | 601 | { |
c9dc05bf | 602 | struct task_struct *thread, *reaper; |
1da177e4 | 603 | |
c9dc05bf ON |
604 | thread = find_alive_thread(father); |
605 | if (thread) | |
950bbabb | 606 | return thread; |
1da177e4 | 607 | |
7d24e2df | 608 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 609 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 610 | /* |
175aed3f | 611 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
612 | * We can't check reaper != child_reaper to ensure we do not |
613 | * cross the namespaces, the exiting parent could be injected | |
614 | * by setns() + fork(). | |
615 | * We check pid->level, this is slightly more efficient than | |
616 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 617 | */ |
c6c70f44 ON |
618 | for (reaper = father->real_parent; |
619 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 620 | reaper = reaper->real_parent) { |
175aed3f | 621 | if (reaper == &init_task) |
ebec18a6 LP |
622 | break; |
623 | if (!reaper->signal->is_child_subreaper) | |
624 | continue; | |
c9dc05bf ON |
625 | thread = find_alive_thread(reaper); |
626 | if (thread) | |
627 | return thread; | |
ebec18a6 | 628 | } |
1da177e4 | 629 | } |
762a24be | 630 | |
1109909c | 631 | return child_reaper; |
950bbabb ON |
632 | } |
633 | ||
5dfc80be ON |
634 | /* |
635 | * Any that need to be release_task'd are put on the @dead list. | |
636 | */ | |
9cd80bbb | 637 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
638 | struct list_head *dead) |
639 | { | |
2831096e | 640 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
641 | return; |
642 | ||
abd50b39 | 643 | /* We don't want people slaying init. */ |
5dfc80be ON |
644 | p->exit_signal = SIGCHLD; |
645 | ||
646 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 647 | if (!p->ptrace && |
5dfc80be | 648 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 649 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 650 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 651 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
652 | } |
653 | } | |
654 | ||
655 | kill_orphaned_pgrp(p, father); | |
656 | } | |
657 | ||
482a3767 ON |
658 | /* |
659 | * This does two things: | |
660 | * | |
661 | * A. Make init inherit all the child processes | |
662 | * B. Check to see if any process groups have become orphaned | |
663 | * as a result of our exiting, and if they have any stopped | |
664 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
665 | */ | |
666 | static void forget_original_parent(struct task_struct *father, | |
667 | struct list_head *dead) | |
1da177e4 | 668 | { |
482a3767 | 669 | struct task_struct *p, *t, *reaper; |
762a24be | 670 | |
7c8bd232 | 671 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 672 | exit_ptrace(father, dead); |
f470021a | 673 | |
7c8bd232 | 674 | /* Can drop and reacquire tasklist_lock */ |
8bca422a | 675 | reaper = find_child_reaper(father, dead); |
ad9e206a | 676 | if (list_empty(&father->children)) |
482a3767 | 677 | return; |
1109909c ON |
678 | |
679 | reaper = find_new_reaper(father, reaper); | |
2831096e | 680 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 681 | for_each_thread(p, t) { |
9cd80bbb | 682 | t->real_parent = reaper; |
57a05918 ON |
683 | BUG_ON((!t->ptrace) != (t->parent == father)); |
684 | if (likely(!t->ptrace)) | |
9cd80bbb | 685 | t->parent = t->real_parent; |
9cd80bbb ON |
686 | if (t->pdeath_signal) |
687 | group_send_sig_info(t->pdeath_signal, | |
688 | SEND_SIG_NOINFO, t); | |
57a05918 | 689 | } |
2831096e ON |
690 | /* |
691 | * If this is a threaded reparent there is no need to | |
692 | * notify anyone anything has happened. | |
693 | */ | |
694 | if (!same_thread_group(reaper, father)) | |
482a3767 | 695 | reparent_leader(father, p, dead); |
1da177e4 | 696 | } |
2831096e | 697 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
698 | } |
699 | ||
700 | /* | |
701 | * Send signals to all our closest relatives so that they know | |
702 | * to properly mourn us.. | |
703 | */ | |
821c7de7 | 704 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 705 | { |
53c8f9f1 | 706 | bool autoreap; |
482a3767 ON |
707 | struct task_struct *p, *n; |
708 | LIST_HEAD(dead); | |
1da177e4 | 709 | |
762a24be | 710 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
711 | forget_original_parent(tsk, &dead); |
712 | ||
821c7de7 ON |
713 | if (group_dead) |
714 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 715 | |
45cdf5cc ON |
716 | if (unlikely(tsk->ptrace)) { |
717 | int sig = thread_group_leader(tsk) && | |
718 | thread_group_empty(tsk) && | |
719 | !ptrace_reparented(tsk) ? | |
720 | tsk->exit_signal : SIGCHLD; | |
721 | autoreap = do_notify_parent(tsk, sig); | |
722 | } else if (thread_group_leader(tsk)) { | |
723 | autoreap = thread_group_empty(tsk) && | |
724 | do_notify_parent(tsk, tsk->exit_signal); | |
725 | } else { | |
726 | autoreap = true; | |
727 | } | |
1da177e4 | 728 | |
53c8f9f1 | 729 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
6c66e7db ON |
730 | if (tsk->exit_state == EXIT_DEAD) |
731 | list_add(&tsk->ptrace_entry, &dead); | |
1da177e4 | 732 | |
9c339168 ON |
733 | /* mt-exec, de_thread() is waiting for group leader */ |
734 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 735 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
736 | write_unlock_irq(&tasklist_lock); |
737 | ||
482a3767 ON |
738 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
