Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/exit.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
1da177e4 LT |
8 | #include <linux/mm.h> |
9 | #include <linux/slab.h> | |
4eb5aaa3 | 10 | #include <linux/sched/autogroup.h> |
6e84f315 | 11 | #include <linux/sched/mm.h> |
03441a34 | 12 | #include <linux/sched/stat.h> |
29930025 | 13 | #include <linux/sched/task.h> |
68db0cf1 | 14 | #include <linux/sched/task_stack.h> |
32ef5517 | 15 | #include <linux/sched/cputime.h> |
1da177e4 | 16 | #include <linux/interrupt.h> |
1da177e4 | 17 | #include <linux/module.h> |
c59ede7b | 18 | #include <linux/capability.h> |
1da177e4 LT |
19 | #include <linux/completion.h> |
20 | #include <linux/personality.h> | |
21 | #include <linux/tty.h> | |
da9cbc87 | 22 | #include <linux/iocontext.h> |
1da177e4 | 23 | #include <linux/key.h> |
1da177e4 LT |
24 | #include <linux/cpu.h> |
25 | #include <linux/acct.h> | |
8f0ab514 | 26 | #include <linux/tsacct_kern.h> |
1da177e4 | 27 | #include <linux/file.h> |
9f3acc31 | 28 | #include <linux/fdtable.h> |
80d26af8 | 29 | #include <linux/freezer.h> |
1da177e4 | 30 | #include <linux/binfmts.h> |
ab516013 | 31 | #include <linux/nsproxy.h> |
84d73786 | 32 | #include <linux/pid_namespace.h> |
1da177e4 LT |
33 | #include <linux/ptrace.h> |
34 | #include <linux/profile.h> | |
35 | #include <linux/mount.h> | |
36 | #include <linux/proc_fs.h> | |
49d769d5 | 37 | #include <linux/kthread.h> |
1da177e4 | 38 | #include <linux/mempolicy.h> |
c757249a | 39 | #include <linux/taskstats_kern.h> |
ca74e92b | 40 | #include <linux/delayacct.h> |
b4f48b63 | 41 | #include <linux/cgroup.h> |
1da177e4 | 42 | #include <linux/syscalls.h> |
7ed20e1a | 43 | #include <linux/signal.h> |
6a14c5c9 | 44 | #include <linux/posix-timers.h> |
9f46080c | 45 | #include <linux/cn_proc.h> |
de5097c2 | 46 | #include <linux/mutex.h> |
0771dfef | 47 | #include <linux/futex.h> |
b92ce558 | 48 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 49 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 50 | #include <linux/resource.h> |
0d67a46d | 51 | #include <linux/blkdev.h> |
6eaeeaba | 52 | #include <linux/task_io_accounting_ops.h> |
30199f5a | 53 | #include <linux/tracehook.h> |
5ad4e53b | 54 | #include <linux/fs_struct.h> |
d84f4f99 | 55 | #include <linux/init_task.h> |
cdd6c482 | 56 | #include <linux/perf_event.h> |
ad8d75ff | 57 | #include <trace/events/sched.h> |
24f1e32c | 58 | #include <linux/hw_breakpoint.h> |
3d5992d2 | 59 | #include <linux/oom.h> |
54848d73 | 60 | #include <linux/writeback.h> |
40401530 | 61 | #include <linux/shm.h> |
5c9a8750 | 62 | #include <linux/kcov.h> |
53d3eaa3 | 63 | #include <linux/random.h> |
8f95c90c | 64 | #include <linux/rcuwait.h> |
7e95a225 | 65 | #include <linux/compat.h> |
1da177e4 | 66 | |
7c0f6ba6 | 67 | #include <linux/uaccess.h> |
1da177e4 LT |
68 | #include <asm/unistd.h> |
69 | #include <asm/pgtable.h> | |
70 | #include <asm/mmu_context.h> | |
71 | ||
d40e48e0 | 72 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
73 | { |
74 | nr_threads--; | |
50d75f8d | 75 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 76 | if (group_dead) { |
6883f81a | 77 | detach_pid(p, PIDTYPE_TGID); |
1da177e4 LT |
78 | detach_pid(p, PIDTYPE_PGID); |
79 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 80 | |
5e85d4ab | 81 | list_del_rcu(&p->tasks); |
9cd80bbb | 82 | list_del_init(&p->sibling); |
909ea964 | 83 | __this_cpu_dec(process_counts); |
1da177e4 | 84 | } |
47e65328 | 85 | list_del_rcu(&p->thread_group); |
0c740d0a | 86 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
87 | } |
88 | ||
6a14c5c9 ON |
89 | /* |
90 | * This function expects the tasklist_lock write-locked. | |
91 | */ | |
92 | static void __exit_signal(struct task_struct *tsk) | |
93 | { | |
94 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 95 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 96 | struct sighand_struct *sighand; |
4ada856f | 97 | struct tty_struct *uninitialized_var(tty); |
5613fda9 | 98 | u64 utime, stime; |
6a14c5c9 | 99 | |
d11c563d | 100 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 101 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
102 | spin_lock(&sighand->siglock); |
103 | ||
baa73d9e | 104 | #ifdef CONFIG_POSIX_TIMERS |
6a14c5c9 | 105 | posix_cpu_timers_exit(tsk); |
d40e48e0 | 106 | if (group_dead) { |
6a14c5c9 | 107 | posix_cpu_timers_exit_group(tsk); |
4a599942 | 108 | } else { |
e0a70217 ON |
109 | /* |
110 | * This can only happen if the caller is de_thread(). | |
111 | * FIXME: this is the temporary hack, we should teach | |
112 | * posix-cpu-timers to handle this case correctly. | |
113 | */ | |
114 | if (unlikely(has_group_leader_pid(tsk))) | |
115 | posix_cpu_timers_exit_group(tsk); | |
baa73d9e NP |
116 | } |
117 | #endif | |
e0a70217 | 118 | |
baa73d9e NP |
119 | if (group_dead) { |
120 | tty = sig->tty; | |
121 | sig->tty = NULL; | |
122 | } else { | |
6a14c5c9 ON |
123 | /* |
124 | * If there is any task waiting for the group exit | |
125 | * then notify it: | |
126 | */ | |
d344193a | 127 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 128 | wake_up_process(sig->group_exit_task); |
6db840fa | 129 | |
6a14c5c9 ON |
130 | if (tsk == sig->curr_target) |
131 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
132 | } |
133 | ||
53d3eaa3 NP |
134 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
135 | sizeof(unsigned long long)); | |
136 | ||
90ed9cbe | 137 | /* |
26e75b5c ON |
138 | * Accumulate here the counters for all threads as they die. We could |
139 | * skip the group leader because it is the last user of signal_struct, | |
140 | * but we want to avoid the race with thread_group_cputime() which can | |
141 | * see the empty ->thread_head list. | |
90ed9cbe RR |
142 | */ |
143 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 144 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
145 | sig->utime += utime; |
146 | sig->stime += stime; | |
147 | sig->gtime += task_gtime(tsk); | |
148 | sig->min_flt += tsk->min_flt; | |
149 | sig->maj_flt += tsk->maj_flt; | |
150 | sig->nvcsw += tsk->nvcsw; | |
151 | sig->nivcsw += tsk->nivcsw; | |
152 | sig->inblock += task_io_get_inblock(tsk); | |
153 | sig->oublock += task_io_get_oublock(tsk); | |
154 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
155 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 156 | sig->nr_threads--; |
d40e48e0 | 157 | __unhash_process(tsk, group_dead); |
e78c3496 | 158 | write_sequnlock(&sig->stats_lock); |
5876700c | 159 | |
da7978b0 ON |
160 | /* |
161 | * Do this under ->siglock, we can race with another thread | |
162 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
163 | */ | |
164 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 165 | tsk->sighand = NULL; |
6a14c5c9 | 166 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 167 | |
a7e5328a | 168 | __cleanup_sighand(sighand); |
a0be55de | 169 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 170 | if (group_dead) { |
6a14c5c9 | 171 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 172 | tty_kref_put(tty); |
6a14c5c9 ON |
173 | } |
174 | } | |
175 | ||
8c7904a0 EB |
176 | static void delayed_put_task_struct(struct rcu_head *rhp) |
177 | { | |
0a16b607 MD |
178 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
179 | ||
4e231c79 | 180 | perf_event_delayed_put(tsk); |
0a16b607 MD |
181 | trace_sched_process_free(tsk); |
182 | put_task_struct(tsk); | |
8c7904a0 EB |
183 | } |
184 | ||
3fbd7ee2 EB |
185 | void put_task_struct_rcu_user(struct task_struct *task) |
186 | { | |
187 | if (refcount_dec_and_test(&task->rcu_users)) | |
188 | call_rcu(&task->rcu, delayed_put_task_struct); | |
189 | } | |
f470021a | 190 | |
a0be55de | 191 | void release_task(struct task_struct *p) |
1da177e4 | 192 | { |
36c8b586 | 193 | struct task_struct *leader; |
1da177e4 | 194 | int zap_leader; |
1f09f974 | 195 | repeat: |
c69e8d9c | 196 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
197 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
198 | rcu_read_lock(); | |
c69e8d9c | 199 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 200 | rcu_read_unlock(); |
c69e8d9c | 201 | |
