Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/exit.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
1da177e4 LT |
7 | #include <linux/mm.h> |
8 | #include <linux/slab.h> | |
9 | #include <linux/interrupt.h> | |
1da177e4 | 10 | #include <linux/module.h> |
c59ede7b | 11 | #include <linux/capability.h> |
1da177e4 LT |
12 | #include <linux/completion.h> |
13 | #include <linux/personality.h> | |
14 | #include <linux/tty.h> | |
da9cbc87 | 15 | #include <linux/iocontext.h> |
1da177e4 LT |
16 | #include <linux/key.h> |
17 | #include <linux/security.h> | |
18 | #include <linux/cpu.h> | |
19 | #include <linux/acct.h> | |
8f0ab514 | 20 | #include <linux/tsacct_kern.h> |
1da177e4 | 21 | #include <linux/file.h> |
9f3acc31 | 22 | #include <linux/fdtable.h> |
80d26af8 | 23 | #include <linux/freezer.h> |
1da177e4 | 24 | #include <linux/binfmts.h> |
ab516013 | 25 | #include <linux/nsproxy.h> |
84d73786 | 26 | #include <linux/pid_namespace.h> |
1da177e4 LT |
27 | #include <linux/ptrace.h> |
28 | #include <linux/profile.h> | |
29 | #include <linux/mount.h> | |
30 | #include <linux/proc_fs.h> | |
49d769d5 | 31 | #include <linux/kthread.h> |
1da177e4 | 32 | #include <linux/mempolicy.h> |
c757249a | 33 | #include <linux/taskstats_kern.h> |
ca74e92b | 34 | #include <linux/delayacct.h> |
b4f48b63 | 35 | #include <linux/cgroup.h> |
1da177e4 | 36 | #include <linux/syscalls.h> |
7ed20e1a | 37 | #include <linux/signal.h> |
6a14c5c9 | 38 | #include <linux/posix-timers.h> |
9f46080c | 39 | #include <linux/cn_proc.h> |
de5097c2 | 40 | #include <linux/mutex.h> |
0771dfef | 41 | #include <linux/futex.h> |
b92ce558 | 42 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 43 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 44 | #include <linux/resource.h> |
0d67a46d | 45 | #include <linux/blkdev.h> |
6eaeeaba | 46 | #include <linux/task_io_accounting_ops.h> |
30199f5a | 47 | #include <linux/tracehook.h> |
5ad4e53b | 48 | #include <linux/fs_struct.h> |
d84f4f99 | 49 | #include <linux/init_task.h> |
cdd6c482 | 50 | #include <linux/perf_event.h> |
ad8d75ff | 51 | #include <trace/events/sched.h> |
24f1e32c | 52 | #include <linux/hw_breakpoint.h> |
3d5992d2 | 53 | #include <linux/oom.h> |
54848d73 | 54 | #include <linux/writeback.h> |
40401530 | 55 | #include <linux/shm.h> |
1da177e4 LT |
56 | |
57 | #include <asm/uaccess.h> | |
58 | #include <asm/unistd.h> | |
59 | #include <asm/pgtable.h> | |
60 | #include <asm/mmu_context.h> | |
61 | ||
a0be55de | 62 | static void exit_mm(struct task_struct *tsk); |
408b664a | 63 | |
d40e48e0 | 64 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
65 | { |
66 | nr_threads--; | |
50d75f8d | 67 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 68 | if (group_dead) { |
1da177e4 LT |
69 | detach_pid(p, PIDTYPE_PGID); |
70 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 71 | |
5e85d4ab | 72 | list_del_rcu(&p->tasks); |
9cd80bbb | 73 | list_del_init(&p->sibling); |
909ea964 | 74 | __this_cpu_dec(process_counts); |
1da177e4 | 75 | } |
47e65328 | 76 | list_del_rcu(&p->thread_group); |
0c740d0a | 77 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
78 | } |
79 | ||
6a14c5c9 ON |
80 | /* |
81 | * This function expects the tasklist_lock write-locked. | |
82 | */ | |
83 | static void __exit_signal(struct task_struct *tsk) | |
84 | { | |
85 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 86 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 87 | struct sighand_struct *sighand; |
4ada856f | 88 | struct tty_struct *uninitialized_var(tty); |
6fac4829 | 89 | cputime_t utime, stime; |
6a14c5c9 | 90 | |
d11c563d | 91 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 92 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
93 | spin_lock(&sighand->siglock); |
94 | ||
95 | posix_cpu_timers_exit(tsk); | |
d40e48e0 | 96 | if (group_dead) { |
6a14c5c9 | 97 | posix_cpu_timers_exit_group(tsk); |
4ada856f ON |
98 | tty = sig->tty; |
99 | sig->tty = NULL; | |
4a599942 | 100 | } else { |
e0a70217 ON |
101 | /* |
102 | * This can only happen if the caller is de_thread(). | |
103 | * FIXME: this is the temporary hack, we should teach | |
104 | * posix-cpu-timers to handle this case correctly. | |
105 | */ | |
106 | if (unlikely(has_group_leader_pid(tsk))) | |
107 | posix_cpu_timers_exit_group(tsk); | |
108 | ||
6a14c5c9 ON |
109 | /* |
110 | * If there is any task waiting for the group exit | |
111 | * then notify it: | |
112 | */ | |
d344193a | 113 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 114 | wake_up_process(sig->group_exit_task); |
6db840fa | 115 | |
6a14c5c9 ON |
116 | if (tsk == sig->curr_target) |
117 | sig->curr_target = next_thread(tsk); | |
6a14c5c9 ON |
118 | } |
119 | ||
90ed9cbe | 120 | /* |
26e75b5c ON |
121 | * Accumulate here the counters for all threads as they die. We could |
122 | * skip the group leader because it is the last user of signal_struct, | |
123 | * but we want to avoid the race with thread_group_cputime() which can | |
124 | * see the empty ->thread_head list. | |
90ed9cbe RR |
125 | */ |
126 | task_cputime(tsk, &utime, &stime); | |
e78c3496 | 127 | write_seqlock(&sig->stats_lock); |
90ed9cbe RR |
128 | sig->utime += utime; |
129 | sig->stime += stime; | |
130 | sig->gtime += task_gtime(tsk); | |
131 | sig->min_flt += tsk->min_flt; | |
132 | sig->maj_flt += tsk->maj_flt; | |
133 | sig->nvcsw += tsk->nvcsw; | |
134 | sig->nivcsw += tsk->nivcsw; | |
135 | sig->inblock += task_io_get_inblock(tsk); | |
136 | sig->oublock += task_io_get_oublock(tsk); | |
137 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | |
138 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | |
b3ac022c | 139 | sig->nr_threads--; |
d40e48e0 | 140 | __unhash_process(tsk, group_dead); |
e78c3496 | 141 | write_sequnlock(&sig->stats_lock); |
5876700c | 142 | |
da7978b0 ON |
143 | /* |
144 | * Do this under ->siglock, we can race with another thread | |
145 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
146 | */ | |
147 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 148 | tsk->sighand = NULL; |
6a14c5c9 | 149 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 150 | |
a7e5328a | 151 | __cleanup_sighand(sighand); |
a0be55de | 152 | clear_tsk_thread_flag(tsk, TIF_SIGPENDING); |
d40e48e0 | 153 | if (group_dead) { |
6a14c5c9 | 154 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 155 | tty_kref_put(tty); |
6a14c5c9 ON |
156 | } |
157 | } | |
158 | ||
8c7904a0 EB |
159 | static void delayed_put_task_struct(struct rcu_head *rhp) |
160 | { | |
0a16b607 MD |
161 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
162 | ||
4e231c79 | 163 | perf_event_delayed_put(tsk); |
0a16b607 MD |
164 | trace_sched_process_free(tsk); |
165 | put_task_struct(tsk); | |
8c7904a0 EB |
166 | } |
167 | ||
f470021a | 168 | |
a0be55de | 169 | void release_task(struct task_struct *p) |
1da177e4 | 170 | { |
36c8b586 | 171 | struct task_struct *leader; |
1da177e4 | 172 | int zap_leader; |
1f09f974 | 173 | repeat: |
c69e8d9c | 174 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
175 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
176 | rcu_read_lock(); | |
c69e8d9c | 177 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 178 | rcu_read_unlock(); |
c69e8d9c | 179 | |
60347f67 | 180 | proc_flush_task(p); |
0203026b | 181 | |
1da177e4 | 182 | write_lock_irq(&tasklist_lock); |
a288eecc | 183 | ptrace_release_task(p); |
1da177e4 | 184 | __exit_signal(p); |
35f5cad8 | 185 | |
1da177e4 LT |
186 | /* |
187 | * If we are the last non-leader member of the thread | |
188 | * group, and the leader is zombie, then notify the | |
189 | * group leader's parent process. (if it wants notification.) | |
190 | */ | |
191 | zap_leader = 0; | |
192 | leader = p->group_leader; | |
a0be55de IA |
193 | if (leader != p && thread_group_empty(leader) |
194 | && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