739 | list_del_init(&p->ptrace_entry); | |
740 | release_task(p); | |
741 | } | |
1da177e4 LT |
742 | } |
743 | ||
e18eecb8 JD |
744 | #ifdef CONFIG_DEBUG_STACK_USAGE |
745 | static void check_stack_usage(void) | |
746 | { | |
747 | static DEFINE_SPINLOCK(low_water_lock); | |
748 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
749 | unsigned long free; |
750 | ||
7c9f8861 | 751 | free = stack_not_used(current); |
e18eecb8 JD |
752 | |
753 | if (free >= lowest_to_date) | |
754 | return; | |
755 | ||
756 | spin_lock(&low_water_lock); | |
757 | if (free < lowest_to_date) { | |
627393d4 | 758 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 759 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
760 | lowest_to_date = free; |
761 | } | |
762 | spin_unlock(&low_water_lock); | |
763 | } | |
764 | #else | |
765 | static inline void check_stack_usage(void) {} | |
766 | #endif | |
767 | ||
9af6528e | 768 | void __noreturn do_exit(long code) |
1da177e4 LT |
769 | { |
770 | struct task_struct *tsk = current; | |
771 | int group_dead; | |
772 | ||
773 | profile_task_exit(tsk); | |
5c9a8750 | 774 | kcov_task_exit(tsk); |
1da177e4 | 775 | |
73c10101 | 776 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 777 | |
1da177e4 LT |
778 | if (unlikely(in_interrupt())) |
779 | panic("Aiee, killing interrupt handler!"); | |
780 | if (unlikely(!tsk->pid)) | |
781 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 782 | |
33dd94ae NE |
783 | /* |
784 | * If do_exit is called because this processes oopsed, it's possible | |
785 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
786 | * continuing. Amongst other possible reasons, this is to prevent | |
787 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
788 | * kernel address. | |
789 | */ | |
790 | set_fs(USER_DS); | |
791 | ||
a288eecc | 792 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 793 | |
e0e81739 DH |
794 | validate_creds_for_do_exit(tsk); |
795 | ||
df164db5 AN |
796 | /* |
797 | * We're taking recursive faults here in do_exit. Safest is to just | |
798 | * leave this task alone and wait for reboot. | |
799 | */ | |
800 | if (unlikely(tsk->flags & PF_EXITING)) { | |
a0be55de | 801 | pr_alert("Fixing recursive fault but reboot is needed!\n"); |
f34df84f | 802 | futex_exit_recursive(tsk); |
df164db5 AN |
803 | set_current_state(TASK_UNINTERRUPTIBLE); |
804 | schedule(); | |
805 | } | |
806 | ||
d12619b5 | 807 | exit_signals(tsk); /* sets PF_EXITING */ |
1da177e4 | 808 | |
1dc0fffc | 809 | if (unlikely(in_atomic())) { |
a0be55de IA |
810 | pr_info("note: %s[%d] exited with preempt_count %d\n", |
811 | current->comm, task_pid_nr(current), | |
812 | preempt_count()); | |
1dc0fffc PZ |
813 | preempt_count_set(PREEMPT_ENABLED); |
814 | } | |
1da177e4 | 815 | |
48d212a2 LT |
816 | /* sync mm's RSS info before statistics gathering */ |
817 | if (tsk->mm) | |
818 | sync_mm_rss(tsk->mm); | |
51229b49 | 819 | acct_update_integrals(tsk); |
1da177e4 | 820 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 821 | if (group_dead) { |
9898ca0d | 822 | /* |
823 | * If the last thread of global init has exited, panic | |
824 | * immediately to get a useable coredump. | |
825 | */ | |
826 | if (unlikely(is_global_init(tsk))) | |
827 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
828 | tsk->signal->group_exit_code ?: (int)code); | |
829 | ||
baa73d9e | 830 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 831 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 832 | exit_itimers(tsk->signal); |
baa73d9e | 833 | #endif |
1f10206c JP |
834 | if (tsk->mm) |
835 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 836 | } |
f6ec29a4 | 837 | acct_collect(code, group_dead); |
522ed776 MT |
838 | if (group_dead) |
839 | tty_audit_exit(); | |
a4ff8dba | 840 | audit_free(tsk); |
115085ea | 841 | |
48d212a2 | 842 | tsk->exit_code = code; |
115085ea | 843 | taskstats_exit(tsk, group_dead); |
c757249a | 844 | |
0039962a | 845 | exit_mm(); |
1da177e4 | 846 | |
0e464814 | 847 | if (group_dead) |
f6ec29a4 | 848 | acct_process(); |
0a16b607 MD |
849 | trace_sched_process_exit(tsk); |
850 | ||
1da177e4 | 851 | exit_sem(tsk); |
b34a6b1d | 852 | exit_shm(tsk); |
1ec7f1dd AV |
853 | exit_files(tsk); |
854 | exit_fs(tsk); | |
c39df5fa ON |
855 | if (group_dead) |
856 | disassociate_ctty(1); | |
8aac6270 | 857 | exit_task_namespaces(tsk); |
ed3e694d | 858 | exit_task_work(tsk); |
e6464694 | 859 | exit_thread(tsk); |
0b3fcf17 SE |
860 | |
861 | /* | |
862 | * Flush inherited counters to the parent - before the parent | |
863 | * gets woken up by child-exit notifications. | |
864 | * | |
865 | * because of cgroup mode, must be called before cgroup_exit() | |
866 | */ | |
867 | perf_event_exit_task(tsk); | |
868 | ||
8e5bfa8c | 869 | sched_autogroup_exit_task(tsk); |
1ec41830 | 870 | cgroup_exit(tsk); |
1da177e4 | 871 | |
24f1e32c FW |
872 | /* |
873 | * FIXME: do that only when needed, using sched_exit tracepoint | |
874 | */ | |
7c8df286 | 875 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 876 | |
ccdd29ff | 877 | exit_tasks_rcu_start(); |
821c7de7 | 878 | exit_notify(tsk, group_dead); |
ef982393 | 879 | proc_exit_connector(tsk); |
c11600e4 | 880 | mpol_put_task_policy(tsk); |
42b2dd0a | 881 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
882 | if (unlikely(current->pi_state_cache)) |
883 | kfree(current->pi_state_cache); | |
42b2dd0a | 884 | #endif |
