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); |
b95e31c0 | 106 | if (group_dead) |
6a14c5c9 | 107 | posix_cpu_timers_exit_group(tsk); |
baa73d9e | 108 | #endif |
e0a70217 | 109 | |
baa73d9e NP |
110 | if (group_dead) { |
111 | tty = sig->tty; | |
112 | sig->tty = NULL; | |
113 | } else { | |
6a14c5c9 ON |
114 | /* |
115 | * If there is any task waiting for the group exit | |
116 | * then notify it: | |
117 | */ | |
d344193a | 118 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 119 | wake_up_process(sig->group_exit_task); |
6db840fa | 120 | |
6a14c5c9 ON |
121 | if (tsk == sig->curr_target) |
122 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
123 | } |
124 | ||
53d3eaa3 NP |
125 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
126 | sizeof(unsigned long long)); | |
127 | ||
90ed9cbe | 128 | /* |
26e75b5c ON |
129 | * Accumulate here the counters for all threads as they die. We could |
130 | * skip the group leader because it is the last user of signal_struct, | |
131 | * but we want to avoid the race with thread_group_cputime() which can | |
132 | * see the empty ->thread_head list. | |
90ed9cbe RR |
133 | */ |
134 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 135 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
136 | sig->utime += utime; |
137 | sig->stime += stime; | |
138 | sig->gtime += task_gtime(tsk); | |
139 | sig->min_flt += tsk->min_flt; | |
140 | sig->maj_flt += tsk->maj_flt; | |
141 | sig->nvcsw += tsk->nvcsw; | |
142 | sig->nivcsw += tsk->nivcsw; | |
143 | sig->inblock += task_io_get_inblock(tsk); | |
144 | sig->oublock += task_io_get_oublock(tsk); | |
145 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
146 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 147 | sig->nr_threads--; |
d40e48e0 | 148 | __unhash_process(tsk, group_dead); |
e78c3496 | 149 | write_sequnlock(&sig->stats_lock); |
5876700c | 150 | |
da7978b0 ON |
151 | /* |
152 | * Do this under ->siglock, we can race with another thread | |
153 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
154 | */ | |
155 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 156 | tsk->sighand = NULL; |
6a14c5c9 | 157 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 158 | |
a7e5328a | 159 | __cleanup_sighand(sighand); |
a0be55de | 160 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 161 | if (group_dead) { |
6a14c5c9 | 162 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 163 | tty_kref_put(tty); |
6a14c5c9 ON |
164 | } |
165 | } | |
166 | ||
8c7904a0 EB |
167 | static void delayed_put_task_struct(struct rcu_head *rhp) |
168 | { | |
0a16b607 MD |
169 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
170 | ||
4e231c79 | 171 | perf_event_delayed_put(tsk); |
0a16b607 MD |
172 | trace_sched_process_free(tsk); |
173 | put_task_struct(tsk); | |
8c7904a0 EB |
174 | } |
175 | ||
3fbd7ee2 EB |
176 | void put_task_struct_rcu_user(struct task_struct *task) |
177 | { | |
178 | if (refcount_dec_and_test(&task->rcu_users)) | |
179 | call_rcu(&task->rcu, delayed_put_task_struct); | |
180 | } | |
f470021a | 181 | |
a0be55de | 182 | void release_task(struct task_struct *p) |
1da177e4 | 183 | { |
36c8b586 | 184 | struct task_struct *leader; |
7bc3e6e5 | 185 | struct pid *thread_pid; |
1da177e4 | 186 | int zap_leader; |
1f09f974 | 187 | repeat: |
c69e8d9c | 188 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
189 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
190 | rcu_read_lock(); | |
c69e8d9c | 191 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 192 | rcu_read_unlock(); |
c69e8d9c | 193 | |
6b115bf5 | 194 | cgroup_release(p); |
0203026b | 195 | |
1da177e4 | 196 | write_lock_irq(&tasklist_lock); |
a288eecc | 197 | ptrace_release_task(p); |
7bc3e6e5 | 198 | thread_pid = get_pid(p->thread_pid); |
1da177e4 | 199 | __exit_signal(p); |
35f5cad8 | 200 | |
1da177e4 LT |
201 | /* |
202 | * If we are the last non-leader member of the thread | |
203 | * group, and the leader is zombie, then notify the | |
204 | * group leader's parent process. (if it wants notification.) | |
205 | */ | |
206 | zap_leader = 0; | |
207 | leader = p->group_leader; | |
a0be55de IA |
208 | if (leader != p && thread_group_empty(leader) |
209 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
210 | /* |
211 | * If we were the last child thread and the leader has | |
212 | * exited already, and the leader's parent ignores SIGCHLD, | |
213 | * then we are the one who should release the leader. | |
dae33574 | 214 | */ |
86773473 | 215 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
216 | if (zap_leader) |
217 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
218 | } |
219 | ||
1da177e4 | 220 | write_unlock_irq(&tasklist_lock); |
7bc3e6e5 | 221 | proc_flush_pid(thread_pid); |
1da177e4 | 222 | release_thread(p); |
3fbd7ee2 | 223 | put_task_struct_rcu_user(p); |
1da177e4 LT |
224 | |
225 | p = leader; | |
226 | if (unlikely(zap_leader)) | |
227 | goto repeat; | |
228 | } | |
229 | ||
8f95c90c DB |
230 | void rcuwait_wake_up(struct rcuwait *w) |
231 | { | |
232 | struct task_struct *task; | |
233 | ||
234 | rcu_read_lock(); | |
235 | ||
236 | /* | |
237 | * Order condition vs @task, such that everything prior to the load | |
238 | * of @task is visible. This is the condition as to why the user called | |
239 | * rcuwait_trywake() in the first place. Pairs with set_current_state() | |
240 | * barrier (A) in rcuwait_wait_event(). | |
241 | * | |
242 | * WAIT WAKE | |
243 | * [S] tsk = current [S] cond = true | |
244 | * MB (A) MB (B) | |
245 | * [L] cond [L] tsk | |
246 | */ | |
6dc080ee | 247 | smp_mb(); /* (B) */ |
8f95c90c | 248 | |
8f95c90c DB |
249 | task = rcu_dereference(w->task); |
250 | if (task) | |
251 | wake_up_process(task); | |
252 | rcu_read_unlock(); | |
253 | } | |
ac8dec42 | 254 | EXPORT_SYMBOL_GPL(rcuwait_wake_up); |
8f95c90c | 255 | |
1da177e4 LT |
256 | /* |
257 | * Determine if a process group is "orphaned", according to the POSIX | |
258 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
259 | * by terminal-generated stop signals. Newly orphaned process groups are | |
260 | * to receive a SIGHUP and a SIGCONT. | |
261 | * | |
262 | * "I ask you, have you ever known what it is to be an orphan?" | |
263 | */ | |
a0be55de IA |
264 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
265 | struct task_struct *ignored_task) | |
1da177e4 LT |
266 | { |
267 | struct task_struct *p; | |
1da177e4 | 268 | |
0475ac08 | 269 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
270 | if ((p == ignored_task) || |
271 | (p->exit_state && thread_group_empty(p)) || | |
272 | is_global_init(p->real_parent)) | |
1da177e4 | 273 | continue; |
05e83df6 | 274 | |
0475ac08 | 275 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
276 | task_session(p->real_parent) == task_session(p)) |
277 | return 0; | |
0475ac08 | 278 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
279 | |
280 | return 1; | |
1da177e4 LT |
281 | } |
282 | ||
3e7cd6c4 | 283 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
284 | { |
285 | int retval; | |
286 | ||
287 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 288 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
289 | read_unlock(&tasklist_lock); |
290 | ||
291 | return retval; | |
292 | } | |
293 | ||
961c4675 | 294 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 295 | { |
1da177e4 LT |
296 | struct task_struct *p; |
297 | ||
0475ac08 | 298 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