195 | /* |
196 | * If we were the last child thread and the leader has | |
197 | * exited already, and the leader's parent ignores SIGCHLD, | |
198 | * then we are the one who should release the leader. | |
dae33574 | 199 | */ |
86773473 | 200 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
201 | if (zap_leader) |
202 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
203 | } |
204 | ||
1da177e4 | 205 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 206 | release_thread(p); |
8c7904a0 | 207 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
208 | |
209 | p = leader; | |
210 | if (unlikely(zap_leader)) | |
211 | goto repeat; | |
212 | } | |
213 | ||
1da177e4 LT |
214 | /* |
215 | * This checks not only the pgrp, but falls back on the pid if no | |
216 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
217 | * without this... | |
04a2e6a5 EB |
218 | * |
219 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 220 | */ |
04a2e6a5 | 221 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
222 | { |
223 | struct task_struct *p; | |
04a2e6a5 | 224 | struct pid *sid = NULL; |
62dfb554 | 225 | |
04a2e6a5 | 226 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 227 | if (p == NULL) |
04a2e6a5 | 228 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 229 | if (p != NULL) |
04a2e6a5 | 230 | sid = task_session(p); |
62dfb554 | 231 | |
1da177e4 LT |
232 | return sid; |
233 | } | |
234 | ||
235 | /* | |
236 | * Determine if a process group is "orphaned", according to the POSIX | |
237 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
238 | * by terminal-generated stop signals. Newly orphaned process groups are | |
239 | * to receive a SIGHUP and a SIGCONT. | |
240 | * | |
241 | * "I ask you, have you ever known what it is to be an orphan?" | |
242 | */ | |
a0be55de IA |
243 | static int will_become_orphaned_pgrp(struct pid *pgrp, |
244 | struct task_struct *ignored_task) | |
1da177e4 LT |
245 | { |
246 | struct task_struct *p; | |
1da177e4 | 247 | |
0475ac08 | 248 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
249 | if ((p == ignored_task) || |
250 | (p->exit_state && thread_group_empty(p)) || | |
251 | is_global_init(p->real_parent)) | |
1da177e4 | 252 | continue; |
05e83df6 | 253 | |
0475ac08 | 254 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
255 | task_session(p->real_parent) == task_session(p)) |
256 | return 0; | |
0475ac08 | 257 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
258 | |
259 | return 1; | |
1da177e4 LT |
260 | } |
261 | ||
3e7cd6c4 | 262 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
263 | { |
264 | int retval; | |
265 | ||
266 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 267 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
268 | read_unlock(&tasklist_lock); |
269 | ||
270 | return retval; | |
271 | } | |
272 | ||
961c4675 | 273 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 274 | { |
1da177e4 LT |
275 | struct task_struct *p; |
276 | ||
0475ac08 | 277 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
278 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
279 | return true; | |
0475ac08 | 280 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
281 | |
282 | return false; | |
1da177e4 LT |
283 | } |
284 | ||
f49ee505 ON |
285 | /* |
286 | * Check to see if any process groups have become orphaned as | |
287 | * a result of our exiting, and if they have any stopped jobs, | |
288 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
289 | */ | |
290 | static void | |
291 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
292 | { | |
293 | struct pid *pgrp = task_pgrp(tsk); | |
294 | struct task_struct *ignored_task = tsk; | |
295 | ||
296 | if (!parent) | |
a0be55de IA |
297 | /* exit: our father is in a different pgrp than |
298 | * we are and we were the only connection outside. | |
299 | */ | |
f49ee505 ON |
300 | parent = tsk->real_parent; |
301 | else | |
302 | /* reparent: our child is in a different pgrp than | |
303 | * we are, and it was the only connection outside. | |
304 | */ | |
305 | ignored_task = NULL; | |
306 | ||
307 | if (task_pgrp(parent) != pgrp && | |
308 | task_session(parent) == task_session(tsk) && | |
309 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
310 | has_stopped_jobs(pgrp)) { | |
311 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
312 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
313 | } | |
314 | } | |
315 | ||
f98bafa0 | 316 | #ifdef CONFIG_MEMCG |
cf475ad2 | 317 | /* |
733eda7a | 318 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 319 | */ |
cf475ad2 BS |
320 | void mm_update_next_owner(struct mm_struct *mm) |
321 | { | |
322 | struct task_struct *c, *g, *p = current; | |
323 | ||
324 | retry: | |
733eda7a KH |
325 | /* |
326 | * If the exiting or execing task is not the owner, it's | |
327 | * someone else's problem. | |
328 | */ | |
329 | if (mm->owner != p) | |
cf475ad2 | 330 | return; |
733eda7a KH |
331 | /* |
332 | * The current owner is exiting/execing and there are no other | |
333 | * candidates. Do not leave the mm pointing to a possibly | |
334 | * freed task structure. | |
335 | */ | |
336 | if (atomic_read(&mm->mm_users) <= 1) { | |
337 | mm->owner = NULL; | |
338 | return; | |
339 | } | |
cf475ad2 BS |
340 | |
341 | read_lock(&tasklist_lock); | |
342 | /* | |
343 | * Search in the children | |
344 | */ | |
345 | list_for_each_entry(c, &p->children, sibling) { | |
346 | if (c->mm == mm) | |
347 | goto assign_new_owner; | |
348 | } | |
349 | ||
350 | /* | |
351 | * Search in the siblings | |
352 | */ | |
dea33cfd | 353 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
354 | if (c->mm == mm) |
355 | goto assign_new_owner; | |
356 | } | |
357 | ||
358 | /* | |
f87fb599 | 359 | * Search through everything else, we should not get here often. |
cf475ad2 | 360 | */ |
39af1765 ON |
361 | for_each_process(g) { |
362 | if (g->flags & PF_KTHREAD) | |
363 | continue; | |
364 | for_each_thread(g, c) { | |
365 | if (c->mm == mm) | |
366 | goto assign_new_owner; | |
367 | if (c->mm) | |
368 | break; | |
369 | } | |
f87fb599 | 370 | } |
cf475ad2 | 371 | read_unlock(&tasklist_lock); |
31a78f23 BS |
372 | /* |
373 | * We found no owner yet mm_users > 1: this implies that we are | |
374 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 375 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 376 | */ |
31a78f23 | 377 | mm->owner = NULL; |
cf475ad2 BS |
378 | return; |
379 | ||
380 | assign_new_owner: | |
381 | BUG_ON(c == p); | |
382 | get_task_struct(c); | |
383 | /* | |
384 | * The task_lock protects c->mm from changing. | |
385 | * We always want mm->owner->mm == mm | |
386 | */ | |
387 | task_lock(c); | |
e5991371 HD |
388 | /* |
389 | * Delay read_unlock() till we have the task_lock() | |
390 | * to ensure that c does not slip away underneath us | |
391 | */ | |
392 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
393 | if (c->mm != mm) { |
394 | task_unlock(c); | |
395 | put_task_struct(c); | |
396 | goto retry; | |
397 | } | |
cf475ad2 BS |
398 | mm->owner = c; |
399 | task_unlock(c); | |
400 | put_task_struct(c); | |
401 | } | |
f98bafa0 | 402 | #endif /* CONFIG_MEMCG */ |
cf475ad2 | 403 | |
1da177e4 LT |
404 | /* |
405 | * Turn us into a lazy TLB process if we | |
406 | * aren't already.. | |
407 | */ | |
a0be55de | 408 | static void exit_mm(struct task_struct *tsk) |
1da177e4 LT |
409 | { |
410 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 411 | struct core_state *core_state; |
1da177e4 | 412 | |
48d212a2 | 413 | mm_release(tsk, mm); |
1da177e4 LT |
414 | if (!mm) |
415 | return; | |
4fe7efdb | 416 | sync_mm_rss(mm); |
1da177e4 LT |
417 | /* |
418 | * Serialize with any possible pending coredump. | |
999d9fc1 | 419 | * We must hold mmap_sem around checking core_state |
