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> |
30199f5a | 52 | #include <linux/tracehook.h> |
5ad4e53b | 53 | #include <linux/fs_struct.h> |
d84f4f99 | 54 | #include <linux/init_task.h> |
cdd6c482 | 55 | #include <linux/perf_event.h> |
ad8d75ff | 56 | #include <trace/events/sched.h> |
24f1e32c | 57 | #include <linux/hw_breakpoint.h> |
3d5992d2 | 58 | #include <linux/oom.h> |
54848d73 | 59 | #include <linux/writeback.h> |
40401530 | 60 | #include <linux/shm.h> |
5c9a8750 | 61 | #include <linux/kcov.h> |
53d3eaa3 | 62 | #include <linux/random.h> |
8f95c90c | 63 | #include <linux/rcuwait.h> |
7e95a225 | 64 | #include <linux/compat.h> |
b1b6b5a3 | 65 | #include <linux/io_uring.h> |
670721c7 | 66 | #include <linux/kprobes.h> |
1da177e4 | 67 | |
7c0f6ba6 | 68 | #include <linux/uaccess.h> |
1da177e4 | 69 | #include <asm/unistd.h> |
1da177e4 LT |
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; |
3f649ab7 | 97 | struct tty_struct *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) |
60700e38 | 119 | wake_up_process(sig->group_exec_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 | ||
670721c7 | 171 | kprobe_flush_task(tsk); |
4e231c79 | 172 | perf_event_delayed_put(tsk); |
0a16b607 MD |
173 | trace_sched_process_free(tsk); |
174 | put_task_struct(tsk); | |
8c7904a0 EB |
175 | } |
176 | ||
3fbd7ee2 EB |
177 | void put_task_struct_rcu_user(struct task_struct *task) |
178 | { | |
179 | if (refcount_dec_and_test(&task->rcu_users)) | |
180 | call_rcu(&task->rcu, delayed_put_task_struct); | |
181 | } | |
f470021a | 182 | |
a0be55de | 183 | void release_task(struct task_struct *p) |
1da177e4 | 184 | { |
36c8b586 | 185 | struct task_struct *leader; |
7bc3e6e5 | 186 | struct pid *thread_pid; |
1da177e4 | 187 | int zap_leader; |
1f09f974 | 188 | repeat: |
c69e8d9c | 189 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
190 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
191 | rcu_read_lock(); | |
21d1c5e3 | 192 | dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1); |
d11c563d | 193 | rcu_read_unlock(); |
c69e8d9c | 194 | |
6b115bf5 | 195 | cgroup_release(p); |
0203026b | 196 | |
1da177e4 | 197 | write_lock_irq(&tasklist_lock); |
a288eecc | 198 | ptrace_release_task(p); |
7bc3e6e5 | 199 | thread_pid = get_pid(p->thread_pid); |
1da177e4 | 200 | __exit_signal(p); |
35f5cad8 | 201 | |
1da177e4 LT |
202 | /* |
203 | * If we are the last non-leader member of the thread | |
204 | * group, and the leader is zombie, then notify the | |
205 | * group leader's parent process. (if it wants notification.) | |
206 | */ | |
207 | zap_leader = 0; | |
208 | leader = p->group_leader; | |
a0be55de IA |
209 | if (leader != p && thread_group_empty(leader) |
210 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
211 | /* |
212 | * If we were the last child thread and the leader has | |
213 | * exited already, and the leader's parent ignores SIGCHLD, | |
214 | * then we are the one who should release the leader. | |
dae33574 | 215 | */ |
86773473 | 216 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
217 | if (zap_leader) |
218 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
219 | } |
220 | ||
1da177e4 | 221 | write_unlock_irq(&tasklist_lock); |
3a15fb6e | 222 | seccomp_filter_release(p); |
7bc3e6e5 | 223 | proc_flush_pid(thread_pid); |
6ade99ec | 224 | put_pid(thread_pid); |
1da177e4 | 225 | release_thread(p); |
3fbd7ee2 | 226 | put_task_struct_rcu_user(p); |
1da177e4 LT |
227 | |
228 | p = leader; | |
229 | if (unlikely(zap_leader)) | |
230 | goto repeat; | |
231 | } | |
232 | ||
9d9a6ebf | 233 | int rcuwait_wake_up(struct rcuwait *w) |
8f95c90c | 234 | { |
9d9a6ebf | 235 | int ret = 0; |
8f95c90c DB |
236 | struct task_struct *task; |
237 | ||
238 | rcu_read_lock(); | |
239 | ||
240 | /* | |
241 | * Order condition vs @task, such that everything prior to the load | |
242 | * of @task is visible. This is the condition as to why the user called | |
c9d64a1b | 243 | * rcuwait_wake() in the first place. Pairs with set_current_state() |
8f95c90c DB |
244 | * barrier (A) in rcuwait_wait_event(). |
245 | * | |
246 | * WAIT WAKE | |
247 | * [S] tsk = current [S] cond = true | |
248 | * MB (A) MB (B) | |
249 | * [L] cond [L] tsk | |
250 | */ | |
6dc080ee | 251 | smp_mb(); /* (B) */ |
8f95c90c | 252 | |
8f95c90c DB |
253 | task = rcu_dereference(w->task); |
254 | if (task) | |
9d9a6ebf | 255 | ret = wake_up_process(task); |
8f95c90c | 256 | rcu_read_unlock(); |
9d9a6ebf DB |
257 | |
258 | return ret; | |
8f95c90c | 259 | } |
ac8dec42 | 260 | EXPORT_SYMBOL_GPL(rcuwait_wake_up); |
8f95c90c | 261 | |
1da177e4 LT |
262 | /* |
263 | * Determine if a process group is "orphaned", according to the POSIX | |
264 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
265 | * by terminal-generated stop signals. Newly orphaned process groups are | |
266 | * to receive a SIGHUP and a SIGCONT. | |
267 | * | |
268 | * "I ask you, have you ever known what it is to be an orphan?" | |
269 | */ | |
a0be55de IA |
270 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
271 | struct task_struct *ignored_task) | |
1da177e4 LT |
272 | { |
273 | struct task_struct *p; | |
1da177e4 | 274 | |
0475ac08 | 275 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
276 | if ((p == ignored_task) || |
277 | (p->exit_state && thread_group_empty(p)) || | |
278 | is_global_init(p->real_parent)) | |
1da177e4 | 279 | continue; |
05e83df6 | 280 | |
0475ac08 | 281 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
282 | task_session(p->real_parent) == task_session(p)) |
283 | return 0; | |
0475ac08 | 284 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
285 | |
286 | return 1; | |
1da177e4 LT |
287 | } |
288 | ||
3e7cd6c4 | 289 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
290 | { |
291 | int retval; | |
292 | ||
293 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 294 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
295 | read_unlock(&tasklist_lock); |
296 | ||
297 | return retval; | |
298 | } | |
299 | ||
961c4675 | 300 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 301 | { |
1da177e4 LT |
302 | struct task_struct *p; |
303 | ||
0475ac08 | 304 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
305 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
306 | return true; | |
0475ac08 | 307 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
308 | |
309 | return false; | |
1da177e4 LT |
310 | } |
311 | ||
f49ee505 ON |
312 | /* |
313 | * Check to see if any process groups have become orphaned as | |
314 | * a result of our exiting, and if they have any stopped jobs, | |
315 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
316 | */ | |
317 | static void | |
318 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
319 | { | |
320 | struct pid *pgrp = task_pgrp(tsk); | |
321 | struct task_struct *ignored_task = tsk; | |
322 | ||
323 | if (!parent) | |
a0be55de IA |
324 | /* exit: our father is in a different pgrp than |
325 | * we are and we were the only connection outside. | |
326 | */ | |
f49ee505 ON |
327 | parent = tsk->real_parent; |
328 | else | |
329 | /* reparent: our child is in a different pgrp than | |
330 | * we are, and it was the only connection outside. | |
331 | */ | |
332 | ignored_task = NULL; | |
333 | ||
334 | if (task_pgrp(parent) != pgrp && | |
335 | task_session(parent) == task_session(tsk) && | |
336 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
337 | has_stopped_jobs(pgrp)) { | |
338 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
339 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
340 | } | |
341 | } | |
342 | ||
92307383 | 343 | static void coredump_task_exit(struct task_struct *tsk) |
d67e03e3 EB |
344 | { |
345 | struct core_state *core_state; | |
346 | ||
347 | /* | |
348 | * Serialize with any possible pending coredump. | |
0258b5fd | 349 | * We must hold siglock around checking core_state |
