]>
Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/exit.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
1da177e4 LT |
8 | #include <linux/mm.h> |
9 | #include <linux/slab.h> | |
4eb5aaa3 | 10 | #include <linux/sched/autogroup.h> |
6e84f315 | 11 | #include <linux/sched/mm.h> |
03441a34 | 12 | #include <linux/sched/stat.h> |
29930025 | 13 | #include <linux/sched/task.h> |
68db0cf1 | 14 | #include <linux/sched/task_stack.h> |
32ef5517 | 15 | #include <linux/sched/cputime.h> |
1da177e4 | 16 | #include <linux/interrupt.h> |
1da177e4 | 17 | #include <linux/module.h> |
c59ede7b | 18 | #include <linux/capability.h> |
1da177e4 LT |
19 | #include <linux/completion.h> |
20 | #include <linux/personality.h> | |
21 | #include <linux/tty.h> | |
da9cbc87 | 22 | #include <linux/iocontext.h> |
1da177e4 | 23 | #include <linux/key.h> |
1da177e4 LT |
24 | #include <linux/cpu.h> |
25 | #include <linux/acct.h> | |
8f0ab514 | 26 | #include <linux/tsacct_kern.h> |
1da177e4 | 27 | #include <linux/file.h> |
9f3acc31 | 28 | #include <linux/fdtable.h> |
80d26af8 | 29 | #include <linux/freezer.h> |
1da177e4 | 30 | #include <linux/binfmts.h> |
ab516013 | 31 | #include <linux/nsproxy.h> |
84d73786 | 32 | #include <linux/pid_namespace.h> |
1da177e4 LT |
33 | #include <linux/ptrace.h> |
34 | #include <linux/profile.h> | |
35 | #include <linux/mount.h> | |
36 | #include <linux/proc_fs.h> | |
49d769d5 | 37 | #include <linux/kthread.h> |
1da177e4 | 38 | #include <linux/mempolicy.h> |
c757249a | 39 | #include <linux/taskstats_kern.h> |
ca74e92b | 40 | #include <linux/delayacct.h> |
b4f48b63 | 41 | #include <linux/cgroup.h> |
1da177e4 | 42 | #include <linux/syscalls.h> |
7ed20e1a | 43 | #include <linux/signal.h> |
6a14c5c9 | 44 | #include <linux/posix-timers.h> |
9f46080c | 45 | #include <linux/cn_proc.h> |
de5097c2 | 46 | #include <linux/mutex.h> |
0771dfef | 47 | #include <linux/futex.h> |
b92ce558 | 48 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 49 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 50 | #include <linux/resource.h> |
0d67a46d | 51 | #include <linux/blkdev.h> |
6eaeeaba | 52 | #include <linux/task_io_accounting_ops.h> |
30199f5a | 53 | #include <linux/tracehook.h> |
5ad4e53b | 54 | #include <linux/fs_struct.h> |
d84f4f99 | 55 | #include <linux/init_task.h> |
cdd6c482 | 56 | #include <linux/perf_event.h> |
ad8d75ff | 57 | #include <trace/events/sched.h> |
24f1e32c | 58 | #include <linux/hw_breakpoint.h> |
3d5992d2 | 59 | #include <linux/oom.h> |
54848d73 | 60 | #include <linux/writeback.h> |
40401530 | 61 | #include <linux/shm.h> |
5c9a8750 | 62 | #include <linux/kcov.h> |
53d3eaa3 | 63 | #include <linux/random.h> |
8f95c90c | 64 | #include <linux/rcuwait.h> |
7e95a225 | 65 | #include <linux/compat.h> |
1da177e4 | 66 | |
7c0f6ba6 | 67 | #include <linux/uaccess.h> |
1da177e4 | 68 | #include <asm/unistd.h> |
1da177e4 LT |
69 | #include <asm/mmu_context.h> |
70 | ||
d40e48e0 | 71 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
72 | { |
73 | nr_threads--; | |
50d75f8d | 74 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 75 | if (group_dead) { |
6883f81a | 76 | detach_pid(p, PIDTYPE_TGID); |
1da177e4 LT |
77 | detach_pid(p, PIDTYPE_PGID); |
78 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 79 | |
5e85d4ab | 80 | list_del_rcu(&p->tasks); |
9cd80bbb | 81 | list_del_init(&p->sibling); |
909ea964 | 82 | __this_cpu_dec(process_counts); |
1da177e4 | 83 | } |
47e65328 | 84 | list_del_rcu(&p->thread_group); |
0c740d0a | 85 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
86 | } |
87 | ||
6a14c5c9 ON |
88 | /* |
89 | * This function expects the tasklist_lock write-locked. | |
90 | */ | |
91 | static void __exit_signal(struct task_struct *tsk) | |
92 | { | |
93 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 94 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 95 | struct sighand_struct *sighand; |
4ada856f | 96 | struct tty_struct *uninitialized_var(tty); |
5613fda9 | 97 | u64 utime, stime; |
6a14c5c9 | 98 | |
d11c563d | 99 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 100 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
101 | spin_lock(&sighand->siglock); |
102 | ||
baa73d9e | 103 | #ifdef CONFIG_POSIX_TIMERS |
6a14c5c9 | 104 | posix_cpu_timers_exit(tsk); |
b95e31c0 | 105 | if (group_dead) |
6a14c5c9 | 106 | posix_cpu_timers_exit_group(tsk); |
baa73d9e | 107 | #endif |
e0a70217 | 108 | |
baa73d9e NP |
109 | if (group_dead) { |
110 | tty = sig->tty; | |
111 | sig->tty = NULL; | |
112 | } else { | |
6a14c5c9 ON |
113 | /* |
114 | * If there is any task waiting for the group exit | |
115 | * then notify it: | |
116 | */ | |
d344193a | 117 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 118 | wake_up_process(sig->group_exit_task); |
6db840fa | 119 | |
6a14c5c9 ON |
120 | if (tsk == sig->curr_target) |
121 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
122 | } |
123 | ||
53d3eaa3 NP |
124 | add_device_randomness((const void*) &tsk->se.sum_exec_runtime, |
125 | sizeof(unsigned long long)); | |
126 | ||
90ed9cbe | 127 | /* |
26e75b5c ON |
128 | * Accumulate here the counters for all threads as they die. We could |
129 | * skip the group leader because it is the last user of signal_struct, | |
130 | * but we want to avoid the race with thread_group_cputime() which can | |
131 | * see the empty ->thread_head list. | |
90ed9cbe RR |
132 | */ |
133 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 134 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
135 | sig->utime += utime; |
136 | sig->stime += stime; | |
137 | sig->gtime += task_gtime(tsk); | |
138 | sig->min_flt += tsk->min_flt; | |
139 | sig->maj_flt += tsk->maj_flt; | |
140 | sig->nvcsw += tsk->nvcsw; | |
141 | sig->nivcsw += tsk->nivcsw; | |
142 | sig->inblock += task_io_get_inblock(tsk); | |
143 | sig->oublock += task_io_get_oublock(tsk); | |
144 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
145 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 146 | sig->nr_threads--; |
d40e48e0 | 147 | __unhash_process(tsk, group_dead); |
e78c3496 | 148 | write_sequnlock(&sig->stats_lock); |
5876700c | 149 | |
da7978b0 ON |
150 | /* |
151 | * Do this under ->siglock, we can race with another thread | |
152 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
153 | */ | |
154 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 155 | tsk->sighand = NULL; |
6a14c5c9 | 156 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 157 | |
a7e5328a | 158 | __cleanup_sighand(sighand); |
a0be55de | 159 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 160 | if (group_dead) { |
6a14c5c9 | 161 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 162 | tty_kref_put(tty); |
6a14c5c9 ON |
163 | } |
164 | } | |
165 | ||
8c7904a0 EB |
166 | static void delayed_put_task_struct(struct rcu_head *rhp) |
167 | { | |
0a16b607 MD |
168 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
169 | ||
4e231c79 | 170 | perf_event_delayed_put(tsk); |
0a16b607 MD |
171 | trace_sched_process_free(tsk); |
172 | put_task_struct(tsk); | |
8c7904a0 EB |
173 | } |
174 | ||
3fbd7ee2 EB |
175 | void put_task_struct_rcu_user(struct task_struct *task) |
176 | { | |
177 | if (refcount_dec_and_test(&task->rcu_users)) | |
178 | call_rcu(&task->rcu, delayed_put_task_struct); | |
179 | } | |
f470021a | 180 | |
a0be55de | 181 | void release_task(struct task_struct *p) |
1da177e4 | 182 | { |
36c8b586 | 183 | struct task_struct *leader; |
7bc3e6e5 | 184 | struct pid *thread_pid; |
1da177e4 | 185 | int zap_leader; |
1f09f974 | 186 | repeat: |
c69e8d9c | 187 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
188 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
189 | rcu_read_lock(); | |
c69e8d9c | 190 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 191 | rcu_read_unlock(); |
c69e8d9c | 192 | |
6b115bf5 | 193 | cgroup_release(p); |
0203026b | 194 | |
