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 | ||
6fa3eb70 S |
62 | #ifdef CONFIG_MT_PRIO_TRACER |
63 | # include <linux/prio_tracer.h> | |
64 | #endif | |
65 | ||
408b664a AB |
66 | static void exit_mm(struct task_struct * tsk); |
67 | ||
d40e48e0 | 68 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
69 | { |
70 | nr_threads--; | |
50d75f8d | 71 | detach_pid(p, PIDTYPE_PID); |
d40e48e0 | 72 | if (group_dead) { |
1da177e4 LT |
73 | detach_pid(p, PIDTYPE_PGID); |
74 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 75 | |
5e85d4ab | 76 | list_del_rcu(&p->tasks); |
9cd80bbb | 77 | list_del_init(&p->sibling); |
909ea964 | 78 | __this_cpu_dec(process_counts); |
1da177e4 | 79 | } |
47e65328 | 80 | list_del_rcu(&p->thread_group); |
6fa3eb70 | 81 | list_del_rcu(&p->thread_node); |
1da177e4 LT |
82 | } |
83 | ||
6a14c5c9 ON |
84 | /* |
85 | * This function expects the tasklist_lock write-locked. | |
86 | */ | |
87 | static void __exit_signal(struct task_struct *tsk) | |
88 | { | |
89 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 90 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 91 | struct sighand_struct *sighand; |
4ada856f | 92 | struct tty_struct *uninitialized_var(tty); |
6fac4829 | 93 | cputime_t utime, stime; |
6a14c5c9 | 94 | |
d11c563d | 95 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 96 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
97 | spin_lock(&sighand->siglock); |
98 | ||
99 | posix_cpu_timers_exit(tsk); | |
d40e48e0 | 100 | if (group_dead) { |
6a14c5c9 | 101 | posix_cpu_timers_exit_group(tsk); |
4ada856f ON |
102 | tty = sig->tty; |
103 | sig->tty = NULL; | |
4a599942 | 104 | } else { |
e0a70217 ON |
105 | /* |
106 | * This can only happen if the caller is de_thread(). | |
107 | * FIXME: this is the temporary hack, we should teach | |
108 | * posix-cpu-timers to handle this case correctly. | |
109 | */ | |
110 | if (unlikely(has_group_leader_pid(tsk))) | |
111 | posix_cpu_timers_exit_group(tsk); | |
112 | ||
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); | |
122 | /* | |
123 | * Accumulate here the counters for all threads but the | |
124 | * group leader as they die, so they can be added into | |
125 | * the process-wide totals when those are taken. | |
126 | * The group leader stays around as a zombie as long | |
127 | * as there are other threads. When it gets reaped, | |
128 | * the exit.c code will add its counts into these totals. | |
129 | * We won't ever get here for the group leader, since it | |
130 | * will have been the last reference on the signal_struct. | |
131 | */ | |
6fac4829 FW |
132 | task_cputime(tsk, &utime, &stime); |
133 | sig->utime += utime; | |
134 | sig->stime += stime; | |
135 | sig->gtime += task_gtime(tsk); | |
6a14c5c9 ON |
136 | sig->min_flt += tsk->min_flt; |
137 | sig->maj_flt += tsk->maj_flt; | |
138 | sig->nvcsw += tsk->nvcsw; | |
139 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
140 | sig->inblock += task_io_get_inblock(tsk); |
141 | sig->oublock += task_io_get_oublock(tsk); | |
5995477a | 142 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
32bd671d | 143 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
144 | } |
145 | ||
b3ac022c | 146 | sig->nr_threads--; |
d40e48e0 | 147 | __unhash_process(tsk, group_dead); |
5876700c | 148 | |
da7978b0 ON |
149 | /* |
150 | * Do this under ->siglock, we can race with another thread | |
151 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
152 | */ | |
153 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 154 | tsk->sighand = NULL; |
6a14c5c9 | 155 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 156 | |
a7e5328a | 157 | __cleanup_sighand(sighand); |
6a14c5c9 | 158 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
d40e48e0 | 159 | if (group_dead) { |
6a14c5c9 | 160 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 161 | tty_kref_put(tty); |
6a14c5c9 ON |
162 | } |
163 | } | |
164 | ||
8c7904a0 EB |
165 | static void delayed_put_task_struct(struct rcu_head *rhp) |
166 | { | |
0a16b607 MD |
167 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
168 | ||
4e231c79 | 169 | perf_event_delayed_put(tsk); |
0a16b607 MD |
170 | trace_sched_process_free(tsk); |
171 | put_task_struct(tsk); | |
8c7904a0 EB |
172 | } |
173 | ||
f470021a | 174 | |
1da177e4 LT |
175 | void release_task(struct task_struct * p) |
176 | { | |
36c8b586 | 177 | struct task_struct *leader; |
1da177e4 | 178 | int zap_leader; |
1f09f974 | 179 | repeat: |
c69e8d9c | 180 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
181 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
182 | rcu_read_lock(); | |
c69e8d9c | 183 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 184 | rcu_read_unlock(); |
c69e8d9c | 185 | |
60347f67 | 186 | proc_flush_task(p); |
0203026b | 187 | |
1da177e4 | 188 | write_lock_irq(&tasklist_lock); |
a288eecc | 189 | ptrace_release_task(p); |
1da177e4 | 190 | __exit_signal(p); |
35f5cad8 | 191 | |
1da177e4 LT |
192 | /* |
193 | * If we are the last non-leader member of the thread | |
194 | * group, and the leader is zombie, then notify the | |
195 | * group leader's parent process. (if it wants notification.) | |
196 | */ | |
197 | zap_leader = 0; | |
198 | leader = p->group_leader; | |
199 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
200 | /* |
201 | * If we were the last child thread and the leader has | |
202 | * exited already, and the leader's parent ignores SIGCHLD, | |
203 | * then we are the one who should release the leader. | |
dae33574 | 204 | */ |
86773473 | 205 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
206 | if (zap_leader) |
207 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
208 | } |
209 | ||
1da177e4 | 210 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 211 | release_thread(p); |
8c7904a0 | 212 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
213 | |
214 | p = leader; | |
215 | if (unlikely(zap_leader)) | |
216 | goto repeat; | |
217 | } | |
218 | ||
1da177e4 LT |
219 | /* |
220 | * This checks not only the pgrp, but falls back on the pid if no | |
221 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
222 | * without this... | |
04a2e6a5 EB |
223 | * |
224 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 225 | */ |
04a2e6a5 | 226 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
227 | { |
228 | struct task_struct *p; | |
04a2e6a5 | 229 | struct pid *sid = NULL; |
62dfb554 | 230 | |
04a2e6a5 | 231 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 232 | if (p == NULL) |
04a2e6a5 | 233 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 234 | if (p != NULL) |
04a2e6a5 | 235 | sid = task_session(p); |
62dfb554 | 236 | |
1da177e4 LT |
237 | return sid; |
238 | } | |
239 | ||
240 | /* | |
241 | * Determine if a process group is "orphaned", according to the POSIX | |
242 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
243 | * by terminal-generated stop signals. Newly orphaned process groups are | |
244 | * to receive a SIGHUP and a SIGCONT. | |
245 | * | |
246 | * "I ask you, have you ever known what it is to be an orphan?" | |
247 | */ | |
0475ac08 | 248 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
249 | { |
250 | struct task_struct *p; | |
1da177e4 | 251 | |
0475ac08 | 252 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
253 | if ((p == ignored_task) || |
254 | (p->exit_state && thread_group_empty(p)) || | |
255 | is_global_init(p->real_parent)) | |
1da177e4 | 256 | continue; |
05e83df6 | 257 | |
0475ac08 | 258 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
259 | task_session(p->real_parent) == task_session(p)) |
260 | return 0; | |
0475ac08 | 261 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
262 | |
263 | return 1; | |
