]>
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); |
909ea964 | 72 | __this_cpu_dec(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 { |
e0a70217 ON |
98 | /* |
99 | * This can only happen if the caller is de_thread(). | |
100 | * FIXME: this is the temporary hack, we should teach | |
101 | * posix-cpu-timers to handle this case correctly. | |
102 | */ | |
103 | if (unlikely(has_group_leader_pid(tsk))) | |
104 | posix_cpu_timers_exit_group(tsk); | |
105 | ||
6a14c5c9 ON |
106 | /* |
107 | * If there is any task waiting for the group exit | |
108 | * then notify it: | |
109 | */ | |
d344193a | 110 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 111 | wake_up_process(sig->group_exit_task); |
6db840fa | 112 | |
6a14c5c9 ON |
113 | if (tsk == sig->curr_target) |
114 | sig->curr_target = next_thread(tsk); | |
115 | /* | |
116 | * Accumulate here the counters for all threads but the | |
117 | * group leader as they die, so they can be added into | |
118 | * the process-wide totals when those are taken. | |
119 | * The group leader stays around as a zombie as long | |
120 | * as there are other threads. When it gets reaped, | |
121 | * the exit.c code will add its counts into these totals. | |
122 | * We won't ever get here for the group leader, since it | |
123 | * will have been the last reference on the signal_struct. | |
124 | */ | |
0cf55e1e HS |
125 | sig->utime = cputime_add(sig->utime, tsk->utime); |
126 | sig->stime = cputime_add(sig->stime, tsk->stime); | |
d5b7c78e | 127 | sig->gtime = cputime_add(sig->gtime, tsk->gtime); |
6a14c5c9 ON |
128 | sig->min_flt += tsk->min_flt; |
129 | sig->maj_flt += tsk->maj_flt; | |
130 | sig->nvcsw += tsk->nvcsw; | |
131 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
132 | sig->inblock += task_io_get_inblock(tsk); |
133 | sig->oublock += task_io_get_oublock(tsk); | |
5995477a | 134 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
32bd671d | 135 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
136 | } |
137 | ||
b3ac022c | 138 | sig->nr_threads--; |
d40e48e0 | 139 | __unhash_process(tsk, group_dead); |
5876700c | 140 | |
da7978b0 ON |
141 | /* |
142 | * Do this under ->siglock, we can race with another thread | |
143 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
144 | */ | |
145 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 146 | tsk->sighand = NULL; |
6a14c5c9 | 147 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 148 | |
a7e5328a | 149 | __cleanup_sighand(sighand); |
6a14c5c9 | 150 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
d40e48e0 | 151 | if (group_dead) { |
6a14c5c9 | 152 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 153 | tty_kref_put(tty); |
6a14c5c9 ON |
154 | } |
155 | } | |
156 | ||
8c7904a0 EB |
157 | static void delayed_put_task_struct(struct rcu_head *rhp) |
158 | { | |
0a16b607 MD |
159 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
160 | ||
4e231c79 | 161 | perf_event_delayed_put(tsk); |
0a16b607 MD |
162 | trace_sched_process_free(tsk); |
163 | put_task_struct(tsk); | |
8c7904a0 EB |
164 | } |
165 | ||
f470021a | 166 | |
1da177e4 LT |
167 | void release_task(struct task_struct * p) |
168 | { | |
36c8b586 | 169 | struct task_struct *leader; |
1da177e4 | 170 | int zap_leader; |
1f09f974 | 171 | repeat: |
c69e8d9c | 172 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
173 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
174 | rcu_read_lock(); | |
c69e8d9c | 175 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 176 | rcu_read_unlock(); |
c69e8d9c | 177 | |
60347f67 | 178 | proc_flush_task(p); |
0203026b | 179 | |
1da177e4 | 180 | write_lock_irq(&tasklist_lock); |
a288eecc | 181 | ptrace_release_task(p); |
1da177e4 | 182 | __exit_signal(p); |
35f5cad8 | 183 | |
1da177e4 LT |
184 | /* |
185 | * If we are the last non-leader member of the thread | |
186 | * group, and the leader is zombie, then notify the | |
187 | * group leader's parent process. (if it wants notification.) | |
188 | */ | |
189 | zap_leader = 0; | |
190 | leader = p->group_leader; | |
191 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
d839fd4d | 192 | BUG_ON(task_detached(leader)); |
1da177e4 LT |
193 | do_notify_parent(leader, leader->exit_signal); |
194 | /* | |
195 | * If we were the last child thread and the leader has | |
196 | * exited already, and the leader's parent ignores SIGCHLD, | |
197 | * then we are the one who should release the leader. | |
198 | * | |
199 | * do_notify_parent() will have marked it self-reaping in | |
200 | * that case. | |
201 | */ | |
d839fd4d | 202 | zap_leader = task_detached(leader); |
dae33574 RM |
203 | |
204 | /* | |
205 | * This maintains the invariant that release_task() | |
206 | * only runs on a task in EXIT_DEAD, just for sanity. | |
207 | */ | |
208 | if (zap_leader) | |
209 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
210 | } |
211 | ||
1da177e4 | 212 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 213 | release_thread(p); |
8c7904a0 | 214 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
215 | |
216 | p = leader; | |
217 | if (unlikely(zap_leader)) | |
218 | goto repeat; | |
219 | } | |
220 | ||
1da177e4 LT |
221 | /* |
222 | * This checks not only the pgrp, but falls back on the pid if no | |
223 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
224 | * without this... | |
04a2e6a5 EB |
225 | * |
226 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 227 | */ |
04a2e6a5 | 228 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
229 | { |
230 | struct task_struct *p; | |
04a2e6a5 | 231 | struct pid *sid = NULL; |
62dfb554 | 232 | |
04a2e6a5 | 233 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 234 | if (p == NULL) |
04a2e6a5 | 235 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 236 | if (p != NULL) |
04a2e6a5 | 237 | sid = task_session(p); |
62dfb554 | 238 | |
1da177e4 LT |
239 | return sid; |
240 | } | |
241 | ||
242 | /* | |
243 | * Determine if a process group is "orphaned", according to the POSIX | |
244 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
245 | * by terminal-generated stop signals. Newly orphaned process groups are | |
246 | * to receive a SIGHUP and a SIGCONT. | |
247 | * | |
248 | * "I ask you, have you ever known what it is to be an orphan?" | |
249 | */ | |
0475ac08 | 250 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
251 | { |
252 | struct task_struct *p; | |
1da177e4 | 253 | |
0475ac08 | 254 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
255 | if ((p == ignored_task) || |
256 | (p->exit_state && thread_group_empty(p)) || | |
257 | is_global_init(p->real_parent)) | |
1da177e4 | 258 | continue; |
05e83df6 | 259 | |
0475ac08 | 260 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
261 | task_session(p->real_parent) == task_session(p)) |
262 | return 0; | |
0475ac08 | 263 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
264 | |
265 | return 1; | |
1da177e4 LT |
266 | } |
267 | ||
3e7cd6c4 | 268 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
269 | { |
270 | int retval; | |
271 | ||
272 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 273 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
274 | read_unlock(&tasklist_lock); |
275 | ||
276 | return retval; | |
277 | } | |
278 | ||
0475ac08 | 279 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
280 | { |
281 | int retval = 0; | |
282 | struct task_struct *p; | |
283 | ||
0475ac08 | 284 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
338077e5 | 285 | if (!task_is_stopped(p)) |
1da177e4 | 286 | continue; |
1da177e4 LT |
287 | retval = 1; |
288 | break; | |
0475ac08 | 289 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
290 | return retval; |
291 | } | |
292 | ||
f49ee505 ON |
293 | /* |
294 | * Check to see if any process groups have become orphaned as | |
295 | * a result of our exiting, and if they have any stopped jobs, | |
296 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
297 | */ | |
298 | static void | |
299 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
300 | { | |
301 | struct pid *pgrp = task_pgrp(tsk); | |
302 | struct task_struct *ignored_task = tsk; | |
303 | ||
304 | if (!parent) | |
305 | /* exit: our father is in a different pgrp than | |
306 | * we are and we were the only connection outside. | |