92307383 | 350 | * and setting PF_POSTCOREDUMP. The core-inducing thread |
d67e03e3 | 351 | * will increment ->nr_threads for each thread in the |
92307383 | 352 | * group without PF_POSTCOREDUMP set. |
d67e03e3 | 353 | */ |
0258b5fd | 354 | spin_lock_irq(&tsk->sighand->siglock); |
92307383 | 355 | tsk->flags |= PF_POSTCOREDUMP; |
0258b5fd EB |
356 | core_state = tsk->signal->core_state; |
357 | spin_unlock_irq(&tsk->sighand->siglock); | |
d67e03e3 EB |
358 | if (core_state) { |
359 | struct core_thread self; | |
360 | ||
d67e03e3 EB |
361 | self.task = current; |
362 | if (self.task->flags & PF_SIGNALED) | |
363 | self.next = xchg(&core_state->dumper.next, &self); | |
364 | else | |
365 | self.task = NULL; | |
366 | /* | |
367 | * Implies mb(), the result of xchg() must be visible | |
368 | * to core_state->dumper. | |
369 | */ | |
370 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
371 | complete(&core_state->startup); | |
372 | ||
373 | for (;;) { | |
374 | set_current_state(TASK_UNINTERRUPTIBLE); | |
375 | if (!self.task) /* see coredump_finish() */ | |
376 | break; | |
377 | freezable_schedule(); | |
378 | } | |
379 | __set_current_state(TASK_RUNNING); | |
d67e03e3 EB |
380 | } |
381 | } | |
382 | ||
f98bafa0 | 383 | #ifdef CONFIG_MEMCG |
cf475ad2 | 384 | /* |
733eda7a | 385 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 386 | */ |
cf475ad2 BS |
387 | void mm_update_next_owner(struct mm_struct *mm) |
388 | { | |
389 | struct task_struct *c, *g, *p = current; | |
390 | ||
391 | retry: | |
733eda7a KH |
392 | /* |
393 | * If the exiting or execing task is not the owner, it's | |
394 | * someone else's problem. | |
395 | */ | |
396 | if (mm->owner != p) | |
cf475ad2 | 397 | return; |
733eda7a KH |
398 | /* |
399 | * The current owner is exiting/execing and there are no other | |
400 | * candidates. Do not leave the mm pointing to a possibly | |
401 | * freed task structure. | |
402 | */ | |
403 | if (atomic_read(&mm->mm_users) <= 1) { | |
987717e5 | 404 | WRITE_ONCE(mm->owner, NULL); |
733eda7a KH |
405 | return; |
406 | } | |
cf475ad2 BS |
407 | |
408 | read_lock(&tasklist_lock); | |
409 | /* | |
410 | * Search in the children | |
411 | */ | |
412 | list_for_each_entry(c, &p->children, sibling) { | |
413 | if (c->mm == mm) | |
414 | goto assign_new_owner; | |
415 | } | |
416 | ||
417 | /* | |
418 | * Search in the siblings | |
419 | */ | |
dea33cfd | 420 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
421 | if (c->mm == mm) |
422 | goto assign_new_owner; | |
423 | } | |
424 | ||
425 | /* | |
f87fb599 | 426 | * Search through everything else, we should not get here often. |
cf475ad2 | 427 | */ |
39af1765 ON |
428 | for_each_process(g) { |
429 | if (g->flags & PF_KTHREAD) | |
430 | continue; | |
431 | for_each_thread(g, c) { | |
432 | if (c->mm == mm) | |
433 | goto assign_new_owner; | |
434 | if (c->mm) | |
435 | break; | |
436 | } | |
f87fb599 | 437 | } |
cf475ad2 | 438 | read_unlock(&tasklist_lock); |
31a78f23 BS |
439 | /* |
440 | * We found no owner yet mm_users > 1: this implies that we are | |
441 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 442 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 443 | */ |
987717e5 | 444 | WRITE_ONCE(mm->owner, NULL); |
cf475ad2 BS |
445 | return; |
446 | ||
447 | assign_new_owner: | |
448 | BUG_ON(c == p); | |
449 | get_task_struct(c); | |
450 | /* | |
451 | * The task_lock protects c->mm from changing. | |
452 | * We always want mm->owner->mm == mm | |
453 | */ | |
454 | task_lock(c); | |
e5991371 HD |
455 | /* |
456 | * Delay read_unlock() till we have the task_lock() | |
457 | * to ensure that c does not slip away underneath us | |
458 | */ | |
459 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
460 | if (c->mm != mm) { |
461 | task_unlock(c); | |
462 | put_task_struct(c); | |
463 | goto retry; | |
464 | } | |
987717e5 | 465 | WRITE_ONCE(mm->owner, c); |
cf475ad2 BS |
466 | task_unlock(c); |
467 | put_task_struct(c); | |
468 | } | |
f98bafa0 | 469 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 470 | |
1da177e4 LT |
471 | /* |
472 | * Turn us into a lazy TLB process if we | |
473 | * aren't already.. | |
474 | */ | |
0039962a | 475 | static void exit_mm(void) |
1da177e4 | 476 | { |
0039962a | 477 | struct mm_struct *mm = current->mm; |
1da177e4 | 478 | |
4610ba7a | 479 | exit_mm_release(current, mm); |
1da177e4 LT |
480 | if (!mm) |
481 | return; | |
4fe7efdb | 482 | sync_mm_rss(mm); |
d8ed45c5 | 483 | mmap_read_lock(mm); |
f1f10076 | 484 | mmgrab(mm); |
0039962a | 485 | BUG_ON(mm != current->active_mm); |
1da177e4 | 486 | /* more a memory barrier than a real lock */ |
0039962a | 487 | task_lock(current); |
5bc78502 MD |
488 | /* |
489 | * When a thread stops operating on an address space, the loop | |
490 | * in membarrier_private_expedited() may not observe that | |
491 | * tsk->mm, and the loop in membarrier_global_expedited() may | |
492 | * not observe a MEMBARRIER_STATE_GLOBAL_EXPEDITED | |
493 | * rq->membarrier_state, so those would not issue an IPI. | |
494 | * Membarrier requires a memory barrier after accessing | |
495 | * user-space memory, before clearing tsk->mm or the | |
496 | * rq->membarrier_state. | |
497 | */ | |
498 | smp_mb__after_spinlock(); | |
499 | local_irq_disable(); | |
0039962a | 500 | current->mm = NULL; |
5bc78502 | 501 | membarrier_update_current_mm(NULL); |
1da177e4 | 502 | enter_lazy_tlb(mm, current); |
5bc78502 | 503 | local_irq_enable(); |
0039962a | 504 | task_unlock(current); |
5bc78502 | 505 | mmap_read_unlock(mm); |
cf475ad2 | 506 | mm_update_next_owner(mm); |
1da177e4 | 507 | mmput(mm); |
c32b3cbe | 508 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 509 | exit_oom_victim(); |
1da177e4 LT |
510 | } |
511 | ||
c9dc05bf ON |
512 | static struct task_struct *find_alive_thread(struct task_struct *p) |
513 | { | |
514 | struct task_struct *t; | |
515 | ||
516 | for_each_thread(p, t) { | |
517 | if (!(t->flags & PF_EXITING)) | |
518 | return t; | |
519 | } | |
520 | return NULL; | |
521 | } | |
522 | ||
8fb335e0 AV |
523 | static struct task_struct *find_child_reaper(struct task_struct *father, |
524 | struct list_head *dead) | |
1109909c ON |
525 | __releases(&tasklist_lock) |
526 | __acquires(&tasklist_lock) | |
527 | { | |
528 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
529 | struct task_struct *reaper = pid_ns->child_reaper; | |
8fb335e0 | 530 | struct task_struct *p, *n; |
1109909c ON |
531 | |
532 | if (likely(reaper != father)) | |
533 | return reaper; | |
534 | ||
c9dc05bf ON |
535 | reaper = find_alive_thread(father); |
536 | if (reaper) { | |
1109909c ON |
537 | pid_ns->child_reaper = reaper; |
538 | return reaper; | |
539 | } | |
540 | ||
541 | write_unlock_irq(&tasklist_lock); | |
8fb335e0 AV |
542 | |
543 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
544 | list_del_init(&p->ptrace_entry); | |
545 | release_task(p); | |
546 | } | |
547 | ||
1109909c ON |
548 | zap_pid_ns_processes(pid_ns); |
549 | write_lock_irq(&tasklist_lock); | |
550 | ||
551 | return father; | |
552 | } | |
553 | ||
1da177e4 | 554 | /* |
ebec18a6 LP |
555 | * When we die, we re-parent all our children, and try to: |
556 | * 1. give them to another thread in our thread group, if such a member exists | |
557 | * 2. give it to the first ancestor process which prctl'd itself as a | |
558 | * child_subreaper for its children (like a service manager) | |
559 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 560 | */ |
1109909c ON |
561 | static struct task_struct *find_new_reaper(struct task_struct *father, |
562 | struct task_struct *child_reaper) | |
1da177e4 | 563 | { |
c9dc05bf | 564 | struct task_struct *thread, *reaper; |
1da177e4 | 565 | |
c9dc05bf ON |
566 | thread = find_alive_thread(father); |
567 | if (thread) | |
950bbabb | 568 | return thread; |
1da177e4 | 569 | |