1da177e4 | 195 | write_lock_irq(&tasklist_lock); |
a288eecc | 196 | ptrace_release_task(p); |
7bc3e6e5 | 197 | thread_pid = get_pid(p->thread_pid); |
1da177e4 | 198 | __exit_signal(p); |
35f5cad8 | 199 | |
1da177e4 LT |
200 | /* |
201 | * If we are the last non-leader member of the thread | |
202 | * group, and the leader is zombie, then notify the | |
203 | * group leader's parent process. (if it wants notification.) | |
204 | */ | |
205 | zap_leader = 0; | |
206 | leader = p->group_leader; | |
a0be55de IA |
207 | if (leader != p && thread_group_empty(leader) |
208 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
209 | /* |
210 | * If we were the last child thread and the leader has | |
211 | * exited already, and the leader's parent ignores SIGCHLD, | |
212 | * then we are the one who should release the leader. | |
dae33574 | 213 | */ |
86773473 | 214 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
215 | if (zap_leader) |
216 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
217 | } |
218 | ||
1da177e4 | 219 | write_unlock_irq(&tasklist_lock); |
7bc3e6e5 | 220 | proc_flush_pid(thread_pid); |
6ade99ec | 221 | put_pid(thread_pid); |
1da177e4 | 222 | release_thread(p); |
3fbd7ee2 | 223 | put_task_struct_rcu_user(p); |
1da177e4 LT |
224 | |
225 | p = leader; | |
226 | if (unlikely(zap_leader)) | |
227 | goto repeat; | |
228 | } | |
229 | ||
9d9a6ebf | 230 | int rcuwait_wake_up(struct rcuwait *w) |
8f95c90c | 231 | { |
9d9a6ebf | 232 | int ret = 0; |
8f95c90c DB |
233 | struct task_struct *task; |
234 | ||
235 | rcu_read_lock(); | |
236 | ||
237 | /* | |
238 | * Order condition vs @task, such that everything prior to the load | |
239 | * of @task is visible. This is the condition as to why the user called | |
c9d64a1b | 240 | * rcuwait_wake() in the first place. Pairs with set_current_state() |
8f95c90c DB |
241 | * barrier (A) in rcuwait_wait_event(). |
242 | * | |
243 | * WAIT WAKE | |
244 | * [S] tsk = current [S] cond = true | |
245 | * MB (A) MB (B) | |
246 | * [L] cond [L] tsk | |
247 | */ | |
6dc080ee | 248 | smp_mb(); /* (B) */ |
8f95c90c | 249 | |
8f95c90c DB |
250 | task = rcu_dereference(w->task); |
251 | if (task) | |
9d9a6ebf | 252 | ret = wake_up_process(task); |
8f95c90c | 253 | rcu_read_unlock(); |
9d9a6ebf DB |
254 | |
255 | return ret; | |
8f95c90c | 256 | } |
ac8dec42 | 257 | EXPORT_SYMBOL_GPL(rcuwait_wake_up); |
8f95c90c | 258 | |
1da177e4 LT |
259 | /* |
260 | * Determine if a process group is "orphaned", according to the POSIX | |
261 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
262 | * by terminal-generated stop signals. Newly orphaned process groups are | |
263 | * to receive a SIGHUP and a SIGCONT. | |
264 | * | |
265 | * "I ask you, have you ever known what it is to be an orphan?" | |
266 | */ | |
a0be55de IA |
267 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
268 | struct task_struct *ignored_task) | |
1da177e4 LT |
269 | { |
270 | struct task_struct *p; | |
1da177e4 | 271 | |
0475ac08 | 272 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
273 | if ((p == ignored_task) || |
274 | (p->exit_state && thread_group_empty(p)) || | |
275 | is_global_init(p->real_parent)) | |
1da177e4 | 276 | continue; |
05e83df6 | 277 | |
0475ac08 | 278 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
279 | task_session(p->real_parent) == task_session(p)) |
280 | return 0; | |
0475ac08 | 281 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
282 | |
283 | return 1; | |
1da177e4 LT |
284 | } |
285 | ||
3e7cd6c4 | 286 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
287 | { |
288 | int retval; | |
289 | ||
290 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 291 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
292 | read_unlock(&tasklist_lock); |
293 | ||
294 | return retval; | |
295 | } | |
296 | ||
961c4675 | 297 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 298 | { |
1da177e4 LT |
299 | struct task_struct *p; |
300 | ||
0475ac08 | 301 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
302 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
303 | return true; | |
0475ac08 | 304 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
305 | |
306 | return false; | |
1da177e4 LT |
307 | } |
308 | ||
f49ee505 ON |
309 | /* |
310 | * Check to see if any process groups have become orphaned as | |
311 | * a result of our exiting, and if they have any stopped jobs, | |
312 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
313 | */ | |
314 | static void | |
315 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
316 | { | |
317 | struct pid *pgrp = task_pgrp(tsk); | |
318 | struct task_struct *ignored_task = tsk; | |
319 | ||
320 | if (!parent) | |
a0be55de IA |
321 | /* exit: our father is in a different pgrp than |
322 | * we are and we were the only connection outside. | |
323 | */ | |
f49ee505 ON |
324 | parent = tsk->real_parent; |
325 | else | |
326 | /* reparent: our child is in a different pgrp than | |
327 | * we are, and it was the only connection outside. | |
328 | */ | |
329 | ignored_task = NULL; | |
330 | ||
331 | if (task_pgrp(parent) != pgrp && | |
332 | task_session(parent) == task_session(tsk) && | |
333 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
334 | has_stopped_jobs(pgrp)) { | |
335 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
336 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
337 | } | |
338 | } | |
339 | ||
f98bafa0 | 340 | #ifdef CONFIG_MEMCG |
cf475ad2 | 341 | /* |
733eda7a | 342 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 343 | */ |
cf475ad2 BS |
344 | void mm_update_next_owner(struct mm_struct *mm) |
345 | { | |
346 | struct task_struct *c, *g, *p = current; | |
347 | ||
348 | retry: | |
733eda7a KH |
349 | /* |
350 | * If the exiting or execing task is not the owner, it's | |
351 | * someone else's problem. | |
352 | */ | |
353 | if (mm->owner != p) | |
cf475ad2 | 354 | return; |
733eda7a KH |
355 | /* |
356 | * The current owner is exiting/execing and there are no other | |
357 | * candidates. Do not leave the mm pointing to a possibly | |
358 | * freed task structure. | |
359 | */ | |
360 | if (atomic_read(&mm->mm_users) <= 1) { | |
987717e5 | 361 | WRITE_ONCE(mm->owner, NULL); |
733eda7a KH |
362 | return; |
363 | } | |
cf475ad2 BS |
364 | |
365 | read_lock(&tasklist_lock); | |
366 | /* | |
367 | * Search in the children | |
368 | */ | |
369 | list_for_each_entry(c, &p->children, sibling) { | |
370 | if (c->mm == mm) | |
371 | goto assign_new_owner; | |
372 | } | |
373 | ||
374 | /* | |
375 | * Search in the siblings | |
376 | */ | |
dea33cfd | 377 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
378 | if (c->mm == mm) |
379 | goto assign_new_owner; | |
380 | } | |
381 | ||
382 | /* | |
f87fb599 | 383 | * Search through everything else, we should not get here often. |
cf475ad2 | 384 | */ |
39af1765 ON |
385 | for_each_process(g) { |
386 | if (g->flags & PF_KTHREAD) | |
387 | continue; | |
388 | for_each_thread(g, c) { | |
389 | if (c->mm == mm) | |
390 | goto assign_new_owner; | |
391 | if (c->mm) | |
392 | break; | |
393 | } | |
f87fb599 | 394 | } |
cf475ad2 | 395 | read_unlock(&tasklist_lock); |
31a78f23 BS |
396 | /* |
397 | * We found no owner yet mm_users > 1: this implies that we are | |
398 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 399 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 400 | */ |
987717e5 | 401 | WRITE_ONCE(mm->owner, NULL); |
cf475ad2 BS |
402 | return; |
403 | ||
404 | assign_new_owner: | |
405 | BUG_ON(c == p); | |
406 | get_task_struct(c); | |
407 | /* | |
408 | * The task_lock protects c->mm from changing. | |
409 | * We always want mm->owner->mm == mm | |
410 | */ | |
411 | task_lock(c); | |
e5991371 HD |
412 | /* |
413 | * Delay read_unlock() till we have the task_lock() | |
414 | * to ensure that c does not slip away underneath us | |
415 | */ | |