1da177e4 LT |
264 | } |
265 | ||
3e7cd6c4 | 266 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
267 | { |
268 | int retval; | |
269 | ||
270 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 271 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
272 | read_unlock(&tasklist_lock); |
273 | ||
274 | return retval; | |
275 | } | |
276 | ||
961c4675 | 277 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 278 | { |
1da177e4 LT |
279 | struct task_struct *p; |
280 | ||
0475ac08 | 281 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
282 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
283 | return true; | |
0475ac08 | 284 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
285 | |
286 | return false; | |
1da177e4 LT |
287 | } |
288 | ||
f49ee505 ON |
289 | /* |
290 | * Check to see if any process groups have become orphaned as | |
291 | * a result of our exiting, and if they have any stopped jobs, | |
292 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
293 | */ | |
294 | static void | |
295 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
296 | { | |
297 | struct pid *pgrp = task_pgrp(tsk); | |
298 | struct task_struct *ignored_task = tsk; | |
299 | ||
300 | if (!parent) | |
301 | /* exit: our father is in a different pgrp than | |
302 | * we are and we were the only connection outside. | |
303 | */ | |
304 | parent = tsk->real_parent; | |
305 | else | |
306 | /* reparent: our child is in a different pgrp than | |
307 | * we are, and it was the only connection outside. | |
308 | */ | |
309 | ignored_task = NULL; | |
310 | ||
311 | if (task_pgrp(parent) != pgrp && | |
312 | task_session(parent) == task_session(tsk) && | |
313 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
314 | has_stopped_jobs(pgrp)) { | |
315 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
316 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
317 | } | |
318 | } | |
319 | ||
8520d7c7 | 320 | void __set_special_pids(struct pid *pid) |
1da177e4 | 321 | { |
e19f247a | 322 | struct task_struct *curr = current->group_leader; |
1da177e4 | 323 | |
0d0df599 | 324 | if (task_session(curr) != pid) |
7d8da096 | 325 | change_pid(curr, PIDTYPE_SID, pid); |
1b0f7ffd ON |
326 | |
327 | if (task_pgrp(curr) != pid) | |
7d8da096 | 328 | change_pid(curr, PIDTYPE_PGID, pid); |
1da177e4 LT |
329 | } |
330 | ||
1da177e4 | 331 | /* |
87245135 ON |
332 | * Let kernel threads use this to say that they allow a certain signal. |
333 | * Must not be used if kthread was cloned with CLONE_SIGHAND. | |
1da177e4 LT |
334 | */ |
335 | int allow_signal(int sig) | |
336 | { | |
7ed20e1a | 337 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
338 | return -EINVAL; |
339 | ||
340 | spin_lock_irq(¤t->sighand->siglock); | |
87245135 | 341 | /* This is only needed for daemonize()'ed kthreads */ |
1da177e4 | 342 | sigdelset(¤t->blocked, sig); |
87245135 ON |
343 | /* |
344 | * Kernel threads handle their own signals. Let the signal code | |
345 | * know it'll be handled, so that they don't get converted to | |
346 | * SIGKILL or just silently dropped. | |
347 | */ | |
348 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
1da177e4 LT |
349 | recalc_sigpending(); |
350 | spin_unlock_irq(¤t->sighand->siglock); | |
351 | return 0; | |
352 | } | |
353 | ||
354 | EXPORT_SYMBOL(allow_signal); | |
355 | ||
356 | int disallow_signal(int sig) | |
357 | { | |
7ed20e1a | 358 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
359 | return -EINVAL; |
360 | ||
361 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 362 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
363 | recalc_sigpending(); |
364 | spin_unlock_irq(¤t->sighand->siglock); | |
365 | return 0; | |
366 | } | |
367 | ||
368 | EXPORT_SYMBOL(disallow_signal); | |
369 | ||
cf475ad2 BS |
370 | #ifdef CONFIG_MM_OWNER |
371 | /* | |
733eda7a | 372 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 373 | */ |
cf475ad2 BS |
374 | void mm_update_next_owner(struct mm_struct *mm) |
375 | { | |
376 | struct task_struct *c, *g, *p = current; | |
377 | ||
378 | retry: | |
733eda7a KH |
379 | /* |
380 | * If the exiting or execing task is not the owner, it's | |
381 | * someone else's problem. | |
382 | */ | |
383 | if (mm->owner != p) | |
cf475ad2 | 384 | return; |
733eda7a KH |
385 | /* |
386 | * The current owner is exiting/execing and there are no other | |
387 | * candidates. Do not leave the mm pointing to a possibly | |
388 | * freed task structure. | |
389 | */ | |
390 | if (atomic_read(&mm->mm_users) <= 1) { | |
391 | mm->owner = NULL; | |
392 | return; | |
393 | } | |
cf475ad2 BS |
394 | |
395 | read_lock(&tasklist_lock); | |
396 | /* | |
397 | * Search in the children | |
398 | */ | |
399 | list_for_each_entry(c, &p->children, sibling) { | |
400 | if (c->mm == mm) | |
401 | goto assign_new_owner; | |
402 | } | |
403 | ||
404 | /* | |
405 | * Search in the siblings | |
406 | */ | |
dea33cfd | 407 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
408 | if (c->mm == mm) |
409 | goto assign_new_owner; | |
410 | } | |
411 | ||
412 | /* | |
413 | * Search through everything else. We should not get | |
414 | * here often | |
415 | */ | |
416 | do_each_thread(g, c) { | |
417 | if (c->mm == mm) | |
418 | goto assign_new_owner; | |
419 | } while_each_thread(g, c); | |
420 | ||
421 | read_unlock(&tasklist_lock); | |
31a78f23 BS |
422 | /* |
423 | * We found no owner yet mm_users > 1: this implies that we are | |
424 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 425 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 426 | */ |
31a78f23 | 427 | mm->owner = NULL; |
cf475ad2 BS |
428 | return; |
429 | ||
430 | assign_new_owner: | |
431 | BUG_ON(c == p); | |
432 | get_task_struct(c); | |
433 | /* | |
434 | * The task_lock protects c->mm from changing. | |
435 | * We always want mm->owner->mm == mm | |
436 | */ | |
437 | task_lock(c); | |
e5991371 HD |
438 | /* |
439 | * Delay read_unlock() till we have the task_lock() | |
440 | * to ensure that c does not slip away underneath us | |
441 | */ | |
442 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
443 | if (c->mm != mm) { |
444 | task_unlock(c); | |
445 | put_task_struct(c); | |
446 | goto retry; | |
447 | } | |
cf475ad2 BS |
448 | mm->owner = c; |
449 | task_unlock(c); | |
450 | put_task_struct(c); | |
451 | } | |
452 | #endif /* CONFIG_MM_OWNER */ | |
453 | ||
1da177e4 LT |
454 | /* |
455 | * Turn us into a lazy TLB process if we | |
456 | * aren't already.. | |
457 | */ | |
408b664a | 458 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
459 | { |
460 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 461 | struct core_state *core_state; |
1da177e4 | 462 | |
48d212a2 | 463 | mm_release(tsk, mm); |
1da177e4 LT |
464 | if (!mm) |
465 | return; | |
4fe7efdb | 466 | sync_mm_rss(mm); |
1da177e4 LT |
467 | /* |
468 | * Serialize with any possible pending coredump. | |
999d9fc1 | 469 | * We must hold mmap_sem around checking core_state |
1da177e4 | 470 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 471 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
472 | * group with ->mm != NULL. |
473 | */ | |
474 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
475 | core_state = mm->core_state; |
476 | if (core_state) { | |
477 | struct core_thread self; | |
1da177e4 | 478 | up_read(&mm->mmap_sem); |
1da177e4 | 479 | |
b564daf8 ON |
480 | self.task = tsk; |
481 | self.next = xchg(&core_state->dumper.next, &self); | |
482 | /* | |
483 | * Implies mb(), the result of xchg() must be visible | |
484 | * to core_state->dumper. | |
485 | */ | |
486 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
487 | complete(&core_state->startup); | |
1da177e4 | 488 | |
a94e2d40 ON |
