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> |
1da177e4 LT |
52 | |
53 | #include <asm/uaccess.h> | |
54 | #include <asm/unistd.h> | |
55 | #include <asm/pgtable.h> | |
56 | #include <asm/mmu_context.h> | |
d84f4f99 | 57 | #include "cred-internals.h" |
1da177e4 | 58 | |
408b664a AB |
59 | static void exit_mm(struct task_struct * tsk); |
60 | ||
1da177e4 LT |
61 | static void __unhash_process(struct task_struct *p) |
62 | { | |
63 | nr_threads--; | |
64 | detach_pid(p, PIDTYPE_PID); | |
1da177e4 LT |
65 | if (thread_group_leader(p)) { |
66 | detach_pid(p, PIDTYPE_PGID); | |
67 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 68 | |
5e85d4ab | 69 | list_del_rcu(&p->tasks); |
73b9ebfe | 70 | __get_cpu_var(process_counts)--; |
1da177e4 | 71 | } |
47e65328 | 72 | list_del_rcu(&p->thread_group); |
f470021a | 73 | list_del_init(&p->sibling); |
1da177e4 LT |
74 | } |
75 | ||
6a14c5c9 ON |
76 | /* |
77 | * This function expects the tasklist_lock write-locked. | |
78 | */ | |
79 | static void __exit_signal(struct task_struct *tsk) | |
80 | { | |
81 | struct signal_struct *sig = tsk->signal; | |
82 | struct sighand_struct *sighand; | |
83 | ||
84 | BUG_ON(!sig); | |
85 | BUG_ON(!atomic_read(&sig->count)); | |
86 | ||
6a14c5c9 ON |
87 | sighand = rcu_dereference(tsk->sighand); |
88 | spin_lock(&sighand->siglock); | |
89 | ||
90 | posix_cpu_timers_exit(tsk); | |
91 | if (atomic_dec_and_test(&sig->count)) | |
92 | posix_cpu_timers_exit_group(tsk); | |
93 | else { | |
94 | /* | |
95 | * If there is any task waiting for the group exit | |
96 | * then notify it: | |
97 | */ | |
6db840fa | 98 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) |
6a14c5c9 | 99 | wake_up_process(sig->group_exit_task); |
6db840fa | 100 | |
6a14c5c9 ON |
101 | if (tsk == sig->curr_target) |
102 | sig->curr_target = next_thread(tsk); | |
103 | /* | |
104 | * Accumulate here the counters for all threads but the | |
105 | * group leader as they die, so they can be added into | |
106 | * the process-wide totals when those are taken. | |
107 | * The group leader stays around as a zombie as long | |
108 | * as there are other threads. When it gets reaped, | |
109 | * the exit.c code will add its counts into these totals. | |
110 | * We won't ever get here for the group leader, since it | |
111 | * will have been the last reference on the signal_struct. | |
112 | */ | |
32bd671d PZ |
113 | sig->utime = cputime_add(sig->utime, task_utime(tsk)); |
114 | sig->stime = cputime_add(sig->stime, task_stime(tsk)); | |
49048622 | 115 | sig->gtime = cputime_add(sig->gtime, task_gtime(tsk)); |
6a14c5c9 ON |
116 | sig->min_flt += tsk->min_flt; |
117 | sig->maj_flt += tsk->maj_flt; | |
118 | sig->nvcsw += tsk->nvcsw; | |
119 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
120 | sig->inblock += task_io_get_inblock(tsk); |
121 | sig->oublock += task_io_get_oublock(tsk); | |
5995477a | 122 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
32bd671d | 123 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
124 | sig = NULL; /* Marker for below. */ |
125 | } | |
126 | ||
5876700c ON |
127 | __unhash_process(tsk); |
128 | ||
da7978b0 ON |
129 | /* |
130 | * Do this under ->siglock, we can race with another thread | |
131 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
132 | */ | |
133 | flush_sigqueue(&tsk->pending); | |
134 | ||
6a14c5c9 | 135 | tsk->signal = NULL; |
a7e5328a | 136 | tsk->sighand = NULL; |
6a14c5c9 | 137 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 138 | |
a7e5328a | 139 | __cleanup_sighand(sighand); |
6a14c5c9 | 140 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
6a14c5c9 ON |
141 | if (sig) { |
142 | flush_sigqueue(&sig->shared_pending); | |
093a8e8a | 143 | taskstats_tgid_free(sig); |
ad474cac ON |
144 | /* |
145 | * Make sure ->signal can't go away under rq->lock, | |
146 | * see account_group_exec_runtime(). | |
147 | */ | |
148 | task_rq_unlock_wait(tsk); | |
6a14c5c9 ON |
149 | __cleanup_signal(sig); |
150 | } | |
151 | } | |
152 | ||
8c7904a0 EB |
153 | static void delayed_put_task_struct(struct rcu_head *rhp) |
154 | { | |
0a16b607 MD |
155 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
156 | ||
cdd6c482 IM |
157 | #ifdef CONFIG_PERF_EVENTS |
158 | WARN_ON_ONCE(tsk->perf_event_ctxp); | |
eef6cbf5 | 159 | #endif |
0a16b607 MD |
160 | trace_sched_process_free(tsk); |
161 | put_task_struct(tsk); | |
8c7904a0 EB |
162 | } |
163 | ||
f470021a | 164 | |
1da177e4 LT |
165 | void release_task(struct task_struct * p) |
166 | { | |
36c8b586 | 167 | struct task_struct *leader; |
1da177e4 | 168 | int zap_leader; |
1f09f974 | 169 | repeat: |
dae33574 | 170 | tracehook_prepare_release_task(p); |
c69e8d9c DH |
171 | /* don't need to get the RCU readlock here - the process is dead and |
172 | * can't be modifying its own credentials */ | |
173 | atomic_dec(&__task_cred(p)->user->processes); | |
174 | ||
60347f67 | 175 | proc_flush_task(p); |
0203026b | 176 | |
1da177e4 | 177 | write_lock_irq(&tasklist_lock); |
dae33574 | 178 | tracehook_finish_release_task(p); |
1da177e4 | 179 | __exit_signal(p); |
35f5cad8 | 180 | |
1da177e4 LT |
181 | /* |
182 | * If we are the last non-leader member of the thread | |
183 | * group, and the leader is zombie, then notify the | |
184 | * group leader's parent process. (if it wants notification.) | |
185 | */ | |
186 | zap_leader = 0; | |
187 | leader = p->group_leader; | |
188 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
d839fd4d | 189 | BUG_ON(task_detached(leader)); |
1da177e4 LT |
190 | do_notify_parent(leader, leader->exit_signal); |
191 | /* | |
192 | * If we were the last child thread and the leader has | |
193 | * exited already, and the leader's parent ignores SIGCHLD, | |
194 | * then we are the one who should release the leader. | |
195 | * | |
196 | * do_notify_parent() will have marked it self-reaping in | |
197 | * that case. | |
198 | */ | |
d839fd4d | 199 | zap_leader = task_detached(leader); |
dae33574 RM |
200 | |
201 | /* | |
202 | * This maintains the invariant that release_task() | |
203 | * only runs on a task in EXIT_DEAD, just for sanity. | |
204 | */ | |
205 | if (zap_leader) | |
206 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
207 | } |
208 | ||
1da177e4 | 209 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 210 | release_thread(p); |
8c7904a0 | 211 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
212 | |
213 | p = leader; | |
214 | if (unlikely(zap_leader)) | |
215 | goto repeat; | |
216 | } | |
217 | ||
1da177e4 LT |
218 | /* |
219 | * This checks not only the pgrp, but falls back on the pid if no | |
220 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
221 | * without this... | |
04a2e6a5 EB |
222 | * |
223 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 224 | */ |
04a2e6a5 | 225 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
226 | { |
227 | struct task_struct *p; | |
04a2e6a5 | 228 | struct pid *sid = NULL; |
62dfb554 | 229 | |
04a2e6a5 | 230 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 231 | if (p == NULL) |
04a2e6a5 | 232 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 233 | if (p != NULL) |
04a2e6a5 | 234 | sid = task_session(p); |
62dfb554 | 235 | |
1da177e4 LT |
236 | return sid; |
237 | } | |
238 | ||
239 | /* | |
240 | * Determine if a process group is "orphaned", according to the POSIX | |
241 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
242 | * by terminal-generated stop signals. Newly orphaned process groups are | |
243 | * to receive a SIGHUP and a SIGCONT. | |
244 | * | |
245 | * "I ask you, have you ever known what it is to be an orphan?" | |
246 | */ | |
0475ac08 | 247 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