299 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
300 | return true; | |
0475ac08 | 301 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
302 | |
303 | return false; | |
1da177e4 LT |
304 | } |
305 | ||
f49ee505 ON |
306 | /* |
307 | * Check to see if any process groups have become orphaned as | |
308 | * a result of our exiting, and if they have any stopped jobs, | |
309 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
310 | */ | |
311 | static void | |
312 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
313 | { | |
314 | struct pid *pgrp = task_pgrp(tsk); | |
315 | struct task_struct *ignored_task = tsk; | |
316 | ||
317 | if (!parent) | |
a0be55de IA |
318 | /* exit: our father is in a different pgrp than |
319 | * we are and we were the only connection outside. | |
320 | */ | |
f49ee505 ON |
321 | parent = tsk->real_parent; |
322 | else | |
323 | /* reparent: our child is in a different pgrp than | |
324 | * we are, and it was the only connection outside. | |
325 | */ | |
326 | ignored_task = NULL; | |
327 | ||
328 | if (task_pgrp(parent) != pgrp && | |
329 | task_session(parent) == task_session(tsk) && | |
330 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
331 | has_stopped_jobs(pgrp)) { | |
332 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
333 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
334 | } | |
335 | } | |
336 | ||
f98bafa0 | 337 | #ifdef CONFIG_MEMCG |
cf475ad2 | 338 | /* |
733eda7a | 339 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 340 | */ |
cf475ad2 BS |
341 | void mm_update_next_owner(struct mm_struct *mm) |
342 | { | |
343 | struct task_struct *c, *g, *p = current; | |
344 | ||
345 | retry: | |
733eda7a KH |
346 | /* |
347 | * If the exiting or execing task is not the owner, it's | |
348 | * someone else's problem. | |
349 | */ | |
350 | if (mm->owner != p) | |
cf475ad2 | 351 | return; |
733eda7a KH |
352 | /* |
353 | * The current owner is exiting/execing and there are no other | |
354 | * candidates. Do not leave the mm pointing to a possibly | |
355 | * freed task structure. | |
356 | */ | |
357 | if (atomic_read(&mm->mm_users) <= 1) { | |
987717e5 | 358 | WRITE_ONCE(mm->owner, NULL); |
733eda7a KH |
359 | return; |
360 | } | |
cf475ad2 BS |
361 | |
362 | read_lock(&tasklist_lock); | |
363 | /* | |
364 | * Search in the children | |
365 | */ | |
366 | list_for_each_entry(c, &p->children, sibling) { | |
367 | if (c->mm == mm) | |
368 | goto assign_new_owner; | |
369 | } | |
370 | ||
371 | /* | |
372 | * Search in the siblings | |
373 | */ | |
dea33cfd | 374 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
375 | if (c->mm == mm) |
376 | goto assign_new_owner; | |
377 | } | |
378 | ||
379 | /* | |
f87fb599 | 380 | * Search through everything else, we should not get here often. |
cf475ad2 | 381 | */ |
39af1765 ON |
382 | for_each_process(g) { |
383 | if (g->flags & PF_KTHREAD) | |
384 | continue; | |
385 | for_each_thread(g, c) { | |
386 | if (c->mm == mm) | |
387 | goto assign_new_owner; | |
388 | if (c->mm) | |
389 | break; | |
390 | } | |
f87fb599 | 391 | } |
cf475ad2 | 392 | read_unlock(&tasklist_lock); |
31a78f23 BS |
393 | /* |
394 | * We found no owner yet mm_users > 1: this implies that we are | |
395 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 396 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 397 | */ |
987717e5 | 398 | WRITE_ONCE(mm->owner, NULL); |
cf475ad2 BS |
399 | return; |
400 | ||
401 | assign_new_owner: | |
402 | BUG_ON(c == p); | |
403 | get_task_struct(c); | |
404 | /* | |
405 | * The task_lock protects c->mm from changing. | |
406 | * We always want mm->owner->mm == mm | |
407 | */ | |
408 | task_lock(c); | |
e5991371 HD |
409 | /* |
410 | * Delay read_unlock() till we have the task_lock() | |
411 | * to ensure that c does not slip away underneath us | |
412 | */ | |
413 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
414 | if (c->mm != mm) { |
415 | task_unlock(c); | |
416 | put_task_struct(c); | |
417 | goto retry; | |
418 | } | |
987717e5 | 419 | WRITE_ONCE(mm->owner, c); |
cf475ad2 BS |
420 | task_unlock(c); |
421 | put_task_struct(c); | |
422 | } | |
f98bafa0 | 423 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 424 | |
1da177e4 LT |
425 | /* |
426 | * Turn us into a lazy TLB process if we | |
427 | * aren't already.. | |
428 | */ | |
0039962a | 429 | static void exit_mm(void) |
1da177e4 | 430 | { |
0039962a | 431 | struct mm_struct *mm = current->mm; |
b564daf8 | 432 | struct core_state *core_state; |
1da177e4 | 433 | |
4610ba7a | 434 | exit_mm_release(current, mm); |
1da177e4 LT |
435 | if (!mm) |
436 | return; | |
4fe7efdb | 437 | sync_mm_rss(mm); |
1da177e4 LT |
438 | /* |
439 | * Serialize with any possible pending coredump. | |
999d9fc1 | 440 | * We must hold mmap_sem around checking core_state |
1da177e4 | 441 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 442 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
443 | * group with ->mm != NULL. |
444 | */ | |
445 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
446 | core_state = mm->core_state; |
447 | if (core_state) { | |
448 | struct core_thread self; | |
a0be55de | 449 | |
1da177e4 | 450 | up_read(&mm->mmap_sem); |
1da177e4 | 451 | |
0039962a | 452 | self.task = current; |
b564daf8 ON |
453 | self.next = xchg(&core_state->dumper.next, &self); |
454 | /* | |
455 | * Implies mb(), the result of xchg() must be visible | |
456 | * to core_state->dumper. | |
457 | */ | |
458 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
459 | complete(&core_state->startup); | |
1da177e4 | 460 | |
a94e2d40 | 461 | for (;;) { |
642fa448 | 462 | set_current_state(TASK_UNINTERRUPTIBLE); |
a94e2d40 ON |
463 | if (!self.task) /* see coredump_finish() */ |
464 | break; | |
80d26af8 | 465 | freezable_schedule(); |
a94e2d40 | 466 | } |
642fa448 | 467 | __set_current_state(TASK_RUNNING); |
1da177e4 LT |
468 | down_read(&mm->mmap_sem); |
469 | } | |
f1f10076 | 470 | mmgrab(mm); |
0039962a | 471 | BUG_ON(mm != current->active_mm); |
1da177e4 | 472 | /* more a memory barrier than a real lock */ |
0039962a DB |
473 | task_lock(current); |
474 | current->mm = NULL; | |
1da177e4 LT |
475 | up_read(&mm->mmap_sem); |
476 | enter_lazy_tlb(mm, current); | |
0039962a | 477 | task_unlock(current); |
cf475ad2 | 478 | mm_update_next_owner(mm); |
1da177e4 | 479 | mmput(mm); |
c32b3cbe | 480 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 481 | exit_oom_victim(); |
1da177e4 LT |
482 | } |
483 | ||
c9dc05bf ON |
484 | static struct task_struct *find_alive_thread(struct task_struct *p) |
485 | { | |
486 | struct task_struct *t; | |
487 | ||
488 | for_each_thread(p, t) { | |
489 | if (!(t->flags & PF_EXITING)) | |
490 | return t; | |
491 | } | |
492 | return NULL; | |
493 | } | |
494 | ||
8fb335e0 AV |
495 | static struct task_struct *find_child_reaper(struct task_struct *father, |
496 | struct list_head *dead) | |
1109909c ON |
497 | __releases(&tasklist_lock) |
498 | __acquires(&tasklist_lock) | |
499 | { | |
500 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
501 | struct task_struct *reaper = pid_ns->child_reaper; | |
8fb335e0 | 502 | struct task_struct *p, *n; |
1109909c ON |
503 | |
504 | if (likely(reaper != father)) | |
505 | return reaper; | |
506 | ||
c9dc05bf ON |
507 | reaper = find_alive_thread(father); |
508 | if (reaper) { | |
1109909c ON |
509 | pid_ns->child_reaper = reaper; |
510 | return reaper; | |
511 | } | |
512 | ||