307 | */ | |
308 | parent = tsk->real_parent; | |
309 | else | |
310 | /* reparent: our child is in a different pgrp than | |
311 | * we are, and it was the only connection outside. | |
312 | */ | |
313 | ignored_task = NULL; | |
314 | ||
315 | if (task_pgrp(parent) != pgrp && | |
316 | task_session(parent) == task_session(tsk) && | |
317 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
318 | has_stopped_jobs(pgrp)) { | |
319 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
320 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
321 | } | |
322 | } | |
323 | ||
1da177e4 | 324 | /** |
49d769d5 | 325 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
326 | * |
327 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
328 | * it ever exits, it should generally reparent itself to kthreadd so it |
329 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
330 | * |
331 | * The various task state such as scheduling policy and priority may have | |
332 | * been inherited from a user process, so we reset them to sane values here. | |
333 | * | |
49d769d5 | 334 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 335 | */ |
49d769d5 | 336 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
337 | { |
338 | write_lock_irq(&tasklist_lock); | |
339 | ||
340 | ptrace_unlink(current); | |
341 | /* Reparent to init */ | |
49d769d5 | 342 | current->real_parent = current->parent = kthreadd_task; |
f470021a | 343 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
1da177e4 LT |
344 | |
345 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
346 | current->exit_signal = SIGCHLD; | |
347 | ||
e05606d3 | 348 | if (task_nice(current) < 0) |
1da177e4 LT |
349 | set_user_nice(current, 0); |
350 | /* cpus_allowed? */ | |
351 | /* rt_priority? */ | |
352 | /* signals? */ | |
1da177e4 LT |
353 | memcpy(current->signal->rlim, init_task.signal->rlim, |
354 | sizeof(current->signal->rlim)); | |
d84f4f99 DH |
355 | |
356 | atomic_inc(&init_cred.usage); | |
357 | commit_creds(&init_cred); | |
1da177e4 | 358 | write_unlock_irq(&tasklist_lock); |
1da177e4 LT |
359 | } |
360 | ||
8520d7c7 | 361 | void __set_special_pids(struct pid *pid) |
1da177e4 | 362 | { |
e19f247a | 363 | struct task_struct *curr = current->group_leader; |
1da177e4 | 364 | |
0d0df599 | 365 | if (task_session(curr) != pid) |
7d8da096 | 366 | change_pid(curr, PIDTYPE_SID, pid); |
1b0f7ffd ON |
367 | |
368 | if (task_pgrp(curr) != pid) | |
7d8da096 | 369 | change_pid(curr, PIDTYPE_PGID, pid); |
1da177e4 LT |
370 | } |
371 | ||
8520d7c7 | 372 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
373 | { |
374 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 375 | __set_special_pids(pid); |
1da177e4 LT |
376 | write_unlock_irq(&tasklist_lock); |
377 | } | |
378 | ||
379 | /* | |
87245135 ON |
380 | * Let kernel threads use this to say that they allow a certain signal. |
381 | * Must not be used if kthread was cloned with CLONE_SIGHAND. | |
1da177e4 LT |
382 | */ |
383 | int allow_signal(int sig) | |
384 | { | |
7ed20e1a | 385 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
386 | return -EINVAL; |
387 | ||
388 | spin_lock_irq(¤t->sighand->siglock); | |
87245135 | 389 | /* This is only needed for daemonize()'ed kthreads */ |
1da177e4 | 390 | sigdelset(¤t->blocked, sig); |
87245135 ON |
391 | /* |
392 | * Kernel threads handle their own signals. Let the signal code | |
393 | * know it'll be handled, so that they don't get converted to | |
394 | * SIGKILL or just silently dropped. | |
395 | */ | |
396 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
1da177e4 LT |
397 | recalc_sigpending(); |
398 | spin_unlock_irq(¤t->sighand->siglock); | |
399 | return 0; | |
400 | } | |
401 | ||
402 | EXPORT_SYMBOL(allow_signal); | |
403 | ||
404 | int disallow_signal(int sig) | |
405 | { | |
7ed20e1a | 406 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
407 | return -EINVAL; |
408 | ||
409 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 410 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
411 | recalc_sigpending(); |
412 | spin_unlock_irq(¤t->sighand->siglock); | |
413 | return 0; | |
414 | } | |
415 | ||
416 | EXPORT_SYMBOL(disallow_signal); | |
417 | ||
418 | /* | |
419 | * Put all the gunge required to become a kernel thread without | |
420 | * attached user resources in one place where it belongs. | |
421 | */ | |
422 | ||
423 | void daemonize(const char *name, ...) | |
424 | { | |
425 | va_list args; | |
1da177e4 LT |
426 | sigset_t blocked; |
427 | ||
428 | va_start(args, name); | |
429 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
430 | va_end(args); | |
431 | ||
432 | /* | |
433 | * If we were started as result of loading a module, close all of the | |
434 | * user space pages. We don't need them, and if we didn't close them | |
435 | * they would be locked into memory. | |
436 | */ | |
437 | exit_mm(current); | |
83144186 RW |
438 | /* |
439 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
440 | * or suspend transition begins right now. | |
441 | */ | |
7b34e428 | 442 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
1da177e4 | 443 | |
8520d7c7 ON |
444 | if (current->nsproxy != &init_nsproxy) { |
445 | get_nsproxy(&init_nsproxy); | |
446 | switch_task_namespaces(current, &init_nsproxy); | |
447 | } | |
297bd42b | 448 | set_special_pids(&init_struct_pid); |
24ec839c | 449 | proc_clear_tty(current); |
1da177e4 LT |
450 | |
451 | /* Block and flush all signals */ | |
452 | sigfillset(&blocked); | |
453 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
454 | flush_signals(current); | |
455 | ||
456 | /* Become as one with the init task */ | |
457 | ||
3e93cd67 | 458 | daemonize_fs_struct(); |
d4c5e41f | 459 | exit_files(current); |
1da177e4 LT |
460 | current->files = init_task.files; |
461 | atomic_inc(¤t->files->count); | |
462 | ||
49d769d5 | 463 | reparent_to_kthreadd(); |
1da177e4 LT |
464 | } |
465 | ||
466 | EXPORT_SYMBOL(daemonize); | |
467 | ||
858119e1 | 468 | static void close_files(struct files_struct * files) |
1da177e4 LT |
469 | { |
470 | int i, j; | |
badf1662 | 471 | struct fdtable *fdt; |
1da177e4 LT |
472 | |
473 | j = 0; | |
4fb3a538 DS |
474 | |
475 | /* | |
476 | * It is safe to dereference the fd table without RCU or | |
477 | * ->file_lock because this is the last reference to the | |
d11c563d | 478 | * files structure. But use RCU to shut RCU-lockdep up. |
4fb3a538 | 479 | */ |
d11c563d | 480 | rcu_read_lock(); |
badf1662 | 481 | fdt = files_fdtable(files); |
d11c563d | 482 | rcu_read_unlock(); |
1da177e4 LT |
483 | for (;;) { |
484 | unsigned long set; | |
485 | i = j * __NFDBITS; | |
bbea9f69 | 486 | if (i >= fdt->max_fds) |
1da177e4 | 487 | break; |
badf1662 | 488 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
489 | while (set) { |
490 | if (set & 1) { | |
badf1662 | 491 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 492 | if (file) { |
1da177e4 | 493 | filp_close(file, files); |
944be0b2 IM |
494 | cond_resched(); |
495 | } | |
1da177e4 LT |
496 | } |
497 | i++; | |
498 | set >>= 1; | |
499 | } | |
500 | } | |
501 | } | |
502 | ||
503 | struct files_struct *get_files_struct(struct task_struct *task) | |
504 | { | |
505 | struct files_struct *files; | |
506 | ||
507 | task_lock(task); | |
508 | files = task->files; | |
509 | if (files) | |
510 | atomic_inc(&files->count); | |
511 | task_unlock(task); | |
512 | ||
513 | return files; | |
514 | } | |
515 | ||
7ad5b3a5 | 516 | void put_files_struct(struct files_struct *files) |
1da177e4 | 517 | { |
badf1662 DS |
518 | struct fdtable *fdt; |
519 | ||
1da177e4 LT |
520 | if (atomic_dec_and_test(&files->count)) { |
521 | close_files(files); | |
522 | /* | |
523 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
524 | * If the fdtable was embedded, pass files for freeing |
525 | * at the end of the RCU grace period. Otherwise, | |
526 | * you can free files immediately. | |
1da177e4 | 527 | */ |
d11c563d | 528 | rcu_read_lock(); |
badf1662 | 529 | fdt = files_fdtable(files); |
4fd45812 | 530 | if (fdt != &files->fdtab) |
ab2af1f5 | 531 | kmem_cache_free(files_cachep, files); |
01b2d93c | 532 | free_fdtable(fdt); |
d11c563d | 533 | rcu_read_unlock(); |
1da177e4 LT |
534 | } |
535 | } | |
536 | ||
3b125388 | 537 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 538 | { |