248 | { |
249 | struct task_struct *p; | |
1da177e4 | 250 | |
0475ac08 | 251 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
252 | if ((p == ignored_task) || |
253 | (p->exit_state && thread_group_empty(p)) || | |
254 | is_global_init(p->real_parent)) | |
1da177e4 | 255 | continue; |
05e83df6 | 256 | |
0475ac08 | 257 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
258 | task_session(p->real_parent) == task_session(p)) |
259 | return 0; | |
0475ac08 | 260 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
261 | |
262 | return 1; | |
1da177e4 LT |
263 | } |
264 | ||
3e7cd6c4 | 265 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
266 | { |
267 | int retval; | |
268 | ||
269 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 270 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
271 | read_unlock(&tasklist_lock); |
272 | ||
273 | return retval; | |
274 | } | |
275 | ||
0475ac08 | 276 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
277 | { |
278 | int retval = 0; | |
279 | struct task_struct *p; | |
280 | ||
0475ac08 | 281 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
338077e5 | 282 | if (!task_is_stopped(p)) |
1da177e4 | 283 | continue; |
1da177e4 LT |
284 | retval = 1; |
285 | break; | |
0475ac08 | 286 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
287 | return retval; |
288 | } | |
289 | ||
f49ee505 ON |
290 | /* |
291 | * Check to see if any process groups have become orphaned as | |
292 | * a result of our exiting, and if they have any stopped jobs, | |
293 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
294 | */ | |
295 | static void | |
296 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
297 | { | |
298 | struct pid *pgrp = task_pgrp(tsk); | |
299 | struct task_struct *ignored_task = tsk; | |
300 | ||
301 | if (!parent) | |
302 | /* exit: our father is in a different pgrp than | |
303 | * we are and we were the only connection outside. | |
304 | */ | |
305 | parent = tsk->real_parent; | |
306 | else | |
307 | /* reparent: our child is in a different pgrp than | |
308 | * we are, and it was the only connection outside. | |
309 | */ | |
310 | ignored_task = NULL; | |
311 | ||
312 | if (task_pgrp(parent) != pgrp && | |
313 | task_session(parent) == task_session(tsk) && | |
314 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
315 | has_stopped_jobs(pgrp)) { | |
316 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
317 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
318 | } | |
319 | } | |
320 | ||
1da177e4 | 321 | /** |
49d769d5 | 322 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
323 | * |
324 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
325 | * it ever exits, it should generally reparent itself to kthreadd so it |
326 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
327 | * |
328 | * The various task state such as scheduling policy and priority may have | |
329 | * been inherited from a user process, so we reset them to sane values here. | |
330 | * | |
49d769d5 | 331 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 332 | */ |
49d769d5 | 333 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
334 | { |
335 | write_lock_irq(&tasklist_lock); | |
336 | ||
337 | ptrace_unlink(current); | |
338 | /* Reparent to init */ | |
49d769d5 | 339 | current->real_parent = current->parent = kthreadd_task; |
f470021a | 340 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
1da177e4 LT |
341 | |
342 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
343 | current->exit_signal = SIGCHLD; | |
344 | ||
e05606d3 | 345 | if (task_nice(current) < 0) |
1da177e4 LT |
346 | set_user_nice(current, 0); |
347 | /* cpus_allowed? */ | |
348 | /* rt_priority? */ | |
349 | /* signals? */ | |
1da177e4 LT |
350 | memcpy(current->signal->rlim, init_task.signal->rlim, |
351 | sizeof(current->signal->rlim)); | |
d84f4f99 DH |
352 | |
353 | atomic_inc(&init_cred.usage); | |
354 | commit_creds(&init_cred); | |
1da177e4 | 355 | write_unlock_irq(&tasklist_lock); |
1da177e4 LT |
356 | } |
357 | ||
8520d7c7 | 358 | void __set_special_pids(struct pid *pid) |
1da177e4 | 359 | { |
e19f247a | 360 | struct task_struct *curr = current->group_leader; |
1da177e4 | 361 | |
02b51df1 | 362 | if (task_session(curr) != pid) { |
7d8da096 | 363 | change_pid(curr, PIDTYPE_SID, pid); |
02b51df1 SJR |
364 | proc_sid_connector(curr); |
365 | } | |
1b0f7ffd ON |
366 | |
367 | if (task_pgrp(curr) != pid) | |
7d8da096 | 368 | change_pid(curr, PIDTYPE_PGID, pid); |
1da177e4 LT |
369 | } |
370 | ||
8520d7c7 | 371 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
372 | { |
373 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 374 | __set_special_pids(pid); |
1da177e4 LT |
375 | write_unlock_irq(&tasklist_lock); |
376 | } | |
377 | ||
378 | /* | |
87245135 ON |
379 | * Let kernel threads use this to say that they allow a certain signal. |
380 | * Must not be used if kthread was cloned with CLONE_SIGHAND. | |
1da177e4 LT |
381 | */ |
382 | int allow_signal(int sig) | |
383 | { | |
7ed20e1a | 384 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
385 | return -EINVAL; |
386 | ||
387 | spin_lock_irq(¤t->sighand->siglock); | |
87245135 | 388 | /* This is only needed for daemonize()'ed kthreads */ |
1da177e4 | 389 | sigdelset(¤t->blocked, sig); |
87245135 ON |
390 | /* |
391 | * Kernel threads handle their own signals. Let the signal code | |
392 | * know it'll be handled, so that they don't get converted to | |
393 | * SIGKILL or just silently dropped. | |
394 | */ | |
395 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
1da177e4 LT |
396 | recalc_sigpending(); |
397 | spin_unlock_irq(¤t->sighand->siglock); | |
398 | return 0; | |
399 | } | |
400 | ||
401 | EXPORT_SYMBOL(allow_signal); | |
402 | ||
403 | int disallow_signal(int sig) | |
404 | { | |
7ed20e1a | 405 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
406 | return -EINVAL; |
407 | ||
408 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 409 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
410 | recalc_sigpending(); |
411 | spin_unlock_irq(¤t->sighand->siglock); | |
412 | return 0; | |
413 | } | |
414 | ||
415 | EXPORT_SYMBOL(disallow_signal); | |
416 | ||
417 | /* | |
418 | * Put all the gunge required to become a kernel thread without | |
419 | * attached user resources in one place where it belongs. | |
420 | */ | |
421 | ||
422 | void daemonize(const char *name, ...) | |
423 | { | |
424 | va_list args; | |
1da177e4 LT |
425 | sigset_t blocked; |
426 | ||
427 | va_start(args, name); | |
428 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
429 | va_end(args); | |
430 | ||
431 | /* | |
432 | * If we were started as result of loading a module, close all of the | |
433 | * user space pages. We don't need them, and if we didn't close them | |
434 | * they would be locked into memory. | |
435 | */ | |
436 | exit_mm(current); | |
83144186 RW |
437 | /* |
438 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
439 | * or suspend transition begins right now. | |
440 | */ | |
7b34e428 | 441 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
1da177e4 | 442 | |
8520d7c7 ON |
443 | if (current->nsproxy != &init_nsproxy) { |
444 | get_nsproxy(&init_nsproxy); | |
445 | switch_task_namespaces(current, &init_nsproxy); | |
446 | } | |
297bd42b | 447 | set_special_pids(&init_struct_pid); |
24ec839c | 448 | proc_clear_tty(current); |
1da177e4 LT |
449 | |
450 | /* Block and flush all signals */ | |
451 | sigfillset(&blocked); | |
452 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
453 | flush_signals(current); | |
454 | ||
455 | /* Become as one with the init task */ | |
456 | ||
