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> |
1da177e4 | 23 | #include <linux/binfmts.h> |
ab516013 | 24 | #include <linux/nsproxy.h> |
84d73786 | 25 | #include <linux/pid_namespace.h> |
1da177e4 LT |
26 | #include <linux/ptrace.h> |
27 | #include <linux/profile.h> | |
28 | #include <linux/mount.h> | |
29 | #include <linux/proc_fs.h> | |
49d769d5 | 30 | #include <linux/kthread.h> |
1da177e4 | 31 | #include <linux/mempolicy.h> |
c757249a | 32 | #include <linux/taskstats_kern.h> |
ca74e92b | 33 | #include <linux/delayacct.h> |
83144186 | 34 | #include <linux/freezer.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> |
1da177e4 LT |
54 | |
55 | #include <asm/uaccess.h> | |
56 | #include <asm/unistd.h> | |
57 | #include <asm/pgtable.h> | |
58 | #include <asm/mmu_context.h> | |
59 | ||
408b664a AB |
60 | static void exit_mm(struct task_struct * tsk); |
61 | ||
d40e48e0 | 62 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
63 | { |
64 | nr_threads--; | |
65 | detach_pid(p, PIDTYPE_PID); | |
d40e48e0 | 66 | if (group_dead) { |
1da177e4 LT |
67 | detach_pid(p, PIDTYPE_PGID); |
68 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 69 | |
5e85d4ab | 70 | list_del_rcu(&p->tasks); |
9cd80bbb | 71 | list_del_init(&p->sibling); |
73b9ebfe | 72 | __get_cpu_var(process_counts)--; |
1da177e4 | 73 | } |
47e65328 | 74 | list_del_rcu(&p->thread_group); |
1da177e4 LT |
75 | } |
76 | ||
6a14c5c9 ON |
77 | /* |
78 | * This function expects the tasklist_lock write-locked. | |
79 | */ | |
80 | static void __exit_signal(struct task_struct *tsk) | |
81 | { | |
82 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 83 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 84 | struct sighand_struct *sighand; |
4ada856f | 85 | struct tty_struct *uninitialized_var(tty); |
6a14c5c9 | 86 | |
d11c563d PM |
87 | sighand = rcu_dereference_check(tsk->sighand, |
88 | rcu_read_lock_held() || | |
db1466b3 | 89 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
90 | spin_lock(&sighand->siglock); |
91 | ||
92 | posix_cpu_timers_exit(tsk); | |
d40e48e0 | 93 | if (group_dead) { |
6a14c5c9 | 94 | posix_cpu_timers_exit_group(tsk); |
4ada856f ON |
95 | tty = sig->tty; |
96 | sig->tty = NULL; | |
4a599942 | 97 | } else { |
6a14c5c9 ON |
98 | /* |
99 | * If there is any task waiting for the group exit | |
100 | * then notify it: | |
101 | */ | |
d344193a | 102 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 103 | wake_up_process(sig->group_exit_task); |
6db840fa | 104 | |
6a14c5c9 ON |
105 | if (tsk == sig->curr_target) |
106 | sig->curr_target = next_thread(tsk); | |
107 | /* | |
108 | * Accumulate here the counters for all threads but the | |
109 | * group leader as they die, so they can be added into | |
110 | * the process-wide totals when those are taken. | |
111 | * The group leader stays around as a zombie as long | |
112 | * as there are other threads. When it gets reaped, | |
113 | * the exit.c code will add its counts into these totals. | |
114 | * We won't ever get here for the group leader, since it | |
115 | * will have been the last reference on the signal_struct. | |
116 | */ | |
0cf55e1e HS |
117 | sig->utime = cputime_add(sig->utime, tsk->utime); |
118 | sig->stime = cputime_add(sig->stime, tsk->stime); | |
d5b7c78e | 119 | sig->gtime = cputime_add(sig->gtime, tsk->gtime); |
6a14c5c9 ON |
120 | sig->min_flt += tsk->min_flt; |
121 | sig->maj_flt += tsk->maj_flt; | |
122 | sig->nvcsw += tsk->nvcsw; | |
123 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
124 | sig->inblock += task_io_get_inblock(tsk); |
125 | sig->oublock += task_io_get_oublock(tsk); | |
5995477a | 126 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
32bd671d | 127 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
128 | } |
129 | ||
b3ac022c | 130 | sig->nr_threads--; |
d40e48e0 | 131 | __unhash_process(tsk, group_dead); |
5876700c | 132 | |
da7978b0 ON |
133 | /* |
134 | * Do this under ->siglock, we can race with another thread | |
135 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
136 | */ | |
137 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 138 | tsk->sighand = NULL; |
6a14c5c9 | 139 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 140 | |
a7e5328a | 141 | __cleanup_sighand(sighand); |
6a14c5c9 | 142 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
d40e48e0 | 143 | if (group_dead) { |
6a14c5c9 | 144 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 145 | tty_kref_put(tty); |
6a14c5c9 ON |
146 | } |
147 | } | |
148 | ||
8c7904a0 EB |
149 | static void delayed_put_task_struct(struct rcu_head *rhp) |
150 | { | |
0a16b607 MD |
151 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
152 | ||
4e231c79 | 153 | perf_event_delayed_put(tsk); |
0a16b607 MD |
154 | trace_sched_process_free(tsk); |
155 | put_task_struct(tsk); | |
8c7904a0 EB |
156 | } |
157 | ||
f470021a | 158 | |
1da177e4 LT |
159 | void release_task(struct task_struct * p) |
160 | { | |
36c8b586 | 161 | struct task_struct *leader; |
1da177e4 | 162 | int zap_leader; |
1f09f974 | 163 | repeat: |
dae33574 | 164 | tracehook_prepare_release_task(p); |
c69e8d9c | 165 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
166 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
167 | rcu_read_lock(); | |
c69e8d9c | 168 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 169 | rcu_read_unlock(); |
c69e8d9c | 170 | |
60347f67 | 171 | proc_flush_task(p); |
0203026b | 172 | |
1da177e4 | 173 | write_lock_irq(&tasklist_lock); |
dae33574 | 174 | tracehook_finish_release_task(p); |
1da177e4 | 175 | __exit_signal(p); |
35f5cad8 | 176 | |
1da177e4 LT |
177 | /* |
178 | * If we are the last non-leader member of the thread | |
179 | * group, and the leader is zombie, then notify the | |
180 | * group leader's parent process. (if it wants notification.) | |
181 | */ | |
182 | zap_leader = 0; | |
183 | leader = p->group_leader; | |
184 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
d839fd4d | 185 | BUG_ON(task_detached(leader)); |
1da177e4 LT |
186 | do_notify_parent(leader, leader->exit_signal); |
187 | /* | |
188 | * If we were the last child thread and the leader has | |
189 | * exited already, and the leader's parent ignores SIGCHLD, | |
190 | * then we are the one who should release the leader. | |
191 | * | |
192 | * do_notify_parent() will have marked it self-reaping in | |
193 | * that case. | |
194 | */ | |
d839fd4d | 195 | zap_leader = task_detached(leader); |
dae33574 RM |
196 | |
197 | /* | |
198 | * This maintains the invariant that release_task() | |
199 | * only runs on a task in EXIT_DEAD, just for sanity. | |
200 | */ | |
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 | */ | |
0475ac08 | 243 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
244 | { |
245 | struct task_struct *p; | |
1da177e4 | 246 | |
0475ac08 | 247 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
248 | if ((p == ignored_task) || |
249 | (p->exit_state && thread_group_empty(p)) || | |
250 | is_global_init(p->real_parent)) | |
1da177e4 | 251 | continue; |
05e83df6 | 252 | |
0475ac08 | 253 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
254 | task_session(p->real_parent) == task_session(p)) |
255 | return 0; | |
0475ac08 | 256 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
257 | |
258 | return 1; | |
1da177e4 LT |
259 | } |
260 | ||
3e7cd6c4 | 261 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
262 | { |
263 | int retval; | |
264 | ||
265 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 266 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
267 | read_unlock(&tasklist_lock); |
268 | ||
269 | return retval; | |
270 | } | |
271 | ||
0475ac08 | 272 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
273 | { |
274 | int retval = 0; | |
275 | struct task_struct *p; | |
276 | ||
0475ac08 | 277 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
