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> |
6eaeeaba | 51 | #include <linux/task_io_accounting_ops.h> |
355f841a EB |
52 | #include <linux/blkdev.h> |
53 | #include <linux/task_work.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> |
50b5e49c | 63 | #include <linux/kmsan.h> |
53d3eaa3 | 64 | #include <linux/random.h> |
8f95c90c | 65 | #include <linux/rcuwait.h> |
7e95a225 | 66 | #include <linux/compat.h> |
b1b6b5a3 | 67 | #include <linux/io_uring.h> |
670721c7 | 68 | #include <linux/kprobes.h> |
54ecbe6f | 69 | #include <linux/rethook.h> |
1da177e4 | 70 | |
7c0f6ba6 | 71 | #include <linux/uaccess.h> |
1da177e4 | 72 | #include <asm/unistd.h> |
1da177e4 LT |
73 | #include <asm/mmu_context.h> |
74 | ||
d40e48e0 | 75 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
76 | { |
77 | nr_threads--; | |
50d75f8d | 78 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 79 | if (group_dead) { |
6883f81a | 80 | detach_pid(p, PIDTYPE_TGID); |
1da177e4 LT |
81 | detach_pid(p, PIDTYPE_PGID); |
82 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 83 | |
5e85d4ab | 84 | list_del_rcu(&p->tasks); |
9cd80bbb | 85 | list_del_init(&p->sibling); |
909ea964 | 86 | __this_cpu_dec(process_counts); |
1da177e4 | 87 | } |
47e65328 | 88 | list_del_rcu(&p->thread_group); |
0c740d0a | 89 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
90 | } |
91 | ||
6a14c5c9 ON |
92 | /* |
93 | * This function expects the tasklist_lock write-locked. | |
94 | */ | |
95 | static void __exit_signal(struct task_struct *tsk) | |
96 | { | |
97 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 98 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 99 | struct sighand_struct *sighand; |
3f649ab7 | 100 | struct tty_struct *tty; |
5613fda9 | 101 | u64 utime, stime; |
6a14c5c9 | 102 | |
d11c563d | 103 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 104 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
105 | spin_lock(&sighand->siglock); |
106 | ||
baa73d9e | 107 | #ifdef CONFIG_POSIX_TIMERS |
6a14c5c9 | 108 | posix_cpu_timers_exit(tsk); |
b95e31c0 | 109 | if (group_dead) |
6a14c5c9 | 110 | posix_cpu_timers_exit_group(tsk); |
baa73d9e | 111 | #endif |
e0a70217 | 112 | |
baa73d9e NP |
113 | if (group_dead) { |
114 | tty = sig->tty; | |
115 | sig->tty = NULL; | |
116 | } else { | |
6a14c5c9 ON |
117 | /* |
118 | * If there is any task waiting for the group exit | |
119 | * then notify it: | |
120 | */ | |
d344193a | 121 | if (sig->notify_count > 0 && !--sig->notify_count) |
60700e38 | 122 | wake_up_process(sig->group_exec_task); |
6db840fa | 123 | |
6a14c5c9 ON |
124 | if (tsk == sig->curr_target) |
125 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
126 | } |
127 | ||
53d3eaa3 NP |
128 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
129 | sizeof(unsigned long long)); | |
130 | ||
90ed9cbe | 131 | /* |
26e75b5c ON |
132 | * Accumulate here the counters for all threads as they die. We could |
133 | * skip the group leader because it is the last user of signal_struct, | |
134 | * but we want to avoid the race with thread_group_cputime() which can | |
135 | * see the empty ->thread_head list. | |
90ed9cbe RR |
136 | */ |
137 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 138 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
139 | sig->utime += utime; |
140 | sig->stime += stime; | |
141 | sig->gtime += task_gtime(tsk); | |
142 | sig->min_flt += tsk->min_flt; | |
143 | sig->maj_flt += tsk->maj_flt; | |
144 | sig->nvcsw += tsk->nvcsw; | |
145 | sig->nivcsw += tsk->nivcsw; | |
146 | sig->inblock += task_io_get_inblock(tsk); | |
147 | sig->oublock += task_io_get_oublock(tsk); | |
148 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
149 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 150 | sig->nr_threads--; |
d40e48e0 | 151 | __unhash_process(tsk, group_dead); |
e78c3496 | 152 | write_sequnlock(&sig->stats_lock); |
5876700c | 153 | |
da7978b0 ON |
154 | /* |
155 | * Do this under ->siglock, we can race with another thread | |
156 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
157 | */ | |
158 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 159 | tsk->sighand = NULL; |
6a14c5c9 | 160 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 161 | |
a7e5328a | 162 | __cleanup_sighand(sighand); |
a0be55de | 163 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 164 | if (group_dead) { |
6a14c5c9 | 165 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 166 | tty_kref_put(tty); |
6a14c5c9 ON |
167 | } |
168 | } | |
169 | ||
8c7904a0 EB |
170 | static void delayed_put_task_struct(struct rcu_head *rhp) |
171 | { | |
0a16b607 MD |
172 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
173 | ||
670721c7 | 174 | kprobe_flush_task(tsk); |
54ecbe6f | 175 | rethook_flush_task(tsk); |
4e231c79 | 176 | perf_event_delayed_put(tsk); |
0a16b607 MD |
177 | trace_sched_process_free(tsk); |
178 | put_task_struct(tsk); | |
8c7904a0 EB |
179 | } |
180 | ||
3fbd7ee2 EB |
181 | void put_task_struct_rcu_user(struct task_struct *task) |
182 | { | |
183 | if (refcount_dec_and_test(&task->rcu_users)) | |
184 | call_rcu(&task->rcu, delayed_put_task_struct); | |
185 | } | |
f470021a | 186 | |
2be9880d KW |
187 | void __weak release_thread(struct task_struct *dead_task) |
188 | { | |
189 | } | |
190 | ||
a0be55de | 191 | void release_task(struct task_struct *p) |
1da177e4 | 192 | { |
36c8b586 | 193 | struct task_struct *leader; |
7bc3e6e5 | 194 | struct pid *thread_pid; |
1da177e4 | 195 | int zap_leader; |
1f09f974 | 196 | repeat: |
c69e8d9c | 197 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
198 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
199 | rcu_read_lock(); | |
21d1c5e3 | 200 | dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1); |
d11c563d | 201 | rcu_read_unlock(); |
c69e8d9c | 202 | |
6b115bf5 | 203 | cgroup_release(p); |
0203026b | 204 | |
1da177e4 | 205 | write_lock_irq(&tasklist_lock); |
a288eecc | 206 | ptrace_release_task(p); |
7bc3e6e5 | 207 | thread_pid = get_pid(p->thread_pid); |
1da177e4 | 208 | __exit_signal(p); |
35f5cad8 | 209 | |
1da177e4 LT |
210 | /* |
211 | * If we are the last non-leader member of the thread | |
212 | * group, and the leader is zombie, then notify the | |
213 | * group leader's parent process. (if it wants notification.) | |
214 | */ | |
215 | zap_leader = 0; | |
216 | leader = p->group_leader; | |
a0be55de IA |
217 | if (leader != p && thread_group_empty(leader) |
218 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
219 | /* |
220 | * If we were the last child thread and the leader has | |
221 | * exited already, and the leader's parent ignores SIGCHLD, | |
222 | * then we are the one who should release the leader. | |
dae33574 | 223 | */ |
86773473 | 224 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
225 | if (zap_leader) |
226 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
227 | } |
228 | ||
1da177e4 | 229 | write_unlock_irq(&tasklist_lock); |
3a15fb6e | 230 | seccomp_filter_release(p); |
7bc3e6e5 | 231 | proc_flush_pid(thread_pid); |
6ade99ec | 232 | put_pid(thread_pid); |
1da177e4 | 233 | release_thread(p); |
3fbd7ee2 | 234 | put_task_struct_rcu_user(p); |
1da177e4 LT |
235 | |
236 | p = leader; | |
237 | if (unlikely(zap_leader)) | |
238 | goto repeat; | |
239 | } | |
240 | ||
9d9a6ebf | 241 | int rcuwait_wake_up(struct rcuwait *w) |
8f95c90c | 242 | { |
9d9a6ebf | 243 | int ret = 0; |
8f95c90c DB |
244 | struct task_struct *task; |
245 | ||
246 | rcu_read_lock(); | |
247 | ||
248 | /* | |
249 | * Order condition vs @task, such that everything prior to the load | |
250 | * of @task is visible. This is the condition as to why the user called | |
c9d64a1b | 251 | * rcuwait_wake() in the first place. Pairs with set_current_state() |
8f95c90c DB |
252 | * barrier (A) in rcuwait_wait_event(). |
253 | * | |
254 | * WAIT WAKE | |
255 | * [S] tsk = current [S] cond = true | |
256 | * MB (A) MB (B) | |
257 | * [L] cond [L] tsk | |
258 | */ | |
6dc080ee | 259 | smp_mb(); /* (B) */ |
8f95c90c | 260 | |
8f95c90c DB |
261 | task = rcu_dereference(w->task); |
262 | if (task) | |
9d9a6ebf | 263 | ret = wake_up_process(task); |
8f95c90c | 264 | rcu_read_unlock(); |
9d9a6ebf DB |
265 | |
266 | return ret; | |
8f95c90c | 267 | } |
ac8dec42 | 268 | EXPORT_SYMBOL_GPL(rcuwait_wake_up); |
8f95c90c | 269 | |
1da177e4 LT |
270 | /* |