3e93cd67 | 457 | daemonize_fs_struct(); |
d4c5e41f | 458 | exit_files(current); |
1da177e4 LT |
459 | current->files = init_task.files; |
460 | atomic_inc(¤t->files->count); | |
461 | ||
49d769d5 | 462 | reparent_to_kthreadd(); |
1da177e4 LT |
463 | } |
464 | ||
465 | EXPORT_SYMBOL(daemonize); | |
466 | ||
858119e1 | 467 | static void close_files(struct files_struct * files) |
1da177e4 LT |
468 | { |
469 | int i, j; | |
badf1662 | 470 | struct fdtable *fdt; |
1da177e4 LT |
471 | |
472 | j = 0; | |
4fb3a538 DS |
473 | |
474 | /* | |
475 | * It is safe to dereference the fd table without RCU or | |
476 | * ->file_lock because this is the last reference to the | |
477 | * files structure. | |
478 | */ | |
badf1662 | 479 | fdt = files_fdtable(files); |
1da177e4 LT |
480 | for (;;) { |
481 | unsigned long set; | |
482 | i = j * __NFDBITS; | |
bbea9f69 | 483 | if (i >= fdt->max_fds) |
1da177e4 | 484 | break; |
badf1662 | 485 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
486 | while (set) { |
487 | if (set & 1) { | |
badf1662 | 488 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 489 | if (file) { |
1da177e4 | 490 | filp_close(file, files); |
944be0b2 IM |
491 | cond_resched(); |
492 | } | |
1da177e4 LT |
493 | } |
494 | i++; | |
495 | set >>= 1; | |
496 | } | |
497 | } | |
498 | } | |
499 | ||
500 | struct files_struct *get_files_struct(struct task_struct *task) | |
501 | { | |
502 | struct files_struct *files; | |
503 | ||
504 | task_lock(task); | |
505 | files = task->files; | |
506 | if (files) | |
507 | atomic_inc(&files->count); | |
508 | task_unlock(task); | |
509 | ||
510 | return files; | |
511 | } | |
512 | ||
7ad5b3a5 | 513 | void put_files_struct(struct files_struct *files) |
1da177e4 | 514 | { |
badf1662 DS |
515 | struct fdtable *fdt; |
516 | ||
1da177e4 LT |
517 | if (atomic_dec_and_test(&files->count)) { |
518 | close_files(files); | |
519 | /* | |
520 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
521 | * If the fdtable was embedded, pass files for freeing |
522 | * at the end of the RCU grace period. Otherwise, | |
523 | * you can free files immediately. | |
1da177e4 | 524 | */ |
badf1662 | 525 | fdt = files_fdtable(files); |
4fd45812 | 526 | if (fdt != &files->fdtab) |
ab2af1f5 | 527 | kmem_cache_free(files_cachep, files); |
01b2d93c | 528 | free_fdtable(fdt); |
1da177e4 LT |
529 | } |
530 | } | |
531 | ||
3b125388 | 532 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 533 | { |
3b125388 | 534 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
535 | struct files_struct *old; |
536 | ||
537 | old = tsk->files; | |
538 | task_lock(tsk); | |
539 | tsk->files = files; | |
540 | task_unlock(tsk); | |
541 | put_files_struct(old); | |
542 | } | |
3b9b8ab6 | 543 | |
1ec7f1dd | 544 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
545 | { |
546 | struct files_struct * files = tsk->files; | |
547 | ||
548 | if (files) { | |
549 | task_lock(tsk); | |
550 | tsk->files = NULL; | |
551 | task_unlock(tsk); | |
552 | put_files_struct(files); | |
553 | } | |
554 | } | |
555 | ||
cf475ad2 BS |
556 | #ifdef CONFIG_MM_OWNER |
557 | /* | |
558 | * Task p is exiting and it owned mm, lets find a new owner for it | |
559 | */ | |
560 | static inline int | |
561 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) | |
562 | { | |
563 | /* | |
564 | * If there are other users of the mm and the owner (us) is exiting | |
565 | * we need to find a new owner to take on the responsibility. | |
566 | */ | |
cf475ad2 BS |
567 | if (atomic_read(&mm->mm_users) <= 1) |
568 | return 0; | |
569 | if (mm->owner != p) | |
570 | return 0; | |
571 | return 1; | |
572 | } | |
573 | ||
574 | void mm_update_next_owner(struct mm_struct *mm) | |
575 | { | |
576 | struct task_struct *c, *g, *p = current; | |
577 | ||
578 | retry: | |
579 | if (!mm_need_new_owner(mm, p)) | |
580 | return; | |
581 | ||
582 | read_lock(&tasklist_lock); | |
583 | /* | |
584 | * Search in the children | |
585 | */ | |
586 | list_for_each_entry(c, &p->children, sibling) { | |
587 | if (c->mm == mm) | |
588 | goto assign_new_owner; | |
589 | } | |
590 | ||
591 | /* | |
592 | * Search in the siblings | |
593 | */ | |
dea33cfd | 594 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
595 | if (c->mm == mm) |
596 | goto assign_new_owner; | |
597 | } | |
598 | ||
599 | /* | |
600 | * Search through everything else. We should not get | |
601 | * here often | |
602 | */ | |
603 | do_each_thread(g, c) { | |
604 | if (c->mm == mm) | |
605 | goto assign_new_owner; | |
606 | } while_each_thread(g, c); | |
607 | ||
608 | read_unlock(&tasklist_lock); | |
31a78f23 BS |
609 | /* |
610 | * We found no owner yet mm_users > 1: this implies that we are | |
611 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 612 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 613 | */ |
31a78f23 | 614 | mm->owner = NULL; |
cf475ad2 BS |
615 | return; |
616 | ||
617 | assign_new_owner: | |
618 | BUG_ON(c == p); | |
619 | get_task_struct(c); | |
620 | /* | |
621 | * The task_lock protects c->mm from changing. | |
622 | * We always want mm->owner->mm == mm | |
623 | */ | |
624 | task_lock(c); | |
e5991371 HD |
625 | /* |
626 | * Delay read_unlock() till we have the task_lock() | |
627 | * to ensure that c does not slip away underneath us | |
628 | */ | |
629 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
630 | if (c->mm != mm) { |
631 | task_unlock(c); | |
632 | put_task_struct(c); | |
633 | goto retry; | |
634 | } | |
cf475ad2 BS |
635 | mm->owner = c; |
636 | task_unlock(c); | |
637 | put_task_struct(c); | |
638 | } | |
639 | #endif /* CONFIG_MM_OWNER */ | |
640 | ||
1da177e4 LT |
641 | /* |
642 | * Turn us into a lazy TLB process if we | |
643 | * aren't already.. | |
644 | */ | |
408b664a | 645 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
646 | { |
647 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 648 | struct core_state *core_state; |
1da177e4 LT |
649 | |
650 | mm_release(tsk, mm); | |
651 | if (!mm) | |
652 | return; | |
653 | /* | |
654 | * Serialize with any possible pending coredump. | |
999d9fc1 | 655 | * We must hold mmap_sem around checking core_state |
1da177e4 | 656 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 657 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
658 | * group with ->mm != NULL. |
659 | */ | |
660 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
661 | core_state = mm->core_state; |
662 | if (core_state) { | |
663 | struct core_thread self; | |
1da177e4 | 664 | up_read(&mm->mmap_sem); |
1da177e4 | 665 | |
b564daf8 ON |
666 | self.task = tsk; |
667 | self.next = xchg(&core_state->dumper.next, &self); | |
668 | /* | |
669 | * Implies mb(), the result of xchg() must be visible | |
670 | * to core_state->dumper. | |
671 | */ | |
672 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
673 | complete(&core_state->startup); | |
1da177e4 | 674 | |
a94e2d40 ON |
675 | for (;;) { |
676 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
677 | if (!self.task) /* see coredump_finish() */ | |
678 | break; | |
679 | schedule(); | |
680 | } | |
681 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
682 | down_read(&mm->mmap_sem); |
683 | } | |
684 | atomic_inc(&mm->mm_count); | |
125e1874 | 685 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
686 | /* more a memory barrier than a real lock */ |
687 | task_lock(tsk); | |
688 | tsk->mm = NULL; | |
689 | up_read(&mm->mmap_sem); | |
690 | enter_lazy_tlb(mm, current); | |
0c1eecfb RW |
691 | /* We don't want this task to be frozen prematurely */ |
692 | clear_freeze_flag(tsk); | |
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 | */ | |
