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> |
54848d73 | 54 | #include <linux/writeback.h> |
40401530 | 55 | #include <linux/shm.h> |
1da177e4 LT |
56 | |
57 | #include <asm/uaccess.h> | |
58 | #include <asm/unistd.h> | |
59 | #include <asm/pgtable.h> | |
60 | #include <asm/mmu_context.h> | |
61 | ||
408b664a AB |
62 | static void exit_mm(struct task_struct * tsk); |
63 | ||
d40e48e0 | 64 | static void __unhash_process(struct task_struct *p, bool group_dead) |
1da177e4 LT |
65 | { |
66 | nr_threads--; | |
67 | detach_pid(p, PIDTYPE_PID); | |
d40e48e0 | 68 | if (group_dead) { |
1da177e4 LT |
69 | detach_pid(p, PIDTYPE_PGID); |
70 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 71 | |
5e85d4ab | 72 | list_del_rcu(&p->tasks); |
9cd80bbb | 73 | list_del_init(&p->sibling); |
909ea964 | 74 | __this_cpu_dec(process_counts); |
1da177e4 | 75 | } |
47e65328 | 76 | list_del_rcu(&p->thread_group); |
1da177e4 LT |
77 | } |
78 | ||
6a14c5c9 ON |
79 | /* |
80 | * This function expects the tasklist_lock write-locked. | |
81 | */ | |
82 | static void __exit_signal(struct task_struct *tsk) | |
83 | { | |
84 | struct signal_struct *sig = tsk->signal; | |
d40e48e0 | 85 | bool group_dead = thread_group_leader(tsk); |
6a14c5c9 | 86 | struct sighand_struct *sighand; |
4ada856f | 87 | struct tty_struct *uninitialized_var(tty); |
6a14c5c9 | 88 | |
d11c563d | 89 | sighand = rcu_dereference_check(tsk->sighand, |
db1466b3 | 90 | lockdep_tasklist_lock_is_held()); |
6a14c5c9 ON |
91 | spin_lock(&sighand->siglock); |
92 | ||
93 | posix_cpu_timers_exit(tsk); | |
d40e48e0 | 94 | if (group_dead) { |
6a14c5c9 | 95 | posix_cpu_timers_exit_group(tsk); |
4ada856f ON |
96 | tty = sig->tty; |
97 | sig->tty = NULL; | |
4a599942 | 98 | } else { |
e0a70217 ON |
99 | /* |
100 | * This can only happen if the caller is de_thread(). | |
101 | * FIXME: this is the temporary hack, we should teach | |
102 | * posix-cpu-timers to handle this case correctly. | |
103 | */ | |
104 | if (unlikely(has_group_leader_pid(tsk))) | |
105 | posix_cpu_timers_exit_group(tsk); | |
106 | ||
6a14c5c9 ON |
107 | /* |
108 | * If there is any task waiting for the group exit | |
109 | * then notify it: | |
110 | */ | |
d344193a | 111 | if (sig->notify_count > 0 && !--sig->notify_count) |
6a14c5c9 | 112 | wake_up_process(sig->group_exit_task); |
6db840fa | 113 | |
6a14c5c9 ON |
114 | if (tsk == sig->curr_target) |
115 | sig->curr_target = next_thread(tsk); | |
116 | /* | |
117 | * Accumulate here the counters for all threads but the | |
118 | * group leader as they die, so they can be added into | |
119 | * the process-wide totals when those are taken. | |
120 | * The group leader stays around as a zombie as long | |
121 | * as there are other threads. When it gets reaped, | |
122 | * the exit.c code will add its counts into these totals. | |
123 | * We won't ever get here for the group leader, since it | |
124 | * will have been the last reference on the signal_struct. | |
125 | */ | |
64861634 MS |
126 | sig->utime += tsk->utime; |
127 | sig->stime += tsk->stime; | |
128 | sig->gtime += tsk->gtime; | |
6a14c5c9 ON |
129 | sig->min_flt += tsk->min_flt; |
130 | sig->maj_flt += tsk->maj_flt; | |
131 | sig->nvcsw += tsk->nvcsw; | |
132 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
133 | sig->inblock += task_io_get_inblock(tsk); |
134 | sig->oublock += task_io_get_oublock(tsk); | |
5995477a | 135 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
32bd671d | 136 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
137 | } |
138 | ||
b3ac022c | 139 | sig->nr_threads--; |
d40e48e0 | 140 | __unhash_process(tsk, group_dead); |
5876700c | 141 | |
da7978b0 ON |
142 | /* |
143 | * Do this under ->siglock, we can race with another thread | |
144 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
145 | */ | |
146 | flush_sigqueue(&tsk->pending); | |
a7e5328a | 147 | tsk->sighand = NULL; |
6a14c5c9 | 148 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 149 | |
a7e5328a | 150 | __cleanup_sighand(sighand); |
6a14c5c9 | 151 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
d40e48e0 | 152 | if (group_dead) { |
6a14c5c9 | 153 | flush_sigqueue(&sig->shared_pending); |
4ada856f | 154 | tty_kref_put(tty); |
6a14c5c9 ON |
155 | } |
156 | } | |
157 | ||
8c7904a0 EB |
158 | static void delayed_put_task_struct(struct rcu_head *rhp) |
159 | { | |
0a16b607 MD |
160 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
161 | ||
4e231c79 | 162 | perf_event_delayed_put(tsk); |
0a16b607 MD |
163 | trace_sched_process_free(tsk); |
164 | put_task_struct(tsk); | |
8c7904a0 EB |
165 | } |
166 | ||
f470021a | 167 | |
1da177e4 LT |
168 | void release_task(struct task_struct * p) |
169 | { | |
36c8b586 | 170 | struct task_struct *leader; |
1da177e4 | 171 | int zap_leader; |
1f09f974 | 172 | repeat: |
c69e8d9c | 173 | /* don't need to get the RCU readlock here - the process is dead and |
d11c563d PM |
174 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
175 | rcu_read_lock(); | |
c69e8d9c | 176 | atomic_dec(&__task_cred(p)->user->processes); |
d11c563d | 177 | rcu_read_unlock(); |
c69e8d9c | 178 | |
60347f67 | 179 | proc_flush_task(p); |
0203026b | 180 | |
1da177e4 | 181 | write_lock_irq(&tasklist_lock); |
a288eecc | 182 | ptrace_release_task(p); |
1da177e4 | 183 | __exit_signal(p); |
35f5cad8 | 184 | |
1da177e4 LT |
185 | /* |
186 | * If we are the last non-leader member of the thread | |
187 | * group, and the leader is zombie, then notify the | |
188 | * group leader's parent process. (if it wants notification.) | |
189 | */ | |
190 | zap_leader = 0; | |
191 | leader = p->group_leader; | |
192 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
1da177e4 LT |
193 | /* |
194 | * If we were the last child thread and the leader has | |
195 | * exited already, and the leader's parent ignores SIGCHLD, | |
196 | * then we are the one who should release the leader. | |
dae33574 | 197 | */ |
86773473 | 198 | zap_leader = do_notify_parent(leader, leader->exit_signal); |
dae33574 RM |
199 | if (zap_leader) |
200 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
201 | } |
202 | ||
1da177e4 | 203 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 204 | release_thread(p); |
8c7904a0 | 205 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
206 | |
207 | p = leader; | |
208 | if (unlikely(zap_leader)) | |
209 | goto repeat; | |
210 | } | |
211 | ||
1da177e4 LT |
212 | /* |