3b125388 | 539 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
540 | struct files_struct *old; |
541 | ||
542 | old = tsk->files; | |
543 | task_lock(tsk); | |
544 | tsk->files = files; | |
545 | task_unlock(tsk); | |
546 | put_files_struct(old); | |
547 | } | |
3b9b8ab6 | 548 | |
1ec7f1dd | 549 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
550 | { |
551 | struct files_struct * files = tsk->files; | |
552 | ||
553 | if (files) { | |
554 | task_lock(tsk); | |
555 | tsk->files = NULL; | |
556 | task_unlock(tsk); | |
557 | put_files_struct(files); | |
558 | } | |
559 | } | |
560 | ||
cf475ad2 BS |
561 | #ifdef CONFIG_MM_OWNER |
562 | /* | |
563 | * Task p is exiting and it owned mm, lets find a new owner for it | |
564 | */ | |
565 | static inline int | |
566 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) | |
567 | { | |
568 | /* | |
569 | * If there are other users of the mm and the owner (us) is exiting | |
570 | * we need to find a new owner to take on the responsibility. | |
571 | */ | |
cf475ad2 BS |
572 | if (atomic_read(&mm->mm_users) <= 1) |
573 | return 0; | |
574 | if (mm->owner != p) | |
575 | return 0; | |
576 | return 1; | |
577 | } | |
578 | ||
579 | void mm_update_next_owner(struct mm_struct *mm) | |
580 | { | |
581 | struct task_struct *c, *g, *p = current; | |
582 | ||
583 | retry: | |
584 | if (!mm_need_new_owner(mm, p)) | |
585 | return; | |
586 | ||
587 | read_lock(&tasklist_lock); | |
588 | /* | |
589 | * Search in the children | |
590 | */ | |
591 | list_for_each_entry(c, &p->children, sibling) { | |
592 | if (c->mm == mm) | |
593 | goto assign_new_owner; | |
594 | } | |
595 | ||
596 | /* | |
597 | * Search in the siblings | |
598 | */ | |
dea33cfd | 599 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
600 | if (c->mm == mm) |
601 | goto assign_new_owner; | |
602 | } | |
603 | ||
604 | /* | |
605 | * Search through everything else. We should not get | |
606 | * here often | |
607 | */ | |
608 | do_each_thread(g, c) { | |
609 | if (c->mm == mm) | |
610 | goto assign_new_owner; | |
611 | } while_each_thread(g, c); | |
612 | ||
613 | read_unlock(&tasklist_lock); | |
31a78f23 BS |
614 | /* |
615 | * We found no owner yet mm_users > 1: this implies that we are | |
616 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 617 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 618 | */ |
31a78f23 | 619 | mm->owner = NULL; |
cf475ad2 BS |
620 | return; |
621 | ||
622 | assign_new_owner: | |
623 | BUG_ON(c == p); | |
624 | get_task_struct(c); | |
625 | /* | |
626 | * The task_lock protects c->mm from changing. | |
627 | * We always want mm->owner->mm == mm | |
628 | */ | |
629 | task_lock(c); | |
e5991371 HD |
630 | /* |
631 | * Delay read_unlock() till we have the task_lock() | |
632 | * to ensure that c does not slip away underneath us | |
633 | */ | |
634 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
635 | if (c->mm != mm) { |
636 | task_unlock(c); | |
637 | put_task_struct(c); | |
638 | goto retry; | |
639 | } | |
cf475ad2 BS |
640 | mm->owner = c; |
641 | task_unlock(c); | |
642 | put_task_struct(c); | |
643 | } | |
644 | #endif /* CONFIG_MM_OWNER */ | |
645 | ||
1da177e4 LT |
646 | /* |
647 | * Turn us into a lazy TLB process if we | |
648 | * aren't already.. | |
649 | */ | |
408b664a | 650 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
651 | { |
652 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 653 | struct core_state *core_state; |
1da177e4 LT |
654 | |
655 | mm_release(tsk, mm); | |
656 | if (!mm) | |
657 | return; | |
658 | /* | |
659 | * Serialize with any possible pending coredump. | |
999d9fc1 | 660 | * We must hold mmap_sem around checking core_state |
1da177e4 | 661 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 662 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
663 | * group with ->mm != NULL. |
664 | */ | |
665 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
666 | core_state = mm->core_state; |
667 | if (core_state) { | |
668 | struct core_thread self; | |
1da177e4 | 669 | up_read(&mm->mmap_sem); |
1da177e4 | 670 | |
b564daf8 ON |
671 | self.task = tsk; |
672 | self.next = xchg(&core_state->dumper.next, &self); | |
673 | /* | |
674 | * Implies mb(), the result of xchg() must be visible | |
675 | * to core_state->dumper. | |
676 | */ | |
677 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
678 | complete(&core_state->startup); | |
1da177e4 | 679 | |
a94e2d40 ON |
680 | for (;;) { |
681 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
682 | if (!self.task) /* see coredump_finish() */ | |
683 | break; | |
684 | schedule(); | |
685 | } | |
686 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
687 | down_read(&mm->mmap_sem); |
688 | } | |
689 | atomic_inc(&mm->mm_count); | |
125e1874 | 690 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
691 | /* more a memory barrier than a real lock */ |
692 | task_lock(tsk); | |
693 | tsk->mm = NULL; | |
694 | up_read(&mm->mmap_sem); | |
695 | enter_lazy_tlb(mm, current); | |
0c1eecfb RW |
696 | /* We don't want this task to be frozen prematurely */ |
697 | clear_freeze_flag(tsk); | |
3d5992d2 YH |
698 | if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) |
699 | atomic_dec(&mm->oom_disable_count); | |
1da177e4 | 700 | task_unlock(tsk); |
cf475ad2 | 701 | mm_update_next_owner(mm); |
1da177e4 LT |
702 | mmput(mm); |
703 | } | |
704 | ||
1da177e4 LT |
705 | /* |
706 | * When we die, we re-parent all our children. | |
707 | * Try to give them to another thread in our thread | |
708 | * group, and if no such member exists, give it to | |
84d73786 SB |
709 | * the child reaper process (ie "init") in our pid |
710 | * space. | |
1da177e4 | 711 | */ |
950bbabb | 712 | static struct task_struct *find_new_reaper(struct task_struct *father) |
d16e15f5 NK |
713 | __releases(&tasklist_lock) |
714 | __acquires(&tasklist_lock) | |
1da177e4 | 715 | { |
950bbabb ON |
716 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
717 | struct task_struct *thread; | |
1da177e4 | 718 | |
950bbabb ON |
719 | thread = father; |
720 | while_each_thread(father, thread) { | |
721 | if (thread->flags & PF_EXITING) | |
722 | continue; | |
723 | if (unlikely(pid_ns->child_reaper == father)) | |
724 | pid_ns->child_reaper = thread; | |
725 | return thread; | |
726 | } | |
1da177e4 | 727 | |
950bbabb ON |
728 | if (unlikely(pid_ns->child_reaper == father)) { |
729 | write_unlock_irq(&tasklist_lock); | |
730 | if (unlikely(pid_ns == &init_pid_ns)) | |
731 | panic("Attempted to kill init!"); | |
1da177e4 | 732 | |
950bbabb ON |
733 | zap_pid_ns_processes(pid_ns); |
734 | write_lock_irq(&tasklist_lock); | |
1da177e4 | 735 | /* |
950bbabb ON |
736 | * We can not clear ->child_reaper or leave it alone. |
737 | * There may by stealth EXIT_DEAD tasks on ->children, | |
738 | * forget_original_parent() must move them somewhere. | |
1da177e4 | 739 | */ |
950bbabb | 740 | pid_ns->child_reaper = init_pid_ns.child_reaper; |
1da177e4 | 741 | } |
762a24be | 742 | |
950bbabb ON |
743 | return pid_ns->child_reaper; |
744 | } | |
745 | ||
5dfc80be ON |
746 | /* |
747 | * Any that need to be release_task'd are put on the @dead list. | |
748 | */ | |
9cd80bbb | 749 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
750 | struct list_head *dead) |
751 | { | |
5dfc80be ON |
752 | list_move_tail(&p->sibling, &p->real_parent->children); |
753 | ||
754 | if (task_detached(p)) | |
755 | return; | |
756 | /* | |
757 | * If this is a threaded reparent there is no need to | |
758 | * notify anyone anything has happened. | |
759 | */ | |
760 | if (same_thread_group(p->real_parent, father)) | |
761 | return; | |
762 | ||
763 | /* We don't want people slaying init. */ | |
764 | p->exit_signal = SIGCHLD; | |
765 | ||
766 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 767 | if (!p->ptrace && |
5dfc80be ON |
768 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
769 | do_notify_parent(p, p->exit_signal); | |
770 | if (task_detached(p)) { | |
771 | p->exit_state = EXIT_DEAD; | |
772 | list_move_tail(&p->sibling, dead); | |
773 | } | |
774 | } | |
775 | ||
776 | kill_orphaned_pgrp(p, father); | |
777 | } | |
778 | ||
762a24be | 779 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 780 | { |
950bbabb | 781 | struct task_struct *p, *n, *reaper; |