416 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
417 | if (c->mm != mm) { |
418 | task_unlock(c); | |
419 | put_task_struct(c); | |
420 | goto retry; | |
421 | } | |
987717e5 | 422 | WRITE_ONCE(mm->owner, c); |
cf475ad2 BS |
423 | task_unlock(c); |
424 | put_task_struct(c); | |
425 | } | |
f98bafa0 | 426 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 427 | |
1da177e4 LT |
428 | /* |
429 | * Turn us into a lazy TLB process if we | |
430 | * aren't already.. | |
431 | */ | |
0039962a | 432 | static void exit_mm(void) |
1da177e4 | 433 | { |
0039962a | 434 | struct mm_struct *mm = current->mm; |
b564daf8 | 435 | struct core_state *core_state; |
1da177e4 | 436 | |
4610ba7a | 437 | exit_mm_release(current, mm); |
1da177e4 LT |
438 | if (!mm) |
439 | return; | |
4fe7efdb | 440 | sync_mm_rss(mm); |
1da177e4 LT |
441 | /* |
442 | * Serialize with any possible pending coredump. | |
999d9fc1 | 443 | * We must hold mmap_sem around checking core_state |
1da177e4 | 444 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 445 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
446 | * group with ->mm != NULL. |
447 | */ | |
d8ed45c5 | 448 | mmap_read_lock(mm); |
b564daf8 ON |
449 | core_state = mm->core_state; |
450 | if (core_state) { | |
451 | struct core_thread self; | |
a0be55de | 452 | |
d8ed45c5 | 453 | mmap_read_unlock(mm); |
1da177e4 | 454 | |
0039962a | 455 | self.task = current; |
b564daf8 ON |
456 | self.next = xchg(&core_state->dumper.next, &self); |
457 | /* | |
458 | * Implies mb(), the result of xchg() must be visible | |
459 | * to core_state->dumper. | |
460 | */ | |
461 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
462 | complete(&core_state->startup); | |
1da177e4 | 463 | |
a94e2d40 | 464 | for (;;) { |
642fa448 | 465 | set_current_state(TASK_UNINTERRUPTIBLE); |
a94e2d40 ON |
466 | if (!self.task) /* see coredump_finish() */ |
467 | break; | |
80d26af8 | 468 | freezable_schedule(); |
a94e2d40 | 469 | } |
642fa448 | 470 | __set_current_state(TASK_RUNNING); |
d8ed45c5 | 471 | mmap_read_lock(mm); |
1da177e4 | 472 | } |
f1f10076 | 473 | mmgrab(mm); |
0039962a | 474 | BUG_ON(mm != current->active_mm); |
1da177e4 | 475 | /* more a memory barrier than a real lock */ |
0039962a DB |
476 | task_lock(current); |
477 | current->mm = NULL; | |
d8ed45c5 | 478 | mmap_read_unlock(mm); |
1da177e4 | 479 | enter_lazy_tlb(mm, current); |
0039962a | 480 | task_unlock(current); |
cf475ad2 | 481 | mm_update_next_owner(mm); |
1da177e4 | 482 | mmput(mm); |
c32b3cbe | 483 | if (test_thread_flag(TIF_MEMDIE)) |
38531201 | 484 | exit_oom_victim(); |
1da177e4 LT |
485 | } |
486 | ||
c9dc05bf ON |
487 | static struct task_struct *find_alive_thread(struct task_struct *p) |
488 | { | |
489 | struct task_struct *t; | |
490 | ||
491 | for_each_thread(p, t) { | |
492 | if (!(t->flags & PF_EXITING)) | |
493 | return t; | |
494 | } | |
495 | return NULL; | |
496 | } | |
497 | ||
8fb335e0 AV |
498 | static struct task_struct *find_child_reaper(struct task_struct *father, |
499 | struct list_head *dead) | |
1109909c ON |
500 | __releases(&tasklist_lock) |
501 | __acquires(&tasklist_lock) | |
502 | { | |
503 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
504 | struct task_struct *reaper = pid_ns->child_reaper; | |
8fb335e0 | 505 | struct task_struct *p, *n; |
1109909c ON |
506 | |
507 | if (likely(reaper != father)) | |
508 | return reaper; | |
509 | ||
c9dc05bf ON |
510 | reaper = find_alive_thread(father); |
511 | if (reaper) { | |
1109909c ON |
512 | pid_ns->child_reaper = reaper; |
513 | return reaper; | |
514 | } | |
515 | ||
516 | write_unlock_irq(&tasklist_lock); | |
8fb335e0 AV |
517 | |
518 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
519 | list_del_init(&p->ptrace_entry); | |
520 | release_task(p); | |
521 | } | |
522 | ||
1109909c ON |
523 | zap_pid_ns_processes(pid_ns); |
524 | write_lock_irq(&tasklist_lock); | |
525 | ||
526 | return father; | |
527 | } | |
528 | ||
1da177e4 | 529 | /* |
ebec18a6 LP |
530 | * When we die, we re-parent all our children, and try to: |
531 | * 1. give them to another thread in our thread group, if such a member exists | |
532 | * 2. give it to the first ancestor process which prctl'd itself as a | |
533 | * child_subreaper for its children (like a service manager) | |
534 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 535 | */ |
1109909c ON |
536 | static struct task_struct *find_new_reaper(struct task_struct *father, |
537 | struct task_struct *child_reaper) | |
1da177e4 | 538 | { |
c9dc05bf | 539 | struct task_struct *thread, *reaper; |
1da177e4 | 540 | |
c9dc05bf ON |
541 | thread = find_alive_thread(father); |
542 | if (thread) | |
950bbabb | 543 | return thread; |
1da177e4 | 544 | |
7d24e2df | 545 | if (father->signal->has_child_subreaper) { |
c6c70f44 | 546 | unsigned int ns_level = task_pid(father)->level; |
ebec18a6 | 547 | /* |
175aed3f | 548 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
c6c70f44 ON |
549 | * We can't check reaper != child_reaper to ensure we do not |
550 | * cross the namespaces, the exiting parent could be injected | |
551 | * by setns() + fork(). | |
552 | * We check pid->level, this is slightly more efficient than | |
553 | * task_active_pid_ns(reaper) != task_active_pid_ns(father). | |
ebec18a6 | 554 | */ |
c6c70f44 ON |
555 | for (reaper = father->real_parent; |
556 | task_pid(reaper)->level == ns_level; | |
ebec18a6 | 557 | reaper = reaper->real_parent) { |
175aed3f | 558 | if (reaper == &init_task) |
ebec18a6 LP |
559 | break; |
560 | if (!reaper->signal->is_child_subreaper) | |
561 | continue; | |
c9dc05bf ON |
562 | thread = find_alive_thread(reaper); |
563 | if (thread) | |
564 | return thread; | |
ebec18a6 | 565 | } |
1da177e4 | 566 | } |
762a24be | 567 | |
1109909c | 568 | return child_reaper; |
950bbabb ON |
569 | } |
570 | ||
5dfc80be ON |
571 | /* |
572 | * Any that need to be release_task'd are put on the @dead list. | |
573 | */ | |
9cd80bbb | 574 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
575 | struct list_head *dead) |
576 | { | |
2831096e | 577 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
578 | return; |
579 | ||
abd50b39 | 580 | /* We don't want people slaying init. */ |
5dfc80be ON |
581 | p->exit_signal = SIGCHLD; |
582 | ||
583 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 584 | if (!p->ptrace && |
5dfc80be | 585 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 586 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 587 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 588 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
589 | } |
590 | } | |
591 | ||
592 | kill_orphaned_pgrp(p, father); | |
593 | } | |
594 | ||
482a3767 ON |
595 | /* |
596 | * This does two things: | |
597 | * | |
598 | * A. Make init inherit all the child processes | |
599 | * B. Check to see if any process groups have become orphaned | |
600 | * as a result of our exiting, and if they have any stopped | |
601 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
602 | */ | |
603 | static void forget_original_parent(struct task_struct *father, | |
604 | struct list_head *dead) | |
1da177e4 | 605 | { |
482a3767 | 606 | struct task_struct *p, *t, *reaper; |
762a24be | 607 | |
7c8bd232 | 608 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 609 | exit_ptrace(father, dead); |
f470021a | 610 | |
7c8bd232 | 611 | /* Can drop and reacquire tasklist_lock */ |
8fb335e0 | 612 | reaper = find_child_reaper(father, dead); |
ad9e206a | 613 | if (list_empty(&father->children)) |
482a3767 | 614 | return; |
1109909c ON |
615 | |
616 | reaper = find_new_reaper(father, reaper); | |
2831096e | 617 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 618 | for_each_thread(p, t) { |
22a34c6f MB |
619 | RCU_INIT_POINTER(t->real_parent, reaper); |
620 | BUG_ON((!t->ptrace) != (rcu_access_pointer(t->parent) == father)); | |