213 | * This checks not only the pgrp, but falls back on the pid if no | |
214 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
215 | * without this... | |
04a2e6a5 EB |
216 | * |
217 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 218 | */ |
04a2e6a5 | 219 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
220 | { |
221 | struct task_struct *p; | |
04a2e6a5 | 222 | struct pid *sid = NULL; |
62dfb554 | 223 | |
04a2e6a5 | 224 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 225 | if (p == NULL) |
04a2e6a5 | 226 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 227 | if (p != NULL) |
04a2e6a5 | 228 | sid = task_session(p); |
62dfb554 | 229 | |
1da177e4 LT |
230 | return sid; |
231 | } | |
232 | ||
233 | /* | |
234 | * Determine if a process group is "orphaned", according to the POSIX | |
235 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
236 | * by terminal-generated stop signals. Newly orphaned process groups are | |
237 | * to receive a SIGHUP and a SIGCONT. | |
238 | * | |
239 | * "I ask you, have you ever known what it is to be an orphan?" | |
240 | */ | |
0475ac08 | 241 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
242 | { |
243 | struct task_struct *p; | |
1da177e4 | 244 | |
0475ac08 | 245 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
246 | if ((p == ignored_task) || |
247 | (p->exit_state && thread_group_empty(p)) || | |
248 | is_global_init(p->real_parent)) | |
1da177e4 | 249 | continue; |
05e83df6 | 250 | |
0475ac08 | 251 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
252 | task_session(p->real_parent) == task_session(p)) |
253 | return 0; | |
0475ac08 | 254 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
255 | |
256 | return 1; | |
1da177e4 LT |
257 | } |
258 | ||
3e7cd6c4 | 259 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
260 | { |
261 | int retval; | |
262 | ||
263 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 264 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
265 | read_unlock(&tasklist_lock); |
266 | ||
267 | return retval; | |
268 | } | |
269 | ||
961c4675 | 270 | static bool has_stopped_jobs(struct pid *pgrp) |
1da177e4 | 271 | { |
1da177e4 LT |
272 | struct task_struct *p; |
273 | ||
0475ac08 | 274 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
961c4675 ON |
275 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
276 | return true; | |
0475ac08 | 277 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
961c4675 ON |
278 | |
279 | return false; | |
1da177e4 LT |
280 | } |
281 | ||
f49ee505 ON |
282 | /* |
283 | * Check to see if any process groups have become orphaned as | |
284 | * a result of our exiting, and if they have any stopped jobs, | |
285 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
286 | */ | |
287 | static void | |
288 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
289 | { | |
290 | struct pid *pgrp = task_pgrp(tsk); | |
291 | struct task_struct *ignored_task = tsk; | |
292 | ||
293 | if (!parent) | |
294 | /* exit: our father is in a different pgrp than | |
295 | * we are and we were the only connection outside. | |
296 | */ | |
297 | parent = tsk->real_parent; | |
298 | else | |
299 | /* reparent: our child is in a different pgrp than | |
300 | * we are, and it was the only connection outside. | |
301 | */ | |
302 | ignored_task = NULL; | |
303 | ||
304 | if (task_pgrp(parent) != pgrp && | |
305 | task_session(parent) == task_session(tsk) && | |
306 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
307 | has_stopped_jobs(pgrp)) { | |
308 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
309 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
310 | } | |
311 | } | |
312 | ||
1da177e4 | 313 | /** |
49d769d5 | 314 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
315 | * |
316 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
317 | * it ever exits, it should generally reparent itself to kthreadd so it |
318 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
319 | * |
320 | * The various task state such as scheduling policy and priority may have | |
321 | * been inherited from a user process, so we reset them to sane values here. | |
322 | * | |
49d769d5 | 323 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 324 | */ |
49d769d5 | 325 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
326 | { |
327 | write_lock_irq(&tasklist_lock); | |
328 | ||
329 | ptrace_unlink(current); | |
330 | /* Reparent to init */ | |
49d769d5 | 331 | current->real_parent = current->parent = kthreadd_task; |
f470021a | 332 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
1da177e4 LT |
333 | |
334 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
335 | current->exit_signal = SIGCHLD; | |
336 | ||
e05606d3 | 337 | if (task_nice(current) < 0) |
1da177e4 LT |
338 | set_user_nice(current, 0); |
339 | /* cpus_allowed? */ | |
340 | /* rt_priority? */ | |
341 | /* signals? */ | |
1da177e4 LT |
342 | memcpy(current->signal->rlim, init_task.signal->rlim, |
343 | sizeof(current->signal->rlim)); | |
d84f4f99 DH |
344 | |
345 | atomic_inc(&init_cred.usage); | |
346 | commit_creds(&init_cred); | |
1da177e4 | 347 | write_unlock_irq(&tasklist_lock); |
1da177e4 LT |
348 | } |
349 | ||
8520d7c7 | 350 | void __set_special_pids(struct pid *pid) |
1da177e4 | 351 | { |
e19f247a | 352 | struct task_struct *curr = current->group_leader; |
1da177e4 | 353 | |
0d0df599 | 354 | if (task_session(curr) != pid) |
7d8da096 | 355 | change_pid(curr, PIDTYPE_SID, pid); |
1b0f7ffd ON |
356 | |
357 | if (task_pgrp(curr) != pid) | |
7d8da096 | 358 | change_pid(curr, PIDTYPE_PGID, pid); |
1da177e4 LT |
359 | } |
360 | ||
8520d7c7 | 361 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
362 | { |
363 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 364 | __set_special_pids(pid); |
1da177e4 LT |
365 | write_unlock_irq(&tasklist_lock); |
366 | } | |
367 | ||
368 | /* | |
87245135 ON |
369 | * Let kernel threads use this to say that they allow a certain signal. |
370 | * Must not be used if kthread was cloned with CLONE_SIGHAND. | |
1da177e4 LT |
371 | */ |
372 | int allow_signal(int sig) | |
373 | { | |
7ed20e1a | 374 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
375 | return -EINVAL; |
376 | ||
377 | spin_lock_irq(¤t->sighand->siglock); | |
87245135 | 378 | /* This is only needed for daemonize()'ed kthreads */ |
1da177e4 | 379 | sigdelset(¤t->blocked, sig); |
87245135 ON |
380 | /* |
381 | * Kernel threads handle their own signals. Let the signal code | |