60347f67 | 202 | proc_flush_task(p); |
6b115bf5 | 203 | cgroup_release(p); |
0203026b | 204 | |
1da177e4 | 205 | write_lock_irq(&tasklist_lock); |
a288eecc | 206 | ptrace_release_task(p); |
1da177e4 | 207 | __exit_signal(p); |
35f5cad8 | 208 | |
1da177e4 LT |
209 | /* |
210 | * If we are the last non-leader member of the thread | |
211 | * group, and the leader is zombie, then notify the | |
212 | * group leader's parent process. (if it wants notification.) | |
213 | */ | |
214 | zap_leader = 0; | |
215 | leader = p->group_leader; | |
a0be55de IA |
216 | if (leader != p && thread_group_empty(leader) |
217 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
218 | /* |
219 | * If we were the last child thread and the leader has | |
220 | * exited already, and the leader's parent ignores SIGCHLD, | |
221 | * then we are the one who should release the leader. | |
dae33574 | 222 | */ |
86773473 | 223 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
224 | if (zap_leader) |
225 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
226 | } |
227 | ||
1da177e4 | 228 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 229 | release_thread(p); |
3fbd7ee2 | 230 | put_task_struct_rcu_user(p); |
1da177e4 LT |
231 | |
232 | p = leader; | |
233 | if (unlikely(zap_leader)) | |
234 | goto repeat; | |
235 | } | |
236 | ||
8f95c90c DB |
237 | void rcuwait_wake_up(struct rcuwait *w) |
238 | { | |
239 | struct task_struct *task; | |
240 | ||
241 | rcu_read_lock(); | |
242 | ||
243 | /* | |
244 | * Order condition vs @task, such that everything prior to the load | |
245 | * of @task is visible. This is the condition as to why the user called | |
246 | * rcuwait_trywake() in the first place. Pairs with set_current_state() | |
247 | * barrier (A) in rcuwait_wait_event(). | |
248 | * | |
249 | * WAIT WAKE | |
250 | * [S] tsk = current [S] cond = true | |
251 | * MB (A) MB (B) | |
252 | * [L] cond [L] tsk | |
253 | */ | |
6dc080ee | 254 | smp_mb(); /* (B) */ |
8f95c90c | 255 | |
8f95c90c DB |
256 | task = rcu_dereference(w->task); |
257 | if (task) | |
258 | wake_up_process(task); | |
259 | rcu_read_unlock(); | |
260 | } | |
261 | ||
1da177e4 LT |
262 | /* |
263 | * Determine if a process group is "orphaned", according to the POSIX | |
264 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
265 | * by terminal-generated stop signals. Newly orphaned process groups are | |
266 | * to receive a SIGHUP and a SIGCONT. | |
267 | * | |
268 | * "I ask you, have you ever known what it is to be an orphan?" | |
269 | */ | |
a0be55de IA |
270 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
271 | struct task_struct *ignored_task) | |
1da177e4 LT |
272 | { |
273 | struct task_struct *p; | |
1da177e4 | 274 | |
0475ac08 | 275 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
276 | if ((p == ignored_task) || |
277 | (p->exit_state && thread_group_empty(p)) || | |
278 | is_global_init(p->real_parent)) | |
1da177e4 | 279 | continue; |
05e83df6 | 280 | |
0475ac08 | 281 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
282 | task_session(p->real_parent) == task_session(p)) |
283 | return 0; | |
0475ac08 | 284 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
285 | |
286 | return 1; | |
1da177e4 LT |
287 | } |
288 | ||
3e7cd6c4 | 289 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
290 | { |
291 | int retval; | |
292 | ||
293 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 294 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
295 | read_unlock(&tasklist_lock); |
296 | ||
297 | return retval; | |
298 | } | |
299 | ||
961c4675 | 300 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 301 | { |
1da177e4 LT |
302 | struct task_struct *p; |
303 | ||
0475ac08 | 304 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
305 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
306 | return true; | |
0475ac08 | 307 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
308 | |
309 | return false; | |
1da177e4 LT |
310 | } |
311 | ||
f49ee505 ON |
312 | /* |
313 | * Check to see if any process groups have become orphaned as | |
314 | * a result of our exiting, and if they have any stopped jobs, | |
315 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
316 | */ | |
317 | static void | |
318 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
319 | { | |
320 | struct pid *pgrp = task_pgrp(tsk); | |
321 | struct task_struct *ignored_task = tsk; | |
322 | ||
323 | if (!parent) | |
a0be55de IA |
324 | /* exit: our father is in a different pgrp than |
325 | * we are and we were the only connection outside. | |
326 | */ | |
f49ee505 ON |
327 | parent = tsk->real_parent; |
328 | else | |
329 | /* reparent: our child is in a different pgrp than | |
330 | * we are, and it was the only connection outside. | |
331 | */ | |
332 | ignored_task = NULL; | |
333 | ||
334 | if (task_pgrp(parent) != pgrp && | |
335 | task_session(parent) == task_session(tsk) && | |
336 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
337 | has_stopped_jobs(pgrp)) { | |
338 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
339 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
340 | } | |
341 | } | |
342 | ||
f98bafa0 | 343 | #ifdef CONFIG_MEMCG |
cf475ad2 | 344 | /* |
733eda7a | 345 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 346 | */ |
cf475ad2 BS |
347 | void mm_update_next_owner(struct mm_struct *mm) |
348 | { | |
349 | struct task_struct *c, *g, *p = current; | |
350 | ||
351 | retry: | |
733eda7a KH |
352 | /* |
353 | * If the exiting or execing task is not the owner, it's | |
354 | * someone else's problem. | |
355 | */ | |
356 | if (mm->owner != p) | |
cf475ad2 | 357 | return; |
733eda7a KH |
358 | /* |
359 | * The current owner is exiting/execing and there are no other | |
360 | * candidates. Do not leave the mm pointing to a possibly | |
361 | * freed task structure. | |
362 | */ | |
363 | if (atomic_read(&mm->mm_users) <= 1) { | |
987717e5 | 364 | WRITE_ONCE(mm->owner, NULL); |
733eda7a KH |
365 | return; |
366 | } | |
cf475ad2 BS |
367 | |
368 | read_lock(&tasklist_lock); | |
369 | /* | |
370 | * Search in the children | |
371 | */ | |
372 | list_for_each_entry(c, &p->children, sibling) { | |
373 | if (c->mm == mm) | |
374 | goto assign_new_owner; | |
375 | } | |
376 | ||
377 | /* | |
378 | * Search in the siblings | |
379 | */ | |
dea33cfd | 380 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
381 | if (c->mm == mm) |
382 | goto assign_new_owner; | |
383 | } | |
384 | ||
385 | /* | |
f87fb599 | 386 | * Search through everything else, we should not get here often. |
cf475ad2 | 387 | */ |
39af1765 ON |
388 | for_each_process(g) { |
389 | if (g->flags & PF_KTHREAD) | |
390 | continue; | |
391 | for_each_thread(g, c) { | |
392 | if (c->mm == mm) | |
393 | goto assign_new_owner; | |
394 | if (c->mm) | |
395 | break; | |
396 | } | |
f87fb599 | 397 | } |
cf475ad2 | 398 | read_unlock(&tasklist_lock); |
31a78f23 BS |
399 | /* |
400 | * We found no owner yet mm_users > 1: this implies that we are | |
401 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 402 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 403 | */ |
987717e5 | 404 | WRITE_ONCE(mm->owner, NULL); |
cf475ad2 BS |
405 | return; |
406 | ||
407 | assign_new_owner: | |
408 | BUG_ON(c == p); | |
409 | get_task_struct(c); | |
410 | /* | |
411 | * The task_lock protects c->mm from changing. | |
412 | * We always want mm->owner->mm == mm | |
413 | */ | |
414 | task_lock(c); | |
e5991371 HD |
415 | /* |
416 | * Delay read_unlock() till we have the task_lock() | |
417 | * to ensure that c does not slip away underneath us | |
418 | */ | |
419 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
420 | if (c->mm != mm) { |
421 | task_unlock(c); | |
422 | put_task_struct(c); | |
423 | goto retry; | |
424 | } | |
987717e5 | 425 | WRITE_ONCE(mm->owner, c); |
cf475ad2 BS |
426 | task_unlock(c); |
427 | put_task_struct(c); | |
428 | } | |
f98bafa0 | 429 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 430 | |
1da177e4 LT |
431 | /* |
432 | * Turn us into a lazy TLB process if we | |
433 | * aren't already.. | |
434 | */ | |
0039962a | 435 | static void exit_mm(void) |
1da177e4 | 436 | { |