de5097c2 | 885 | /* |
9a11b49a | 886 | * Make sure we are holding no locks: |
de5097c2 | 887 | */ |
1b1d2fb4 | 888 | debug_check_no_locks_held(); |
1da177e4 | 889 | |
afc847b7 | 890 | if (tsk->io_context) |
b69f2292 | 891 | exit_io_context(tsk); |
afc847b7 | 892 | |
b92ce558 | 893 | if (tsk->splice_pipe) |
4b8a8f1e | 894 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 895 | |
5640f768 ED |
896 | if (tsk->task_frag.page) |
897 | put_page(tsk->task_frag.page); | |
898 | ||
e0e81739 DH |
899 | validate_creds_for_do_exit(tsk); |
900 | ||
4bcb8232 | 901 | check_stack_usage(); |
7407251a | 902 | preempt_disable(); |
54848d73 WF |
903 | if (tsk->nr_dirtied) |
904 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 905 | exit_rcu(); |
ccdd29ff | 906 | exit_tasks_rcu_finish(); |
b5740f4b | 907 | |
b09be676 | 908 | lockdep_free_task(tsk); |
9af6528e | 909 | do_task_dead(); |
1da177e4 | 910 | } |
012914da RA |
911 | EXPORT_SYMBOL_GPL(do_exit); |
912 | ||
9402c95f | 913 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
914 | { |
915 | if (comp) | |
916 | complete(comp); | |
55a101f8 | 917 | |
1da177e4 LT |
918 | do_exit(code); |
919 | } | |
1da177e4 LT |
920 | EXPORT_SYMBOL(complete_and_exit); |
921 | ||
754fe8d2 | 922 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
923 | { |
924 | do_exit((error_code&0xff)<<8); | |
925 | } | |
926 | ||
1da177e4 LT |
927 | /* |
928 | * Take down every thread in the group. This is called by fatal signals | |
929 | * as well as by sys_exit_group (below). | |
930 | */ | |
9402c95f | 931 | void |
1da177e4 LT |
932 | do_group_exit(int exit_code) |
933 | { | |
bfc4b089 ON |
934 | struct signal_struct *sig = current->signal; |
935 | ||
1da177e4 LT |
936 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
937 | ||
bfc4b089 ON |
938 | if (signal_group_exit(sig)) |
939 | exit_code = sig->group_exit_code; | |
1da177e4 | 940 | else if (!thread_group_empty(current)) { |
1da177e4 | 941 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 942 | |
1da177e4 | 943 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 944 | if (signal_group_exit(sig)) |
1da177e4 LT |
945 | /* Another thread got here before we took the lock. */ |
946 | exit_code = sig->group_exit_code; | |
947 | else { | |
1da177e4 | 948 | sig->group_exit_code = exit_code; |
ed5d2cac | 949 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
950 | zap_other_threads(current); |
951 | } | |
952 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
953 | } |
954 | ||
955 | do_exit(exit_code); | |
956 | /* NOTREACHED */ | |
957 | } | |
958 | ||
959 | /* | |
960 | * this kills every thread in the thread group. Note that any externally | |
961 | * wait4()-ing process will get the correct exit code - even if this | |
962 | * thread is not the thread group leader. | |
963 | */ | |
754fe8d2 | 964 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
965 | { |
966 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
967 | /* NOTREACHED */ |
968 | return 0; | |
1da177e4 LT |
969 | } |
970 | ||
67d7ddde AV |
971 | struct waitid_info { |
972 | pid_t pid; | |
973 | uid_t uid; | |
974 | int status; | |
975 | int cause; | |
976 | }; | |
977 | ||
9e8ae01d ON |
978 | struct wait_opts { |
979 | enum pid_type wo_type; | |
9e8ae01d | 980 | int wo_flags; |
e1eb1ebc | 981 | struct pid *wo_pid; |
9e8ae01d | 982 | |
67d7ddde | 983 | struct waitid_info *wo_info; |
359566fa | 984 | int wo_stat; |
ce72a16f | 985 | struct rusage *wo_rusage; |
9e8ae01d | 986 | |
ac6424b9 | 987 | wait_queue_entry_t child_wait; |
9e8ae01d ON |
988 | int notask_error; |
989 | }; | |
990 | ||
989264f4 ON |
991 | static inline |
992 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 993 | { |
989264f4 ON |
994 | if (type != PIDTYPE_PID) |
995 | task = task->group_leader; | |
996 | return task->pids[type].pid; | |
161550d7 EB |
997 | } |
998 | ||
989264f4 | 999 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1000 | { |
5c01ba49 ON |
1001 | return wo->wo_type == PIDTYPE_MAX || |
1002 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1003 | } | |
1da177e4 | 1004 | |
bf959931 ON |
1005 | static int |
1006 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
1007 | { |
1008 | if (!eligible_pid(wo, p)) | |
1009 | return 0; | |
bf959931 ON |
1010 | |
1011 | /* | |
1012 | * Wait for all children (clone and not) if __WALL is set or | |
1013 | * if it is traced by us. | |
1014 | */ | |
1015 | if (ptrace || (wo->wo_flags & __WALL)) | |
1016 | return 1; | |
1017 | ||
1018 | /* | |
1019 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
1020 | * otherwise, wait for non-clone children *only*. | |
1021 | * | |
1022 | * Note: a "clone" child here is one that reports to its parent | |
1023 | * using a signal other than SIGCHLD, or a non-leader thread which | |
1024 | * we can only see if it is traced by us. | |
1025 | */ | |
1026 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 1027 | return 0; |
1da177e4 | 1028 | |
14dd0b81 | 1029 | return 1; |
1da177e4 LT |
1030 | } |
1031 | ||
1da177e4 LT |
1032 | /* |
1033 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1034 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1035 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1036 | * released the lock and the system call should return. | |
1037 | */ | |
9e8ae01d | 1038 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1039 | { |
67d7ddde | 1040 | int state, status; |