7d24e2df | 570 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 571 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 572 | /* |
175aed3f | 573 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
574 | * We can't check reaper != child_reaper to ensure we do not |
575 | * cross the namespaces, the exiting parent could be injected | |
576 | * by setns() + fork(). | |
577 | * We check pid->level, this is slightly more efficient than | |
578 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 579 | */ |
c6c70f44 ON |
580 | for (reaper = father->real_parent; |
581 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 582 | reaper = reaper->real_parent) { |
175aed3f | 583 | if (reaper == &init_task) |
ebec18a6 LP |
584 | break; |
585 | if (!reaper->signal->is_child_subreaper) | |
586 | continue; | |
c9dc05bf ON |
587 | thread = find_alive_thread(reaper); |
588 | if (thread) | |
589 | return thread; | |
ebec18a6 | 590 | } |
1da177e4 | 591 | } |
762a24be | 592 | |
1109909c | 593 | return child_reaper; |
950bbabb ON |
594 | } |
595 | ||
5dfc80be ON |
596 | /* |
597 | * Any that need to be release_task'd are put on the @dead list. | |
598 | */ | |
9cd80bbb | 599 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
600 | struct list_head *dead) |
601 | { | |
2831096e | 602 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
603 | return; |
604 | ||
abd50b39 | 605 | /* We don't want people slaying init. */ |
5dfc80be ON |
606 | p->exit_signal = SIGCHLD; |
607 | ||
608 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 609 | if (!p->ptrace && |
5dfc80be | 610 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 611 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 612 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 613 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
614 | } |
615 | } | |
616 | ||
617 | kill_orphaned_pgrp(p, father); | |
618 | } | |
619 | ||
482a3767 ON |
620 | /* |
621 | * This does two things: | |
622 | * | |
623 | * A. Make init inherit all the child processes | |
624 | * B. Check to see if any process groups have become orphaned | |
625 | * as a result of our exiting, and if they have any stopped | |
626 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
627 | */ | |
628 | static void forget_original_parent(struct task_struct *father, | |
629 | struct list_head *dead) | |
1da177e4 | 630 | { |
482a3767 | 631 | struct task_struct *p, *t, *reaper; |
762a24be | 632 | |
7c8bd232 | 633 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 634 | exit_ptrace(father, dead); |
f470021a | 635 | |
7c8bd232 | 636 | /* Can drop and reacquire tasklist_lock */ |
8fb335e0 | 637 | reaper = find_child_reaper(father, dead); |
ad9e206a | 638 | if (list_empty(&father->children)) |
482a3767 | 639 | return; |
1109909c ON |
640 | |
641 | reaper = find_new_reaper(father, reaper); | |
2831096e | 642 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 643 | for_each_thread(p, t) { |
22a34c6f MB |
644 | RCU_INIT_POINTER(t->real_parent, reaper); |
645 | BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father)); | |
57a05918 | 646 | if (likely(!t->ptrace)) |
9cd80bbb | 647 | t->parent = t->real_parent; |
9cd80bbb ON |
648 | if (t->pdeath_signal) |
649 | group_send_sig_info(t->pdeath_signal, | |
01024980 EB |
650 | SEND_SIG_NOINFO, t, |
651 | PIDTYPE_TGID); | |
57a05918 | 652 | } |
2831096e ON |
653 | /* |
654 | * If this is a threaded reparent there is no need to | |
655 | * notify anyone anything has happened. | |
656 | */ | |
657 | if (!same_thread_group(reaper, father)) | |
482a3767 | 658 | reparent_leader(father, p, dead); |
1da177e4 | 659 | } |
2831096e | 660 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
661 | } |
662 | ||
663 | /* | |
664 | * Send signals to all our closest relatives so that they know | |
665 | * to properly mourn us.. | |
666 | */ | |
821c7de7 | 667 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 668 | { |
53c8f9f1 | 669 | bool autoreap; |
482a3767 ON |
670 | struct task_struct *p, *n; |
671 | LIST_HEAD(dead); | |
1da177e4 | 672 | |
762a24be | 673 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
674 | forget_original_parent(tsk, &dead); |
675 | ||
821c7de7 ON |
676 | if (group_dead) |
677 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 678 | |
b191d649 | 679 | tsk->exit_state = EXIT_ZOMBIE; |
45cdf5cc ON |
680 | if (unlikely(tsk->ptrace)) { |
681 | int sig = thread_group_leader(tsk) && | |
682 | thread_group_empty(tsk) && | |
683 | !ptrace_reparented(tsk) ? | |
684 | tsk->exit_signal : SIGCHLD; | |
685 | autoreap = do_notify_parent(tsk, sig); | |
686 | } else if (thread_group_leader(tsk)) { | |
687 | autoreap = thread_group_empty(tsk) && | |
688 | do_notify_parent(tsk, tsk->exit_signal); | |
689 | } else { | |
690 | autoreap = true; | |
691 | } | |
1da177e4 | 692 | |
30b692d3 CB |
693 | if (autoreap) { |
694 | tsk->exit_state = EXIT_DEAD; | |
6c66e7db | 695 | list_add(&tsk->ptrace_entry, &dead); |
30b692d3 | 696 | } |
1da177e4 | 697 | |
9c339168 ON |
698 | /* mt-exec, de_thread() is waiting for group leader */ |
699 | if (unlikely(tsk->signal->notify_count < 0)) | |
60700e38 | 700 | wake_up_process(tsk->signal->group_exec_task); |
1da177e4 LT |
701 | write_unlock_irq(&tasklist_lock); |
702 | ||
482a3767 ON |
703 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
704 | list_del_init(&p->ptrace_entry); | |
705 | release_task(p); | |
706 | } | |
1da177e4 LT |
707 | } |
708 | ||
e18eecb8 JD |
709 | #ifdef CONFIG_DEBUG_STACK_USAGE |
710 | static void check_stack_usage(void) | |
711 | { | |
712 | static DEFINE_SPINLOCK(low_water_lock); | |
713 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
714 | unsigned long free; |
715 | ||
7c9f8861 | 716 | free = stack_not_used(current); |
e18eecb8 JD |
717 | |
718 | if (free >= lowest_to_date) | |
719 | return; | |
720 | ||
721 | spin_lock(&low_water_lock); | |
722 | if (free < lowest_to_date) { | |
627393d4 | 723 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 724 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
725 | lowest_to_date = free; |
726 | } | |
727 | spin_unlock(&low_water_lock); | |
728 | } | |
729 | #else | |
730 | static inline void check_stack_usage(void) {} | |
731 | #endif | |
732 | ||
9af6528e | 733 | void __noreturn do_exit(long code) |
1da177e4 LT |
734 | { |
735 | struct task_struct *tsk = current; | |
736 | int group_dead; | |
737 | ||
73c10101 | 738 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 739 | |
de77c3a5 EB |
740 | /* |
741 | * If do_dead is called because this processes oopsed, it's possible | |
742 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
743 | * continuing. Amongst other possible reasons, this is to prevent | |
744 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
745 | * kernel address. | |
746 | * | |
747 | * On uptodate architectures force_uaccess_begin is a noop. On | |
748 | * architectures that still have set_fs/get_fs in addition to handling | |
749 | * oopses handles kernel threads that run as set_fs(KERNEL_DS) by | |
750 | * default. | |
751 | */ | |
752 | force_uaccess_begin(); | |
753 | ||
586b58ca JH |
754 | kcov_task_exit(tsk); |
755 | ||
92307383 | 756 | coredump_task_exit(tsk); |
a288eecc | 757 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 758 | |
e0e81739 DH |
759 | validate_creds_for_do_exit(tsk); |
760 | ||
f552a27a | 761 | io_uring_files_cancel(); |
d12619b5 | 762 | exit_signals(tsk); /* sets PF_EXITING */ |
1da177e4 | 763 | |
48d212a2 LT |
764 | /* sync mm's RSS info before statistics gathering */ |
765 | if (tsk->mm) | |
766 | sync_mm_rss(tsk->mm); | |
51229b49 | 767 | acct_update_integrals(tsk); |
1da177e4 | 768 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 769 | if (group_dead) { |
43cf75d9 | 770 | /* |
771 | * If the last thread of global init has exited, panic | |
772 | * immediately to get a useable coredump. | |