338077e5 | 278 | if (!task_is_stopped(p)) |
1da177e4 | 279 | continue; |
1da177e4 LT |
280 | retval = 1; |
281 | break; | |
0475ac08 | 282 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
283 | return retval; |
284 | } | |
285 | ||
f49ee505 ON |
286 | /* |
287 | * Check to see if any process groups have become orphaned as | |
288 | * a result of our exiting, and if they have any stopped jobs, | |
289 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
290 | */ | |
291 | static void | |
292 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
293 | { | |
294 | struct pid *pgrp = task_pgrp(tsk); | |
295 | struct task_struct *ignored_task = tsk; | |
296 | ||
297 | if (!parent) | |
298 | /* exit: our father is in a different pgrp than | |
299 | * we are and we were the only connection outside. | |
300 | */ | |
301 | parent = tsk->real_parent; | |
302 | else | |
303 | /* reparent: our child is in a different pgrp than | |
304 | * we are, and it was the only connection outside. | |
305 | */ | |
306 | ignored_task = NULL; | |
307 | ||
308 | if (task_pgrp(parent) != pgrp && | |
309 | task_session(parent) == task_session(tsk) && | |
310 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
311 | has_stopped_jobs(pgrp)) { | |
312 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
313 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
314 | } | |
315 | } | |
316 | ||
1da177e4 | 317 | /** |
49d769d5 | 318 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
319 | * |
320 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
321 | * it ever exits, it should generally reparent itself to kthreadd so it |
322 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
323 | * |
324 | * The various task state such as scheduling policy and priority may have | |
325 | * been inherited from a user process, so we reset them to sane values here. | |
326 | * | |
49d769d5 | 327 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 328 | */ |
49d769d5 | 329 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
330 | { |
331 | write_lock_irq(&tasklist_lock); | |
332 | ||
333 | ptrace_unlink(current); | |
334 | /* Reparent to init */ | |
49d769d5 | 335 | current->real_parent = current->parent = kthreadd_task; |
f470021a | 336 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
1da177e4 LT |
337 | |
338 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
339 | current->exit_signal = SIGCHLD; | |
340 | ||
e05606d3 | 341 | if (task_nice(current) < 0) |
1da177e4 LT |
342 | set_user_nice(current, 0); |
343 | /* cpus_allowed? */ | |
344 | /* rt_priority? */ | |
345 | /* signals? */ | |
1da177e4 LT |
346 | memcpy(current->signal->rlim, init_task.signal->rlim, |
347 | sizeof(current->signal->rlim)); | |
d84f4f99 DH |
348 | |
349 | atomic_inc(&init_cred.usage); | |
350 | commit_creds(&init_cred); | |
1da177e4 | 351 | write_unlock_irq(&tasklist_lock); |
1da177e4 LT |
352 | } |
353 | ||
8520d7c7 | 354 | void __set_special_pids(struct pid *pid) |
1da177e4 | 355 | { |
e19f247a | 356 | struct task_struct *curr = current->group_leader; |
1da177e4 | 357 | |
0d0df599 | 358 | if (task_session(curr) != pid) |
7d8da096 | 359 | change_pid(curr, PIDTYPE_SID, pid); |
1b0f7ffd ON |
360 | |
361 | if (task_pgrp(curr) != pid) | |
7d8da096 | 362 | change_pid(curr, PIDTYPE_PGID, pid); |
1da177e4 LT |
363 | } |
364 | ||
8520d7c7 | 365 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
366 | { |
367 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 368 | __set_special_pids(pid); |
1da177e4 LT |
369 | write_unlock_irq(&tasklist_lock); |
370 | } | |
371 | ||
372 | /* | |
87245135 ON |
373 | * Let kernel threads use this to say that they allow a certain signal. |
374 | * Must not be used if kthread was cloned with CLONE_SIGHAND. | |
1da177e4 LT |
375 | */ |
376 | int allow_signal(int sig) | |
377 | { | |
7ed20e1a | 378 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
379 | return -EINVAL; |
380 | ||
381 | spin_lock_irq(¤t->sighand->siglock); | |
87245135 | 382 | /* This is only needed for daemonize()'ed kthreads */ |
1da177e4 | 383 | sigdelset(¤t->blocked, sig); |
87245135 ON |
384 | /* |
385 | * Kernel threads handle their own signals. Let the signal code | |
386 | * know it'll be handled, so that they don't get converted to | |
387 | * SIGKILL or just silently dropped. | |
388 | */ | |
389 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
1da177e4 LT |
390 | recalc_sigpending(); |
391 | spin_unlock_irq(¤t->sighand->siglock); | |
392 | return 0; | |
393 | } | |
394 | ||
395 | EXPORT_SYMBOL(allow_signal); | |
396 | ||
397 | int disallow_signal(int sig) | |
398 | { | |
7ed20e1a | 399 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
400 | return -EINVAL; |
401 | ||
402 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 403 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
404 | recalc_sigpending(); |
405 | spin_unlock_irq(¤t->sighand->siglock); | |
406 | return 0; | |
407 | } | |
408 | ||
409 | EXPORT_SYMBOL(disallow_signal); | |
410 | ||
411 | /* | |
412 | * Put all the gunge required to become a kernel thread without | |
413 | * attached user resources in one place where it belongs. | |
414 | */ | |
415 | ||
416 | void daemonize(const char *name, ...) | |
417 | { | |
418 | va_list args; | |
1da177e4 LT |
419 | sigset_t blocked; |
420 | ||
421 | va_start(args, name); | |
422 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
423 | va_end(args); | |
424 | ||
425 | /* | |
426 | * If we were started as result of loading a module, close all of the | |
427 | * user space pages. We don't need them, and if we didn't close them | |
428 | * they would be locked into memory. | |
429 | */ | |
430 | exit_mm(current); | |
83144186 RW |
431 | /* |
432 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
433 | * or suspend transition begins right now. | |
434 | */ | |
7b34e428 | 435 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
1da177e4 | 436 | |
8520d7c7 ON |
437 | if (current->nsproxy != &init_nsproxy) { |
438 | get_nsproxy(&init_nsproxy); | |
439 | switch_task_namespaces(current, &init_nsproxy); | |
440 | } | |
297bd42b | 441 | set_special_pids(&init_struct_pid); |
24ec839c | 442 | proc_clear_tty(current); |
1da177e4 LT |
443 | |
444 | /* Block and flush all signals */ | |
445 | sigfillset(&blocked); | |
446 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
447 | flush_signals(current); | |
448 | ||
449 | /* Become as one with the init task */ | |
450 | ||
3e93cd67 | 451 | daemonize_fs_struct(); |
d4c5e41f | 452 | exit_files(current); |
1da177e4 LT |
453 | current->files = init_task.files; |
454 | atomic_inc(¤t->files->count); | |
455 | ||
49d769d5 | 456 | reparent_to_kthreadd(); |
1da177e4 LT |
457 | } |
458 | ||
459 | EXPORT_SYMBOL(daemonize); | |
460 | ||
858119e1 | 461 | static void close_files(struct files_struct * files) |
1da177e4 LT |
462 | { |
463 | int i, j; | |
badf1662 | 464 | struct fdtable *fdt; |
1da177e4 LT |
465 | |
466 | j = 0; | |
4fb3a538 DS |
467 | |
468 | /* | |
469 | * It is safe to dereference the fd table without RCU or | |
470 | * ->file_lock because this is the last reference to the | |
d11c563d | 471 | * files structure. But use RCU to shut RCU-lockdep up. |
4fb3a538 | 472 | */ |
d11c563d | 473 | rcu_read_lock(); |
badf1662 | 474 | fdt = files_fdtable(files); |
d11c563d | 475 | rcu_read_unlock(); |
1da177e4 LT |
476 | for (;;) { |
477 | unsigned long set; | |
478 | i = j * __NFDBITS; | |
bbea9f69 | 479 | if (i >= fdt->max_fds) |
1da177e4 | 480 | break; |
badf1662 | 481 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
482 | while (set) { |
483 | if (set & 1) { | |
badf1662 | 484 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 485 | if (file) { |
1da177e4 | 486 | filp_close(file, files); |
944be0b2 IM |
487 | cond_resched(); |
488 | } | |
1da177e4 LT |
489 | } |
490 | i++; | |