489 | for (;;) { |
490 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
491 | if (!self.task) /* see coredump_finish() */ | |
492 | break; | |
80d26af8 | 493 | freezable_schedule(); |
a94e2d40 ON |
494 | } |
495 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
496 | down_read(&mm->mmap_sem); |
497 | } | |
498 | atomic_inc(&mm->mm_count); | |
125e1874 | 499 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
500 | /* more a memory barrier than a real lock */ |
501 | task_lock(tsk); | |
502 | tsk->mm = NULL; | |
503 | up_read(&mm->mmap_sem); | |
504 | enter_lazy_tlb(mm, current); | |
505 | task_unlock(tsk); | |
cf475ad2 | 506 | mm_update_next_owner(mm); |
1da177e4 LT |
507 | mmput(mm); |
508 | } | |
509 | ||
1da177e4 | 510 | /* |
ebec18a6 LP |
511 | * When we die, we re-parent all our children, and try to: |
512 | * 1. give them to another thread in our thread group, if such a member exists | |
513 | * 2. give it to the first ancestor process which prctl'd itself as a | |
514 | * child_subreaper for its children (like a service manager) | |
515 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 516 | */ |
950bbabb | 517 | static struct task_struct *find_new_reaper(struct task_struct *father) |
d16e15f5 NK |
518 | __releases(&tasklist_lock) |
519 | __acquires(&tasklist_lock) | |
1da177e4 | 520 | { |
950bbabb ON |
521 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
522 | struct task_struct *thread; | |
1da177e4 | 523 | |
950bbabb ON |
524 | thread = father; |
525 | while_each_thread(father, thread) { | |
526 | if (thread->flags & PF_EXITING) | |
527 | continue; | |
528 | if (unlikely(pid_ns->child_reaper == father)) | |
529 | pid_ns->child_reaper = thread; | |
530 | return thread; | |
531 | } | |
1da177e4 | 532 | |
950bbabb ON |
533 | if (unlikely(pid_ns->child_reaper == father)) { |
534 | write_unlock_irq(&tasklist_lock); | |
397a21f2 DV |
535 | if (unlikely(pid_ns == &init_pid_ns)) { |
536 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
537 | father->signal->group_exit_code ?: | |
538 | father->exit_code); | |
539 | } | |
1da177e4 | 540 | |
950bbabb ON |
541 | zap_pid_ns_processes(pid_ns); |
542 | write_lock_irq(&tasklist_lock); | |
ebec18a6 LP |
543 | } else if (father->signal->has_child_subreaper) { |
544 | struct task_struct *reaper; | |
545 | ||
546 | /* | |
547 | * Find the first ancestor marked as child_subreaper. | |
548 | * Note that the code below checks same_thread_group(reaper, | |
549 | * pid_ns->child_reaper). This is what we need to DTRT in a | |
550 | * PID namespace. However we still need the check above, see | |
551 | * http://marc.info/?l=linux-kernel&m=131385460420380 | |
552 | */ | |
553 | for (reaper = father->real_parent; | |
554 | reaper != &init_task; | |
555 | reaper = reaper->real_parent) { | |
556 | if (same_thread_group(reaper, pid_ns->child_reaper)) | |
557 | break; | |
558 | if (!reaper->signal->is_child_subreaper) | |
559 | continue; | |
560 | thread = reaper; | |
561 | do { | |
562 | if (!(thread->flags & PF_EXITING)) | |
563 | return reaper; | |
564 | } while_each_thread(reaper, thread); | |
565 | } | |
1da177e4 | 566 | } |
762a24be | 567 | |
950bbabb ON |
568 | return pid_ns->child_reaper; |
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 | { | |
5dfc80be | 577 | list_move_tail(&p->sibling, &p->real_parent->children); |
5dfc80be ON |
578 | /* |
579 | * If this is a threaded reparent there is no need to | |
580 | * notify anyone anything has happened. | |
581 | */ | |
582 | if (same_thread_group(p->real_parent, father)) | |
583 | return; | |
584 | ||
a756e3d4 ON |
585 | /* |
586 | * We don't want people slaying init. | |
587 | * | |
588 | * Note: we do this even if it is EXIT_DEAD, wait_task_zombie() | |
589 | * can change ->exit_state to EXIT_ZOMBIE. If this is the final | |
590 | * state, do_notify_parent() was already called and ->exit_signal | |
591 | * doesn't matter. | |
592 | */ | |
5dfc80be ON |
593 | p->exit_signal = SIGCHLD; |
594 | ||
a756e3d4 ON |
595 | if (p->exit_state == EXIT_DEAD) |
596 | return; | |
597 | ||
5dfc80be | 598 | /* If it has exited notify the new parent about this child's death. */ |
d21142ec | 599 | if (!p->ptrace && |
5dfc80be | 600 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 601 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be ON |
602 | p->exit_state = EXIT_DEAD; |
603 | list_move_tail(&p->sibling, dead); | |
604 | } | |
605 | } | |
606 | ||
607 | kill_orphaned_pgrp(p, father); | |
608 | } | |
609 | ||
762a24be | 610 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 611 | { |
950bbabb | 612 | struct task_struct *p, *n, *reaper; |
5dfc80be | 613 | LIST_HEAD(dead_children); |
762a24be ON |
614 | |
615 | write_lock_irq(&tasklist_lock); | |
c7e49c14 ON |
616 | /* |
617 | * Note that exit_ptrace() and find_new_reaper() might | |
618 | * drop tasklist_lock and reacquire it. | |
619 | */ | |
620 | exit_ptrace(father); | |
950bbabb | 621 | reaper = find_new_reaper(father); |
f470021a | 622 | |
03ff1797 | 623 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
9cd80bbb ON |
624 | struct task_struct *t = p; |
625 | do { | |
626 | t->real_parent = reaper; | |
627 | if (t->parent == father) { | |
d21142ec | 628 | BUG_ON(t->ptrace); |
9cd80bbb ON |
629 | t->parent = t->real_parent; |
630 | } | |
631 | if (t->pdeath_signal) | |
632 | group_send_sig_info(t->pdeath_signal, | |
633 | SEND_SIG_NOINFO, t); | |
634 | } while_each_thread(p, t); | |
635 | reparent_leader(father, p, &dead_children); | |
1da177e4 | 636 | } |
762a24be | 637 | write_unlock_irq(&tasklist_lock); |
5dfc80be | 638 | |
762a24be | 639 | BUG_ON(!list_empty(&father->children)); |
762a24be | 640 | |
5dfc80be ON |
641 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
642 | list_del_init(&p->sibling); | |
39c626ae ON |
643 | release_task(p); |
644 | } | |
1da177e4 LT |
645 | } |
646 | ||
647 | /* | |
648 | * Send signals to all our closest relatives so that they know | |
649 | * to properly mourn us.. | |
650 | */ | |
821c7de7 | 651 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 652 | { |
53c8f9f1 | 653 | bool autoreap; |
1da177e4 | 654 | |
1da177e4 LT |
655 | /* |
656 | * This does two things: | |
657 | * | |
658 | * A. Make init inherit all the child processes | |
659 | * B. Check to see if any process groups have become orphaned | |
660 | * as a result of our exiting, and if they have any stopped | |
661 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
662 | */ | |
762a24be | 663 | forget_original_parent(tsk); |
1da177e4 | 664 | |
762a24be | 665 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
666 | if (group_dead) |
667 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 668 | |
45cdf5cc ON |
669 | if (unlikely(tsk->ptrace)) { |
670 | int sig = thread_group_leader(tsk) && | |
671 | thread_group_empty(tsk) && | |
672 | !ptrace_reparented(tsk) ? | |
673 | tsk->exit_signal : SIGCHLD; | |
674 | autoreap = do_notify_parent(tsk, sig); | |
675 | } else if (thread_group_leader(tsk)) { | |
676 | autoreap = thread_group_empty(tsk) && | |
677 | do_notify_parent(tsk, tsk->exit_signal); | |
678 | } else { | |
679 | autoreap = true; | |
680 | } | |
1da177e4 | 681 | |
53c8f9f1 | 682 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
1da177e4 | 683 | |
9c339168 ON |
684 | /* mt-exec, de_thread() is waiting for group leader */ |
685 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 686 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
687 | write_unlock_irq(&tasklist_lock); |
688 | ||
1da177e4 | 689 | /* If the process is dead, release it - nobody will wait for it */ |
53c8f9f1 | 690 | if (autoreap) |
1da177e4 | 691 | release_task(tsk); |
1da177e4 LT |
692 | } |
693 | ||
e18eecb8 JD |