271 | * Determine if a process group is "orphaned", according to the POSIX | |
272 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
273 | * by terminal-generated stop signals. Newly orphaned process groups are | |
274 | * to receive a SIGHUP and a SIGCONT. | |
275 | * | |
276 | * "I ask you, have you ever known what it is to be an orphan?" | |
277 | */ | |
a0be55de IA |
278 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
279 | struct task_struct *ignored_task) | |
1da177e4 LT |
280 | { |
281 | struct task_struct *p; | |
1da177e4 | 282 | |
0475ac08 | 283 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
284 | if ((p == ignored_task) || |
285 | (p->exit_state && thread_group_empty(p)) || | |
286 | is_global_init(p->real_parent)) | |
1da177e4 | 287 | continue; |
05e83df6 | 288 | |
0475ac08 | 289 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
290 | task_session(p->real_parent) == task_session(p)) |
291 | return 0; | |
0475ac08 | 292 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
293 | |
294 | return 1; | |
1da177e4 LT |
295 | } |
296 | ||
3e7cd6c4 | 297 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
298 | { |
299 | int retval; | |
300 | ||
301 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 302 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
303 | read_unlock(&tasklist_lock); |
304 | ||
305 | return retval; | |
306 | } | |
307 | ||
961c4675 | 308 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 309 | { |
1da177e4 LT |
310 | struct task_struct *p; |
311 | ||
0475ac08 | 312 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
313 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
314 | return true; | |
0475ac08 | 315 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
316 | |
317 | return false; | |
1da177e4 LT |
318 | } |
319 | ||
f49ee505 ON |
320 | /* |
321 | * Check to see if any process groups have become orphaned as | |
322 | * a result of our exiting, and if they have any stopped jobs, | |
323 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
324 | */ | |
325 | static void | |
326 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
327 | { | |
328 | struct pid *pgrp = task_pgrp(tsk); | |
329 | struct task_struct *ignored_task = tsk; | |
330 | ||
331 | if (!parent) | |
a0be55de IA |
332 | /* exit: our father is in a different pgrp than |
333 | * we are and we were the only connection outside. | |
334 | */ | |
f49ee505 ON |
335 | parent = tsk->real_parent; |
336 | else | |
337 | /* reparent: our child is in a different pgrp than | |
338 | * we are, and it was the only connection outside. | |
339 | */ | |
340 | ignored_task = NULL; | |
341 | ||
342 | if (task_pgrp(parent) != pgrp && | |
343 | task_session(parent) == task_session(tsk) && | |
344 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
345 | has_stopped_jobs(pgrp)) { | |
346 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
347 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
348 | } | |
349 | } | |
350 | ||
92307383 | 351 | static void coredump_task_exit(struct task_struct *tsk) |
d67e03e3 EB |
352 | { |
353 | struct core_state *core_state; | |
354 | ||
355 | /* | |
356 | * Serialize with any possible pending coredump. | |
0258b5fd | 357 | * We must hold siglock around checking core_state |
92307383 | 358 | * and setting PF_POSTCOREDUMP. The core-inducing thread |
d67e03e3 | 359 | * will increment ->nr_threads for each thread in the |
92307383 | 360 | * group without PF_POSTCOREDUMP set. |
d67e03e3 | 361 | */ |
0258b5fd | 362 | spin_lock_irq(&tsk->sighand->siglock); |
92307383 | 363 | tsk->flags |= PF_POSTCOREDUMP; |
0258b5fd EB |
364 | core_state = tsk->signal->core_state; |
365 | spin_unlock_irq(&tsk->sighand->siglock); | |
d67e03e3 EB |
366 | if (core_state) { |
367 | struct core_thread self; | |
368 | ||
d67e03e3 EB |
369 | self.task = current; |
370 | if (self.task->flags & PF_SIGNALED) | |
371 | self.next = xchg(&core_state->dumper.next, &self); | |
372 | else | |
373 | self.task = NULL; | |
374 | /* | |
375 | * Implies mb(), the result of xchg() must be visible | |
376 | * to core_state->dumper. | |
377 | */ | |
378 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
379 | complete(&core_state->startup); | |
380 | ||
381 | for (;;) { | |
f5d39b02 | 382 | set_current_state(TASK_UNINTERRUPTIBLE|TASK_FREEZABLE); |
d67e03e3 EB |
383 | if (!self.task) /* see coredump_finish() */ |
384 | break; | |
f5d39b02 | 385 | schedule(); |
d67e03e3 EB |
386 | } |
387 | __set_current_state(TASK_RUNNING); | |
d67e03e3 EB |
388 | } |
389 | } | |
390 | ||
f98bafa0 | 391 | #ifdef CONFIG_MEMCG |
cf475ad2 | 392 | /* |
733eda7a | 393 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 394 | */ |
cf475ad2 BS |
395 | void mm_update_next_owner(struct mm_struct *mm) |
396 | { | |
397 | struct task_struct *c, *g, *p = current; | |
398 | ||
399 | retry: | |
733eda7a KH |
400 | /* |
401 | * If the exiting or execing task is not the owner, it's | |
402 | * someone else's problem. | |
403 | */ | |
404 | if (mm->owner != p) | |
cf475ad2 | 405 | return; |
733eda7a KH |
406 | /* |
407 | * The current owner is exiting/execing and there are no other | |
408 | * candidates. Do not leave the mm pointing to a possibly | |
409 | * freed task structure. | |
410 | */ | |
411 | if (atomic_read(&mm->mm_users) <= 1) { | |
987717e5 | 412 | WRITE_ONCE(mm->owner, NULL); |
733eda7a KH |
413 | return; |
414 | } | |
cf475ad2 BS |
415 | |
416 | read_lock(&tasklist_lock); | |
417 | /* | |
418 | * Search in the children | |
419 | */ | |
420 | list_for_each_entry(c, &p->children, sibling) { | |
421 | if (c->mm == mm) | |
422 | goto assign_new_owner; | |
423 | } | |
424 | ||
425 | /* | |
426 | * Search in the siblings | |
427 | */ | |
dea33cfd | 428 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
429 | if (c->mm == mm) |
430 | goto assign_new_owner; | |
431 | } | |
432 | ||
433 | /* | |
f87fb599 | 434 | * Search through everything else, we should not get here often. |
cf475ad2 | 435 | */ |
39af1765 ON |
436 | for_each_process(g) { |
437 | if (g->flags & PF_KTHREAD) | |
438 | continue; | |
439 | for_each_thread(g, c) { | |
440 | if (c->mm == mm) | |
441 | goto assign_new_owner; | |
442 | if (c->mm) | |
443 | break; | |
444 | } | |
f87fb599 | 445 | } |
cf475ad2 | 446 | read_unlock(&tasklist_lock); |
31a78f23 BS |
447 | /* |
448 | * We found no owner yet mm_users > 1: this implies that we are | |
449 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 450 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 451 | */ |
987717e5 | 452 | WRITE_ONCE(mm->owner, NULL); |
cf475ad2 BS |
453 | return; |
454 | ||
455 | assign_new_owner: | |
456 | BUG_ON(c == p); | |
457 | get_task_struct(c); | |
458 | /* | |
459 | * The task_lock protects c->mm from changing. | |
460 | * We always want mm->owner->mm == mm | |
461 | */ | |
462 | task_lock(c); | |
e5991371 HD |
463 | /* |
464 | * Delay read_unlock() till we have the task_lock() | |
465 | * to ensure that c does not slip away underneath us | |
466 | */ | |
467 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
468 | if (c->mm != mm) { |
469 | task_unlock(c); | |
470 | put_task_struct(c); | |
471 | goto retry; | |
472 | } | |
987717e5 | 473 | WRITE_ONCE(mm->owner, c); |
bd74fdae | 474 | lru_gen_migrate_mm(mm); |
cf475ad2 BS |
475 | task_unlock(c); |
476 | put_task_struct(c); | |
477 | } | |
f98bafa0 | 478 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 479 | |
1da177e4 LT |
480 | /* |
481 | * Turn us into a lazy TLB process if we | |
482 | * aren't already.. | |
483 | */ | |
0039962a | 484 | static void exit_mm(void) |
1da177e4 | 485 | { |
0039962a | 486 | struct mm_struct *mm = current->mm; |
1da177e4 | 487 | |
4610ba7a | 488 | exit_mm_release(current, mm); |
1da177e4 LT |
489 | if (!mm) |
490 | return; | |
4fe7efdb | 491 | sync_mm_rss(mm); |
d8ed45c5 | 492 | mmap_read_lock(mm); |
f1f10076 | 493 | mmgrab(mm); |
0039962a | 494 | BUG_ON(mm != current->active_mm); |
1da177e4 | 495 | /* more a memory barrier than a real lock */ |
0039962a | 496 | task_lock(current); |
5bc78502 MD |
497 | /* |
498 | * When a thread stops operating on an address space, the loop | |
499 | * in membarrier_private_expedited() may not observe that | |
500 | * tsk->mm, and the loop in membarrier_global_expedited() may | |
501 | * not observe a MEMBARRIER_STATE_GLOBAL_EXPEDITED | |