491 | set >>= 1; | |
492 | } | |
493 | } | |
494 | } | |
495 | ||
496 | struct files_struct *get_files_struct(struct task_struct *task) | |
497 | { | |
498 | struct files_struct *files; | |
499 | ||
500 | task_lock(task); | |
501 | files = task->files; | |
502 | if (files) | |
503 | atomic_inc(&files->count); | |
504 | task_unlock(task); | |
505 | ||
506 | return files; | |
507 | } | |
508 | ||
7ad5b3a5 | 509 | void put_files_struct(struct files_struct *files) |
1da177e4 | 510 | { |
badf1662 DS |
511 | struct fdtable *fdt; |
512 | ||
1da177e4 LT |
513 | if (atomic_dec_and_test(&files->count)) { |
514 | close_files(files); | |
515 | /* | |
516 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
517 | * If the fdtable was embedded, pass files for freeing |
518 | * at the end of the RCU grace period. Otherwise, | |
519 | * you can free files immediately. | |
1da177e4 | 520 | */ |
d11c563d | 521 | rcu_read_lock(); |
badf1662 | 522 | fdt = files_fdtable(files); |
4fd45812 | 523 | if (fdt != &files->fdtab) |
ab2af1f5 | 524 | kmem_cache_free(files_cachep, files); |
01b2d93c | 525 | free_fdtable(fdt); |
d11c563d | 526 | rcu_read_unlock(); |
1da177e4 LT |
527 | } |
528 | } | |
529 | ||
3b125388 | 530 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 531 | { |
3b125388 | 532 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
533 | struct files_struct *old; |
534 | ||
535 | old = tsk->files; | |
536 | task_lock(tsk); | |
537 | tsk->files = files; | |
538 | task_unlock(tsk); | |
539 | put_files_struct(old); | |
540 | } | |
3b9b8ab6 | 541 | |
1ec7f1dd | 542 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
543 | { |
544 | struct files_struct * files = tsk->files; | |
545 | ||
546 | if (files) { | |
547 | task_lock(tsk); | |
548 | tsk->files = NULL; | |
549 | task_unlock(tsk); | |
550 | put_files_struct(files); | |
551 | } | |
552 | } | |
553 | ||
cf475ad2 BS |
554 | #ifdef CONFIG_MM_OWNER |
555 | /* | |
556 | * Task p is exiting and it owned mm, lets find a new owner for it | |
557 | */ | |
558 | static inline int | |
559 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) | |
560 | { | |
561 | /* | |
562 | * If there are other users of the mm and the owner (us) is exiting | |
563 | * we need to find a new owner to take on the responsibility. | |
564 | */ | |
cf475ad2 BS |
565 | if (atomic_read(&mm->mm_users) <= 1) |
566 | return 0; | |
567 | if (mm->owner != p) | |
568 | return 0; | |
569 | return 1; | |
570 | } | |
571 | ||
572 | void mm_update_next_owner(struct mm_struct *mm) | |
573 | { | |
574 | struct task_struct *c, *g, *p = current; | |
575 | ||
576 | retry: | |
577 | if (!mm_need_new_owner(mm, p)) | |
578 | return; | |
579 | ||
580 | read_lock(&tasklist_lock); | |
581 | /* | |
582 | * Search in the children | |
583 | */ | |
584 | list_for_each_entry(c, &p->children, sibling) { | |
585 | if (c->mm == mm) | |
586 | goto assign_new_owner; | |
587 | } | |
588 | ||
589 | /* | |
590 | * Search in the siblings | |
591 | */ | |
dea33cfd | 592 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
593 | if (c->mm == mm) |
594 | goto assign_new_owner; | |
595 | } | |
596 | ||
597 | /* | |
598 | * Search through everything else. We should not get | |
599 | * here often | |
600 | */ | |
601 | do_each_thread(g, c) { | |
602 | if (c->mm == mm) | |
603 | goto assign_new_owner; | |
604 | } while_each_thread(g, c); | |
605 | ||
606 | read_unlock(&tasklist_lock); | |
31a78f23 BS |
607 | /* |
608 | * We found no owner yet mm_users > 1: this implies that we are | |
609 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 610 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 611 | */ |
31a78f23 | 612 | mm->owner = NULL; |
cf475ad2 BS |
613 | return; |
614 | ||
615 | assign_new_owner: | |
616 | BUG_ON(c == p); | |
617 | get_task_struct(c); | |
618 | /* | |
619 | * The task_lock protects c->mm from changing. | |
620 | * We always want mm->owner->mm == mm | |
621 | */ | |
622 | task_lock(c); | |
e5991371 HD |
623 | /* |
624 | * Delay read_unlock() till we have the task_lock() | |
625 | * to ensure that c does not slip away underneath us | |
626 | */ | |
627 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
628 | if (c->mm != mm) { |
629 | task_unlock(c); | |
630 | put_task_struct(c); | |
631 | goto retry; | |
632 | } | |
cf475ad2 BS |
633 | mm->owner = c; |
634 | task_unlock(c); | |
635 | put_task_struct(c); | |
636 | } | |
637 | #endif /* CONFIG_MM_OWNER */ | |
638 | ||
1da177e4 LT |
639 | /* |
640 | * Turn us into a lazy TLB process if we | |
641 | * aren't already.. | |
642 | */ | |
408b664a | 643 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
644 | { |
645 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 646 | struct core_state *core_state; |
1da177e4 LT |
647 | |
648 | mm_release(tsk, mm); | |
649 | if (!mm) | |
650 | return; | |
651 | /* | |
652 | * Serialize with any possible pending coredump. | |
999d9fc1 | 653 | * We must hold mmap_sem around checking core_state |
1da177e4 | 654 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 655 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
656 | * group with ->mm != NULL. |
657 | */ | |
658 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
659 | core_state = mm->core_state; |
660 | if (core_state) { | |
661 | struct core_thread self; | |
1da177e4 | 662 | up_read(&mm->mmap_sem); |
1da177e4 | 663 | |
b564daf8 ON |
664 | self.task = tsk; |
665 | self.next = xchg(&core_state->dumper.next, &self); | |
666 | /* | |
667 | * Implies mb(), the result of xchg() must be visible | |
668 | * to core_state->dumper. | |
669 | */ | |
670 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
671 | complete(&core_state->startup); | |
1da177e4 | 672 | |
a94e2d40 ON |
673 | for (;;) { |
674 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
675 | if (!self.task) /* see coredump_finish() */ | |
676 | break; | |
677 | schedule(); | |
678 | } | |
679 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
680 | down_read(&mm->mmap_sem); |
681 | } | |
682 | atomic_inc(&mm->mm_count); | |
125e1874 | 683 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
684 | /* more a memory barrier than a real lock */ |
685 | task_lock(tsk); | |
686 | tsk->mm = NULL; | |
687 | up_read(&mm->mmap_sem); | |
688 | enter_lazy_tlb(mm, current); | |
0c1eecfb RW |
689 | /* We don't want this task to be frozen prematurely */ |
690 | clear_freeze_flag(tsk); | |
3d5992d2 YH |
691 | if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) |
692 | atomic_dec(&mm->oom_disable_count); | |
1da177e4 | 693 | task_unlock(tsk); |
cf475ad2 | 694 | mm_update_next_owner(mm); |
1da177e4 LT |
695 | mmput(mm); |
696 | } | |
697 | ||
1da177e4 LT |
698 | /* |
699 | * When we die, we re-parent all our children. | |
700 | * Try to give them to another thread in our thread | |
701 | * group, and if no such member exists, give it to | |
84d73786 SB |
702 | * the child reaper process (ie "init") in our pid |
703 | * space. | |
1da177e4 | 704 | */ |
950bbabb | 705 | static struct task_struct *find_new_reaper(struct task_struct *father) |
1da177e4 | 706 | { |
950bbabb ON |
707 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
708 | struct task_struct *thread; | |
1da177e4 | 709 | |
950bbabb ON |
710 | thread = father; |
711 | while_each_thread(father, thread) { | |
712 | if (thread->flags & PF_EXITING) | |
713 | continue; | |
714 | if (unlikely(pid_ns->child_reaper == father)) | |
715 | pid_ns->child_reaper = thread; | |
716 | return thread; | |
717 | } | |
1da177e4 | 718 | |
950bbabb ON |
719 | if (unlikely(pid_ns->child_reaper == father)) { |
720 | write_unlock_irq(&tasklist_lock); | |
721 | if (unlikely(pid_ns == &init_pid_ns)) | |
722 | panic("Attempted to kill init!"); | |
1da177e4 | 723 | |
950bbabb ON |
724 | zap_pid_ns_processes(pid_ns); |
725 | write_lock_irq(&tasklist_lock); | |
1da177e4 | 726 | /* |
950bbabb ON |