57a05918 | 621 | if (likely(!t->ptrace)) |
9cd80bbb | 622 | t->parent = t->real_parent; |
9cd80bbb ON |
623 | if (t->pdeath_signal) |
624 | group_send_sig_info(t->pdeath_signal, | |
01024980 EB |
625 | SEND_SIG_NOINFO, t, |
626 | PIDTYPE_TGID); | |
57a05918 | 627 | } |
2831096e ON |
628 | /* |
629 | * If this is a threaded reparent there is no need to | |
630 | * notify anyone anything has happened. | |
631 | */ | |
632 | if (!same_thread_group(reaper, father)) | |
482a3767 | 633 | reparent_leader(father, p, dead); |
1da177e4 | 634 | } |
2831096e | 635 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
636 | } |
637 | ||
638 | /* | |
639 | * Send signals to all our closest relatives so that they know | |
640 | * to properly mourn us.. | |
641 | */ | |
821c7de7 | 642 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 643 | { |
53c8f9f1 | 644 | bool autoreap; |
482a3767 ON |
645 | struct task_struct *p, *n; |
646 | LIST_HEAD(dead); | |
1da177e4 | 647 | |
762a24be | 648 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
649 | forget_original_parent(tsk, &dead); |
650 | ||
821c7de7 ON |
651 | if (group_dead) |
652 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 653 | |
b191d649 | 654 | tsk->exit_state = EXIT_ZOMBIE; |
45cdf5cc ON |
655 | if (unlikely(tsk->ptrace)) { |
656 | int sig = thread_group_leader(tsk) && | |
657 | thread_group_empty(tsk) && | |
658 | !ptrace_reparented(tsk) ? | |
659 | tsk->exit_signal : SIGCHLD; | |
660 | autoreap = do_notify_parent(tsk, sig); | |
661 | } else if (thread_group_leader(tsk)) { | |
662 | autoreap = thread_group_empty(tsk) && | |
663 | do_notify_parent(tsk, tsk->exit_signal); | |
664 | } else { | |
665 | autoreap = true; | |
666 | } | |
1da177e4 | 667 | |
30b692d3 CB |
668 | if (autoreap) { |
669 | tsk->exit_state = EXIT_DEAD; | |
6c66e7db | 670 | list_add(&tsk->ptrace_entry, &dead); |
30b692d3 | 671 | } |
1da177e4 | 672 | |
9c339168 ON |
673 | /* mt-exec, de_thread() is waiting for group leader */ |
674 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 675 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
676 | write_unlock_irq(&tasklist_lock); |
677 | ||
482a3767 ON |
678 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
679 | list_del_init(&p->ptrace_entry); | |
680 | release_task(p); | |
681 | } | |
1da177e4 LT |
682 | } |
683 | ||
e18eecb8 JD |
684 | #ifdef CONFIG_DEBUG_STACK_USAGE |
685 | static void check_stack_usage(void) | |
686 | { | |
687 | static DEFINE_SPINLOCK(low_water_lock); | |
688 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
689 | unsigned long free; |
690 | ||
7c9f8861 | 691 | free = stack_not_used(current); |
e18eecb8 JD |
692 | |
693 | if (free >= lowest_to_date) | |
694 | return; | |
695 | ||
696 | spin_lock(&low_water_lock); | |
697 | if (free < lowest_to_date) { | |
627393d4 | 698 | pr_info("%s (%d) used greatest stack depth: %lu bytes left\n", |
a0be55de | 699 | current->comm, task_pid_nr(current), free); |
e18eecb8 JD |
700 | lowest_to_date = free; |
701 | } | |
702 | spin_unlock(&low_water_lock); | |
703 | } | |
704 | #else | |
705 | static inline void check_stack_usage(void) {} | |
706 | #endif | |
707 | ||
9af6528e | 708 | void __noreturn do_exit(long code) |
1da177e4 LT |
709 | { |
710 | struct task_struct *tsk = current; | |
711 | int group_dead; | |
712 | ||
586b58ca JH |
713 | /* |
714 | * We can get here from a kernel oops, sometimes with preemption off. | |
715 | * Start by checking for critical errors. | |
716 | * Then fix up important state like USER_DS and preemption. | |
717 | * Then do everything else. | |
718 | */ | |
1da177e4 | 719 | |
73c10101 | 720 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 721 | |
1da177e4 LT |
722 | if (unlikely(in_interrupt())) |
723 | panic("Aiee, killing interrupt handler!"); | |
724 | if (unlikely(!tsk->pid)) | |
725 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 726 | |
33dd94ae NE |
727 | /* |
728 | * If do_exit is called because this processes oopsed, it's possible | |
729 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
730 | * continuing. Amongst other possible reasons, this is to prevent | |
731 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
732 | * kernel address. | |
733 | */ | |
734 | set_fs(USER_DS); | |
735 | ||
586b58ca JH |
736 | if (unlikely(in_atomic())) { |
737 | pr_info("note: %s[%d] exited with preempt_count %d\n", | |
738 | current->comm, task_pid_nr(current), | |
739 | preempt_count()); | |
740 | preempt_count_set(PREEMPT_ENABLED); | |
741 | } | |
742 | ||
743 | profile_task_exit(tsk); | |
744 | kcov_task_exit(tsk); | |
745 | ||
a288eecc | 746 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 747 | |
e0e81739 DH |
748 | validate_creds_for_do_exit(tsk); |
749 | ||
df164db5 AN |
750 | /* |
751 | * We're taking recursive faults here in do_exit. Safest is to just | |
752 | * leave this task alone and wait for reboot. | |
753 | */ | |
754 | if (unlikely(tsk->flags & PF_EXITING)) { | |
a0be55de | 755 | pr_alert("Fixing recursive fault but reboot is needed!\n"); |
18f69438 | 756 | futex_exit_recursive(tsk); |
df164db5 AN |
757 | set_current_state(TASK_UNINTERRUPTIBLE); |
758 | schedule(); | |
759 | } | |
760 | ||
d12619b5 | 761 | exit_signals(tsk); /* sets PF_EXITING */ |
1da177e4 | 762 | |
48d212a2 LT |
763 | /* sync mm's RSS info before statistics gathering */ |
764 | if (tsk->mm) | |
765 | sync_mm_rss(tsk->mm); | |
51229b49 | 766 | acct_update_integrals(tsk); |
1da177e4 | 767 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 768 | if (group_dead) { |
43cf75d9 | 769 | /* |
770 | * If the last thread of global init has exited, panic | |
771 | * immediately to get a useable coredump. | |
772 | */ | |
773 | if (unlikely(is_global_init(tsk))) | |
774 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
775 | tsk->signal->group_exit_code ?: (int)code); | |
776 | ||
baa73d9e | 777 | #ifdef CONFIG_POSIX_TIMERS |
778e9a9c | 778 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 779 | exit_itimers(tsk->signal); |
baa73d9e | 780 | #endif |
1f10206c JP |
781 | if (tsk->mm) |
782 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 783 | } |
f6ec29a4 | 784 | acct_collect(code, group_dead); |
522ed776 MT |
785 | if (group_dead) |
786 | tty_audit_exit(); | |
a4ff8dba | 787 | audit_free(tsk); |
115085ea | 788 | |
48d212a2 | 789 | tsk->exit_code = code; |
115085ea | 790 | taskstats_exit(tsk, group_dead); |
c757249a | 791 | |
0039962a | 792 | exit_mm(); |
1da177e4 | 793 | |
0e464814 | 794 | if (group_dead) |
f6ec29a4 | 795 | acct_process(); |
0a16b607 MD |
796 | trace_sched_process_exit(tsk); |
797 | ||
1da177e4 | 798 | exit_sem(tsk); |
b34a6b1d | 799 | exit_shm(tsk); |
1ec7f1dd AV |
800 | exit_files(tsk); |
801 | exit_fs(tsk); | |
c39df5fa ON |
802 | if (group_dead) |
803 | disassociate_ctty(1); | |
8aac6270 | 804 | exit_task_namespaces(tsk); |
ed3e694d | 805 | exit_task_work(tsk); |
e6464694 | 806 | exit_thread(tsk); |
73ab1cb2 | 807 | exit_umh(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 RA |
859 | EXPORT_SYMBOL_GPL(do_exit); |
860 | ||
9402c95f | 861 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
862 | { |
863 | if (comp) | |
864 | complete(comp); | |
55a101f8 | 865 | |
1da177e4 LT |
866 | do_exit(code); |
867 | } | |
1da177e4 LT |
868 | EXPORT_SYMBOL(complete_and_exit); |
869 | ||
754fe8d2 | 870 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
871 | { |
872 | do_exit((error_code&0xff)<<8); | |
873 | } | |
874 | ||
1da177e4 LT |
875 | /* |
876 | * Take down every thread in the group. This is called by fatal signals | |
877 | * as well as by sys_exit_group (below). | |
878 | */ | |
9402c95f | 879 | void |
1da177e4 LT |
880 | do_group_exit(int exit_code) |
881 | { | |
bfc4b089 ON |
882 | struct signal_struct *sig = current->signal; |
883 | ||
1da177e4 LT |
884 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
885 | ||
bfc4b089 ON |