502 | * rq->membarrier_state, so those would not issue an IPI. | |
503 | * Membarrier requires a memory barrier after accessing | |
504 | * user-space memory, before clearing tsk->mm or the | |
505 | * rq->membarrier_state. | |
506 | */ | |
507 | smp_mb__after_spinlock(); | |
508 | local_irq_disable(); | |
0039962a | 509 | current->mm = NULL; |
5bc78502 | 510 | membarrier_update_current_mm(NULL); |
1da177e4 | 511 | enter_lazy_tlb(mm, current); |
5bc78502 | 512 | local_irq_enable(); |
0039962a | 513 | task_unlock(current); |
5bc78502 | 514 | mmap_read_unlock(mm); |
cf475ad2 | 515 | mm_update_next_owner(mm); |
1da177e4 | 516 | mmput(mm); |
c32b3cbe | 517 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 518 | exit_oom_victim(); |
1da177e4 LT |
519 | } |
520 | ||
c9dc05bf ON |
521 | static struct task_struct *find_alive_thread(struct task_struct *p) |
522 | { | |
523 | struct task_struct *t; | |
524 | ||
525 | for_each_thread(p, t) { | |
526 | if (!(t->flags & PF_EXITING)) | |
527 | return t; | |
528 | } | |
529 | return NULL; | |
530 | } | |
531 | ||
8fb335e0 AV |
532 | static struct task_struct *find_child_reaper(struct task_struct *father, |
533 | struct list_head *dead) | |
1109909c ON |
534 | __releases(&tasklist_lock) |
535 | __acquires(&tasklist_lock) | |
536 | { | |
537 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
538 | struct task_struct *reaper = pid_ns->child_reaper; | |
8fb335e0 | 539 | struct task_struct *p, *n; |
1109909c ON |
540 | |
541 | if (likely(reaper != father)) | |
542 | return reaper; | |
543 | ||
c9dc05bf ON |
544 | reaper = find_alive_thread(father); |
545 | if (reaper) { | |
1109909c ON |
546 | pid_ns->child_reaper = reaper; |
547 | return reaper; | |
548 | } | |
549 | ||
550 | write_unlock_irq(&tasklist_lock); | |
8fb335e0 AV |
551 | |
552 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
553 | list_del_init(&p->ptrace_entry); | |
554 | release_task(p); | |
555 | } | |
556 | ||
1109909c ON |
557 | zap_pid_ns_processes(pid_ns); |
558 | write_lock_irq(&tasklist_lock); | |
559 | ||
560 | return father; | |
561 | } | |
562 | ||
1da177e4 | 563 | /* |
ebec18a6 LP |
564 | * When we die, we re-parent all our children, and try to: |
565 | * 1. give them to another thread in our thread group, if such a member exists | |
566 | * 2. give it to the first ancestor process which prctl'd itself as a | |
567 | * child_subreaper for its children (like a service manager) | |
568 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 569 | */ |
1109909c ON |
570 | static struct task_struct *find_new_reaper(struct task_struct *father, |
571 | struct task_struct *child_reaper) | |
1da177e4 | 572 | { |
c9dc05bf | 573 | struct task_struct *thread, *reaper; |
1da177e4 | 574 | |
c9dc05bf ON |
575 | thread = find_alive_thread(father); |
576 | if (thread) | |
950bbabb | 577 | return thread; |
1da177e4 | 578 | |
7d24e2df | 579 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 580 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 581 | /* |
175aed3f | 582 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
583 | * We can't check reaper != child_reaper to ensure we do not |
584 | * cross the namespaces, the exiting parent could be injected | |
585 | * by setns() + fork(). | |
586 | * We check pid->level, this is slightly more efficient than | |
587 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 588 | */ |
c6c70f44 ON |
589 | for (reaper = father->real_parent; |
590 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 591 | reaper = reaper->real_parent) { |
175aed3f | 592 | if (reaper == &init_task) |
ebec18a6 LP |
593 | break; |
594 | if (!reaper->signal->is_child_subreaper) | |
595 | continue; | |
c9dc05bf ON |
596 | thread = find_alive_thread(reaper); |
597 | if (thread) | |
598 | return thread; | |
ebec18a6 | 599 | } |
1da177e4 | 600 | } |
762a24be | 601 | |
1109909c | 602 | return child_reaper; |
950bbabb ON |
603 | } |
604 | ||
5dfc80be ON |
605 | /* |
606 | * Any that need to be release_task'd are put on the @dead list. | |
607 | */ | |
9cd80bbb | 608 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
609 | struct list_head *dead) |
610 | { | |
2831096e | 611 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
612 | return; |
613 | ||
abd50b39 | 614 | /* We don't want people slaying init. */ |
5dfc80be ON |
615 | p->exit_signal = SIGCHLD; |
616 | ||
617 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 618 | if (!p->ptrace && |
5dfc80be | 619 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 620 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 621 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 622 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
623 | } |
624 | } | |
625 | ||
626 | kill_orphaned_pgrp(p, father); | |
627 | } | |
628 | ||
482a3767 ON |
629 | /* |
630 | * This does two things: | |
631 | * | |
632 | * A. Make init inherit all the child processes | |
633 | * B. Check to see if any process groups have become orphaned | |
634 | * as a result of our exiting, and if they have any stopped | |
635 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
636 | */ | |
637 | static void forget_original_parent(struct task_struct *father, | |
638 | struct list_head *dead) | |
1da177e4 | 639 | { |
482a3767 | 640 | struct task_struct *p, *t, *reaper; |
762a24be | 641 | |
7c8bd232 | 642 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 643 | exit_ptrace(father, dead); |
f470021a | 644 | |
7c8bd232 | 645 | /* Can drop and reacquire tasklist_lock */ |
8fb335e0 | 646 | reaper = find_child_reaper(father, dead); |
ad9e206a | 647 | if (list_empty(&father->children)) |
482a3767 | 648 | return; |
1109909c ON |
649 | |
650 | reaper = find_new_reaper(father, reaper); | |
2831096e | 651 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 652 | for_each_thread(p, t) { |
22a34c6f MB |
653 | RCU_INIT_POINTER(t->real_parent, reaper); |
654 | BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father)); | |
57a05918 | 655 | if (likely(!t->ptrace)) |
9cd80bbb | 656 | t->parent = t->real_parent; |
9cd80bbb ON |
657 | if (t->pdeath_signal) |
658 | group_send_sig_info(t->pdeath_signal, | |
01024980 EB |
659 | SEND_SIG_NOINFO, t, |
660 | PIDTYPE_TGID); | |
57a05918 | 661 | } |
2831096e ON |
662 | /* |
663 | * If this is a threaded reparent there is no need to | |
664 | * notify anyone anything has happened. | |
665 | */ | |
666 | if (!same_thread_group(reaper, father)) | |
482a3767 | 667 | reparent_leader(father, p, dead); |
1da177e4 | 668 | } |
2831096e | 669 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
670 | } |
671 | ||
672 | /* | |
673 | * Send signals to all our closest relatives so that they know | |
674 | * to properly mourn us.. | |
675 | */ | |
821c7de7 | 676 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 677 | { |
53c8f9f1 | 678 | bool autoreap; |
482a3767 ON |
679 | struct task_struct *p, *n; |
680 | LIST_HEAD(dead); | |
1da177e4 | 681 | |
762a24be | 682 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
683 | forget_original_parent(tsk, &dead); |
684 | ||
821c7de7 ON |
685 | if (group_dead) |
686 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 687 | |
b191d649 | 688 | tsk->exit_state = EXIT_ZOMBIE; |
45cdf5cc ON |
689 | if (unlikely(tsk->ptrace)) { |
690 | int sig = thread_group_leader(tsk) && | |
691 | thread_group_empty(tsk) && | |
692 | !ptrace_reparented(tsk) ? | |
693 | tsk->exit_signal : SIGCHLD; | |
694 | autoreap = do_notify_parent(tsk, sig); | |
695 | } else if (thread_group_leader(tsk)) { | |
696 | autoreap = thread_group_empty(tsk) && | |
697 | do_notify_parent(tsk, tsk->exit_signal); | |
698 | } else { | |
699 | autoreap = true; | |
700 | } | |
1da177e4 | 701 | |
30b692d3 CB |
702 | if (autoreap) { |
703 | tsk->exit_state = EXIT_DEAD; | |
6c66e7db | 704 | list_add(&tsk->ptrace_entry, &dead); |
30b692d3 | 705 | } |
1da177e4 | 706 | |
9c339168 ON |
707 | /* mt-exec, de_thread() is waiting for group leader */ |
708 | if (unlikely(tsk->signal->notify_count < 0)) | |
60700e38 | 709 | wake_up_process(tsk->signal->group_exec_task); |
1da177e4 LT |
710 | write_unlock_irq(&tasklist_lock); |
711 | ||
482a3767 ON |
712 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
713 | list_del_init(&p->ptrace_entry); | |
714 | release_task(p); | |
715 | } | |
1da177e4 LT |
716 | } |
717 | ||
e18eecb8 JD |
718 | #ifdef CONFIG_DEBUG_STACK_USAGE |