5dfc80be | 782 | LIST_HEAD(dead_children); |
762a24be ON |
783 | |
784 | write_lock_irq(&tasklist_lock); | |
c7e49c14 ON |
785 | /* |
786 | * Note that exit_ptrace() and find_new_reaper() might | |
787 | * drop tasklist_lock and reacquire it. | |
788 | */ | |
789 | exit_ptrace(father); | |
950bbabb | 790 | reaper = find_new_reaper(father); |
f470021a | 791 | |
03ff1797 | 792 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
9cd80bbb ON |
793 | struct task_struct *t = p; |
794 | do { | |
795 | t->real_parent = reaper; | |
796 | if (t->parent == father) { | |
d21142ec | 797 | BUG_ON(t->ptrace); |
9cd80bbb ON |
798 | t->parent = t->real_parent; |
799 | } | |
800 | if (t->pdeath_signal) | |
801 | group_send_sig_info(t->pdeath_signal, | |
802 | SEND_SIG_NOINFO, t); | |
803 | } while_each_thread(p, t); | |
804 | reparent_leader(father, p, &dead_children); | |
1da177e4 | 805 | } |
762a24be | 806 | write_unlock_irq(&tasklist_lock); |
5dfc80be | 807 | |
762a24be | 808 | BUG_ON(!list_empty(&father->children)); |
762a24be | 809 | |
5dfc80be ON |
810 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
811 | list_del_init(&p->sibling); | |
39c626ae ON |
812 | release_task(p); |
813 | } | |
1da177e4 LT |
814 | } |
815 | ||
816 | /* | |
817 | * Send signals to all our closest relatives so that they know | |
818 | * to properly mourn us.. | |
819 | */ | |
821c7de7 | 820 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 821 | { |
53c8f9f1 | 822 | bool autoreap; |
1da177e4 | 823 | |
1da177e4 LT |
824 | /* |
825 | * This does two things: | |
826 | * | |
827 | * A. Make init inherit all the child processes | |
828 | * B. Check to see if any process groups have become orphaned | |
829 | * as a result of our exiting, and if they have any stopped | |
830 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
831 | */ | |
762a24be | 832 | forget_original_parent(tsk); |
2e4a7072 | 833 | exit_task_namespaces(tsk); |
1da177e4 | 834 | |
762a24be | 835 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
836 | if (group_dead) |
837 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 838 | |
24728448 | 839 | /* Let father know we died |
1da177e4 LT |
840 | * |
841 | * Thread signals are configurable, but you aren't going to use | |
25985edc | 842 | * that to send signals to arbitrary processes. |
1da177e4 LT |
843 | * That stops right now. |
844 | * | |
845 | * If the parent exec id doesn't match the exec id we saved | |
846 | * when we started then we know the parent has changed security | |
847 | * domain. | |
848 | * | |
849 | * If our self_exec id doesn't match our parent_exec_id then | |
850 | * we have changed execution domain as these two values started | |
851 | * the same after a fork. | |
1da177e4 | 852 | */ |
45cdf5cc | 853 | if (thread_group_leader(tsk) && tsk->exit_signal != SIGCHLD && |
f49ee505 | 854 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
432870da | 855 | tsk->self_exec_id != tsk->parent_exec_id)) |
1da177e4 LT |
856 | tsk->exit_signal = SIGCHLD; |
857 | ||
45cdf5cc ON |
858 | if (unlikely(tsk->ptrace)) { |
859 | int sig = thread_group_leader(tsk) && | |
860 | thread_group_empty(tsk) && | |
861 | !ptrace_reparented(tsk) ? | |
862 | tsk->exit_signal : SIGCHLD; | |
863 | autoreap = do_notify_parent(tsk, sig); | |
864 | } else if (thread_group_leader(tsk)) { | |
865 | autoreap = thread_group_empty(tsk) && | |
866 | do_notify_parent(tsk, tsk->exit_signal); | |
867 | } else { | |
868 | autoreap = true; | |
869 | } | |
1da177e4 | 870 | |
53c8f9f1 | 871 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
1da177e4 | 872 | |
9c339168 ON |
873 | /* mt-exec, de_thread() is waiting for group leader */ |
874 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 875 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
876 | write_unlock_irq(&tasklist_lock); |
877 | ||
1da177e4 | 878 | /* If the process is dead, release it - nobody will wait for it */ |
53c8f9f1 | 879 | if (autoreap) |
1da177e4 | 880 | release_task(tsk); |
1da177e4 LT |
881 | } |
882 | ||
e18eecb8 JD |
883 | #ifdef CONFIG_DEBUG_STACK_USAGE |
884 | static void check_stack_usage(void) | |
885 | { | |
886 | static DEFINE_SPINLOCK(low_water_lock); | |
887 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
888 | unsigned long free; |
889 | ||
7c9f8861 | 890 | free = stack_not_used(current); |
e18eecb8 JD |
891 | |
892 | if (free >= lowest_to_date) | |
893 | return; | |
894 | ||
895 | spin_lock(&low_water_lock); | |
896 | if (free < lowest_to_date) { | |
897 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
898 | "left\n", | |
899 | current->comm, free); | |
900 | lowest_to_date = free; | |
901 | } | |
902 | spin_unlock(&low_water_lock); | |
903 | } | |
904 | #else | |
905 | static inline void check_stack_usage(void) {} | |
906 | #endif | |
907 | ||
7ad5b3a5 | 908 | NORET_TYPE void do_exit(long code) |
1da177e4 LT |
909 | { |
910 | struct task_struct *tsk = current; | |
911 | int group_dead; | |
912 | ||
913 | profile_task_exit(tsk); | |
914 | ||
22e2c507 | 915 | WARN_ON(atomic_read(&tsk->fs_excl)); |
73c10101 | 916 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 917 | |
1da177e4 LT |
918 | if (unlikely(in_interrupt())) |
919 | panic("Aiee, killing interrupt handler!"); | |
920 | if (unlikely(!tsk->pid)) | |
921 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 922 | |
33dd94ae NE |
923 | /* |
924 | * If do_exit is called because this processes oopsed, it's possible | |
925 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
926 | * continuing. Amongst other possible reasons, this is to prevent | |
927 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
928 | * kernel address. | |
929 | */ | |
930 | set_fs(USER_DS); | |
931 | ||
a288eecc | 932 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 933 | |
e0e81739 DH |
934 | validate_creds_for_do_exit(tsk); |
935 | ||
df164db5 AN |
936 | /* |
937 | * We're taking recursive faults here in do_exit. Safest is to just | |
938 | * leave this task alone and wait for reboot. | |
939 | */ | |
940 | if (unlikely(tsk->flags & PF_EXITING)) { | |
941 | printk(KERN_ALERT | |
942 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
943 | /* |
944 | * We can do this unlocked here. The futex code uses | |
945 | * this flag just to verify whether the pi state | |
946 | * cleanup has been done or not. In the worst case it | |
947 | * loops once more. We pretend that the cleanup was | |
948 | * done as there is no way to return. Either the | |
949 | * OWNER_DIED bit is set by now or we push the blocked | |
950 | * task into the wait for ever nirwana as well. | |
951 | */ | |
952 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
953 | set_current_state(TASK_UNINTERRUPTIBLE); |
954 | schedule(); | |
955 | } | |
956 | ||
3aa551c9 TG |
957 | exit_irq_thread(); |
958 | ||
d12619b5 | 959 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
960 | /* |
961 | * tsk->flags are checked in the futex code to protect against | |
962 | * an exiting task cleaning up the robust pi futexes. | |
963 | */ | |
d2ee7198 | 964 | smp_mb(); |
1d615482 | 965 | raw_spin_unlock_wait(&tsk->pi_lock); |
1da177e4 | 966 | |
1da177e4 LT |
967 | if (unlikely(in_atomic())) |
968 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 969 | current->comm, task_pid_nr(current), |
1da177e4 LT |
970 | preempt_count()); |
971 | ||
972 | acct_update_integrals(tsk); | |
34e55232 | 973 | /* sync mm's RSS info before statistics gathering */ |
a3a2e76c KH |
974 | if (tsk->mm) |
975 | sync_mm_rss(tsk, tsk->mm); | |
1da177e4 | 976 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 977 | if (group_dead) { |
778e9a9c | 978 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 979 | exit_itimers(tsk->signal); |
1f10206c JP |
980 | if (tsk->mm) |
981 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 982 | } |
f6ec29a4 | 983 | acct_collect(code, group_dead); |
522ed776 MT |
984 | if (group_dead) |
985 | tty_audit_exit(); | |
fa84cb93 AV |
986 | if (unlikely(tsk->audit_context)) |
987 | audit_free(tsk); | |
115085ea | 988 | |
f2ab6d88 | 989 | tsk->exit_code = code; |
115085ea | 990 | taskstats_exit(tsk, group_dead); |
c757249a | 991 | |
1da177e4 LT |
992 | exit_mm(tsk); |
993 | ||
0e464814 | 994 | if (group_dead) |
f6ec29a4 | 995 | acct_process(); |