0039962a | 437 | struct mm_struct *mm = current->mm; |
b564daf8 | 438 | struct core_state *core_state; |
1da177e4 | 439 | |
0039962a | 440 | mm_release(current, mm); |
1da177e4 LT |
441 | if (!mm) |
442 | return; | |
4fe7efdb | 443 | sync_mm_rss(mm); |
1da177e4 LT |
444 | /* |
445 | * Serialize with any possible pending coredump. | |
999d9fc1 | 446 | * We must hold mmap_sem around checking core_state |
1da177e4 | 447 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 448 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
449 | * group with ->mm != NULL. |
450 | */ | |
451 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
452 | core_state = mm->core_state; |
453 | if (core_state) { | |
454 | struct core_thread self; | |
a0be55de | 455 | |
1da177e4 | 456 | up_read(&mm->mmap_sem); |
1da177e4 | 457 | |
0039962a | 458 | self.task = current; |
b564daf8 ON |
459 | self.next = xchg(&core_state->dumper.next, &self); |
460 | /* | |
461 | * Implies mb(), the result of xchg() must be visible | |
462 | * to core_state->dumper. | |
463 | */ | |
464 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
465 | complete(&core_state->startup); | |
1da177e4 | 466 | |
a94e2d40 | 467 | for (;;) { |
642fa448 | 468 | set_current_state(TASK_UNINTERRUPTIBLE); |
a94e2d40 ON |
469 | if (!self.task) /* see coredump_finish() */ |
470 | break; | |
80d26af8 | 471 | freezable_schedule(); |
a94e2d40 | 472 | } |
642fa448 | 473 | __set_current_state(TASK_RUNNING); |
1da177e4 LT |
474 | down_read(&mm->mmap_sem); |
475 | } | |
f1f10076 | 476 | mmgrab(mm); |
0039962a | 477 | BUG_ON(mm != current->active_mm); |
1da177e4 | 478 | /* more a memory barrier than a real lock */ |
0039962a DB |
479 | task_lock(current); |
480 | current->mm = NULL; | |
1da177e4 LT |
481 | up_read(&mm->mmap_sem); |
482 | enter_lazy_tlb(mm, current); | |
0039962a | 483 | task_unlock(current); |
cf475ad2 | 484 | mm_update_next_owner(mm); |
1da177e4 | 485 | mmput(mm); |
c32b3cbe | 486 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 487 | exit_oom_victim(); |
1da177e4 LT |
488 | } |
489 | ||
c9dc05bf ON |
490 | static struct task_struct *find_alive_thread(struct task_struct *p) |
491 | { | |
492 | struct task_struct *t; | |
493 | ||
494 | for_each_thread(p, t) { | |
495 | if (!(t->flags & PF_EXITING)) | |
496 | return t; | |
497 | } | |
498 | return NULL; | |
499 | } | |
500 | ||
8fb335e0 AV |
501 | static struct task_struct *find_child_reaper(struct task_struct *father, |
502 | struct list_head *dead) | |
1109909c ON |
503 | __releases(&tasklist_lock) |
504 | __acquires(&tasklist_lock) | |
505 | { | |
506 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
507 | struct task_struct *reaper = pid_ns->child_reaper; | |
8fb335e0 | 508 | struct task_struct *p, *n; |
1109909c ON |
509 | |
510 | if (likely(reaper != father)) | |
511 | return reaper; | |
512 | ||
c9dc05bf ON |
513 | reaper = find_alive_thread(father); |
514 | if (reaper) { | |
1109909c ON |
515 | pid_ns->child_reaper = reaper; |
516 | return reaper; | |
517 | } | |
518 | ||
519 | write_unlock_irq(&tasklist_lock); | |
520 | if (unlikely(pid_ns == &init_pid_ns)) { | |
521 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
522 | father->signal->group_exit_code ?: father->exit_code); | |
523 | } | |
8fb335e0 AV |
524 | |
525 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
526 | list_del_init(&p->ptrace_entry); | |
527 | release_task(p); | |
528 | } | |
529 | ||
1109909c ON |
530 | zap_pid_ns_processes(pid_ns); |
531 | write_lock_irq(&tasklist_lock); | |
532 | ||
533 | return father; | |
534 | } | |
535 | ||
1da177e4 | 536 | /* |
ebec18a6 LP |
537 | * When we die, we re-parent all our children, and try to: |
538 | * 1. give them to another thread in our thread group, if such a member exists | |
539 | * 2. give it to the first ancestor process which prctl'd itself as a | |
540 | * child_subreaper for its children (like a service manager) | |
541 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 542 | */ |
1109909c ON |
543 | static struct task_struct *find_new_reaper(struct task_struct *father, |
544 | struct task_struct *child_reaper) | |
1da177e4 | 545 | { |
c9dc05bf | 546 | struct task_struct *thread, *reaper; |
1da177e4 | 547 | |
c9dc05bf ON |
548 | thread = find_alive_thread(father); |
549 | if (thread) | |
950bbabb | 550 | return thread; |
1da177e4 | 551 | |
7d24e2df | 552 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 553 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 554 | /* |
175aed3f | 555 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
556 | * We can't check reaper != child_reaper to ensure we do not |
557 | * cross the namespaces, the exiting parent could be injected | |
558 | * by setns() + fork(). | |
559 | * We check pid->level, this is slightly more efficient than | |
560 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 561 | */ |
c6c70f44 ON |
562 | for (reaper = father->real_parent; |
563 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 564 | reaper = reaper->real_parent) { |
175aed3f | 565 | if (reaper == &init_task) |
ebec18a6 LP |
566 | break; |
567 | if (!reaper->signal->is_child_subreaper) | |
568 | continue; | |
c9dc05bf ON |
569 | thread = find_alive_thread(reaper); |
570 | if (thread) | |
571 | return thread; | |
ebec18a6 | 572 | } |
1da177e4 | 573 | } |
762a24be | 574 | |
1109909c | 575 | return child_reaper; |
950bbabb ON |
576 | } |
577 | ||
5dfc80be ON |
578 | /* |
579 | * Any that need to be release_task'd are put on the @dead list. | |
580 | */ | |
9cd80bbb | 581 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
582 | struct list_head *dead) |
583 | { | |
2831096e | 584 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
585 | return; |
586 | ||
abd50b39 | 587 | /* We don't want people slaying init. */ |
5dfc80be ON |
588 | p->exit_signal = SIGCHLD; |
589 | ||
590 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 591 | if (!p->ptrace && |
5dfc80be | 592 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 593 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 594 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 595 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
596 | } |
597 | } | |
598 | ||
599 | kill_orphaned_pgrp(p, father); | |
600 | } | |
601 | ||
482a3767 ON |
602 | /* |
603 | * This does two things: | |
604 | * | |
605 | * A. Make init inherit all the child processes | |
606 | * B. Check to see if any process groups have become orphaned | |
607 | * as a result of our exiting, and if they have any stopped | |
608 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
609 | */ | |
610 | static void forget_original_parent(struct task_struct *father, | |
611 | struct list_head *dead) | |
1da177e4 | 612 | { |
482a3767 | 613 | struct task_struct *p, *t, *reaper; |
762a24be | 614 | |
7c8bd232 | 615 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 616 | exit_ptrace(father, dead); |
f470021a | 617 | |
7c8bd232 | 618 | /* Can drop and reacquire tasklist_lock */ |
8fb335e0 | 619 | reaper = find_child_reaper(father, dead); |
ad9e206a | 620 | if (list_empty(&father->children)) |
482a3767 | 621 | return; |
1109909c ON |
622 | |
623 | reaper = find_new_reaper(father, reaper); | |
2831096e | 624 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 625 | for_each_thread(p, t) { |
9cd80bbb | 626 | t->real_parent = reaper; |
57a05918 ON |
627 | BUG_ON((!t->ptrace) != (t->parent == father)); |
628 | if (likely(!t->ptrace)) | |
9cd80bbb | 629 | t->parent = t->real_parent; |
9cd80bbb ON |
630 | if (t->pdeath_signal) |
631 | group_send_sig_info(t->pdeath_signal, | |
01024980 EB |
632 | SEND_SIG_NOINFO, t, |
633 | PIDTYPE_TGID); | |
57a05918 | 634 | } |
2831096e ON |
635 | /* |
636 | * If this is a threaded reparent there is no need to | |
637 | * notify anyone anything has happened. | |
638 | */ | |
639 | if (!same_thread_group(reaper, father)) | |
482a3767 | 640 | reparent_leader(father, p, dead); |
1da177e4 | 641 | } |
2831096e | 642 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
643 | } |
644 | ||
645 | /* | |
646 | * Send signals to all our closest relatives so that they know | |