382 | * know it'll be handled, so that they don't get converted to | |
383 | * SIGKILL or just silently dropped. | |
384 | */ | |
385 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
1da177e4 LT |
386 | recalc_sigpending(); |
387 | spin_unlock_irq(¤t->sighand->siglock); | |
388 | return 0; | |
389 | } | |
390 | ||
391 | EXPORT_SYMBOL(allow_signal); | |
392 | ||
393 | int disallow_signal(int sig) | |
394 | { | |
7ed20e1a | 395 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
396 | return -EINVAL; |
397 | ||
398 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 399 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
400 | recalc_sigpending(); |
401 | spin_unlock_irq(¤t->sighand->siglock); | |
402 | return 0; | |
403 | } | |
404 | ||
405 | EXPORT_SYMBOL(disallow_signal); | |
406 | ||
407 | /* | |
408 | * Put all the gunge required to become a kernel thread without | |
409 | * attached user resources in one place where it belongs. | |
410 | */ | |
411 | ||
412 | void daemonize(const char *name, ...) | |
413 | { | |
414 | va_list args; | |
1da177e4 LT |
415 | sigset_t blocked; |
416 | ||
417 | va_start(args, name); | |
418 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
419 | va_end(args); | |
420 | ||
421 | /* | |
422 | * If we were started as result of loading a module, close all of the | |
423 | * user space pages. We don't need them, and if we didn't close them | |
424 | * they would be locked into memory. | |
425 | */ | |
426 | exit_mm(current); | |
83144186 | 427 | /* |
37f08be1 | 428 | * We don't want to get frozen, in case system-wide hibernation |
83144186 RW |
429 | * or suspend transition begins right now. |
430 | */ | |
7b34e428 | 431 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
1da177e4 | 432 | |
8520d7c7 ON |
433 | if (current->nsproxy != &init_nsproxy) { |
434 | get_nsproxy(&init_nsproxy); | |
435 | switch_task_namespaces(current, &init_nsproxy); | |
436 | } | |
297bd42b | 437 | set_special_pids(&init_struct_pid); |
24ec839c | 438 | proc_clear_tty(current); |
1da177e4 LT |
439 | |
440 | /* Block and flush all signals */ | |
441 | sigfillset(&blocked); | |
442 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
443 | flush_signals(current); | |
444 | ||
445 | /* Become as one with the init task */ | |
446 | ||
3e93cd67 | 447 | daemonize_fs_struct(); |
d4c5e41f | 448 | exit_files(current); |
1da177e4 LT |
449 | current->files = init_task.files; |
450 | atomic_inc(¤t->files->count); | |
451 | ||
49d769d5 | 452 | reparent_to_kthreadd(); |
1da177e4 LT |
453 | } |
454 | ||
455 | EXPORT_SYMBOL(daemonize); | |
456 | ||
858119e1 | 457 | static void close_files(struct files_struct * files) |
1da177e4 LT |
458 | { |
459 | int i, j; | |
badf1662 | 460 | struct fdtable *fdt; |
1da177e4 LT |
461 | |
462 | j = 0; | |
4fb3a538 DS |
463 | |
464 | /* | |
465 | * It is safe to dereference the fd table without RCU or | |
466 | * ->file_lock because this is the last reference to the | |
d11c563d | 467 | * files structure. But use RCU to shut RCU-lockdep up. |
4fb3a538 | 468 | */ |
d11c563d | 469 | rcu_read_lock(); |
badf1662 | 470 | fdt = files_fdtable(files); |
d11c563d | 471 | rcu_read_unlock(); |
1da177e4 LT |
472 | for (;;) { |
473 | unsigned long set; | |
474 | i = j * __NFDBITS; | |
bbea9f69 | 475 | if (i >= fdt->max_fds) |
1da177e4 | 476 | break; |
1fd36adc | 477 | set = fdt->open_fds[j++]; |
1da177e4 LT |
478 | while (set) { |
479 | if (set & 1) { | |
badf1662 | 480 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 481 | if (file) { |
1da177e4 | 482 | filp_close(file, files); |
944be0b2 IM |
483 | cond_resched(); |
484 | } | |
1da177e4 LT |
485 | } |
486 | i++; | |
487 | set >>= 1; | |
488 | } | |
489 | } | |
490 | } | |
491 | ||
492 | struct files_struct *get_files_struct(struct task_struct *task) | |
493 | { | |
494 | struct files_struct *files; | |
495 | ||
496 | task_lock(task); | |
497 | files = task->files; | |
498 | if (files) | |
499 | atomic_inc(&files->count); | |
500 | task_unlock(task); | |
501 | ||
502 | return files; | |
503 | } | |
504 | ||
7ad5b3a5 | 505 | void put_files_struct(struct files_struct *files) |
1da177e4 | 506 | { |
badf1662 DS |
507 | struct fdtable *fdt; |
508 | ||
1da177e4 LT |
509 | if (atomic_dec_and_test(&files->count)) { |
510 | close_files(files); | |
511 | /* | |
512 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
513 | * If the fdtable was embedded, pass files for freeing |
514 | * at the end of the RCU grace period. Otherwise, | |
515 | * you can free files immediately. | |
1da177e4 | 516 | */ |
d11c563d | 517 | rcu_read_lock(); |
badf1662 | 518 | fdt = files_fdtable(files); |
4fd45812 | 519 | if (fdt != &files->fdtab) |
ab2af1f5 | 520 | kmem_cache_free(files_cachep, files); |
01b2d93c | 521 | free_fdtable(fdt); |
d11c563d | 522 | rcu_read_unlock(); |
1da177e4 LT |
523 | } |
524 | } | |
525 | ||
3b125388 | 526 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 527 | { |
3b125388 | 528 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
529 | struct files_struct *old; |
530 | ||
531 | old = tsk->files; | |
532 | task_lock(tsk); | |
533 | tsk->files = files; | |
534 | task_unlock(tsk); | |
535 | put_files_struct(old); | |
536 | } | |
3b9b8ab6 | 537 | |
1ec7f1dd | 538 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
539 | { |
540 | struct files_struct * files = tsk->files; | |
541 | ||
542 | if (files) { | |
543 | task_lock(tsk); | |
544 | tsk->files = NULL; | |
545 | task_unlock(tsk); | |
546 | put_files_struct(files); | |
547 | } | |
548 | } | |
549 | ||
cf475ad2 BS |
550 | #ifdef CONFIG_MM_OWNER |
551 | /* | |
733eda7a | 552 | * A task is exiting. If it owned this mm, find a new owner for the mm. |
cf475ad2 | 553 | */ |
cf475ad2 BS |
554 | void mm_update_next_owner(struct mm_struct *mm) |
555 | { | |
556 | struct task_struct *c, *g, *p = current; | |
557 | ||
558 | retry: | |
733eda7a KH |
559 | /* |
560 | * If the exiting or execing task is not the owner, it's | |
561 | * someone else's problem. | |
562 | */ | |
563 | if (mm->owner != p) | |
cf475ad2 | 564 | return; |
733eda7a KH |
565 | /* |
566 | * The current owner is exiting/execing and there are no other | |
567 | * candidates. Do not leave the mm pointing to a possibly | |
568 | * freed task structure. | |
569 | */ | |
570 | if (atomic_read(&mm->mm_users) <= 1) { | |
571 | mm->owner = NULL; | |
572 | return; | |
573 | } | |
cf475ad2 BS |
574 | |