694 | #ifdef CONFIG_DEBUG_STACK_USAGE |
695 | static void check_stack_usage(void) | |
696 | { | |
697 | static DEFINE_SPINLOCK(low_water_lock); | |
698 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
699 | unsigned long free; |
700 | ||
7c9f8861 | 701 | free = stack_not_used(current); |
e18eecb8 JD |
702 | |
703 | if (free >= lowest_to_date) | |
704 | return; | |
705 | ||
706 | spin_lock(&low_water_lock); | |
707 | if (free < lowest_to_date) { | |
168eeccb TB |
708 | printk(KERN_WARNING "%s (%d) used greatest stack depth: " |
709 | "%lu bytes left\n", | |
710 | current->comm, task_pid_nr(current), free); | |
e18eecb8 JD |
711 | lowest_to_date = free; |
712 | } | |
713 | spin_unlock(&low_water_lock); | |
714 | } | |
715 | #else | |
716 | static inline void check_stack_usage(void) {} | |
717 | #endif | |
718 | ||
6fa3eb70 S |
719 | #ifdef CONFIG_SCHEDSTATS |
720 | /* mt shceduler profiling*/ | |
721 | extern void end_mtproc_info(struct task_struct *p); | |
722 | #endif | |
9402c95f | 723 | void do_exit(long code) |
1da177e4 LT |
724 | { |
725 | struct task_struct *tsk = current; | |
726 | int group_dead; | |
727 | ||
728 | profile_task_exit(tsk); | |
6fa3eb70 S |
729 | #ifdef CONFIG_SCHEDSTATS |
730 | /* mt shceduler profiling*/ | |
731 | printk(KERN_DEBUG "[%d:%s] exit\n", tsk->pid, tsk->comm); | |
732 | end_mtproc_info(tsk); | |
733 | #endif | |
734 | ||
735 | #ifdef CONFIG_MT_PRIO_TRACER | |
736 | delete_prio_tracer(tsk->pid); | |
737 | #endif | |
1da177e4 | 738 | |
73c10101 | 739 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 740 | |
1da177e4 LT |
741 | if (unlikely(in_interrupt())) |
742 | panic("Aiee, killing interrupt handler!"); | |
743 | if (unlikely(!tsk->pid)) | |
744 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 745 | |
33dd94ae NE |
746 | /* |
747 | * If do_exit is called because this processes oopsed, it's possible | |
748 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
749 | * continuing. Amongst other possible reasons, this is to prevent | |
750 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
751 | * kernel address. | |
752 | */ | |
753 | set_fs(USER_DS); | |
754 | ||
a288eecc | 755 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 756 | |
e0e81739 DH |
757 | validate_creds_for_do_exit(tsk); |
758 | ||
df164db5 AN |
759 | /* |
760 | * We're taking recursive faults here in do_exit. Safest is to just | |
761 | * leave this task alone and wait for reboot. | |
762 | */ | |
763 | if (unlikely(tsk->flags & PF_EXITING)) { | |
764 | printk(KERN_ALERT | |
765 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
766 | /* |
767 | * We can do this unlocked here. The futex code uses | |
768 | * this flag just to verify whether the pi state | |
769 | * cleanup has been done or not. In the worst case it | |
770 | * loops once more. We pretend that the cleanup was | |
771 | * done as there is no way to return. Either the | |
772 | * OWNER_DIED bit is set by now or we push the blocked | |
773 | * task into the wait for ever nirwana as well. | |
774 | */ | |
775 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
776 | set_current_state(TASK_UNINTERRUPTIBLE); |
777 | schedule(); | |
778 | } | |
779 | ||
d12619b5 | 780 | exit_signals(tsk); /* sets PF_EXITING */ |
672c40a1 TM |
781 | |
782 | if (tsk->flags & PF_SU) { | |
783 | su_exit(); | |
784 | } | |
785 | ||
778e9a9c AK |
786 | /* |
787 | * tsk->flags are checked in the futex code to protect against | |
ed3e694d | 788 | * an exiting task cleaning up the robust pi futexes. |
778e9a9c | 789 | */ |
d2ee7198 | 790 | smp_mb(); |
1d615482 | 791 | raw_spin_unlock_wait(&tsk->pi_lock); |
1da177e4 | 792 | |
1da177e4 LT |
793 | if (unlikely(in_atomic())) |
794 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 795 | current->comm, task_pid_nr(current), |
1da177e4 LT |
796 | preempt_count()); |
797 | ||
798 | acct_update_integrals(tsk); | |
48d212a2 LT |
799 | /* sync mm's RSS info before statistics gathering */ |
800 | if (tsk->mm) | |
801 | sync_mm_rss(tsk->mm); | |
1da177e4 | 802 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 803 | if (group_dead) { |
778e9a9c | 804 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 805 | exit_itimers(tsk->signal); |
1f10206c JP |
806 | if (tsk->mm) |
807 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 808 | } |
f6ec29a4 | 809 | acct_collect(code, group_dead); |
522ed776 MT |
810 | if (group_dead) |
811 | tty_audit_exit(); | |
a4ff8dba | 812 | audit_free(tsk); |
115085ea | 813 | |
48d212a2 | 814 | tsk->exit_code = code; |
115085ea | 815 | taskstats_exit(tsk, group_dead); |
c757249a | 816 | |
1da177e4 LT |
817 | exit_mm(tsk); |
818 | ||
0e464814 | 819 | if (group_dead) |
f6ec29a4 | 820 | acct_process(); |
0a16b607 MD |
821 | trace_sched_process_exit(tsk); |
822 | ||
1da177e4 | 823 | exit_sem(tsk); |
b34a6b1d | 824 | exit_shm(tsk); |
1ec7f1dd AV |
825 | exit_files(tsk); |
826 | exit_fs(tsk); | |
16739781 ON |
827 | if (group_dead) |
828 | disassociate_ctty(1); | |
8aac6270 | 829 | exit_task_namespaces(tsk); |
ed3e694d | 830 | exit_task_work(tsk); |
e18eecb8 | 831 | check_stack_usage(); |
1da177e4 | 832 | exit_thread(); |
0b3fcf17 SE |
833 | |
834 | /* | |
835 | * Flush inherited counters to the parent - before the parent | |
836 | * gets woken up by child-exit notifications. | |
837 | * | |
838 | * because of cgroup mode, must be called before cgroup_exit() | |
839 | */ | |
840 | perf_event_exit_task(tsk); | |
841 | ||
b4f48b63 | 842 | cgroup_exit(tsk, 1); |
1da177e4 | 843 | |
a1261f54 | 844 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 | 845 | |
9f46080c | 846 | proc_exit_connector(tsk); |
24f1e32c FW |
847 | /* |
848 | * FIXME: do that only when needed, using sched_exit tracepoint | |
849 | */ | |
bf26c018 | 850 | ptrace_put_breakpoints(tsk); |
33b2fb30 | 851 | |
821c7de7 | 852 | exit_notify(tsk, group_dead); |
1da177e4 | 853 | #ifdef CONFIG_NUMA |
c0ff7453 | 854 | task_lock(tsk); |
f0be3d32 | 855 | mpol_put(tsk->mempolicy); |
1da177e4 | 856 | tsk->mempolicy = NULL; |
c0ff7453 | 857 | task_unlock(tsk); |
1da177e4 | 858 | #endif |
42b2dd0a | 859 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
860 | if (unlikely(current->pi_state_cache)) |
861 | kfree(current->pi_state_cache); | |
42b2dd0a | 862 | #endif |
de5097c2 | 863 | /* |
9a11b49a | 864 | * Make sure we are holding no locks: |
de5097c2 | 865 | */ |
6fa3eb70 | 866 | debug_check_no_locks_held(); |
778e9a9c AK |
867 | /* |
868 | * We can do this unlocked here. The futex code uses this flag | |
869 | * just to verify whether the pi state cleanup has been done | |
870 | * or not. In the worst case it loops once more. | |
871 | */ | |
872 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 873 | |
afc847b7 | 874 | if (tsk->io_context) |
b69f2292 | 875 | exit_io_context(tsk); |
afc847b7 | 876 | |
b92ce558 | 877 | if (tsk->splice_pipe) |
4b8a8f1e | 878 | free_pipe_info(tsk->splice_pipe); |
b92ce558 | 879 | |
5640f768 ED |
880 | if (tsk->task_frag.page) |
881 | put_page(tsk->task_frag.page); | |
882 | ||
e0e81739 DH |
883 | validate_creds_for_do_exit(tsk); |
884 | ||
7407251a | 885 | preempt_disable(); |
54848d73 WF |
886 | if (tsk->nr_dirtied) |
887 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 888 | exit_rcu(); |
b5740f4b YG |
889 | |
890 | /* | |
891 | * The setting of TASK_RUNNING by try_to_wake_up() may be delayed | |
892 | * when the following two conditions become true. | |
893 | * - There is race condition of mmap_sem (It is acquired by | |
894 | * exit_mm()), and | |
895 | * - SMI occurs before setting TASK_RUNINNG. | |
896 | * (or hypervisor of virtual machine switches to other guest) | |
897 | * As a result, we may become TASK_RUNNING after becoming TASK_DEAD | |
898 | * | |