773 | */ | |
774 | if (unlikely(is_global_init(tsk))) | |
775 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
776 | tsk->signal->group_exit_code ?: (int)code); | |
777 | ||
baa73d9e | 778 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 779 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 780 | exit_itimers(tsk->signal); |
baa73d9e | 781 | #endif |
1f10206c JP |
782 | if (tsk->mm) |
783 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 784 | } |
f6ec29a4 | 785 | acct_collect(code, group_dead); |
522ed776 MT |
786 | if (group_dead) |
787 | tty_audit_exit(); | |
a4ff8dba | 788 | audit_free(tsk); |
115085ea | 789 | |
48d212a2 | 790 | tsk->exit_code = code; |
115085ea | 791 | taskstats_exit(tsk, group_dead); |
c757249a | 792 | |
0039962a | 793 | exit_mm(); |
1da177e4 | 794 | |
0e464814 | 795 | if (group_dead) |
f6ec29a4 | 796 | acct_process(); |
0a16b607 MD |
797 | trace_sched_process_exit(tsk); |
798 | ||
1da177e4 | 799 | exit_sem(tsk); |
b34a6b1d | 800 | exit_shm(tsk); |
1ec7f1dd AV |
801 | exit_files(tsk); |
802 | exit_fs(tsk); | |
c39df5fa ON |
803 | if (group_dead) |
804 | disassociate_ctty(1); | |
8aac6270 | 805 | exit_task_namespaces(tsk); |
ed3e694d | 806 | exit_task_work(tsk); |
e6464694 | 807 | exit_thread(tsk); |
0b3fcf17 SE |
808 | |
809 | /* | |
810 | * Flush inherited counters to the parent - before the parent | |
811 | * gets woken up by child-exit notifications. | |
812 | * | |
813 | * because of cgroup mode, must be called before cgroup_exit() | |
814 | */ | |
815 | perf_event_exit_task(tsk); | |
816 | ||
8e5bfa8c | 817 | sched_autogroup_exit_task(tsk); |
1ec41830 | 818 | cgroup_exit(tsk); |
1da177e4 | 819 | |
24f1e32c FW |
820 | /* |
821 | * FIXME: do that only when needed, using sched_exit tracepoint | |
822 | */ | |
7c8df286 | 823 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 824 | |
ccdd29ff | 825 | exit_tasks_rcu_start(); |
821c7de7 | 826 | exit_notify(tsk, group_dead); |
ef982393 | 827 | proc_exit_connector(tsk); |
c11600e4 | 828 | mpol_put_task_policy(tsk); |
42b2dd0a | 829 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
830 | if (unlikely(current->pi_state_cache)) |
831 | kfree(current->pi_state_cache); | |
42b2dd0a | 832 | #endif |
de5097c2 | 833 | /* |
9a11b49a | 834 | * Make sure we are holding no locks: |
de5097c2 | 835 | */ |
1b1d2fb4 | 836 | debug_check_no_locks_held(); |
1da177e4 | 837 | |
afc847b7 | 838 | if (tsk->io_context) |
b69f2292 | 839 | exit_io_context(tsk); |
afc847b7 | 840 | |
b92ce558 | 841 | if (tsk->splice_pipe) |
4b8a8f1e | 842 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 843 | |
5640f768 ED |
844 | if (tsk->task_frag.page) |
845 | put_page(tsk->task_frag.page); | |
846 | ||
e0e81739 DH |
847 | validate_creds_for_do_exit(tsk); |
848 | ||
4bcb8232 | 849 | check_stack_usage(); |
7407251a | 850 | preempt_disable(); |
54848d73 WF |
851 | if (tsk->nr_dirtied) |
852 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 853 | exit_rcu(); |
ccdd29ff | 854 | exit_tasks_rcu_finish(); |
b5740f4b | 855 | |
b09be676 | 856 | lockdep_free_task(tsk); |
9af6528e | 857 | do_task_dead(); |
1da177e4 | 858 | } |
012914da | 859 | |
0e25498f EB |
860 | void __noreturn make_task_dead(int signr) |
861 | { | |
862 | /* | |
863 | * Take the task off the cpu after something catastrophic has | |
864 | * happened. | |
05ea0424 EB |
865 | * |
866 | * We can get here from a kernel oops, sometimes with preemption off. | |
867 | * Start by checking for critical errors. | |
868 | * Then fix up important state like USER_DS and preemption. | |
869 | * Then do everything else. | |
0e25498f | 870 | */ |
05ea0424 EB |
871 | struct task_struct *tsk = current; |
872 | ||
873 | if (unlikely(in_interrupt())) | |
874 | panic("Aiee, killing interrupt handler!"); | |
875 | if (unlikely(!tsk->pid)) | |
876 | panic("Attempted to kill the idle task!"); | |
877 | ||
05ea0424 EB |
878 | if (unlikely(in_atomic())) { |
879 | pr_info("note: %s[%d] exited with preempt_count %d\n", | |
880 | current->comm, task_pid_nr(current), | |
881 | preempt_count()); | |
882 | preempt_count_set(PREEMPT_ENABLED); | |
883 | } | |
884 | ||
885 | /* | |
886 | * We're taking recursive faults here in make_task_dead. Safest is to just | |
887 | * leave this task alone and wait for reboot. | |
888 | */ | |
889 | if (unlikely(tsk->flags & PF_EXITING)) { | |
890 | pr_alert("Fixing recursive fault but reboot is needed!\n"); | |
891 | futex_exit_recursive(tsk); | |
912616f1 EB |
892 | tsk->exit_state = EXIT_DEAD; |
893 | refcount_inc(&tsk->rcu_users); | |
7f80a2fd | 894 | do_task_dead(); |
05ea0424 EB |
895 | } |
896 | ||
0e25498f EB |
897 | do_exit(signr); |
898 | } | |
899 | ||
754fe8d2 | 900 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
901 | { |
902 | do_exit((error_code&0xff)<<8); | |
903 | } | |
904 | ||
1da177e4 LT |
905 | /* |
906 | * Take down every thread in the group. This is called by fatal signals | |
907 | * as well as by sys_exit_group (below). | |
908 | */ | |
9402c95f | 909 | void |
1da177e4 LT |
910 | do_group_exit(int exit_code) |
911 | { | |
bfc4b089 ON |
912 | struct signal_struct *sig = current->signal; |
913 | ||
49697335 | 914 | if (sig->flags & SIGNAL_GROUP_EXIT) |
bfc4b089 | 915 | exit_code = sig->group_exit_code; |
49697335 EB |
916 | else if (sig->group_exec_task) |
917 | exit_code = 0; | |
1da177e4 | 918 | else if (!thread_group_empty(current)) { |
1da177e4 | 919 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 920 | |
1da177e4 | 921 | spin_lock_irq(&sighand->siglock); |
49697335 | 922 | if (sig->flags & SIGNAL_GROUP_EXIT) |
1da177e4 LT |
923 | /* Another thread got here before we took the lock. */ |
924 | exit_code = sig->group_exit_code; | |
49697335 EB |
925 | else if (sig->group_exec_task) |
926 | exit_code = 0; | |
1da177e4 | 927 | else { |
1da177e4 | 928 | sig->group_exit_code = exit_code; |
ed5d2cac | 929 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
930 | zap_other_threads(current); |
931 | } | |
932 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
933 | } |
934 | ||
935 | do_exit(exit_code); | |
936 | /* NOTREACHED */ | |
937 | } | |
938 | ||
939 | /* | |
940 | * this kills every thread in the thread group. Note that any externally | |
941 | * wait4()-ing process will get the correct exit code - even if this | |
942 | * thread is not the thread group leader. | |
943 | */ | |
754fe8d2 | 944 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
945 | { |
946 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
947 | /* NOTREACHED */ |
948 | return 0; | |
1da177e4 LT |
949 | } |
950 | ||
67d7ddde AV |
951 | struct waitid_info { |
952 | pid_t pid; | |
953 | uid_t uid; | |
954 | int status; | |
955 | int cause; | |
956 | }; | |
957 | ||
9e8ae01d ON |
958 | struct wait_opts { |
959 | enum pid_type wo_type; | |
9e8ae01d | 960 | int wo_flags; |
e1eb1ebc | 961 | struct pid *wo_pid; |
9e8ae01d | 962 | |
67d7ddde | 963 | struct waitid_info *wo_info; |
359566fa | 964 | int wo_stat; |
ce72a16f | 965 | struct rusage *wo_rusage; |
9e8ae01d | 966 | |
ac6424b9 | 967 | wait_queue_entry_t child_wait; |
9e8ae01d ON |
968 | int notask_error; |
969 | }; | |
970 | ||
989264f4 | 971 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 972 | { |
5c01ba49 ON |
973 | return wo->wo_type == PIDTYPE_MAX || |
974 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
975 | } | |
1da177e4 | 976 | |
bf959931 ON |
977 | static int |
978 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
979 | { |
980 | if (!eligible_pid(wo, p)) | |
981 | return 0; | |
bf959931 ON |
982 | |
983 | /* | |
984 | * Wait for all children (clone and not) if __WALL is set or | |
985 | * if it is traced by us. | |
986 | */ | |
987 | if (ptrace || (wo->wo_flags & __WALL)) | |
988 | return 1; | |
989 | ||
990 | /* | |
991 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