727 | * We can not clear ->child_reaper or leave it alone. |
728 | * There may by stealth EXIT_DEAD tasks on ->children, | |
729 | * forget_original_parent() must move them somewhere. | |
1da177e4 | 730 | */ |
950bbabb | 731 | pid_ns->child_reaper = init_pid_ns.child_reaper; |
1da177e4 | 732 | } |
762a24be | 733 | |
950bbabb ON |
734 | return pid_ns->child_reaper; |
735 | } | |
736 | ||
5dfc80be ON |
737 | /* |
738 | * Any that need to be release_task'd are put on the @dead list. | |
739 | */ | |
9cd80bbb | 740 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
741 | struct list_head *dead) |
742 | { | |
5dfc80be ON |
743 | list_move_tail(&p->sibling, &p->real_parent->children); |
744 | ||
745 | if (task_detached(p)) | |
746 | return; | |
747 | /* | |
748 | * If this is a threaded reparent there is no need to | |
749 | * notify anyone anything has happened. | |
750 | */ | |
751 | if (same_thread_group(p->real_parent, father)) | |
752 | return; | |
753 | ||
754 | /* We don't want people slaying init. */ | |
755 | p->exit_signal = SIGCHLD; | |
756 | ||
757 | /* If it has exited notify the new parent about this child's death. */ | |
5cb11446 | 758 | if (!task_ptrace(p) && |
5dfc80be ON |
759 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
760 | do_notify_parent(p, p->exit_signal); | |
761 | if (task_detached(p)) { | |
762 | p->exit_state = EXIT_DEAD; | |
763 | list_move_tail(&p->sibling, dead); | |
764 | } | |
765 | } | |
766 | ||
767 | kill_orphaned_pgrp(p, father); | |
768 | } | |
769 | ||
762a24be | 770 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 771 | { |
950bbabb | 772 | struct task_struct *p, *n, *reaper; |
5dfc80be | 773 | LIST_HEAD(dead_children); |
762a24be ON |
774 | |
775 | write_lock_irq(&tasklist_lock); | |
c7e49c14 ON |
776 | /* |
777 | * Note that exit_ptrace() and find_new_reaper() might | |
778 | * drop tasklist_lock and reacquire it. | |
779 | */ | |
780 | exit_ptrace(father); | |
950bbabb | 781 | reaper = find_new_reaper(father); |
f470021a | 782 | |
03ff1797 | 783 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
9cd80bbb ON |
784 | struct task_struct *t = p; |
785 | do { | |
786 | t->real_parent = reaper; | |
787 | if (t->parent == father) { | |
788 | BUG_ON(task_ptrace(t)); | |
789 | t->parent = t->real_parent; | |
790 | } | |
791 | if (t->pdeath_signal) | |
792 | group_send_sig_info(t->pdeath_signal, | |
793 | SEND_SIG_NOINFO, t); | |
794 | } while_each_thread(p, t); | |
795 | reparent_leader(father, p, &dead_children); | |
1da177e4 | 796 | } |
762a24be | 797 | write_unlock_irq(&tasklist_lock); |
5dfc80be | 798 | |
762a24be | 799 | BUG_ON(!list_empty(&father->children)); |
762a24be | 800 | |
5dfc80be ON |
801 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
802 | list_del_init(&p->sibling); | |
39c626ae ON |
803 | release_task(p); |
804 | } | |
1da177e4 LT |
805 | } |
806 | ||
807 | /* | |
808 | * Send signals to all our closest relatives so that they know | |
809 | * to properly mourn us.. | |
810 | */ | |
821c7de7 | 811 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 812 | { |
2b2a1ff6 RM |
813 | int signal; |
814 | void *cookie; | |
1da177e4 | 815 | |
1da177e4 LT |
816 | /* |
817 | * This does two things: | |
818 | * | |
819 | * A. Make init inherit all the child processes | |
820 | * B. Check to see if any process groups have become orphaned | |
821 | * as a result of our exiting, and if they have any stopped | |
822 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
823 | */ | |
762a24be | 824 | forget_original_parent(tsk); |
2e4a7072 | 825 | exit_task_namespaces(tsk); |
1da177e4 | 826 | |
762a24be | 827 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
828 | if (group_dead) |
829 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 830 | |
24728448 | 831 | /* Let father know we died |
1da177e4 LT |
832 | * |
833 | * Thread signals are configurable, but you aren't going to use | |
d4c5e41f | 834 | * that to send signals to arbitary processes. |
1da177e4 LT |
835 | * That stops right now. |
836 | * | |
837 | * If the parent exec id doesn't match the exec id we saved | |
838 | * when we started then we know the parent has changed security | |
839 | * domain. | |
840 | * | |
841 | * If our self_exec id doesn't match our parent_exec_id then | |
842 | * we have changed execution domain as these two values started | |
843 | * the same after a fork. | |
1da177e4 | 844 | */ |
d839fd4d | 845 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
f49ee505 | 846 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
432870da | 847 | tsk->self_exec_id != tsk->parent_exec_id)) |
1da177e4 LT |
848 | tsk->exit_signal = SIGCHLD; |
849 | ||
2b2a1ff6 | 850 | signal = tracehook_notify_death(tsk, &cookie, group_dead); |
5c7edcd7 | 851 | if (signal >= 0) |
2b2a1ff6 | 852 | signal = do_notify_parent(tsk, signal); |
1da177e4 | 853 | |
5c7edcd7 | 854 | tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE; |
1da177e4 | 855 | |
9c339168 ON |
856 | /* mt-exec, de_thread() is waiting for group leader */ |
857 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 858 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
859 | write_unlock_irq(&tasklist_lock); |
860 | ||
2b2a1ff6 RM |
861 | tracehook_report_death(tsk, signal, cookie, group_dead); |
862 | ||
1da177e4 | 863 | /* If the process is dead, release it - nobody will wait for it */ |
5c7edcd7 | 864 | if (signal == DEATH_REAP) |
1da177e4 | 865 | release_task(tsk); |
1da177e4 LT |
866 | } |
867 | ||
e18eecb8 JD |
868 | #ifdef CONFIG_DEBUG_STACK_USAGE |
869 | static void check_stack_usage(void) | |
870 | { | |
871 | static DEFINE_SPINLOCK(low_water_lock); | |
872 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
873 | unsigned long free; |
874 | ||
7c9f8861 | 875 | free = stack_not_used(current); |
e18eecb8 JD |
876 | |
877 | if (free >= lowest_to_date) | |
878 | return; | |
879 | ||
880 | spin_lock(&low_water_lock); | |
881 | if (free < lowest_to_date) { | |
882 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
883 | "left\n", | |
884 | current->comm, free); | |
885 | lowest_to_date = free; | |
886 | } | |
887 | spin_unlock(&low_water_lock); | |
888 | } | |
889 | #else | |
890 | static inline void check_stack_usage(void) {} | |
891 | #endif | |
892 | ||
7ad5b3a5 | 893 | NORET_TYPE void do_exit(long code) |
1da177e4 LT |
894 | { |
895 | struct task_struct *tsk = current; | |
896 | int group_dead; | |
897 | ||
898 | profile_task_exit(tsk); | |
899 | ||
22e2c507 JA |
900 | WARN_ON(atomic_read(&tsk->fs_excl)); |
901 | ||
1da177e4 LT |
902 | if (unlikely(in_interrupt())) |
903 | panic("Aiee, killing interrupt handler!"); | |
904 | if (unlikely(!tsk->pid)) | |
905 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 906 | |
30199f5a | 907 | tracehook_report_exit(&code); |
1da177e4 | 908 | |
e0e81739 DH |
909 | validate_creds_for_do_exit(tsk); |
910 | ||
df164db5 AN |
911 | /* |
912 | * We're taking recursive faults here in do_exit. Safest is to just | |
913 | * leave this task alone and wait for reboot. | |
914 | */ | |
915 | if (unlikely(tsk->flags & PF_EXITING)) { | |
916 | printk(KERN_ALERT | |
917 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
918 | /* |
919 | * We can do this unlocked here. The futex code uses | |
920 | * this flag just to verify whether the pi state | |
921 | * cleanup has been done or not. In the worst case it | |
922 | * loops once more. We pretend that the cleanup was | |
923 | * done as there is no way to return. Either the | |
924 | * OWNER_DIED bit is set by now or we push the blocked | |
925 | * task into the wait for ever nirwana as well. | |
926 | */ | |
927 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
928 | set_current_state(TASK_UNINTERRUPTIBLE); |
929 | schedule(); | |
930 | } | |
931 | ||
3aa551c9 TG |
932 | exit_irq_thread(); |
933 | ||