647 | * to properly mourn us.. | |
648 | */ | |
821c7de7 | 649 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 650 | { |
53c8f9f1 | 651 | bool autoreap; |
482a3767 ON |
652 | struct task_struct *p, *n; |
653 | LIST_HEAD(dead); | |
1da177e4 | 654 | |
762a24be | 655 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
656 | forget_original_parent(tsk, &dead); |
657 | ||
821c7de7 ON |
658 | if (group_dead) |
659 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 660 | |
b191d649 | 661 | tsk->exit_state = EXIT_ZOMBIE; |
45cdf5cc ON |
662 | if (unlikely(tsk->ptrace)) { |
663 | int sig = thread_group_leader(tsk) && | |
664 | thread_group_empty(tsk) && | |
665 | !ptrace_reparented(tsk) ? | |
666 | tsk->exit_signal : SIGCHLD; | |
667 | autoreap = do_notify_parent(tsk, sig); | |
668 | } else if (thread_group_leader(tsk)) { | |
669 | autoreap = thread_group_empty(tsk) && | |
670 | do_notify_parent(tsk, tsk->exit_signal); | |
671 | } else { | |
672 | autoreap = true; | |
673 | } | |
1da177e4 | 674 | |
30b692d3 CB |
675 | if (autoreap) { |
676 | tsk->exit_state = EXIT_DEAD; | |
6c66e7db | 677 | list_add(&tsk->ptrace_entry, &dead); |
30b692d3 | 678 | } |
1da177e4 | 679 | |
9c339168 ON |
680 | /* mt-exec, de_thread() is waiting for group leader */ |
681 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 682 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
683 | write_unlock_irq(&tasklist_lock); |
684 | ||
482a3767 ON |
685 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
686 | list_del_init(&p->ptrace_entry); | |
687 | release_task(p); | |
688 | } | |
1da177e4 LT |
689 | } |
690 | ||
e18eecb8 JD |
691 | #ifdef CONFIG_DEBUG_STACK_USAGE |
692 | static void check_stack_usage(void) | |
693 | { | |
694 | static DEFINE_SPINLOCK(low_water_lock); | |
695 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
696 | unsigned long free; |
697 | ||
7c9f8861 | 698 | free = stack_not_used(current); |
e18eecb8 JD |
699 | |
700 | if (free >= lowest_to_date) | |
701 | return; | |
702 | ||
703 | spin_lock(&low_water_lock); | |
704 | if (free < lowest_to_date) { | |
627393d4 | 705 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 706 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
707 | lowest_to_date = free; |
708 | } | |
709 | spin_unlock(&low_water_lock); | |
710 | } | |
711 | #else | |
712 | static inline void check_stack_usage(void) {} | |
713 | #endif | |
714 | ||
9af6528e | 715 | void __noreturn do_exit(long code) |
1da177e4 LT |
716 | { |
717 | struct task_struct *tsk = current; | |
718 | int group_dead; | |
719 | ||
720 | profile_task_exit(tsk); | |
5c9a8750 | 721 | kcov_task_exit(tsk); |
1da177e4 | 722 | |
73c10101 | 723 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 724 | |
1da177e4 LT |
725 | if (unlikely(in_interrupt())) |
726 | panic("Aiee, killing interrupt handler!"); | |
727 | if (unlikely(!tsk->pid)) | |
728 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 729 | |
33dd94ae NE |
730 | /* |
731 | * If do_exit is called because this processes oopsed, it's possible | |
732 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
733 | * continuing. Amongst other possible reasons, this is to prevent | |
734 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
735 | * kernel address. | |
736 | */ | |
737 | set_fs(USER_DS); | |
738 | ||
a288eecc | 739 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 740 | |
e0e81739 DH |
741 | validate_creds_for_do_exit(tsk); |
742 | ||
df164db5 AN |
743 | /* |
744 | * We're taking recursive faults here in do_exit. Safest is to just | |
745 | * leave this task alone and wait for reboot. | |
746 | */ | |
747 | if (unlikely(tsk->flags & PF_EXITING)) { | |
a0be55de | 748 | pr_alert("Fixing recursive fault but reboot is needed!\n"); |
778e9a9c AK |
749 | /* |
750 | * We can do this unlocked here. The futex code uses | |
751 | * this flag just to verify whether the pi state | |
752 | * cleanup has been done or not. In the worst case it | |
753 | * loops once more. We pretend that the cleanup was | |
754 | * done as there is no way to return. Either the | |
755 | * OWNER_DIED bit is set by now or we push the blocked | |
756 | * task into the wait for ever nirwana as well. | |
757 | */ | |
758 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
759 | set_current_state(TASK_UNINTERRUPTIBLE); |
760 | schedule(); | |
761 | } | |
762 | ||
d12619b5 | 763 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c | 764 | /* |
be3e7844 PZ |
765 | * Ensure that all new tsk->pi_lock acquisitions must observe |
766 | * PF_EXITING. Serializes against futex.c:attach_to_pi_owner(). | |
778e9a9c | 767 | */ |
d2ee7198 | 768 | smp_mb(); |
be3e7844 PZ |
769 | /* |
770 | * Ensure that we must observe the pi_state in exit_mm() -> | |
771 | * mm_release() -> exit_pi_state_list(). | |
772 | */ | |
8083f293 PM |
773 | raw_spin_lock_irq(&tsk->pi_lock); |
774 | raw_spin_unlock_irq(&tsk->pi_lock); | |
1da177e4 | 775 | |
1dc0fffc | 776 | if (unlikely(in_atomic())) { |
a0be55de IA |
777 | pr_info("note: %s[%d] exited with preempt_count %d\n", |
778 | current->comm, task_pid_nr(current), | |
779 | preempt_count()); | |
1dc0fffc PZ |
780 | preempt_count_set(PREEMPT_ENABLED); |
781 | } | |
1da177e4 | 782 | |
48d212a2 LT |
783 | /* sync mm's RSS info before statistics gathering */ |
784 | if (tsk->mm) | |
785 | sync_mm_rss(tsk->mm); | |
51229b49 | 786 | acct_update_integrals(tsk); |
1da177e4 | 787 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 788 | if (group_dead) { |
baa73d9e | 789 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 790 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 791 | exit_itimers(tsk->signal); |
baa73d9e | 792 | #endif |
1f10206c JP |
793 | if (tsk->mm) |
794 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 795 | } |
f6ec29a4 | 796 | acct_collect(code, group_dead); |
522ed776 MT |
797 | if (group_dead) |
798 | tty_audit_exit(); | |
a4ff8dba | 799 | audit_free(tsk); |
115085ea | 800 | |
48d212a2 | 801 | tsk->exit_code = code; |
115085ea | 802 | taskstats_exit(tsk, group_dead); |
c757249a | 803 | |
0039962a | 804 | exit_mm(); |
1da177e4 | 805 | |
0e464814 | 806 | if (group_dead) |
f6ec29a4 | 807 | acct_process(); |
0a16b607 MD |
808 | trace_sched_process_exit(tsk); |
809 | ||
1da177e4 | 810 | exit_sem(tsk); |
b34a6b1d | 811 | exit_shm(tsk); |
1ec7f1dd AV |
812 | exit_files(tsk); |
813 | exit_fs(tsk); | |
c39df5fa ON |
814 | if (group_dead) |
815 | disassociate_ctty(1); | |
8aac6270 | 816 | exit_task_namespaces(tsk); |
ed3e694d | 817 | exit_task_work(tsk); |
e6464694 | 818 | exit_thread(tsk); |
73ab1cb2 | 819 | exit_umh(tsk); |
0b3fcf17 SE |
820 | |
821 | /* | |
822 | * Flush inherited counters to the parent - before the parent | |
823 | * gets woken up by child-exit notifications. | |
824 | * | |
825 | * because of cgroup mode, must be called before cgroup_exit() | |
826 | */ | |
827 | perf_event_exit_task(tsk); | |
828 | ||
8e5bfa8c | 829 | sched_autogroup_exit_task(tsk); |
1ec41830 | 830 | cgroup_exit(tsk); |
1da177e4 | 831 | |
24f1e32c FW |
832 | /* |
833 | * FIXME: do that only when needed, using sched_exit tracepoint | |
834 | */ | |
7c8df286 | 835 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 836 | |
ccdd29ff | 837 | exit_tasks_rcu_start(); |
821c7de7 | 838 | exit_notify(tsk, group_dead); |
ef982393 | 839 | proc_exit_connector(tsk); |
c11600e4 | 840 | mpol_put_task_policy(tsk); |
42b2dd0a | 841 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
842 | if (unlikely(current->pi_state_cache)) |
843 | kfree(current->pi_state_cache); | |
42b2dd0a | 844 | #endif |
de5097c2 | 845 | /* |
9a11b49a | 846 | * Make sure we are holding no locks: |
de5097c2 | 847 | */ |
1b1d2fb4 | 848 | debug_check_no_locks_held(); |
778e9a9c AK |
849 | /* |
850 | * We can do this unlocked here. The futex code uses this flag | |
851 | * just to verify whether the pi state cleanup has been done | |
852 | * or not. In the worst case it loops once more. | |
853 | */ | |
854 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 855 | |
afc847b7 | 856 | if (tsk->io_context) |
b69f2292 | 857 | exit_io_context(tsk); |
afc847b7 | 858 | |
b92ce558 | 859 | if (tsk->splice_pipe) |