1da177e4 | 420 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 421 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
422 | * group with ->mm != NULL. |
423 | */ | |
424 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
425 | core_state = mm->core_state; |
426 | if (core_state) { | |
427 | struct core_thread self; | |
a0be55de | 428 | |
1da177e4 | 429 | up_read(&mm->mmap_sem); |
1da177e4 | 430 | |
b564daf8 ON |
431 | self.task = tsk; |
432 | self.next = xchg(&core_state->dumper.next, &self); | |
433 | /* | |
434 | * Implies mb(), the result of xchg() must be visible | |
435 | * to core_state->dumper. | |
436 | */ | |
437 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
438 | complete(&core_state->startup); | |
1da177e4 | 439 | |
a94e2d40 ON |
440 | for (;;) { |
441 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
442 | if (!self.task) /* see coredump_finish() */ | |
443 | break; | |
80d26af8 | 444 | freezable_schedule(); |
a94e2d40 ON |
445 | } |
446 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
447 | down_read(&mm->mmap_sem); |
448 | } | |
449 | atomic_inc(&mm->mm_count); | |
125e1874 | 450 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
451 | /* more a memory barrier than a real lock */ |
452 | task_lock(tsk); | |
453 | tsk->mm = NULL; | |
454 | up_read(&mm->mmap_sem); | |
455 | enter_lazy_tlb(mm, current); | |
456 | task_unlock(tsk); | |
cf475ad2 | 457 | mm_update_next_owner(mm); |
1da177e4 | 458 | mmput(mm); |
fb794bcb | 459 | clear_thread_flag(TIF_MEMDIE); |
1da177e4 LT |
460 | } |
461 | ||
c9dc05bf ON |
462 | static struct task_struct *find_alive_thread(struct task_struct *p) |
463 | { | |
464 | struct task_struct *t; | |
465 | ||
466 | for_each_thread(p, t) { | |
467 | if (!(t->flags & PF_EXITING)) | |
468 | return t; | |
469 | } | |
470 | return NULL; | |
471 | } | |
472 | ||
1109909c ON |
473 | static struct task_struct *find_child_reaper(struct task_struct *father) |
474 | __releases(&tasklist_lock) | |
475 | __acquires(&tasklist_lock) | |
476 | { | |
477 | struct pid_namespace *pid_ns = task_active_pid_ns(father); | |
478 | struct task_struct *reaper = pid_ns->child_reaper; | |
479 | ||
480 | if (likely(reaper != father)) | |
481 | return reaper; | |
482 | ||
c9dc05bf ON |
483 | reaper = find_alive_thread(father); |
484 | if (reaper) { | |
1109909c ON |
485 | pid_ns->child_reaper = reaper; |
486 | return reaper; | |
487 | } | |
488 | ||
489 | write_unlock_irq(&tasklist_lock); | |
490 | if (unlikely(pid_ns == &init_pid_ns)) { | |
491 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
492 | father->signal->group_exit_code ?: father->exit_code); | |
493 | } | |
494 | zap_pid_ns_processes(pid_ns); | |
495 | write_lock_irq(&tasklist_lock); | |
496 | ||
497 | return father; | |
498 | } | |
499 | ||
1da177e4 | 500 | /* |
ebec18a6 LP |
501 | * When we die, we re-parent all our children, and try to: |
502 | * 1. give them to another thread in our thread group, if such a member exists | |
503 | * 2. give it to the first ancestor process which prctl'd itself as a | |
504 | * child_subreaper for its children (like a service manager) | |
505 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 506 | */ |
1109909c ON |
507 | static struct task_struct *find_new_reaper(struct task_struct *father, |
508 | struct task_struct *child_reaper) | |
1da177e4 | 509 | { |
c9dc05bf | 510 | struct task_struct *thread, *reaper; |
1da177e4 | 511 | |
c9dc05bf ON |
512 | thread = find_alive_thread(father); |
513 | if (thread) | |
950bbabb | 514 | return thread; |
1da177e4 | 515 | |
7d24e2df | 516 | if (father->signal->has_child_subreaper) { |
ebec18a6 | 517 | /* |
175aed3f ON |
518 | * Find the first ->is_child_subreaper ancestor in our pid_ns. |
519 | * We start from father to ensure we can not look into another | |
520 | * namespace, this is safe because all its threads are dead. | |
ebec18a6 | 521 | */ |
7d24e2df | 522 | for (reaper = father; |
1109909c | 523 | !same_thread_group(reaper, child_reaper); |
ebec18a6 | 524 | reaper = reaper->real_parent) { |
175aed3f ON |
525 | /* call_usermodehelper() descendants need this check */ |
526 | if (reaper == &init_task) | |
ebec18a6 LP |
527 | break; |
528 | if (!reaper->signal->is_child_subreaper) | |
529 | continue; | |
c9dc05bf ON |
530 | thread = find_alive_thread(reaper); |
531 | if (thread) | |
532 | return thread; | |
ebec18a6 | 533 | } |
1da177e4 | 534 | } |
762a24be | 535 | |
1109909c | 536 | return child_reaper; |
950bbabb ON |
537 | } |
538 | ||
5dfc80be ON |
539 | /* |
540 | * Any that need to be release_task'd are put on the @dead list. | |
541 | */ | |
9cd80bbb | 542 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
543 | struct list_head *dead) |
544 | { | |
2831096e | 545 | if (unlikely(p->exit_state == EXIT_DEAD)) |
5dfc80be ON |
546 | return; |
547 | ||
abd50b39 | 548 | /* We don't want people slaying init. */ |
5dfc80be ON |
549 | p->exit_signal = SIGCHLD; |
550 | ||
551 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 552 | if (!p->ptrace && |
5dfc80be | 553 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 554 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be | 555 | p->exit_state = EXIT_DEAD; |
dc2fd4b0 | 556 | list_add(&p->ptrace_entry, dead); |
5dfc80be ON |
557 | } |
558 | } | |
559 | ||
560 | kill_orphaned_pgrp(p, father); | |
561 | } | |
562 | ||
482a3767 ON |
563 | /* |
564 | * This does two things: | |
565 | * | |
566 | * A. Make init inherit all the child processes | |
567 | * B. Check to see if any process groups have become orphaned | |
568 | * as a result of our exiting, and if they have any stopped | |
569 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
570 | */ | |
571 | static void forget_original_parent(struct task_struct *father, | |
572 | struct list_head *dead) | |
1da177e4 | 573 | { |
482a3767 | 574 | struct task_struct *p, *t, *reaper; |
762a24be | 575 | |
7c8bd232 | 576 | if (unlikely(!list_empty(&father->ptraced))) |
482a3767 | 577 | exit_ptrace(father, dead); |
f470021a | 578 | |
7c8bd232 | 579 | /* Can drop and reacquire tasklist_lock */ |
1109909c | 580 | reaper = find_child_reaper(father); |
ad9e206a | 581 | if (list_empty(&father->children)) |
482a3767 | 582 | return; |
1109909c ON |
583 | |
584 | reaper = find_new_reaper(father, reaper); | |
2831096e | 585 | list_for_each_entry(p, &father->children, sibling) { |
57a05918 | 586 | for_each_thread(p, t) { |
9cd80bbb | 587 | t->real_parent = reaper; |
57a05918 ON |
588 | BUG_ON((!t->ptrace) != (t->parent == father)); |
589 | if (likely(!t->ptrace)) | |
9cd80bbb | 590 | t->parent = t->real_parent; |
9cd80bbb ON |
591 | if (t->pdeath_signal) |
592 | group_send_sig_info(t->pdeath_signal, | |
593 | SEND_SIG_NOINFO, t); | |
57a05918 | 594 | } |
2831096e ON |
595 | /* |
596 | * If this is a threaded reparent there is no need to | |
597 | * notify anyone anything has happened. | |
598 | */ | |
599 | if (!same_thread_group(reaper, father)) | |
482a3767 | 600 | reparent_leader(father, p, dead); |
1da177e4 | 601 | } |
2831096e | 602 | list_splice_tail_init(&father->children, &reaper->children); |
1da177e4 LT |
603 | } |
604 | ||
605 | /* | |
606 | * Send signals to all our closest relatives so that they know | |
607 | * to properly mourn us.. | |
608 | */ | |
821c7de7 | 609 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 610 | { |
53c8f9f1 | 611 | bool autoreap; |
482a3767 ON |
612 | struct task_struct *p, *n; |
613 | LIST_HEAD(dead); | |
1da177e4 | 614 | |
762a24be | 615 | write_lock_irq(&tasklist_lock); |
482a3767 ON |
616 | forget_original_parent(tsk, &dead); |
617 | ||
821c7de7 ON |
618 | if (group_dead) |
619 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 620 | |
45cdf5cc ON |
621 | if (unlikely(tsk->ptrace)) { |
622 | int sig = thread_group_leader(tsk) && | |
623 | thread_group_empty(tsk) && | |
624 | !ptrace_reparented(tsk) ? | |
625 | tsk->exit_signal : SIGCHLD; | |