719 | static void check_stack_usage(void) | |
720 | { | |
721 | static DEFINE_SPINLOCK(low_water_lock); | |
722 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
723 | unsigned long free; |
724 | ||
7c9f8861 | 725 | free = stack_not_used(current); |
e18eecb8 JD |
726 | |
727 | if (free >= lowest_to_date) | |
728 | return; | |
729 | ||
730 | spin_lock(&low_water_lock); | |
731 | if (free < lowest_to_date) { | |
627393d4 | 732 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 733 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
734 | lowest_to_date = free; |
735 | } | |
736 | spin_unlock(&low_water_lock); | |
737 | } | |
738 | #else | |
739 | static inline void check_stack_usage(void) {} | |
740 | #endif | |
741 | ||
d80f7d7b EB |
742 | static void synchronize_group_exit(struct task_struct *tsk, long code) |
743 | { | |
744 | struct sighand_struct *sighand = tsk->sighand; | |
745 | struct signal_struct *signal = tsk->signal; | |
746 | ||
747 | spin_lock_irq(&sighand->siglock); | |
748 | signal->quick_threads--; | |
749 | if ((signal->quick_threads == 0) && | |
750 | !(signal->flags & SIGNAL_GROUP_EXIT)) { | |
751 | signal->flags = SIGNAL_GROUP_EXIT; | |
752 | signal->group_exit_code = code; | |
753 | signal->group_stop_count = 0; | |
754 | } | |
755 | spin_unlock_irq(&sighand->siglock); | |
756 | } | |
757 | ||
9af6528e | 758 | void __noreturn do_exit(long code) |
1da177e4 LT |
759 | { |
760 | struct task_struct *tsk = current; | |
761 | int group_dead; | |
762 | ||
d80f7d7b EB |
763 | synchronize_group_exit(tsk, code); |
764 | ||
b1f866b0 | 765 | WARN_ON(tsk->plug); |
22e2c507 | 766 | |
586b58ca | 767 | kcov_task_exit(tsk); |
50b5e49c | 768 | kmsan_task_exit(tsk); |
586b58ca | 769 | |
92307383 | 770 | coredump_task_exit(tsk); |
a288eecc | 771 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 772 | |
e0e81739 DH |
773 | validate_creds_for_do_exit(tsk); |
774 | ||
f552a27a | 775 | io_uring_files_cancel(); |
d12619b5 | 776 | exit_signals(tsk); /* sets PF_EXITING */ |
1da177e4 | 777 | |
48d212a2 LT |
778 | /* sync mm's RSS info before statistics gathering */ |
779 | if (tsk->mm) | |
780 | sync_mm_rss(tsk->mm); | |
51229b49 | 781 | acct_update_integrals(tsk); |
1da177e4 | 782 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 783 | if (group_dead) { |
43cf75d9 | 784 | /* |
785 | * If the last thread of global init has exited, panic | |
786 | * immediately to get a useable coredump. | |
787 | */ | |
788 | if (unlikely(is_global_init(tsk))) | |
789 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
790 | tsk->signal->group_exit_code ?: (int)code); | |
791 | ||
baa73d9e | 792 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 793 | hrtimer_cancel(&tsk->signal->real_timer); |
d5b36a4d | 794 | exit_itimers(tsk); |
baa73d9e | 795 | #endif |
1f10206c JP |
796 | if (tsk->mm) |
797 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 798 | } |
f6ec29a4 | 799 | acct_collect(code, group_dead); |
522ed776 MT |
800 | if (group_dead) |
801 | tty_audit_exit(); | |
a4ff8dba | 802 | audit_free(tsk); |
115085ea | 803 | |
48d212a2 | 804 | tsk->exit_code = code; |
115085ea | 805 | taskstats_exit(tsk, group_dead); |
c757249a | 806 | |
0039962a | 807 | exit_mm(); |
1da177e4 | 808 | |
0e464814 | 809 | if (group_dead) |
f6ec29a4 | 810 | acct_process(); |
0a16b607 MD |
811 | trace_sched_process_exit(tsk); |
812 | ||
1da177e4 | 813 | exit_sem(tsk); |
b34a6b1d | 814 | exit_shm(tsk); |
1ec7f1dd AV |
815 | exit_files(tsk); |
816 | exit_fs(tsk); | |
c39df5fa ON |
817 | if (group_dead) |
818 | disassociate_ctty(1); | |
8aac6270 | 819 | exit_task_namespaces(tsk); |
ed3e694d | 820 | exit_task_work(tsk); |
e6464694 | 821 | exit_thread(tsk); |
0b3fcf17 SE |
822 | |
823 | /* | |
824 | * Flush inherited counters to the parent - before the parent | |
825 | * gets woken up by child-exit notifications. | |
826 | * | |
827 | * because of cgroup mode, must be called before cgroup_exit() | |
828 | */ | |
829 | perf_event_exit_task(tsk); | |
830 | ||
8e5bfa8c | 831 | sched_autogroup_exit_task(tsk); |
1ec41830 | 832 | cgroup_exit(tsk); |
1da177e4 | 833 | |
24f1e32c FW |
834 | /* |
835 | * FIXME: do that only when needed, using sched_exit tracepoint | |
836 | */ | |
7c8df286 | 837 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 838 | |
ccdd29ff | 839 | exit_tasks_rcu_start(); |
821c7de7 | 840 | exit_notify(tsk, group_dead); |
ef982393 | 841 | proc_exit_connector(tsk); |
c11600e4 | 842 | mpol_put_task_policy(tsk); |
42b2dd0a | 843 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
844 | if (unlikely(current->pi_state_cache)) |
845 | kfree(current->pi_state_cache); | |
42b2dd0a | 846 | #endif |
de5097c2 | 847 | /* |
9a11b49a | 848 | * Make sure we are holding no locks: |
de5097c2 | 849 | */ |
1b1d2fb4 | 850 | debug_check_no_locks_held(); |
1da177e4 | 851 | |
afc847b7 | 852 | if (tsk->io_context) |
b69f2292 | 853 | exit_io_context(tsk); |
afc847b7 | 854 | |
b92ce558 | 855 | if (tsk->splice_pipe) |
4b8a8f1e | 856 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 857 | |
5640f768 ED |
858 | if (tsk->task_frag.page) |
859 | put_page(tsk->task_frag.page); | |
860 | ||
e0e81739 | 861 | validate_creds_for_do_exit(tsk); |
1a03d3f1 | 862 | exit_task_stack_account(tsk); |
e0e81739 | 863 | |
4bcb8232 | 864 | check_stack_usage(); |
7407251a | 865 | preempt_disable(); |
54848d73 WF |
866 | if (tsk->nr_dirtied) |
867 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 868 | exit_rcu(); |
ccdd29ff | 869 | exit_tasks_rcu_finish(); |
b5740f4b | 870 | |
b09be676 | 871 | lockdep_free_task(tsk); |
9af6528e | 872 | do_task_dead(); |
1da177e4 | 873 | } |
012914da | 874 | |
0e25498f EB |
875 | void __noreturn make_task_dead(int signr) |
876 | { | |
877 | /* | |
878 | * Take the task off the cpu after something catastrophic has | |
879 | * happened. | |
05ea0424 EB |
880 | * |
881 | * We can get here from a kernel oops, sometimes with preemption off. | |
882 | * Start by checking for critical errors. | |
883 | * Then fix up important state like USER_DS and preemption. | |
884 | * Then do everything else. | |
0e25498f | 885 | */ |
05ea0424 EB |
886 | struct task_struct *tsk = current; |
887 | ||
888 | if (unlikely(in_interrupt())) | |
889 | panic("Aiee, killing interrupt handler!"); | |
890 | if (unlikely(!tsk->pid)) | |
891 | panic("Attempted to kill the idle task!"); | |
892 | ||
05ea0424 EB |
893 | if (unlikely(in_atomic())) { |
894 | pr_info("note: %s[%d] exited with preempt_count %d\n", | |
895 | current->comm, task_pid_nr(current), | |
896 | preempt_count()); | |
897 | preempt_count_set(PREEMPT_ENABLED); | |
898 | } | |
899 | ||
900 | /* | |
901 | * We're taking recursive faults here in make_task_dead. Safest is to just | |
902 | * leave this task alone and wait for reboot. | |
903 | */ | |
904 | if (unlikely(tsk->flags & PF_EXITING)) { | |
905 | pr_alert("Fixing recursive fault but reboot is needed!\n"); | |
906 | futex_exit_recursive(tsk); | |
912616f1 EB |
907 | tsk->exit_state = EXIT_DEAD; |
908 | refcount_inc(&tsk->rcu_users); | |
7f80a2fd | 909 | do_task_dead(); |
05ea0424 EB |
910 | } |
911 | ||
0e25498f EB |
912 | do_exit(signr); |
913 | } | |
914 | ||
754fe8d2 | 915 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
916 | { |
917 | do_exit((error_code&0xff)<<8); | |
918 | } | |
919 | ||
1da177e4 LT |
920 | /* |
921 | * Take down every thread in the group. This is called by fatal signals | |
922 | * as well as by sys_exit_group (below). | |
923 | */ | |
eae654f1 | 924 | void __noreturn |
1da177e4 LT |
925 | do_group_exit(int exit_code) |
926 | { | |
bfc4b089 ON |
927 | struct signal_struct *sig = current->signal; |
928 | ||
49697335 | 929 | if (sig->flags & SIGNAL_GROUP_EXIT) |
bfc4b089 | 930 | exit_code = sig->group_exit_code; |
49697335 EB |
931 | else if (sig->group_exec_task) |
932 | exit_code = 0; | |
cbe9dac3 | 933 | else { |
1da177e4 | 934 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 935 | |
1da177e4 | 936 | spin_lock_irq(&sighand->siglock); |
49697335 | 937 | if (sig->flags & SIGNAL_GROUP_EXIT) |
1da177e4 LT |
938 | /* Another thread got here before we took the lock. */ |
939 | exit_code = sig->group_exit_code; | |
49697335 EB |
940 | else if (sig->group_exec_task) |
941 | exit_code = 0; | |
1da177e4 | 942 | else { |
1da177e4 | 943 | sig->group_exit_code = exit_code; |
ed5d2cac | 944 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