d12619b5 | 934 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
935 | /* |
936 | * tsk->flags are checked in the futex code to protect against | |
937 | * an exiting task cleaning up the robust pi futexes. | |
938 | */ | |
d2ee7198 | 939 | smp_mb(); |
1d615482 | 940 | raw_spin_unlock_wait(&tsk->pi_lock); |
1da177e4 | 941 | |
1da177e4 LT |
942 | if (unlikely(in_atomic())) |
943 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 944 | current->comm, task_pid_nr(current), |
1da177e4 LT |
945 | preempt_count()); |
946 | ||
947 | acct_update_integrals(tsk); | |
34e55232 | 948 | /* sync mm's RSS info before statistics gathering */ |
a3a2e76c KH |
949 | if (tsk->mm) |
950 | sync_mm_rss(tsk, tsk->mm); | |
1da177e4 | 951 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 952 | if (group_dead) { |
778e9a9c | 953 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 954 | exit_itimers(tsk->signal); |
1f10206c JP |
955 | if (tsk->mm) |
956 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 957 | } |
f6ec29a4 | 958 | acct_collect(code, group_dead); |
522ed776 MT |
959 | if (group_dead) |
960 | tty_audit_exit(); | |
fa84cb93 AV |
961 | if (unlikely(tsk->audit_context)) |
962 | audit_free(tsk); | |
115085ea | 963 | |
f2ab6d88 | 964 | tsk->exit_code = code; |
115085ea | 965 | taskstats_exit(tsk, group_dead); |
c757249a | 966 | |
1da177e4 LT |
967 | exit_mm(tsk); |
968 | ||
0e464814 | 969 | if (group_dead) |
f6ec29a4 | 970 | acct_process(); |
0a16b607 MD |
971 | trace_sched_process_exit(tsk); |
972 | ||
1da177e4 | 973 | exit_sem(tsk); |
1ec7f1dd AV |
974 | exit_files(tsk); |
975 | exit_fs(tsk); | |
e18eecb8 | 976 | check_stack_usage(); |
1da177e4 | 977 | exit_thread(); |
b4f48b63 | 978 | cgroup_exit(tsk, 1); |
1da177e4 | 979 | |
5ec93d11 | 980 | if (group_dead) |
1da177e4 LT |
981 | disassociate_ctty(1); |
982 | ||
a1261f54 | 983 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 | 984 | |
9f46080c | 985 | proc_exit_connector(tsk); |
33b2fb30 | 986 | |
24f1e32c FW |
987 | /* |
988 | * FIXME: do that only when needed, using sched_exit tracepoint | |
989 | */ | |
990 | flush_ptrace_hw_breakpoint(tsk); | |
33b2fb30 IM |
991 | /* |
992 | * Flush inherited counters to the parent - before the parent | |
993 | * gets woken up by child-exit notifications. | |
994 | */ | |
cdd6c482 | 995 | perf_event_exit_task(tsk); |
33b2fb30 | 996 | |
821c7de7 | 997 | exit_notify(tsk, group_dead); |
1da177e4 | 998 | #ifdef CONFIG_NUMA |
c0ff7453 | 999 | task_lock(tsk); |
f0be3d32 | 1000 | mpol_put(tsk->mempolicy); |
1da177e4 | 1001 | tsk->mempolicy = NULL; |
c0ff7453 | 1002 | task_unlock(tsk); |
1da177e4 | 1003 | #endif |
42b2dd0a | 1004 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
1005 | if (unlikely(current->pi_state_cache)) |
1006 | kfree(current->pi_state_cache); | |
42b2dd0a | 1007 | #endif |
de5097c2 | 1008 | /* |
9a11b49a | 1009 | * Make sure we are holding no locks: |
de5097c2 | 1010 | */ |
9a11b49a | 1011 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1012 | /* |
1013 | * We can do this unlocked here. The futex code uses this flag | |
1014 | * just to verify whether the pi state cleanup has been done | |
1015 | * or not. In the worst case it loops once more. | |
1016 | */ | |
1017 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1018 | |
afc847b7 | 1019 | if (tsk->io_context) |
b69f2292 | 1020 | exit_io_context(tsk); |
afc847b7 | 1021 | |
b92ce558 JA |
1022 | if (tsk->splice_pipe) |
1023 | __free_pipe_info(tsk->splice_pipe); | |
1024 | ||
e0e81739 DH |
1025 | validate_creds_for_do_exit(tsk); |
1026 | ||
7407251a | 1027 | preempt_disable(); |
f41d911f | 1028 | exit_rcu(); |
55a101f8 | 1029 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1030 | tsk->state = TASK_DEAD; |
1da177e4 LT |
1031 | schedule(); |
1032 | BUG(); | |
1033 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1034 | for (;;) |
1035 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1036 | } |
1037 | ||
012914da RA |
1038 | EXPORT_SYMBOL_GPL(do_exit); |
1039 | ||
1da177e4 LT |
1040 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1041 | { | |
1042 | if (comp) | |
1043 | complete(comp); | |
55a101f8 | 1044 | |
1da177e4 LT |
1045 | do_exit(code); |
1046 | } | |
1047 | ||
1048 | EXPORT_SYMBOL(complete_and_exit); | |
1049 | ||
754fe8d2 | 1050 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
1051 | { |
1052 | do_exit((error_code&0xff)<<8); | |
1053 | } | |
1054 | ||
1da177e4 LT |
1055 | /* |
1056 | * Take down every thread in the group. This is called by fatal signals | |
1057 | * as well as by sys_exit_group (below). | |
1058 | */ | |
1059 | NORET_TYPE void | |
1060 | do_group_exit(int exit_code) | |
1061 | { | |
bfc4b089 ON |
1062 | struct signal_struct *sig = current->signal; |
1063 | ||
1da177e4 LT |
1064 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1065 | ||
bfc4b089 ON |
1066 | if (signal_group_exit(sig)) |
1067 | exit_code = sig->group_exit_code; | |
1da177e4 | 1068 | else if (!thread_group_empty(current)) { |
1da177e4 | 1069 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1070 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1071 | if (signal_group_exit(sig)) |
1da177e4 LT |
1072 | /* Another thread got here before we took the lock. */ |
1073 | exit_code = sig->group_exit_code; | |
1074 | else { | |
1da177e4 | 1075 | sig->group_exit_code = exit_code; |
ed5d2cac | 1076 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1077 | zap_other_threads(current); |
1078 | } | |
1079 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1080 | } |
1081 | ||
1082 | do_exit(exit_code); | |
1083 | /* NOTREACHED */ | |
1084 | } | |
1085 | ||
1086 | /* | |
1087 | * this kills every thread in the thread group. Note that any externally | |
1088 | * wait4()-ing process will get the correct exit code - even if this | |
1089 | * thread is not the thread group leader. | |
1090 | */ | |
754fe8d2 | 1091 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
1092 | { |
1093 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
1094 | /* NOTREACHED */ |
1095 | return 0; | |
1da177e4 LT |
1096 | } |
1097 | ||
9e8ae01d ON |
1098 | struct wait_opts { |
1099 | enum pid_type wo_type; | |
9e8ae01d | 1100 | int wo_flags; |
e1eb1ebc | 1101 | struct pid *wo_pid; |
9e8ae01d ON |
1102 | |
1103 | struct siginfo __user *wo_info; | |
1104 | int __user *wo_stat; | |
1105 | struct rusage __user *wo_rusage; | |
1106 | ||
0b7570e7 | 1107 | wait_queue_t child_wait; |
9e8ae01d ON |
1108 | int notask_error; |
1109 | }; | |
1110 | ||
989264f4 ON |
1111 | static inline |
1112 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 1113 | { |
989264f4 ON |
1114 | if (type != PIDTYPE_PID) |
1115 | task = task->group_leader; | |
1116 | return task->pids[type].pid; | |
161550d7 EB |
1117 | } |
1118 | ||
989264f4 | 1119 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1120 | { |
5c01ba49 ON |
1121 | return wo->wo_type == PIDTYPE_MAX || |
1122 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1123 | } | |
1da177e4 | 1124 | |
5c01ba49 ON |
1125 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
1126 | { | |
1127 | if (!eligible_pid(wo, p)) | |
1128 | return 0; | |
1da177e4 LT |
1129 | /* Wait for all children (clone and not) if __WALL is set; |
1130 | * otherwise, wait for clone children *only* if __WCLONE is | |
1131 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1132 | * A "clone" child here is one that reports to its parent | |
1133 | * using a signal other than SIGCHLD.) */ | |
9e8ae01d ON |
1134 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
1135 | && !(wo->wo_flags & __WALL)) | |
1da177e4 | 1136 | return 0; |
1da177e4 | 1137 | |
14dd0b81 | 1138 | return 1; |
1da177e4 LT |
1139 | } |
1140 | ||
9e8ae01d ON |
1141 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
1142 | pid_t pid, uid_t uid, int why, int status) | |
1da177e4 | 1143 | { |
9e8ae01d ON |