6c5f3e7b | 1041 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1042 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
67d7ddde | 1043 | struct waitid_info *infop; |
1da177e4 | 1044 | |
9e8ae01d | 1045 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1046 | return 0; |
1047 | ||
9e8ae01d | 1048 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
76d9871e | 1049 | status = p->exit_code; |
1da177e4 LT |
1050 | get_task_struct(p); |
1051 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1052 | sched_annotate_sleep(); |
e61a2502 AV |
1053 | if (wo->wo_rusage) |
1054 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1055 | put_task_struct(p); |
76d9871e | 1056 | goto out_info; |
1da177e4 | 1057 | } |
1da177e4 | 1058 | /* |
abd50b39 | 1059 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1060 | */ |
f6507f83 ON |
1061 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1062 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1063 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1064 | return 0; |
986094df ON |
1065 | /* |
1066 | * We own this thread, nobody else can reap it. | |
1067 | */ | |
1068 | read_unlock(&tasklist_lock); | |
1069 | sched_annotate_sleep(); | |
f6507f83 | 1070 | |
befca967 | 1071 | /* |
f6507f83 | 1072 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1073 | */ |
f6507f83 | 1074 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1075 | struct signal_struct *sig = p->signal; |
1076 | struct signal_struct *psig = current->signal; | |
1f10206c | 1077 | unsigned long maxrss; |
5613fda9 | 1078 | u64 tgutime, tgstime; |
3795e161 | 1079 | |
1da177e4 LT |
1080 | /* |
1081 | * The resource counters for the group leader are in its | |
1082 | * own task_struct. Those for dead threads in the group | |
1083 | * are in its signal_struct, as are those for the child | |
1084 | * processes it has previously reaped. All these | |
1085 | * accumulate in the parent's signal_struct c* fields. | |
1086 | * | |
1087 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1088 | * p->signal fields because the whole thread group is dead |
1089 | * and nobody can change them. | |
1090 | * | |
1091 | * psig->stats_lock also protects us from our sub-theads | |
1092 | * which can reap other children at the same time. Until | |
1093 | * we change k_getrusage()-like users to rely on this lock | |
1094 | * we have to take ->siglock as well. | |
0cf55e1e | 1095 | * |
a0be55de IA |
1096 | * We use thread_group_cputime_adjusted() to get times for |
1097 | * the thread group, which consolidates times for all threads | |
1098 | * in the group including the group leader. | |
1da177e4 | 1099 | */ |
e80d0a1a | 1100 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1101 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1102 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1103 | psig->cutime += tgutime + sig->cutime; |
1104 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1105 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1106 | psig->cmin_flt += |
1107 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1108 | psig->cmaj_flt += | |
1109 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1110 | psig->cnvcsw += | |
1111 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1112 | psig->cnivcsw += | |
1113 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1114 | psig->cinblock += |
1115 | task_io_get_inblock(p) + | |
1116 | sig->inblock + sig->cinblock; | |
1117 | psig->coublock += | |
1118 | task_io_get_oublock(p) + | |
1119 | sig->oublock + sig->coublock; | |
1f10206c JP |
1120 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1121 | if (psig->cmaxrss < maxrss) | |
1122 | psig->cmaxrss = maxrss; | |
5995477a AR |
1123 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1124 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1125 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1126 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1127 | } |
1128 | ||
ce72a16f AV |
1129 | if (wo->wo_rusage) |
1130 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
1da177e4 LT |
1131 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1132 | ? p->signal->group_exit_code : p->exit_code; | |
359566fa | 1133 | wo->wo_stat = status; |
2f4e6e2a | 1134 | |
b4360690 | 1135 | if (state == EXIT_TRACE) { |
1da177e4 | 1136 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1137 | /* We dropped tasklist, ptracer could die and untrace */ |
1138 | ptrace_unlink(p); | |
b4360690 ON |
1139 | |
1140 | /* If parent wants a zombie, don't release it now */ | |
1141 | state = EXIT_ZOMBIE; | |
1142 | if (do_notify_parent(p, p->exit_signal)) | |
1143 | state = EXIT_DEAD; | |
abd50b39 | 1144 | p->exit_state = state; |
1da177e4 LT |
1145 | write_unlock_irq(&tasklist_lock); |
1146 | } | |
abd50b39 | 1147 | if (state == EXIT_DEAD) |
1da177e4 | 1148 | release_task(p); |
2f4e6e2a | 1149 | |
76d9871e AV |
1150 | out_info: |
1151 | infop = wo->wo_info; | |
1152 | if (infop) { | |
1153 | if ((status & 0x7f) == 0) { | |
1154 | infop->cause = CLD_EXITED; | |
1155 | infop->status = status >> 8; | |
1156 | } else { | |
1157 | infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1158 | infop->status = status & 0x7f; | |
1159 | } | |
1160 | infop->pid = pid; | |
1161 | infop->uid = uid; | |
1162 | } | |
1163 | ||
67d7ddde | 1164 | return pid; |
1da177e4 LT |
1165 | } |
1166 | ||
90bc8d8b ON |