626 | autoreap = do_notify_parent(tsk, sig); | |
627 | } else if (thread_group_leader(tsk)) { | |
628 | autoreap = thread_group_empty(tsk) && | |
629 | do_notify_parent(tsk, tsk->exit_signal); | |
630 | } else { | |
631 | autoreap = true; | |
632 | } | |
1da177e4 | 633 | |
53c8f9f1 | 634 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
1da177e4 | 635 | |
9c339168 ON |
636 | /* mt-exec, de_thread() is waiting for group leader */ |
637 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 638 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
639 | write_unlock_irq(&tasklist_lock); |
640 | ||
482a3767 ON |
641 | list_for_each_entry_safe(p, n, &dead, ptrace_entry) { |
642 | list_del_init(&p->ptrace_entry); | |
643 | release_task(p); | |
644 | } | |
645 | ||
1da177e4 | 646 | /* If the process is dead, release it - nobody will wait for it */ |
53c8f9f1 | 647 | if (autoreap) |
1da177e4 | 648 | release_task(tsk); |
1da177e4 LT |
649 | } |
650 | ||
e18eecb8 JD |
651 | #ifdef CONFIG_DEBUG_STACK_USAGE |
652 | static void check_stack_usage(void) | |
653 | { | |
654 | static DEFINE_SPINLOCK(low_water_lock); | |
655 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
656 | unsigned long free; |
657 | ||
7c9f8861 | 658 | free = stack_not_used(current); |
e18eecb8 JD |
659 | |
660 | if (free >= lowest_to_date) | |
661 | return; | |
662 | ||
663 | spin_lock(&low_water_lock); | |
664 | if (free < lowest_to_date) { | |
a0be55de IA |
665 | pr_warn("%s (%d) used greatest stack depth: %lu bytes left\n", |
666 | current->comm, task_pid_nr(current), free); | |
e18eecb8 JD |
667 | lowest_to_date = free; |
668 | } | |
669 | spin_unlock(&low_water_lock); | |
670 | } | |
671 | #else | |
672 | static inline void check_stack_usage(void) {} | |
673 | #endif | |
674 | ||
9402c95f | 675 | void do_exit(long code) |
1da177e4 LT |
676 | { |
677 | struct task_struct *tsk = current; | |
678 | int group_dead; | |
3f95aa81 | 679 | TASKS_RCU(int tasks_rcu_i); |
1da177e4 LT |
680 | |
681 | profile_task_exit(tsk); | |
682 | ||
73c10101 | 683 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 684 | |
1da177e4 LT |
685 | if (unlikely(in_interrupt())) |
686 | panic("Aiee, killing interrupt handler!"); | |
687 | if (unlikely(!tsk->pid)) | |
688 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 689 | |
33dd94ae NE |
690 | /* |
691 | * If do_exit is called because this processes oopsed, it's possible | |
692 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
693 | * continuing. Amongst other possible reasons, this is to prevent | |
694 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
695 | * kernel address. | |
696 | */ | |
697 | set_fs(USER_DS); | |
698 | ||
a288eecc | 699 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 700 | |
e0e81739 DH |
701 | validate_creds_for_do_exit(tsk); |
702 | ||
df164db5 AN |
703 | /* |
704 | * We're taking recursive faults here in do_exit. Safest is to just | |
705 | * leave this task alone and wait for reboot. | |
706 | */ | |
707 | if (unlikely(tsk->flags & PF_EXITING)) { | |
a0be55de | 708 | pr_alert("Fixing recursive fault but reboot is needed!\n"); |
778e9a9c AK |
709 | /* |
710 | * We can do this unlocked here. The futex code uses | |
711 | * this flag just to verify whether the pi state | |
712 | * cleanup has been done or not. In the worst case it | |
713 | * loops once more. We pretend that the cleanup was | |
714 | * done as there is no way to return. Either the | |
715 | * OWNER_DIED bit is set by now or we push the blocked | |
716 | * task into the wait for ever nirwana as well. | |
717 | */ | |
718 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
719 | set_current_state(TASK_UNINTERRUPTIBLE); |
720 | schedule(); | |
721 | } | |
722 | ||
d12619b5 | 723 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
724 | /* |
725 | * tsk->flags are checked in the futex code to protect against | |
ed3e694d | 726 | * an exiting task cleaning up the robust pi futexes. |
778e9a9c | 727 | */ |
d2ee7198 | 728 | smp_mb(); |
1d615482 | 729 | raw_spin_unlock_wait(&tsk->pi_lock); |
1da177e4 | 730 | |
1da177e4 | 731 | if (unlikely(in_atomic())) |
a0be55de IA |
732 | pr_info("note: %s[%d] exited with preempt_count %d\n", |
733 | current->comm, task_pid_nr(current), | |
734 | preempt_count()); | |
1da177e4 LT |
735 | |
736 | acct_update_integrals(tsk); | |
48d212a2 LT |
737 | /* sync mm's RSS info before statistics gathering */ |
738 | if (tsk->mm) | |
739 | sync_mm_rss(tsk->mm); | |
1da177e4 | 740 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 741 | if (group_dead) { |
778e9a9c | 742 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 743 | exit_itimers(tsk->signal); |
1f10206c JP |
744 | if (tsk->mm) |
745 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 746 | } |
f6ec29a4 | 747 | acct_collect(code, group_dead); |
522ed776 MT |
748 | if (group_dead) |
749 | tty_audit_exit(); | |
a4ff8dba | 750 | audit_free(tsk); |
115085ea | 751 | |
48d212a2 | 752 | tsk->exit_code = code; |
115085ea | 753 | taskstats_exit(tsk, group_dead); |
c757249a | 754 | |
1da177e4 LT |
755 | exit_mm(tsk); |
756 | ||
0e464814 | 757 | if (group_dead) |
f6ec29a4 | 758 | acct_process(); |
0a16b607 MD |
759 | trace_sched_process_exit(tsk); |
760 | ||
1da177e4 | 761 | exit_sem(tsk); |
b34a6b1d | 762 | exit_shm(tsk); |
1ec7f1dd AV |
763 | exit_files(tsk); |
764 | exit_fs(tsk); | |
c39df5fa ON |
765 | if (group_dead) |
766 | disassociate_ctty(1); | |
8aac6270 | 767 | exit_task_namespaces(tsk); |
ed3e694d | 768 | exit_task_work(tsk); |
1da177e4 | 769 | exit_thread(); |
0b3fcf17 SE |
770 | |
771 | /* | |
772 | * Flush inherited counters to the parent - before the parent | |
773 | * gets woken up by child-exit notifications. | |
774 | * | |
775 | * because of cgroup mode, must be called before cgroup_exit() | |
776 | */ | |
777 | perf_event_exit_task(tsk); | |
778 | ||
1ec41830 | 779 | cgroup_exit(tsk); |
1da177e4 | 780 | |
a1261f54 | 781 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 | 782 | |
24f1e32c FW |
783 | /* |
784 | * FIXME: do that only when needed, using sched_exit tracepoint | |
785 | */ | |
7c8df286 | 786 | flush_ptrace_hw_breakpoint(tsk); |
33b2fb30 | 787 | |
3f95aa81 | 788 | TASKS_RCU(tasks_rcu_i = __srcu_read_lock(&tasks_rcu_exit_srcu)); |
821c7de7 | 789 | exit_notify(tsk, group_dead); |
ef982393 | 790 | proc_exit_connector(tsk); |
1da177e4 | 791 | #ifdef CONFIG_NUMA |
c0ff7453 | 792 | task_lock(tsk); |
f0be3d32 | 793 | mpol_put(tsk->mempolicy); |
1da177e4 | 794 | tsk->mempolicy = NULL; |
c0ff7453 | 795 | task_unlock(tsk); |
1da177e4 | 796 | #endif |
42b2dd0a | 797 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
798 | if (unlikely(current->pi_state_cache)) |
799 | kfree(current->pi_state_cache); | |
42b2dd0a | 800 | #endif |
de5097c2 | 801 | /* |
9a11b49a | 802 | * Make sure we are holding no locks: |
de5097c2 | 803 | */ |
1b1d2fb4 | 804 | debug_check_no_locks_held(); |
778e9a9c AK |
805 | /* |
806 | * We can do this unlocked here. The futex code uses this flag | |
807 | * just to verify whether the pi state cleanup has been done | |
808 | * or not. In the worst case it loops once more. | |
809 | */ | |
810 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 811 | |
afc847b7 | 812 | if (tsk->io_context) |
b69f2292 | 813 | exit_io_context(tsk); |
afc847b7 | 814 | |
b92ce558 | 815 | if (tsk->splice_pipe) |
4b8a8f1e | 816 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 817 | |
5640f768 ED |
818 | if (tsk->task_frag.page) |
819 | put_page(tsk->task_frag.page); | |
820 | ||
e0e81739 DH |
821 | validate_creds_for_do_exit(tsk); |
822 | ||
4bcb8232 | 823 | check_stack_usage(); |
7407251a | 824 | preempt_disable(); |
54848d73 WF |
825 | if (tsk->nr_dirtied) |
826 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 827 | exit_rcu(); |
3f95aa81 | 828 | TASKS_RCU(__srcu_read_unlock(&tasks_rcu_exit_srcu, tasks_rcu_i)); |