4b8a8f1e | 860 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 861 | |
5640f768 ED |
862 | if (tsk->task_frag.page) |
863 | put_page(tsk->task_frag.page); | |
864 | ||
e0e81739 DH |
865 | validate_creds_for_do_exit(tsk); |
866 | ||
4bcb8232 | 867 | check_stack_usage(); |
7407251a | 868 | preempt_disable(); |
54848d73 WF |
869 | if (tsk->nr_dirtied) |
870 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 871 | exit_rcu(); |
ccdd29ff | 872 | exit_tasks_rcu_finish(); |
b5740f4b | 873 | |
b09be676 | 874 | lockdep_free_task(tsk); |
9af6528e | 875 | do_task_dead(); |
1da177e4 | 876 | } |
012914da RA |
877 | EXPORT_SYMBOL_GPL(do_exit); |
878 | ||
9402c95f | 879 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
880 | { |
881 | if (comp) | |
882 | complete(comp); | |
55a101f8 | 883 | |
1da177e4 LT |
884 | do_exit(code); |
885 | } | |
1da177e4 LT |
886 | EXPORT_SYMBOL(complete_and_exit); |
887 | ||
754fe8d2 | 888 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
889 | { |
890 | do_exit((error_code&0xff)<<8); | |
891 | } | |
892 | ||
1da177e4 LT |
893 | /* |
894 | * Take down every thread in the group. This is called by fatal signals | |
895 | * as well as by sys_exit_group (below). | |
896 | */ | |
9402c95f | 897 | void |
1da177e4 LT |
898 | do_group_exit(int exit_code) |
899 | { | |
bfc4b089 ON |
900 | struct signal_struct *sig = current->signal; |
901 | ||
1da177e4 LT |
902 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
903 | ||
bfc4b089 ON |
904 | if (signal_group_exit(sig)) |
905 | exit_code = sig->group_exit_code; | |
1da177e4 | 906 | else if (!thread_group_empty(current)) { |
1da177e4 | 907 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 908 | |
1da177e4 | 909 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 910 | if (signal_group_exit(sig)) |
1da177e4 LT |
911 | /* Another thread got here before we took the lock. */ |
912 | exit_code = sig->group_exit_code; | |
913 | else { | |
1da177e4 | 914 | sig->group_exit_code = exit_code; |
ed5d2cac | 915 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
916 | zap_other_threads(current); |
917 | } | |
918 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
919 | } |
920 | ||
921 | do_exit(exit_code); | |
922 | /* NOTREACHED */ | |
923 | } | |
924 | ||
925 | /* | |
926 | * this kills every thread in the thread group. Note that any externally | |
927 | * wait4()-ing process will get the correct exit code - even if this | |
928 | * thread is not the thread group leader. | |
929 | */ | |
754fe8d2 | 930 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
931 | { |
932 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
933 | /* NOTREACHED */ |
934 | return 0; | |
1da177e4 LT |
935 | } |
936 | ||
67d7ddde AV |
937 | struct waitid_info { |
938 | pid_t pid; | |
939 | uid_t uid; | |
940 | int status; | |
941 | int cause; | |
942 | }; | |
943 | ||
9e8ae01d ON |
944 | struct wait_opts { |
945 | enum pid_type wo_type; | |
9e8ae01d | 946 | int wo_flags; |
e1eb1ebc | 947 | struct pid *wo_pid; |
9e8ae01d | 948 | |
67d7ddde | 949 | struct waitid_info *wo_info; |
359566fa | 950 | int wo_stat; |
ce72a16f | 951 | struct rusage *wo_rusage; |
9e8ae01d | 952 | |
ac6424b9 | 953 | wait_queue_entry_t child_wait; |
9e8ae01d ON |
954 | int notask_error; |
955 | }; | |
956 | ||
989264f4 | 957 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 958 | { |
5c01ba49 ON |
959 | return wo->wo_type == PIDTYPE_MAX || |
960 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
961 | } | |
1da177e4 | 962 | |
bf959931 ON |
963 | static int |
964 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
965 | { |
966 | if (!eligible_pid(wo, p)) | |
967 | return 0; | |
bf959931 ON |
968 | |
969 | /* | |
970 | * Wait for all children (clone and not) if __WALL is set or | |
971 | * if it is traced by us. | |
972 | */ | |
973 | if (ptrace || (wo->wo_flags & __WALL)) | |
974 | return 1; | |
975 | ||
976 | /* | |
977 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
978 | * otherwise, wait for non-clone children *only*. | |
979 | * | |
980 | * Note: a "clone" child here is one that reports to its parent | |
981 | * using a signal other than SIGCHLD, or a non-leader thread which | |
982 | * we can only see if it is traced by us. | |
983 | */ | |
984 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 985 | return 0; |
1da177e4 | 986 | |
14dd0b81 | 987 | return 1; |
1da177e4 LT |
988 | } |
989 | ||
1da177e4 LT |
990 | /* |
991 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
992 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
993 | * the lock and this task is uninteresting. If we return nonzero, we have | |
994 | * released the lock and the system call should return. | |
995 | */ | |
9e8ae01d | 996 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 997 | { |
67d7ddde | 998 | int state, status; |
6c5f3e7b | 999 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1000 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
67d7ddde | 1001 | struct waitid_info *infop; |
1da177e4 | 1002 | |
9e8ae01d | 1003 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1004 | return 0; |
1005 | ||
9e8ae01d | 1006 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
76d9871e | 1007 | status = p->exit_code; |
1da177e4 LT |
1008 | get_task_struct(p); |
1009 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1010 | sched_annotate_sleep(); |
e61a2502 AV |
1011 | if (wo->wo_rusage) |
1012 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1013 | put_task_struct(p); |
76d9871e | 1014 | goto out_info; |
1da177e4 | 1015 | } |
1da177e4 | 1016 | /* |
abd50b39 | 1017 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1018 | */ |
f6507f83 ON |
1019 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1020 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1021 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1022 | return 0; |
986094df ON |
1023 | /* |
1024 | * We own this thread, nobody else can reap it. | |
1025 | */ | |
1026 | read_unlock(&tasklist_lock); | |
1027 | sched_annotate_sleep(); | |
f6507f83 | 1028 | |
befca967 | 1029 | /* |
f6507f83 | 1030 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1031 | */ |
f6507f83 | 1032 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1033 | struct signal_struct *sig = p->signal; |
1034 | struct signal_struct *psig = current->signal; | |
1f10206c | 1035 | unsigned long maxrss; |
5613fda9 | 1036 | u64 tgutime, tgstime; |
3795e161 | 1037 | |
1da177e4 LT |
1038 | /* |
1039 | * The resource counters for the group leader are in its | |
1040 | * own task_struct. Those for dead threads in the group | |
1041 | * are in its signal_struct, as are those for the child | |
1042 | * processes it has previously reaped. All these | |
1043 | * accumulate in the parent's signal_struct c* fields. | |
1044 | * | |
1045 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1046 | * p->signal fields because the whole thread group is dead |
1047 | * and nobody can change them. | |
1048 | * | |
1049 | * psig->stats_lock also protects us from our sub-theads | |
1050 | * which can reap other children at the same time. Until | |
1051 | * we change k_getrusage()-like users to rely on this lock | |
1052 | * we have to take ->siglock as well. | |
0cf55e1e | 1053 | * |
a0be55de IA |
1054 | * We use thread_group_cputime_adjusted() to get times for |
1055 | * the thread group, which consolidates times for all threads | |
1056 | * in the group including the group leader. | |
1da177e4 | 1057 | */ |
e80d0a1a | 1058 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1059 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1060 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1061 | psig->cutime += tgutime + sig->cutime; |
1062 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1063 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1064 | psig->cmin_flt += |
1065 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1066 | psig->cmaj_flt += | |
1067 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1068 | psig->cnvcsw += | |
1069 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1070 | psig->cnivcsw += | |
1071 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1072 | psig->cinblock += |