899 | * To avoid it, we have to wait for releasing tsk->pi_lock which | |
900 | * is held by try_to_wake_up() | |
901 | */ | |
902 | smp_mb(); | |
903 | raw_spin_unlock_wait(&tsk->pi_lock); | |
904 | ||
55a101f8 | 905 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 906 | tsk->state = TASK_DEAD; |
a585042f | 907 | tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */ |
1da177e4 LT |
908 | schedule(); |
909 | BUG(); | |
910 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
911 | for (;;) |
912 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
913 | } |
914 | ||
012914da RA |
915 | EXPORT_SYMBOL_GPL(do_exit); |
916 | ||
9402c95f | 917 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
918 | { |
919 | if (comp) | |
920 | complete(comp); | |
55a101f8 | 921 | |
1da177e4 LT |
922 | do_exit(code); |
923 | } | |
924 | ||
925 | EXPORT_SYMBOL(complete_and_exit); | |
926 | ||
754fe8d2 | 927 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
928 | { |
929 | do_exit((error_code&0xff)<<8); | |
930 | } | |
931 | ||
1da177e4 LT |
932 | /* |
933 | * Take down every thread in the group. This is called by fatal signals | |
934 | * as well as by sys_exit_group (below). | |
935 | */ | |
9402c95f | 936 | void |
1da177e4 LT |
937 | do_group_exit(int exit_code) |
938 | { | |
bfc4b089 ON |
939 | struct signal_struct *sig = current->signal; |
940 | ||
1da177e4 LT |
941 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
942 | ||
bfc4b089 ON |
943 | if (signal_group_exit(sig)) |
944 | exit_code = sig->group_exit_code; | |
1da177e4 | 945 | else if (!thread_group_empty(current)) { |
1da177e4 | 946 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 947 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 948 | if (signal_group_exit(sig)) |
1da177e4 LT |
949 | /* Another thread got here before we took the lock. */ |
950 | exit_code = sig->group_exit_code; | |
951 | else { | |
1da177e4 | 952 | sig->group_exit_code = exit_code; |
ed5d2cac | 953 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
954 | zap_other_threads(current); |
955 | } | |
956 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
957 | } |
958 | ||
959 | do_exit(exit_code); | |
960 | /* NOTREACHED */ | |
961 | } | |
962 | ||
963 | /* | |
964 | * this kills every thread in the thread group. Note that any externally | |
965 | * wait4()-ing process will get the correct exit code - even if this | |
966 | * thread is not the thread group leader. | |
967 | */ | |
754fe8d2 | 968 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
969 | { |
970 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
971 | /* NOTREACHED */ |
972 | return 0; | |
1da177e4 LT |
973 | } |
974 | ||
9e8ae01d ON |
975 | struct wait_opts { |
976 | enum pid_type wo_type; | |
9e8ae01d | 977 | int wo_flags; |
e1eb1ebc | 978 | struct pid *wo_pid; |
9e8ae01d ON |
979 | |
980 | struct siginfo __user *wo_info; | |
981 | int __user *wo_stat; | |
982 | struct rusage __user *wo_rusage; | |
983 | ||
0b7570e7 | 984 | wait_queue_t child_wait; |
9e8ae01d ON |
985 | int notask_error; |
986 | }; | |
987 | ||
989264f4 ON |
988 | static inline |
989 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 990 | { |
989264f4 ON |
991 | if (type != PIDTYPE_PID) |
992 | task = task->group_leader; | |
993 | return task->pids[type].pid; | |
161550d7 EB |
994 | } |
995 | ||
989264f4 | 996 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 997 | { |
5c01ba49 ON |
998 | return wo->wo_type == PIDTYPE_MAX || |
999 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1000 | } | |
1da177e4 | 1001 | |
5c01ba49 ON |
1002 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
1003 | { | |
1004 | if (!eligible_pid(wo, p)) | |
1005 | return 0; | |
1da177e4 LT |
1006 | /* Wait for all children (clone and not) if __WALL is set; |
1007 | * otherwise, wait for clone children *only* if __WCLONE is | |
1008 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1009 | * A "clone" child here is one that reports to its parent | |
1010 | * using a signal other than SIGCHLD.) */ | |
9e8ae01d ON |
1011 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
1012 | && !(wo->wo_flags & __WALL)) | |
1da177e4 | 1013 | return 0; |
1da177e4 | 1014 | |
14dd0b81 | 1015 | return 1; |
1da177e4 LT |
1016 | } |
1017 | ||
9e8ae01d ON |
1018 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
1019 | pid_t pid, uid_t uid, int why, int status) | |
1da177e4 | 1020 | { |
9e8ae01d ON |
1021 | struct siginfo __user *infop; |
1022 | int retval = wo->wo_rusage | |
1023 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
36c8b586 | 1024 | |
1da177e4 | 1025 | put_task_struct(p); |
9e8ae01d | 1026 | infop = wo->wo_info; |
b6fe2d11 VM |
1027 | if (infop) { |
1028 | if (!retval) | |
1029 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1030 | if (!retval) | |
1031 | retval = put_user(0, &infop->si_errno); | |
1032 | if (!retval) | |
1033 | retval = put_user((short)why, &infop->si_code); | |
1034 | if (!retval) | |
1035 | retval = put_user(pid, &infop->si_pid); | |
1036 | if (!retval) | |
1037 | retval = put_user(uid, &infop->si_uid); | |
1038 | if (!retval) | |
1039 | retval = put_user(status, &infop->si_status); | |
1040 | } | |
1da177e4 LT |
1041 | if (!retval) |
1042 | retval = pid; | |
1043 | return retval; | |
1044 | } | |
1045 | ||
1046 | /* | |
1047 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1048 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1049 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1050 | * released the lock and the system call should return. | |
1051 | */ | |
9e8ae01d | 1052 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1053 | { |
1054 | unsigned long state; | |
2f4e6e2a | 1055 | int retval, status, traced; |
6c5f3e7b | 1056 | pid_t pid = task_pid_vnr(p); |
43e13cc1 | 1057 | uid_t uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
9e8ae01d | 1058 | struct siginfo __user *infop; |
1da177e4 | 1059 | |
9e8ae01d | 1060 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1061 | return 0; |
1062 | ||
9e8ae01d | 1063 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1da177e4 | 1064 | int exit_code = p->exit_code; |
f3abd4f9 | 1065 | int why; |
1da177e4 | 1066 | |
1da177e4 LT |
1067 | get_task_struct(p); |
1068 | read_unlock(&tasklist_lock); | |
1069 | if ((exit_code & 0x7f) == 0) { | |
1070 | why = CLD_EXITED; | |
1071 | status = exit_code >> 8; | |
1072 | } else { | |
1073 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1074 | status = exit_code & 0x7f; | |
1075 | } | |
9e8ae01d | 1076 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1da177e4 LT |
1077 | } |
1078 | ||
1079 | /* | |
1080 | * Try to move the task's state to DEAD | |
1081 | * only one thread is allowed to do this: | |
1082 | */ | |
1083 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1084 | if (state != EXIT_ZOMBIE) { | |
1085 | BUG_ON(state != EXIT_DEAD); | |
1086 | return 0; | |
1087 | } | |
1da177e4 | 1088 | |
53b6f9fb | 1089 | traced = ptrace_reparented(p); |
befca967 ON |
1090 | /* |
1091 | * It can be ptraced but not reparented, check | |
e550f14d | 1092 | * thread_group_leader() to filter out sub-threads. |
befca967 | 1093 | */ |
e550f14d | 1094 | if (likely(!traced) && thread_group_leader(p)) { |
3795e161 JJ |
1095 | struct signal_struct *psig; |
1096 | struct signal_struct *sig; | |
1f10206c | 1097 | unsigned long maxrss; |
0cf55e1e | 1098 | cputime_t tgutime, tgstime; |
3795e161 | 1099 | |
1da177e4 LT |
1100 | /* |
1101 | * The resource counters for the group leader are in its | |
1102 | * own task_struct. Those for dead threads in the group | |
1103 | * are in its signal_struct, as are those for the child | |
1104 | * processes it has previously reaped. All these | |