886 | if (signal_group_exit(sig)) |
887 | exit_code = sig->group_exit_code; | |
1da177e4 | 888 | else if (!thread_group_empty(current)) { |
1da177e4 | 889 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 890 | |
1da177e4 | 891 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 892 | if (signal_group_exit(sig)) |
1da177e4 LT |
893 | /* Another thread got here before we took the lock. */ |
894 | exit_code = sig->group_exit_code; | |
895 | else { | |
1da177e4 | 896 | sig->group_exit_code = exit_code; |
ed5d2cac | 897 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
898 | zap_other_threads(current); |
899 | } | |
900 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
901 | } |
902 | ||
903 | do_exit(exit_code); | |
904 | /* NOTREACHED */ | |
905 | } | |
906 | ||
907 | /* | |
908 | * this kills every thread in the thread group. Note that any externally | |
909 | * wait4()-ing process will get the correct exit code - even if this | |
910 | * thread is not the thread group leader. | |
911 | */ | |
754fe8d2 | 912 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
913 | { |
914 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
915 | /* NOTREACHED */ |
916 | return 0; | |
1da177e4 LT |
917 | } |
918 | ||
67d7ddde AV |
919 | struct waitid_info { |
920 | pid_t pid; | |
921 | uid_t uid; | |
922 | int status; | |
923 | int cause; | |
924 | }; | |
925 | ||
9e8ae01d ON |
926 | struct wait_opts { |
927 | enum pid_type wo_type; | |
9e8ae01d | 928 | int wo_flags; |
e1eb1ebc | 929 | struct pid *wo_pid; |
9e8ae01d | 930 | |
67d7ddde | 931 | struct waitid_info *wo_info; |
359566fa | 932 | int wo_stat; |
ce72a16f | 933 | struct rusage *wo_rusage; |
9e8ae01d | 934 | |
ac6424b9 | 935 | wait_queue_entry_t child_wait; |
9e8ae01d ON |
936 | int notask_error; |
937 | }; | |
938 | ||
989264f4 | 939 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 940 | { |
5c01ba49 ON |
941 | return wo->wo_type == PIDTYPE_MAX || |
942 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
943 | } | |
1da177e4 | 944 | |
bf959931 ON |
945 | static int |
946 | eligible_child(struct wait_opts *wo, bool ptrace, struct task_struct *p) | |
5c01ba49 ON |
947 | { |
948 | if (!eligible_pid(wo, p)) | |
949 | return 0; | |
bf959931 ON |
950 | |
951 | /* | |
952 | * Wait for all children (clone and not) if __WALL is set or | |
953 | * if it is traced by us. | |
954 | */ | |
955 | if (ptrace || (wo->wo_flags & __WALL)) | |
956 | return 1; | |
957 | ||
958 | /* | |
959 | * Otherwise, wait for clone children *only* if __WCLONE is set; | |
960 | * otherwise, wait for non-clone children *only*. | |
961 | * | |
962 | * Note: a "clone" child here is one that reports to its parent | |
963 | * using a signal other than SIGCHLD, or a non-leader thread which | |
964 | * we can only see if it is traced by us. | |
965 | */ | |
966 | if ((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) | |
1da177e4 | 967 | return 0; |
1da177e4 | 968 | |
14dd0b81 | 969 | return 1; |
1da177e4 LT |
970 | } |
971 | ||
1da177e4 LT |
972 | /* |
973 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
974 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
975 | * the lock and this task is uninteresting. If we return nonzero, we have | |
976 | * released the lock and the system call should return. | |
977 | */ | |
9e8ae01d | 978 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 979 | { |
67d7ddde | 980 | int state, status; |
6c5f3e7b | 981 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 982 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
67d7ddde | 983 | struct waitid_info *infop; |
1da177e4 | 984 | |
9e8ae01d | 985 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
986 | return 0; |
987 | ||
9e8ae01d | 988 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
76d9871e | 989 | status = p->exit_code; |
1da177e4 LT |
990 | get_task_struct(p); |
991 | read_unlock(&tasklist_lock); | |
1029a2b5 | 992 | sched_annotate_sleep(); |
e61a2502 AV |
993 | if (wo->wo_rusage) |
994 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 995 | put_task_struct(p); |
76d9871e | 996 | goto out_info; |
1da177e4 | 997 | } |
1da177e4 | 998 | /* |
abd50b39 | 999 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1000 | */ |
f6507f83 ON |
1001 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1002 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1003 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1004 | return 0; |
986094df ON |
1005 | /* |
1006 | * We own this thread, nobody else can reap it. | |
1007 | */ | |
1008 | read_unlock(&tasklist_lock); | |
1009 | sched_annotate_sleep(); | |
f6507f83 | 1010 | |
befca967 | 1011 | /* |
f6507f83 | 1012 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1013 | */ |
f6507f83 | 1014 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1015 | struct signal_struct *sig = p->signal; |
1016 | struct signal_struct *psig = current->signal; | |
1f10206c | 1017 | unsigned long maxrss; |
5613fda9 | 1018 | u64 tgutime, tgstime; |
3795e161 | 1019 | |
1da177e4 LT |
1020 | /* |
1021 | * The resource counters for the group leader are in its | |
1022 | * own task_struct. Those for dead threads in the group | |
1023 | * are in its signal_struct, as are those for the child | |
1024 | * processes it has previously reaped. All these | |
1025 | * accumulate in the parent's signal_struct c* fields. | |
1026 | * | |
1027 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1028 | * p->signal fields because the whole thread group is dead |
1029 | * and nobody can change them. | |
1030 | * | |
1031 | * psig->stats_lock also protects us from our sub-theads | |
1032 | * which can reap other children at the same time. Until | |
1033 | * we change k_getrusage()-like users to rely on this lock | |
1034 | * we have to take ->siglock as well. | |
0cf55e1e | 1035 | * |
a0be55de IA |
1036 | * We use thread_group_cputime_adjusted() to get times for |
1037 | * the thread group, which consolidates times for all threads | |
1038 | * in the group including the group leader. | |
1da177e4 | 1039 | */ |
e80d0a1a | 1040 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1041 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1042 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1043 | psig->cutime += tgutime + sig->cutime; |
1044 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1045 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1046 | psig->cmin_flt += |
1047 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1048 | psig->cmaj_flt += | |
1049 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1050 | psig->cnvcsw += | |
1051 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1052 | psig->cnivcsw += | |
1053 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1054 | psig->cinblock += |
1055 | task_io_get_inblock(p) + | |
1056 | sig->inblock + sig->cinblock; | |
1057 | psig->coublock += | |
1058 | task_io_get_oublock(p) + | |
1059 | sig->oublock + sig->coublock; | |
1f10206c JP |
1060 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1061 | if (psig->cmaxrss < maxrss) | |
1062 | psig->cmaxrss = maxrss; | |
5995477a AR |
1063 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1064 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1065 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1066 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1067 | } |
1068 | ||
ce72a16f AV |
1069 | if (wo->wo_rusage) |
1070 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
1da177e4 LT |
1071 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1072 | ? p->signal->group_exit_code : p->exit_code; | |
359566fa | 1073 | wo->wo_stat = status; |
2f4e6e2a | 1074 | |
b4360690 | 1075 | if (state == EXIT_TRACE) { |
1da177e4 | 1076 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1077 | /* We dropped tasklist, ptracer could die and untrace */ |
1078 | ptrace_unlink(p); | |