1073 | task_io_get_inblock(p) + | |
1074 | sig->inblock + sig->cinblock; | |
1075 | psig->coublock += | |
1076 | task_io_get_oublock(p) + | |
1077 | sig->oublock + sig->coublock; | |
1f10206c JP |
1078 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1079 | if (psig->cmaxrss < maxrss) | |
1080 | psig->cmaxrss = maxrss; | |
5995477a AR |
1081 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1082 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1083 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1084 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1085 | } |
1086 | ||
ce72a16f AV |
1087 | if (wo->wo_rusage) |
1088 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
1da177e4 LT |
1089 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1090 | ? p->signal->group_exit_code : p->exit_code; | |
359566fa | 1091 | wo->wo_stat = status; |
2f4e6e2a | 1092 | |
b4360690 | 1093 | if (state == EXIT_TRACE) { |
1da177e4 | 1094 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1095 | /* We dropped tasklist, ptracer could die and untrace */ |
1096 | ptrace_unlink(p); | |
b4360690 ON |
1097 | |
1098 | /* If parent wants a zombie, don't release it now */ | |
1099 | state = EXIT_ZOMBIE; | |
1100 | if (do_notify_parent(p, p->exit_signal)) | |
1101 | state = EXIT_DEAD; | |
abd50b39 | 1102 | p->exit_state = state; |
1da177e4 LT |
1103 | write_unlock_irq(&tasklist_lock); |
1104 | } | |
abd50b39 | 1105 | if (state == EXIT_DEAD) |
1da177e4 | 1106 | release_task(p); |
2f4e6e2a | 1107 | |
76d9871e AV |
1108 | out_info: |
1109 | infop = wo->wo_info; | |
1110 | if (infop) { | |
1111 | if ((status & 0x7f) == 0) { | |
1112 | infop->cause = CLD_EXITED; | |
1113 | infop->status = status >> 8; | |
1114 | } else { | |
1115 | infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1116 | infop->status = status & 0x7f; | |
1117 | } | |
1118 | infop->pid = pid; | |
1119 | infop->uid = uid; | |
1120 | } | |
1121 | ||
67d7ddde | 1122 | return pid; |
1da177e4 LT |
1123 | } |
1124 | ||
90bc8d8b ON |
1125 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1126 | { | |
1127 | if (ptrace) { | |
570ac933 | 1128 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1129 | return &p->exit_code; |
1130 | } else { | |
1131 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1132 | return &p->signal->group_exit_code; | |
1133 | } | |
1134 | return NULL; | |
1135 | } | |
1136 | ||
19e27463 TH |
1137 | /** |
1138 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1139 | * @wo: wait options | |
1140 | * @ptrace: is the wait for ptrace | |
1141 | * @p: task to wait for | |
1142 | * | |
1143 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1144 | * | |
1145 | * CONTEXT: | |
1146 | * read_lock(&tasklist_lock), which is released if return value is | |
1147 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1148 | * | |
1149 | * RETURNS: | |
1150 | * 0 if wait condition didn't exist and search for other wait conditions | |
1151 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1152 | * success, implies that tasklist_lock is released and wait condition | |
1153 | * search should terminate. | |
1da177e4 | 1154 | */ |
9e8ae01d ON |
1155 | static int wait_task_stopped(struct wait_opts *wo, |
1156 | int ptrace, struct task_struct *p) | |
1da177e4 | 1157 | { |
67d7ddde AV |
1158 | struct waitid_info *infop; |
1159 | int exit_code, *p_code, why; | |
ee7c82da | 1160 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1161 | pid_t pid; |
1da177e4 | 1162 | |
47918025 ON |
1163 | /* |
1164 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1165 | */ | |
9e8ae01d | 1166 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1167 | return 0; |
1168 | ||
19e27463 TH |
1169 | if (!task_stopped_code(p, ptrace)) |
1170 | return 0; | |
1171 | ||
ee7c82da ON |
1172 | exit_code = 0; |
1173 | spin_lock_irq(&p->sighand->siglock); | |
1174 | ||
90bc8d8b ON |
1175 | p_code = task_stopped_code(p, ptrace); |
1176 | if (unlikely(!p_code)) | |
ee7c82da ON |
1177 | goto unlock_sig; |
1178 | ||
90bc8d8b | 1179 | exit_code = *p_code; |
ee7c82da ON |
1180 | if (!exit_code) |
1181 | goto unlock_sig; | |
1182 | ||
9e8ae01d | 1183 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1184 | *p_code = 0; |
ee7c82da | 1185 | |
8ca937a6 | 1186 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1187 | unlock_sig: |
1188 | spin_unlock_irq(&p->sighand->siglock); | |
1189 | if (!exit_code) | |
1da177e4 LT |
1190 | return 0; |
1191 | ||
1192 | /* | |
1193 | * Now we are pretty sure this task is interesting. | |
1194 | * Make sure it doesn't get reaped out from under us while we | |
1195 | * give up the lock and then examine it below. We don't want to | |
1196 | * keep holding onto the tasklist_lock while we call getrusage and | |
1197 | * possibly take page faults for user memory. | |
1198 | */ | |
1199 | get_task_struct(p); | |
6c5f3e7b | 1200 | pid = task_pid_vnr(p); |
f470021a | 1201 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1202 | read_unlock(&tasklist_lock); |
1029a2b5 | 1203 | sched_annotate_sleep(); |
e61a2502 AV |
1204 | if (wo->wo_rusage) |
1205 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1206 | put_task_struct(p); |
1da177e4 | 1207 | |
bb380ec3 AV |
1208 | if (likely(!(wo->wo_flags & WNOWAIT))) |
1209 | wo->wo_stat = (exit_code << 8) | 0x7f; | |
1da177e4 | 1210 | |
9e8ae01d | 1211 | infop = wo->wo_info; |
67d7ddde AV |
1212 | if (infop) { |
1213 | infop->cause = why; | |
1214 | infop->status = exit_code; | |
1215 | infop->pid = pid; | |
1216 | infop->uid = uid; | |
1217 | } | |
67d7ddde | 1218 | return pid; |
1da177e4 LT |
1219 | } |
1220 | ||
1221 | /* | |
1222 | * Handle do_wait work for one task in a live, non-stopped state. | |
1223 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1224 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1225 | * released the lock and the system call should return. | |
1226 | */ | |
9e8ae01d | 1227 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1228 | { |
bb380ec3 | 1229 | struct waitid_info *infop; |
1da177e4 LT |
1230 | pid_t pid; |
1231 | uid_t uid; | |
1232 | ||
9e8ae01d | 1233 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1234 | return 0; |
1235 | ||
1da177e4 LT |
1236 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1237 | return 0; | |
1238 | ||
1239 | spin_lock_irq(&p->sighand->siglock); | |
1240 | /* Re-check with the lock held. */ | |
1241 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1242 | spin_unlock_irq(&p->sighand->siglock); | |
1243 | return 0; | |
1244 | } | |
9e8ae01d | 1245 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1246 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1247 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1248 | spin_unlock_irq(&p->sighand->siglock); |
1249 | ||
6c5f3e7b | 1250 | pid = task_pid_vnr(p); |
1da177e4 LT |
1251 | get_task_struct(p); |
1252 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1253 | sched_annotate_sleep(); |
e61a2502 AV |
1254 | if (wo->wo_rusage) |
1255 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1256 | put_task_struct(p); |
1da177e4 | 1257 | |
bb380ec3 AV |
1258 | infop = wo->wo_info; |
1259 | if (!infop) { | |
359566fa | 1260 | wo->wo_stat = 0xffff; |
1da177e4 | 1261 | } else { |
bb380ec3 AV |
1262 | infop->cause = CLD_CONTINUED; |
1263 | infop->pid = pid; | |
1264 | infop->uid = uid; | |
1265 | infop->status = SIGCONT; | |
1da177e4 | 1266 | } |
bb380ec3 | 1267 | return pid; |
1da177e4 LT |
1268 | } |
1269 | ||
98abed02 RM |
1270 | /* |
1271 | * Consider @p for a wait by @parent. | |
1272 | * | |
9e8ae01d | 1273 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1274 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1275 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1276 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1277 | * or still -ECHILD. |
98abed02 | 1278 | */ |
b6e763f0 ON |
1279 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1280 | struct task_struct *p) | |
98abed02 | 1281 | { |
3245d6ac ON |
1282 | /* |
1283 | * We can race with wait_task_zombie() from another thread. | |