513 | write_unlock_irq(&tasklist_lock); | |
8fb335e0 AV |
514 | |
515 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
516 | list_del_init(&p->ptrace_entry); | |
517 | release_task(p); | |
518 | } | |
519 | ||
1109909c ON |
520 | zap_pid_ns_processes(pid_ns); |
521 | write_lock_irq(&tasklist_lock); | |
522 | ||
523 | return father; | |
524 | } | |
525 | ||
1da177e4 | 526 | /* |
ebec18a6 LP |
527 | * When we die, we re-parent all our children, and try to: |
528 | * 1. give them to another thread in our thread group, if such a member exists | |
529 | * 2. give it to the first ancestor process which prctl'd itself as a | |
530 | * child_subreaper for its children (like a service manager) | |
531 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 532 | */ |
1109909c ON |
533 | static struct task_struct *find_new_reaper(struct task_struct *father, |
534 | struct task_struct *child_reaper) | |
1da177e4 | 535 | { |
c9dc05bf | 536 | struct task_struct *thread, *reaper; |
1da177e4 | 537 | |
c9dc05bf ON |
538 | thread = find_alive_thread(father); |
539 | if (thread) | |
950bbabb | 540 | return thread; |
1da177e4 | 541 | |
7d24e2df | 542 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 543 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 544 | /* |
175aed3f | 545 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
546 | * We can't check reaper != child_reaper to ensure we do not |
547 | * cross the namespaces, the exiting parent could be injected | |
548 | * by setns() + fork(). | |
549 | * We check pid->level, this is slightly more efficient than | |
550 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 551 | */ |
c6c70f44 ON |
552 | for (reaper = father->real_parent; |
553 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 554 | reaper = reaper->real_parent) { |
175aed3f | 555 | if (reaper == &init_task) |
ebec18a6 LP |
556 | break; |
557 | if (!reaper->signal->is_child_subreaper) | |
558 | continue; | |
c9dc05bf ON |
559 | thread = find_alive_thread(reaper); |
560 | if (thread) | |
561 | return thread; | |
ebec18a6 | 562 | } |
1da177e4 | 563 | } |
762a24be | 564 | |
1109909c | 565 | return child_reaper; |
950bbabb ON |
566 | } |
567 | ||
5dfc80be ON |
568 | /* |
569 | * Any that need to be release_task'd are put on the @dead list. | |
570 | */ | |
9cd80bbb | 571 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
572 | struct list_head *dead) |
573 | { | |
2831096e | 574 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
575 | return; |
576 | ||
abd50b39 | 577 | /* We don't want people slaying init. */ |
5dfc80be ON |
578 | p->exit_signal = SIGCHLD; |
579 | ||
580 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 581 | if (!p->ptrace && |
5dfc80be | 582 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 583 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 584 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 585 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
586 | } |
587 | } | |
588 | ||
589 | kill_orphaned_pgrp(p, father); | |
590 | } | |
591 | ||
482a3767 ON |
592 | /* |
593 | * This does two things: | |
594 | * | |
595 | * A. Make init inherit all the child processes | |
596 | * B. Check to see if any process groups have become orphaned | |
597 | * as a result of our exiting, and if they have any stopped | |
598 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
599 | */ | |
600 | static void forget_original_parent(struct task_struct *father, | |
601 | struct list_head *dead) | |
1da177e4 | 602 | { |
482a3767 | 603 | struct task_struct *p, *t, *reaper; |
762a24be | 604 | |
7c8bd232 | 605 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 606 | exit_ptrace(father, dead); |
f470021a | 607 | |
7c8bd232 | 608 | /* Can drop and reacquire tasklist_lock */ |
8fb335e0 | 609 | reaper = find_child_reaper(father, dead); |
ad9e206a | 610 | if (list_empty(&father->children)) |
482a3767 | 611 | return; |
1109909c ON |
612 | |
613 | reaper = find_new_reaper(father, reaper); | |
2831096e | 614 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 615 | for_each_thread(p, t) { |
22a34c6f MB |
616 | RCU_INIT_POINTER(t->real_parent, reaper); |
617 | BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father)); | |
57a05918 | 618 | if (likely(!t->ptrace)) |
9cd80bbb | 619 | t->parent = t->real_parent; |
9cd80bbb ON |
620 | if (t->pdeath_signal) |
621 | group_send_sig_info(t->pdeath_signal, | |
01024980 EB |
622 | SEND_SIG_NOINFO, t, |
623 | PIDTYPE_TGID); | |
57a05918 | 624 | } |
2831096e ON |
625 | /* |
626 | * If this is a threaded reparent there is no need to | |
627 | * notify anyone anything has happened. | |
628 | */ | |
629 | if (!same_thread_group(reaper, father)) | |
482a3767 | 630 | reparent_leader(father, p, dead); |
1da177e4 | 631 | } |
2831096e | 632 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
633 | } |
634 | ||
635 | /* | |
636 | * Send signals to all our closest relatives so that they know | |
637 | * to properly mourn us.. | |
638 | */ | |
821c7de7 | 639 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 640 | { |
53c8f9f1 | 641 | bool autoreap; |
482a3767 ON |
642 | struct task_struct *p, *n; |
643 | LIST_HEAD(dead); | |
1da177e4 | 644 | |
762a24be | 645 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
646 | forget_original_parent(tsk, &dead); |
647 | ||
821c7de7 ON |
648 | if (group_dead) |
649 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 650 | |
b191d649 | 651 | tsk->exit_state = EXIT_ZOMBIE; |
45cdf5cc ON |
652 | if (unlikely(tsk->ptrace)) { |
653 | int sig = thread_group_leader(tsk) && | |
654 | thread_group_empty(tsk) && | |
655 | !ptrace_reparented(tsk) ? | |
656 | tsk->exit_signal : SIGCHLD; | |
657 | autoreap = do_notify_parent(tsk, sig); | |
658 | } else if (thread_group_leader(tsk)) { | |
659 | autoreap = thread_group_empty(tsk) && | |
660 | do_notify_parent(tsk, tsk->exit_signal); | |
661 | } else { | |
662 | autoreap = true; | |
663 | } | |
1da177e4 | 664 | |
30b692d3 CB |
665 | if (autoreap) { |
666 | tsk->exit_state = EXIT_DEAD; | |
6c66e7db | 667 | list_add(&tsk->ptrace_entry, &dead); |
30b692d3 | 668 | } |
1da177e4 | 669 | |
9c339168 ON |
670 | /* mt-exec, de_thread() is waiting for group leader */ |
671 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 672 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
673 | write_unlock_irq(&tasklist_lock); |
674 | ||
482a3767 ON |
675 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
676 | list_del_init(&p->ptrace_entry); | |
677 | release_task(p); | |
678 | } | |
1da177e4 LT |
679 | } |
680 | ||
e18eecb8 JD |
681 | #ifdef CONFIG_DEBUG_STACK_USAGE |
682 | static void check_stack_usage(void) | |
683 | { | |
684 | static DEFINE_SPINLOCK(low_water_lock); | |
685 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
686 | unsigned long free; |
687 | ||
7c9f8861 | 688 | free = stack_not_used(current); |
e18eecb8 JD |
689 | |
690 | if (free >= lowest_to_date) | |
691 | return; | |
692 | ||
693 | spin_lock(&low_water_lock); | |
694 | if (free < lowest_to_date) { | |
627393d4 | 695 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 696 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
697 | lowest_to_date = free; |
698 | } | |
699 | spin_unlock(&low_water_lock); | |
700 | } | |
701 | #else | |
702 | static inline void check_stack_usage(void) {} | |
703 | #endif | |
704 | ||
9af6528e | 705 | void __noreturn do_exit(long code) |
1da177e4 LT |
706 | { |
707 | struct task_struct *tsk = current; | |