992 | * otherwise, wait for non-clone children *only*. | |
993 | * | |
994 | * Note: a "clone" child here is one that reports to its parent | |
995 | * using a signal other than SIGCHLD, or a non-leader thread which | |
996 | * we can only see if it is traced by us. | |
997 | */ | |
998 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 999 | return 0; |
1da177e4 | 1000 | |
14dd0b81 | 1001 | return 1; |
1da177e4 LT |
1002 | } |
1003 | ||
1da177e4 LT |
1004 | /* |
1005 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1006 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1007 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1008 | * released the lock and the system call should return. | |
1009 | */ | |
9e8ae01d | 1010 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1011 | { |
67d7ddde | 1012 | int state, status; |
6c5f3e7b | 1013 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1014 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
67d7ddde | 1015 | struct waitid_info *infop; |
1da177e4 | 1016 | |
9e8ae01d | 1017 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1018 | return 0; |
1019 | ||
9e8ae01d | 1020 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
907c311f EB |
1021 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1022 | ? p->signal->group_exit_code : p->exit_code; | |
1da177e4 LT |
1023 | get_task_struct(p); |
1024 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1025 | sched_annotate_sleep(); |
e61a2502 AV |
1026 | if (wo->wo_rusage) |
1027 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1028 | put_task_struct(p); |
76d9871e | 1029 | goto out_info; |
1da177e4 | 1030 | } |
1da177e4 | 1031 | /* |
abd50b39 | 1032 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1033 | */ |
f6507f83 ON |
1034 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1035 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1036 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1037 | return 0; |
986094df ON |
1038 | /* |
1039 | * We own this thread, nobody else can reap it. | |
1040 | */ | |
1041 | read_unlock(&tasklist_lock); | |
1042 | sched_annotate_sleep(); | |
f6507f83 | 1043 | |
befca967 | 1044 | /* |
f6507f83 | 1045 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1046 | */ |
f6507f83 | 1047 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1048 | struct signal_struct *sig = p->signal; |
1049 | struct signal_struct *psig = current->signal; | |
1f10206c | 1050 | unsigned long maxrss; |
5613fda9 | 1051 | u64 tgutime, tgstime; |
3795e161 | 1052 | |
1da177e4 LT |
1053 | /* |
1054 | * The resource counters for the group leader are in its | |
1055 | * own task_struct. Those for dead threads in the group | |
1056 | * are in its signal_struct, as are those for the child | |
1057 | * processes it has previously reaped. All these | |
1058 | * accumulate in the parent's signal_struct c* fields. | |
1059 | * | |
1060 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1061 | * p->signal fields because the whole thread group is dead |
1062 | * and nobody can change them. | |
1063 | * | |
1064 | * psig->stats_lock also protects us from our sub-theads | |
1065 | * which can reap other children at the same time. Until | |
1066 | * we change k_getrusage()-like users to rely on this lock | |
1067 | * we have to take ->siglock as well. | |
0cf55e1e | 1068 | * |
a0be55de IA |
1069 | * We use thread_group_cputime_adjusted() to get times for |
1070 | * the thread group, which consolidates times for all threads | |
1071 | * in the group including the group leader. | |
1da177e4 | 1072 | */ |
e80d0a1a | 1073 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1074 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1075 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1076 | psig->cutime += tgutime + sig->cutime; |
1077 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1078 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1079 | psig->cmin_flt += |
1080 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1081 | psig->cmaj_flt += | |
1082 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1083 | psig->cnvcsw += | |
1084 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1085 | psig->cnivcsw += | |
1086 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1087 | psig->cinblock += |
1088 | task_io_get_inblock(p) + | |
1089 | sig->inblock + sig->cinblock; | |
1090 | psig->coublock += | |
1091 | task_io_get_oublock(p) + | |
1092 | sig->oublock + sig->coublock; | |
1f10206c JP |
1093 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1094 | if (psig->cmaxrss < maxrss) | |
1095 | psig->cmaxrss = maxrss; | |
5995477a AR |
1096 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1097 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1098 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1099 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1100 | } |
1101 | ||
ce72a16f AV |
1102 | if (wo->wo_rusage) |
1103 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
1da177e4 LT |
1104 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1105 | ? p->signal->group_exit_code : p->exit_code; | |
359566fa | 1106 | wo->wo_stat = status; |
2f4e6e2a | 1107 | |
b4360690 | 1108 | if (state == EXIT_TRACE) { |
1da177e4 | 1109 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1110 | /* We dropped tasklist, ptracer could die and untrace */ |
1111 | ptrace_unlink(p); | |
b4360690 ON |
1112 | |
1113 | /* If parent wants a zombie, don't release it now */ | |
1114 | state = EXIT_ZOMBIE; | |
1115 | if (do_notify_parent(p, p->exit_signal)) | |
1116 | state = EXIT_DEAD; | |
abd50b39 | 1117 | p->exit_state = state; |
1da177e4 LT |
1118 | write_unlock_irq(&tasklist_lock); |
1119 | } | |
abd50b39 | 1120 | if (state == EXIT_DEAD) |
1da177e4 | 1121 | release_task(p); |
2f4e6e2a | 1122 | |
76d9871e AV |
1123 | out_info: |
1124 | infop = wo->wo_info; | |
1125 | if (infop) { | |
1126 | if ((status & 0x7f) == 0) { | |
1127 | infop->cause = CLD_EXITED; | |
1128 | infop->status = status >> 8; | |
1129 | } else { | |
1130 | infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1131 | infop->status = status & 0x7f; | |
1132 | } | |
1133 | infop->pid = pid; | |
1134 | infop->uid = uid; | |
1135 | } | |
1136 | ||
67d7ddde | 1137 | return pid; |
1da177e4 LT |
1138 | } |
1139 | ||
90bc8d8b ON |
1140 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1141 | { | |
1142 | if (ptrace) { | |
570ac933 | 1143 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1144 | return &p->exit_code; |
1145 | } else { | |
1146 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1147 | return &p->signal->group_exit_code; | |
1148 | } | |
1149 | return NULL; | |
1150 | } | |
1151 | ||
19e27463 TH |
1152 | /** |
1153 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1154 | * @wo: wait options | |
1155 | * @ptrace: is the wait for ptrace | |
1156 | * @p: task to wait for | |
1157 | * | |
1158 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1159 | * | |
1160 | * CONTEXT: | |
1161 | * read_lock(&tasklist_lock), which is released if return value is | |
1162 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1163 | * | |
1164 | * RETURNS: | |
1165 | * 0 if wait condition didn't exist and search for other wait conditions | |
1166 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1167 | * success, implies that tasklist_lock is released and wait condition | |
1168 | * search should terminate. | |
1da177e4 | 1169 | */ |
9e8ae01d ON |
1170 | static int wait_task_stopped(struct wait_opts *wo, |
1171 | int ptrace, struct task_struct *p) | |
1da177e4 | 1172 | { |
67d7ddde AV |
1173 | struct waitid_info *infop; |
1174 | int exit_code, *p_code, why; | |
ee7c82da | 1175 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1176 | pid_t pid; |
1da177e4 | 1177 | |
47918025 ON |
1178 | /* |
1179 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1180 | */ | |