0a16b607 MD |
996 | trace_sched_process_exit(tsk); |
997 | ||
1da177e4 | 998 | exit_sem(tsk); |
1ec7f1dd AV |
999 | exit_files(tsk); |
1000 | exit_fs(tsk); | |
e18eecb8 | 1001 | check_stack_usage(); |
1da177e4 | 1002 | exit_thread(); |
0b3fcf17 SE |
1003 | |
1004 | /* | |
1005 | * Flush inherited counters to the parent - before the parent | |
1006 | * gets woken up by child-exit notifications. | |
1007 | * | |
1008 | * because of cgroup mode, must be called before cgroup_exit() | |
1009 | */ | |
1010 | perf_event_exit_task(tsk); | |
1011 | ||
b4f48b63 | 1012 | cgroup_exit(tsk, 1); |
1da177e4 | 1013 | |
5ec93d11 | 1014 | if (group_dead) |
1da177e4 LT |
1015 | disassociate_ctty(1); |
1016 | ||
a1261f54 | 1017 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 | 1018 | |
9f46080c | 1019 | proc_exit_connector(tsk); |
33b2fb30 | 1020 | |
24f1e32c FW |
1021 | /* |
1022 | * FIXME: do that only when needed, using sched_exit tracepoint | |
1023 | */ | |
bf26c018 | 1024 | ptrace_put_breakpoints(tsk); |
33b2fb30 | 1025 | |
821c7de7 | 1026 | exit_notify(tsk, group_dead); |
1da177e4 | 1027 | #ifdef CONFIG_NUMA |
c0ff7453 | 1028 | task_lock(tsk); |
f0be3d32 | 1029 | mpol_put(tsk->mempolicy); |
1da177e4 | 1030 | tsk->mempolicy = NULL; |
c0ff7453 | 1031 | task_unlock(tsk); |
1da177e4 | 1032 | #endif |
42b2dd0a | 1033 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
1034 | if (unlikely(current->pi_state_cache)) |
1035 | kfree(current->pi_state_cache); | |
42b2dd0a | 1036 | #endif |
de5097c2 | 1037 | /* |
9a11b49a | 1038 | * Make sure we are holding no locks: |
de5097c2 | 1039 | */ |
9a11b49a | 1040 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1041 | /* |
1042 | * We can do this unlocked here. The futex code uses this flag | |
1043 | * just to verify whether the pi state cleanup has been done | |
1044 | * or not. In the worst case it loops once more. | |
1045 | */ | |
1046 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1047 | |
afc847b7 | 1048 | if (tsk->io_context) |
b69f2292 | 1049 | exit_io_context(tsk); |
afc847b7 | 1050 | |
b92ce558 JA |
1051 | if (tsk->splice_pipe) |
1052 | __free_pipe_info(tsk->splice_pipe); | |
1053 | ||
e0e81739 DH |
1054 | validate_creds_for_do_exit(tsk); |
1055 | ||
7407251a | 1056 | preempt_disable(); |
f41d911f | 1057 | exit_rcu(); |
55a101f8 | 1058 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1059 | tsk->state = TASK_DEAD; |
1da177e4 LT |
1060 | schedule(); |
1061 | BUG(); | |
1062 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1063 | for (;;) |
1064 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1065 | } |
1066 | ||
012914da RA |
1067 | EXPORT_SYMBOL_GPL(do_exit); |
1068 | ||
1da177e4 LT |
1069 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1070 | { | |
1071 | if (comp) | |
1072 | complete(comp); | |
55a101f8 | 1073 | |
1da177e4 LT |
1074 | do_exit(code); |
1075 | } | |
1076 | ||
1077 | EXPORT_SYMBOL(complete_and_exit); | |
1078 | ||
754fe8d2 | 1079 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
1080 | { |
1081 | do_exit((error_code&0xff)<<8); | |
1082 | } | |
1083 | ||
1da177e4 LT |
1084 | /* |
1085 | * Take down every thread in the group. This is called by fatal signals | |
1086 | * as well as by sys_exit_group (below). | |
1087 | */ | |
1088 | NORET_TYPE void | |
1089 | do_group_exit(int exit_code) | |
1090 | { | |
bfc4b089 ON |
1091 | struct signal_struct *sig = current->signal; |
1092 | ||
1da177e4 LT |
1093 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1094 | ||
bfc4b089 ON |
1095 | if (signal_group_exit(sig)) |
1096 | exit_code = sig->group_exit_code; | |
1da177e4 | 1097 | else if (!thread_group_empty(current)) { |
1da177e4 | 1098 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1099 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1100 | if (signal_group_exit(sig)) |
1da177e4 LT |
1101 | /* Another thread got here before we took the lock. */ |
1102 | exit_code = sig->group_exit_code; | |
1103 | else { | |
1da177e4 | 1104 | sig->group_exit_code = exit_code; |
ed5d2cac | 1105 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1106 | zap_other_threads(current); |
1107 | } | |
1108 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1109 | } |
1110 | ||
1111 | do_exit(exit_code); | |
1112 | /* NOTREACHED */ | |
1113 | } | |
1114 | ||
1115 | /* | |
1116 | * this kills every thread in the thread group. Note that any externally | |
1117 | * wait4()-ing process will get the correct exit code - even if this | |
1118 | * thread is not the thread group leader. | |
1119 | */ | |
754fe8d2 | 1120 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
1121 | { |
1122 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
1123 | /* NOTREACHED */ |
1124 | return 0; | |
1da177e4 LT |
1125 | } |
1126 | ||
9e8ae01d ON |
1127 | struct wait_opts { |
1128 | enum pid_type wo_type; | |
9e8ae01d | 1129 | int wo_flags; |
e1eb1ebc | 1130 | struct pid *wo_pid; |
9e8ae01d ON |
1131 | |
1132 | struct siginfo __user *wo_info; | |
1133 | int __user *wo_stat; | |
1134 | struct rusage __user *wo_rusage; | |
1135 | ||
0b7570e7 | 1136 | wait_queue_t child_wait; |
9e8ae01d ON |
1137 | int notask_error; |
1138 | }; | |
1139 | ||
989264f4 ON |
1140 | static inline |
1141 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 1142 | { |
989264f4 ON |
1143 | if (type != PIDTYPE_PID) |
1144 | task = task->group_leader; | |
1145 | return task->pids[type].pid; | |
161550d7 EB |
1146 | } |
1147 | ||
989264f4 | 1148 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1149 | { |
5c01ba49 ON |
1150 | return wo->wo_type == PIDTYPE_MAX || |
1151 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1152 | } | |
1da177e4 | 1153 | |
5c01ba49 ON |
1154 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
1155 | { | |
1156 | if (!eligible_pid(wo, p)) | |
1157 | return 0; | |
1da177e4 LT |
1158 | /* Wait for all children (clone and not) if __WALL is set; |
1159 | * otherwise, wait for clone children *only* if __WCLONE is | |
1160 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1161 | * A "clone" child here is one that reports to its parent | |
1162 | * using a signal other than SIGCHLD.) */ | |
9e8ae01d ON |
1163 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
1164 | && !(wo->wo_flags & __WALL)) | |
1da177e4 | 1165 | return 0; |
1da177e4 | 1166 | |
14dd0b81 | 1167 | return 1; |
1da177e4 LT |
1168 | } |
1169 | ||
9e8ae01d ON |
1170 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
1171 | pid_t pid, uid_t uid, int why, int status) | |
1da177e4 | 1172 | { |
9e8ae01d ON |
1173 | struct siginfo __user *infop; |
1174 | int retval = wo->wo_rusage | |
1175 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
36c8b586 | 1176 | |
1da177e4 | 1177 | put_task_struct(p); |
9e8ae01d | 1178 | infop = wo->wo_info; |
b6fe2d11 VM |
1179 | if (infop) { |
1180 | if (!retval) | |
1181 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1182 | if (!retval) | |
1183 | retval = put_user(0, &infop->si_errno); | |
1184 | if (!retval) | |
1185 | retval = put_user((short)why, &infop->si_code); | |
1186 | if (!retval) | |
1187 | retval = put_user(pid, &infop->si_pid); | |
1188 | if (!retval) | |
1189 | retval = put_user(uid, &infop->si_uid); | |
1190 | if (!retval) | |
1191 | retval = put_user(status, &infop->si_status); | |
1192 | } | |
1da177e4 LT |
1193 | if (!retval) |
1194 | retval = pid; | |
1195 | return retval; | |
1196 | } | |
1197 | ||
1198 | /* | |
1199 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1200 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1201 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1202 | * released the lock and the system call should return. | |
1203 | */ | |
9e8ae01d | 1204 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1205 | { |
1206 | unsigned long state; | |
2f4e6e2a | 1207 | int retval, status, traced; |
6c5f3e7b | 1208 | pid_t pid = task_pid_vnr(p); |
c69e8d9c | 1209 | uid_t uid = __task_cred(p)->uid; |
9e8ae01d | 1210 | struct siginfo __user *infop; |
1da177e4 | 1211 | |
9e8ae01d | 1212 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1213 | return 0; |
1214 | ||
9e8ae01d | 1215 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1da177e4 | 1216 | int exit_code = p->exit_code; |
f3abd4f9 | 1217 | int why; |
1da177e4 | 1218 | |
1da177e4 LT |