1167 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1168 | { | |
1169 | if (ptrace) { | |
570ac933 | 1170 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1171 | return &p->exit_code; |
1172 | } else { | |
1173 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1174 | return &p->signal->group_exit_code; | |
1175 | } | |
1176 | return NULL; | |
1177 | } | |
1178 | ||
19e27463 TH |
1179 | /** |
1180 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1181 | * @wo: wait options | |
1182 | * @ptrace: is the wait for ptrace | |
1183 | * @p: task to wait for | |
1184 | * | |
1185 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1186 | * | |
1187 | * CONTEXT: | |
1188 | * read_lock(&tasklist_lock), which is released if return value is | |
1189 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1190 | * | |
1191 | * RETURNS: | |
1192 | * 0 if wait condition didn't exist and search for other wait conditions | |
1193 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1194 | * success, implies that tasklist_lock is released and wait condition | |
1195 | * search should terminate. | |
1da177e4 | 1196 | */ |
9e8ae01d ON |
1197 | static int wait_task_stopped(struct wait_opts *wo, |
1198 | int ptrace, struct task_struct *p) | |
1da177e4 | 1199 | { |
67d7ddde AV |
1200 | struct waitid_info *infop; |
1201 | int exit_code, *p_code, why; | |
ee7c82da | 1202 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1203 | pid_t pid; |
1da177e4 | 1204 | |
47918025 ON |
1205 | /* |
1206 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1207 | */ | |
9e8ae01d | 1208 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1209 | return 0; |
1210 | ||
19e27463 TH |
1211 | if (!task_stopped_code(p, ptrace)) |
1212 | return 0; | |
1213 | ||
ee7c82da ON |
1214 | exit_code = 0; |
1215 | spin_lock_irq(&p->sighand->siglock); | |
1216 | ||
90bc8d8b ON |
1217 | p_code = task_stopped_code(p, ptrace); |
1218 | if (unlikely(!p_code)) | |
ee7c82da ON |
1219 | goto unlock_sig; |
1220 | ||
90bc8d8b | 1221 | exit_code = *p_code; |
ee7c82da ON |
1222 | if (!exit_code) |
1223 | goto unlock_sig; | |
1224 | ||
9e8ae01d | 1225 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1226 | *p_code = 0; |
ee7c82da | 1227 | |
8ca937a6 | 1228 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1229 | unlock_sig: |
1230 | spin_unlock_irq(&p->sighand->siglock); | |
1231 | if (!exit_code) | |
1da177e4 LT |
1232 | return 0; |
1233 | ||
1234 | /* | |
1235 | * Now we are pretty sure this task is interesting. | |
1236 | * Make sure it doesn't get reaped out from under us while we | |
1237 | * give up the lock and then examine it below. We don't want to | |
1238 | * keep holding onto the tasklist_lock while we call getrusage and | |
1239 | * possibly take page faults for user memory. | |
1240 | */ | |
1241 | get_task_struct(p); | |
6c5f3e7b | 1242 | pid = task_pid_vnr(p); |
f470021a | 1243 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1244 | read_unlock(&tasklist_lock); |
1029a2b5 | 1245 | sched_annotate_sleep(); |
e61a2502 AV |
1246 | if (wo->wo_rusage) |
1247 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1248 | put_task_struct(p); |
1da177e4 | 1249 | |
bb380ec3 AV |
1250 | if (likely(!(wo->wo_flags & WNOWAIT))) |
1251 | wo->wo_stat = (exit_code << 8) | 0x7f; | |
1da177e4 | 1252 | |
9e8ae01d | 1253 | infop = wo->wo_info; |
67d7ddde AV |
1254 | if (infop) { |
1255 | infop->cause = why; | |
1256 | infop->status = exit_code; | |
1257 | infop->pid = pid; | |
1258 | infop->uid = uid; | |
1259 | } | |
67d7ddde | 1260 | return pid; |
1da177e4 LT |
1261 | } |
1262 | ||
1263 | /* | |
1264 | * Handle do_wait work for one task in a live, non-stopped state. | |
1265 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1266 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1267 | * released the lock and the system call should return. | |
1268 | */ | |
9e8ae01d | 1269 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1270 | { |
bb380ec3 | 1271 | struct waitid_info *infop; |
1da177e4 LT |
1272 | pid_t pid; |
1273 | uid_t uid; | |
1274 | ||
9e8ae01d | 1275 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1276 | return 0; |
1277 | ||
1da177e4 LT |
1278 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1279 | return 0; | |
1280 | ||
1281 | spin_lock_irq(&p->sighand->siglock); | |
1282 | /* Re-check with the lock held. */ | |
1283 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1284 | spin_unlock_irq(&p->sighand->siglock); | |
1285 | return 0; | |
1286 | } | |
9e8ae01d | 1287 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1288 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1289 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1290 | spin_unlock_irq(&p->sighand->siglock); |
1291 | ||
6c5f3e7b | 1292 | pid = task_pid_vnr(p); |
1da177e4 LT |
1293 | get_task_struct(p); |
1294 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1295 | sched_annotate_sleep(); |
e61a2502 AV |
1296 | if (wo->wo_rusage) |
1297 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1298 | put_task_struct(p); |
1da177e4 | 1299 | |
bb380ec3 AV |
1300 | infop = wo->wo_info; |
1301 | if (!infop) { | |
359566fa | 1302 | wo->wo_stat = 0xffff; |
1da177e4 | 1303 | } else { |
bb380ec3 AV |
1304 | infop->cause = CLD_CONTINUED; |
1305 | infop->pid = pid; | |
1306 | infop->uid = uid; | |
1307 | infop->status = SIGCONT; | |
1da177e4 | 1308 | } |
bb380ec3 | 1309 | return pid; |