1105 | * accumulate in the parent's signal_struct c* fields. | |
1106 | * | |
1107 | * We don't bother to take a lock here to protect these | |
1108 | * p->signal fields, because they are only touched by | |
1109 | * __exit_signal, which runs with tasklist_lock | |
1110 | * write-locked anyway, and so is excluded here. We do | |
d1e98f42 | 1111 | * need to protect the access to parent->signal fields, |
1da177e4 LT |
1112 | * as other threads in the parent group can be right |
1113 | * here reaping other children at the same time. | |
0cf55e1e | 1114 | * |
e80d0a1a | 1115 | * We use thread_group_cputime_adjusted() to get times for the thread |
0cf55e1e HS |
1116 | * group, which consolidates times for all threads in the |
1117 | * group including the group leader. | |
1da177e4 | 1118 | */ |
e80d0a1a | 1119 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
d1e98f42 ON |
1120 | spin_lock_irq(&p->real_parent->sighand->siglock); |
1121 | psig = p->real_parent->signal; | |
3795e161 | 1122 | sig = p->signal; |
64861634 MS |
1123 | psig->cutime += tgutime + sig->cutime; |
1124 | psig->cstime += tgstime + sig->cstime; | |
6fac4829 | 1125 | psig->cgtime += task_gtime(p) + sig->gtime + sig->cgtime; |
3795e161 JJ |
1126 | psig->cmin_flt += |
1127 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1128 | psig->cmaj_flt += | |
1129 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1130 | psig->cnvcsw += | |
1131 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1132 | psig->cnivcsw += | |
1133 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1134 | psig->cinblock += |
1135 | task_io_get_inblock(p) + | |
1136 | sig->inblock + sig->cinblock; | |
1137 | psig->coublock += | |
1138 | task_io_get_oublock(p) + | |
1139 | sig->oublock + sig->coublock; | |
1f10206c JP |
1140 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1141 | if (psig->cmaxrss < maxrss) | |
1142 | psig->cmaxrss = maxrss; | |
5995477a AR |
1143 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1144 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
d1e98f42 | 1145 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
1da177e4 LT |
1146 | } |
1147 | ||
1148 | /* | |
1149 | * Now we are sure this task is interesting, and no other | |
1150 | * thread can reap it because we set its state to EXIT_DEAD. | |
1151 | */ | |
1152 | read_unlock(&tasklist_lock); | |
1153 | ||
9e8ae01d ON |
1154 | retval = wo->wo_rusage |
1155 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 LT |
1156 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1157 | ? p->signal->group_exit_code : p->exit_code; | |
9e8ae01d ON |
1158 | if (!retval && wo->wo_stat) |
1159 | retval = put_user(status, wo->wo_stat); | |
1160 | ||
1161 | infop = wo->wo_info; | |
1da177e4 LT |
1162 | if (!retval && infop) |
1163 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1164 | if (!retval && infop) | |
1165 | retval = put_user(0, &infop->si_errno); | |
1166 | if (!retval && infop) { | |
1167 | int why; | |
1168 | ||
1169 | if ((status & 0x7f) == 0) { | |
1170 | why = CLD_EXITED; | |
1171 | status >>= 8; | |
1172 | } else { | |
1173 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1174 | status &= 0x7f; | |
1175 | } | |
1176 | retval = put_user((short)why, &infop->si_code); | |
1177 | if (!retval) | |
1178 | retval = put_user(status, &infop->si_status); | |
1179 | } | |
1180 | if (!retval && infop) | |
3a515e4a | 1181 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1182 | if (!retval && infop) |
c69e8d9c | 1183 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1184 | if (!retval) |
3a515e4a | 1185 | retval = pid; |
2f4e6e2a ON |
1186 | |
1187 | if (traced) { | |
1da177e4 | 1188 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1189 | /* We dropped tasklist, ptracer could die and untrace */ |
1190 | ptrace_unlink(p); | |
1191 | /* | |
86773473 ON |
1192 | * If this is not a sub-thread, notify the parent. |
1193 | * If parent wants a zombie, don't release it now. | |
2f4e6e2a | 1194 | */ |
86773473 ON |
1195 | if (thread_group_leader(p) && |
1196 | !do_notify_parent(p, p->exit_signal)) { | |
1197 | p->exit_state = EXIT_ZOMBIE; | |
1198 | p = NULL; | |
1da177e4 LT |
1199 | } |
1200 | write_unlock_irq(&tasklist_lock); | |
1201 | } | |
1202 | if (p != NULL) | |
1203 | release_task(p); | |
2f4e6e2a | 1204 | |
1da177e4 LT |
1205 | return retval; |
1206 | } | |
1207 | ||
90bc8d8b ON |
1208 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1209 | { | |
1210 | if (ptrace) { | |
544b2c91 TH |
1211 | if (task_is_stopped_or_traced(p) && |
1212 | !(p->jobctl & JOBCTL_LISTENING)) | |
90bc8d8b ON |
1213 | return &p->exit_code; |
1214 | } else { | |
1215 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1216 | return &p->signal->group_exit_code; | |
1217 | } | |
1218 | return NULL; | |
1219 | } | |
1220 | ||
19e27463 TH |
1221 | /** |
1222 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1223 | * @wo: wait options | |
1224 | * @ptrace: is the wait for ptrace | |
1225 | * @p: task to wait for | |
1226 | * | |
1227 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1228 | * | |
1229 | * CONTEXT: | |
1230 | * read_lock(&tasklist_lock), which is released if return value is | |
1231 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1232 | * | |
1233 | * RETURNS: | |
1234 | * 0 if wait condition didn't exist and search for other wait conditions | |
1235 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1236 | * success, implies that tasklist_lock is released and wait condition | |
1237 | * search should terminate. | |
1da177e4 | 1238 | */ |
9e8ae01d ON |
1239 | static int wait_task_stopped(struct wait_opts *wo, |
1240 | int ptrace, struct task_struct *p) | |
1da177e4 | 1241 | { |
9e8ae01d | 1242 | struct siginfo __user *infop; |
90bc8d8b | 1243 | int retval, exit_code, *p_code, why; |
ee7c82da | 1244 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1245 | pid_t pid; |
1da177e4 | 1246 | |
47918025 ON |
1247 | /* |
1248 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1249 | */ | |
9e8ae01d | 1250 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1251 | return 0; |
1252 | ||
19e27463 TH |
1253 | if (!task_stopped_code(p, ptrace)) |
1254 | return 0; | |
1255 | ||
ee7c82da ON |
1256 | exit_code = 0; |
1257 | spin_lock_irq(&p->sighand->siglock); | |
1258 | ||
90bc8d8b ON |
1259 | p_code = task_stopped_code(p, ptrace); |
1260 | if (unlikely(!p_code)) | |
ee7c82da ON |
1261 | goto unlock_sig; |
1262 | ||
90bc8d8b | 1263 | exit_code = *p_code; |
ee7c82da ON |
1264 | if (!exit_code) |
1265 | goto unlock_sig; | |
1266 | ||
9e8ae01d | 1267 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1268 | *p_code = 0; |
ee7c82da | 1269 | |
8ca937a6 | 1270 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1271 | unlock_sig: |
1272 | spin_unlock_irq(&p->sighand->siglock); | |
1273 | if (!exit_code) | |
1da177e4 LT |
1274 | return 0; |
1275 | ||
1276 | /* | |
1277 | * Now we are pretty sure this task is interesting. | |
1278 | * Make sure it doesn't get reaped out from under us while we | |
1279 | * give up the lock and then examine it below. We don't want to | |
1280 | * keep holding onto the tasklist_lock while we call getrusage and | |
1281 | * possibly take page faults for user memory. | |
1282 | */ | |
1283 | get_task_struct(p); | |
6c5f3e7b | 1284 | pid = task_pid_vnr(p); |
f470021a | 1285 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1286 | read_unlock(&tasklist_lock); |
1287 | ||
9e8ae01d ON |
1288 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1289 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | |
1290 | ||
1291 | retval = wo->wo_rusage | |
1292 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1293 | if (!retval && wo->wo_stat) | |