945 | zap_other_threads(current); |
946 | } | |
947 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
948 | } |
949 | ||
950 | do_exit(exit_code); | |
951 | /* NOTREACHED */ | |
952 | } | |
953 | ||
954 | /* | |
955 | * this kills every thread in the thread group. Note that any externally | |
956 | * wait4()-ing process will get the correct exit code - even if this | |
957 | * thread is not the thread group leader. | |
958 | */ | |
754fe8d2 | 959 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
960 | { |
961 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
962 | /* NOTREACHED */ |
963 | return 0; | |
1da177e4 LT |
964 | } |
965 | ||
67d7ddde AV |
966 | struct waitid_info { |
967 | pid_t pid; | |
968 | uid_t uid; | |
969 | int status; | |
970 | int cause; | |
971 | }; | |
972 | ||
9e8ae01d ON |
973 | struct wait_opts { |
974 | enum pid_type wo_type; | |
9e8ae01d | 975 | int wo_flags; |
e1eb1ebc | 976 | struct pid *wo_pid; |
9e8ae01d | 977 | |
67d7ddde | 978 | struct waitid_info *wo_info; |
359566fa | 979 | int wo_stat; |
ce72a16f | 980 | struct rusage *wo_rusage; |
9e8ae01d | 981 | |
ac6424b9 | 982 | wait_queue_entry_t child_wait; |
9e8ae01d ON |
983 | int notask_error; |
984 | }; | |
985 | ||
989264f4 | 986 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 987 | { |
5c01ba49 ON |
988 | return wo->wo_type == PIDTYPE_MAX || |
989 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
990 | } | |
1da177e4 | 991 | |
bf959931 ON |
992 | static int |
993 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
994 | { |
995 | if (!eligible_pid(wo, p)) | |
996 | return 0; | |
bf959931 ON |
997 | |
998 | /* | |
999 | * Wait for all children (clone and not) if __WALL is set or | |
1000 | * if it is traced by us. | |
1001 | */ | |
1002 | if (ptrace || (wo->wo_flags & __WALL)) | |
1003 | return 1; | |
1004 | ||
1005 | /* | |
1006 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
1007 | * otherwise, wait for non-clone children *only*. | |
1008 | * | |
1009 | * Note: a "clone" child here is one that reports to its parent | |
1010 | * using a signal other than SIGCHLD, or a non-leader thread which | |
1011 | * we can only see if it is traced by us. | |
1012 | */ | |
1013 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 1014 | return 0; |
1da177e4 | 1015 | |
14dd0b81 | 1016 | return 1; |
1da177e4 LT |
1017 | } |
1018 | ||
1da177e4 LT |
1019 | /* |
1020 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1021 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1022 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1023 | * released the lock and the system call should return. | |
1024 | */ | |
9e8ae01d | 1025 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1026 | { |
67d7ddde | 1027 | int state, status; |
6c5f3e7b | 1028 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1029 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
67d7ddde | 1030 | struct waitid_info *infop; |
1da177e4 | 1031 | |
9e8ae01d | 1032 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1033 | return 0; |
1034 | ||
9e8ae01d | 1035 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
907c311f EB |
1036 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1037 | ? p->signal->group_exit_code : p->exit_code; | |
1da177e4 LT |
1038 | get_task_struct(p); |
1039 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1040 | sched_annotate_sleep(); |
e61a2502 AV |
1041 | if (wo->wo_rusage) |
1042 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1043 | put_task_struct(p); |
76d9871e | 1044 | goto out_info; |
1da177e4 | 1045 | } |
1da177e4 | 1046 | /* |
abd50b39 | 1047 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1048 | */ |
f6507f83 ON |
1049 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1050 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1051 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1052 | return 0; |
986094df ON |
1053 | /* |
1054 | * We own this thread, nobody else can reap it. | |
1055 | */ | |
1056 | read_unlock(&tasklist_lock); | |
1057 | sched_annotate_sleep(); | |
f6507f83 | 1058 | |
befca967 | 1059 | /* |
f6507f83 | 1060 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1061 | */ |
f6507f83 | 1062 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1063 | struct signal_struct *sig = p->signal; |
1064 | struct signal_struct *psig = current->signal; | |
1f10206c | 1065 | unsigned long maxrss; |
5613fda9 | 1066 | u64 tgutime, tgstime; |
3795e161 | 1067 | |
1da177e4 LT |
1068 | /* |
1069 | * The resource counters for the group leader are in its | |
1070 | * own task_struct. Those for dead threads in the group | |
1071 | * are in its signal_struct, as are those for the child | |
1072 | * processes it has previously reaped. All these | |
1073 | * accumulate in the parent's signal_struct c* fields. | |
1074 | * | |
1075 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1076 | * p->signal fields because the whole thread group is dead |
1077 | * and nobody can change them. | |
1078 | * | |
dcca3475 | 1079 | * psig->stats_lock also protects us from our sub-threads |
f953ccd0 ON |
1080 | * which can reap other children at the same time. Until |
1081 | * we change k_getrusage()-like users to rely on this lock | |
1082 | * we have to take ->siglock as well. | |
0cf55e1e | 1083 | * |
a0be55de IA |
1084 | * We use thread_group_cputime_adjusted() to get times for |
1085 | * the thread group, which consolidates times for all threads | |
1086 | * in the group including the group leader. | |
1da177e4 | 1087 | */ |
e80d0a1a | 1088 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1089 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1090 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1091 | psig->cutime += tgutime + sig->cutime; |
1092 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1093 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1094 | psig->cmin_flt += |
1095 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1096 | psig->cmaj_flt += | |
1097 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1098 | psig->cnvcsw += | |
1099 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1100 | psig->cnivcsw += | |
1101 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1102 | psig->cinblock += |
1103 | task_io_get_inblock(p) + | |
1104 | sig->inblock + sig->cinblock; | |
1105 | psig->coublock += | |
1106 | task_io_get_oublock(p) + | |
1107 | sig->oublock + sig->coublock; | |
1f10206c JP |
1108 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1109 | if (psig->cmaxrss < maxrss) | |
1110 | psig->cmaxrss = maxrss; | |
5995477a AR |
1111 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1112 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1113 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1114 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1115 | } |
1116 | ||
ce72a16f AV |
1117 | if (wo->wo_rusage) |
1118 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
1da177e4 LT |
1119 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1120 | ? p->signal->group_exit_code : p->exit_code; | |
359566fa | 1121 | wo->wo_stat = status; |
2f4e6e2a | 1122 | |
b4360690 | 1123 | if (state == EXIT_TRACE) { |
1da177e4 | 1124 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1125 | /* We dropped tasklist, ptracer could die and untrace */ |
1126 | ptrace_unlink(p); | |
b4360690 ON |
1127 | |
1128 | /* If parent wants a zombie, don't release it now */ | |
1129 | state = EXIT_ZOMBIE; | |
1130 | if (do_notify_parent(p, p->exit_signal)) | |
1131 | state = EXIT_DEAD; | |
abd50b39 | 1132 | p->exit_state = state; |
1da177e4 LT |
1133 | write_unlock_irq(&tasklist_lock); |
1134 | } | |
abd50b39 | 1135 | if (state == EXIT_DEAD) |
1da177e4 | 1136 | release_task(p); |
2f4e6e2a | 1137 | |
76d9871e AV |
1138 | out_info: |
1139 | infop = wo->wo_info; | |
1140 | if (infop) { | |
1141 | if ((status & 0x7f) == 0) { | |
1142 | infop->cause = CLD_EXITED; | |
1143 | infop->status = status >> 8; | |
1144 | } else { | |
1145 | infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1146 | infop->status = status & 0x7f; | |
1147 | } | |
1148 | infop->pid = pid; | |
1149 | infop->uid = uid; | |
1150 | } | |
1151 | ||
67d7ddde | 1152 | return pid; |
1da177e4 LT |
1153 | } |
1154 | ||
90bc8d8b ON |
1155 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1156 | { | |