b5740f4b YG |
829 | |
830 | /* | |
831 | * The setting of TASK_RUNNING by try_to_wake_up() may be delayed | |
832 | * when the following two conditions become true. | |
833 | * - There is race condition of mmap_sem (It is acquired by | |
834 | * exit_mm()), and | |
835 | * - SMI occurs before setting TASK_RUNINNG. | |
836 | * (or hypervisor of virtual machine switches to other guest) | |
837 | * As a result, we may become TASK_RUNNING after becoming TASK_DEAD | |
838 | * | |
839 | * To avoid it, we have to wait for releasing tsk->pi_lock which | |
840 | * is held by try_to_wake_up() | |
841 | */ | |
842 | smp_mb(); | |
843 | raw_spin_unlock_wait(&tsk->pi_lock); | |
844 | ||
55a101f8 | 845 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 846 | tsk->state = TASK_DEAD; |
a585042f | 847 | tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */ |
1da177e4 LT |
848 | schedule(); |
849 | BUG(); | |
850 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
851 | for (;;) |
852 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 | 853 | } |
012914da RA |
854 | EXPORT_SYMBOL_GPL(do_exit); |
855 | ||
9402c95f | 856 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
857 | { |
858 | if (comp) | |
859 | complete(comp); | |
55a101f8 | 860 | |
1da177e4 LT |
861 | do_exit(code); |
862 | } | |
1da177e4 LT |
863 | EXPORT_SYMBOL(complete_and_exit); |
864 | ||
754fe8d2 | 865 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
866 | { |
867 | do_exit((error_code&0xff)<<8); | |
868 | } | |
869 | ||
1da177e4 LT |
870 | /* |
871 | * Take down every thread in the group. This is called by fatal signals | |
872 | * as well as by sys_exit_group (below). | |
873 | */ | |
9402c95f | 874 | void |
1da177e4 LT |
875 | do_group_exit(int exit_code) |
876 | { | |
bfc4b089 ON |
877 | struct signal_struct *sig = current->signal; |
878 | ||
1da177e4 LT |
879 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
880 | ||
bfc4b089 ON |
881 | if (signal_group_exit(sig)) |
882 | exit_code = sig->group_exit_code; | |
1da177e4 | 883 | else if (!thread_group_empty(current)) { |
1da177e4 | 884 | struct sighand_struct *const sighand = current->sighand; |
a0be55de | 885 | |
1da177e4 | 886 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 887 | if (signal_group_exit(sig)) |
1da177e4 LT |
888 | /* Another thread got here before we took the lock. */ |
889 | exit_code = sig->group_exit_code; | |
890 | else { | |
1da177e4 | 891 | sig->group_exit_code = exit_code; |
ed5d2cac | 892 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
893 | zap_other_threads(current); |
894 | } | |
895 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
896 | } |
897 | ||
898 | do_exit(exit_code); | |
899 | /* NOTREACHED */ | |
900 | } | |
901 | ||
902 | /* | |
903 | * this kills every thread in the thread group. Note that any externally | |
904 | * wait4()-ing process will get the correct exit code - even if this | |
905 | * thread is not the thread group leader. | |
906 | */ | |
754fe8d2 | 907 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
908 | { |
909 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
910 | /* NOTREACHED */ |
911 | return 0; | |
1da177e4 LT |
912 | } |
913 | ||
9e8ae01d ON |
914 | struct wait_opts { |
915 | enum pid_type wo_type; | |
9e8ae01d | 916 | int wo_flags; |
e1eb1ebc | 917 | struct pid *wo_pid; |
9e8ae01d ON |
918 | |
919 | struct siginfo __user *wo_info; | |
920 | int __user *wo_stat; | |
921 | struct rusage __user *wo_rusage; | |
922 | ||
0b7570e7 | 923 | wait_queue_t child_wait; |
9e8ae01d ON |
924 | int notask_error; |
925 | }; | |
926 | ||
989264f4 ON |
927 | static inline |
928 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 929 | { |
989264f4 ON |
930 | if (type != PIDTYPE_PID) |
931 | task = task->group_leader; | |
932 | return task->pids[type].pid; | |
161550d7 EB |
933 | } |
934 | ||
989264f4 | 935 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 936 | { |
5c01ba49 ON |
937 | return wo->wo_type == PIDTYPE_MAX || |
938 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
939 | } | |
1da177e4 | 940 | |
5c01ba49 ON |
941 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
942 | { | |
943 | if (!eligible_pid(wo, p)) | |
944 | return 0; | |
1da177e4 LT |
945 | /* Wait for all children (clone and not) if __WALL is set; |
946 | * otherwise, wait for clone children *only* if __WCLONE is | |
947 | * set; otherwise, wait for non-clone children *only*. (Note: | |
948 | * A "clone" child here is one that reports to its parent | |
949 | * using a signal other than SIGCHLD.) */ | |
9e8ae01d ON |
950 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
951 | && !(wo->wo_flags & __WALL)) | |
1da177e4 | 952 | return 0; |
1da177e4 | 953 | |
14dd0b81 | 954 | return 1; |
1da177e4 LT |
955 | } |
956 | ||
9e8ae01d ON |
957 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
958 | pid_t pid, uid_t uid, int why, int status) | |
1da177e4 | 959 | { |
9e8ae01d ON |
960 | struct siginfo __user *infop; |
961 | int retval = wo->wo_rusage | |
962 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
36c8b586 | 963 | |
1da177e4 | 964 | put_task_struct(p); |
9e8ae01d | 965 | infop = wo->wo_info; |
b6fe2d11 VM |
966 | if (infop) { |
967 | if (!retval) | |
968 | retval = put_user(SIGCHLD, &infop->si_signo); | |
969 | if (!retval) | |
970 | retval = put_user(0, &infop->si_errno); | |
971 | if (!retval) | |
972 | retval = put_user((short)why, &infop->si_code); | |
973 | if (!retval) | |
974 | retval = put_user(pid, &infop->si_pid); | |
975 | if (!retval) | |
976 | retval = put_user(uid, &infop->si_uid); | |
977 | if (!retval) | |
978 | retval = put_user(status, &infop->si_status); | |
979 | } | |
1da177e4 LT |
980 | if (!retval) |
981 | retval = pid; | |
982 | return retval; | |
983 | } | |
984 | ||
985 | /* | |
986 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
987 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
988 | * the lock and this task is uninteresting. If we return nonzero, we have | |
989 | * released the lock and the system call should return. | |
990 | */ | |
9e8ae01d | 991 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 992 | { |
f6507f83 | 993 | int state, retval, status; |
6c5f3e7b | 994 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 995 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
9e8ae01d | 996 | struct siginfo __user *infop; |
1da177e4 | 997 | |
9e8ae01d | 998 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
999 | return 0; |
1000 | ||
9e8ae01d | 1001 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1da177e4 | 1002 | int exit_code = p->exit_code; |
f3abd4f9 | 1003 | int why; |
1da177e4 | 1004 | |
1da177e4 LT |
1005 | get_task_struct(p); |
1006 | read_unlock(&tasklist_lock); | |
1029a2b5 PZ |
1007 | sched_annotate_sleep(); |
1008 | ||
1da177e4 LT |
1009 | if ((exit_code & 0x7f) == 0) { |
1010 | why = CLD_EXITED; | |
1011 | status = exit_code >> 8; | |
1012 | } else { | |
1013 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1014 | status = exit_code & 0x7f; | |
1015 | } | |
9e8ae01d | 1016 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1da177e4 | 1017 | } |
1da177e4 | 1018 | /* |
abd50b39 | 1019 | * Move the task's state to DEAD/TRACE, only one thread can do this. |
1da177e4 | 1020 | */ |
f6507f83 ON |
1021 | state = (ptrace_reparented(p) && thread_group_leader(p)) ? |
1022 | EXIT_TRACE : EXIT_DEAD; | |
abd50b39 | 1023 | if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE) |
1da177e4 | 1024 | return 0; |
986094df ON |
1025 | /* |
1026 | * We own this thread, nobody else can reap it. | |
1027 | */ | |
1028 | read_unlock(&tasklist_lock); | |
1029 | sched_annotate_sleep(); | |
f6507f83 | 1030 | |
befca967 | 1031 | /* |
f6507f83 | 1032 | * Check thread_group_leader() to exclude the traced sub-threads. |
befca967 | 1033 | */ |