575 | read_lock(&tasklist_lock); | |
576 | /* | |
577 | * Search in the children | |
578 | */ | |
579 | list_for_each_entry(c, &p->children, sibling) { | |
580 | if (c->mm == mm) | |
581 | goto assign_new_owner; | |
582 | } | |
583 | ||
584 | /* | |
585 | * Search in the siblings | |
586 | */ | |
dea33cfd | 587 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
cf475ad2 BS |
588 | if (c->mm == mm) |
589 | goto assign_new_owner; | |
590 | } | |
591 | ||
592 | /* | |
593 | * Search through everything else. We should not get | |
594 | * here often | |
595 | */ | |
596 | do_each_thread(g, c) { | |
597 | if (c->mm == mm) | |
598 | goto assign_new_owner; | |
599 | } while_each_thread(g, c); | |
600 | ||
601 | read_unlock(&tasklist_lock); | |
31a78f23 BS |
602 | /* |
603 | * We found no owner yet mm_users > 1: this implies that we are | |
604 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 605 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 606 | */ |
31a78f23 | 607 | mm->owner = NULL; |
cf475ad2 BS |
608 | return; |
609 | ||
610 | assign_new_owner: | |
611 | BUG_ON(c == p); | |
612 | get_task_struct(c); | |
613 | /* | |
614 | * The task_lock protects c->mm from changing. | |
615 | * We always want mm->owner->mm == mm | |
616 | */ | |
617 | task_lock(c); | |
e5991371 HD |
618 | /* |
619 | * Delay read_unlock() till we have the task_lock() | |
620 | * to ensure that c does not slip away underneath us | |
621 | */ | |
622 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
623 | if (c->mm != mm) { |
624 | task_unlock(c); | |
625 | put_task_struct(c); | |
626 | goto retry; | |
627 | } | |
cf475ad2 BS |
628 | mm->owner = c; |
629 | task_unlock(c); | |
630 | put_task_struct(c); | |
631 | } | |
632 | #endif /* CONFIG_MM_OWNER */ | |
633 | ||
1da177e4 LT |
634 | /* |
635 | * Turn us into a lazy TLB process if we | |
636 | * aren't already.. | |
637 | */ | |
408b664a | 638 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
639 | { |
640 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 641 | struct core_state *core_state; |
1da177e4 LT |
642 | |
643 | mm_release(tsk, mm); | |
644 | if (!mm) | |
645 | return; | |
646 | /* | |
647 | * Serialize with any possible pending coredump. | |
999d9fc1 | 648 | * We must hold mmap_sem around checking core_state |
1da177e4 | 649 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 650 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
651 | * group with ->mm != NULL. |
652 | */ | |
653 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
654 | core_state = mm->core_state; |
655 | if (core_state) { | |
656 | struct core_thread self; | |
1da177e4 | 657 | up_read(&mm->mmap_sem); |
1da177e4 | 658 | |
b564daf8 ON |
659 | self.task = tsk; |
660 | self.next = xchg(&core_state->dumper.next, &self); | |
661 | /* | |
662 | * Implies mb(), the result of xchg() must be visible | |
663 | * to core_state->dumper. | |
664 | */ | |
665 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
666 | complete(&core_state->startup); | |
1da177e4 | 667 | |
a94e2d40 ON |
668 | for (;;) { |
669 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
670 | if (!self.task) /* see coredump_finish() */ | |
671 | break; | |
672 | schedule(); | |
673 | } | |
674 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
675 | down_read(&mm->mmap_sem); |
676 | } | |
677 | atomic_inc(&mm->mm_count); | |
125e1874 | 678 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
679 | /* more a memory barrier than a real lock */ |
680 | task_lock(tsk); | |
681 | tsk->mm = NULL; | |
682 | up_read(&mm->mmap_sem); | |
683 | enter_lazy_tlb(mm, current); | |
684 | task_unlock(tsk); | |
cf475ad2 | 685 | mm_update_next_owner(mm); |
1da177e4 LT |
686 | mmput(mm); |
687 | } | |
688 | ||
1da177e4 | 689 | /* |
ebec18a6 LP |
690 | * When we die, we re-parent all our children, and try to: |
691 | * 1. give them to another thread in our thread group, if such a member exists | |
692 | * 2. give it to the first ancestor process which prctl'd itself as a | |
693 | * child_subreaper for its children (like a service manager) | |
694 | * 3. give it to the init process (PID 1) in our pid namespace | |
1da177e4 | 695 | */ |
950bbabb | 696 | static struct task_struct *find_new_reaper(struct task_struct *father) |
d16e15f5 NK |
697 | __releases(&tasklist_lock) |
698 | __acquires(&tasklist_lock) | |
1da177e4 | 699 | { |
950bbabb ON |
700 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
701 | struct task_struct *thread; | |
1da177e4 | 702 | |
950bbabb ON |
703 | thread = father; |
704 | while_each_thread(father, thread) { | |
705 | if (thread->flags & PF_EXITING) | |
706 | continue; | |
707 | if (unlikely(pid_ns->child_reaper == father)) | |
708 | pid_ns->child_reaper = thread; | |
709 | return thread; | |
710 | } | |
1da177e4 | 711 | |
950bbabb ON |
712 | if (unlikely(pid_ns->child_reaper == father)) { |
713 | write_unlock_irq(&tasklist_lock); | |
397a21f2 DV |
714 | if (unlikely(pid_ns == &init_pid_ns)) { |
715 | panic("Attempted to kill init! exitcode=0x%08x\n", | |
716 | father->signal->group_exit_code ?: | |
717 | father->exit_code); | |
718 | } | |
1da177e4 | 719 | |
950bbabb ON |
720 | zap_pid_ns_processes(pid_ns); |
721 | write_lock_irq(&tasklist_lock); | |
1da177e4 | 722 | /* |
950bbabb ON |
723 | * We can not clear ->child_reaper or leave it alone. |
724 | * There may by stealth EXIT_DEAD tasks on ->children, | |
725 | * forget_original_parent() must move them somewhere. | |
1da177e4 | 726 | */ |
950bbabb | 727 | pid_ns->child_reaper = init_pid_ns.child_reaper; |
ebec18a6 LP |
728 | } else if (father->signal->has_child_subreaper) { |
729 | struct task_struct *reaper; | |
730 | ||
731 | /* | |
732 | * Find the first ancestor marked as child_subreaper. | |
733 | * Note that the code below checks same_thread_group(reaper, | |
734 | * pid_ns->child_reaper). This is what we need to DTRT in a | |
735 | * PID namespace. However we still need the check above, see | |
736 | * http://marc.info/?l=linux-kernel&m=131385460420380 | |
737 | */ | |
738 | for (reaper = father->real_parent; | |
739 | reaper != &init_task; | |
740 | reaper = reaper->real_parent) { | |
741 | if (same_thread_group(reaper, pid_ns->child_reaper)) | |
742 | break; | |
743 | if (!reaper->signal->is_child_subreaper) | |
744 | continue; | |
745 | thread = reaper; | |
746 | do { | |
747 | if (!(thread->flags & PF_EXITING)) | |
748 | return reaper; | |
749 | } while_each_thread(reaper, thread); | |