708 | int group_dead; | |
709 | ||
710 | profile_task_exit(tsk); | |
5c9a8750 | 711 | kcov_task_exit(tsk); |
1da177e4 | 712 | |
73c10101 | 713 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 714 | |
1da177e4 LT |
715 | if (unlikely(in_interrupt())) |
716 | panic("Aiee, killing interrupt handler!"); | |
717 | if (unlikely(!tsk->pid)) | |
718 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 719 | |
33dd94ae NE |
720 | /* |
721 | * If do_exit is called because this processes oopsed, it's possible | |
722 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
723 | * continuing. Amongst other possible reasons, this is to prevent | |
724 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
725 | * kernel address. | |
726 | */ | |
727 | set_fs(USER_DS); | |
728 | ||
a288eecc | 729 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 730 | |
e0e81739 DH |
731 | validate_creds_for_do_exit(tsk); |
732 | ||
df164db5 AN |
733 | /* |
734 | * We're taking recursive faults here in do_exit. Safest is to just | |
735 | * leave this task alone and wait for reboot. | |
736 | */ | |
737 | if (unlikely(tsk->flags & PF_EXITING)) { | |
a0be55de | 738 | pr_alert("Fixing recursive fault but reboot is needed!\n"); |
18f69438 | 739 | futex_exit_recursive(tsk); |
df164db5 AN |
740 | set_current_state(TASK_UNINTERRUPTIBLE); |
741 | schedule(); | |
742 | } | |
743 | ||
d12619b5 | 744 | exit_signals(tsk); /* sets PF_EXITING */ |
1da177e4 | 745 | |
1dc0fffc | 746 | if (unlikely(in_atomic())) { |
a0be55de IA |
747 | pr_info("note: %s[%d] exited with preempt_count %d\n", |
748 | current->comm, task_pid_nr(current), | |
749 | preempt_count()); | |
1dc0fffc PZ |
750 | preempt_count_set(PREEMPT_ENABLED); |
751 | } | |
1da177e4 | 752 | |
48d212a2 LT |
753 | /* sync mm's RSS info before statistics gathering */ |
754 | if (tsk->mm) | |
755 | sync_mm_rss(tsk->mm); | |
51229b49 | 756 | acct_update_integrals(tsk); |
1da177e4 | 757 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 758 | if (group_dead) { |
43cf75d9 | 759 | /* |
760 | * If the last thread of global init has exited, panic | |
761 | * immediately to get a useable coredump. | |
762 | */ | |
763 | if (unlikely(is_global_init(tsk))) | |
764 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
765 | tsk->signal->group_exit_code ?: (int)code); | |
766 | ||
baa73d9e | 767 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 768 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 769 | exit_itimers(tsk->signal); |
baa73d9e | 770 | #endif |
1f10206c JP |
771 | if (tsk->mm) |
772 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 773 | } |
f6ec29a4 | 774 | acct_collect(code, group_dead); |
522ed776 MT |
775 | if (group_dead) |
776 | tty_audit_exit(); | |
a4ff8dba | 777 | audit_free(tsk); |
115085ea | 778 | |
48d212a2 | 779 | tsk->exit_code = code; |
115085ea | 780 | taskstats_exit(tsk, group_dead); |
c757249a | 781 | |
0039962a | 782 | exit_mm(); |
1da177e4 | 783 | |
0e464814 | 784 | if (group_dead) |
f6ec29a4 | 785 | acct_process(); |
0a16b607 MD |
786 | trace_sched_process_exit(tsk); |
787 | ||
1da177e4 | 788 | exit_sem(tsk); |
b34a6b1d | 789 | exit_shm(tsk); |
1ec7f1dd AV |
790 | exit_files(tsk); |
791 | exit_fs(tsk); | |
c39df5fa ON |
792 | if (group_dead) |
793 | disassociate_ctty(1); | |
8aac6270 | 794 | exit_task_namespaces(tsk); |
ed3e694d | 795 | exit_task_work(tsk); |
e6464694 | 796 | exit_thread(tsk); |
73ab1cb2 | 797 | exit_umh(tsk); |
0b3fcf17 SE |
798 | |
799 | /* | |
800 | * Flush inherited counters to the parent - before the parent | |
801 | * gets woken up by child-exit notifications. | |
802 | * | |
803 | * because of cgroup mode, must be called before cgroup_exit() | |
804 | */ | |
805 | perf_event_exit_task(tsk); | |
806 | ||
8e5bfa8c | 807 | sched_autogroup_exit_task(tsk); |
1ec41830 | 808 | cgroup_exit(tsk); |
1da177e4 | 809 | |
24f1e32c FW |
810 | /* |
811 | * FIXME: do that only when needed, using sched_exit tracepoint | |
812 | */ | |
7c8df286 | 813 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 814 | |
ccdd29ff | 815 | exit_tasks_rcu_start(); |
821c7de7 | 816 | exit_notify(tsk, group_dead); |
ef982393 | 817 | proc_exit_connector(tsk); |
c11600e4 | 818 | mpol_put_task_policy(tsk); |
42b2dd0a | 819 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
820 | if (unlikely(current->pi_state_cache)) |
821 | kfree(current->pi_state_cache); | |
42b2dd0a | 822 | #endif |
de5097c2 | 823 | /* |
9a11b49a | 824 | * Make sure we are holding no locks: |
de5097c2 | 825 | */ |
1b1d2fb4 | 826 | debug_check_no_locks_held(); |
1da177e4 | 827 | |
afc847b7 | 828 | if (tsk->io_context) |
b69f2292 | 829 | exit_io_context(tsk); |
afc847b7 | 830 | |
b92ce558 | 831 | if (tsk->splice_pipe) |
4b8a8f1e | 832 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 833 | |
5640f768 ED |
834 | if (tsk->task_frag.page) |
835 | put_page(tsk->task_frag.page); | |
836 | ||
e0e81739 DH |
837 | validate_creds_for_do_exit(tsk); |
838 | ||
4bcb8232 | 839 | check_stack_usage(); |
7407251a | 840 | preempt_disable(); |
54848d73 WF |
841 | if (tsk->nr_dirtied) |
842 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 843 | exit_rcu(); |
ccdd29ff | 844 | exit_tasks_rcu_finish(); |
b5740f4b | 845 | |
b09be676 | 846 | lockdep_free_task(tsk); |
9af6528e | 847 | do_task_dead(); |
1da177e4 | 848 | } |
012914da RA |
849 | EXPORT_SYMBOL_GPL(do_exit); |
850 | ||
9402c95f | 851 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
852 | { |
853 | if (comp) | |
854 | complete(comp); | |
55a101f8 | 855 | |
1da177e4 LT |
856 | do_exit(code); |
857 | } | |
1da177e4 LT |
858 | EXPORT_SYMBOL(complete_and_exit); |
859 | ||
754fe8d2 | 860 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
861 | { |
862 | do_exit((error_code&0xff)<<8); | |
863 | } | |
864 | ||
1da177e4 LT |
865 | /* |
866 | * Take down every thread in the group. This is called by fatal signals | |
867 | * as well as by sys_exit_group (below). | |
868 | */ | |
9402c95f | 869 | void |
1da177e4 LT |
870 | do_group_exit(int exit_code) |
871 | { | |
bfc4b089 ON |
872 | struct signal_struct *sig = current->signal; |
873 | ||
1da177e4 LT |
874 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
875 | ||
bfc4b089 ON |
876 | if (signal_group_exit(sig)) |
877 | exit_code = sig->group_exit_code; | |
1da177e4 | 878 | else if (!thread_group_empty(current)) { |
1da177e4 | 879 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 880 | |
1da177e4 | 881 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 882 | if (signal_group_exit(sig)) |
1da177e4 LT |
883 | /* Another thread got here before we took the lock. */ |
884 | exit_code = sig->group_exit_code; | |
885 | else { | |
1da177e4 | 886 | sig->group_exit_code = exit_code; |
ed5d2cac | 887 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
888 | zap_other_threads(current); |
889 | } | |
890 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
891 | } |
892 | ||
893 | do_exit(exit_code); | |
894 | /* NOTREACHED */ | |
895 | } | |
896 | ||
897 | /* | |
898 | * this kills every thread in the thread group. Note that any externally | |
899 | * wait4()-ing process will get the correct exit code - even if this | |
900 | * thread is not the thread group leader. | |
901 | */ | |
754fe8d2 | 902 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
903 | { |
904 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
905 | /* NOTREACHED */ |
906 | return 0; | |
1da177e4 LT |
907 | } |
908 | ||
67d7ddde AV |
909 | struct waitid_info { |