1284 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1285 | * can't confuse the checks below. | |
1286 | */ | |
6aa7de05 | 1287 | int exit_state = READ_ONCE(p->exit_state); |
b3ab0316 ON |
1288 | int ret; |
1289 | ||
3245d6ac | 1290 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1291 | return 0; |
1292 | ||
bf959931 | 1293 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1294 | if (!ret) |
98abed02 RM |
1295 | return ret; |
1296 | ||
3245d6ac | 1297 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1298 | /* |
abd50b39 ON |
1299 | * ptrace == 0 means we are the natural parent. In this case |
1300 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1301 | */ |
abd50b39 | 1302 | if (likely(!ptrace)) |
50b8d257 | 1303 | wo->notask_error = 0; |
823b018e | 1304 | return 0; |
50b8d257 | 1305 | } |
823b018e | 1306 | |
377d75da ON |
1307 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1308 | /* | |
1309 | * If it is traced by its real parent's group, just pretend | |
1310 | * the caller is ptrace_do_wait() and reap this child if it | |
1311 | * is zombie. | |
1312 | * | |
1313 | * This also hides group stop state from real parent; otherwise | |
1314 | * a single stop can be reported twice as group and ptrace stop. | |
1315 | * If a ptracer wants to distinguish these two events for its | |
1316 | * own children it should create a separate process which takes | |
1317 | * the role of real parent. | |
1318 | */ | |
1319 | if (!ptrace_reparented(p)) | |
1320 | ptrace = 1; | |
1321 | } | |
1322 | ||
45cb24a1 | 1323 | /* slay zombie? */ |
3245d6ac | 1324 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1325 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1326 | if (!delay_group_leader(p)) { |
1327 | /* | |
1328 | * A zombie ptracee is only visible to its ptracer. | |
1329 | * Notification and reaping will be cascaded to the | |
1330 | * real parent when the ptracer detaches. | |
1331 | */ | |
1332 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1333 | return wait_task_zombie(wo, p); | |
1334 | } | |
98abed02 | 1335 | |
f470021a | 1336 | /* |
9b84cca2 TH |
1337 | * Allow access to stopped/continued state via zombie by |
1338 | * falling through. Clearing of notask_error is complex. | |
1339 | * | |
1340 | * When !@ptrace: | |
1341 | * | |
1342 | * If WEXITED is set, notask_error should naturally be | |
1343 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1344 | * so, if there are live subthreads, there are events to | |
1345 | * wait for. If all subthreads are dead, it's still safe | |
1346 | * to clear - this function will be called again in finite | |
1347 | * amount time once all the subthreads are released and | |
1348 | * will then return without clearing. | |
1349 | * | |
1350 | * When @ptrace: | |
1351 | * | |
1352 | * Stopped state is per-task and thus can't change once the | |
1353 | * target task dies. Only continued and exited can happen. | |
1354 | * Clear notask_error if WCONTINUED | WEXITED. | |
1355 | */ | |
1356 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1357 | wo->notask_error = 0; | |
1358 | } else { | |
1359 | /* | |
1360 | * @p is alive and it's gonna stop, continue or exit, so | |
1361 | * there always is something to wait for. | |
f470021a | 1362 | */ |
9e8ae01d | 1363 | wo->notask_error = 0; |
f470021a RM |
1364 | } |
1365 | ||
98abed02 | 1366 | /* |
45cb24a1 TH |
1367 | * Wait for stopped. Depending on @ptrace, different stopped state |
1368 | * is used and the two don't interact with each other. | |
98abed02 | 1369 | */ |
19e27463 TH |
1370 | ret = wait_task_stopped(wo, ptrace, p); |
1371 | if (ret) | |
1372 | return ret; | |
98abed02 RM |
1373 | |
1374 | /* | |
45cb24a1 TH |
1375 | * Wait for continued. There's only one continued state and the |
1376 | * ptracer can consume it which can confuse the real parent. Don't | |
1377 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1378 | */ |
9e8ae01d | 1379 | return wait_task_continued(wo, p); |
98abed02 RM |
1380 | } |
1381 | ||
1382 | /* | |
1383 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1384 | * | |
9e8ae01d | 1385 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1386 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1387 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1388 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1389 | * or still -ECHILD. |
98abed02 | 1390 | */ |
9e8ae01d | 1391 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1392 | { |
1393 | struct task_struct *p; | |
1394 | ||
1395 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1396 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1397 | |
9cd80bbb ON |
1398 | if (ret) |
1399 | return ret; | |
98abed02 RM |
1400 | } |
1401 | ||
1402 | return 0; | |
1403 | } | |
1404 | ||
9e8ae01d | 1405 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1406 | { |
1407 | struct task_struct *p; | |
1408 | ||
f470021a | 1409 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1410 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1411 | |
f470021a | 1412 | if (ret) |
98abed02 | 1413 | return ret; |
98abed02 RM |
1414 | } |
1415 | ||
1416 | return 0; | |
1417 | } | |
1418 | ||
ac6424b9 | 1419 | static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode, |
0b7570e7 ON |
1420 | int sync, void *key) |
1421 | { | |
1422 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1423 | child_wait); | |
1424 | struct task_struct *p = key; | |
1425 | ||
5c01ba49 | 1426 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1427 | return 0; |
1428 | ||
b4fe5182 ON |
1429 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1430 | return 0; | |
1431 | ||
0b7570e7 ON |
1432 | return default_wake_function(wait, mode, sync, key); |
1433 | } | |
1434 | ||
a7f0765e ON |
1435 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1436 | { | |
0b7570e7 ON |
1437 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1438 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1439 | } |
1440 | ||
9e8ae01d | 1441 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1442 | { |
1da177e4 | 1443 | struct task_struct *tsk; |
98abed02 | 1444 | int retval; |
1da177e4 | 1445 | |
9e8ae01d | 1446 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1447 | |
0b7570e7 ON |
1448 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1449 | wo->child_wait.private = current; | |
1450 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1451 | repeat: |
98abed02 | 1452 | /* |
3da56d16 | 1453 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1454 | * We will clear ->notask_error to zero if we see any child that |
1455 | * might later match our criteria, even if we are not able to reap | |
1456 | * it yet. | |
98abed02 | 1457 | */ |
64a16caf | 1458 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1459 | if ((wo->wo_type < PIDTYPE_MAX) && |
1460 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1461 | goto notask; |
161550d7 | 1462 | |
f95d39d1 | 1463 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1464 | read_lock(&tasklist_lock); |
1465 | tsk = current; | |
1466 | do { | |
64a16caf ON |
1467 | retval = do_wait_thread(wo, tsk); |
1468 | if (retval) | |
1469 | goto end; | |
9e8ae01d | 1470 | |
64a16caf ON |
1471 | retval = ptrace_do_wait(wo, tsk); |
1472 | if (retval) | |
98abed02 | 1473 | goto end; |
98abed02 | 1474 | |
9e8ae01d | 1475 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1476 | break; |
a3f6dfb7 | 1477 | } while_each_thread(current, tsk); |
1da177e4 | 1478 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1479 | |
64a16caf | 1480 | notask: |
9e8ae01d ON |
1481 | retval = wo->notask_error; |
1482 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1483 | retval = -ERESTARTSYS; |
98abed02 RM |
1484 | if (!signal_pending(current)) { |
1485 | schedule(); | |
1486 | goto repeat; | |
1487 | } | |
1da177e4 | 1488 | } |
1da177e4 | 1489 | end: |
f95d39d1 | 1490 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1491 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1492 | return retval; |
1493 | } | |
1494 | ||
3695eae5 CB |
1495 | static struct pid *pidfd_get_pid(unsigned int fd) |
1496 | { | |
1497 | struct fd f; | |
1498 | struct pid *pid; | |
1499 | ||