1da177e4 LT |
1310 | } |
1311 | ||
98abed02 RM |
1312 | /* |
1313 | * Consider @p for a wait by @parent. | |
1314 | * | |
9e8ae01d | 1315 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1316 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1317 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1318 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1319 | * or still -ECHILD. |
98abed02 | 1320 | */ |
b6e763f0 ON |
1321 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1322 | struct task_struct *p) | |
98abed02 | 1323 | { |
3245d6ac ON |
1324 | /* |
1325 | * We can race with wait_task_zombie() from another thread. | |
1326 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1327 | * can't confuse the checks below. | |
1328 | */ | |
6aa7de05 | 1329 | int exit_state = READ_ONCE(p->exit_state); |
b3ab0316 ON |
1330 | int ret; |
1331 | ||
3245d6ac | 1332 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1333 | return 0; |
1334 | ||
bf959931 | 1335 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1336 | if (!ret) |
98abed02 RM |
1337 | return ret; |
1338 | ||
3245d6ac | 1339 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1340 | /* |
abd50b39 ON |
1341 | * ptrace == 0 means we are the natural parent. In this case |
1342 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1343 | */ |
abd50b39 | 1344 | if (likely(!ptrace)) |
50b8d257 | 1345 | wo->notask_error = 0; |
823b018e | 1346 | return 0; |
50b8d257 | 1347 | } |
823b018e | 1348 | |
377d75da ON |
1349 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1350 | /* | |
1351 | * If it is traced by its real parent's group, just pretend | |
1352 | * the caller is ptrace_do_wait() and reap this child if it | |
1353 | * is zombie. | |
1354 | * | |
1355 | * This also hides group stop state from real parent; otherwise | |
1356 | * a single stop can be reported twice as group and ptrace stop. | |
1357 | * If a ptracer wants to distinguish these two events for its | |
1358 | * own children it should create a separate process which takes | |
1359 | * the role of real parent. | |
1360 | */ | |
1361 | if (!ptrace_reparented(p)) | |
1362 | ptrace = 1; | |
1363 | } | |
1364 | ||
45cb24a1 | 1365 | /* slay zombie? */ |
3245d6ac | 1366 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1367 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1368 | if (!delay_group_leader(p)) { |
1369 | /* | |
1370 | * A zombie ptracee is only visible to its ptracer. | |
1371 | * Notification and reaping will be cascaded to the | |
1372 | * real parent when the ptracer detaches. | |
1373 | */ | |
1374 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1375 | return wait_task_zombie(wo, p); | |
1376 | } | |
98abed02 | 1377 | |
f470021a | 1378 | /* |
9b84cca2 TH |
1379 | * Allow access to stopped/continued state via zombie by |
1380 | * falling through. Clearing of notask_error is complex. | |
1381 | * | |
1382 | * When !@ptrace: | |
1383 | * | |
1384 | * If WEXITED is set, notask_error should naturally be | |
1385 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1386 | * so, if there are live subthreads, there are events to | |
1387 | * wait for. If all subthreads are dead, it's still safe | |
1388 | * to clear - this function will be called again in finite | |
1389 | * amount time once all the subthreads are released and | |
1390 | * will then return without clearing. | |
1391 | * | |
1392 | * When @ptrace: | |
1393 | * | |
1394 | * Stopped state is per-task and thus can't change once the | |
1395 | * target task dies. Only continued and exited can happen. | |
1396 | * Clear notask_error if WCONTINUED | WEXITED. | |
1397 | */ | |
1398 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1399 | wo->notask_error = 0; | |
1400 | } else { | |
1401 | /* | |
1402 | * @p is alive and it's gonna stop, continue or exit, so | |
1403 | * there always is something to wait for. | |
f470021a | 1404 | */ |
9e8ae01d | 1405 | wo->notask_error = 0; |
f470021a RM |
1406 | } |
1407 | ||
98abed02 | 1408 | /* |
45cb24a1 TH |
1409 | * Wait for stopped. Depending on @ptrace, different stopped state |
1410 | * is used and the two don't interact with each other. | |
98abed02 | 1411 | */ |
19e27463 TH |
1412 | ret = wait_task_stopped(wo, ptrace, p); |
1413 | if (ret) | |
1414 | return ret; | |
98abed02 RM |
1415 | |
1416 | /* | |
45cb24a1 TH |
1417 | * Wait for continued. There's only one continued state and the |
1418 | * ptracer can consume it which can confuse the real parent. Don't | |
1419 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1420 | */ |
9e8ae01d | 1421 | return wait_task_continued(wo, p); |
98abed02 RM |
1422 | } |
1423 | ||
1424 | /* | |
1425 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1426 | * | |
9e8ae01d | 1427 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1428 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1429 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1430 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1431 | * or still -ECHILD. |
98abed02 | 1432 | */ |
9e8ae01d | 1433 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1434 | { |
1435 | struct task_struct *p; | |
1436 | ||
1437 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1438 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1439 | |
9cd80bbb ON |
1440 | if (ret) |
1441 | return ret; | |
98abed02 RM |
1442 | } |
1443 | ||
1444 | return 0; | |
1445 | } | |
1446 | ||