1144 | struct siginfo __user *infop; |
1145 | int retval = wo->wo_rusage | |
1146 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
36c8b586 | 1147 | |
1da177e4 | 1148 | put_task_struct(p); |
9e8ae01d | 1149 | infop = wo->wo_info; |
b6fe2d11 VM |
1150 | if (infop) { |
1151 | if (!retval) | |
1152 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1153 | if (!retval) | |
1154 | retval = put_user(0, &infop->si_errno); | |
1155 | if (!retval) | |
1156 | retval = put_user((short)why, &infop->si_code); | |
1157 | if (!retval) | |
1158 | retval = put_user(pid, &infop->si_pid); | |
1159 | if (!retval) | |
1160 | retval = put_user(uid, &infop->si_uid); | |
1161 | if (!retval) | |
1162 | retval = put_user(status, &infop->si_status); | |
1163 | } | |
1da177e4 LT |
1164 | if (!retval) |
1165 | retval = pid; | |
1166 | return retval; | |
1167 | } | |
1168 | ||
1169 | /* | |
1170 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1171 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1172 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1173 | * released the lock and the system call should return. | |
1174 | */ | |
9e8ae01d | 1175 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1176 | { |
1177 | unsigned long state; | |
2f4e6e2a | 1178 | int retval, status, traced; |
6c5f3e7b | 1179 | pid_t pid = task_pid_vnr(p); |
c69e8d9c | 1180 | uid_t uid = __task_cred(p)->uid; |
9e8ae01d | 1181 | struct siginfo __user *infop; |
1da177e4 | 1182 | |
9e8ae01d | 1183 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1184 | return 0; |
1185 | ||
9e8ae01d | 1186 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1da177e4 | 1187 | int exit_code = p->exit_code; |
f3abd4f9 | 1188 | int why; |
1da177e4 | 1189 | |
1da177e4 LT |
1190 | get_task_struct(p); |
1191 | read_unlock(&tasklist_lock); | |
1192 | if ((exit_code & 0x7f) == 0) { | |
1193 | why = CLD_EXITED; | |
1194 | status = exit_code >> 8; | |
1195 | } else { | |
1196 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1197 | status = exit_code & 0x7f; | |
1198 | } | |
9e8ae01d | 1199 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1da177e4 LT |
1200 | } |
1201 | ||
1202 | /* | |
1203 | * Try to move the task's state to DEAD | |
1204 | * only one thread is allowed to do this: | |
1205 | */ | |
1206 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1207 | if (state != EXIT_ZOMBIE) { | |
1208 | BUG_ON(state != EXIT_DEAD); | |
1209 | return 0; | |
1210 | } | |
1da177e4 | 1211 | |
53b6f9fb | 1212 | traced = ptrace_reparented(p); |
befca967 ON |
1213 | /* |
1214 | * It can be ptraced but not reparented, check | |
1215 | * !task_detached() to filter out sub-threads. | |
1216 | */ | |
1217 | if (likely(!traced) && likely(!task_detached(p))) { | |
3795e161 JJ |
1218 | struct signal_struct *psig; |
1219 | struct signal_struct *sig; | |
1f10206c | 1220 | unsigned long maxrss; |
0cf55e1e | 1221 | cputime_t tgutime, tgstime; |
3795e161 | 1222 | |
1da177e4 LT |
1223 | /* |
1224 | * The resource counters for the group leader are in its | |
1225 | * own task_struct. Those for dead threads in the group | |
1226 | * are in its signal_struct, as are those for the child | |
1227 | * processes it has previously reaped. All these | |
1228 | * accumulate in the parent's signal_struct c* fields. | |
1229 | * | |
1230 | * We don't bother to take a lock here to protect these | |
1231 | * p->signal fields, because they are only touched by | |
1232 | * __exit_signal, which runs with tasklist_lock | |
1233 | * write-locked anyway, and so is excluded here. We do | |
d1e98f42 | 1234 | * need to protect the access to parent->signal fields, |
1da177e4 LT |
1235 | * as other threads in the parent group can be right |
1236 | * here reaping other children at the same time. | |
0cf55e1e HS |
1237 | * |
1238 | * We use thread_group_times() to get times for the thread | |
1239 | * group, which consolidates times for all threads in the | |
1240 | * group including the group leader. | |
1da177e4 | 1241 | */ |
0cf55e1e | 1242 | thread_group_times(p, &tgutime, &tgstime); |
d1e98f42 ON |
1243 | spin_lock_irq(&p->real_parent->sighand->siglock); |
1244 | psig = p->real_parent->signal; | |
3795e161 JJ |
1245 | sig = p->signal; |
1246 | psig->cutime = | |
1247 | cputime_add(psig->cutime, | |
0cf55e1e HS |
1248 | cputime_add(tgutime, |
1249 | sig->cutime)); | |
3795e161 JJ |
1250 | psig->cstime = |
1251 | cputime_add(psig->cstime, | |
0cf55e1e HS |
1252 | cputime_add(tgstime, |
1253 | sig->cstime)); | |
9ac52315 LV |
1254 | psig->cgtime = |
1255 | cputime_add(psig->cgtime, | |
1256 | cputime_add(p->gtime, | |
1257 | cputime_add(sig->gtime, | |
1258 | sig->cgtime))); | |
3795e161 JJ |
1259 | psig->cmin_flt += |
1260 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1261 | psig->cmaj_flt += | |
1262 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1263 | psig->cnvcsw += | |
1264 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1265 | psig->cnivcsw += | |
1266 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1267 | psig->cinblock += |
1268 | task_io_get_inblock(p) + | |
1269 | sig->inblock + sig->cinblock; | |
1270 | psig->coublock += | |
1271 | task_io_get_oublock(p) + | |
1272 | sig->oublock + sig->coublock; | |
1f10206c JP |
1273 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1274 | if (psig->cmaxrss < maxrss) | |
1275 | psig->cmaxrss = maxrss; | |
5995477a AR |
1276 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1277 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
d1e98f42 | 1278 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
1da177e4 LT |
1279 | } |
1280 | ||
1281 | /* | |
1282 | * Now we are sure this task is interesting, and no other | |
1283 | * thread can reap it because we set its state to EXIT_DEAD. | |
1284 | */ | |
1285 | read_unlock(&tasklist_lock); | |
1286 | ||
9e8ae01d ON |
1287 | retval = wo->wo_rusage |
1288 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 LT |
1289 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1290 | ? p->signal->group_exit_code : p->exit_code; | |
9e8ae01d ON |
1291 | if (!retval && wo->wo_stat) |
1292 | retval = put_user(status, wo->wo_stat); | |
1293 | ||
1294 | infop = wo->wo_info; | |
1da177e4 LT |
1295 | if (!retval && infop) |
1296 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1297 | if (!retval && infop) | |
1298 | retval = put_user(0, &infop->si_errno); | |
1299 | if (!retval && infop) { | |
1300 | int why; | |
1301 | ||
1302 | if ((status & 0x7f) == 0) { | |
1303 | why = CLD_EXITED; | |
1304 | status >>= 8; | |
1305 | } else { | |
1306 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1307 | status &= 0x7f; | |
1308 | } | |
1309 | retval = put_user((short)why, &infop->si_code); | |
1310 | if (!retval) | |
1311 | retval = put_user(status, &infop->si_status); | |
1312 | } | |
1313 | if (!retval && infop) | |
3a515e4a | 1314 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1315 | if (!retval && infop) |
c69e8d9c | 1316 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1317 | if (!retval) |
3a515e4a | 1318 | retval = pid; |
2f4e6e2a ON |
1319 | |
1320 | if (traced) { | |
1da177e4 | 1321 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1322 | /* We dropped tasklist, ptracer could die and untrace */ |
1323 | ptrace_unlink(p); | |
1324 | /* | |
1325 | * If this is not a detached task, notify the parent. | |
1326 | * If it's still not detached after that, don't release | |
1327 | * it now. | |
1328 | */ | |
d839fd4d | 1329 | if (!task_detached(p)) { |
2f4e6e2a | 1330 | do_notify_parent(p, p->exit_signal); |
d839fd4d | 1331 | if (!task_detached(p)) { |
2f4e6e2a ON |
1332 | p->exit_state = EXIT_ZOMBIE; |
1333 | p = NULL; | |
1da177e4 LT |
1334 | } |
1335 | } | |
1336 | write_unlock_irq(&tasklist_lock); | |
1337 | } | |
1338 | if (p != NULL) | |
1339 | release_task(p); | |
2f4e6e2a | 1340 | |
1da177e4 LT |
1341 | return retval; |