f6507f83 | 1034 | if (state == EXIT_DEAD && thread_group_leader(p)) { |
f953ccd0 ON |
1035 | struct signal_struct *sig = p->signal; |
1036 | struct signal_struct *psig = current->signal; | |
1f10206c | 1037 | unsigned long maxrss; |
0cf55e1e | 1038 | cputime_t tgutime, tgstime; |
3795e161 | 1039 | |
1da177e4 LT |
1040 | /* |
1041 | * The resource counters for the group leader are in its | |
1042 | * own task_struct. Those for dead threads in the group | |
1043 | * are in its signal_struct, as are those for the child | |
1044 | * processes it has previously reaped. All these | |
1045 | * accumulate in the parent's signal_struct c* fields. | |
1046 | * | |
1047 | * We don't bother to take a lock here to protect these | |
f953ccd0 ON |
1048 | * p->signal fields because the whole thread group is dead |
1049 | * and nobody can change them. | |
1050 | * | |
1051 | * psig->stats_lock also protects us from our sub-theads | |
1052 | * which can reap other children at the same time. Until | |
1053 | * we change k_getrusage()-like users to rely on this lock | |
1054 | * we have to take ->siglock as well. | |
0cf55e1e | 1055 | * |
a0be55de IA |
1056 | * We use thread_group_cputime_adjusted() to get times for |
1057 | * the thread group, which consolidates times for all threads | |
1058 | * in the group including the group leader. | |
1da177e4 | 1059 | */ |
e80d0a1a | 1060 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
f953ccd0 | 1061 | spin_lock_irq(¤t->sighand->siglock); |
e78c3496 | 1062 | write_seqlock(&psig->stats_lock); |
64861634 MS |
1063 | psig->cutime += tgutime + sig->cutime; |
1064 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1065 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1066 | psig->cmin_flt += |
1067 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1068 | psig->cmaj_flt += | |
1069 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1070 | psig->cnvcsw += | |
1071 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1072 | psig->cnivcsw += | |
1073 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1074 | psig->cinblock += |
1075 | task_io_get_inblock(p) + | |
1076 | sig->inblock + sig->cinblock; | |
1077 | psig->coublock += | |
1078 | task_io_get_oublock(p) + | |
1079 | sig->oublock + sig->coublock; | |
1f10206c JP |
1080 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1081 | if (psig->cmaxrss < maxrss) | |
1082 | psig->cmaxrss = maxrss; | |
5995477a AR |
1083 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1084 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
e78c3496 | 1085 | write_sequnlock(&psig->stats_lock); |
f953ccd0 | 1086 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1087 | } |
1088 | ||
9e8ae01d ON |
1089 | retval = wo->wo_rusage |
1090 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 LT |
1091 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1092 | ? p->signal->group_exit_code : p->exit_code; | |
9e8ae01d ON |
1093 | if (!retval && wo->wo_stat) |
1094 | retval = put_user(status, wo->wo_stat); | |
1095 | ||
1096 | infop = wo->wo_info; | |
1da177e4 LT |
1097 | if (!retval && infop) |
1098 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1099 | if (!retval && infop) | |
1100 | retval = put_user(0, &infop->si_errno); | |
1101 | if (!retval && infop) { | |
1102 | int why; | |
1103 | ||
1104 | if ((status & 0x7f) == 0) { | |
1105 | why = CLD_EXITED; | |
1106 | status >>= 8; | |
1107 | } else { | |
1108 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1109 | status &= 0x7f; | |
1110 | } | |
1111 | retval = put_user((short)why, &infop->si_code); | |
1112 | if (!retval) | |
1113 | retval = put_user(status, &infop->si_status); | |
1114 | } | |
1115 | if (!retval && infop) | |
3a515e4a | 1116 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1117 | if (!retval && infop) |
c69e8d9c | 1118 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1119 | if (!retval) |
3a515e4a | 1120 | retval = pid; |
2f4e6e2a | 1121 | |
b4360690 | 1122 | if (state == EXIT_TRACE) { |
1da177e4 | 1123 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1124 | /* We dropped tasklist, ptracer could die and untrace */ |
1125 | ptrace_unlink(p); | |
b4360690 ON |
1126 | |
1127 | /* If parent wants a zombie, don't release it now */ | |
1128 | state = EXIT_ZOMBIE; | |
1129 | if (do_notify_parent(p, p->exit_signal)) | |
1130 | state = EXIT_DEAD; | |
abd50b39 | 1131 | p->exit_state = state; |
1da177e4 LT |
1132 | write_unlock_irq(&tasklist_lock); |
1133 | } | |
abd50b39 | 1134 | if (state == EXIT_DEAD) |
1da177e4 | 1135 | release_task(p); |
2f4e6e2a | 1136 | |
1da177e4 LT |
1137 | return retval; |
1138 | } | |
1139 | ||
90bc8d8b ON |
1140 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1141 | { | |
1142 | if (ptrace) { | |
544b2c91 TH |
1143 | if (task_is_stopped_or_traced(p) && |
1144 | !(p->jobctl & JOBCTL_LISTENING)) | |
90bc8d8b ON |
1145 | return &p->exit_code; |
1146 | } else { | |
1147 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1148 | return &p->signal->group_exit_code; | |
1149 | } | |
1150 | return NULL; | |
1151 | } | |
1152 | ||
19e27463 TH |
1153 | /** |
1154 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1155 | * @wo: wait options | |
1156 | * @ptrace: is the wait for ptrace | |
1157 | * @p: task to wait for | |
1158 | * | |
1159 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1160 | * | |
1161 | * CONTEXT: | |
1162 | * read_lock(&tasklist_lock), which is released if return value is | |
1163 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1164 | * | |
1165 | * RETURNS: | |
1166 | * 0 if wait condition didn't exist and search for other wait conditions | |
1167 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1168 | * success, implies that tasklist_lock is released and wait condition | |
1169 | * search should terminate. | |
1da177e4 | 1170 | */ |
9e8ae01d ON |
1171 | static int wait_task_stopped(struct wait_opts *wo, |
1172 | int ptrace, struct task_struct *p) | |
1da177e4 | 1173 | { |
9e8ae01d | 1174 | struct siginfo __user *infop; |
90bc8d8b | 1175 | int retval, exit_code, *p_code, why; |
ee7c82da | 1176 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1177 | pid_t pid; |
1da177e4 | 1178 | |
47918025 ON |
1179 | /* |
1180 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1181 | */ | |
9e8ae01d | 1182 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1183 | return 0; |
1184 | ||
19e27463 TH |
1185 | if (!task_stopped_code(p, ptrace)) |
1186 | return 0; | |
1187 | ||
ee7c82da ON |
1188 | exit_code = 0; |
1189 | spin_lock_irq(&p->sighand->siglock); | |
1190 | ||
90bc8d8b ON |
1191 | p_code = task_stopped_code(p, ptrace); |
1192 | if (unlikely(!p_code)) | |
ee7c82da ON |
1193 | goto unlock_sig; |
1194 | ||
90bc8d8b | 1195 | exit_code = *p_code; |
ee7c82da ON |
1196 | if (!exit_code) |
1197 | goto unlock_sig; | |
1198 | ||
9e8ae01d | 1199 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1200 | *p_code = 0; |
ee7c82da | 1201 | |
8ca937a6 | 1202 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1203 | unlock_sig: |
1204 | spin_unlock_irq(&p->sighand->siglock); | |
1205 | if (!exit_code) | |
1da177e4 LT |
1206 | return 0; |
1207 | ||
1208 | /* | |
1209 | * Now we are pretty sure this task is interesting. | |
1210 | * Make sure it doesn't get reaped out from under us while we | |
1211 | * give up the lock and then examine it below. We don't want to | |
1212 | * keep holding onto the tasklist_lock while we call getrusage and | |
1213 | * possibly take page faults for user memory. | |
1214 | */ | |
1215 | get_task_struct(p); | |
6c5f3e7b | 1216 | pid = task_pid_vnr(p); |
f470021a | 1217 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 | 1218 | read_unlock(&tasklist_lock); |
1029a2b5 | 1219 | sched_annotate_sleep(); |
1da177e4 | 1220 | |
9e8ae01d ON |
1221 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1222 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | |
1223 | ||
1224 | retval = wo->wo_rusage | |
1225 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1226 | if (!retval && wo->wo_stat) | |
1227 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | |
1da177e4 | 1228 | |
9e8ae01d | 1229 | infop = wo->wo_info; |
1da177e4 LT |
1230 | if (!retval && infop) |
1231 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1232 | if (!retval && infop) | |
1233 | retval = put_user(0, &infop->si_errno); | |
1234 | if (!retval && infop) | |
6efcae46 | 1235 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1236 | if (!retval && infop) |
1237 | retval = put_user(exit_code, &infop->si_status); | |
1238 | if (!retval && infop) | |
c8950783 | 1239 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1240 | if (!retval && infop) |
ee7c82da | 1241 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1242 | if (!retval) |
c8950783 | 1243 | retval = pid; |
1da177e4 LT |
1244 | put_task_struct(p); |
1245 | ||
1246 | BUG_ON(!retval); | |
1247 | return retval; | |
1248 | } | |
1249 | ||
1250 | /* | |
1251 | * Handle do_wait work for one task in a live, non-stopped state. | |
1252 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1253 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1254 | * released the lock and the system call should return. | |
1255 | */ | |
9e8ae01d | 1256 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1257 | { |
1258 | int retval; | |
1259 | pid_t pid; | |
1260 | uid_t uid; | |
1261 | ||
9e8ae01d | 1262 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1263 | return 0; |
1264 | ||
1da177e4 LT |
1265 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1266 | return 0; | |
1267 | ||
1268 | spin_lock_irq(&p->sighand->siglock); | |
1269 | /* Re-check with the lock held. */ | |
1270 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1271 | spin_unlock_irq(&p->sighand->siglock); | |
1272 | return 0; | |
1273 | } | |
9e8ae01d | 1274 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1275 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1276 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1277 | spin_unlock_irq(&p->sighand->siglock); |
1278 | ||
6c5f3e7b | 1279 | pid = task_pid_vnr(p); |
1da177e4 LT |
1280 | get_task_struct(p); |
1281 | read_unlock(&tasklist_lock); | |
1029a2b5 | 1282 | sched_annotate_sleep(); |
1da177e4 | 1283 | |
9e8ae01d ON |
1284 | if (!wo->wo_info) { |
1285 | retval = wo->wo_rusage | |
1286 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 | 1287 | put_task_struct(p); |
9e8ae01d ON |
1288 | if (!retval && wo->wo_stat) |
1289 | retval = put_user(0xffff, wo->wo_stat); | |
1da177e4 | 1290 | if (!retval) |
3a515e4a | 1291 | retval = pid; |
1da177e4 | 1292 | } else { |
9e8ae01d ON |
1293 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1294 | CLD_CONTINUED, SIGCONT); | |
1da177e4 LT |
1295 | BUG_ON(retval == 0); |
1296 | } | |
1297 | ||
1298 | return retval; | |
1299 | } | |
1300 | ||
98abed02 RM |
1301 | /* |
1302 | * Consider @p for a wait by @parent. | |
1303 | * | |
9e8ae01d | 1304 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1305 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1306 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1307 | * then ->notask_error is 0 if @p is an eligible child, |
14dd0b81 | 1308 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1309 | */ |
b6e763f0 ON |
1310 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1311 | struct task_struct *p) | |
98abed02 | 1312 | { |
b3ab0316 ON |
1313 | int ret; |
1314 | ||
1315 | if (unlikely(p->exit_state == EXIT_DEAD)) | |
1316 | return 0; | |
1317 | ||
1318 | ret = eligible_child(wo, p); | |
14dd0b81 | 1319 | if (!ret) |
98abed02 RM |
1320 | return ret; |
1321 | ||
a2322e1d | 1322 | ret = security_task_wait(p); |
14dd0b81 RM |
1323 | if (unlikely(ret < 0)) { |
1324 | /* | |
1325 | * If we have not yet seen any eligible child, | |
1326 | * then let this error code replace -ECHILD. | |
1327 | * A permission error will give the user a clue | |
1328 | * to look for security policy problems, rather | |
1329 | * than for mysterious wait bugs. | |
1330 | */ | |
9e8ae01d ON |
1331 | if (wo->notask_error) |
1332 | wo->notask_error = ret; | |
78a3d9d5 | 1333 | return 0; |
14dd0b81 RM |
1334 | } |
1335 | ||
abd50b39 | 1336 | if (unlikely(p->exit_state == EXIT_TRACE)) { |
50b8d257 | 1337 | /* |
abd50b39 ON |
1338 | * ptrace == 0 means we are the natural parent. In this case |
1339 | * we should clear notask_error, debugger will notify us. | |
50b8d257 | 1340 | */ |
abd50b39 | 1341 | if (likely(!ptrace)) |
50b8d257 | 1342 | wo->notask_error = 0; |
823b018e | 1343 | return 0; |
50b8d257 | 1344 | } |
823b018e | 1345 | |
377d75da ON |
1346 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1347 | /* | |
1348 | * If it is traced by its real parent's group, just pretend | |
1349 | * the caller is ptrace_do_wait() and reap this child if it | |
1350 | * is zombie. | |
1351 | * | |
1352 | * This also hides group stop state from real parent; otherwise | |
1353 | * a single stop can be reported twice as group and ptrace stop. | |
1354 | * If a ptracer wants to distinguish these two events for its | |
1355 | * own children it should create a separate process which takes | |
1356 | * the role of real parent. | |
1357 | */ | |
1358 | if (!ptrace_reparented(p)) | |
1359 | ptrace = 1; | |
1360 | } | |
1361 | ||
45cb24a1 TH |
1362 | /* slay zombie? */ |
1363 | if (p->exit_state == EXIT_ZOMBIE) { | |
9b84cca2 | 1364 | /* we don't reap group leaders with subthreads */ |
7c733eb3 ON |
1365 | if (!delay_group_leader(p)) { |
1366 | /* | |
1367 | * A zombie ptracee is only visible to its ptracer. | |
1368 | * Notification and reaping will be cascaded to the | |
1369 | * real parent when the ptracer detaches. | |
1370 | */ | |
1371 | if (unlikely(ptrace) || likely(!p->ptrace)) | |
1372 | return wait_task_zombie(wo, p); | |
1373 | } | |
98abed02 | 1374 | |
f470021a | 1375 | /* |
9b84cca2 TH |
1376 | * Allow access to stopped/continued state via zombie by |
1377 | * falling through. Clearing of notask_error is complex. | |
1378 | * | |
1379 | * When !@ptrace: | |
1380 | * | |
1381 | * If WEXITED is set, notask_error should naturally be | |
1382 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1383 | * so, if there are live subthreads, there are events to | |
1384 | * wait for. If all subthreads are dead, it's still safe | |
1385 | * to clear - this function will be called again in finite | |
1386 | * amount time once all the subthreads are released and | |
1387 | * will then return without clearing. | |
1388 | * | |
1389 | * When @ptrace: | |
1390 | * | |
1391 | * Stopped state is per-task and thus can't change once the | |
1392 | * target task dies. Only continued and exited can happen. | |
1393 | * Clear notask_error if WCONTINUED | WEXITED. | |
1394 | */ | |
1395 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1396 | wo->notask_error = 0; | |
1397 | } else { | |
1398 | /* | |
1399 | * @p is alive and it's gonna stop, continue or exit, so | |
1400 | * there always is something to wait for. | |
f470021a | 1401 | */ |
9e8ae01d | 1402 | wo->notask_error = 0; |
f470021a RM |
1403 | } |
1404 | ||
98abed02 | 1405 | /* |
45cb24a1 TH |
1406 | * Wait for stopped. Depending on @ptrace, different stopped state |
1407 | * is used and the two don't interact with each other. | |
98abed02 | 1408 | */ |
19e27463 TH |
1409 | ret = wait_task_stopped(wo, ptrace, p); |
1410 | if (ret) | |
1411 | return ret; | |
98abed02 RM |
1412 | |
1413 | /* | |
45cb24a1 TH |
1414 | * Wait for continued. There's only one continued state and the |
1415 | * ptracer can consume it which can confuse the real parent. Don't | |
1416 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1417 | */ |
9e8ae01d | 1418 | return wait_task_continued(wo, p); |