740 | static void reparent_thread(struct task_struct *father, struct task_struct *p, | |
741 | struct list_head *dead) | |
742 | { | |
743 | if (p->pdeath_signal) | |
744 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | |
745 | ||
746 | list_move_tail(&p->sibling, &p->real_parent->children); | |
747 | ||
748 | if (task_detached(p)) | |
749 | return; | |
750 | /* | |
751 | * If this is a threaded reparent there is no need to | |
752 | * notify anyone anything has happened. | |
753 | */ | |
754 | if (same_thread_group(p->real_parent, father)) | |
755 | return; | |
756 | ||
757 | /* We don't want people slaying init. */ | |
758 | p->exit_signal = SIGCHLD; | |
759 | ||
760 | /* If it has exited notify the new parent about this child's death. */ | |
5cb11446 | 761 | if (!task_ptrace(p) && |
5dfc80be ON |
762 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
763 | do_notify_parent(p, p->exit_signal); | |
764 | if (task_detached(p)) { | |
765 | p->exit_state = EXIT_DEAD; | |
766 | list_move_tail(&p->sibling, dead); | |
767 | } | |
768 | } | |
769 | ||
770 | kill_orphaned_pgrp(p, father); | |
771 | } | |
772 | ||
762a24be | 773 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 774 | { |
950bbabb | 775 | struct task_struct *p, *n, *reaper; |
5dfc80be | 776 | LIST_HEAD(dead_children); |
762a24be | 777 | |
39c626ae ON |
778 | exit_ptrace(father); |
779 | ||
762a24be | 780 | write_lock_irq(&tasklist_lock); |
950bbabb | 781 | reaper = find_new_reaper(father); |
f470021a | 782 | |
03ff1797 | 783 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
84eb646b | 784 | p->real_parent = reaper; |
f470021a | 785 | if (p->parent == father) { |
5cb11446 | 786 | BUG_ON(task_ptrace(p)); |
f470021a RM |
787 | p->parent = p->real_parent; |
788 | } | |
5dfc80be | 789 | reparent_thread(father, p, &dead_children); |
1da177e4 | 790 | } |
762a24be | 791 | write_unlock_irq(&tasklist_lock); |
5dfc80be | 792 | |
762a24be | 793 | BUG_ON(!list_empty(&father->children)); |
762a24be | 794 | |
5dfc80be ON |
795 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
796 | list_del_init(&p->sibling); | |
39c626ae ON |
797 | release_task(p); |
798 | } | |
1da177e4 LT |
799 | } |
800 | ||
801 | /* | |
802 | * Send signals to all our closest relatives so that they know | |
803 | * to properly mourn us.. | |
804 | */ | |
821c7de7 | 805 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 806 | { |
2b2a1ff6 RM |
807 | int signal; |
808 | void *cookie; | |
1da177e4 | 809 | |
1da177e4 LT |
810 | /* |
811 | * This does two things: | |
812 | * | |
813 | * A. Make init inherit all the child processes | |
814 | * B. Check to see if any process groups have become orphaned | |
815 | * as a result of our exiting, and if they have any stopped | |
816 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
817 | */ | |
762a24be | 818 | forget_original_parent(tsk); |
2e4a7072 | 819 | exit_task_namespaces(tsk); |
1da177e4 | 820 | |
762a24be | 821 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
822 | if (group_dead) |
823 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 824 | |
24728448 | 825 | /* Let father know we died |
1da177e4 LT |
826 | * |
827 | * Thread signals are configurable, but you aren't going to use | |
d4c5e41f | 828 | * that to send signals to arbitary processes. |
1da177e4 LT |
829 | * That stops right now. |
830 | * | |
831 | * If the parent exec id doesn't match the exec id we saved | |
832 | * when we started then we know the parent has changed security | |
833 | * domain. | |
834 | * | |
835 | * If our self_exec id doesn't match our parent_exec_id then | |
836 | * we have changed execution domain as these two values started | |
837 | * the same after a fork. | |
1da177e4 | 838 | */ |
d839fd4d | 839 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
f49ee505 | 840 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
432870da | 841 | tsk->self_exec_id != tsk->parent_exec_id)) |
1da177e4 LT |
842 | tsk->exit_signal = SIGCHLD; |
843 | ||
2b2a1ff6 | 844 | signal = tracehook_notify_death(tsk, &cookie, group_dead); |
5c7edcd7 | 845 | if (signal >= 0) |
2b2a1ff6 | 846 | signal = do_notify_parent(tsk, signal); |
1da177e4 | 847 | |
5c7edcd7 | 848 | tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE; |
1da177e4 | 849 | |
2800d8d1 | 850 | /* mt-exec, de_thread() is waiting for us */ |
6db840fa | 851 | if (thread_group_leader(tsk) && |
2633f0e5 SV |
852 | tsk->signal->group_exit_task && |
853 | tsk->signal->notify_count < 0) | |
6db840fa ON |
854 | wake_up_process(tsk->signal->group_exit_task); |
855 | ||
1da177e4 LT |
856 | write_unlock_irq(&tasklist_lock); |
857 | ||
2b2a1ff6 RM |
858 | tracehook_report_death(tsk, signal, cookie, group_dead); |
859 | ||
1da177e4 | 860 | /* If the process is dead, release it - nobody will wait for it */ |
5c7edcd7 | 861 | if (signal == DEATH_REAP) |
1da177e4 | 862 | release_task(tsk); |
1da177e4 LT |
863 | } |
864 | ||
e18eecb8 JD |
865 | #ifdef CONFIG_DEBUG_STACK_USAGE |
866 | static void check_stack_usage(void) | |
867 | { | |
868 | static DEFINE_SPINLOCK(low_water_lock); | |
869 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
870 | unsigned long free; |
871 | ||
7c9f8861 | 872 | free = stack_not_used(current); |
e18eecb8 JD |
873 | |
874 | if (free >= lowest_to_date) | |
875 | return; | |
876 | ||
877 | spin_lock(&low_water_lock); | |
878 | if (free < lowest_to_date) { | |
879 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
880 | "left\n", | |
881 | current->comm, free); | |
882 | lowest_to_date = free; | |
883 | } | |
884 | spin_unlock(&low_water_lock); | |
885 | } | |
886 | #else | |
887 | static inline void check_stack_usage(void) {} | |
888 | #endif | |
889 | ||
7ad5b3a5 | 890 | NORET_TYPE void do_exit(long code) |
1da177e4 LT |
891 | { |
892 | struct task_struct *tsk = current; | |
893 | int group_dead; | |
894 | ||
895 | profile_task_exit(tsk); | |
896 | ||
22e2c507 JA |
897 | WARN_ON(atomic_read(&tsk->fs_excl)); |
898 | ||
1da177e4 LT |
899 | if (unlikely(in_interrupt())) |
900 | panic("Aiee, killing interrupt handler!"); | |
901 | if (unlikely(!tsk->pid)) | |
902 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 903 | |
30199f5a | 904 | tracehook_report_exit(&code); |
1da177e4 | 905 | |
e0e81739 DH |
906 | validate_creds_for_do_exit(tsk); |
907 | ||
df164db5 AN |
908 | /* |
909 | * We're taking recursive faults here in do_exit. Safest is to just | |
910 | * leave this task alone and wait for reboot. | |
911 | */ | |
912 | if (unlikely(tsk->flags & PF_EXITING)) { | |
913 | printk(KERN_ALERT | |
914 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
915 | /* |
916 | * We can do this unlocked here. The futex code uses | |
917 | * this flag just to verify whether the pi state | |
918 | * cleanup has been done or not. In the worst case it | |
919 | * loops once more. We pretend that the cleanup was | |
920 | * done as there is no way to return. Either the | |
921 | * OWNER_DIED bit is set by now or we push the blocked | |
922 | * task into the wait for ever nirwana as well. | |
923 | */ | |
924 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
925 | set_current_state(TASK_UNINTERRUPTIBLE); |
926 | schedule(); | |
927 | } | |
928 | ||
3aa551c9 TG |
929 | exit_irq_thread(); |
930 | ||
d12619b5 | 931 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
932 | /* |
933 | * tsk->flags are checked in the futex code to protect against | |
934 | * an exiting task cleaning up the robust pi futexes. | |
935 | */ | |
d2ee7198 ON |
936 | smp_mb(); |
937 | spin_unlock_wait(&tsk->pi_lock); | |
1da177e4 | 938 | |
1da177e4 LT |
939 | if (unlikely(in_atomic())) |
940 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 941 | current->comm, task_pid_nr(current), |