910 | pid_t pid; | |
911 | uid_t uid; | |
912 | int status; | |
913 | int cause; | |
914 | }; | |
915 | ||
9e8ae01d ON |
916 | struct wait_opts { |
917 | enum pid_type wo_type; | |
9e8ae01d | 918 | int wo_flags; |
e1eb1ebc | 919 | struct pid *wo_pid; |
9e8ae01d | 920 | |
67d7ddde | 921 | struct waitid_info *wo_info; |
359566fa | 922 | int wo_stat; |
ce72a16f | 923 | struct rusage *wo_rusage; |
9e8ae01d | 924 | |
ac6424b9 | 925 | wait_queue_entry_t child_wait; |
9e8ae01d ON |
926 | int notask_error; |
927 | }; | |
928 | ||
989264f4 | 929 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 930 | { |
5c01ba49 ON |
931 | return wo->wo_type == PIDTYPE_MAX || |
932 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
933 | } | |
1da177e4 | 934 | |
bf959931 ON |
935 | static int |
936 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
937 | { |
938 | if (!eligible_pid(wo, p)) | |
939 | return 0; | |
bf959931 ON |
940 | |
941 | /* | |
942 | * Wait for all children (clone and not) if __WALL is set or | |
943 | * if it is traced by us. | |
944 | */ | |
945 | if (ptrace || (wo->wo_flags & __WALL)) | |
946 | return 1; | |
947 | ||
948 | /* | |
949 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
950 | * otherwise, wait for non-clone children *only*. | |
951 | * | |
952 | * Note: a "clone" child here is one that reports to its parent | |
953 | * using a signal other than SIGCHLD, or a non-leader thread which | |
954 | * we can only see if it is traced by us. | |
955 | */ | |
956 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 957 | return 0; |
1da177e4 | 958 | |
14dd0b81 | 959 | return 1; |
1da177e4 LT |
960 | } |
961 | ||
1da177e4 LT |
962 | /* |
963 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
964 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
965 | * the lock and this task is uninteresting. If we return nonzero, we have | |
966 | * released the lock and the system call should return. | |
967 | */ | |
9e8ae01d | 968 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 969 | { |
67d7ddde | 970 | int state, status; |
6c5f3e7b | 971 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 972 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
67d7ddde | 973 | struct waitid_info *infop; |
1da177e4 | 974 | |
9e8ae01d | 975 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
976 | return 0; |
977 | ||
9e8ae01d | 978 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
76d9871e | 979 | status = p->exit_code; |
1da177e4 LT |
980 | get_task_struct(p); |
981 | read_unlock(&tasklist_lock); | |
1029a2b5 | 982 | sched_annotate_sleep(); |
e61a2502 AV |
983 | if (wo->wo_rusage) |
984 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 985 | put_task_struct(p); |
76d9871e | 986 | goto out_info; |
1da177e4 | 987 | } |
1da177e4 | 988 | /* |
abd50b39 | 989 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 990 | */ |
f6507f83 ON |
991 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
992 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 993 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 994 | return 0; |
986094df ON |
995 | /* |
996 | * We own this thread, nobody else can reap it. | |
997 | */ | |
998 | read_unlock(&tasklist_lock); | |
999 | sched_annotate_sleep(); | |
f6507f83 | 1000 | |
befca967 | 1001 | /* |
f6507f83 | 1002 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1003 | */ |
f6507f83 | 1004 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1005 | struct signal_struct *sig = p->signal; |
1006 | struct signal_struct *psig = current->signal; | |
1f10206c | 1007 | unsigned long maxrss; |
5613fda9 | 1008 | u64 tgutime, tgstime; |
3795e161 | 1009 | |
1da177e4 LT |
1010 | /* |
1011 | * The resource counters for the group leader are in its | |
1012 | * own task_struct. Those for dead threads in the group | |
1013 | * are in its signal_struct, as are those for the child | |
1014 | * processes it has previously reaped. All these | |
1015 | * accumulate in the parent's signal_struct c* fields. | |
1016 | * | |
1017 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1018 | * p->signal fields because the whole thread group is dead |
1019 | * and nobody can change them. | |
1020 | * | |
1021 | * psig->stats_lock also protects us from our sub-theads | |
1022 | * which can reap other children at the same time. Until | |
1023 | * we change k_getrusage()-like users to rely on this lock | |
1024 | * we have to take ->siglock as well. | |
0cf55e1e | 1025 | * |
a0be55de IA |
1026 | * We use thread_group_cputime_adjusted() to get times for |
1027 | * the thread group, which consolidates times for all threads | |
1028 | * in the group including the group leader. | |
1da177e4 | 1029 | */ |
e80d0a1a | 1030 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1031 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1032 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1033 | psig->cutime += tgutime + sig->cutime; |
1034 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1035 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1036 | psig->cmin_flt += |
1037 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1038 | psig->cmaj_flt += | |
1039 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1040 | psig->cnvcsw += | |
1041 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1042 | psig->cnivcsw += | |
1043 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1044 | psig->cinblock += |
1045 | task_io_get_inblock(p) + | |
1046 | sig->inblock + sig->cinblock; | |
1047 | psig->coublock += | |
1048 | task_io_get_oublock(p) + | |
1049 | sig->oublock + sig->coublock; | |
1f10206c JP |
1050 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1051 | if (psig->cmaxrss < maxrss) | |
1052 | psig->cmaxrss = maxrss; | |
5995477a AR |
1053 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1054 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1055 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1056 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1057 | } |
1058 | ||
ce72a16f AV |
1059 | if (wo->wo_rusage) |
1060 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
1da177e4 LT |
1061 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1062 | ? p->signal->group_exit_code : p->exit_code; | |
359566fa | 1063 | wo->wo_stat = status; |
2f4e6e2a | 1064 | |
b4360690 | 1065 | if (state == EXIT_TRACE) { |
1da177e4 | 1066 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1067 | /* We dropped tasklist, ptracer could die and untrace */ |
1068 | ptrace_unlink(p); | |
b4360690 ON |
1069 | |
1070 | /* If parent wants a zombie, don't release it now */ | |
1071 | state = EXIT_ZOMBIE; | |
1072 | if (do_notify_parent(p, p->exit_signal)) | |
1073 | state = EXIT_DEAD; | |
abd50b39 | 1074 | p->exit_state = state; |
1da177e4 LT |
1075 | write_unlock_irq(&tasklist_lock); |
1076 | } | |
abd50b39 | 1077 | if (state == EXIT_DEAD) |
1da177e4 | 1078 | release_task(p); |
2f4e6e2a | 1079 | |
76d9871e AV |
1080 | out_info: |
1081 | infop = wo->wo_info; | |
1082 | if (infop) { | |
1083 | if ((status & 0x7f) == 0) { | |
1084 | infop->cause = CLD_EXITED; | |
1085 | infop->status = status >> 8; | |
1086 | } else { | |
1087 | infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1088 | infop->status = status & 0x7f; | |
1089 | } | |
1090 | infop->pid = pid; | |
1091 | infop->uid = uid; | |
1092 | } | |
1093 | ||
67d7ddde | 1094 | return pid; |
1da177e4 LT |
1095 | } |
1096 | ||
90bc8d8b ON |