b4360690 ON |
1079 | |
1080 | /* If parent wants a zombie, don't release it now */ | |
1081 | state = EXIT_ZOMBIE; | |
1082 | if (do_notify_parent(p, p->exit_signal)) | |
1083 | state = EXIT_DEAD; | |
abd50b39 | 1084 | p->exit_state = state; |
1da177e4 LT |
1085 | write_unlock_irq(&tasklist_lock); |
1086 | } | |
abd50b39 | 1087 | if (state == EXIT_DEAD) |
1da177e4 | 1088 | release_task(p); |
2f4e6e2a | 1089 | |
76d9871e AV |
1090 | out_info: |
1091 | infop = wo->wo_info; | |
1092 | if (infop) { | |
1093 | if ((status & 0x7f) == 0) { | |
1094 | infop->cause = CLD_EXITED; | |
1095 | infop->status = status >> 8; | |
1096 | } else { | |
1097 | infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1098 | infop->status = status & 0x7f; | |
1099 | } | |
1100 | infop->pid = pid; | |
1101 | infop->uid = uid; | |
1102 | } | |
1103 | ||
67d7ddde | 1104 | return pid; |
1da177e4 LT |
1105 | } |
1106 | ||
90bc8d8b ON |
1107 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1108 | { | |
1109 | if (ptrace) { | |
570ac933 | 1110 | if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING)) |
90bc8d8b ON |
1111 | return &p->exit_code; |
1112 | } else { | |
1113 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1114 | return &p->signal->group_exit_code; | |
1115 | } | |
1116 | return NULL; | |
1117 | } | |
1118 | ||
19e27463 TH |
1119 | /** |
1120 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1121 | * @wo: wait options | |
1122 | * @ptrace: is the wait for ptrace | |
1123 | * @p: task to wait for | |
1124 | * | |
1125 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1126 | * | |
1127 | * CONTEXT: | |
1128 | * read_lock(&tasklist_lock), which is released if return value is | |
1129 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1130 | * | |
1131 | * RETURNS: | |
1132 | * 0 if wait condition didn't exist and search for other wait conditions | |
1133 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1134 | * success, implies that tasklist_lock is released and wait condition | |
1135 | * search should terminate. | |
1da177e4 | 1136 | */ |
9e8ae01d ON |
1137 | static int wait_task_stopped(struct wait_opts *wo, |
1138 | int ptrace, struct task_struct *p) | |
1da177e4 | 1139 | { |
67d7ddde AV |
1140 | struct waitid_info *infop; |
1141 | int exit_code, *p_code, why; | |
ee7c82da | 1142 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1143 | pid_t pid; |
1da177e4 | 1144 | |
47918025 ON |
1145 | /* |
1146 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1147 | */ | |
9e8ae01d | 1148 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1149 | return 0; |
1150 | ||
19e27463 TH |
1151 | if (!task_stopped_code(p, ptrace)) |
1152 | return 0; | |
1153 | ||
ee7c82da ON |
1154 | exit_code = 0; |
1155 | spin_lock_irq(&p->sighand->siglock); | |
1156 | ||
90bc8d8b ON |
1157 | p_code = task_stopped_code(p, ptrace); |
1158 | if (unlikely(!p_code)) | |
ee7c82da ON |
1159 | goto unlock_sig; |
1160 | ||
90bc8d8b | 1161 | exit_code = *p_code; |
ee7c82da ON |
1162 | if (!exit_code) |
1163 | goto unlock_sig; | |
1164 | ||
9e8ae01d | 1165 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1166 | *p_code = 0; |
ee7c82da | 1167 | |
8ca937a6 | 1168 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1169 | unlock_sig: |
1170 | spin_unlock_irq(&p->sighand->siglock); | |
1171 | if (!exit_code) | |
1da177e4 LT |
1172 | return 0; |
1173 | ||
1174 | /* | |
1175 | * Now we are pretty sure this task is interesting. | |
1176 | * Make sure it doesn't get reaped out from under us while we | |
1177 | * give up the lock and then examine it below. We don't want to | |
1178 | * keep holding onto the tasklist_lock while we call getrusage and | |
1179 | * possibly take page faults for user memory. | |
1180 | */ | |
1181 | get_task_struct(p); | |
6c5f3e7b | 1182 | pid = task_pid_vnr(p); |
f470021a | 1183 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1184 | read_unlock(&tasklist_lock); |
1029a2b5 | 1185 | sched_annotate_sleep(); |
e61a2502 AV |
1186 | if (wo->wo_rusage) |
1187 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1188 | put_task_struct(p); |
1da177e4 | 1189 | |
bb380ec3 AV |
1190 | if (likely(!(wo->wo_flags & WNOWAIT))) |
1191 | wo->wo_stat = (exit_code << 8) | 0x7f; | |
1da177e4 | 1192 | |
9e8ae01d | 1193 | infop = wo->wo_info; |
67d7ddde AV |
1194 | if (infop) { |
1195 | infop->cause = why; | |
1196 | infop->status = exit_code; | |
1197 | infop->pid = pid; | |
1198 | infop->uid = uid; | |
1199 | } | |
67d7ddde | 1200 | return pid; |
1da177e4 LT |
1201 | } |
1202 | ||
1203 | /* | |
1204 | * Handle do_wait work for one task in a live, non-stopped state. | |
1205 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1206 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1207 | * released the lock and the system call should return. | |
1208 | */ | |
9e8ae01d | 1209 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1210 | { |
bb380ec3 | 1211 | struct waitid_info *infop; |
1da177e4 LT |
1212 | pid_t pid; |
1213 | uid_t uid; | |
1214 | ||
9e8ae01d | 1215 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1216 | return 0; |
1217 | ||
1da177e4 LT |
1218 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1219 | return 0; | |
1220 | ||
1221 | spin_lock_irq(&p->sighand->siglock); | |
1222 | /* Re-check with the lock held. */ | |
1223 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1224 | spin_unlock_irq(&p->sighand->siglock); | |
1225 | return 0; | |
1226 | } | |
9e8ae01d | 1227 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1228 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1229 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1230 | spin_unlock_irq(&p->sighand->siglock); |
1231 | ||
6c5f3e7b | 1232 | pid = task_pid_vnr(p); |
1da177e4 LT |
1233 | get_task_struct(p); |
1234 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1235 | sched_annotate_sleep(); |
e61a2502 AV |
1236 | if (wo->wo_rusage) |
1237 | getrusage(p, RUSAGE_BOTH, wo->wo_rusage); | |
bb380ec3 | 1238 | put_task_struct(p); |
1da177e4 | 1239 | |
bb380ec3 AV |
1240 | infop = wo->wo_info; |
1241 | if (!infop) { | |
359566fa | 1242 | wo->wo_stat = 0xffff; |
1da177e4 | 1243 | } else { |
bb380ec3 AV |
1244 | infop->cause = CLD_CONTINUED; |
1245 | infop->pid = pid; | |
1246 | infop->uid = uid; | |
1247 | infop->status = SIGCONT; | |
1da177e4 | 1248 | } |
bb380ec3 | 1249 | return pid; |
1da177e4 LT |
1250 | } |
1251 | ||
98abed02 RM |
1252 | /* |
1253 | * Consider @p for a wait by @parent. | |
1254 | * | |
9e8ae01d | 1255 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1256 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1257 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1258 | * then ->notask_error is 0 if @p is an eligible child, |
3a2f5a59 | 1259 | * or still -ECHILD. |
98abed02 | 1260 | */ |
b6e763f0 ON |
1261 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1262 | struct task_struct *p) | |
98abed02 | 1263 | { |
3245d6ac ON |
1264 | /* |
1265 | * We can race with wait_task_zombie() from another thread. | |
1266 | * Ensure that EXIT_ZOMBIE -> EXIT_DEAD/EXIT_TRACE transition | |
1267 | * can't confuse the checks below. | |
1268 | */ | |
6aa7de05 | 1269 | int exit_state = READ_ONCE(p->exit_state); |
b3ab0316 ON |
1270 | int ret; |
1271 | ||
3245d6ac | 1272 | if (unlikely(exit_state == EXIT_DEAD)) |
b3ab0316 ON |
1273 | return 0; |
1274 | ||
bf959931 | 1275 | ret = eligible_child(wo, ptrace, p); |
14dd0b81 | 1276 | if (!ret) |
98abed02 RM |
1277 | return ret; |
1278 | ||
3245d6ac | 1279 | if (unlikely(exit_state == EXIT_TRACE)) { |
50b8d257 | 1280 | /* |
abd50b39 ON |
1281 | * ptrace == 0 means we are the natural parent. In this case |
1282 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1283 | */ |
abd50b39 | 1284 | if (likely(!ptrace)) |
50b8d257 | 1285 | wo->notask_error = 0; |
823b018e | 1286 | return 0; |
50b8d257 | 1287 | } |