1342 | } | |
1343 | ||
90bc8d8b ON |
1344 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1345 | { | |
1346 | if (ptrace) { | |
1347 | if (task_is_stopped_or_traced(p)) | |
1348 | return &p->exit_code; | |
1349 | } else { | |
1350 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1351 | return &p->signal->group_exit_code; | |
1352 | } | |
1353 | return NULL; | |
1354 | } | |
1355 | ||
1da177e4 LT |
1356 | /* |
1357 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1358 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1359 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1360 | * released the lock and the system call should return. | |
1361 | */ | |
9e8ae01d ON |
1362 | static int wait_task_stopped(struct wait_opts *wo, |
1363 | int ptrace, struct task_struct *p) | |
1da177e4 | 1364 | { |
9e8ae01d | 1365 | struct siginfo __user *infop; |
90bc8d8b | 1366 | int retval, exit_code, *p_code, why; |
ee7c82da | 1367 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1368 | pid_t pid; |
1da177e4 | 1369 | |
47918025 ON |
1370 | /* |
1371 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1372 | */ | |
9e8ae01d | 1373 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1374 | return 0; |
1375 | ||
ee7c82da ON |
1376 | exit_code = 0; |
1377 | spin_lock_irq(&p->sighand->siglock); | |
1378 | ||
90bc8d8b ON |
1379 | p_code = task_stopped_code(p, ptrace); |
1380 | if (unlikely(!p_code)) | |
ee7c82da ON |
1381 | goto unlock_sig; |
1382 | ||
90bc8d8b | 1383 | exit_code = *p_code; |
ee7c82da ON |
1384 | if (!exit_code) |
1385 | goto unlock_sig; | |
1386 | ||
9e8ae01d | 1387 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1388 | *p_code = 0; |
ee7c82da | 1389 | |
f362b732 | 1390 | uid = task_uid(p); |
ee7c82da ON |
1391 | unlock_sig: |
1392 | spin_unlock_irq(&p->sighand->siglock); | |
1393 | if (!exit_code) | |
1da177e4 LT |
1394 | return 0; |
1395 | ||
1396 | /* | |
1397 | * Now we are pretty sure this task is interesting. | |
1398 | * Make sure it doesn't get reaped out from under us while we | |
1399 | * give up the lock and then examine it below. We don't want to | |
1400 | * keep holding onto the tasklist_lock while we call getrusage and | |
1401 | * possibly take page faults for user memory. | |
1402 | */ | |
1403 | get_task_struct(p); | |
6c5f3e7b | 1404 | pid = task_pid_vnr(p); |
f470021a | 1405 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1406 | read_unlock(&tasklist_lock); |
1407 | ||
9e8ae01d ON |
1408 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1409 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | |
1410 | ||
1411 | retval = wo->wo_rusage | |
1412 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1413 | if (!retval && wo->wo_stat) | |
1414 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | |
1da177e4 | 1415 | |
9e8ae01d | 1416 | infop = wo->wo_info; |
1da177e4 LT |
1417 | if (!retval && infop) |
1418 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1419 | if (!retval && infop) | |
1420 | retval = put_user(0, &infop->si_errno); | |
1421 | if (!retval && infop) | |
6efcae46 | 1422 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1423 | if (!retval && infop) |
1424 | retval = put_user(exit_code, &infop->si_status); | |
1425 | if (!retval && infop) | |
c8950783 | 1426 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1427 | if (!retval && infop) |
ee7c82da | 1428 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1429 | if (!retval) |
c8950783 | 1430 | retval = pid; |
1da177e4 LT |
1431 | put_task_struct(p); |
1432 | ||
1433 | BUG_ON(!retval); | |
1434 | return retval; | |
1435 | } | |
1436 | ||
1437 | /* | |
1438 | * Handle do_wait work for one task in a live, non-stopped state. | |
1439 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1440 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1441 | * released the lock and the system call should return. | |
1442 | */ | |
9e8ae01d | 1443 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1444 | { |
1445 | int retval; | |
1446 | pid_t pid; | |
1447 | uid_t uid; | |
1448 | ||
9e8ae01d | 1449 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1450 | return 0; |
1451 | ||
1da177e4 LT |
1452 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1453 | return 0; | |
1454 | ||
1455 | spin_lock_irq(&p->sighand->siglock); | |
1456 | /* Re-check with the lock held. */ | |
1457 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1458 | spin_unlock_irq(&p->sighand->siglock); | |
1459 | return 0; | |
1460 | } | |
9e8ae01d | 1461 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1462 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
f362b732 | 1463 | uid = task_uid(p); |
1da177e4 LT |
1464 | spin_unlock_irq(&p->sighand->siglock); |
1465 | ||
6c5f3e7b | 1466 | pid = task_pid_vnr(p); |
1da177e4 LT |
1467 | get_task_struct(p); |
1468 | read_unlock(&tasklist_lock); | |
1469 | ||
9e8ae01d ON |
1470 | if (!wo->wo_info) { |
1471 | retval = wo->wo_rusage | |
1472 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 | 1473 | put_task_struct(p); |
9e8ae01d ON |
1474 | if (!retval && wo->wo_stat) |
1475 | retval = put_user(0xffff, wo->wo_stat); | |
1da177e4 | 1476 | if (!retval) |
3a515e4a | 1477 | retval = pid; |
1da177e4 | 1478 | } else { |
9e8ae01d ON |
1479 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1480 | CLD_CONTINUED, SIGCONT); | |
1da177e4 LT |
1481 | BUG_ON(retval == 0); |
1482 | } | |
1483 | ||
1484 | return retval; | |
1485 | } | |
1486 | ||
98abed02 RM |
1487 | /* |
1488 | * Consider @p for a wait by @parent. | |
1489 | * | |
9e8ae01d | 1490 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1491 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1492 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1493 | * then ->notask_error is 0 if @p is an eligible child, |
14dd0b81 | 1494 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1495 | */ |
b6e763f0 ON |
1496 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1497 | struct task_struct *p) | |
98abed02 | 1498 | { |
9e8ae01d | 1499 | int ret = eligible_child(wo, p); |
14dd0b81 | 1500 | if (!ret) |
98abed02 RM |
1501 | return ret; |
1502 | ||
a2322e1d | 1503 | ret = security_task_wait(p); |
14dd0b81 RM |
1504 | if (unlikely(ret < 0)) { |
1505 | /* | |
1506 | * If we have not yet seen any eligible child, | |
1507 | * then let this error code replace -ECHILD. | |
1508 | * A permission error will give the user a clue | |
1509 | * to look for security policy problems, rather | |
1510 | * than for mysterious wait bugs. | |
1511 | */ | |
9e8ae01d ON |
1512 | if (wo->notask_error) |
1513 | wo->notask_error = ret; | |
78a3d9d5 | 1514 | return 0; |
14dd0b81 RM |
1515 | } |
1516 | ||
5cb11446 | 1517 | if (likely(!ptrace) && unlikely(task_ptrace(p))) { |
f470021a RM |
1518 | /* |
1519 | * This child is hidden by ptrace. | |
1520 | * We aren't allowed to see it now, but eventually we will. | |
1521 | */ | |
9e8ae01d | 1522 | wo->notask_error = 0; |
f470021a RM |
1523 | return 0; |
1524 | } | |
1525 | ||
98abed02 RM |
1526 | if (p->exit_state == EXIT_DEAD) |
1527 | return 0; | |
1528 | ||
1529 | /* | |
1530 | * We don't reap group leaders with subthreads. | |
1531 | */ | |
1532 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) | |
9e8ae01d | 1533 | return wait_task_zombie(wo, p); |
98abed02 RM |
1534 | |
1535 | /* | |
1536 | * It's stopped or running now, so it might | |
1537 | * later continue, exit, or stop again. | |
1538 | */ | |
9e8ae01d | 1539 | wo->notask_error = 0; |
98abed02 | 1540 | |
90bc8d8b | 1541 | if (task_stopped_code(p, ptrace)) |
9e8ae01d | 1542 | return wait_task_stopped(wo, ptrace, p); |
98abed02 | 1543 | |
9e8ae01d | 1544 | return wait_task_continued(wo, p); |
98abed02 RM |
1545 | } |
1546 | ||
1547 | /* | |
1548 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1549 | * | |