1157 | if (ptrace) { | |
570ac933 | 1158 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1159 | return &p->exit_code; |
1160 | } else { | |
1161 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1162 | return &p->signal->group_exit_code; | |
1163 | } | |
1164 | return NULL; | |
1165 | } | |
1166 | ||
19e27463 TH |
1167 | /** |
1168 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1169 | * @wo: wait options | |
1170 | * @ptrace: is the wait for ptrace | |
1171 | * @p: task to wait for | |
1172 | * | |
1173 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1174 | * | |
1175 | * CONTEXT: | |
1176 | * read_lock(&tasklist_lock), which is released if return value is | |
1177 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1178 | * | |
1179 | * RETURNS: | |
1180 | * 0 if wait condition didn't exist and search for other wait conditions | |
1181 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1182 | * success, implies that tasklist_lock is released and wait condition | |
1183 | * search should terminate. | |
1da177e4 | 1184 | */ |
9e8ae01d ON |
1185 | static int wait_task_stopped(struct wait_opts *wo, |
1186 | int ptrace, struct task_struct *p) | |
1da177e4 | 1187 | { |
67d7ddde AV |
1188 | struct waitid_info *infop; |
1189 | int exit_code, *p_code, why; | |
ee7c82da | 1190 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1191 | pid_t pid; |
1da177e4 | 1192 | |
47918025 ON |
1193 | /* |
1194 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1195 | */ | |
9e8ae01d | 1196 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1197 | return 0; |
1198 | ||
19e27463 TH |
1199 | if (!task_stopped_code(p, ptrace)) |
1200 | return 0; | |
1201 | ||
ee7c82da ON |
1202 | exit_code = 0; |
1203 | spin_lock_irq(&p->sighand->siglock); | |
1204 | ||
90bc8d8b ON |
1205 | p_code = task_stopped_code(p, ptrace); |
1206 | if (unlikely(!p_code)) | |
ee7c82da ON |
1207 | goto unlock_sig; |
1208 | ||
90bc8d8b | 1209 | exit_code = *p_code; |
ee7c82da ON |
1210 | if (!exit_code) |
1211 | goto unlock_sig; | |
1212 | ||
9e8ae01d | 1213 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1214 | *p_code = 0; |
ee7c82da | 1215 | |
8ca937a6 | 1216 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1217 | unlock_sig: |
1218 | spin_unlock_irq(&p->sighand->siglock); | |
1219 | if (!exit_code) | |
1da177e4 LT |
1220 | return 0; |
1221 | ||
1222 | /* | |
1223 | * Now we are pretty sure this task is interesting. | |
1224 | * Make sure it doesn't get reaped out from under us while we | |
1225 | * give up the lock and then examine it below. We don't want to | |
1226 | * keep holding onto the tasklist_lock while we call getrusage and | |
1227 | * possibly take page faults for user memory. | |
1228 | */ | |
1229 | get_task_struct(p); | |
6c5f3e7b | 1230 | pid = task_pid_vnr(p); |
f470021a | 1231 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1232 | read_unlock(&tasklist_lock); |
1029a2b5 | 1233 | sched_annotate_sleep(); |
e61a2502 AV |
1234 | if (wo->wo_rusage) |
1235 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1236 | put_task_struct(p); |
1da177e4 | 1237 | |
bb380ec3 AV |
1238 | if (likely(!(wo->wo_flags & WNOWAIT))) |
1239 | wo->wo_stat = (exit_code << 8) | 0x7f; | |
1da177e4 | 1240 | |
9e8ae01d | 1241 | infop = wo->wo_info; |
67d7ddde AV |
1242 | if (infop) { |
1243 | infop->cause = why; | |
1244 | infop->status = exit_code; | |
1245 | infop->pid = pid; | |
1246 | infop->uid = uid; | |
1247 | } | |
67d7ddde | 1248 | return pid; |
1da177e4 LT |
1249 | } |
1250 | ||
1251 | /* | |
1252 | * Handle do_wait work for one task in a live, non-stopped state. | |
1253 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1254 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1255 | * released the lock and the system call should return. | |
1256 | */ | |
9e8ae01d | 1257 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1258 | { |
bb380ec3 | 1259 | struct waitid_info *infop; |
1da177e4 LT |
1260 | pid_t pid; |
1261 | uid_t uid; | |
1262 | ||
9e8ae01d | 1263 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1264 | return 0; |
1265 | ||
1da177e4 LT |
1266 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1267 | return 0; | |
1268 | ||
1269 | spin_lock_irq(&p->sighand->siglock); | |
1270 | /* Re-check with the lock held. */ | |
1271 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1272 | spin_unlock_irq(&p->sighand->siglock); | |
1273 | return 0; | |
1274 | } | |
9e8ae01d | 1275 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1276 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1277 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1278 | spin_unlock_irq(&p->sighand->siglock); |
1279 | ||
6c5f3e7b | 1280 | pid = task_pid_vnr(p); |
1da177e4 LT |
1281 | get_task_struct(p); |
1282 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1283 | sched_annotate_sleep(); |
e61a2502 AV |
1284 | if (wo->wo_rusage) |
1285 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1286 | put_task_struct(p); |
1da177e4 | 1287 | |
bb380ec3 AV |
1288 | infop = wo->wo_info; |
1289 | if (!infop) { | |
359566fa | 1290 | wo->wo_stat = 0xffff; |
1da177e4 | 1291 | } else { |
bb380ec3 AV |
1292 | infop->cause = CLD_CONTINUED; |
1293 | infop->pid = pid; | |
1294 | infop->uid = uid; | |
1295 | infop->status = SIGCONT; | |
1da177e4 | 1296 | } |
bb380ec3 | 1297 | return pid; |
1da177e4 LT |
1298 | } |
1299 | ||
98abed02 RM |
1300 | /* |
1301 | * Consider @p for a wait by @parent. | |
1302 | * | |
9e8ae01d | 1303 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1304 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1305 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1306 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1307 | * or still -ECHILD. |
98abed02 | 1308 | */ |
b6e763f0 ON |
1309 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1310 | struct task_struct *p) | |
98abed02 | 1311 | { |
3245d6ac ON |
1312 | /* |
1313 | * We can race with wait_task_zombie() from another thread. | |
1314 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1315 | * can't confuse the checks below. | |
1316 | */ | |
6aa7de05 | 1317 | int exit_state = READ_ONCE(p->exit_state); |
b3ab0316 ON |
1318 | int ret; |
1319 | ||
3245d6ac | 1320 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1321 | return 0; |
1322 | ||
bf959931 | 1323 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1324 | if (!ret) |
98abed02 RM |
1325 | return ret; |
1326 | ||
3245d6ac | 1327 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1328 | /* |
abd50b39 ON |
1329 | * ptrace == 0 means we are the natural parent. In this case |
1330 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1331 | */ |
abd50b39 | 1332 | if (likely(!ptrace)) |
50b8d257 | 1333 | wo->notask_error = 0; |
823b018e | 1334 | return 0; |
50b8d257 | 1335 | } |
823b018e | 1336 | |
377d75da ON |
1337 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1338 | /* | |
1339 | * If it is traced by its real parent's group, just pretend | |
1340 | * the caller is ptrace_do_wait() and reap this child if it | |
1341 | * is zombie. | |
1342 | * | |
1343 | * This also hides group stop state from real parent; otherwise | |
1344 | * a single stop can be reported twice as group and ptrace stop. | |
1345 | * If a ptracer wants to distinguish these two events for its | |
1346 | * own children it should create a separate process which takes | |
1347 | * the role of real parent. | |
1348 | */ | |
1349 | if (!ptrace_reparented(p)) | |
1350 | ptrace = 1; | |
1351 | } | |
1352 | ||
45cb24a1 | 1353 | /* slay zombie? */ |
3245d6ac | 1354 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1355 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1356 | if (!delay_group_leader(p)) { |
1357 | /* | |
1358 | * A zombie ptracee is only visible to its ptracer. | |
1359 | * Notification and reaping will be cascaded to the | |
1360 | * real parent when the ptracer detaches. | |
1361 | */ | |
1362 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1363 | return wait_task_zombie(wo, p); | |
1364 | } | |
98abed02 | 1365 | |
f470021a | 1366 | /* |
9b84cca2 TH |
1367 | * Allow access to stopped/continued state via zombie by |
1368 | * falling through. Clearing of notask_error is complex. | |
1369 | * | |
1370 | * When !@ptrace: | |
1371 | * | |
1372 | * If WEXITED is set, notask_error should naturally be | |