1da177e4 LT |
942 | preempt_count()); |
943 | ||
944 | acct_update_integrals(tsk); | |
901608d9 | 945 | |
1da177e4 | 946 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 947 | if (group_dead) { |
778e9a9c | 948 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 949 | exit_itimers(tsk->signal); |
1f10206c JP |
950 | if (tsk->mm) |
951 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 952 | } |
f6ec29a4 | 953 | acct_collect(code, group_dead); |
522ed776 MT |
954 | if (group_dead) |
955 | tty_audit_exit(); | |
fa84cb93 AV |
956 | if (unlikely(tsk->audit_context)) |
957 | audit_free(tsk); | |
115085ea | 958 | |
f2ab6d88 | 959 | tsk->exit_code = code; |
115085ea | 960 | taskstats_exit(tsk, group_dead); |
c757249a | 961 | |
1da177e4 LT |
962 | exit_mm(tsk); |
963 | ||
0e464814 | 964 | if (group_dead) |
f6ec29a4 | 965 | acct_process(); |
0a16b607 MD |
966 | trace_sched_process_exit(tsk); |
967 | ||
1da177e4 | 968 | exit_sem(tsk); |
1ec7f1dd AV |
969 | exit_files(tsk); |
970 | exit_fs(tsk); | |
e18eecb8 | 971 | check_stack_usage(); |
1da177e4 | 972 | exit_thread(); |
b4f48b63 | 973 | cgroup_exit(tsk, 1); |
1da177e4 LT |
974 | |
975 | if (group_dead && tsk->signal->leader) | |
976 | disassociate_ctty(1); | |
977 | ||
a1261f54 | 978 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 | 979 | |
9f46080c | 980 | proc_exit_connector(tsk); |
33b2fb30 IM |
981 | |
982 | /* | |
983 | * Flush inherited counters to the parent - before the parent | |
984 | * gets woken up by child-exit notifications. | |
985 | */ | |
cdd6c482 | 986 | perf_event_exit_task(tsk); |
33b2fb30 | 987 | |
821c7de7 | 988 | exit_notify(tsk, group_dead); |
1da177e4 | 989 | #ifdef CONFIG_NUMA |
f0be3d32 | 990 | mpol_put(tsk->mempolicy); |
1da177e4 LT |
991 | tsk->mempolicy = NULL; |
992 | #endif | |
42b2dd0a | 993 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
994 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
995 | exit_pi_state_list(tsk); | |
996 | if (unlikely(current->pi_state_cache)) | |
997 | kfree(current->pi_state_cache); | |
42b2dd0a | 998 | #endif |
de5097c2 | 999 | /* |
9a11b49a | 1000 | * Make sure we are holding no locks: |
de5097c2 | 1001 | */ |
9a11b49a | 1002 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1003 | /* |
1004 | * We can do this unlocked here. The futex code uses this flag | |
1005 | * just to verify whether the pi state cleanup has been done | |
1006 | * or not. In the worst case it loops once more. | |
1007 | */ | |
1008 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1009 | |
afc847b7 AV |
1010 | if (tsk->io_context) |
1011 | exit_io_context(); | |
1012 | ||
b92ce558 JA |
1013 | if (tsk->splice_pipe) |
1014 | __free_pipe_info(tsk->splice_pipe); | |
1015 | ||
e0e81739 DH |
1016 | validate_creds_for_do_exit(tsk); |
1017 | ||
7407251a | 1018 | preempt_disable(); |
f41d911f | 1019 | exit_rcu(); |
55a101f8 | 1020 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1021 | tsk->state = TASK_DEAD; |
1da177e4 LT |
1022 | schedule(); |
1023 | BUG(); | |
1024 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1025 | for (;;) |
1026 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1027 | } |
1028 | ||
012914da RA |
1029 | EXPORT_SYMBOL_GPL(do_exit); |
1030 | ||
1da177e4 LT |
1031 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1032 | { | |
1033 | if (comp) | |
1034 | complete(comp); | |
55a101f8 | 1035 | |
1da177e4 LT |
1036 | do_exit(code); |
1037 | } | |
1038 | ||
1039 | EXPORT_SYMBOL(complete_and_exit); | |
1040 | ||
754fe8d2 | 1041 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
1042 | { |
1043 | do_exit((error_code&0xff)<<8); | |
1044 | } | |
1045 | ||
1da177e4 LT |
1046 | /* |
1047 | * Take down every thread in the group. This is called by fatal signals | |
1048 | * as well as by sys_exit_group (below). | |
1049 | */ | |
1050 | NORET_TYPE void | |
1051 | do_group_exit(int exit_code) | |
1052 | { | |
bfc4b089 ON |
1053 | struct signal_struct *sig = current->signal; |
1054 | ||
1da177e4 LT |
1055 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1056 | ||
bfc4b089 ON |
1057 | if (signal_group_exit(sig)) |
1058 | exit_code = sig->group_exit_code; | |
1da177e4 | 1059 | else if (!thread_group_empty(current)) { |
1da177e4 | 1060 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1061 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1062 | if (signal_group_exit(sig)) |
1da177e4 LT |
1063 | /* Another thread got here before we took the lock. */ |
1064 | exit_code = sig->group_exit_code; | |
1065 | else { | |
1da177e4 | 1066 | sig->group_exit_code = exit_code; |
ed5d2cac | 1067 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1068 | zap_other_threads(current); |
1069 | } | |
1070 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1071 | } |
1072 | ||
1073 | do_exit(exit_code); | |
1074 | /* NOTREACHED */ | |
1075 | } | |
1076 | ||
1077 | /* | |
1078 | * this kills every thread in the thread group. Note that any externally | |
1079 | * wait4()-ing process will get the correct exit code - even if this | |
1080 | * thread is not the thread group leader. | |
1081 | */ | |
754fe8d2 | 1082 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
1083 | { |
1084 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
1085 | /* NOTREACHED */ |
1086 | return 0; | |
1da177e4 LT |
1087 | } |
1088 | ||
9e8ae01d ON |
1089 | struct wait_opts { |
1090 | enum pid_type wo_type; | |
9e8ae01d | 1091 | int wo_flags; |
e1eb1ebc | 1092 | struct pid *wo_pid; |
9e8ae01d ON |
1093 | |
1094 | struct siginfo __user *wo_info; | |
1095 | int __user *wo_stat; | |
1096 | struct rusage __user *wo_rusage; | |
1097 | ||
0b7570e7 | 1098 | wait_queue_t child_wait; |
9e8ae01d ON |
1099 | int notask_error; |
1100 | }; | |
1101 | ||
989264f4 ON |
1102 | static inline |
1103 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 1104 | { |
989264f4 ON |
1105 | if (type != PIDTYPE_PID) |
1106 | task = task->group_leader; | |
1107 | return task->pids[type].pid; | |
161550d7 EB |
1108 | } |
1109 | ||
989264f4 | 1110 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1111 | { |
5c01ba49 ON |
1112 | return wo->wo_type == PIDTYPE_MAX || |
1113 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1114 | } | |
1da177e4 | 1115 | |
5c01ba49 ON |
1116 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
1117 | { | |
1118 | if (!eligible_pid(wo, p)) | |
1119 | return 0; | |
1da177e4 LT |
1120 | /* Wait for all children (clone and not) if __WALL is set; |
1121 | * otherwise, wait for clone children *only* if __WCLONE is | |
1122 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1123 | * A "clone" child here is one that reports to its parent | |
1124 | * using a signal other than SIGCHLD.) */ | |
9e8ae01d ON |
1125 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
1126 | && !(wo->wo_flags & __WALL)) | |
1da177e4 | 1127 | return 0; |
1da177e4 | 1128 | |
14dd0b81 | 1129 | return 1; |
1da177e4 LT |
1130 | } |
1131 | ||
9e8ae01d ON |
1132 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
1133 | pid_t pid, uid_t uid, int why, int status) | |
1da177e4 | 1134 | { |
9e8ae01d ON |
1135 | struct siginfo __user *infop; |
1136 | int retval = wo->wo_rusage | |
1137 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
36c8b586 | 1138 | |
1da177e4 | 1139 | put_task_struct(p); |
9e8ae01d | 1140 | infop = wo->wo_info; |
b6fe2d11 VM |
1141 | if (infop) { |
1142 | if (!retval) | |
1143 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1144 | if (!retval) | |
1145 | retval = put_user(0, &infop->si_errno); | |
1146 | if (!retval) | |
1147 | retval = put_user((short)why, &infop->si_code); | |
1148 | if (!retval) | |