1294 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | |
1da177e4 | 1295 | |
9e8ae01d | 1296 | infop = wo->wo_info; |
1da177e4 LT |
1297 | if (!retval && infop) |
1298 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1299 | if (!retval && infop) | |
1300 | retval = put_user(0, &infop->si_errno); | |
1301 | if (!retval && infop) | |
6efcae46 | 1302 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1303 | if (!retval && infop) |
1304 | retval = put_user(exit_code, &infop->si_status); | |
1305 | if (!retval && infop) | |
c8950783 | 1306 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1307 | if (!retval && infop) |
ee7c82da | 1308 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1309 | if (!retval) |
c8950783 | 1310 | retval = pid; |
1da177e4 LT |
1311 | put_task_struct(p); |
1312 | ||
1313 | BUG_ON(!retval); | |
1314 | return retval; | |
1315 | } | |
1316 | ||
1317 | /* | |
1318 | * Handle do_wait work for one task in a live, non-stopped state. | |
1319 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1320 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1321 | * released the lock and the system call should return. | |
1322 | */ | |
9e8ae01d | 1323 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1324 | { |
1325 | int retval; | |
1326 | pid_t pid; | |
1327 | uid_t uid; | |
1328 | ||
9e8ae01d | 1329 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1330 | return 0; |
1331 | ||
1da177e4 LT |
1332 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1333 | return 0; | |
1334 | ||
1335 | spin_lock_irq(&p->sighand->siglock); | |
1336 | /* Re-check with the lock held. */ | |
1337 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1338 | spin_unlock_irq(&p->sighand->siglock); | |
1339 | return 0; | |
1340 | } | |
9e8ae01d | 1341 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1342 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1343 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1344 | spin_unlock_irq(&p->sighand->siglock); |
1345 | ||
6c5f3e7b | 1346 | pid = task_pid_vnr(p); |
1da177e4 LT |
1347 | get_task_struct(p); |
1348 | read_unlock(&tasklist_lock); | |
1349 | ||
9e8ae01d ON |
1350 | if (!wo->wo_info) { |
1351 | retval = wo->wo_rusage | |
1352 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 | 1353 | put_task_struct(p); |
9e8ae01d ON |
1354 | if (!retval && wo->wo_stat) |
1355 | retval = put_user(0xffff, wo->wo_stat); | |
1da177e4 | 1356 | if (!retval) |
3a515e4a | 1357 | retval = pid; |
1da177e4 | 1358 | } else { |
9e8ae01d ON |
1359 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1360 | CLD_CONTINUED, SIGCONT); | |
1da177e4 LT |
1361 | BUG_ON(retval == 0); |
1362 | } | |
1363 | ||
1364 | return retval; | |
1365 | } | |
1366 | ||
98abed02 RM |
1367 | /* |
1368 | * Consider @p for a wait by @parent. | |
1369 | * | |
9e8ae01d | 1370 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1371 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1372 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1373 | * then ->notask_error is 0 if @p is an eligible child, |
14dd0b81 | 1374 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1375 | */ |
b6e763f0 ON |
1376 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1377 | struct task_struct *p) | |
98abed02 | 1378 | { |
9e8ae01d | 1379 | int ret = eligible_child(wo, p); |
14dd0b81 | 1380 | if (!ret) |
98abed02 RM |
1381 | return ret; |
1382 | ||
a2322e1d | 1383 | ret = security_task_wait(p); |
14dd0b81 RM |
1384 | if (unlikely(ret < 0)) { |
1385 | /* | |
1386 | * If we have not yet seen any eligible child, | |
1387 | * then let this error code replace -ECHILD. | |
1388 | * A permission error will give the user a clue | |
1389 | * to look for security policy problems, rather | |
1390 | * than for mysterious wait bugs. | |
1391 | */ | |
9e8ae01d ON |
1392 | if (wo->notask_error) |
1393 | wo->notask_error = ret; | |
78a3d9d5 | 1394 | return 0; |
14dd0b81 RM |
1395 | } |
1396 | ||
823b018e | 1397 | /* dead body doesn't have much to contribute */ |
50b8d257 ON |
1398 | if (unlikely(p->exit_state == EXIT_DEAD)) { |
1399 | /* | |
1400 | * But do not ignore this task until the tracer does | |
1401 | * wait_task_zombie()->do_notify_parent(). | |
1402 | */ | |
1403 | if (likely(!ptrace) && unlikely(ptrace_reparented(p))) | |
1404 | wo->notask_error = 0; | |
823b018e | 1405 | return 0; |
50b8d257 | 1406 | } |
823b018e | 1407 | |
45cb24a1 TH |
1408 | /* slay zombie? */ |
1409 | if (p->exit_state == EXIT_ZOMBIE) { | |
f470021a | 1410 | /* |
45cb24a1 TH |
1411 | * A zombie ptracee is only visible to its ptracer. |
1412 | * Notification and reaping will be cascaded to the real | |
1413 | * parent when the ptracer detaches. | |
f470021a | 1414 | */ |
d21142ec | 1415 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
45cb24a1 TH |
1416 | /* it will become visible, clear notask_error */ |
1417 | wo->notask_error = 0; | |
1418 | return 0; | |
1419 | } | |
f470021a | 1420 | |
9b84cca2 TH |
1421 | /* we don't reap group leaders with subthreads */ |
1422 | if (!delay_group_leader(p)) | |
1423 | return wait_task_zombie(wo, p); | |
98abed02 | 1424 | |
f470021a | 1425 | /* |
9b84cca2 TH |
1426 | * Allow access to stopped/continued state via zombie by |
1427 | * falling through. Clearing of notask_error is complex. | |
1428 | * | |
1429 | * When !@ptrace: | |
1430 | * | |
1431 | * If WEXITED is set, notask_error should naturally be | |
1432 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1433 | * so, if there are live subthreads, there are events to | |
1434 | * wait for. If all subthreads are dead, it's still safe | |
1435 | * to clear - this function will be called again in finite | |
1436 | * amount time once all the subthreads are released and | |
1437 | * will then return without clearing. | |
1438 | * | |
1439 | * When @ptrace: | |
1440 | * | |
1441 | * Stopped state is per-task and thus can't change once the | |
1442 | * target task dies. Only continued and exited can happen. | |
1443 | * Clear notask_error if WCONTINUED | WEXITED. | |
1444 | */ | |
1445 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1446 | wo->notask_error = 0; | |
1447 | } else { | |
45cb24a1 TH |
1448 | /* |
1449 | * If @p is ptraced by a task in its real parent's group, | |
1450 | * hide group stop/continued state when looking at @p as | |
1451 | * the real parent; otherwise, a single stop can be | |
1452 | * reported twice as group and ptrace stops. | |
1453 | * | |
1454 | * If a ptracer wants to distinguish the two events for its | |
1455 | * own children, it should create a separate process which | |
1456 | * takes the role of real parent. | |
1457 | */ | |
479bf98c | 1458 | if (likely(!ptrace) && p->ptrace && !ptrace_reparented(p)) |
45cb24a1 TH |
1459 | return 0; |
1460 | ||
9b84cca2 TH |
1461 | /* |
1462 | * @p is alive and it's gonna stop, continue or exit, so | |
1463 | * there always is something to wait for. | |
f470021a | 1464 | */ |
9e8ae01d | 1465 | wo->notask_error = 0; |
f470021a RM |
1466 | } |
1467 | ||
98abed02 | 1468 | /* |
45cb24a1 TH |
1469 | * Wait for stopped. Depending on @ptrace, different stopped state |
1470 | * is used and the two don't interact with each other. | |
98abed02 | 1471 | */ |
19e27463 TH |
1472 | ret = wait_task_stopped(wo, ptrace, p); |
1473 | if (ret) | |
1474 | return ret; | |
98abed02 RM |
1475 | |
1476 | /* | |
45cb24a1 TH |
1477 | * Wait for continued. There's only one continued state and the |
1478 | * ptracer can consume it which can confuse the real parent. Don't | |
1479 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1480 | */ |
9e8ae01d | 1481 | return wait_task_continued(wo, p); |
98abed02 RM |
1482 | } |
1483 | ||
1484 | /* | |