9e8ae01d | 1447 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1448 | { |
1449 | struct task_struct *p; | |
1450 | ||
f470021a | 1451 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1452 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1453 | |
f470021a | 1454 | if (ret) |
98abed02 | 1455 | return ret; |
98abed02 RM |
1456 | } |
1457 | ||
1458 | return 0; | |
1459 | } | |
1460 | ||
ac6424b9 | 1461 | static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode, |
0b7570e7 ON |
1462 | int sync, void *key) |
1463 | { | |
1464 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1465 | child_wait); | |
1466 | struct task_struct *p = key; | |
1467 | ||
5c01ba49 | 1468 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1469 | return 0; |
1470 | ||
b4fe5182 ON |
1471 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1472 | return 0; | |
1473 | ||
0b7570e7 ON |
1474 | return default_wake_function(wait, mode, sync, key); |
1475 | } | |
1476 | ||
a7f0765e ON |
1477 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1478 | { | |
0b7570e7 ON |
1479 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1480 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1481 | } |
1482 | ||
9e8ae01d | 1483 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1484 | { |
1da177e4 | 1485 | struct task_struct *tsk; |
98abed02 | 1486 | int retval; |
1da177e4 | 1487 | |
9e8ae01d | 1488 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1489 | |
0b7570e7 ON |
1490 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1491 | wo->child_wait.private = current; | |
1492 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1493 | repeat: |
98abed02 | 1494 | /* |
3da56d16 | 1495 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1496 | * We will clear ->notask_error to zero if we see any child that |
1497 | * might later match our criteria, even if we are not able to reap | |
1498 | * it yet. | |
98abed02 | 1499 | */ |
64a16caf | 1500 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1501 | if ((wo->wo_type < PIDTYPE_MAX) && |
1502 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1503 | goto notask; |
161550d7 | 1504 | |
f95d39d1 | 1505 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1506 | read_lock(&tasklist_lock); |
1507 | tsk = current; | |
1508 | do { | |
64a16caf ON |
1509 | retval = do_wait_thread(wo, tsk); |
1510 | if (retval) | |
1511 | goto end; | |
9e8ae01d | 1512 | |
64a16caf ON |
1513 | retval = ptrace_do_wait(wo, tsk); |
1514 | if (retval) | |
98abed02 | 1515 | goto end; |
98abed02 | 1516 | |
9e8ae01d | 1517 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1518 | break; |
a3f6dfb7 | 1519 | } while_each_thread(current, tsk); |
1da177e4 | 1520 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1521 | |
64a16caf | 1522 | notask: |
9e8ae01d ON |
1523 | retval = wo->notask_error; |
1524 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1525 | retval = -ERESTARTSYS; |
98abed02 RM |
1526 | if (!signal_pending(current)) { |
1527 | schedule(); | |
1528 | goto repeat; | |
1529 | } | |
1da177e4 | 1530 | } |
1da177e4 | 1531 | end: |
f95d39d1 | 1532 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1533 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1534 | return retval; |
1535 | } | |
1536 | ||
67d7ddde | 1537 | static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, |
ce72a16f | 1538 | int options, struct rusage *ru) |
1da177e4 | 1539 | { |
9e8ae01d | 1540 | struct wait_opts wo; |
161550d7 EB |
1541 | struct pid *pid = NULL; |
1542 | enum pid_type type; | |
1da177e4 LT |
1543 | long ret; |
1544 | ||
91c4e8ea ON |
1545 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1546 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1547 | return -EINVAL; |
1548 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1549 | return -EINVAL; | |
1550 | ||
1551 | switch (which) { | |
1552 | case P_ALL: | |
161550d7 | 1553 | type = PIDTYPE_MAX; |
1da177e4 LT |
1554 | break; |
1555 | case P_PID: | |
161550d7 EB |
1556 | type = PIDTYPE_PID; |
1557 | if (upid <= 0) | |
1da177e4 LT |
1558 | return -EINVAL; |
1559 | break; | |
1560 | case P_PGID: | |
161550d7 EB |
1561 | type = PIDTYPE_PGID; |
1562 | if (upid <= 0) | |
1da177e4 | 1563 | return -EINVAL; |
1da177e4 LT |
1564 | break; |
1565 | default: | |
1566 | return -EINVAL; | |
1567 | } | |
1568 | ||
161550d7 EB |
1569 | if (type < PIDTYPE_MAX) |
1570 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1571 | |
1572 | wo.wo_type = type; | |
1573 | wo.wo_pid = pid; | |
1574 | wo.wo_flags = options; | |
1575 | wo.wo_info = infop; | |
9e8ae01d ON |
1576 | wo.wo_rusage = ru; |
1577 | ret = do_wait(&wo); | |
dfe16dfa | 1578 | |
161550d7 | 1579 | put_pid(pid); |
1da177e4 LT |
1580 | return ret; |
1581 | } | |
1582 | ||
ce72a16f AV |
1583 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1584 | infop, int, options, struct rusage __user *, ru) | |
1585 | { | |
1586 | struct rusage r; | |
67d7ddde AV |
1587 | struct waitid_info info = {.status = 0}; |
1588 | long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL); | |
634a8160 | 1589 | int signo = 0; |
6c85501f | 1590 | |
634a8160 AV |
1591 | if (err > 0) { |
1592 | signo = SIGCHLD; | |
1593 | err = 0; | |
ce72a16f AV |
1594 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) |
1595 | return -EFAULT; | |
1596 | } | |
67d7ddde AV |
1597 | if (!infop) |
1598 | return err; | |
1599 | ||