1373 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1374 | * so, if there are live subthreads, there are events to | |
1375 | * wait for. If all subthreads are dead, it's still safe | |
1376 | * to clear - this function will be called again in finite | |
1377 | * amount time once all the subthreads are released and | |
1378 | * will then return without clearing. | |
1379 | * | |
1380 | * When @ptrace: | |
1381 | * | |
1382 | * Stopped state is per-task and thus can't change once the | |
1383 | * target task dies. Only continued and exited can happen. | |
1384 | * Clear notask_error if WCONTINUED | WEXITED. | |
1385 | */ | |
1386 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1387 | wo->notask_error = 0; | |
1388 | } else { | |
1389 | /* | |
1390 | * @p is alive and it's gonna stop, continue or exit, so | |
1391 | * there always is something to wait for. | |
f470021a | 1392 | */ |
9e8ae01d | 1393 | wo->notask_error = 0; |
f470021a RM |
1394 | } |
1395 | ||
98abed02 | 1396 | /* |
45cb24a1 TH |
1397 | * Wait for stopped. Depending on @ptrace, different stopped state |
1398 | * is used and the two don't interact with each other. | |
98abed02 | 1399 | */ |
19e27463 TH |
1400 | ret = wait_task_stopped(wo, ptrace, p); |
1401 | if (ret) | |
1402 | return ret; | |
98abed02 RM |
1403 | |
1404 | /* | |
45cb24a1 TH |
1405 | * Wait for continued. There's only one continued state and the |
1406 | * ptracer can consume it which can confuse the real parent. Don't | |
1407 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1408 | */ |
9e8ae01d | 1409 | return wait_task_continued(wo, p); |
98abed02 RM |
1410 | } |
1411 | ||
1412 | /* | |
1413 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1414 | * | |
9e8ae01d | 1415 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1416 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1417 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1418 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1419 | * or still -ECHILD. |
98abed02 | 1420 | */ |
9e8ae01d | 1421 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1422 | { |
1423 | struct task_struct *p; | |
1424 | ||
1425 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1426 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1427 | |
9cd80bbb ON |
1428 | if (ret) |
1429 | return ret; | |
98abed02 RM |
1430 | } |
1431 | ||
1432 | return 0; | |
1433 | } | |
1434 | ||
9e8ae01d | 1435 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1436 | { |
1437 | struct task_struct *p; | |
1438 | ||
f470021a | 1439 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1440 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1441 | |
f470021a | 1442 | if (ret) |
98abed02 | 1443 | return ret; |
98abed02 RM |
1444 | } |
1445 | ||
1446 | return 0; | |
1447 | } | |
1448 | ||
ac6424b9 | 1449 | static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode, |
0b7570e7 ON |
1450 | int sync, void *key) |
1451 | { | |
1452 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1453 | child_wait); | |
1454 | struct task_struct *p = key; | |
1455 | ||
5c01ba49 | 1456 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1457 | return 0; |
1458 | ||
b4fe5182 ON |
1459 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1460 | return 0; | |
1461 | ||
0b7570e7 ON |
1462 | return default_wake_function(wait, mode, sync, key); |
1463 | } | |
1464 | ||
a7f0765e ON |
1465 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1466 | { | |
0b7570e7 | 1467 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
ce4dd442 | 1468 | TASK_INTERRUPTIBLE, p); |
a7f0765e ON |
1469 | } |
1470 | ||
5449162a JN |
1471 | static bool is_effectively_child(struct wait_opts *wo, bool ptrace, |
1472 | struct task_struct *target) | |
1473 | { | |
1474 | struct task_struct *parent = | |
1475 | !ptrace ? target->real_parent : target->parent; | |
1476 | ||
1477 | return current == parent || (!(wo->wo_flags & __WNOTHREAD) && | |
1478 | same_thread_group(current, parent)); | |
1479 | } | |
1480 | ||
1481 | /* | |
1482 | * Optimization for waiting on PIDTYPE_PID. No need to iterate through child | |
1483 | * and tracee lists to find the target task. | |
1484 | */ | |
1485 | static int do_wait_pid(struct wait_opts *wo) | |
1486 | { | |
1487 | bool ptrace; | |
1488 | struct task_struct *target; | |
1489 | int retval; | |
1490 | ||
1491 | ptrace = false; | |
1492 | target = pid_task(wo->wo_pid, PIDTYPE_TGID); | |
1493 | if (target && is_effectively_child(wo, ptrace, target)) { | |
1494 | retval = wait_consider_task(wo, ptrace, target); | |
1495 | if (retval) | |
1496 | return retval; | |
1497 | } | |
1498 | ||
1499 | ptrace = true; | |
1500 | target = pid_task(wo->wo_pid, PIDTYPE_PID); | |
1501 | if (target && target->ptrace && | |
1502 | is_effectively_child(wo, ptrace, target)) { | |
1503 | retval = wait_consider_task(wo, ptrace, target); | |
1504 | if (retval) | |
1505 | return retval; | |
1506 | } | |
1507 | ||
1508 | return 0; | |
1509 | } | |
1510 | ||
9e8ae01d | 1511 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1512 | { |
98abed02 | 1513 | int retval; |
1da177e4 | 1514 | |
9e8ae01d | 1515 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1516 | |
0b7570e7 ON |
1517 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1518 | wo->child_wait.private = current; | |
1519 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1520 | repeat: |
98abed02 | 1521 | /* |
3da56d16 | 1522 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1523 | * We will clear ->notask_error to zero if we see any child that |
1524 | * might later match our criteria, even if we are not able to reap | |
1525 | * it yet. | |
98abed02 | 1526 | */ |
64a16caf | 1527 | wo->notask_error = -ECHILD; |
9e8ae01d | 1528 | if ((wo->wo_type < PIDTYPE_MAX) && |
1722c14a | 1529 | (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type))) |
64a16caf | 1530 | goto notask; |
161550d7 | 1531 | |
f95d39d1 | 1532 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 | 1533 | read_lock(&tasklist_lock); |
9e8ae01d | 1534 | |
5449162a JN |
1535 | if (wo->wo_type == PIDTYPE_PID) { |
1536 | retval = do_wait_pid(wo); | |
64a16caf | 1537 | if (retval) |
98abed02 | 1538 | goto end; |
5449162a JN |
1539 | } else { |
1540 | struct task_struct *tsk = current; | |
1541 | ||
1542 | do { | |
1543 | retval = do_wait_thread(wo, tsk); | |
1544 | if (retval) | |
1545 | goto end; | |
98abed02 | 1546 | |
5449162a JN |
1547 | retval = ptrace_do_wait(wo, tsk); |
1548 | if (retval) | |
1549 | goto end; | |
1550 | ||
1551 | if (wo->wo_flags & __WNOTHREAD) | |
1552 | break; | |
1553 | } while_each_thread(current, tsk); | |
1554 | } | |
1da177e4 | 1555 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1556 | |
64a16caf | 1557 | notask: |
9e8ae01d ON |
1558 | retval = wo->notask_error; |
1559 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1560 | retval = -ERESTARTSYS; |
98abed02 RM |
1561 | if (!signal_pending(current)) { |
1562 | schedule(); | |
1563 | goto repeat; | |
1564 | } | |
1da177e4 | 1565 | } |
1da177e4 | 1566 | end: |
f95d39d1 | 1567 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1568 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1569 | return retval; |
1570 | } | |
1571 | ||
67d7ddde | 1572 | static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, |
ce72a16f | 1573 | int options, struct rusage *ru) |
1da177e4 | 1574 | { |
9e8ae01d | 1575 | struct wait_opts wo; |
161550d7 EB |
1576 | struct pid *pid = NULL; |
1577 | enum pid_type type; | |
1da177e4 | 1578 | long ret; |
ba7d25f3 | 1579 | unsigned int f_flags = 0; |
1da177e4 | 1580 | |
91c4e8ea ON |
1581 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1582 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1583 | return -EINVAL; |
1584 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1585 | return -EINVAL; | |
1586 | ||
1587 | switch (which) { | |
1588 | case P_ALL: | |
161550d7 | 1589 | type = PIDTYPE_MAX; |
1da177e4 LT |
1590 | break; |
1591 | case P_PID: | |
161550d7 EB |
1592 | type = PIDTYPE_PID; |
1593 | if (upid <= 0) | |
1da177e4 | 1594 | return -EINVAL; |
3695eae5 CB |
1595 | |
1596 | pid = find_get_pid(upid); | |
1da177e4 LT |
1597 | break; |
1598 | case P_PGID: | |
161550d7 | 1599 | type = PIDTYPE_PGID; |
821cc7b0 | 1600 | if (upid < 0) |
1da177e4 | 1601 | return -EINVAL; |
3695eae5 | 1602 | |
821cc7b0 EB |
1603 | if (upid) |