1219 | get_task_struct(p); |
1220 | read_unlock(&tasklist_lock); | |
1221 | if ((exit_code & 0x7f) == 0) { | |
1222 | why = CLD_EXITED; | |
1223 | status = exit_code >> 8; | |
1224 | } else { | |
1225 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1226 | status = exit_code & 0x7f; | |
1227 | } | |
9e8ae01d | 1228 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1da177e4 LT |
1229 | } |
1230 | ||
1231 | /* | |
1232 | * Try to move the task's state to DEAD | |
1233 | * only one thread is allowed to do this: | |
1234 | */ | |
1235 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1236 | if (state != EXIT_ZOMBIE) { | |
1237 | BUG_ON(state != EXIT_DEAD); | |
1238 | return 0; | |
1239 | } | |
1da177e4 | 1240 | |
53b6f9fb | 1241 | traced = ptrace_reparented(p); |
befca967 ON |
1242 | /* |
1243 | * It can be ptraced but not reparented, check | |
1244 | * !task_detached() to filter out sub-threads. | |
1245 | */ | |
1246 | if (likely(!traced) && likely(!task_detached(p))) { | |
3795e161 JJ |
1247 | struct signal_struct *psig; |
1248 | struct signal_struct *sig; | |
1f10206c | 1249 | unsigned long maxrss; |
0cf55e1e | 1250 | cputime_t tgutime, tgstime; |
3795e161 | 1251 | |
1da177e4 LT |
1252 | /* |
1253 | * The resource counters for the group leader are in its | |
1254 | * own task_struct. Those for dead threads in the group | |
1255 | * are in its signal_struct, as are those for the child | |
1256 | * processes it has previously reaped. All these | |
1257 | * accumulate in the parent's signal_struct c* fields. | |
1258 | * | |
1259 | * We don't bother to take a lock here to protect these | |
1260 | * p->signal fields, because they are only touched by | |
1261 | * __exit_signal, which runs with tasklist_lock | |
1262 | * write-locked anyway, and so is excluded here. We do | |
d1e98f42 | 1263 | * need to protect the access to parent->signal fields, |
1da177e4 LT |
1264 | * as other threads in the parent group can be right |
1265 | * here reaping other children at the same time. | |
0cf55e1e HS |
1266 | * |
1267 | * We use thread_group_times() to get times for the thread | |
1268 | * group, which consolidates times for all threads in the | |
1269 | * group including the group leader. | |
1da177e4 | 1270 | */ |
0cf55e1e | 1271 | thread_group_times(p, &tgutime, &tgstime); |
d1e98f42 ON |
1272 | spin_lock_irq(&p->real_parent->sighand->siglock); |
1273 | psig = p->real_parent->signal; | |
3795e161 JJ |
1274 | sig = p->signal; |
1275 | psig->cutime = | |
1276 | cputime_add(psig->cutime, | |
0cf55e1e HS |
1277 | cputime_add(tgutime, |
1278 | sig->cutime)); | |
3795e161 JJ |
1279 | psig->cstime = |
1280 | cputime_add(psig->cstime, | |
0cf55e1e HS |
1281 | cputime_add(tgstime, |
1282 | sig->cstime)); | |
9ac52315 LV |
1283 | psig->cgtime = |
1284 | cputime_add(psig->cgtime, | |
1285 | cputime_add(p->gtime, | |
1286 | cputime_add(sig->gtime, | |
1287 | sig->cgtime))); | |
3795e161 JJ |
1288 | psig->cmin_flt += |
1289 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1290 | psig->cmaj_flt += | |
1291 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1292 | psig->cnvcsw += | |
1293 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1294 | psig->cnivcsw += | |
1295 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1296 | psig->cinblock += |
1297 | task_io_get_inblock(p) + | |
1298 | sig->inblock + sig->cinblock; | |
1299 | psig->coublock += | |
1300 | task_io_get_oublock(p) + | |
1301 | sig->oublock + sig->coublock; | |
1f10206c JP |
1302 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1303 | if (psig->cmaxrss < maxrss) | |
1304 | psig->cmaxrss = maxrss; | |
5995477a AR |
1305 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1306 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
d1e98f42 | 1307 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
1da177e4 LT |
1308 | } |
1309 | ||
1310 | /* | |
1311 | * Now we are sure this task is interesting, and no other | |
1312 | * thread can reap it because we set its state to EXIT_DEAD. | |
1313 | */ | |
1314 | read_unlock(&tasklist_lock); | |
1315 | ||
9e8ae01d ON |
1316 | retval = wo->wo_rusage |
1317 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 LT |
1318 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1319 | ? p->signal->group_exit_code : p->exit_code; | |
9e8ae01d ON |
1320 | if (!retval && wo->wo_stat) |
1321 | retval = put_user(status, wo->wo_stat); | |
1322 | ||
1323 | infop = wo->wo_info; | |
1da177e4 LT |
1324 | if (!retval && infop) |
1325 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1326 | if (!retval && infop) | |
1327 | retval = put_user(0, &infop->si_errno); | |
1328 | if (!retval && infop) { | |
1329 | int why; | |
1330 | ||
1331 | if ((status & 0x7f) == 0) { | |
1332 | why = CLD_EXITED; | |
1333 | status >>= 8; | |
1334 | } else { | |
1335 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1336 | status &= 0x7f; | |
1337 | } | |
1338 | retval = put_user((short)why, &infop->si_code); | |
1339 | if (!retval) | |
1340 | retval = put_user(status, &infop->si_status); | |
1341 | } | |
1342 | if (!retval && infop) | |
3a515e4a | 1343 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1344 | if (!retval && infop) |
c69e8d9c | 1345 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1346 | if (!retval) |
3a515e4a | 1347 | retval = pid; |
2f4e6e2a ON |
1348 | |
1349 | if (traced) { | |
1da177e4 | 1350 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1351 | /* We dropped tasklist, ptracer could die and untrace */ |
1352 | ptrace_unlink(p); | |
1353 | /* | |
1354 | * If this is not a detached task, notify the parent. | |
1355 | * If it's still not detached after that, don't release | |
1356 | * it now. | |
1357 | */ | |
d839fd4d | 1358 | if (!task_detached(p)) { |
2f4e6e2a | 1359 | do_notify_parent(p, p->exit_signal); |
d839fd4d | 1360 | if (!task_detached(p)) { |
2f4e6e2a ON |
1361 | p->exit_state = EXIT_ZOMBIE; |
1362 | p = NULL; | |
1da177e4 LT |
1363 | } |
1364 | } | |
1365 | write_unlock_irq(&tasklist_lock); | |
1366 | } | |
1367 | if (p != NULL) | |
1368 | release_task(p); | |
2f4e6e2a | 1369 | |
1da177e4 LT |
1370 | return retval; |
1371 | } | |
1372 | ||
90bc8d8b ON |
1373 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1374 | { | |
1375 | if (ptrace) { | |
544b2c91 TH |
1376 | if (task_is_stopped_or_traced(p) && |
1377 | !(p->jobctl & JOBCTL_LISTENING)) | |
90bc8d8b ON |
1378 | return &p->exit_code; |
1379 | } else { | |
1380 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1381 | return &p->signal->group_exit_code; | |
1382 | } | |
1383 | return NULL; | |
1384 | } | |
1385 | ||
19e27463 TH |
1386 | /** |
1387 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1388 | * @wo: wait options | |
1389 | * @ptrace: is the wait for ptrace | |
1390 | * @p: task to wait for | |
1391 | * | |
1392 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1393 | * | |
1394 | * CONTEXT: | |
1395 | * read_lock(&tasklist_lock), which is released if return value is | |
1396 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1397 | * | |
1398 | * RETURNS: | |
1399 | * 0 if wait condition didn't exist and search for other wait conditions | |
1400 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1401 | * success, implies that tasklist_lock is released and wait condition | |
1402 | * search should terminate. | |
1da177e4 | 1403 | */ |
9e8ae01d ON |
1404 | static int wait_task_stopped(struct wait_opts *wo, |
1405 | int ptrace, struct task_struct *p) | |
1da177e4 | 1406 | { |
9e8ae01d | 1407 | struct siginfo __user *infop; |
90bc8d8b | 1408 | int retval, exit_code, *p_code, why; |
ee7c82da | 1409 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1410 | pid_t pid; |
1da177e4 | 1411 | |
47918025 ON |
1412 | /* |
1413 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1414 | */ | |
9e8ae01d | 1415 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1416 | return 0; |
1417 | ||
19e27463 TH |
1418 | if (!task_stopped_code(p, ptrace)) |
1419 | return 0; | |
1420 | ||
ee7c82da ON |
1421 | exit_code = 0; |
1422 | spin_lock_irq(&p->sighand->siglock); | |
1423 | ||
90bc8d8b ON |
1424 | p_code = task_stopped_code(p, ptrace); |
1425 | if (unlikely(!p_code)) | |
ee7c82da ON |
1426 | goto unlock_sig; |
1427 | ||
90bc8d8b | 1428 | exit_code = *p_code; |
ee7c82da ON |