1149 | retval = put_user(pid, &infop->si_pid); | |
1150 | if (!retval) | |
1151 | retval = put_user(uid, &infop->si_uid); | |
1152 | if (!retval) | |
1153 | retval = put_user(status, &infop->si_status); | |
1154 | } | |
1da177e4 LT |
1155 | if (!retval) |
1156 | retval = pid; | |
1157 | return retval; | |
1158 | } | |
1159 | ||
1160 | /* | |
1161 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1162 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1163 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1164 | * released the lock and the system call should return. | |
1165 | */ | |
9e8ae01d | 1166 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1167 | { |
1168 | unsigned long state; | |
2f4e6e2a | 1169 | int retval, status, traced; |
6c5f3e7b | 1170 | pid_t pid = task_pid_vnr(p); |
c69e8d9c | 1171 | uid_t uid = __task_cred(p)->uid; |
9e8ae01d | 1172 | struct siginfo __user *infop; |
1da177e4 | 1173 | |
9e8ae01d | 1174 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1175 | return 0; |
1176 | ||
9e8ae01d | 1177 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1da177e4 LT |
1178 | int exit_code = p->exit_code; |
1179 | int why, status; | |
1180 | ||
1da177e4 LT |
1181 | get_task_struct(p); |
1182 | read_unlock(&tasklist_lock); | |
1183 | if ((exit_code & 0x7f) == 0) { | |
1184 | why = CLD_EXITED; | |
1185 | status = exit_code >> 8; | |
1186 | } else { | |
1187 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1188 | status = exit_code & 0x7f; | |
1189 | } | |
9e8ae01d | 1190 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1da177e4 LT |
1191 | } |
1192 | ||
1193 | /* | |
1194 | * Try to move the task's state to DEAD | |
1195 | * only one thread is allowed to do this: | |
1196 | */ | |
1197 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1198 | if (state != EXIT_ZOMBIE) { | |
1199 | BUG_ON(state != EXIT_DEAD); | |
1200 | return 0; | |
1201 | } | |
1da177e4 | 1202 | |
53b6f9fb | 1203 | traced = ptrace_reparented(p); |
befca967 ON |
1204 | /* |
1205 | * It can be ptraced but not reparented, check | |
1206 | * !task_detached() to filter out sub-threads. | |
1207 | */ | |
1208 | if (likely(!traced) && likely(!task_detached(p))) { | |
3795e161 JJ |
1209 | struct signal_struct *psig; |
1210 | struct signal_struct *sig; | |
1f10206c | 1211 | unsigned long maxrss; |
3795e161 | 1212 | |
1da177e4 LT |
1213 | /* |
1214 | * The resource counters for the group leader are in its | |
1215 | * own task_struct. Those for dead threads in the group | |
1216 | * are in its signal_struct, as are those for the child | |
1217 | * processes it has previously reaped. All these | |
1218 | * accumulate in the parent's signal_struct c* fields. | |
1219 | * | |
1220 | * We don't bother to take a lock here to protect these | |
1221 | * p->signal fields, because they are only touched by | |
1222 | * __exit_signal, which runs with tasklist_lock | |
1223 | * write-locked anyway, and so is excluded here. We do | |
d1e98f42 | 1224 | * need to protect the access to parent->signal fields, |
1da177e4 LT |
1225 | * as other threads in the parent group can be right |
1226 | * here reaping other children at the same time. | |
1227 | */ | |
d1e98f42 ON |
1228 | spin_lock_irq(&p->real_parent->sighand->siglock); |
1229 | psig = p->real_parent->signal; | |
3795e161 JJ |
1230 | sig = p->signal; |
1231 | psig->cutime = | |
1232 | cputime_add(psig->cutime, | |
77d1ef79 ON |
1233 | cputime_add(p->utime, |
1234 | cputime_add(sig->utime, | |
1235 | sig->cutime))); | |
3795e161 JJ |
1236 | psig->cstime = |
1237 | cputime_add(psig->cstime, | |
77d1ef79 ON |
1238 | cputime_add(p->stime, |
1239 | cputime_add(sig->stime, | |
1240 | sig->cstime))); | |
9ac52315 LV |
1241 | psig->cgtime = |
1242 | cputime_add(psig->cgtime, | |
1243 | cputime_add(p->gtime, | |
1244 | cputime_add(sig->gtime, | |
1245 | sig->cgtime))); | |
3795e161 JJ |
1246 | psig->cmin_flt += |
1247 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1248 | psig->cmaj_flt += | |
1249 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1250 | psig->cnvcsw += | |
1251 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1252 | psig->cnivcsw += | |
1253 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1254 | psig->cinblock += |
1255 | task_io_get_inblock(p) + | |
1256 | sig->inblock + sig->cinblock; | |
1257 | psig->coublock += | |
1258 | task_io_get_oublock(p) + | |
1259 | sig->oublock + sig->coublock; | |
1f10206c JP |
1260 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1261 | if (psig->cmaxrss < maxrss) | |
1262 | psig->cmaxrss = maxrss; | |
5995477a AR |
1263 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1264 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
d1e98f42 | 1265 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
1da177e4 LT |
1266 | } |
1267 | ||
1268 | /* | |
1269 | * Now we are sure this task is interesting, and no other | |
1270 | * thread can reap it because we set its state to EXIT_DEAD. | |
1271 | */ | |
1272 | read_unlock(&tasklist_lock); | |
1273 | ||
9e8ae01d ON |
1274 | retval = wo->wo_rusage |
1275 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 LT |
1276 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1277 | ? p->signal->group_exit_code : p->exit_code; | |
9e8ae01d ON |
1278 | if (!retval && wo->wo_stat) |
1279 | retval = put_user(status, wo->wo_stat); | |
1280 | ||
1281 | infop = wo->wo_info; | |
1da177e4 LT |
1282 | if (!retval && infop) |
1283 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1284 | if (!retval && infop) | |
1285 | retval = put_user(0, &infop->si_errno); | |
1286 | if (!retval && infop) { | |
1287 | int why; | |
1288 | ||
1289 | if ((status & 0x7f) == 0) { | |
1290 | why = CLD_EXITED; | |
1291 | status >>= 8; | |
1292 | } else { | |
1293 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1294 | status &= 0x7f; | |
1295 | } | |
1296 | retval = put_user((short)why, &infop->si_code); | |
1297 | if (!retval) | |
1298 | retval = put_user(status, &infop->si_status); | |
1299 | } | |
1300 | if (!retval && infop) | |
3a515e4a | 1301 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1302 | if (!retval && infop) |
c69e8d9c | 1303 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1304 | if (!retval) |
3a515e4a | 1305 | retval = pid; |
2f4e6e2a ON |
1306 | |
1307 | if (traced) { | |
1da177e4 | 1308 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1309 | /* We dropped tasklist, ptracer could die and untrace */ |
1310 | ptrace_unlink(p); | |
1311 | /* | |
1312 | * If this is not a detached task, notify the parent. | |
1313 | * If it's still not detached after that, don't release | |
1314 | * it now. | |
1315 | */ | |
d839fd4d | 1316 | if (!task_detached(p)) { |
2f4e6e2a | 1317 | do_notify_parent(p, p->exit_signal); |
d839fd4d | 1318 | if (!task_detached(p)) { |
2f4e6e2a ON |
1319 | p->exit_state = EXIT_ZOMBIE; |
1320 | p = NULL; | |
1da177e4 LT |
1321 | } |
1322 | } | |
1323 | write_unlock_irq(&tasklist_lock); | |
1324 | } | |
1325 | if (p != NULL) | |
1326 | release_task(p); | |
2f4e6e2a | 1327 | |
1da177e4 LT |
1328 | return retval; |
1329 | } | |
1330 | ||
90bc8d8b ON |
1331 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1332 | { | |
1333 | if (ptrace) { | |
1334 | if (task_is_stopped_or_traced(p)) | |
1335 | return &p->exit_code; | |
1336 | } else { | |
1337 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1338 | return &p->signal->group_exit_code; | |
1339 | } | |
1340 | return NULL; | |
1341 | } | |
1342 | ||
1da177e4 LT |
1343 | /* |
1344 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1345 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1346 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1347 | * released the lock and the system call should return. | |