1604 | pid = find_get_pid(upid); | |
1605 | else | |
1606 | pid = get_task_pid(current, PIDTYPE_PGID); | |
3695eae5 CB |
1607 | break; |
1608 | case P_PIDFD: | |
1609 | type = PIDTYPE_PID; | |
1610 | if (upid < 0) | |
1da177e4 | 1611 | return -EINVAL; |
3695eae5 | 1612 | |
ba7d25f3 | 1613 | pid = pidfd_get_pid(upid, &f_flags); |
3695eae5 CB |
1614 | if (IS_ERR(pid)) |
1615 | return PTR_ERR(pid); | |
ba7d25f3 | 1616 | |
1da177e4 LT |
1617 | break; |
1618 | default: | |
1619 | return -EINVAL; | |
1620 | } | |
1621 | ||
9e8ae01d ON |
1622 | wo.wo_type = type; |
1623 | wo.wo_pid = pid; | |
1624 | wo.wo_flags = options; | |
1625 | wo.wo_info = infop; | |
9e8ae01d | 1626 | wo.wo_rusage = ru; |
ba7d25f3 CB |
1627 | if (f_flags & O_NONBLOCK) |
1628 | wo.wo_flags |= WNOHANG; | |
1629 | ||
9e8ae01d | 1630 | ret = do_wait(&wo); |
ba7d25f3 CB |
1631 | if (!ret && !(options & WNOHANG) && (f_flags & O_NONBLOCK)) |
1632 | ret = -EAGAIN; | |
dfe16dfa | 1633 | |
161550d7 | 1634 | put_pid(pid); |
1da177e4 LT |
1635 | return ret; |
1636 | } | |
1637 | ||
ce72a16f AV |
1638 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1639 | infop, int, options, struct rusage __user *, ru) | |
1640 | { | |
1641 | struct rusage r; | |
67d7ddde AV |
1642 | struct waitid_info info = {.status = 0}; |
1643 | long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL); | |
634a8160 | 1644 | int signo = 0; |
6c85501f | 1645 | |
634a8160 AV |
1646 | if (err > 0) { |
1647 | signo = SIGCHLD; | |
1648 | err = 0; | |
ce72a16f AV |
1649 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) |
1650 | return -EFAULT; | |
1651 | } | |
67d7ddde AV |
1652 | if (!infop) |
1653 | return err; | |
1654 | ||
41cd7805 | 1655 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1656 | return -EFAULT; |
96ca579a | 1657 | |
634a8160 | 1658 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1659 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1660 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1661 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1662 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1663 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1664 | user_write_access_end(); |
ce72a16f | 1665 | return err; |
4c48abe9 | 1666 | Efault: |
41cd7805 | 1667 | user_write_access_end(); |
4c48abe9 | 1668 | return -EFAULT; |
ce72a16f AV |
1669 | } |
1670 | ||
92ebce5a AV |
1671 | long kernel_wait4(pid_t upid, int __user *stat_addr, int options, |
1672 | struct rusage *ru) | |
1da177e4 | 1673 | { |
9e8ae01d | 1674 | struct wait_opts wo; |
161550d7 EB |
1675 | struct pid *pid = NULL; |
1676 | enum pid_type type; | |
1da177e4 LT |
1677 | long ret; |
1678 | ||
1679 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1680 | __WNOTHREAD|__WCLONE|__WALL)) | |
1681 | return -EINVAL; | |
161550d7 | 1682 | |
dd83c161 | 1683 | /* -INT_MIN is not defined */ |
1684 | if (upid == INT_MIN) | |
1685 | return -ESRCH; | |
1686 | ||
161550d7 EB |
1687 | if (upid == -1) |
1688 | type = PIDTYPE_MAX; | |
1689 | else if (upid < 0) { | |
1690 | type = PIDTYPE_PGID; | |
1691 | pid = find_get_pid(-upid); | |
1692 | } else if (upid == 0) { | |
1693 | type = PIDTYPE_PGID; | |
2ae448ef | 1694 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1695 | } else /* upid > 0 */ { |
1696 | type = PIDTYPE_PID; | |
1697 | pid = find_get_pid(upid); | |
1698 | } | |
1699 | ||
9e8ae01d ON |
1700 | wo.wo_type = type; |
1701 | wo.wo_pid = pid; | |
1702 | wo.wo_flags = options | WEXITED; | |
1703 | wo.wo_info = NULL; | |
359566fa | 1704 | wo.wo_stat = 0; |
9e8ae01d ON |
1705 | wo.wo_rusage = ru; |
1706 | ret = do_wait(&wo); | |
161550d7 | 1707 | put_pid(pid); |
359566fa AV |
1708 | if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr)) |
1709 | ret = -EFAULT; | |
1da177e4 | 1710 | |
1da177e4 LT |
1711 | return ret; |
1712 | } | |
1713 | ||
8043fc14 CH |
1714 | int kernel_wait(pid_t pid, int *stat) |
1715 | { | |
1716 | struct wait_opts wo = { | |
1717 | .wo_type = PIDTYPE_PID, | |
1718 | .wo_pid = find_get_pid(pid), | |
1719 | .wo_flags = WEXITED, | |
1720 | }; | |
1721 | int ret; | |
1722 | ||
1723 | ret = do_wait(&wo); | |
1724 | if (ret > 0 && wo.wo_stat) | |
1725 | *stat = wo.wo_stat; | |
1726 | put_pid(wo.wo_pid); | |
1727 | return ret; | |
1728 | } | |
1729 | ||
ce72a16f AV |
1730 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1731 | int, options, struct rusage __user *, ru) | |
1732 | { | |
1733 | struct rusage r; | |
1734 | long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL); | |
1735 | ||
1736 | if (err > 0) { | |
1737 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) | |
1738 | return -EFAULT; | |
1739 | } | |
1740 | return err; | |
1741 | } | |
1742 | ||
1da177e4 LT |
1743 | #ifdef __ARCH_WANT_SYS_WAITPID |
1744 | ||
1745 | /* | |
1746 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1747 | * implemented by calling sys_wait4() from libc.a. | |
1748 | */ | |
17da2bd9 | 1749 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 | 1750 | { |
d300b610 | 1751 | return kernel_wait4(pid, stat_addr, options, NULL); |
1da177e4 LT |
1752 | } |
1753 | ||
1754 | #endif | |
7e95a225 AV |
1755 | |
1756 | #ifdef CONFIG_COMPAT | |
1757 | COMPAT_SYSCALL_DEFINE4(wait4, | |
1758 | compat_pid_t, pid, | |
1759 | compat_uint_t __user *, stat_addr, | |
1760 | int, options, | |
1761 | struct compat_rusage __user *, ru) | |
1762 | { | |
ce72a16f AV |
1763 | struct rusage r; |
1764 | long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL); | |
1765 | if (err > 0) { | |
1766 | if (ru && put_compat_rusage(&r, ru)) | |
1767 | return -EFAULT; | |
7e95a225 | 1768 | } |
ce72a16f | 1769 | return err; |
7e95a225 AV |
1770 | } |
1771 | ||
1772 | COMPAT_SYSCALL_DEFINE5(waitid, | |
1773 | int, which, compat_pid_t, pid, | |
1774 | struct compat_siginfo __user *, infop, int, options, | |
1775 | struct compat_rusage __user *, uru) | |
1776 | { | |
7e95a225 | 1777 | struct rusage ru; |
67d7ddde AV |
1778 | struct waitid_info info = {.status = 0}; |
1779 | long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL); | |
634a8160 AV |
1780 | int signo = 0; |
1781 | if (err > 0) { | |
1782 | signo = SIGCHLD; | |
1783 | err = 0; | |
6c85501f AV |
1784 | if (uru) { |
1785 | /* kernel_waitid() overwrites everything in ru */ | |
1786 | if (COMPAT_USE_64BIT_TIME) | |
1787 | err = copy_to_user(uru, &ru, sizeof(ru)); | |
1788 | else | |
1789 | err = put_compat_rusage(&ru, uru); | |
1790 | if (err) | |
1791 | return -EFAULT; | |
1792 | } | |
7e95a225 AV |
1793 | } |
1794 | ||
4c48abe9 AV |
1795 | if (!infop) |
1796 | return err; | |
1797 | ||
41cd7805 | 1798 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1799 | return -EFAULT; |
96ca579a | 1800 | |
634a8160 | 1801 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1802 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1803 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1804 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1805 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1806 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1807 | user_write_access_end(); |
67d7ddde | 1808 | return err; |
4c48abe9 | 1809 | Efault: |
41cd7805 | 1810 | user_write_access_end(); |
4c48abe9 | 1811 | return -EFAULT; |
7e95a225 AV |
1812 | } |
1813 | #endif | |
7c2c11b2 | 1814 | |
38fd525a EB |
1815 | /** |
1816 | * thread_group_exited - check that a thread group has exited | |
1817 | * @pid: tgid of thread group to be checked. | |
1818 | * | |
1819 | * Test if the thread group represented by tgid has exited (all | |
1820 | * threads are zombies, dead or completely gone). | |
1821 | * | |
1822 | * Return: true if the thread group has exited. false otherwise. | |
1823 | */ | |
1824 | bool thread_group_exited(struct pid *pid) | |
1825 | { | |
1826 | struct task_struct *task; | |
1827 | bool exited; | |
1828 | ||
1829 | rcu_read_lock(); | |
1830 | task = pid_task(pid, PIDTYPE_PID); | |
1831 | exited = !task || | |
1832 | (READ_ONCE(task->exit_state) && thread_group_empty(task)); | |
1833 | rcu_read_unlock(); | |
1834 | ||
1835 | return exited; | |
1836 | } | |
1837 | EXPORT_SYMBOL(thread_group_exited); | |
1838 | ||
7c2c11b2 SM |
1839 | __weak void abort(void) |
1840 | { | |
1841 | BUG(); | |
1842 | ||
1843 | /* if that doesn't kill us, halt */ | |
1844 | panic("Oops failed to kill thread"); | |
1845 | } | |
dc8635b7 | 1846 | EXPORT_SYMBOL(abort); |