1429 | if (!exit_code) |
1430 | goto unlock_sig; | |
1431 | ||
9e8ae01d | 1432 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1433 | *p_code = 0; |
ee7c82da | 1434 | |
f362b732 | 1435 | uid = task_uid(p); |
ee7c82da ON |
1436 | unlock_sig: |
1437 | spin_unlock_irq(&p->sighand->siglock); | |
1438 | if (!exit_code) | |
1da177e4 LT |
1439 | return 0; |
1440 | ||
1441 | /* | |
1442 | * Now we are pretty sure this task is interesting. | |
1443 | * Make sure it doesn't get reaped out from under us while we | |
1444 | * give up the lock and then examine it below. We don't want to | |
1445 | * keep holding onto the tasklist_lock while we call getrusage and | |
1446 | * possibly take page faults for user memory. | |
1447 | */ | |
1448 | get_task_struct(p); | |
6c5f3e7b | 1449 | pid = task_pid_vnr(p); |
f470021a | 1450 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1451 | read_unlock(&tasklist_lock); |
1452 | ||
9e8ae01d ON |
1453 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1454 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | |
1455 | ||
1456 | retval = wo->wo_rusage | |
1457 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1458 | if (!retval && wo->wo_stat) | |
1459 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | |
1da177e4 | 1460 | |
9e8ae01d | 1461 | infop = wo->wo_info; |
1da177e4 LT |
1462 | if (!retval && infop) |
1463 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1464 | if (!retval && infop) | |
1465 | retval = put_user(0, &infop->si_errno); | |
1466 | if (!retval && infop) | |
6efcae46 | 1467 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1468 | if (!retval && infop) |
1469 | retval = put_user(exit_code, &infop->si_status); | |
1470 | if (!retval && infop) | |
c8950783 | 1471 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1472 | if (!retval && infop) |
ee7c82da | 1473 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1474 | if (!retval) |
c8950783 | 1475 | retval = pid; |
1da177e4 LT |
1476 | put_task_struct(p); |
1477 | ||
1478 | BUG_ON(!retval); | |
1479 | return retval; | |
1480 | } | |
1481 | ||
1482 | /* | |
1483 | * Handle do_wait work for one task in a live, non-stopped state. | |
1484 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1485 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1486 | * released the lock and the system call should return. | |
1487 | */ | |
9e8ae01d | 1488 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1489 | { |
1490 | int retval; | |
1491 | pid_t pid; | |
1492 | uid_t uid; | |
1493 | ||
9e8ae01d | 1494 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1495 | return 0; |
1496 | ||
1da177e4 LT |
1497 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1498 | return 0; | |
1499 | ||
1500 | spin_lock_irq(&p->sighand->siglock); | |
1501 | /* Re-check with the lock held. */ | |
1502 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1503 | spin_unlock_irq(&p->sighand->siglock); | |
1504 | return 0; | |
1505 | } | |
9e8ae01d | 1506 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1507 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
f362b732 | 1508 | uid = task_uid(p); |
1da177e4 LT |
1509 | spin_unlock_irq(&p->sighand->siglock); |
1510 | ||
6c5f3e7b | 1511 | pid = task_pid_vnr(p); |
1da177e4 LT |
1512 | get_task_struct(p); |
1513 | read_unlock(&tasklist_lock); | |
1514 | ||
9e8ae01d ON |
1515 | if (!wo->wo_info) { |
1516 | retval = wo->wo_rusage | |
1517 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 | 1518 | put_task_struct(p); |
9e8ae01d ON |
1519 | if (!retval && wo->wo_stat) |
1520 | retval = put_user(0xffff, wo->wo_stat); | |
1da177e4 | 1521 | if (!retval) |
3a515e4a | 1522 | retval = pid; |
1da177e4 | 1523 | } else { |
9e8ae01d ON |
1524 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1525 | CLD_CONTINUED, SIGCONT); | |
1da177e4 LT |
1526 | BUG_ON(retval == 0); |
1527 | } | |
1528 | ||
1529 | return retval; | |
1530 | } | |
1531 | ||
98abed02 RM |
1532 | /* |
1533 | * Consider @p for a wait by @parent. | |
1534 | * | |
9e8ae01d | 1535 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1536 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1537 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1538 | * then ->notask_error is 0 if @p is an eligible child, |
14dd0b81 | 1539 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1540 | */ |
b6e763f0 ON |
1541 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1542 | struct task_struct *p) | |
98abed02 | 1543 | { |
9e8ae01d | 1544 | int ret = eligible_child(wo, p); |
14dd0b81 | 1545 | if (!ret) |
98abed02 RM |
1546 | return ret; |
1547 | ||
a2322e1d | 1548 | ret = security_task_wait(p); |
14dd0b81 RM |
1549 | if (unlikely(ret < 0)) { |
1550 | /* | |
1551 | * If we have not yet seen any eligible child, | |
1552 | * then let this error code replace -ECHILD. | |
1553 | * A permission error will give the user a clue | |
1554 | * to look for security policy problems, rather | |
1555 | * than for mysterious wait bugs. | |
1556 | */ | |
9e8ae01d ON |
1557 | if (wo->notask_error) |
1558 | wo->notask_error = ret; | |
78a3d9d5 | 1559 | return 0; |
14dd0b81 RM |
1560 | } |
1561 | ||
823b018e TH |
1562 | /* dead body doesn't have much to contribute */ |
1563 | if (p->exit_state == EXIT_DEAD) | |
1564 | return 0; | |
1565 | ||
45cb24a1 TH |
1566 | /* slay zombie? */ |
1567 | if (p->exit_state == EXIT_ZOMBIE) { | |
f470021a | 1568 | /* |
45cb24a1 TH |
1569 | * A zombie ptracee is only visible to its ptracer. |
1570 | * Notification and reaping will be cascaded to the real | |
1571 | * parent when the ptracer detaches. | |
f470021a | 1572 | */ |
d21142ec | 1573 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
45cb24a1 TH |
1574 | /* it will become visible, clear notask_error */ |
1575 | wo->notask_error = 0; | |
1576 | return 0; | |
1577 | } | |
f470021a | 1578 | |
9b84cca2 TH |
1579 | /* we don't reap group leaders with subthreads */ |
1580 | if (!delay_group_leader(p)) | |
1581 | return wait_task_zombie(wo, p); | |
98abed02 | 1582 | |
f470021a | 1583 | /* |
9b84cca2 TH |
1584 | * Allow access to stopped/continued state via zombie by |
1585 | * falling through. Clearing of notask_error is complex. | |
1586 | * | |
1587 | * When !@ptrace: | |
1588 | * | |
1589 | * If WEXITED is set, notask_error should naturally be | |
1590 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1591 | * so, if there are live subthreads, there are events to | |
1592 | * wait for. If all subthreads are dead, it's still safe | |
1593 | * to clear - this function will be called again in finite | |
1594 | * amount time once all the subthreads are released and | |
1595 | * will then return without clearing. | |
1596 | * | |
1597 | * When @ptrace: | |
1598 | * | |
1599 | * Stopped state is per-task and thus can't change once the | |
1600 | * target task dies. Only continued and exited can happen. | |
1601 | * Clear notask_error if WCONTINUED | WEXITED. | |
1602 | */ | |
1603 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1604 | wo->notask_error = 0; | |
1605 | } else { | |
45cb24a1 TH |
1606 | /* |
1607 | * If @p is ptraced by a task in its real parent's group, | |
1608 | * hide group stop/continued state when looking at @p as | |
1609 | * the real parent; otherwise, a single stop can be | |
1610 | * reported twice as group and ptrace stops. | |
1611 | * | |
1612 | * If a ptracer wants to distinguish the two events for its | |
1613 | * own children, it should create a separate process which | |
1614 | * takes the role of real parent. | |
1615 | */ | |
d21142ec | 1616 | if (likely(!ptrace) && p->ptrace && |
45cb24a1 TH |
1617 | same_thread_group(p->parent, p->real_parent)) |
1618 | return 0; | |
1619 | ||
9b84cca2 TH |
1620 | /* |
1621 | * @p is alive and it's gonna stop, continue or exit, so | |
1622 | * there always is something to wait for. | |
f470021a | 1623 | */ |
9e8ae01d | 1624 | wo->notask_error = 0; |
f470021a RM |
1625 | } |
1626 | ||
98abed02 | 1627 | /* |
45cb24a1 TH |
1628 | * Wait for stopped. Depending on @ptrace, different stopped state |
1629 | * is used and the two don't interact with each other. | |
98abed02 | 1630 | */ |
19e27463 TH |
1631 | ret = wait_task_stopped(wo, ptrace, p); |
1632 | if (ret) | |
1633 | return ret; | |
98abed02 RM |
1634 | |
1635 | /* | |
45cb24a1 TH |