9e8ae01d | 1550 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1551 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1552 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1553 | * ->notask_error is 0 if there were any eligible children, |
14dd0b81 | 1554 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1555 | */ |
9e8ae01d | 1556 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1557 | { |
1558 | struct task_struct *p; | |
1559 | ||
1560 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb ON |
1561 | int ret = wait_consider_task(wo, 0, p); |
1562 | if (ret) | |
1563 | return ret; | |
98abed02 RM |
1564 | } |
1565 | ||
1566 | return 0; | |
1567 | } | |
1568 | ||
9e8ae01d | 1569 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1570 | { |
1571 | struct task_struct *p; | |
1572 | ||
f470021a | 1573 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1574 | int ret = wait_consider_task(wo, 1, p); |
f470021a | 1575 | if (ret) |
98abed02 | 1576 | return ret; |
98abed02 RM |
1577 | } |
1578 | ||
1579 | return 0; | |
1580 | } | |
1581 | ||
0b7570e7 ON |
1582 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1583 | int sync, void *key) | |
1584 | { | |
1585 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1586 | child_wait); | |
1587 | struct task_struct *p = key; | |
1588 | ||
5c01ba49 | 1589 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1590 | return 0; |
1591 | ||
b4fe5182 ON |
1592 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1593 | return 0; | |
1594 | ||
0b7570e7 ON |
1595 | return default_wake_function(wait, mode, sync, key); |
1596 | } | |
1597 | ||
a7f0765e ON |
1598 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1599 | { | |
0b7570e7 ON |
1600 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1601 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1602 | } |
1603 | ||
9e8ae01d | 1604 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1605 | { |
1da177e4 | 1606 | struct task_struct *tsk; |
98abed02 | 1607 | int retval; |
1da177e4 | 1608 | |
9e8ae01d | 1609 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1610 | |
0b7570e7 ON |
1611 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1612 | wo->child_wait.private = current; | |
1613 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1614 | repeat: |
98abed02 RM |
1615 | /* |
1616 | * If there is nothing that can match our critiera just get out. | |
9e8ae01d ON |
1617 | * We will clear ->notask_error to zero if we see any child that |
1618 | * might later match our criteria, even if we are not able to reap | |
1619 | * it yet. | |
98abed02 | 1620 | */ |
64a16caf | 1621 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1622 | if ((wo->wo_type < PIDTYPE_MAX) && |
1623 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1624 | goto notask; |
161550d7 | 1625 | |
f95d39d1 | 1626 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1627 | read_lock(&tasklist_lock); |
1628 | tsk = current; | |
1629 | do { | |
64a16caf ON |
1630 | retval = do_wait_thread(wo, tsk); |
1631 | if (retval) | |
1632 | goto end; | |
9e8ae01d | 1633 | |
64a16caf ON |
1634 | retval = ptrace_do_wait(wo, tsk); |
1635 | if (retval) | |
98abed02 | 1636 | goto end; |
98abed02 | 1637 | |
9e8ae01d | 1638 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1639 | break; |
a3f6dfb7 | 1640 | } while_each_thread(current, tsk); |
1da177e4 | 1641 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1642 | |
64a16caf | 1643 | notask: |
9e8ae01d ON |
1644 | retval = wo->notask_error; |
1645 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1646 | retval = -ERESTARTSYS; |
98abed02 RM |
1647 | if (!signal_pending(current)) { |
1648 | schedule(); | |
1649 | goto repeat; | |
1650 | } | |
1da177e4 | 1651 | } |
1da177e4 | 1652 | end: |
f95d39d1 | 1653 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1654 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1655 | return retval; |
1656 | } | |
1657 | ||
17da2bd9 HC |
1658 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1659 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1660 | { |
9e8ae01d | 1661 | struct wait_opts wo; |
161550d7 EB |
1662 | struct pid *pid = NULL; |
1663 | enum pid_type type; | |
1da177e4 LT |
1664 | long ret; |
1665 | ||
1666 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1667 | return -EINVAL; | |
1668 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1669 | return -EINVAL; | |
1670 | ||
1671 | switch (which) { | |
1672 | case P_ALL: | |
161550d7 | 1673 | type = PIDTYPE_MAX; |
1da177e4 LT |
1674 | break; |
1675 | case P_PID: | |
161550d7 EB |
1676 | type = PIDTYPE_PID; |
1677 | if (upid <= 0) | |
1da177e4 LT |
1678 | return -EINVAL; |
1679 | break; | |
1680 | case P_PGID: | |
161550d7 EB |
1681 | type = PIDTYPE_PGID; |
1682 | if (upid <= 0) | |
1da177e4 | 1683 | return -EINVAL; |
1da177e4 LT |
1684 | break; |
1685 | default: | |
1686 | return -EINVAL; | |
1687 | } | |
1688 | ||
161550d7 EB |
1689 | if (type < PIDTYPE_MAX) |
1690 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1691 | |
1692 | wo.wo_type = type; | |
1693 | wo.wo_pid = pid; | |
1694 | wo.wo_flags = options; | |
1695 | wo.wo_info = infop; | |
1696 | wo.wo_stat = NULL; | |
1697 | wo.wo_rusage = ru; | |
1698 | ret = do_wait(&wo); | |
dfe16dfa VM |
1699 | |
1700 | if (ret > 0) { | |
1701 | ret = 0; | |
1702 | } else if (infop) { | |
1703 | /* | |
1704 | * For a WNOHANG return, clear out all the fields | |
1705 | * we would set so the user can easily tell the | |
1706 | * difference. | |
1707 | */ | |
1708 | if (!ret) | |
1709 | ret = put_user(0, &infop->si_signo); | |
1710 | if (!ret) | |
1711 | ret = put_user(0, &infop->si_errno); | |
1712 | if (!ret) | |
1713 | ret = put_user(0, &infop->si_code); | |
1714 | if (!ret) | |
1715 | ret = put_user(0, &infop->si_pid); | |
1716 | if (!ret) | |
1717 | ret = put_user(0, &infop->si_uid); | |
1718 | if (!ret) | |
1719 | ret = put_user(0, &infop->si_status); | |
1720 | } | |
1721 | ||
161550d7 | 1722 | put_pid(pid); |
1da177e4 LT |
1723 | |
1724 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1725 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1726 | return ret; |
1727 | } | |
1728 | ||
754fe8d2 HC |
1729 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1730 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1731 | { |
9e8ae01d | 1732 | struct wait_opts wo; |
161550d7 EB |
1733 | struct pid *pid = NULL; |
1734 | enum pid_type type; | |
1da177e4 LT |
1735 | long ret; |
1736 | ||
1737 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1738 | __WNOTHREAD|__WCLONE|__WALL)) | |
1739 | return -EINVAL; | |
161550d7 EB |
1740 | |
1741 | if (upid == -1) | |
1742 | type = PIDTYPE_MAX; | |
1743 | else if (upid < 0) { | |
1744 | type = PIDTYPE_PGID; | |
1745 | pid = find_get_pid(-upid); | |
1746 | } else if (upid == 0) { | |
1747 | type = PIDTYPE_PGID; | |
2ae448ef | 1748 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1749 | } else /* upid > 0 */ { |
1750 | type = PIDTYPE_PID; | |
1751 | pid = find_get_pid(upid); | |
1752 | } | |
1753 | ||
9e8ae01d ON |
1754 | wo.wo_type = type; |
1755 | wo.wo_pid = pid; | |
1756 | wo.wo_flags = options | WEXITED; | |
1757 | wo.wo_info = NULL; | |
1758 | wo.wo_stat = stat_addr; | |
1759 | wo.wo_rusage = ru; | |
1760 | ret = do_wait(&wo); | |
161550d7 | 1761 | put_pid(pid); |
1da177e4 LT |
1762 | |
1763 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1764 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1765 | return ret; |
1766 | } | |
1767 | ||
1768 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1769 | ||
1770 | /* | |
1771 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1772 | * implemented by calling sys_wait4() from libc.a. | |
1773 | */ | |
17da2bd9 | 1774 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1775 | { |
1776 | return sys_wait4(pid, stat_addr, options, NULL); | |
1777 | } | |
1778 | ||
1779 | #endif |