750 | } | |
1da177e4 | 751 | } |
762a24be | 752 | |
950bbabb ON |
753 | return pid_ns->child_reaper; |
754 | } | |
755 | ||
5dfc80be ON |
756 | /* |
757 | * Any that need to be release_task'd are put on the @dead list. | |
758 | */ | |
9cd80bbb | 759 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
5dfc80be ON |
760 | struct list_head *dead) |
761 | { | |
5dfc80be ON |
762 | list_move_tail(&p->sibling, &p->real_parent->children); |
763 | ||
0976a03e | 764 | if (p->exit_state == EXIT_DEAD) |
5dfc80be ON |
765 | return; |
766 | /* | |
767 | * If this is a threaded reparent there is no need to | |
768 | * notify anyone anything has happened. | |
769 | */ | |
770 | if (same_thread_group(p->real_parent, father)) | |
771 | return; | |
772 | ||
773 | /* We don't want people slaying init. */ | |
774 | p->exit_signal = SIGCHLD; | |
775 | ||
776 | /* If it has exited notify the new parent about this child's death. */ | |
d21142ec | 777 | if (!p->ptrace && |
5dfc80be | 778 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
86773473 | 779 | if (do_notify_parent(p, p->exit_signal)) { |
5dfc80be ON |
780 | p->exit_state = EXIT_DEAD; |
781 | list_move_tail(&p->sibling, dead); | |
782 | } | |
783 | } | |
784 | ||
785 | kill_orphaned_pgrp(p, father); | |
786 | } | |
787 | ||
762a24be | 788 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 789 | { |
950bbabb | 790 | struct task_struct *p, *n, *reaper; |
5dfc80be | 791 | LIST_HEAD(dead_children); |
762a24be ON |
792 | |
793 | write_lock_irq(&tasklist_lock); | |
c7e49c14 ON |
794 | /* |
795 | * Note that exit_ptrace() and find_new_reaper() might | |
796 | * drop tasklist_lock and reacquire it. | |
797 | */ | |
798 | exit_ptrace(father); | |
950bbabb | 799 | reaper = find_new_reaper(father); |
f470021a | 800 | |
03ff1797 | 801 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
9cd80bbb ON |
802 | struct task_struct *t = p; |
803 | do { | |
804 | t->real_parent = reaper; | |
805 | if (t->parent == father) { | |
d21142ec | 806 | BUG_ON(t->ptrace); |
9cd80bbb ON |
807 | t->parent = t->real_parent; |
808 | } | |
809 | if (t->pdeath_signal) | |
810 | group_send_sig_info(t->pdeath_signal, | |
811 | SEND_SIG_NOINFO, t); | |
812 | } while_each_thread(p, t); | |
813 | reparent_leader(father, p, &dead_children); | |
1da177e4 | 814 | } |
762a24be | 815 | write_unlock_irq(&tasklist_lock); |
5dfc80be | 816 | |
762a24be | 817 | BUG_ON(!list_empty(&father->children)); |
762a24be | 818 | |
5dfc80be ON |
819 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
820 | list_del_init(&p->sibling); | |
39c626ae ON |
821 | release_task(p); |
822 | } | |
1da177e4 LT |
823 | } |
824 | ||
825 | /* | |
826 | * Send signals to all our closest relatives so that they know | |
827 | * to properly mourn us.. | |
828 | */ | |
821c7de7 | 829 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 830 | { |
53c8f9f1 | 831 | bool autoreap; |
1da177e4 | 832 | |
1da177e4 LT |
833 | /* |
834 | * This does two things: | |
835 | * | |
836 | * A. Make init inherit all the child processes | |
837 | * B. Check to see if any process groups have become orphaned | |
838 | * as a result of our exiting, and if they have any stopped | |
839 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
840 | */ | |
762a24be | 841 | forget_original_parent(tsk); |
2e4a7072 | 842 | exit_task_namespaces(tsk); |
1da177e4 | 843 | |
762a24be | 844 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
845 | if (group_dead) |
846 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 847 | |
45cdf5cc ON |
848 | if (unlikely(tsk->ptrace)) { |
849 | int sig = thread_group_leader(tsk) && | |
850 | thread_group_empty(tsk) && | |
851 | !ptrace_reparented(tsk) ? | |
852 | tsk->exit_signal : SIGCHLD; | |
853 | autoreap = do_notify_parent(tsk, sig); | |
854 | } else if (thread_group_leader(tsk)) { | |
855 | autoreap = thread_group_empty(tsk) && | |
856 | do_notify_parent(tsk, tsk->exit_signal); | |
857 | } else { | |
858 | autoreap = true; | |
859 | } | |
1da177e4 | 860 | |
53c8f9f1 | 861 | tsk->exit_state = autoreap ? EXIT_DEAD : EXIT_ZOMBIE; |
1da177e4 | 862 | |
9c339168 ON |
863 | /* mt-exec, de_thread() is waiting for group leader */ |
864 | if (unlikely(tsk->signal->notify_count < 0)) | |
6db840fa | 865 | wake_up_process(tsk->signal->group_exit_task); |
1da177e4 LT |
866 | write_unlock_irq(&tasklist_lock); |
867 | ||
1da177e4 | 868 | /* If the process is dead, release it - nobody will wait for it */ |
53c8f9f1 | 869 | if (autoreap) |
1da177e4 | 870 | release_task(tsk); |
1da177e4 LT |
871 | } |
872 | ||
e18eecb8 JD |
873 | #ifdef CONFIG_DEBUG_STACK_USAGE |
874 | static void check_stack_usage(void) | |
875 | { | |
876 | static DEFINE_SPINLOCK(low_water_lock); | |
877 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
878 | unsigned long free; |
879 | ||
7c9f8861 | 880 | free = stack_not_used(current); |
e18eecb8 JD |
881 | |
882 | if (free >= lowest_to_date) | |
883 | return; | |
884 | ||
885 | spin_lock(&low_water_lock); | |
886 | if (free < lowest_to_date) { | |
887 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
888 | "left\n", | |
889 | current->comm, free); | |
890 | lowest_to_date = free; | |
891 | } | |
892 | spin_unlock(&low_water_lock); | |
893 | } | |
894 | #else | |
895 | static inline void check_stack_usage(void) {} | |
896 | #endif | |
897 | ||
9402c95f | 898 | void do_exit(long code) |
1da177e4 LT |
899 | { |
900 | struct task_struct *tsk = current; | |
901 | int group_dead; | |
902 | ||
903 | profile_task_exit(tsk); | |
904 | ||
73c10101 | 905 | WARN_ON(blk_needs_flush_plug(tsk)); |
22e2c507 | 906 | |
1da177e4 LT |
907 | if (unlikely(in_interrupt())) |
908 | panic("Aiee, killing interrupt handler!"); | |
909 | if (unlikely(!tsk->pid)) | |
910 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 911 | |
33dd94ae NE |
912 | /* |
913 | * If do_exit is called because this processes oopsed, it's possible | |
914 | * that get_fs() was left as KERNEL_DS, so reset it to USER_DS before | |
915 | * continuing. Amongst other possible reasons, this is to prevent | |
916 | * mm_release()->clear_child_tid() from writing to a user-controlled | |
917 | * kernel address. | |
918 | */ | |
919 | set_fs(USER_DS); | |
920 | ||
a288eecc | 921 | ptrace_event(PTRACE_EVENT_EXIT, code); |
1da177e4 | 922 | |
e0e81739 DH |
923 | validate_creds_for_do_exit(tsk); |
924 | ||
df164db5 AN |
925 | /* |
926 | * We're taking recursive faults here in do_exit. Safest is to just | |