1636 | * Wait for continued. There's only one continued state and the |
1637 | * ptracer can consume it which can confuse the real parent. Don't | |
1638 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1639 | */ |
9e8ae01d | 1640 | return wait_task_continued(wo, p); |
98abed02 RM |
1641 | } |
1642 | ||
1643 | /* | |
1644 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1645 | * | |
9e8ae01d | 1646 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1647 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1648 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1649 | * ->notask_error is 0 if there were any eligible children, |
14dd0b81 | 1650 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1651 | */ |
9e8ae01d | 1652 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1653 | { |
1654 | struct task_struct *p; | |
1655 | ||
1656 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb ON |
1657 | int ret = wait_consider_task(wo, 0, p); |
1658 | if (ret) | |
1659 | return ret; | |
98abed02 RM |
1660 | } |
1661 | ||
1662 | return 0; | |
1663 | } | |
1664 | ||
9e8ae01d | 1665 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1666 | { |
1667 | struct task_struct *p; | |
1668 | ||
f470021a | 1669 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1670 | int ret = wait_consider_task(wo, 1, p); |
f470021a | 1671 | if (ret) |
98abed02 | 1672 | return ret; |
98abed02 RM |
1673 | } |
1674 | ||
1675 | return 0; | |
1676 | } | |
1677 | ||
0b7570e7 ON |
1678 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1679 | int sync, void *key) | |
1680 | { | |
1681 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1682 | child_wait); | |
1683 | struct task_struct *p = key; | |
1684 | ||
5c01ba49 | 1685 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1686 | return 0; |
1687 | ||
b4fe5182 ON |
1688 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1689 | return 0; | |
1690 | ||
0b7570e7 ON |
1691 | return default_wake_function(wait, mode, sync, key); |
1692 | } | |
1693 | ||
a7f0765e ON |
1694 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1695 | { | |
0b7570e7 ON |
1696 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1697 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1698 | } |
1699 | ||
9e8ae01d | 1700 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1701 | { |
1da177e4 | 1702 | struct task_struct *tsk; |
98abed02 | 1703 | int retval; |
1da177e4 | 1704 | |
9e8ae01d | 1705 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1706 | |
0b7570e7 ON |
1707 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1708 | wo->child_wait.private = current; | |
1709 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1710 | repeat: |
98abed02 RM |
1711 | /* |
1712 | * If there is nothing that can match our critiera just get out. | |
9e8ae01d ON |
1713 | * We will clear ->notask_error to zero if we see any child that |
1714 | * might later match our criteria, even if we are not able to reap | |
1715 | * it yet. | |
98abed02 | 1716 | */ |
64a16caf | 1717 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1718 | if ((wo->wo_type < PIDTYPE_MAX) && |
1719 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1720 | goto notask; |
161550d7 | 1721 | |
f95d39d1 | 1722 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1723 | read_lock(&tasklist_lock); |
1724 | tsk = current; | |
1725 | do { | |
64a16caf ON |
1726 | retval = do_wait_thread(wo, tsk); |
1727 | if (retval) | |
1728 | goto end; | |
9e8ae01d | 1729 | |
64a16caf ON |
1730 | retval = ptrace_do_wait(wo, tsk); |
1731 | if (retval) | |
98abed02 | 1732 | goto end; |
98abed02 | 1733 | |
9e8ae01d | 1734 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1735 | break; |
a3f6dfb7 | 1736 | } while_each_thread(current, tsk); |
1da177e4 | 1737 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1738 | |
64a16caf | 1739 | notask: |
9e8ae01d ON |
1740 | retval = wo->notask_error; |
1741 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1742 | retval = -ERESTARTSYS; |
98abed02 RM |
1743 | if (!signal_pending(current)) { |
1744 | schedule(); | |
1745 | goto repeat; | |
1746 | } | |
1da177e4 | 1747 | } |
1da177e4 | 1748 | end: |
f95d39d1 | 1749 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1750 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1751 | return retval; |
1752 | } | |
1753 | ||
17da2bd9 HC |
1754 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1755 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1756 | { |
9e8ae01d | 1757 | struct wait_opts wo; |
161550d7 EB |
1758 | struct pid *pid = NULL; |
1759 | enum pid_type type; | |
1da177e4 LT |
1760 | long ret; |
1761 | ||
1762 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1763 | return -EINVAL; | |
1764 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1765 | return -EINVAL; | |
1766 | ||
1767 | switch (which) { | |
1768 | case P_ALL: | |
161550d7 | 1769 | type = PIDTYPE_MAX; |
1da177e4 LT |
1770 | break; |
1771 | case P_PID: | |
161550d7 EB |
1772 | type = PIDTYPE_PID; |
1773 | if (upid <= 0) | |
1da177e4 LT |
1774 | return -EINVAL; |
1775 | break; | |
1776 | case P_PGID: | |
161550d7 EB |
1777 | type = PIDTYPE_PGID; |
1778 | if (upid <= 0) | |
1da177e4 | 1779 | return -EINVAL; |
1da177e4 LT |
1780 | break; |
1781 | default: | |
1782 | return -EINVAL; | |
1783 | } | |
1784 | ||
161550d7 EB |
1785 | if (type < PIDTYPE_MAX) |
1786 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1787 | |
1788 | wo.wo_type = type; | |
1789 | wo.wo_pid = pid; | |
1790 | wo.wo_flags = options; | |
1791 | wo.wo_info = infop; | |
1792 | wo.wo_stat = NULL; | |
1793 | wo.wo_rusage = ru; | |
1794 | ret = do_wait(&wo); | |
dfe16dfa VM |
1795 | |
1796 | if (ret > 0) { | |
1797 | ret = 0; | |
1798 | } else if (infop) { | |
1799 | /* | |
1800 | * For a WNOHANG return, clear out all the fields | |
1801 | * we would set so the user can easily tell the | |
1802 | * difference. | |
1803 | */ | |
1804 | if (!ret) | |
1805 | ret = put_user(0, &infop->si_signo); | |
1806 | if (!ret) | |
1807 | ret = put_user(0, &infop->si_errno); | |
1808 | if (!ret) | |
1809 | ret = put_user(0, &infop->si_code); | |
1810 | if (!ret) | |
1811 | ret = put_user(0, &infop->si_pid); | |
1812 | if (!ret) | |
1813 | ret = put_user(0, &infop->si_uid); | |
1814 | if (!ret) | |
1815 | ret = put_user(0, &infop->si_status); | |
1816 | } | |
1817 | ||
161550d7 | 1818 | put_pid(pid); |
1da177e4 LT |
1819 | |
1820 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1821 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1822 | return ret; |
1823 | } | |
1824 | ||
754fe8d2 HC |
1825 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1826 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1827 | { |
9e8ae01d | 1828 | struct wait_opts wo; |
161550d7 EB |
1829 | struct pid *pid = NULL; |
1830 | enum pid_type type; | |
1da177e4 LT |
1831 | long ret; |
1832 | ||
1833 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1834 | __WNOTHREAD|__WCLONE|__WALL)) | |
1835 | return -EINVAL; | |
161550d7 EB |
1836 | |
1837 | if (upid == -1) | |
1838 | type = PIDTYPE_MAX; | |
1839 | else if (upid < 0) { | |
1840 | type = PIDTYPE_PGID; | |
1841 | pid = find_get_pid(-upid); | |
1842 | } else if (upid == 0) { | |
1843 | type = PIDTYPE_PGID; | |
2ae448ef | 1844 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1845 | } else /* upid > 0 */ { |
1846 | type = PIDTYPE_PID; | |
1847 | pid = find_get_pid(upid); | |
1848 | } | |
1849 | ||
9e8ae01d ON |
1850 | wo.wo_type = type; |
1851 | wo.wo_pid = pid; | |
1852 | wo.wo_flags = options | WEXITED; | |
1853 | wo.wo_info = NULL; | |
1854 | wo.wo_stat = stat_addr; | |
1855 | wo.wo_rusage = ru; | |
1856 | ret = do_wait(&wo); | |
161550d7 | 1857 | put_pid(pid); |
1da177e4 LT |
1858 | |
1859 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1860 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1861 | return ret; |
1862 | } | |
1863 | ||
1864 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1865 | ||
1866 | /* | |
1867 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1868 | * implemented by calling sys_wait4() from libc.a. | |
1869 | */ | |
17da2bd9 | 1870 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1871 | { |
1872 | return sys_wait4(pid, stat_addr, options, NULL); | |
1873 | } | |
1874 | ||
1875 | #endif |