9e8ae01d | 1181 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1182 | return 0; |
1183 | ||
19e27463 TH |
1184 | if (!task_stopped_code(p, ptrace)) |
1185 | return 0; | |
1186 | ||
ee7c82da ON |
1187 | exit_code = 0; |
1188 | spin_lock_irq(&p->sighand->siglock); | |
1189 | ||
90bc8d8b ON |
1190 | p_code = task_stopped_code(p, ptrace); |
1191 | if (unlikely(!p_code)) | |
ee7c82da ON |
1192 | goto unlock_sig; |
1193 | ||
90bc8d8b | 1194 | exit_code = *p_code; |
ee7c82da ON |
1195 | if (!exit_code) |
1196 | goto unlock_sig; | |
1197 | ||
9e8ae01d | 1198 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1199 | *p_code = 0; |
ee7c82da | 1200 | |
8ca937a6 | 1201 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1202 | unlock_sig: |
1203 | spin_unlock_irq(&p->sighand->siglock); | |
1204 | if (!exit_code) | |
1da177e4 LT |
1205 | return 0; |
1206 | ||
1207 | /* | |
1208 | * Now we are pretty sure this task is interesting. | |
1209 | * Make sure it doesn't get reaped out from under us while we | |
1210 | * give up the lock and then examine it below. We don't want to | |
1211 | * keep holding onto the tasklist_lock while we call getrusage and | |
1212 | * possibly take page faults for user memory. | |
1213 | */ | |
1214 | get_task_struct(p); | |
6c5f3e7b | 1215 | pid = task_pid_vnr(p); |
f470021a | 1216 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1217 | read_unlock(&tasklist_lock); |
1029a2b5 | 1218 | sched_annotate_sleep(); |
e61a2502 AV |
1219 | if (wo->wo_rusage) |
1220 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1221 | put_task_struct(p); |
1da177e4 | 1222 | |
bb380ec3 AV |
1223 | if (likely(!(wo->wo_flags & WNOWAIT))) |
1224 | wo->wo_stat = (exit_code << 8) | 0x7f; | |
1da177e4 | 1225 | |
9e8ae01d | 1226 | infop = wo->wo_info; |
67d7ddde AV |
1227 | if (infop) { |
1228 | infop->cause = why; | |
1229 | infop->status = exit_code; | |
1230 | infop->pid = pid; | |
1231 | infop->uid = uid; | |
1232 | } | |
67d7ddde | 1233 | return pid; |
1da177e4 LT |
1234 | } |
1235 | ||
1236 | /* | |
1237 | * Handle do_wait work for one task in a live, non-stopped state. | |
1238 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1239 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1240 | * released the lock and the system call should return. | |
1241 | */ | |
9e8ae01d | 1242 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1243 | { |
bb380ec3 | 1244 | struct waitid_info *infop; |
1da177e4 LT |
1245 | pid_t pid; |
1246 | uid_t uid; | |
1247 | ||
9e8ae01d | 1248 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1249 | return 0; |
1250 | ||
1da177e4 LT |
1251 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1252 | return 0; | |
1253 | ||
1254 | spin_lock_irq(&p->sighand->siglock); | |
1255 | /* Re-check with the lock held. */ | |
1256 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1257 | spin_unlock_irq(&p->sighand->siglock); | |
1258 | return 0; | |
1259 | } | |
9e8ae01d | 1260 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1261 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1262 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1263 | spin_unlock_irq(&p->sighand->siglock); |
1264 | ||
6c5f3e7b | 1265 | pid = task_pid_vnr(p); |
1da177e4 LT |
1266 | get_task_struct(p); |
1267 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1268 | sched_annotate_sleep(); |
e61a2502 AV |
1269 | if (wo->wo_rusage) |
1270 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1271 | put_task_struct(p); |
1da177e4 | 1272 | |
bb380ec3 AV |
1273 | infop = wo->wo_info; |
1274 | if (!infop) { | |
359566fa | 1275 | wo->wo_stat = 0xffff; |
1da177e4 | 1276 | } else { |
bb380ec3 AV |
1277 | infop->cause = CLD_CONTINUED; |
1278 | infop->pid = pid; | |
1279 | infop->uid = uid; | |
1280 | infop->status = SIGCONT; | |
1da177e4 | 1281 | } |
bb380ec3 | 1282 | return pid; |
1da177e4 LT |
1283 | } |
1284 | ||
98abed02 RM |
1285 | /* |
1286 | * Consider @p for a wait by @parent. | |
1287 | * | |
9e8ae01d | 1288 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1289 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1290 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1291 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1292 | * or still -ECHILD. |
98abed02 | 1293 | */ |
b6e763f0 ON |
1294 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1295 | struct task_struct *p) | |
98abed02 | 1296 | { |
3245d6ac ON |
1297 | /* |
1298 | * We can race with wait_task_zombie() from another thread. | |
1299 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1300 | * can't confuse the checks below. | |
1301 | */ | |
6aa7de05 | 1302 | int exit_state = READ_ONCE(p->exit_state); |
b3ab0316 ON |
1303 | int ret; |
1304 | ||
3245d6ac | 1305 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1306 | return 0; |
1307 | ||
bf959931 | 1308 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1309 | if (!ret) |
98abed02 RM |
1310 | return ret; |
1311 | ||
3245d6ac | 1312 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1313 | /* |
abd50b39 ON |
1314 | * ptrace == 0 means we are the natural parent. In this case |
1315 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1316 | */ |
abd50b39 | 1317 | if (likely(!ptrace)) |
50b8d257 | 1318 | wo->notask_error = 0; |
823b018e | 1319 | return 0; |
50b8d257 | 1320 | } |
823b018e | 1321 | |
377d75da ON |
1322 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1323 | /* | |
1324 | * If it is traced by its real parent's group, just pretend | |
1325 | * the caller is ptrace_do_wait() and reap this child if it | |
1326 | * is zombie. | |
1327 | * | |
1328 | * This also hides group stop state from real parent; otherwise | |
1329 | * a single stop can be reported twice as group and ptrace stop. | |
1330 | * If a ptracer wants to distinguish these two events for its | |
1331 | * own children it should create a separate process which takes | |
1332 | * the role of real parent. | |
1333 | */ | |
1334 | if (!ptrace_reparented(p)) | |
1335 | ptrace = 1; | |
1336 | } | |
1337 | ||
45cb24a1 | 1338 | /* slay zombie? */ |
3245d6ac | 1339 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1340 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1341 | if (!delay_group_leader(p)) { |
1342 | /* | |
1343 | * A zombie ptracee is only visible to its ptracer. | |
1344 | * Notification and reaping will be cascaded to the | |
1345 | * real parent when the ptracer detaches. | |
1346 | */ | |
1347 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1348 | return wait_task_zombie(wo, p); | |
1349 | } | |
98abed02 | 1350 | |
f470021a | 1351 | /* |
9b84cca2 TH |
1352 | * Allow access to stopped/continued state via zombie by |
1353 | * falling through. Clearing of notask_error is complex. | |
1354 | * | |
1355 | * When !@ptrace: | |
1356 | * | |
1357 | * If WEXITED is set, notask_error should naturally be | |
1358 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1359 | * so, if there are live subthreads, there are events to | |
1360 | * wait for. If all subthreads are dead, it's still safe | |
1361 | * to clear - this function will be called again in finite | |
1362 | * amount time once all the subthreads are released and | |
1363 | * will then return without clearing. | |
1364 | * | |
1365 | * When @ptrace: | |
1366 | * | |
1367 | * Stopped state is per-task and thus can't change once the | |
1368 | * target task dies. Only continued and exited can happen. | |
1369 | * Clear notask_error if WCONTINUED | WEXITED. | |
1370 | */ | |
1371 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1372 | wo->notask_error = 0; | |
1373 | } else { | |
1374 | /* | |
1375 | * @p is alive and it's gonna stop, continue or exit, so | |
1376 | * there always is something to wait for. | |
f470021a | 1377 | */ |
9e8ae01d | 1378 | wo->notask_error = 0; |
f470021a RM |
1379 | } |
1380 | ||
98abed02 | 1381 | /* |