927 | * leave this task alone and wait for reboot. | |
928 | */ | |
929 | if (unlikely(tsk->flags & PF_EXITING)) { | |
930 | printk(KERN_ALERT | |
931 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
932 | /* |
933 | * We can do this unlocked here. The futex code uses | |
934 | * this flag just to verify whether the pi state | |
935 | * cleanup has been done or not. In the worst case it | |
936 | * loops once more. We pretend that the cleanup was | |
937 | * done as there is no way to return. Either the | |
938 | * OWNER_DIED bit is set by now or we push the blocked | |
939 | * task into the wait for ever nirwana as well. | |
940 | */ | |
941 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
942 | set_current_state(TASK_UNINTERRUPTIBLE); |
943 | schedule(); | |
944 | } | |
945 | ||
d12619b5 | 946 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
947 | /* |
948 | * tsk->flags are checked in the futex code to protect against | |
949 | * an exiting task cleaning up the robust pi futexes. | |
950 | */ | |
d2ee7198 | 951 | smp_mb(); |
1d615482 | 952 | raw_spin_unlock_wait(&tsk->pi_lock); |
1da177e4 | 953 | |
5234ffb9 AG |
954 | exit_irq_thread(); |
955 | ||
1da177e4 LT |
956 | if (unlikely(in_atomic())) |
957 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 958 | current->comm, task_pid_nr(current), |
1da177e4 LT |
959 | preempt_count()); |
960 | ||
961 | acct_update_integrals(tsk); | |
34e55232 | 962 | /* sync mm's RSS info before statistics gathering */ |
a3a2e76c | 963 | if (tsk->mm) |
05af2e10 | 964 | sync_mm_rss(tsk->mm); |
1da177e4 | 965 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 966 | if (group_dead) { |
778e9a9c | 967 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 968 | exit_itimers(tsk->signal); |
1f10206c JP |
969 | if (tsk->mm) |
970 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); | |
c3068951 | 971 | } |
f6ec29a4 | 972 | acct_collect(code, group_dead); |
522ed776 MT |
973 | if (group_dead) |
974 | tty_audit_exit(); | |
a4ff8dba | 975 | audit_free(tsk); |
115085ea | 976 | |
f2ab6d88 | 977 | tsk->exit_code = code; |
115085ea | 978 | taskstats_exit(tsk, group_dead); |
c757249a | 979 | |
1da177e4 LT |
980 | exit_mm(tsk); |
981 | ||
0e464814 | 982 | if (group_dead) |
f6ec29a4 | 983 | acct_process(); |
0a16b607 MD |
984 | trace_sched_process_exit(tsk); |
985 | ||
1da177e4 | 986 | exit_sem(tsk); |
b34a6b1d | 987 | exit_shm(tsk); |
1ec7f1dd AV |
988 | exit_files(tsk); |
989 | exit_fs(tsk); | |
e18eecb8 | 990 | check_stack_usage(); |
1da177e4 | 991 | exit_thread(); |
0b3fcf17 SE |
992 | |
993 | /* | |
994 | * Flush inherited counters to the parent - before the parent | |
995 | * gets woken up by child-exit notifications. | |
996 | * | |
997 | * because of cgroup mode, must be called before cgroup_exit() | |
998 | */ | |
999 | perf_event_exit_task(tsk); | |
1000 | ||
b4f48b63 | 1001 | cgroup_exit(tsk, 1); |
1da177e4 | 1002 | |
5ec93d11 | 1003 | if (group_dead) |
1da177e4 LT |
1004 | disassociate_ctty(1); |
1005 | ||
a1261f54 | 1006 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 | 1007 | |
9f46080c | 1008 | proc_exit_connector(tsk); |
33b2fb30 | 1009 | |
24f1e32c FW |
1010 | /* |
1011 | * FIXME: do that only when needed, using sched_exit tracepoint | |
1012 | */ | |
bf26c018 | 1013 | ptrace_put_breakpoints(tsk); |
33b2fb30 | 1014 | |
821c7de7 | 1015 | exit_notify(tsk, group_dead); |
1da177e4 | 1016 | #ifdef CONFIG_NUMA |
c0ff7453 | 1017 | task_lock(tsk); |
f0be3d32 | 1018 | mpol_put(tsk->mempolicy); |
1da177e4 | 1019 | tsk->mempolicy = NULL; |
c0ff7453 | 1020 | task_unlock(tsk); |
1da177e4 | 1021 | #endif |
42b2dd0a | 1022 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
1023 | if (unlikely(current->pi_state_cache)) |
1024 | kfree(current->pi_state_cache); | |
42b2dd0a | 1025 | #endif |
de5097c2 | 1026 | /* |
9a11b49a | 1027 | * Make sure we are holding no locks: |
de5097c2 | 1028 | */ |
9a11b49a | 1029 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1030 | /* |
1031 | * We can do this unlocked here. The futex code uses this flag | |
1032 | * just to verify whether the pi state cleanup has been done | |
1033 | * or not. In the worst case it loops once more. | |
1034 | */ | |
1035 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1036 | |
afc847b7 | 1037 | if (tsk->io_context) |
b69f2292 | 1038 | exit_io_context(tsk); |
afc847b7 | 1039 | |
b92ce558 JA |
1040 | if (tsk->splice_pipe) |
1041 | __free_pipe_info(tsk->splice_pipe); | |
1042 | ||
e0e81739 DH |
1043 | validate_creds_for_do_exit(tsk); |
1044 | ||
7407251a | 1045 | preempt_disable(); |
54848d73 WF |
1046 | if (tsk->nr_dirtied) |
1047 | __this_cpu_add(dirty_throttle_leaks, tsk->nr_dirtied); | |
f41d911f | 1048 | exit_rcu(); |
b5740f4b YG |
1049 | |
1050 | /* | |
1051 | * The setting of TASK_RUNNING by try_to_wake_up() may be delayed | |
1052 | * when the following two conditions become true. | |
1053 | * - There is race condition of mmap_sem (It is acquired by | |
1054 | * exit_mm()), and | |
1055 | * - SMI occurs before setting TASK_RUNINNG. | |
1056 | * (or hypervisor of virtual machine switches to other guest) | |
1057 | * As a result, we may become TASK_RUNNING after becoming TASK_DEAD | |
1058 | * | |
1059 | * To avoid it, we have to wait for releasing tsk->pi_lock which | |
1060 | * is held by try_to_wake_up() | |
1061 | */ | |
1062 | smp_mb(); | |
1063 | raw_spin_unlock_wait(&tsk->pi_lock); | |
1064 | ||
55a101f8 | 1065 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1066 | tsk->state = TASK_DEAD; |
a585042f | 1067 | tsk->flags |= PF_NOFREEZE; /* tell freezer to ignore us */ |
1da177e4 LT |
1068 | schedule(); |
1069 | BUG(); | |
1070 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1071 | for (;;) |
1072 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1073 | } |
1074 | ||
012914da RA |
1075 | EXPORT_SYMBOL_GPL(do_exit); |
1076 | ||
9402c95f | 1077 | void complete_and_exit(struct completion *comp, long code) |
1da177e4 LT |
1078 | { |
1079 | if (comp) | |
1080 | complete(comp); | |
55a101f8 | 1081 | |
1da177e4 LT |
1082 | do_exit(code); |
1083 | } | |
1084 | ||
1085 | EXPORT_SYMBOL(complete_and_exit); | |
1086 | ||
754fe8d2 | 1087 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
1088 | { |
1089 | do_exit((error_code&0xff)<<8); | |
1090 | } | |
1091 | ||
1da177e4 LT |
1092 | /* |
1093 | * Take down every thread in the group. This is called by fatal signals | |