823b018e | 1288 | |
377d75da ON |
1289 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1290 | /* | |
1291 | * If it is traced by its real parent's group, just pretend | |
1292 | * the caller is ptrace_do_wait() and reap this child if it | |
1293 | * is zombie. | |
1294 | * | |
1295 | * This also hides group stop state from real parent; otherwise | |
1296 | * a single stop can be reported twice as group and ptrace stop. | |
1297 | * If a ptracer wants to distinguish these two events for its | |
1298 | * own children it should create a separate process which takes | |
1299 | * the role of real parent. | |
1300 | */ | |
1301 | if (!ptrace_reparented(p)) | |
1302 | ptrace = 1; | |
1303 | } | |
1304 | ||
45cb24a1 | 1305 | /* slay zombie? */ |
3245d6ac | 1306 | if (exit_state == EXIT_ZOMBIE) { |
9b84cca2 | 1307 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1308 | if (!delay_group_leader(p)) { |
1309 | /* | |
1310 | * A zombie ptracee is only visible to its ptracer. | |
1311 | * Notification and reaping will be cascaded to the | |
1312 | * real parent when the ptracer detaches. | |
1313 | */ | |
1314 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1315 | return wait_task_zombie(wo, p); | |
1316 | } | |
98abed02 | 1317 | |
f470021a | 1318 | /* |
9b84cca2 TH |
1319 | * Allow access to stopped/continued state via zombie by |
1320 | * falling through. Clearing of notask_error is complex. | |
1321 | * | |
1322 | * When !@ptrace: | |
1323 | * | |
1324 | * If WEXITED is set, notask_error should naturally be | |
1325 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1326 | * so, if there are live subthreads, there are events to | |
1327 | * wait for. If all subthreads are dead, it's still safe | |
1328 | * to clear - this function will be called again in finite | |
1329 | * amount time once all the subthreads are released and | |
1330 | * will then return without clearing. | |
1331 | * | |
1332 | * When @ptrace: | |
1333 | * | |
1334 | * Stopped state is per-task and thus can't change once the | |
1335 | * target task dies. Only continued and exited can happen. | |
1336 | * Clear notask_error if WCONTINUED | WEXITED. | |
1337 | */ | |
1338 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1339 | wo->notask_error = 0; | |
1340 | } else { | |
1341 | /* | |
1342 | * @p is alive and it's gonna stop, continue or exit, so | |
1343 | * there always is something to wait for. | |
f470021a | 1344 | */ |
9e8ae01d | 1345 | wo->notask_error = 0; |
f470021a RM |
1346 | } |
1347 | ||
98abed02 | 1348 | /* |
45cb24a1 TH |
1349 | * Wait for stopped. Depending on @ptrace, different stopped state |
1350 | * is used and the two don't interact with each other. | |
98abed02 | 1351 | */ |
19e27463 TH |
1352 | ret = wait_task_stopped(wo, ptrace, p); |
1353 | if (ret) | |
1354 | return ret; | |
98abed02 RM |
1355 | |
1356 | /* | |
45cb24a1 TH |
1357 | * Wait for continued. There's only one continued state and the |
1358 | * ptracer can consume it which can confuse the real parent. Don't | |
1359 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1360 | */ |
9e8ae01d | 1361 | return wait_task_continued(wo, p); |
98abed02 RM |
1362 | } |
1363 | ||
1364 | /* | |
1365 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1366 | * | |
9e8ae01d | 1367 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1368 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1369 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1370 | * ->notask_error is 0 if there were any eligible children, |
3a2f5a59 | 1371 | * or still -ECHILD. |
98abed02 | 1372 | */ |
9e8ae01d | 1373 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1374 | { |
1375 | struct task_struct *p; | |
1376 | ||
1377 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1378 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1379 | |
9cd80bbb ON |
1380 | if (ret) |
1381 | return ret; | |
98abed02 RM |
1382 | } |
1383 | ||
1384 | return 0; | |
1385 | } | |
1386 | ||
9e8ae01d | 1387 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1388 | { |
1389 | struct task_struct *p; | |
1390 | ||
f470021a | 1391 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1392 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1393 | |
f470021a | 1394 | if (ret) |
98abed02 | 1395 | return ret; |
98abed02 RM |
1396 | } |
1397 | ||
1398 | return 0; | |
1399 | } | |
1400 | ||
ac6424b9 | 1401 | static int child_wait_callback(wait_queue_entry_t *wait, unsigned mode, |
0b7570e7 ON |
1402 | int sync, void *key) |
1403 | { | |
1404 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1405 | child_wait); | |
1406 | struct task_struct *p = key; | |
1407 | ||
5c01ba49 | 1408 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1409 | return 0; |
1410 | ||
b4fe5182 ON |
1411 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1412 | return 0; | |
1413 | ||
0b7570e7 ON |
1414 | return default_wake_function(wait, mode, sync, key); |
1415 | } | |
1416 | ||
a7f0765e ON |
1417 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1418 | { | |
0b7570e7 | 1419 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
ce4dd442 | 1420 | TASK_INTERRUPTIBLE, p); |
a7f0765e ON |
1421 | } |
1422 | ||
9e8ae01d | 1423 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1424 | { |
1da177e4 | 1425 | struct task_struct *tsk; |
98abed02 | 1426 | int retval; |
1da177e4 | 1427 | |
9e8ae01d | 1428 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1429 | |
0b7570e7 ON |
1430 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1431 | wo->child_wait.private = current; | |
1432 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1433 | repeat: |
98abed02 | 1434 | /* |
3da56d16 | 1435 | * If there is nothing that can match our criteria, just get out. |
9e8ae01d ON |
1436 | * We will clear ->notask_error to zero if we see any child that |
1437 | * might later match our criteria, even if we are not able to reap | |
1438 | * it yet. | |
98abed02 | 1439 | */ |
64a16caf | 1440 | wo->notask_error = -ECHILD; |
9e8ae01d | 1441 | if ((wo->wo_type < PIDTYPE_MAX) && |
1722c14a | 1442 | (!wo->wo_pid || !pid_has_task(wo->wo_pid, wo->wo_type))) |
64a16caf | 1443 | goto notask; |
161550d7 | 1444 | |
f95d39d1 | 1445 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1446 | read_lock(&tasklist_lock); |
1447 | tsk = current; | |
1448 | do { | |
64a16caf ON |
1449 | retval = do_wait_thread(wo, tsk); |
1450 | if (retval) | |
1451 | goto end; | |
9e8ae01d | 1452 | |
64a16caf ON |
1453 | retval = ptrace_do_wait(wo, tsk); |
1454 | if (retval) | |
98abed02 | 1455 | goto end; |
98abed02 | 1456 | |
9e8ae01d | 1457 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1458 | break; |
a3f6dfb7 | 1459 | } while_each_thread(current, tsk); |
1da177e4 | 1460 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1461 | |
64a16caf | 1462 | notask: |
9e8ae01d ON |
1463 | retval = wo->notask_error; |
1464 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1465 | retval = -ERESTARTSYS; |
98abed02 RM |
1466 | if (!signal_pending(current)) { |
1467 | schedule(); | |
1468 | goto repeat; | |
1469 | } | |
1da177e4 | 1470 | } |
1da177e4 | 1471 | end: |
f95d39d1 | 1472 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1473 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1474 | return retval; |
1475 | } | |
1476 | ||
3695eae5 CB |
1477 | static struct pid *pidfd_get_pid(unsigned int fd) |
1478 | { | |
1479 | struct fd f; | |
1480 | struct pid *pid; | |
1481 | ||
1482 | f = fdget(fd); | |
1483 | if (!f.file) | |
1484 | return ERR_PTR(-EBADF); | |
1485 | ||
1486 | pid = pidfd_pid(f.file); | |
1487 | if (!IS_ERR(pid)) | |
1488 | get_pid(pid); | |
1489 | ||
1490 | fdput(f); | |
1491 | return pid; | |
1492 | } | |
1493 | ||
67d7ddde | 1494 | static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, |
ce72a16f | 1495 | int options, struct rusage *ru) |
1da177e4 | 1496 | { |