96ca579a | 1600 | if (!access_ok(VERIFY_WRITE, infop, sizeof(*infop))) |
1c9fec47 | 1601 | return -EFAULT; |
96ca579a | 1602 | |
4c48abe9 | 1603 | user_access_begin(); |
634a8160 | 1604 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1605 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1606 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1607 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1608 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1609 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
1610 | user_access_end(); | |
ce72a16f | 1611 | return err; |
4c48abe9 AV |
1612 | Efault: |
1613 | user_access_end(); | |
1614 | return -EFAULT; | |
ce72a16f AV |
1615 | } |
1616 | ||
92ebce5a AV |
1617 | long kernel_wait4(pid_t upid, int __user *stat_addr, int options, |
1618 | struct rusage *ru) | |
1da177e4 | 1619 | { |
9e8ae01d | 1620 | struct wait_opts wo; |
161550d7 EB |
1621 | struct pid *pid = NULL; |
1622 | enum pid_type type; | |
1da177e4 LT |
1623 | long ret; |
1624 | ||
1625 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1626 | __WNOTHREAD|__WCLONE|__WALL)) | |
1627 | return -EINVAL; | |
161550d7 | 1628 | |
dd83c161 | 1629 | /* -INT_MIN is not defined */ |
1630 | if (upid == INT_MIN) | |
1631 | return -ESRCH; | |
1632 | ||
161550d7 EB |
1633 | if (upid == -1) |
1634 | type = PIDTYPE_MAX; | |
1635 | else if (upid < 0) { | |
1636 | type = PIDTYPE_PGID; | |
1637 | pid = find_get_pid(-upid); | |
1638 | } else if (upid == 0) { | |
1639 | type = PIDTYPE_PGID; | |
2ae448ef | 1640 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1641 | } else /* upid > 0 */ { |
1642 | type = PIDTYPE_PID; | |
1643 | pid = find_get_pid(upid); | |
1644 | } | |
1645 | ||
9e8ae01d ON |
1646 | wo.wo_type = type; |
1647 | wo.wo_pid = pid; | |
1648 | wo.wo_flags = options | WEXITED; | |
1649 | wo.wo_info = NULL; | |
359566fa | 1650 | wo.wo_stat = 0; |
9e8ae01d ON |
1651 | wo.wo_rusage = ru; |
1652 | ret = do_wait(&wo); | |
161550d7 | 1653 | put_pid(pid); |
359566fa AV |
1654 | if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr)) |
1655 | ret = -EFAULT; | |
1da177e4 | 1656 | |
1da177e4 LT |
1657 | return ret; |
1658 | } | |
1659 | ||
ce72a16f AV |
1660 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1661 | int, options, struct rusage __user *, ru) | |
1662 | { | |
1663 | struct rusage r; | |
1664 | long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL); | |
1665 | ||
1666 | if (err > 0) { | |
1667 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) | |
1668 | return -EFAULT; | |
1669 | } | |
1670 | return err; | |
1671 | } | |
1672 | ||
1da177e4 LT |
1673 | #ifdef __ARCH_WANT_SYS_WAITPID |
1674 | ||
1675 | /* | |
1676 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1677 | * implemented by calling sys_wait4() from libc.a. | |
1678 | */ | |
17da2bd9 | 1679 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1680 | { |
1681 | return sys_wait4(pid, stat_addr, options, NULL); | |
1682 | } | |
1683 | ||
1684 | #endif | |
7e95a225 AV |
1685 | |
1686 | #ifdef CONFIG_COMPAT | |
1687 | COMPAT_SYSCALL_DEFINE4(wait4, | |
1688 | compat_pid_t, pid, | |
1689 | compat_uint_t __user *, stat_addr, | |
1690 | int, options, | |
1691 | struct compat_rusage __user *, ru) | |
1692 | { | |
ce72a16f AV |
1693 | struct rusage r; |
1694 | long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL); | |
1695 | if (err > 0) { | |
1696 | if (ru && put_compat_rusage(&r, ru)) | |
1697 | return -EFAULT; | |
7e95a225 | 1698 | } |
ce72a16f | 1699 | return err; |
7e95a225 AV |
1700 | } |
1701 | ||
1702 | COMPAT_SYSCALL_DEFINE5(waitid, | |
1703 | int, which, compat_pid_t, pid, | |
1704 | struct compat_siginfo __user *, infop, int, options, | |
1705 | struct compat_rusage __user *, uru) | |
1706 | { | |
7e95a225 | 1707 | struct rusage ru; |
67d7ddde AV |
1708 | struct waitid_info info = {.status = 0}; |
1709 | long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL); | |
634a8160 AV |
1710 | int signo = 0; |
1711 | if (err > 0) { | |
1712 | signo = SIGCHLD; | |
1713 | err = 0; | |
6c85501f AV |
1714 | if (uru) { |
1715 | /* kernel_waitid() overwrites everything in ru */ | |
1716 | if (COMPAT_USE_64BIT_TIME) | |
1717 | err = copy_to_user(uru, &ru, sizeof(ru)); | |
1718 | else | |
1719 | err = put_compat_rusage(&ru, uru); | |
1720 | if (err) | |
1721 | return -EFAULT; | |
1722 | } | |
7e95a225 AV |
1723 | } |
1724 | ||
4c48abe9 AV |
1725 | if (!infop) |
1726 | return err; | |
1727 | ||
96ca579a | 1728 | if (!access_ok(VERIFY_WRITE, infop, sizeof(*infop))) |
1c9fec47 | 1729 | return -EFAULT; |
96ca579a | 1730 | |
4c48abe9 | 1731 | user_access_begin(); |
634a8160 | 1732 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1733 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1734 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1735 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1736 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1737 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
1738 | user_access_end(); | |
67d7ddde | 1739 | return err; |
4c48abe9 AV |
1740 | Efault: |
1741 | user_access_end(); | |
1742 | return -EFAULT; | |
7e95a225 AV |
1743 | } |
1744 | #endif | |
7c2c11b2 SM |
1745 | |
1746 | __weak void abort(void) | |
1747 | { | |
1748 | BUG(); | |
1749 | ||
1750 | /* if that doesn't kill us, halt */ | |
1751 | panic("Oops failed to kill thread"); | |
1752 | } | |
dc8635b7 | 1753 | EXPORT_SYMBOL(abort); |