1097 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1098 | { | |
1099 | if (ptrace) { | |
570ac933 | 1100 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1101 | return &p->exit_code; |
1102 | } else { | |
1103 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1104 | return &p->signal->group_exit_code; | |
1105 | } | |
1106 | return NULL; | |
1107 | } | |
1108 | ||
19e27463 TH |
1109 | /** |
1110 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1111 | * @wo: wait options | |
1112 | * @ptrace: is the wait for ptrace | |
1113 | * @p: task to wait for | |
1114 | * | |
1115 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1116 | * | |
1117 | * CONTEXT: | |
1118 | * read_lock(&tasklist_lock), which is released if return value is | |
1119 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1120 | * | |
1121 | * RETURNS: | |
1122 | * 0 if wait condition didn't exist and search for other wait conditions | |
1123 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1124 | * success, implies that tasklist_lock is released and wait condition | |
1125 | * search should terminate. | |
1da177e4 | 1126 | */ |
9e8ae01d ON |
1127 | static int wait_task_stopped(struct wait_opts *wo, |
1128 | int ptrace, struct task_struct *p) | |
1da177e4 | 1129 | { |
67d7ddde AV |
1130 | struct waitid_info *infop; |
1131 | int exit_code, *p_code, why; | |
ee7c82da | 1132 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1133 | pid_t pid; |
1da177e4 | 1134 | |
47918025 ON |
1135 | /* |
1136 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1137 | */ | |
9e8ae01d | 1138 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1139 | return 0; |
1140 | ||
19e27463 TH |
1141 | if (!task_stopped_code(p, ptrace)) |
1142 | return 0; | |
1143 | ||
ee7c82da ON |
1144 | exit_code = 0; |
1145 | spin_lock_irq(&p->sighand->siglock); | |
1146 | ||
90bc8d8b ON |
1147 | p_code = task_stopped_code(p, ptrace); |
1148 | if (unlikely(!p_code)) | |
ee7c82da ON |
1149 | goto unlock_sig; |
1150 | ||
90bc8d8b | 1151 | exit_code = *p_code; |
ee7c82da ON |
1152 | if (!exit_code) |
1153 | goto unlock_sig; | |
1154 | ||
9e8ae01d | 1155 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1156 | *p_code = 0; |
ee7c82da | 1157 | |
8ca937a6 | 1158 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1159 | unlock_sig: |
1160 | spin_unlock_irq(&p->sighand->siglock); | |
1161 | if (!exit_code) | |
1da177e4 LT |
1162 | return 0; |
1163 | ||
1164 | /* | |
1165 | * Now we are pretty sure this task is interesting. | |
1166 | * Make sure it doesn't get reaped out from under us while we | |
1167 | * give up the lock and then examine it below. We don't want to | |
1168 | * keep holding onto the tasklist_lock while we call getrusage and | |
1169 | * possibly take page faults for user memory. | |
1170 | */ | |
1171 | get_task_struct(p); | |
6c5f3e7b | 1172 | pid = task_pid_vnr(p); |
f470021a | 1173 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1174 | read_unlock(&tasklist_lock); |
1029a2b5 | 1175 | sched_annotate_sleep(); |
e61a2502 AV |
1176 | if (wo->wo_rusage) |
1177 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1178 | put_task_struct(p); |
1da177e4 | 1179 | |
bb380ec3 AV |
1180 | if (likely(!(wo->wo_flags & WNOWAIT))) |
1181 | wo->wo_stat = (exit_code << 8) | 0x7f; | |
1da177e4 | 1182 | |
9e8ae01d | 1183 | infop = wo->wo_info; |
67d7ddde AV |
1184 | if (infop) { |
1185 | infop->cause = why; | |
1186 | infop->status = exit_code; | |
1187 | infop->pid = pid; | |
1188 | infop->uid = uid; | |
1189 | } | |
67d7ddde | 1190 | return pid; |
1da177e4 LT |
1191 | } |
1192 | ||
1193 | /* | |
1194 | * Handle do_wait work for one task in a live, non-stopped state. | |
1195 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1196 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1197 | * released the lock and the system call should return. | |
1198 | */ | |
9e8ae01d | 1199 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1200 | { |
bb380ec3 | 1201 | struct waitid_info *infop; |
1da177e4 LT |
1202 | pid_t pid; |
1203 | uid_t uid; | |
1204 | ||
9e8ae01d | 1205 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1206 | return 0; |
1207 | ||
1da177e4 LT |
1208 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1209 | return 0; | |
1210 | ||
1211 | spin_lock_irq(&p->sighand->siglock); | |
1212 | /* Re-check with the lock held. */ | |
1213 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1214 | spin_unlock_irq(&p->sighand->siglock); | |
1215 | return 0; | |
1216 | } | |
9e8ae01d | 1217 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1218 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1219 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1220 | spin_unlock_irq(&p->sighand->siglock); |
1221 | ||
6c5f3e7b | 1222 | pid = task_pid_vnr(p); |
1da177e4 LT |
1223 | get_task_struct(p); |
1224 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1225 | sched_annotate_sleep(); |
e61a2502 AV |
1226 | if (wo->wo_rusage) |
1227 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1228 | put_task_struct(p); |
1da177e4 | 1229 | |
bb380ec3 AV |
1230 | infop = wo->wo_info; |
1231 | if (!infop) { | |
359566fa | 1232 | wo->wo_stat = 0xffff; |
1da177e4 | 1233 | } else { |
bb380ec3 AV |
1234 | infop->cause = CLD_CONTINUED; |
1235 | infop->pid = pid; | |
1236 | infop->uid = uid; | |
1237 | infop->status = SIGCONT; | |
1da177e4 | 1238 | } |
bb380ec3 | 1239 | return pid; |
1da177e4 LT |
1240 | } |
1241 | ||
98abed02 RM |
1242 | /* |
1243 | * Consider @p for a wait by @parent. | |
1244 | * | |
9e8ae01d | 1245 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1246 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1247 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1248 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1249 | * or still -ECHILD. |
98abed02 | 1250 | */ |
b6e763f0 ON |
1251 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1252 | struct task_struct *p) | |
98abed02 | 1253 | { |
3245d6ac ON |
1254 | /* |
1255 | * We can race with wait_task_zombie() from another thread. | |
1256 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1257 | * can't confuse the checks below. | |
1258 | */ | |
6aa7de05 | 1259 | int exit_state = READ_ONCE(p->exit_state); |
b3ab0316 ON |
1260 | int ret; |
1261 | ||
3245d6ac | 1262 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1263 | return 0; |
1264 | ||
bf959931 | 1265 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1266 | if (!ret) |
98abed02 RM |
1267 | return ret; |
1268 | ||
3245d6ac | 1269 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1270 | /* |
abd50b39 ON |
1271 | * ptrace == 0 means we are the natural parent. In this case |
1272 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1273 | */ |
abd50b39 | 1274 | if (likely(!ptrace)) |
50b8d257 | 1275 | wo->notask_error = 0; |
823b018e | 1276 | return 0; |
50b8d257 | 1277 | } |
823b018e | 1278 | |
377d75da ON |
1279 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1280 | /* | |
1281 | * If it is traced by its real parent's group, just pretend | |
1282 | * the caller is ptrace_do_wait() and reap this child if it | |
1283 | * is zombie. | |
1284 | * | |
1285 | * This also hides group stop state from real parent; otherwise | |
1286 | * a single stop can be reported twice as group and ptrace stop. | |
1287 | * If a ptracer wants to distinguish these two events for its | |
1288 | * own children it should create a separate process which takes | |
1289 | * the role of real parent. | |