1348 | */ | |
9e8ae01d ON |
1349 | static int wait_task_stopped(struct wait_opts *wo, |
1350 | int ptrace, struct task_struct *p) | |
1da177e4 | 1351 | { |
9e8ae01d | 1352 | struct siginfo __user *infop; |
90bc8d8b | 1353 | int retval, exit_code, *p_code, why; |
ee7c82da | 1354 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1355 | pid_t pid; |
1da177e4 | 1356 | |
47918025 ON |
1357 | /* |
1358 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1359 | */ | |
9e8ae01d | 1360 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1361 | return 0; |
1362 | ||
ee7c82da ON |
1363 | exit_code = 0; |
1364 | spin_lock_irq(&p->sighand->siglock); | |
1365 | ||
90bc8d8b ON |
1366 | p_code = task_stopped_code(p, ptrace); |
1367 | if (unlikely(!p_code)) | |
ee7c82da ON |
1368 | goto unlock_sig; |
1369 | ||
90bc8d8b | 1370 | exit_code = *p_code; |
ee7c82da ON |
1371 | if (!exit_code) |
1372 | goto unlock_sig; | |
1373 | ||
9e8ae01d | 1374 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1375 | *p_code = 0; |
ee7c82da | 1376 | |
c69e8d9c DH |
1377 | /* don't need the RCU readlock here as we're holding a spinlock */ |
1378 | uid = __task_cred(p)->uid; | |
ee7c82da ON |
1379 | unlock_sig: |
1380 | spin_unlock_irq(&p->sighand->siglock); | |
1381 | if (!exit_code) | |
1da177e4 LT |
1382 | return 0; |
1383 | ||
1384 | /* | |
1385 | * Now we are pretty sure this task is interesting. | |
1386 | * Make sure it doesn't get reaped out from under us while we | |
1387 | * give up the lock and then examine it below. We don't want to | |
1388 | * keep holding onto the tasklist_lock while we call getrusage and | |
1389 | * possibly take page faults for user memory. | |
1390 | */ | |
1391 | get_task_struct(p); | |
6c5f3e7b | 1392 | pid = task_pid_vnr(p); |
f470021a | 1393 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1394 | read_unlock(&tasklist_lock); |
1395 | ||
9e8ae01d ON |
1396 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1397 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | |
1398 | ||
1399 | retval = wo->wo_rusage | |
1400 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1401 | if (!retval && wo->wo_stat) | |
1402 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | |
1da177e4 | 1403 | |
9e8ae01d | 1404 | infop = wo->wo_info; |
1da177e4 LT |
1405 | if (!retval && infop) |
1406 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1407 | if (!retval && infop) | |
1408 | retval = put_user(0, &infop->si_errno); | |
1409 | if (!retval && infop) | |
6efcae46 | 1410 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1411 | if (!retval && infop) |
1412 | retval = put_user(exit_code, &infop->si_status); | |
1413 | if (!retval && infop) | |
c8950783 | 1414 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1415 | if (!retval && infop) |
ee7c82da | 1416 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1417 | if (!retval) |
c8950783 | 1418 | retval = pid; |
1da177e4 LT |
1419 | put_task_struct(p); |
1420 | ||
1421 | BUG_ON(!retval); | |
1422 | return retval; | |
1423 | } | |
1424 | ||
1425 | /* | |
1426 | * Handle do_wait work for one task in a live, non-stopped state. | |
1427 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1428 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1429 | * released the lock and the system call should return. | |
1430 | */ | |
9e8ae01d | 1431 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1432 | { |
1433 | int retval; | |
1434 | pid_t pid; | |
1435 | uid_t uid; | |
1436 | ||
9e8ae01d | 1437 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1438 | return 0; |
1439 | ||
1da177e4 LT |
1440 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1441 | return 0; | |
1442 | ||
1443 | spin_lock_irq(&p->sighand->siglock); | |
1444 | /* Re-check with the lock held. */ | |
1445 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1446 | spin_unlock_irq(&p->sighand->siglock); | |
1447 | return 0; | |
1448 | } | |
9e8ae01d | 1449 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1450 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
c69e8d9c | 1451 | uid = __task_cred(p)->uid; |
1da177e4 LT |
1452 | spin_unlock_irq(&p->sighand->siglock); |
1453 | ||
6c5f3e7b | 1454 | pid = task_pid_vnr(p); |
1da177e4 LT |
1455 | get_task_struct(p); |
1456 | read_unlock(&tasklist_lock); | |
1457 | ||
9e8ae01d ON |
1458 | if (!wo->wo_info) { |
1459 | retval = wo->wo_rusage | |
1460 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 | 1461 | put_task_struct(p); |
9e8ae01d ON |
1462 | if (!retval && wo->wo_stat) |
1463 | retval = put_user(0xffff, wo->wo_stat); | |
1da177e4 | 1464 | if (!retval) |
3a515e4a | 1465 | retval = pid; |
1da177e4 | 1466 | } else { |
9e8ae01d ON |
1467 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1468 | CLD_CONTINUED, SIGCONT); | |
1da177e4 LT |
1469 | BUG_ON(retval == 0); |
1470 | } | |
1471 | ||
1472 | return retval; | |
1473 | } | |
1474 | ||
98abed02 RM |
1475 | /* |
1476 | * Consider @p for a wait by @parent. | |
1477 | * | |
9e8ae01d | 1478 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1479 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1480 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1481 | * then ->notask_error is 0 if @p is an eligible child, |
14dd0b81 | 1482 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1483 | */ |
b6e763f0 ON |
1484 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1485 | struct task_struct *p) | |
98abed02 | 1486 | { |
9e8ae01d | 1487 | int ret = eligible_child(wo, p); |
14dd0b81 | 1488 | if (!ret) |
98abed02 RM |
1489 | return ret; |
1490 | ||
a2322e1d | 1491 | ret = security_task_wait(p); |
14dd0b81 RM |
1492 | if (unlikely(ret < 0)) { |
1493 | /* | |
1494 | * If we have not yet seen any eligible child, | |
1495 | * then let this error code replace -ECHILD. | |
1496 | * A permission error will give the user a clue | |
1497 | * to look for security policy problems, rather | |
1498 | * than for mysterious wait bugs. | |
1499 | */ | |
9e8ae01d ON |
1500 | if (wo->notask_error) |
1501 | wo->notask_error = ret; | |
78a3d9d5 | 1502 | return 0; |
14dd0b81 RM |
1503 | } |
1504 | ||
5cb11446 | 1505 | if (likely(!ptrace) && unlikely(task_ptrace(p))) { |
f470021a RM |
1506 | /* |
1507 | * This child is hidden by ptrace. | |
1508 | * We aren't allowed to see it now, but eventually we will. | |
1509 | */ | |
9e8ae01d | 1510 | wo->notask_error = 0; |
f470021a RM |
1511 | return 0; |
1512 | } | |
1513 | ||
98abed02 RM |
1514 | if (p->exit_state == EXIT_DEAD) |
1515 | return 0; | |
1516 | ||
1517 | /* | |
1518 | * We don't reap group leaders with subthreads. | |
1519 | */ | |
1520 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) | |
9e8ae01d | 1521 | return wait_task_zombie(wo, p); |
98abed02 RM |
1522 | |
1523 | /* | |
1524 | * It's stopped or running now, so it might | |
1525 | * later continue, exit, or stop again. | |
1526 | */ | |
9e8ae01d | 1527 | wo->notask_error = 0; |
98abed02 | 1528 | |
90bc8d8b | 1529 | if (task_stopped_code(p, ptrace)) |
9e8ae01d | 1530 | return wait_task_stopped(wo, ptrace, p); |
98abed02 | 1531 | |
9e8ae01d | 1532 | return wait_task_continued(wo, p); |
98abed02 RM |
1533 | } |
1534 | ||
1535 | /* | |
1536 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1537 | * | |
9e8ae01d | 1538 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1539 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1540 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1541 | * ->notask_error is 0 if there were any eligible children, |
14dd0b81 | 1542 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1543 | */ |
9e8ae01d | 1544 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1545 | { |