45cb24a1 TH |
1382 | * Wait for stopped. Depending on @ptrace, different stopped state |
1383 | * is used and the two don't interact with each other. | |
98abed02 | 1384 | */ |
19e27463 TH |
1385 | ret = wait_task_stopped(wo, ptrace, p); |
1386 | if (ret) | |
1387 | return ret; | |
98abed02 RM |
1388 | |
1389 | /* | |
45cb24a1 TH |
1390 | * Wait for continued. There's only one continued state and the |
1391 | * ptracer can consume it which can confuse the real parent. Don't | |
1392 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1393 | */ |
9e8ae01d | 1394 | return wait_task_continued(wo, p); |
98abed02 RM |
1395 | } |
1396 | ||
1397 | /* | |
1398 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1399 | * | |
9e8ae01d | 1400 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1401 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1402 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1403 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1404 | * or still -ECHILD. |
98abed02 | 1405 | */ |
9e8ae01d | 1406 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1407 | { |
1408 | struct task_struct *p; | |
1409 | ||
1410 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1411 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1412 | |
9cd80bbb ON |
1413 | if (ret) |
1414 | return ret; | |
98abed02 RM |
1415 | } |
1416 | ||
1417 | return 0; | |
1418 | } | |
1419 | ||
9e8ae01d | 1420 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1421 | { |
1422 | struct task_struct *p; | |
1423 | ||
f470021a | 1424 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1425 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1426 | |
f470021a | 1427 | if (ret) |
98abed02 | 1428 | return ret; |
98abed02 RM |
1429 | } |
1430 | ||
1431 | return 0; | |
1432 | } | |
1433 | ||
ac6424b9 | 1434 | static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode, |
0b7570e7 ON |
1435 | int sync, void *key) |
1436 | { | |
1437 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1438 | child_wait); | |
1439 | struct task_struct *p = key; | |
1440 | ||
5c01ba49 | 1441 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1442 | return 0; |
1443 | ||
b4fe5182 ON |
1444 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1445 | return 0; | |
1446 | ||
0b7570e7 ON |
1447 | return default_wake_function(wait, mode, sync, key); |
1448 | } | |
1449 | ||
a7f0765e ON |
1450 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1451 | { | |
0b7570e7 | 1452 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
ce4dd442 | 1453 | TASK_INTERRUPTIBLE, p); |
a7f0765e ON |
1454 | } |
1455 | ||
5449162a JN |
1456 | static bool is_effectively_child(struct wait_opts *wo, bool ptrace, |
1457 | struct task_struct *target) | |
1458 | { | |
1459 | struct task_struct *parent = | |
1460 | !ptrace ? target->real_parent : target->parent; | |
1461 | ||
1462 | return current == parent || (!(wo->wo_flags & __WNOTHREAD) && | |
1463 | same_thread_group(current, parent)); | |
1464 | } | |
1465 | ||
1466 | /* | |
1467 | * Optimization for waiting on PIDTYPE_PID. No need to iterate through child | |
1468 | * and tracee lists to find the target task. | |
1469 | */ | |
1470 | static int do_wait_pid(struct wait_opts *wo) | |
1471 | { | |
1472 | bool ptrace; | |
1473 | struct task_struct *target; | |
1474 | int retval; | |
1475 | ||
1476 | ptrace = false; | |
1477 | target = pid_task(wo->wo_pid, PIDTYPE_TGID); | |
1478 | if (target && is_effectively_child(wo, ptrace, target)) { | |
1479 | retval = wait_consider_task(wo, ptrace, target); | |
1480 | if (retval) | |
1481 | return retval; | |
1482 | } | |
1483 | ||
1484 | ptrace = true; | |
1485 | target = pid_task(wo->wo_pid, PIDTYPE_PID); | |
1486 | if (target && target->ptrace && | |
1487 | is_effectively_child(wo, ptrace, target)) { | |
1488 | retval = wait_consider_task(wo, ptrace, target); | |
1489 | if (retval) | |
1490 | return retval; | |
1491 | } | |
1492 | ||
1493 | return 0; | |
1494 | } | |
1495 | ||
9e8ae01d | 1496 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1497 | { |
98abed02 | 1498 | int retval; |
1da177e4 | 1499 | |
9e8ae01d | 1500 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1501 | |
0b7570e7 ON |
1502 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1503 | wo->child_wait.private = current; | |
1504 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1505 | repeat: |
98abed02 | 1506 | /* |
3da56d16 | 1507 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1508 | * We will clear ->notask_error to zero if we see any child that |
1509 | * might later match our criteria, even if we are not able to reap | |
1510 | * it yet. | |
98abed02 | 1511 | */ |
64a16caf | 1512 | wo->notask_error = -ECHILD; |
9e8ae01d | 1513 | if ((wo->wo_type < PIDTYPE_MAX) && |
1722c14a | 1514 | (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type))) |
64a16caf | 1515 | goto notask; |
161550d7 | 1516 | |
f95d39d1 | 1517 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 | 1518 | read_lock(&tasklist_lock); |
9e8ae01d | 1519 | |
5449162a JN |
1520 | if (wo->wo_type == PIDTYPE_PID) { |
1521 | retval = do_wait_pid(wo); | |
64a16caf | 1522 | if (retval) |
98abed02 | 1523 | goto end; |
5449162a JN |
1524 | } else { |
1525 | struct task_struct *tsk = current; | |
1526 | ||
1527 | do { | |
1528 | retval = do_wait_thread(wo, tsk); | |
1529 | if (retval) | |
1530 | goto end; | |
98abed02 | 1531 | |
5449162a JN |
1532 | retval = ptrace_do_wait(wo, tsk); |
1533 | if (retval) | |
1534 | goto end; | |
1535 | ||
1536 | if (wo->wo_flags & __WNOTHREAD) | |
1537 | break; | |
1538 | } while_each_thread(current, tsk); | |
1539 | } | |
1da177e4 | 1540 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1541 | |
64a16caf | 1542 | notask: |
9e8ae01d ON |
1543 | retval = wo->notask_error; |
1544 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1545 | retval = -ERESTARTSYS; |
98abed02 RM |
1546 | if (!signal_pending(current)) { |
1547 | schedule(); | |
1548 | goto repeat; | |
1549 | } | |
1da177e4 | 1550 | } |
1da177e4 | 1551 | end: |
f95d39d1 | 1552 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1553 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1554 | return retval; |
1555 | } | |
1556 | ||
67d7ddde | 1557 | static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, |
ce72a16f | 1558 | int options, struct rusage *ru) |
1da177e4 | 1559 | { |
9e8ae01d | 1560 | struct wait_opts wo; |
161550d7 EB |
1561 | struct pid *pid = NULL; |
1562 | enum pid_type type; | |
1da177e4 | 1563 | long ret; |
ba7d25f3 | 1564 | unsigned int f_flags = 0; |
1da177e4 | 1565 | |
91c4e8ea ON |
1566 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1567 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1568 | return -EINVAL; |
1569 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1570 | return -EINVAL; | |
1571 | ||
1572 | switch (which) { | |
1573 | case P_ALL: | |
161550d7 | 1574 | type = PIDTYPE_MAX; |
1da177e4 LT |
1575 | break; |
1576 | case P_PID: | |
161550d7 EB |
1577 | type = PIDTYPE_PID; |
1578 | if (upid <= 0) | |
1da177e4 | 1579 | return -EINVAL; |
3695eae5 CB |
1580 | |
1581 | pid = find_get_pid(upid); | |
1da177e4 LT |
1582 | break; |
1583 | case P_PGID: | |
161550d7 | 1584 | type = PIDTYPE_PGID; |
821cc7b0 | 1585 | if (upid < 0) |
1da177e4 | 1586 | return -EINVAL; |
3695eae5 | 1587 | |
821cc7b0 EB |
1588 | if (upid) |
1589 | pid = find_get_pid(upid); | |
1590 | else | |
1591 | pid = get_task_pid(current, PIDTYPE_PGID); | |
3695eae5 CB |
1592 | break; |
1593 | case P_PIDFD: | |
1594 | type = PIDTYPE_PID; | |
1595 | if (upid < 0) | |
1da177e4 | 1596 | return -EINVAL; |
3695eae5 | 1597 | |
ba7d25f3 | 1598 | pid = pidfd_get_pid(upid, &f_flags); |
3695eae5 CB |
1599 | if (IS_ERR(pid)) |
1600 | return PTR_ERR(pid); | |
ba7d25f3 | 1601 | |
1da177e4 LT |
1602 | break; |
1603 | default: | |
1604 | return -EINVAL; | |
1605 | } | |