1290 | */ | |
1291 | if (!ptrace_reparented(p)) | |
1292 | ptrace = 1; | |
1293 | } | |
1294 | ||
45cb24a1 | 1295 | /* slay zombie? */ |
3245d6ac | 1296 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1297 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1298 | if (!delay_group_leader(p)) { |
1299 | /* | |
1300 | * A zombie ptracee is only visible to its ptracer. | |
1301 | * Notification and reaping will be cascaded to the | |
1302 | * real parent when the ptracer detaches. | |
1303 | */ | |
1304 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1305 | return wait_task_zombie(wo, p); | |
1306 | } | |
98abed02 | 1307 | |
f470021a | 1308 | /* |
9b84cca2 TH |
1309 | * Allow access to stopped/continued state via zombie by |
1310 | * falling through. Clearing of notask_error is complex. | |
1311 | * | |
1312 | * When !@ptrace: | |
1313 | * | |
1314 | * If WEXITED is set, notask_error should naturally be | |
1315 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1316 | * so, if there are live subthreads, there are events to | |
1317 | * wait for. If all subthreads are dead, it's still safe | |
1318 | * to clear - this function will be called again in finite | |
1319 | * amount time once all the subthreads are released and | |
1320 | * will then return without clearing. | |
1321 | * | |
1322 | * When @ptrace: | |
1323 | * | |
1324 | * Stopped state is per-task and thus can't change once the | |
1325 | * target task dies. Only continued and exited can happen. | |
1326 | * Clear notask_error if WCONTINUED | WEXITED. | |
1327 | */ | |
1328 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1329 | wo->notask_error = 0; | |
1330 | } else { | |
1331 | /* | |
1332 | * @p is alive and it's gonna stop, continue or exit, so | |
1333 | * there always is something to wait for. | |
f470021a | 1334 | */ |
9e8ae01d | 1335 | wo->notask_error = 0; |
f470021a RM |
1336 | } |
1337 | ||
98abed02 | 1338 | /* |
45cb24a1 TH |
1339 | * Wait for stopped. Depending on @ptrace, different stopped state |
1340 | * is used and the two don't interact with each other. | |
98abed02 | 1341 | */ |
19e27463 TH |
1342 | ret = wait_task_stopped(wo, ptrace, p); |
1343 | if (ret) | |
1344 | return ret; | |
98abed02 RM |
1345 | |
1346 | /* | |
45cb24a1 TH |
1347 | * Wait for continued. There's only one continued state and the |
1348 | * ptracer can consume it which can confuse the real parent. Don't | |
1349 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1350 | */ |
9e8ae01d | 1351 | return wait_task_continued(wo, p); |
98abed02 RM |
1352 | } |
1353 | ||
1354 | /* | |
1355 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1356 | * | |
9e8ae01d | 1357 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1358 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1359 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1360 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1361 | * or still -ECHILD. |
98abed02 | 1362 | */ |
9e8ae01d | 1363 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1364 | { |
1365 | struct task_struct *p; | |
1366 | ||
1367 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1368 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1369 | |
9cd80bbb ON |
1370 | if (ret) |
1371 | return ret; | |
98abed02 RM |
1372 | } |
1373 | ||
1374 | return 0; | |
1375 | } | |
1376 | ||
9e8ae01d | 1377 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1378 | { |
1379 | struct task_struct *p; | |
1380 | ||
f470021a | 1381 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1382 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1383 | |
f470021a | 1384 | if (ret) |
98abed02 | 1385 | return ret; |
98abed02 RM |
1386 | } |
1387 | ||
1388 | return 0; | |
1389 | } | |
1390 | ||
ac6424b9 | 1391 | static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode, |
0b7570e7 ON |
1392 | int sync, void *key) |
1393 | { | |
1394 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1395 | child_wait); | |
1396 | struct task_struct *p = key; | |
1397 | ||
5c01ba49 | 1398 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1399 | return 0; |
1400 | ||
b4fe5182 ON |
1401 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1402 | return 0; | |
1403 | ||
0b7570e7 ON |
1404 | return default_wake_function(wait, mode, sync, key); |
1405 | } | |
1406 | ||
a7f0765e ON |
1407 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1408 | { | |
0b7570e7 | 1409 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
ce4dd442 | 1410 | TASK_INTERRUPTIBLE, p); |
a7f0765e ON |
1411 | } |
1412 | ||
9e8ae01d | 1413 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1414 | { |
1da177e4 | 1415 | struct task_struct *tsk; |
98abed02 | 1416 | int retval; |
1da177e4 | 1417 | |
9e8ae01d | 1418 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1419 | |
0b7570e7 ON |
1420 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1421 | wo->child_wait.private = current; | |
1422 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1423 | repeat: |
98abed02 | 1424 | /* |
3da56d16 | 1425 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1426 | * We will clear ->notask_error to zero if we see any child that |
1427 | * might later match our criteria, even if we are not able to reap | |
1428 | * it yet. | |
98abed02 | 1429 | */ |
64a16caf | 1430 | wo->notask_error = -ECHILD; |
9e8ae01d | 1431 | if ((wo->wo_type < PIDTYPE_MAX) && |
1722c14a | 1432 | (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type))) |
64a16caf | 1433 | goto notask; |
161550d7 | 1434 | |
f95d39d1 | 1435 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1436 | read_lock(&tasklist_lock); |
1437 | tsk = current; | |
1438 | do { | |
64a16caf ON |
1439 | retval = do_wait_thread(wo, tsk); |
1440 | if (retval) | |
1441 | goto end; | |
9e8ae01d | 1442 | |
64a16caf ON |
1443 | retval = ptrace_do_wait(wo, tsk); |
1444 | if (retval) | |
98abed02 | 1445 | goto end; |
98abed02 | 1446 | |
9e8ae01d | 1447 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1448 | break; |
a3f6dfb7 | 1449 | } while_each_thread(current, tsk); |
1da177e4 | 1450 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1451 | |
64a16caf | 1452 | notask: |
9e8ae01d ON |
1453 | retval = wo->notask_error; |
1454 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1455 | retval = -ERESTARTSYS; |
98abed02 RM |
1456 | if (!signal_pending(current)) { |
1457 | schedule(); | |
1458 | goto repeat; | |
1459 | } | |
1da177e4 | 1460 | } |
1da177e4 | 1461 | end: |
f95d39d1 | 1462 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1463 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1464 | return retval; |
1465 | } | |
1466 | ||
3695eae5 CB |
1467 | static struct pid *pidfd_get_pid(unsigned int fd) |
1468 | { | |
1469 | struct fd f; | |
1470 | struct pid *pid; | |
1471 | ||
1472 | f = fdget(fd); | |
1473 | if (!f.file) | |
1474 | return ERR_PTR(-EBADF); | |
1475 | ||
1476 | pid = pidfd_pid(f.file); | |
1477 | if (!IS_ERR(pid)) | |
1478 | get_pid(pid); | |
1479 | ||
1480 | fdput(f); | |
1481 | return pid; | |
1482 | } | |
1483 | ||
67d7ddde | 1484 | static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, |
ce72a16f | 1485 | int options, struct rusage *ru) |
1da177e4 | 1486 | { |
9e8ae01d | 1487 | struct wait_opts wo; |
161550d7 EB |
1488 | struct pid *pid = NULL; |
1489 | enum pid_type type; | |
1da177e4 LT |
1490 | long ret; |
1491 | ||
91c4e8ea ON |
1492 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1493 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1494 | return -EINVAL; |