9e8ae01d | 1497 | struct wait_opts wo; |
161550d7 EB |
1498 | struct pid *pid = NULL; |
1499 | enum pid_type type; | |
1da177e4 LT |
1500 | long ret; |
1501 | ||
91c4e8ea ON |
1502 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED| |
1503 | __WNOTHREAD|__WCLONE|__WALL)) | |
1da177e4 LT |
1504 | return -EINVAL; |
1505 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1506 | return -EINVAL; | |
1507 | ||
1508 | switch (which) { | |
1509 | case P_ALL: | |
161550d7 | 1510 | type = PIDTYPE_MAX; |
1da177e4 LT |
1511 | break; |
1512 | case P_PID: | |
161550d7 EB |
1513 | type = PIDTYPE_PID; |
1514 | if (upid <= 0) | |
1da177e4 | 1515 | return -EINVAL; |
3695eae5 CB |
1516 | |
1517 | pid = find_get_pid(upid); | |
1da177e4 LT |
1518 | break; |
1519 | case P_PGID: | |
161550d7 | 1520 | type = PIDTYPE_PGID; |
821cc7b0 | 1521 | if (upid < 0) |
1da177e4 | 1522 | return -EINVAL; |
3695eae5 | 1523 | |
821cc7b0 EB |
1524 | if (upid) |
1525 | pid = find_get_pid(upid); | |
1526 | else | |
1527 | pid = get_task_pid(current, PIDTYPE_PGID); | |
3695eae5 CB |
1528 | break; |
1529 | case P_PIDFD: | |
1530 | type = PIDTYPE_PID; | |
1531 | if (upid < 0) | |
1da177e4 | 1532 | return -EINVAL; |
3695eae5 CB |
1533 | |
1534 | pid = pidfd_get_pid(upid); | |
1535 | if (IS_ERR(pid)) | |
1536 | return PTR_ERR(pid); | |
1da177e4 LT |
1537 | break; |
1538 | default: | |
1539 | return -EINVAL; | |
1540 | } | |
1541 | ||
9e8ae01d ON |
1542 | wo.wo_type = type; |
1543 | wo.wo_pid = pid; | |
1544 | wo.wo_flags = options; | |
1545 | wo.wo_info = infop; | |
9e8ae01d ON |
1546 | wo.wo_rusage = ru; |
1547 | ret = do_wait(&wo); | |
dfe16dfa | 1548 | |
161550d7 | 1549 | put_pid(pid); |
1da177e4 LT |
1550 | return ret; |
1551 | } | |
1552 | ||
ce72a16f AV |
1553 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1554 | infop, int, options, struct rusage __user *, ru) | |
1555 | { | |
1556 | struct rusage r; | |
67d7ddde AV |
1557 | struct waitid_info info = {.status = 0}; |
1558 | long err = kernel_waitid(which, upid, &info, options, ru ? &r : NULL); | |
634a8160 | 1559 | int signo = 0; |
6c85501f | 1560 | |
634a8160 AV |
1561 | if (err > 0) { |
1562 | signo = SIGCHLD; | |
1563 | err = 0; | |
ce72a16f AV |
1564 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) |
1565 | return -EFAULT; | |
1566 | } | |
67d7ddde AV |
1567 | if (!infop) |
1568 | return err; | |
1569 | ||
41cd7805 | 1570 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1571 | return -EFAULT; |
96ca579a | 1572 | |
634a8160 | 1573 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1574 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1575 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1576 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1577 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1578 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1579 | user_write_access_end(); |
ce72a16f | 1580 | return err; |
4c48abe9 | 1581 | Efault: |
41cd7805 | 1582 | user_write_access_end(); |
4c48abe9 | 1583 | return -EFAULT; |
ce72a16f AV |
1584 | } |
1585 | ||
92ebce5a AV |
1586 | long kernel_wait4(pid_t upid, int __user *stat_addr, int options, |
1587 | struct rusage *ru) | |
1da177e4 | 1588 | { |
9e8ae01d | 1589 | struct wait_opts wo; |
161550d7 EB |
1590 | struct pid *pid = NULL; |
1591 | enum pid_type type; | |
1da177e4 LT |
1592 | long ret; |
1593 | ||
1594 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1595 | __WNOTHREAD|__WCLONE|__WALL)) | |
1596 | return -EINVAL; | |
161550d7 | 1597 | |
dd83c161 | 1598 | /* -INT_MIN is not defined */ |
1599 | if (upid == INT_MIN) | |
1600 | return -ESRCH; | |
1601 | ||
161550d7 EB |
1602 | if (upid == -1) |
1603 | type = PIDTYPE_MAX; | |
1604 | else if (upid < 0) { | |
1605 | type = PIDTYPE_PGID; | |
1606 | pid = find_get_pid(-upid); | |
1607 | } else if (upid == 0) { | |
1608 | type = PIDTYPE_PGID; | |
2ae448ef | 1609 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1610 | } else /* upid > 0 */ { |
1611 | type = PIDTYPE_PID; | |
1612 | pid = find_get_pid(upid); | |
1613 | } | |
1614 | ||
9e8ae01d ON |
1615 | wo.wo_type = type; |
1616 | wo.wo_pid = pid; | |
1617 | wo.wo_flags = options | WEXITED; | |
1618 | wo.wo_info = NULL; | |
359566fa | 1619 | wo.wo_stat = 0; |
9e8ae01d ON |
1620 | wo.wo_rusage = ru; |
1621 | ret = do_wait(&wo); | |
161550d7 | 1622 | put_pid(pid); |
359566fa AV |
1623 | if (ret > 0 && stat_addr && put_user(wo.wo_stat, stat_addr)) |
1624 | ret = -EFAULT; | |
1da177e4 | 1625 | |
1da177e4 LT |
1626 | return ret; |
1627 | } | |
1628 | ||
ce72a16f AV |
1629 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1630 | int, options, struct rusage __user *, ru) | |
1631 | { | |
1632 | struct rusage r; | |
1633 | long err = kernel_wait4(upid, stat_addr, options, ru ? &r : NULL); | |
1634 | ||
1635 | if (err > 0) { | |
1636 | if (ru && copy_to_user(ru, &r, sizeof(struct rusage))) | |
1637 | return -EFAULT; | |
1638 | } | |
1639 | return err; | |
1640 | } | |
1641 | ||
1da177e4 LT |
1642 | #ifdef __ARCH_WANT_SYS_WAITPID |
1643 | ||
1644 | /* | |
1645 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1646 | * implemented by calling sys_wait4() from libc.a. | |
1647 | */ | |
17da2bd9 | 1648 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 | 1649 | { |
d300b610 | 1650 | return kernel_wait4(pid, stat_addr, options, NULL); |
1da177e4 LT |
1651 | } |
1652 | ||
1653 | #endif | |
7e95a225 AV |
1654 | |
1655 | #ifdef CONFIG_COMPAT | |
1656 | COMPAT_SYSCALL_DEFINE4(wait4, | |
1657 | compat_pid_t, pid, | |
1658 | compat_uint_t __user *, stat_addr, | |
1659 | int, options, | |
1660 | struct compat_rusage __user *, ru) | |
1661 | { | |
ce72a16f AV |
1662 | struct rusage r; |
1663 | long err = kernel_wait4(pid, stat_addr, options, ru ? &r : NULL); | |
1664 | if (err > 0) { | |
1665 | if (ru && put_compat_rusage(&r, ru)) | |
1666 | return -EFAULT; | |
7e95a225 | 1667 | } |
ce72a16f | 1668 | return err; |
7e95a225 AV |
1669 | } |
1670 | ||
1671 | COMPAT_SYSCALL_DEFINE5(waitid, | |
1672 | int, which, compat_pid_t, pid, | |
1673 | struct compat_siginfo __user *, infop, int, options, | |
1674 | struct compat_rusage __user *, uru) | |
1675 | { | |
7e95a225 | 1676 | struct rusage ru; |
67d7ddde AV |
1677 | struct waitid_info info = {.status = 0}; |
1678 | long err = kernel_waitid(which, pid, &info, options, uru ? &ru : NULL); | |
634a8160 AV |
1679 | int signo = 0; |
1680 | if (err > 0) { | |
1681 | signo = SIGCHLD; | |
1682 | err = 0; | |
6c85501f AV |
1683 | if (uru) { |
1684 | /* kernel_waitid() overwrites everything in ru */ | |
1685 | if (COMPAT_USE_64BIT_TIME) | |
1686 | err = copy_to_user(uru, &ru, sizeof(ru)); | |
1687 | else | |
1688 | err = put_compat_rusage(&ru, uru); | |
1689 | if (err) | |
1690 | return -EFAULT; | |
1691 | } | |
7e95a225 AV |
1692 | } |
1693 | ||
4c48abe9 AV |
1694 | if (!infop) |
1695 | return err; | |
1696 | ||
41cd7805 | 1697 | if (!user_write_access_begin(infop, sizeof(*infop))) |
1c9fec47 | 1698 | return -EFAULT; |
96ca579a | 1699 | |
634a8160 | 1700 | unsafe_put_user(signo, &infop->si_signo, Efault); |
4c48abe9 | 1701 | unsafe_put_user(0, &infop->si_errno, Efault); |
cc731525 | 1702 | unsafe_put_user(info.cause, &infop->si_code, Efault); |
4c48abe9 AV |
1703 | unsafe_put_user(info.pid, &infop->si_pid, Efault); |
1704 | unsafe_put_user(info.uid, &infop->si_uid, Efault); | |
1705 | unsafe_put_user(info.status, &infop->si_status, Efault); | |
41cd7805 | 1706 | user_write_access_end(); |
67d7ddde | 1707 | return err; |
4c48abe9 | 1708 | Efault: |
41cd7805 | 1709 | user_write_access_end(); |
4c48abe9 | 1710 | return -EFAULT; |
7e95a225 AV |
1711 | } |
1712 | #endif | |
7c2c11b2 SM |
1713 | |
1714 | __weak void abort(void) | |
1715 | { | |
1716 | BUG(); | |
1717 | ||
1718 | /* if that doesn't kill us, halt */ | |
1719 | panic("Oops failed to kill thread"); | |
1720 | } | |
dc8635b7 | 1721 | EXPORT_SYMBOL(abort); |