1094 | * as well as by sys_exit_group (below). | |
1095 | */ | |
9402c95f | 1096 | void |
1da177e4 LT |
1097 | do_group_exit(int exit_code) |
1098 | { | |
bfc4b089 ON |
1099 | struct signal_struct *sig = current->signal; |
1100 | ||
1da177e4 LT |
1101 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1102 | ||
bfc4b089 ON |
1103 | if (signal_group_exit(sig)) |
1104 | exit_code = sig->group_exit_code; | |
1da177e4 | 1105 | else if (!thread_group_empty(current)) { |
1da177e4 | 1106 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1107 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1108 | if (signal_group_exit(sig)) |
1da177e4 LT |
1109 | /* Another thread got here before we took the lock. */ |
1110 | exit_code = sig->group_exit_code; | |
1111 | else { | |
1da177e4 | 1112 | sig->group_exit_code = exit_code; |
ed5d2cac | 1113 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1114 | zap_other_threads(current); |
1115 | } | |
1116 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1117 | } |
1118 | ||
1119 | do_exit(exit_code); | |
1120 | /* NOTREACHED */ | |
1121 | } | |
1122 | ||
1123 | /* | |
1124 | * this kills every thread in the thread group. Note that any externally | |
1125 | * wait4()-ing process will get the correct exit code - even if this | |
1126 | * thread is not the thread group leader. | |
1127 | */ | |
754fe8d2 | 1128 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
1129 | { |
1130 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
1131 | /* NOTREACHED */ |
1132 | return 0; | |
1da177e4 LT |
1133 | } |
1134 | ||
9e8ae01d ON |
1135 | struct wait_opts { |
1136 | enum pid_type wo_type; | |
9e8ae01d | 1137 | int wo_flags; |
e1eb1ebc | 1138 | struct pid *wo_pid; |
9e8ae01d ON |
1139 | |
1140 | struct siginfo __user *wo_info; | |
1141 | int __user *wo_stat; | |
1142 | struct rusage __user *wo_rusage; | |
1143 | ||
0b7570e7 | 1144 | wait_queue_t child_wait; |
9e8ae01d ON |
1145 | int notask_error; |
1146 | }; | |
1147 | ||
989264f4 ON |
1148 | static inline |
1149 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) | |
161550d7 | 1150 | { |
989264f4 ON |
1151 | if (type != PIDTYPE_PID) |
1152 | task = task->group_leader; | |
1153 | return task->pids[type].pid; | |
161550d7 EB |
1154 | } |
1155 | ||
989264f4 | 1156 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
1da177e4 | 1157 | { |
5c01ba49 ON |
1158 | return wo->wo_type == PIDTYPE_MAX || |
1159 | task_pid_type(p, wo->wo_type) == wo->wo_pid; | |
1160 | } | |
1da177e4 | 1161 | |
5c01ba49 ON |
1162 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
1163 | { | |
1164 | if (!eligible_pid(wo, p)) | |
1165 | return 0; | |
1da177e4 LT |
1166 | /* Wait for all children (clone and not) if __WALL is set; |
1167 | * otherwise, wait for clone children *only* if __WCLONE is | |
1168 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1169 | * A "clone" child here is one that reports to its parent | |
1170 | * using a signal other than SIGCHLD.) */ | |
9e8ae01d ON |
1171 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
1172 | && !(wo->wo_flags & __WALL)) | |
1da177e4 | 1173 | return 0; |
1da177e4 | 1174 | |
14dd0b81 | 1175 | return 1; |
1da177e4 LT |
1176 | } |
1177 | ||
9e8ae01d ON |
1178 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
1179 | pid_t pid, uid_t uid, int why, int status) | |
1da177e4 | 1180 | { |
9e8ae01d ON |
1181 | struct siginfo __user *infop; |
1182 | int retval = wo->wo_rusage | |
1183 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
36c8b586 | 1184 | |
1da177e4 | 1185 | put_task_struct(p); |
9e8ae01d | 1186 | infop = wo->wo_info; |
b6fe2d11 VM |
1187 | if (infop) { |
1188 | if (!retval) | |
1189 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1190 | if (!retval) | |
1191 | retval = put_user(0, &infop->si_errno); | |
1192 | if (!retval) | |
1193 | retval = put_user((short)why, &infop->si_code); | |
1194 | if (!retval) | |
1195 | retval = put_user(pid, &infop->si_pid); | |
1196 | if (!retval) | |
1197 | retval = put_user(uid, &infop->si_uid); | |
1198 | if (!retval) | |
1199 | retval = put_user(status, &infop->si_status); | |
1200 | } | |
1da177e4 LT |
1201 | if (!retval) |
1202 | retval = pid; | |
1203 | return retval; | |
1204 | } | |
1205 | ||
1206 | /* | |
1207 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1208 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1209 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1210 | * released the lock and the system call should return. | |
1211 | */ | |
9e8ae01d | 1212 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1213 | { |
1214 | unsigned long state; | |
2f4e6e2a | 1215 | int retval, status, traced; |
6c5f3e7b | 1216 | pid_t pid = task_pid_vnr(p); |
a29c33f4 | 1217 | uid_t uid = from_kuid_munged(current_user_ns(), __task_cred(p)->uid); |
9e8ae01d | 1218 | struct siginfo __user *infop; |
1da177e4 | 1219 | |
9e8ae01d | 1220 | if (!likely(wo->wo_flags & WEXITED)) |
98abed02 RM |
1221 | return 0; |
1222 | ||
9e8ae01d | 1223 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
1da177e4 | 1224 | int exit_code = p->exit_code; |
f3abd4f9 | 1225 | int why; |
1da177e4 | 1226 | |
1da177e4 LT |
1227 | get_task_struct(p); |
1228 | read_unlock(&tasklist_lock); | |
1229 | if ((exit_code & 0x7f) == 0) { | |
1230 | why = CLD_EXITED; | |
1231 | status = exit_code >> 8; | |
1232 | } else { | |
1233 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1234 | status = exit_code & 0x7f; | |
1235 | } | |
9e8ae01d | 1236 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
1da177e4 LT |
1237 | } |
1238 | ||
1239 | /* | |
1240 | * Try to move the task's state to DEAD | |
1241 | * only one thread is allowed to do this: | |
1242 | */ | |
1243 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1244 | if (state != EXIT_ZOMBIE) { | |
1245 | BUG_ON(state != EXIT_DEAD); | |
1246 | return 0; | |
1247 | } | |
1da177e4 | 1248 | |
53b6f9fb | 1249 | traced = ptrace_reparented(p); |
befca967 ON |
1250 | /* |
1251 | * It can be ptraced but not reparented, check | |
e550f14d | 1252 | * thread_group_leader() to filter out sub-threads. |
befca967 | 1253 | */ |
e550f14d | 1254 | if (likely(!traced) && thread_group_leader(p)) { |
3795e161 JJ |
1255 | struct signal_struct *psig; |
1256 | struct signal_struct *sig; | |
1f10206c | 1257 | unsigned long maxrss; |
0cf55e1e | 1258 | cputime_t tgutime, tgstime; |
3795e161 | 1259 | |
1da177e4 LT |
1260 | /* |