1495 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1496 | return -EINVAL; | |
1497 | ||
1498 | switch (which) { | |
1499 | case P_ALL: | |
161550d7 | 1500 | type = PIDTYPE_MAX; |
1da177e4 LT |
1501 | break; |
1502 | case P_PID: | |
161550d7 EB |
1503 | type = PIDTYPE_PID; |
1504 | if (upid <= 0) | |
1da177e4 | 1505 | return -EINVAL; |
3695eae5 CB |
1506 | |
1507 | pid = find_get_pid(upid); | |
1da177e4 LT |
1508 | break; |
1509 | case P_PGID: | |
161550d7 | 1510 | type = PIDTYPE_PGID; |
821cc7b0 | 1511 | if (upid < 0) |
1da177e4 | 1512 | return -EINVAL; |
3695eae5 | 1513 | |
821cc7b0 EB |
1514 | if (upid) |
1515 | pid = find_get_pid(upid); | |
1516 | else | |
1517 | pid = get_task_pid(current, PIDTYPE_PGID); | |
3695eae5 CB |
1518 | break; |
1519 | case P_PIDFD: | |
1520 | type = PIDTYPE_PID; | |
1521 | if (upid < 0) | |
1da177e4 | 1522 | return -EINVAL; |
3695eae5 CB |
1523 | |
1524 | pid = pidfd_get_pid(upid); | |
1525 | if (IS_ERR(pid)) | |
1526 | return PTR_ERR(pid); | |
1da177e4 LT |
1527 | break; |
1528 | default: | |
1529 | return -EINVAL; | |
1530 | } | |
1531 | ||
9e8ae01d ON |
1532 | wo.wo_type = type; |
1533 | wo.wo_pid = pid; | |
1534 | wo.wo_flags = options; | |
1535 | wo.wo_info = infop; | |
9e8ae01d ON |
1536 | wo.wo_rusage = ru; |
1537 | ret = do_wait(&wo); | |
dfe16dfa | 1538 | |
161550d7 | 1539 | put_pid(pid); |
1da177e4 LT |
1540 | return ret; |
1541 | } | |
1542 | ||
ce72a16f AV |
1543 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1544 | infop, int, options, struct rusage __user *, ru) | |
1545 | { | |
1546 | struct rusage r; | |
67d7ddde AV |
1547 | struct waitid_info info = {.status = 0}; |
1548 | long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL); | |
634a8160 | 1549 | int signo = 0; |
6c85501f | 1550 | |
634a8160 AV |
1551 | if (err > 0) { |
1552 | signo = SIGCHLD; | |
1553 | err = 0; | |
ce72a16f AV |
1554 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) |
1555 | return -EFAULT; | |
1556 | } | |
67d7ddde AV |
1557 | if (!infop) |
1558 | return err; | |
1559 | ||
594cc251 | 1560 | if (!user_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1561 | return -EFAULT; |
96ca579a | 1562 | |
634a8160 | 1563 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1564 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1565 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1566 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1567 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1568 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
1569 | user_access_end(); | |
ce72a16f | 1570 | return err; |
4c48abe9 AV |
1571 | Efault: |
1572 | user_access_end(); | |
1573 | return -EFAULT; | |
ce72a16f AV |
1574 | } |
1575 | ||
92ebce5a AV |
1576 | long kernel_wait4(pid_t upid, int __user *stat_addr, int options, |
1577 | struct rusage *ru) | |
1da177e4 | 1578 | { |
9e8ae01d | 1579 | struct wait_opts wo; |
161550d7 EB |
1580 | struct pid *pid = NULL; |
1581 | enum pid_type type; | |
1da177e4 LT |
1582 | long ret; |
1583 | ||
1584 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1585 | __WNOTHREAD|__WCLONE|__WALL)) | |
1586 | return -EINVAL; | |
161550d7 | 1587 | |
dd83c161 | 1588 | /* -INT_MIN is not defined */ |
1589 | if (upid == INT_MIN) | |
1590 | return -ESRCH; | |
1591 | ||
161550d7 EB |
1592 | if (upid == -1) |
1593 | type = PIDTYPE_MAX; | |
1594 | else if (upid < 0) { | |
1595 | type = PIDTYPE_PGID; | |
1596 | pid = find_get_pid(-upid); | |
1597 | } else if (upid == 0) { | |
1598 | type = PIDTYPE_PGID; | |
2ae448ef | 1599 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1600 | } else /* upid > 0 */ { |
1601 | type = PIDTYPE_PID; | |
1602 | pid = find_get_pid(upid); | |
1603 | } | |
1604 | ||
9e8ae01d ON |
1605 | wo.wo_type = type; |
1606 | wo.wo_pid = pid; | |
1607 | wo.wo_flags = options | WEXITED; | |
1608 | wo.wo_info = NULL; | |
359566fa | 1609 | wo.wo_stat = 0; |
9e8ae01d ON |
1610 | wo.wo_rusage = ru; |
1611 | ret = do_wait(&wo); | |
161550d7 | 1612 | put_pid(pid); |
359566fa AV |
1613 | if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr)) |
1614 | ret = -EFAULT; | |
1da177e4 | 1615 | |
1da177e4 LT |
1616 | return ret; |
1617 | } | |
1618 | ||
ce72a16f AV |
1619 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1620 | int, options, struct rusage __user *, ru) | |
1621 | { | |
1622 | struct rusage r; | |
1623 | long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL); | |
1624 | ||
1625 | if (err > 0) { | |
1626 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) | |
1627 | return -EFAULT; | |
1628 | } | |
1629 | return err; | |
1630 | } | |
1631 | ||
1da177e4 LT |
1632 | #ifdef __ARCH_WANT_SYS_WAITPID |
1633 | ||
1634 | /* | |
1635 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1636 | * implemented by calling sys_wait4() from libc.a. | |
1637 | */ | |
17da2bd9 | 1638 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 | 1639 | { |
d300b610 | 1640 | return kernel_wait4(pid, stat_addr, options, NULL); |
1da177e4 LT |
1641 | } |
1642 | ||
1643 | #endif | |
7e95a225 AV |
1644 | |
1645 | #ifdef CONFIG_COMPAT | |
1646 | COMPAT_SYSCALL_DEFINE4(wait4, | |
1647 | compat_pid_t, pid, | |
1648 | compat_uint_t __user *, stat_addr, | |
1649 | int, options, | |
1650 | struct compat_rusage __user *, ru) | |
1651 | { | |
ce72a16f AV |
1652 | struct rusage r; |
1653 | long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL); | |
1654 | if (err > 0) { | |
1655 | if (ru && put_compat_rusage(&r, ru)) | |
1656 | return -EFAULT; | |
7e95a225 | 1657 | } |
ce72a16f | 1658 | return err; |
7e95a225 AV |
1659 | } |
1660 | ||
1661 | COMPAT_SYSCALL_DEFINE5(waitid, | |
1662 | int, which, compat_pid_t, pid, | |
1663 | struct compat_siginfo __user *, infop, int, options, | |
1664 | struct compat_rusage __user *, uru) | |
1665 | { | |
7e95a225 | 1666 | struct rusage ru; |
67d7ddde AV |
1667 | struct waitid_info info = {.status = 0}; |
1668 | long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL); | |
634a8160 AV |
1669 | int signo = 0; |
1670 | if (err > 0) { | |
1671 | signo = SIGCHLD; | |
1672 | err = 0; | |
6c85501f AV |
1673 | if (uru) { |
1674 | /* kernel_waitid() overwrites everything in ru */ | |
1675 | if (COMPAT_USE_64BIT_TIME) | |
1676 | err = copy_to_user(uru, &ru, sizeof(ru)); | |
1677 | else | |
1678 | err = put_compat_rusage(&ru, uru); | |
1679 | if (err) | |
1680 | return -EFAULT; | |
1681 | } | |
7e95a225 AV |
1682 | } |
1683 | ||
4c48abe9 AV |
1684 | if (!infop) |
1685 | return err; | |
1686 | ||
594cc251 | 1687 | if (!user_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1688 | return -EFAULT; |
96ca579a | 1689 | |
634a8160 | 1690 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1691 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1692 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1693 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1694 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1695 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
1696 | user_access_end(); | |
67d7ddde | 1697 | return err; |
4c48abe9 AV |
1698 | Efault: |
1699 | user_access_end(); | |
1700 | return -EFAULT; | |
7e95a225 AV |
1701 | } |
1702 | #endif | |
7c2c11b2 SM |
1703 | |
1704 | __weak void abort(void) | |
1705 | { | |
1706 | BUG(); | |
1707 | ||
1708 | /* if that doesn't kill us, halt */ | |
1709 | panic("Oops failed to kill thread"); | |
1710 | } | |
dc8635b7 | 1711 | EXPORT_SYMBOL(abort); |