1485 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1486 | * | |
9e8ae01d | 1487 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1488 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1489 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1490 | * ->notask_error is 0 if there were any eligible children, |
14dd0b81 | 1491 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1492 | */ |
9e8ae01d | 1493 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1494 | { |
1495 | struct task_struct *p; | |
1496 | ||
1497 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb ON |
1498 | int ret = wait_consider_task(wo, 0, p); |
1499 | if (ret) | |
1500 | return ret; | |
98abed02 RM |
1501 | } |
1502 | ||
1503 | return 0; | |
1504 | } | |
1505 | ||
9e8ae01d | 1506 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1507 | { |
1508 | struct task_struct *p; | |
1509 | ||
f470021a | 1510 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1511 | int ret = wait_consider_task(wo, 1, p); |
f470021a | 1512 | if (ret) |
98abed02 | 1513 | return ret; |
98abed02 RM |
1514 | } |
1515 | ||
1516 | return 0; | |
1517 | } | |
1518 | ||
0b7570e7 ON |
1519 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1520 | int sync, void *key) | |
1521 | { | |
1522 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1523 | child_wait); | |
1524 | struct task_struct *p = key; | |
1525 | ||
5c01ba49 | 1526 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1527 | return 0; |
1528 | ||
b4fe5182 ON |
1529 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1530 | return 0; | |
1531 | ||
0b7570e7 ON |
1532 | return default_wake_function(wait, mode, sync, key); |
1533 | } | |
1534 | ||
a7f0765e ON |
1535 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1536 | { | |
0b7570e7 ON |
1537 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1538 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1539 | } |
1540 | ||
9e8ae01d | 1541 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1542 | { |
1da177e4 | 1543 | struct task_struct *tsk; |
98abed02 | 1544 | int retval; |
1da177e4 | 1545 | |
9e8ae01d | 1546 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1547 | |
0b7570e7 ON |
1548 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1549 | wo->child_wait.private = current; | |
1550 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1551 | repeat: |
98abed02 RM |
1552 | /* |
1553 | * If there is nothing that can match our critiera just get out. | |
9e8ae01d ON |
1554 | * We will clear ->notask_error to zero if we see any child that |
1555 | * might later match our criteria, even if we are not able to reap | |
1556 | * it yet. | |
98abed02 | 1557 | */ |
64a16caf | 1558 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1559 | if ((wo->wo_type < PIDTYPE_MAX) && |
1560 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1561 | goto notask; |
161550d7 | 1562 | |
f95d39d1 | 1563 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1564 | read_lock(&tasklist_lock); |
1565 | tsk = current; | |
1566 | do { | |
64a16caf ON |
1567 | retval = do_wait_thread(wo, tsk); |
1568 | if (retval) | |
1569 | goto end; | |
9e8ae01d | 1570 | |
64a16caf ON |
1571 | retval = ptrace_do_wait(wo, tsk); |
1572 | if (retval) | |
98abed02 | 1573 | goto end; |
98abed02 | 1574 | |
9e8ae01d | 1575 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1576 | break; |
a3f6dfb7 | 1577 | } while_each_thread(current, tsk); |
1da177e4 | 1578 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1579 | |
64a16caf | 1580 | notask: |
9e8ae01d ON |
1581 | retval = wo->notask_error; |
1582 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1583 | retval = -ERESTARTSYS; |
98abed02 RM |
1584 | if (!signal_pending(current)) { |
1585 | schedule(); | |
1586 | goto repeat; | |
1587 | } | |
1da177e4 | 1588 | } |
1da177e4 | 1589 | end: |
f95d39d1 | 1590 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1591 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1592 | return retval; |
1593 | } | |
1594 | ||
17da2bd9 HC |
1595 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1596 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1597 | { |
9e8ae01d | 1598 | struct wait_opts wo; |
161550d7 EB |
1599 | struct pid *pid = NULL; |
1600 | enum pid_type type; | |
1da177e4 LT |
1601 | long ret; |
1602 | ||
1603 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1604 | return -EINVAL; | |
1605 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1606 | return -EINVAL; | |
1607 | ||
1608 | switch (which) { | |
1609 | case P_ALL: | |
161550d7 | 1610 | type = PIDTYPE_MAX; |
1da177e4 LT |
1611 | break; |
1612 | case P_PID: | |
161550d7 EB |
1613 | type = PIDTYPE_PID; |
1614 | if (upid <= 0) | |
1da177e4 LT |
1615 | return -EINVAL; |
1616 | break; | |
1617 | case P_PGID: | |
161550d7 EB |
1618 | type = PIDTYPE_PGID; |
1619 | if (upid <= 0) | |
1da177e4 | 1620 | return -EINVAL; |
1da177e4 LT |
1621 | break; |
1622 | default: | |
1623 | return -EINVAL; | |
1624 | } | |
1625 | ||
161550d7 EB |
1626 | if (type < PIDTYPE_MAX) |
1627 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1628 | |
1629 | wo.wo_type = type; | |
1630 | wo.wo_pid = pid; | |
1631 | wo.wo_flags = options; | |
1632 | wo.wo_info = infop; | |
1633 | wo.wo_stat = NULL; | |
1634 | wo.wo_rusage = ru; | |
1635 | ret = do_wait(&wo); | |
dfe16dfa VM |
1636 | |
1637 | if (ret > 0) { | |
1638 | ret = 0; | |
1639 | } else if (infop) { | |
1640 | /* | |
1641 | * For a WNOHANG return, clear out all the fields | |
1642 | * we would set so the user can easily tell the | |
1643 | * difference. | |
1644 | */ | |
1645 | if (!ret) | |
1646 | ret = put_user(0, &infop->si_signo); | |
1647 | if (!ret) | |
1648 | ret = put_user(0, &infop->si_errno); | |
1649 | if (!ret) | |
1650 | ret = put_user(0, &infop->si_code); | |
1651 | if (!ret) | |
1652 | ret = put_user(0, &infop->si_pid); | |
1653 | if (!ret) | |
1654 | ret = put_user(0, &infop->si_uid); | |
1655 | if (!ret) | |
1656 | ret = put_user(0, &infop->si_status); | |
1657 | } | |
1658 | ||
161550d7 | 1659 | put_pid(pid); |
1da177e4 LT |
1660 | return ret; |
1661 | } | |
1662 | ||
754fe8d2 HC |
1663 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1664 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1665 | { |
9e8ae01d | 1666 | struct wait_opts wo; |
161550d7 EB |
1667 | struct pid *pid = NULL; |
1668 | enum pid_type type; | |
1da177e4 LT |
1669 | long ret; |
1670 | ||
1671 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1672 | __WNOTHREAD|__WCLONE|__WALL)) | |
1673 | return -EINVAL; | |
161550d7 EB |
1674 | |
1675 | if (upid == -1) | |
1676 | type = PIDTYPE_MAX; | |
1677 | else if (upid < 0) { | |
1678 | type = PIDTYPE_PGID; | |
1679 | pid = find_get_pid(-upid); | |
1680 | } else if (upid == 0) { | |
1681 | type = PIDTYPE_PGID; | |
2ae448ef | 1682 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1683 | } else /* upid > 0 */ { |
1684 | type = PIDTYPE_PID; | |
1685 | pid = find_get_pid(upid); | |
1686 | } | |
1687 | ||
9e8ae01d ON |
1688 | wo.wo_type = type; |
1689 | wo.wo_pid = pid; | |
1690 | wo.wo_flags = options | WEXITED; | |
1691 | wo.wo_info = NULL; | |
1692 | wo.wo_stat = stat_addr; | |
1693 | wo.wo_rusage = ru; | |
1694 | ret = do_wait(&wo); | |
161550d7 | 1695 | put_pid(pid); |
1da177e4 | 1696 | |
1da177e4 LT |
1697 | return ret; |
1698 | } | |
1699 | ||
1700 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1701 | ||
1702 | /* | |
1703 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1704 | * implemented by calling sys_wait4() from libc.a. | |
1705 | */ | |
17da2bd9 | 1706 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1707 | { |
1708 | return sys_wait4(pid, stat_addr, options, NULL); | |
1709 | } | |
1710 | ||
1711 | #endif |