1546 | struct task_struct *p; | |
1547 | ||
1548 | list_for_each_entry(p, &tsk->children, sibling) { | |
f470021a RM |
1549 | /* |
1550 | * Do not consider detached threads. | |
1551 | */ | |
1552 | if (!task_detached(p)) { | |
b6e763f0 | 1553 | int ret = wait_consider_task(wo, 0, p); |
f470021a RM |
1554 | if (ret) |
1555 | return ret; | |
1556 | } | |
98abed02 RM |
1557 | } |
1558 | ||
1559 | return 0; | |
1560 | } | |
1561 | ||
9e8ae01d | 1562 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1563 | { |
1564 | struct task_struct *p; | |
1565 | ||
f470021a | 1566 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1567 | int ret = wait_consider_task(wo, 1, p); |
f470021a | 1568 | if (ret) |
98abed02 | 1569 | return ret; |
98abed02 RM |
1570 | } |
1571 | ||
1572 | return 0; | |
1573 | } | |
1574 | ||
0b7570e7 ON |
1575 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1576 | int sync, void *key) | |
1577 | { | |
1578 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1579 | child_wait); | |
1580 | struct task_struct *p = key; | |
1581 | ||
5c01ba49 | 1582 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1583 | return 0; |
1584 | ||
b4fe5182 ON |
1585 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1586 | return 0; | |
1587 | ||
0b7570e7 ON |
1588 | return default_wake_function(wait, mode, sync, key); |
1589 | } | |
1590 | ||
a7f0765e ON |
1591 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1592 | { | |
0b7570e7 ON |
1593 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1594 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1595 | } |
1596 | ||
9e8ae01d | 1597 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1598 | { |
1da177e4 | 1599 | struct task_struct *tsk; |
98abed02 | 1600 | int retval; |
1da177e4 | 1601 | |
9e8ae01d | 1602 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1603 | |
0b7570e7 ON |
1604 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1605 | wo->child_wait.private = current; | |
1606 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1607 | repeat: |
98abed02 RM |
1608 | /* |
1609 | * If there is nothing that can match our critiera just get out. | |
9e8ae01d ON |
1610 | * We will clear ->notask_error to zero if we see any child that |
1611 | * might later match our criteria, even if we are not able to reap | |
1612 | * it yet. | |
98abed02 | 1613 | */ |
64a16caf | 1614 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1615 | if ((wo->wo_type < PIDTYPE_MAX) && |
1616 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1617 | goto notask; |
161550d7 | 1618 | |
f95d39d1 | 1619 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1620 | read_lock(&tasklist_lock); |
1621 | tsk = current; | |
1622 | do { | |
64a16caf ON |
1623 | retval = do_wait_thread(wo, tsk); |
1624 | if (retval) | |
1625 | goto end; | |
9e8ae01d | 1626 | |
64a16caf ON |
1627 | retval = ptrace_do_wait(wo, tsk); |
1628 | if (retval) | |
98abed02 | 1629 | goto end; |
98abed02 | 1630 | |
9e8ae01d | 1631 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1632 | break; |
a3f6dfb7 | 1633 | } while_each_thread(current, tsk); |
1da177e4 | 1634 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1635 | |
64a16caf | 1636 | notask: |
9e8ae01d ON |
1637 | retval = wo->notask_error; |
1638 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1639 | retval = -ERESTARTSYS; |
98abed02 RM |
1640 | if (!signal_pending(current)) { |
1641 | schedule(); | |
1642 | goto repeat; | |
1643 | } | |
1da177e4 | 1644 | } |
1da177e4 | 1645 | end: |
f95d39d1 | 1646 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1647 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1648 | return retval; |
1649 | } | |
1650 | ||
17da2bd9 HC |
1651 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1652 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1653 | { |
9e8ae01d | 1654 | struct wait_opts wo; |
161550d7 EB |
1655 | struct pid *pid = NULL; |
1656 | enum pid_type type; | |
1da177e4 LT |
1657 | long ret; |
1658 | ||
1659 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1660 | return -EINVAL; | |
1661 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1662 | return -EINVAL; | |
1663 | ||
1664 | switch (which) { | |
1665 | case P_ALL: | |
161550d7 | 1666 | type = PIDTYPE_MAX; |
1da177e4 LT |
1667 | break; |
1668 | case P_PID: | |
161550d7 EB |
1669 | type = PIDTYPE_PID; |
1670 | if (upid <= 0) | |
1da177e4 LT |
1671 | return -EINVAL; |
1672 | break; | |
1673 | case P_PGID: | |
161550d7 EB |
1674 | type = PIDTYPE_PGID; |
1675 | if (upid <= 0) | |
1da177e4 | 1676 | return -EINVAL; |
1da177e4 LT |
1677 | break; |
1678 | default: | |
1679 | return -EINVAL; | |
1680 | } | |
1681 | ||
161550d7 EB |
1682 | if (type < PIDTYPE_MAX) |
1683 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1684 | |
1685 | wo.wo_type = type; | |
1686 | wo.wo_pid = pid; | |
1687 | wo.wo_flags = options; | |
1688 | wo.wo_info = infop; | |
1689 | wo.wo_stat = NULL; | |
1690 | wo.wo_rusage = ru; | |
1691 | ret = do_wait(&wo); | |
dfe16dfa VM |
1692 | |
1693 | if (ret > 0) { | |
1694 | ret = 0; | |
1695 | } else if (infop) { | |
1696 | /* | |
1697 | * For a WNOHANG return, clear out all the fields | |
1698 | * we would set so the user can easily tell the | |
1699 | * difference. | |
1700 | */ | |
1701 | if (!ret) | |
1702 | ret = put_user(0, &infop->si_signo); | |
1703 | if (!ret) | |
1704 | ret = put_user(0, &infop->si_errno); | |
1705 | if (!ret) | |
1706 | ret = put_user(0, &infop->si_code); | |
1707 | if (!ret) | |
1708 | ret = put_user(0, &infop->si_pid); | |
1709 | if (!ret) | |
1710 | ret = put_user(0, &infop->si_uid); | |
1711 | if (!ret) | |
1712 | ret = put_user(0, &infop->si_status); | |
1713 | } | |
1714 | ||
161550d7 | 1715 | put_pid(pid); |
1da177e4 LT |
1716 | |
1717 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1718 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1719 | return ret; |
1720 | } | |
1721 | ||
754fe8d2 HC |
1722 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1723 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1724 | { |
9e8ae01d | 1725 | struct wait_opts wo; |
161550d7 EB |
1726 | struct pid *pid = NULL; |
1727 | enum pid_type type; | |
1da177e4 LT |
1728 | long ret; |
1729 | ||
1730 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1731 | __WNOTHREAD|__WCLONE|__WALL)) | |
1732 | return -EINVAL; | |
161550d7 EB |
1733 | |
1734 | if (upid == -1) | |
1735 | type = PIDTYPE_MAX; | |
1736 | else if (upid < 0) { | |
1737 | type = PIDTYPE_PGID; | |
1738 | pid = find_get_pid(-upid); | |
1739 | } else if (upid == 0) { | |
1740 | type = PIDTYPE_PGID; | |
2ae448ef | 1741 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1742 | } else /* upid > 0 */ { |
1743 | type = PIDTYPE_PID; | |
1744 | pid = find_get_pid(upid); | |
1745 | } | |
1746 | ||
9e8ae01d ON |
1747 | wo.wo_type = type; |
1748 | wo.wo_pid = pid; | |
1749 | wo.wo_flags = options | WEXITED; | |
1750 | wo.wo_info = NULL; | |
1751 | wo.wo_stat = stat_addr; | |
1752 | wo.wo_rusage = ru; | |
1753 | ret = do_wait(&wo); | |
161550d7 | 1754 | put_pid(pid); |
1da177e4 LT |
1755 | |
1756 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1757 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1758 | return ret; |
1759 | } | |
1760 | ||
1761 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1762 | ||
1763 | /* | |
1764 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1765 | * implemented by calling sys_wait4() from libc.a. | |
1766 | */ | |
17da2bd9 | 1767 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1768 | { |
1769 | return sys_wait4(pid, stat_addr, options, NULL); | |
1770 | } | |
1771 | ||
1772 | #endif |