1606 | ||
9e8ae01d ON |
1607 | wo.wo_type = type; |
1608 | wo.wo_pid = pid; | |
1609 | wo.wo_flags = options; | |
1610 | wo.wo_info = infop; | |
9e8ae01d | 1611 | wo.wo_rusage = ru; |
ba7d25f3 CB |
1612 | if (f_flags & O_NONBLOCK) |
1613 | wo.wo_flags |= WNOHANG; | |
1614 | ||
9e8ae01d | 1615 | ret = do_wait(&wo); |
ba7d25f3 CB |
1616 | if (!ret && !(options & WNOHANG) && (f_flags & O_NONBLOCK)) |
1617 | ret = -EAGAIN; | |
dfe16dfa | 1618 | |
161550d7 | 1619 | put_pid(pid); |
1da177e4 LT |
1620 | return ret; |
1621 | } | |
1622 | ||
ce72a16f AV |
1623 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1624 | infop, int, options, struct rusage __user *, ru) | |
1625 | { | |
1626 | struct rusage r; | |
67d7ddde AV |
1627 | struct waitid_info info = {.status = 0}; |
1628 | long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL); | |
634a8160 | 1629 | int signo = 0; |
6c85501f | 1630 | |
634a8160 AV |
1631 | if (err > 0) { |
1632 | signo = SIGCHLD; | |
1633 | err = 0; | |
ce72a16f AV |
1634 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) |
1635 | return -EFAULT; | |
1636 | } | |
67d7ddde AV |
1637 | if (!infop) |
1638 | return err; | |
1639 | ||
41cd7805 | 1640 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1641 | return -EFAULT; |
96ca579a | 1642 | |
634a8160 | 1643 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1644 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1645 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1646 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1647 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1648 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1649 | user_write_access_end(); |
ce72a16f | 1650 | return err; |
4c48abe9 | 1651 | Efault: |
41cd7805 | 1652 | user_write_access_end(); |
4c48abe9 | 1653 | return -EFAULT; |
ce72a16f AV |
1654 | } |
1655 | ||
92ebce5a AV |
1656 | long kernel_wait4(pid_t upid, int __user *stat_addr, int options, |
1657 | struct rusage *ru) | |
1da177e4 | 1658 | { |
9e8ae01d | 1659 | struct wait_opts wo; |
161550d7 EB |
1660 | struct pid *pid = NULL; |
1661 | enum pid_type type; | |
1da177e4 LT |
1662 | long ret; |
1663 | ||
1664 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1665 | __WNOTHREAD|__WCLONE|__WALL)) | |
1666 | return -EINVAL; | |
161550d7 | 1667 | |
dd83c161 | 1668 | /* -INT_MIN is not defined */ |
1669 | if (upid == INT_MIN) | |
1670 | return -ESRCH; | |
1671 | ||
161550d7 EB |
1672 | if (upid == -1) |
1673 | type = PIDTYPE_MAX; | |
1674 | else if (upid < 0) { | |
1675 | type = PIDTYPE_PGID; | |
1676 | pid = find_get_pid(-upid); | |
1677 | } else if (upid == 0) { | |
1678 | type = PIDTYPE_PGID; | |
2ae448ef | 1679 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1680 | } else /* upid > 0 */ { |
1681 | type = PIDTYPE_PID; | |
1682 | pid = find_get_pid(upid); | |
1683 | } | |
1684 | ||
9e8ae01d ON |
1685 | wo.wo_type = type; |
1686 | wo.wo_pid = pid; | |
1687 | wo.wo_flags = options | WEXITED; | |
1688 | wo.wo_info = NULL; | |
359566fa | 1689 | wo.wo_stat = 0; |
9e8ae01d ON |
1690 | wo.wo_rusage = ru; |
1691 | ret = do_wait(&wo); | |
161550d7 | 1692 | put_pid(pid); |
359566fa AV |
1693 | if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr)) |
1694 | ret = -EFAULT; | |
1da177e4 | 1695 | |
1da177e4 LT |
1696 | return ret; |
1697 | } | |
1698 | ||
8043fc14 CH |
1699 | int kernel_wait(pid_t pid, int *stat) |
1700 | { | |
1701 | struct wait_opts wo = { | |
1702 | .wo_type = PIDTYPE_PID, | |
1703 | .wo_pid = find_get_pid(pid), | |
1704 | .wo_flags = WEXITED, | |
1705 | }; | |
1706 | int ret; | |
1707 | ||
1708 | ret = do_wait(&wo); | |
1709 | if (ret > 0 && wo.wo_stat) | |
1710 | *stat = wo.wo_stat; | |
1711 | put_pid(wo.wo_pid); | |
1712 | return ret; | |
1713 | } | |
1714 | ||
ce72a16f AV |
1715 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1716 | int, options, struct rusage __user *, ru) | |
1717 | { | |
1718 | struct rusage r; | |
1719 | long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL); | |
1720 | ||
1721 | if (err > 0) { | |
1722 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) | |
1723 | return -EFAULT; | |
1724 | } | |
1725 | return err; | |
1726 | } | |
1727 | ||
1da177e4 LT |
1728 | #ifdef __ARCH_WANT_SYS_WAITPID |
1729 | ||
1730 | /* | |
1731 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1732 | * implemented by calling sys_wait4() from libc.a. | |
1733 | */ | |
17da2bd9 | 1734 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 | 1735 | { |
d300b610 | 1736 | return kernel_wait4(pid, stat_addr, options, NULL); |
1da177e4 LT |
1737 | } |
1738 | ||
1739 | #endif | |
7e95a225 AV |
1740 | |
1741 | #ifdef CONFIG_COMPAT | |
1742 | COMPAT_SYSCALL_DEFINE4(wait4, | |
1743 | compat_pid_t, pid, | |
1744 | compat_uint_t __user *, stat_addr, | |
1745 | int, options, | |
1746 | struct compat_rusage __user *, ru) | |
1747 | { | |
ce72a16f AV |
1748 | struct rusage r; |
1749 | long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL); | |
1750 | if (err > 0) { | |
1751 | if (ru && put_compat_rusage(&r, ru)) | |
1752 | return -EFAULT; | |
7e95a225 | 1753 | } |
ce72a16f | 1754 | return err; |
7e95a225 AV |
1755 | } |
1756 | ||
1757 | COMPAT_SYSCALL_DEFINE5(waitid, | |
1758 | int, which, compat_pid_t, pid, | |
1759 | struct compat_siginfo __user *, infop, int, options, | |
1760 | struct compat_rusage __user *, uru) | |
1761 | { | |
7e95a225 | 1762 | struct rusage ru; |
67d7ddde AV |
1763 | struct waitid_info info = {.status = 0}; |
1764 | long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL); | |
634a8160 AV |
1765 | int signo = 0; |
1766 | if (err > 0) { | |
1767 | signo = SIGCHLD; | |
1768 | err = 0; | |
6c85501f AV |
1769 | if (uru) { |
1770 | /* kernel_waitid() overwrites everything in ru */ | |
1771 | if (COMPAT_USE_64BIT_TIME) | |
1772 | err = copy_to_user(uru, &ru, sizeof(ru)); | |
1773 | else | |
1774 | err = put_compat_rusage(&ru, uru); | |
1775 | if (err) | |
1776 | return -EFAULT; | |
1777 | } | |
7e95a225 AV |
1778 | } |
1779 | ||
4c48abe9 AV |
1780 | if (!infop) |
1781 | return err; | |
1782 | ||
41cd7805 | 1783 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1784 | return -EFAULT; |
96ca579a | 1785 | |
634a8160 | 1786 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1787 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1788 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1789 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1790 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1791 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1792 | user_write_access_end(); |
67d7ddde | 1793 | return err; |
4c48abe9 | 1794 | Efault: |
41cd7805 | 1795 | user_write_access_end(); |
4c48abe9 | 1796 | return -EFAULT; |
7e95a225 AV |
1797 | } |
1798 | #endif | |
7c2c11b2 | 1799 | |
38fd525a EB |
1800 | /** |
1801 | * thread_group_exited - check that a thread group has exited | |
1802 | * @pid: tgid of thread group to be checked. | |
1803 | * | |
1804 | * Test if the thread group represented by tgid has exited (all | |
1805 | * threads are zombies, dead or completely gone). | |
1806 | * | |
1807 | * Return: true if the thread group has exited. false otherwise. | |
1808 | */ | |
1809 | bool thread_group_exited(struct pid *pid) | |
1810 | { | |
1811 | struct task_struct *task; | |
1812 | bool exited; | |
1813 | ||
1814 | rcu_read_lock(); | |
1815 | task = pid_task(pid, PIDTYPE_PID); | |
1816 | exited = !task || | |
1817 | (READ_ONCE(task->exit_state) && thread_group_empty(task)); | |
1818 | rcu_read_unlock(); | |
1819 | ||
1820 | return exited; | |
1821 | } | |
1822 | EXPORT_SYMBOL(thread_group_exited); | |
1823 | ||
7c2c11b2 SM |
1824 | __weak void abort(void) |
1825 | { | |
1826 | BUG(); | |
1827 | ||
1828 | /* if that doesn't kill us, halt */ | |
1829 | panic("Oops failed to kill thread"); | |
1830 | } | |
dc8635b7 | 1831 | EXPORT_SYMBOL(abort); |