1500 | f = fdget(fd); | |
1501 | if (!f.file) | |
1502 | return ERR_PTR(-EBADF); | |
1503 | ||
1504 | pid = pidfd_pid(f.file); | |
1505 | if (!IS_ERR(pid)) | |
1506 | get_pid(pid); | |
1507 | ||
1508 | fdput(f); | |
1509 | return pid; | |
1510 | } | |
1511 | ||
67d7ddde | 1512 | static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, |
ce72a16f | 1513 | int options, struct rusage *ru) |
1da177e4 | 1514 | { |
9e8ae01d | 1515 | struct wait_opts wo; |
161550d7 EB |
1516 | struct pid *pid = NULL; |
1517 | enum pid_type type; | |
1da177e4 LT |
1518 | long ret; |
1519 | ||
91c4e8ea ON |
1520 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1521 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1522 | return -EINVAL; |
1523 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1524 | return -EINVAL; | |
1525 | ||
1526 | switch (which) { | |
1527 | case P_ALL: | |
161550d7 | 1528 | type = PIDTYPE_MAX; |
1da177e4 LT |
1529 | break; |
1530 | case P_PID: | |
161550d7 EB |
1531 | type = PIDTYPE_PID; |
1532 | if (upid <= 0) | |
1da177e4 | 1533 | return -EINVAL; |
3695eae5 CB |
1534 | |
1535 | pid = find_get_pid(upid); | |
1da177e4 LT |
1536 | break; |
1537 | case P_PGID: | |
161550d7 | 1538 | type = PIDTYPE_PGID; |
821cc7b0 | 1539 | if (upid < 0) |
1da177e4 | 1540 | return -EINVAL; |
3695eae5 | 1541 | |
821cc7b0 EB |
1542 | if (upid) |
1543 | pid = find_get_pid(upid); | |
1544 | else | |
1545 | pid = get_task_pid(current, PIDTYPE_PGID); | |
3695eae5 CB |
1546 | break; |
1547 | case P_PIDFD: | |
1548 | type = PIDTYPE_PID; | |
1549 | if (upid < 0) | |
1da177e4 | 1550 | return -EINVAL; |
3695eae5 CB |
1551 | |
1552 | pid = pidfd_get_pid(upid); | |
1553 | if (IS_ERR(pid)) | |
1554 | return PTR_ERR(pid); | |
1da177e4 LT |
1555 | break; |
1556 | default: | |
1557 | return -EINVAL; | |
1558 | } | |
1559 | ||
9e8ae01d ON |
1560 | wo.wo_type = type; |
1561 | wo.wo_pid = pid; | |
1562 | wo.wo_flags = options; | |
1563 | wo.wo_info = infop; | |
9e8ae01d ON |
1564 | wo.wo_rusage = ru; |
1565 | ret = do_wait(&wo); | |
dfe16dfa | 1566 | |
161550d7 | 1567 | put_pid(pid); |
1da177e4 LT |
1568 | return ret; |
1569 | } | |
1570 | ||
ce72a16f AV |
1571 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1572 | infop, int, options, struct rusage __user *, ru) | |
1573 | { | |
1574 | struct rusage r; | |
67d7ddde AV |
1575 | struct waitid_info info = {.status = 0}; |
1576 | long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL); | |
634a8160 | 1577 | int signo = 0; |
6c85501f | 1578 | |
634a8160 AV |
1579 | if (err > 0) { |
1580 | signo = SIGCHLD; | |
1581 | err = 0; | |
ce72a16f AV |
1582 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) |
1583 | return -EFAULT; | |
1584 | } | |
67d7ddde AV |
1585 | if (!infop) |
1586 | return err; | |
1587 | ||
594cc251 | 1588 | if (!user_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1589 | return -EFAULT; |
96ca579a | 1590 | |
634a8160 | 1591 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1592 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1593 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1594 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1595 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1596 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
1597 | user_access_end(); | |
ce72a16f | 1598 | return err; |
4c48abe9 AV |
1599 | Efault: |
1600 | user_access_end(); | |
1601 | return -EFAULT; | |
ce72a16f AV |
1602 | } |
1603 | ||
92ebce5a AV |
1604 | long kernel_wait4(pid_t upid, int __user *stat_addr, int options, |
1605 | struct rusage *ru) | |
1da177e4 | 1606 | { |
9e8ae01d | 1607 | struct wait_opts wo; |
161550d7 EB |
1608 | struct pid *pid = NULL; |
1609 | enum pid_type type; | |
1da177e4 LT |
1610 | long ret; |
1611 | ||
1612 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1613 | __WNOTHREAD|__WCLONE|__WALL)) | |
1614 | return -EINVAL; | |
161550d7 | 1615 | |
dd83c161 | 1616 | /* -INT_MIN is not defined */ |
1617 | if (upid == INT_MIN) | |
1618 | return -ESRCH; | |
1619 | ||
161550d7 EB |
1620 | if (upid == -1) |
1621 | type = PIDTYPE_MAX; | |
1622 | else if (upid < 0) { | |
1623 | type = PIDTYPE_PGID; | |
1624 | pid = find_get_pid(-upid); | |
1625 | } else if (upid == 0) { | |
1626 | type = PIDTYPE_PGID; | |
2ae448ef | 1627 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1628 | } else /* upid > 0 */ { |
1629 | type = PIDTYPE_PID; | |
1630 | pid = find_get_pid(upid); | |
1631 | } | |
1632 | ||
9e8ae01d ON |
1633 | wo.wo_type = type; |
1634 | wo.wo_pid = pid; | |
1635 | wo.wo_flags = options | WEXITED; | |
1636 | wo.wo_info = NULL; | |
359566fa | 1637 | wo.wo_stat = 0; |
9e8ae01d ON |
1638 | wo.wo_rusage = ru; |
1639 | ret = do_wait(&wo); | |
161550d7 | 1640 | put_pid(pid); |
359566fa AV |
1641 | if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr)) |
1642 | ret = -EFAULT; | |
1da177e4 | 1643 | |
1da177e4 LT |
1644 | return ret; |
1645 | } | |
1646 | ||
ce72a16f AV |
1647 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1648 | int, options, struct rusage __user *, ru) | |
1649 | { | |
1650 | struct rusage r; | |
1651 | long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL); | |
1652 | ||
1653 | if (err > 0) { | |
1654 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) | |
1655 | return -EFAULT; | |
1656 | } | |
1657 | return err; | |
1658 | } | |
1659 | ||
1da177e4 LT |
1660 | #ifdef __ARCH_WANT_SYS_WAITPID |
1661 | ||
1662 | /* | |
1663 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1664 | * implemented by calling sys_wait4() from libc.a. | |
1665 | */ | |
17da2bd9 | 1666 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 | 1667 | { |
d300b610 | 1668 | return kernel_wait4(pid, stat_addr, options, NULL); |
1da177e4 LT |
1669 | } |
1670 | ||
1671 | #endif | |
7e95a225 AV |
1672 | |
1673 | #ifdef CONFIG_COMPAT | |
1674 | COMPAT_SYSCALL_DEFINE4(wait4, | |
1675 | compat_pid_t, pid, | |
1676 | compat_uint_t __user *, stat_addr, | |
1677 | int, options, | |
1678 | struct compat_rusage __user *, ru) | |
1679 | { | |
ce72a16f AV |
1680 | struct rusage r; |
1681 | long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL); | |
1682 | if (err > 0) { | |
1683 | if (ru && put_compat_rusage(&r, ru)) | |
1684 | return -EFAULT; | |
7e95a225 | 1685 | } |
ce72a16f | 1686 | return err; |
7e95a225 AV |
1687 | } |
1688 | ||
1689 | COMPAT_SYSCALL_DEFINE5(waitid, | |
1690 | int, which, compat_pid_t, pid, | |
1691 | struct compat_siginfo __user *, infop, int, options, | |
1692 | struct compat_rusage __user *, uru) | |
1693 | { | |
7e95a225 | 1694 | struct rusage ru; |
67d7ddde AV |
1695 | struct waitid_info info = {.status = 0}; |
1696 | long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL); | |
634a8160 AV |
1697 | int signo = 0; |
1698 | if (err > 0) { | |
1699 | signo = SIGCHLD; | |
1700 | err = 0; | |
6c85501f AV |
1701 | if (uru) { |
1702 | /* kernel_waitid() overwrites everything in ru */ | |
1703 | if (COMPAT_USE_64BIT_TIME) | |
1704 | err = copy_to_user(uru, &ru, sizeof(ru)); | |
1705 | else | |
1706 | err = put_compat_rusage(&ru, uru); | |
1707 | if (err) | |
1708 | return -EFAULT; | |
1709 | } | |
7e95a225 AV |
1710 | } |
1711 | ||
4c48abe9 AV |
1712 | if (!infop) |
1713 | return err; | |
1714 | ||
594cc251 | 1715 | if (!user_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1716 | return -EFAULT; |
96ca579a | 1717 | |
634a8160 | 1718 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1719 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1720 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1721 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1722 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1723 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
1724 | user_access_end(); | |
67d7ddde | 1725 | return err; |
4c48abe9 AV |
1726 | Efault: |
1727 | user_access_end(); | |
1728 | return -EFAULT; | |
7e95a225 AV |
1729 | } |
1730 | #endif | |
7c2c11b2 SM |
1731 | |
1732 | __weak void abort(void) | |
1733 | { | |
1734 | BUG(); | |
1735 | ||
1736 | /* if that doesn't kill us, halt */ | |
1737 | panic("Oops failed to kill thread"); | |
1738 | } | |
dc8635b7 | 1739 | EXPORT_SYMBOL(abort); |