98abed02 RM |
1419 | } |
1420 | ||
1421 | /* | |
1422 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1423 | * | |
9e8ae01d | 1424 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1425 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1426 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1427 | * ->notask_error is 0 if there were any eligible children, |
14dd0b81 | 1428 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1429 | */ |
9e8ae01d | 1430 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1431 | { |
1432 | struct task_struct *p; | |
1433 | ||
1434 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb | 1435 | int ret = wait_consider_task(wo, 0, p); |
a0be55de | 1436 | |
9cd80bbb ON |
1437 | if (ret) |
1438 | return ret; | |
98abed02 RM |
1439 | } |
1440 | ||
1441 | return 0; | |
1442 | } | |
1443 | ||
9e8ae01d | 1444 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1445 | { |
1446 | struct task_struct *p; | |
1447 | ||
f470021a | 1448 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1449 | int ret = wait_consider_task(wo, 1, p); |
a0be55de | 1450 | |
f470021a | 1451 | if (ret) |
98abed02 | 1452 | return ret; |
98abed02 RM |
1453 | } |
1454 | ||
1455 | return 0; | |
1456 | } | |
1457 | ||
0b7570e7 ON |
1458 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1459 | int sync, void *key) | |
1460 | { | |
1461 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1462 | child_wait); | |
1463 | struct task_struct *p = key; | |
1464 | ||
5c01ba49 | 1465 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1466 | return 0; |
1467 | ||
b4fe5182 ON |
1468 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1469 | return 0; | |
1470 | ||
0b7570e7 ON |
1471 | return default_wake_function(wait, mode, sync, key); |
1472 | } | |
1473 | ||
a7f0765e ON |
1474 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1475 | { | |
0b7570e7 ON |
1476 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1477 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1478 | } |
1479 | ||
9e8ae01d | 1480 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1481 | { |
1da177e4 | 1482 | struct task_struct *tsk; |
98abed02 | 1483 | int retval; |
1da177e4 | 1484 | |
9e8ae01d | 1485 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1486 | |
0b7570e7 ON |
1487 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1488 | wo->child_wait.private = current; | |
1489 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1490 | repeat: |
98abed02 RM |
1491 | /* |
1492 | * If there is nothing that can match our critiera just get out. | |
9e8ae01d ON |
1493 | * We will clear ->notask_error to zero if we see any child that |
1494 | * might later match our criteria, even if we are not able to reap | |
1495 | * it yet. | |
98abed02 | 1496 | */ |
64a16caf | 1497 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1498 | if ((wo->wo_type < PIDTYPE_MAX) && |
1499 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1500 | goto notask; |
161550d7 | 1501 | |
f95d39d1 | 1502 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1503 | read_lock(&tasklist_lock); |
1504 | tsk = current; | |
1505 | do { | |
64a16caf ON |
1506 | retval = do_wait_thread(wo, tsk); |
1507 | if (retval) | |
1508 | goto end; | |
9e8ae01d | 1509 | |
64a16caf ON |
1510 | retval = ptrace_do_wait(wo, tsk); |
1511 | if (retval) | |
98abed02 | 1512 | goto end; |
98abed02 | 1513 | |
9e8ae01d | 1514 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1515 | break; |
a3f6dfb7 | 1516 | } while_each_thread(current, tsk); |
1da177e4 | 1517 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1518 | |
64a16caf | 1519 | notask: |
9e8ae01d ON |
1520 | retval = wo->notask_error; |
1521 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1522 | retval = -ERESTARTSYS; |
98abed02 RM |
1523 | if (!signal_pending(current)) { |
1524 | schedule(); | |
1525 | goto repeat; | |
1526 | } | |
1da177e4 | 1527 | } |
1da177e4 | 1528 | end: |
f95d39d1 | 1529 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1530 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1531 | return retval; |
1532 | } | |
1533 | ||
17da2bd9 HC |
1534 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1535 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1536 | { |
9e8ae01d | 1537 | struct wait_opts wo; |
161550d7 EB |
1538 | struct pid *pid = NULL; |
1539 | enum pid_type type; | |
1da177e4 LT |
1540 | long ret; |
1541 | ||
1542 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1543 | return -EINVAL; | |
1544 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1545 | return -EINVAL; | |
1546 | ||
1547 | switch (which) { | |
1548 | case P_ALL: | |
161550d7 | 1549 | type = PIDTYPE_MAX; |
1da177e4 LT |
1550 | break; |
1551 | case P_PID: | |
161550d7 EB |
1552 | type = PIDTYPE_PID; |
1553 | if (upid <= 0) | |
1da177e4 LT |
1554 | return -EINVAL; |
1555 | break; | |
1556 | case P_PGID: | |
161550d7 EB |
1557 | type = PIDTYPE_PGID; |
1558 | if (upid <= 0) | |
1da177e4 | 1559 | return -EINVAL; |
1da177e4 LT |
1560 | break; |
1561 | default: | |
1562 | return -EINVAL; | |
1563 | } | |
1564 | ||
161550d7 EB |
1565 | if (type < PIDTYPE_MAX) |
1566 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1567 | |
1568 | wo.wo_type = type; | |
1569 | wo.wo_pid = pid; | |
1570 | wo.wo_flags = options; | |
1571 | wo.wo_info = infop; | |
1572 | wo.wo_stat = NULL; | |
1573 | wo.wo_rusage = ru; | |
1574 | ret = do_wait(&wo); | |
dfe16dfa VM |
1575 | |
1576 | if (ret > 0) { | |
1577 | ret = 0; | |
1578 | } else if (infop) { | |
1579 | /* | |
1580 | * For a WNOHANG return, clear out all the fields | |
1581 | * we would set so the user can easily tell the | |
1582 | * difference. | |
1583 | */ | |
1584 | if (!ret) | |
1585 | ret = put_user(0, &infop->si_signo); | |
1586 | if (!ret) | |
1587 | ret = put_user(0, &infop->si_errno); | |
1588 | if (!ret) | |
1589 | ret = put_user(0, &infop->si_code); | |
1590 | if (!ret) | |
1591 | ret = put_user(0, &infop->si_pid); | |
1592 | if (!ret) | |
1593 | ret = put_user(0, &infop->si_uid); | |
1594 | if (!ret) | |
1595 | ret = put_user(0, &infop->si_status); | |
1596 | } | |
1597 | ||
161550d7 | 1598 | put_pid(pid); |
1da177e4 LT |
1599 | return ret; |
1600 | } | |
1601 | ||
754fe8d2 HC |
1602 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1603 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1604 | { |
9e8ae01d | 1605 | struct wait_opts wo; |
161550d7 EB |
1606 | struct pid *pid = NULL; |
1607 | enum pid_type type; | |
1da177e4 LT |
1608 | long ret; |
1609 | ||
1610 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1611 | __WNOTHREAD|__WCLONE|__WALL)) | |
1612 | return -EINVAL; | |
161550d7 EB |
1613 | |
1614 | if (upid == -1) | |
1615 | type = PIDTYPE_MAX; | |
1616 | else if (upid < 0) { | |
1617 | type = PIDTYPE_PGID; | |
1618 | pid = find_get_pid(-upid); | |
1619 | } else if (upid == 0) { | |
1620 | type = PIDTYPE_PGID; | |
2ae448ef | 1621 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1622 | } else /* upid > 0 */ { |
1623 | type = PIDTYPE_PID; | |
1624 | pid = find_get_pid(upid); | |
1625 | } | |
1626 | ||
9e8ae01d ON |
1627 | wo.wo_type = type; |
1628 | wo.wo_pid = pid; | |
1629 | wo.wo_flags = options | WEXITED; | |
1630 | wo.wo_info = NULL; | |
1631 | wo.wo_stat = stat_addr; | |
1632 | wo.wo_rusage = ru; | |
1633 | ret = do_wait(&wo); | |
161550d7 | 1634 | put_pid(pid); |
1da177e4 | 1635 | |
1da177e4 LT |
1636 | return ret; |
1637 | } | |
1638 | ||
1639 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1640 | ||
1641 | /* | |
1642 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1643 | * implemented by calling sys_wait4() from libc.a. | |
1644 | */ | |
17da2bd9 | 1645 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1646 | { |
1647 | return sys_wait4(pid, stat_addr, options, NULL); | |
1648 | } | |
1649 | ||
1650 | #endif |