1261 | * The resource counters for the group leader are in its | |
1262 | * own task_struct. Those for dead threads in the group | |
1263 | * are in its signal_struct, as are those for the child | |
1264 | * processes it has previously reaped. All these | |
1265 | * accumulate in the parent's signal_struct c* fields. | |
1266 | * | |
1267 | * We don't bother to take a lock here to protect these | |
1268 | * p->signal fields, because they are only touched by | |
1269 | * __exit_signal, which runs with tasklist_lock | |
1270 | * write-locked anyway, and so is excluded here. We do | |
d1e98f42 | 1271 | * need to protect the access to parent->signal fields, |
1da177e4 LT |
1272 | * as other threads in the parent group can be right |
1273 | * here reaping other children at the same time. | |
0cf55e1e HS |
1274 | * |
1275 | * We use thread_group_times() to get times for the thread | |
1276 | * group, which consolidates times for all threads in the | |
1277 | * group including the group leader. | |
1da177e4 | 1278 | */ |
0cf55e1e | 1279 | thread_group_times(p, &tgutime, &tgstime); |
d1e98f42 ON |
1280 | spin_lock_irq(&p->real_parent->sighand->siglock); |
1281 | psig = p->real_parent->signal; | |
3795e161 | 1282 | sig = p->signal; |
64861634 MS |
1283 | psig->cutime += tgutime + sig->cutime; |
1284 | psig->cstime += tgstime + sig->cstime; | |
1285 | psig->cgtime += p->gtime + sig->gtime + sig->cgtime; | |
3795e161 JJ |
1286 | psig->cmin_flt += |
1287 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1288 | psig->cmaj_flt += | |
1289 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1290 | psig->cnvcsw += | |
1291 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1292 | psig->cnivcsw += | |
1293 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1294 | psig->cinblock += |
1295 | task_io_get_inblock(p) + | |
1296 | sig->inblock + sig->cinblock; | |
1297 | psig->coublock += | |
1298 | task_io_get_oublock(p) + | |
1299 | sig->oublock + sig->coublock; | |
1f10206c JP |
1300 | maxrss = max(sig->maxrss, sig->cmaxrss); |
1301 | if (psig->cmaxrss < maxrss) | |
1302 | psig->cmaxrss = maxrss; | |
5995477a AR |
1303 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1304 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
d1e98f42 | 1305 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
1da177e4 LT |
1306 | } |
1307 | ||
1308 | /* | |
1309 | * Now we are sure this task is interesting, and no other | |
1310 | * thread can reap it because we set its state to EXIT_DEAD. | |
1311 | */ | |
1312 | read_unlock(&tasklist_lock); | |
1313 | ||
9e8ae01d ON |
1314 | retval = wo->wo_rusage |
1315 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 LT |
1316 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
1317 | ? p->signal->group_exit_code : p->exit_code; | |
9e8ae01d ON |
1318 | if (!retval && wo->wo_stat) |
1319 | retval = put_user(status, wo->wo_stat); | |
1320 | ||
1321 | infop = wo->wo_info; | |
1da177e4 LT |
1322 | if (!retval && infop) |
1323 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1324 | if (!retval && infop) | |
1325 | retval = put_user(0, &infop->si_errno); | |
1326 | if (!retval && infop) { | |
1327 | int why; | |
1328 | ||
1329 | if ((status & 0x7f) == 0) { | |
1330 | why = CLD_EXITED; | |
1331 | status >>= 8; | |
1332 | } else { | |
1333 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1334 | status &= 0x7f; | |
1335 | } | |
1336 | retval = put_user((short)why, &infop->si_code); | |
1337 | if (!retval) | |
1338 | retval = put_user(status, &infop->si_status); | |
1339 | } | |
1340 | if (!retval && infop) | |
3a515e4a | 1341 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1342 | if (!retval && infop) |
c69e8d9c | 1343 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1344 | if (!retval) |
3a515e4a | 1345 | retval = pid; |
2f4e6e2a ON |
1346 | |
1347 | if (traced) { | |
1da177e4 | 1348 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1349 | /* We dropped tasklist, ptracer could die and untrace */ |
1350 | ptrace_unlink(p); | |
1351 | /* | |
86773473 ON |
1352 | * If this is not a sub-thread, notify the parent. |
1353 | * If parent wants a zombie, don't release it now. | |
2f4e6e2a | 1354 | */ |
86773473 ON |
1355 | if (thread_group_leader(p) && |
1356 | !do_notify_parent(p, p->exit_signal)) { | |
1357 | p->exit_state = EXIT_ZOMBIE; | |
1358 | p = NULL; | |
1da177e4 LT |
1359 | } |
1360 | write_unlock_irq(&tasklist_lock); | |
1361 | } | |
1362 | if (p != NULL) | |
1363 | release_task(p); | |
2f4e6e2a | 1364 | |
1da177e4 LT |
1365 | return retval; |
1366 | } | |
1367 | ||
90bc8d8b ON |
1368 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
1369 | { | |
1370 | if (ptrace) { | |
544b2c91 TH |
1371 | if (task_is_stopped_or_traced(p) && |
1372 | !(p->jobctl & JOBCTL_LISTENING)) | |
90bc8d8b ON |
1373 | return &p->exit_code; |
1374 | } else { | |
1375 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | |
1376 | return &p->signal->group_exit_code; | |
1377 | } | |
1378 | return NULL; | |
1379 | } | |
1380 | ||
19e27463 TH |
1381 | /** |
1382 | * wait_task_stopped - Wait for %TASK_STOPPED or %TASK_TRACED | |
1383 | * @wo: wait options | |
1384 | * @ptrace: is the wait for ptrace | |
1385 | * @p: task to wait for | |
1386 | * | |
1387 | * Handle sys_wait4() work for %p in state %TASK_STOPPED or %TASK_TRACED. | |
1388 | * | |
1389 | * CONTEXT: | |
1390 | * read_lock(&tasklist_lock), which is released if return value is | |
1391 | * non-zero. Also, grabs and releases @p->sighand->siglock. | |
1392 | * | |
1393 | * RETURNS: | |
1394 | * 0 if wait condition didn't exist and search for other wait conditions | |
1395 | * should continue. Non-zero return, -errno on failure and @p's pid on | |
1396 | * success, implies that tasklist_lock is released and wait condition | |
1397 | * search should terminate. | |
1da177e4 | 1398 | */ |
9e8ae01d ON |
1399 | static int wait_task_stopped(struct wait_opts *wo, |
1400 | int ptrace, struct task_struct *p) | |
1da177e4 | 1401 | { |
9e8ae01d | 1402 | struct siginfo __user *infop; |
90bc8d8b | 1403 | int retval, exit_code, *p_code, why; |
ee7c82da | 1404 | uid_t uid = 0; /* unneeded, required by compiler */ |
c8950783 | 1405 | pid_t pid; |
1da177e4 | 1406 | |
47918025 ON |
1407 | /* |
1408 | * Traditionally we see ptrace'd stopped tasks regardless of options. | |
1409 | */ | |
9e8ae01d | 1410 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
98abed02 RM |
1411 | return 0; |
1412 | ||
19e27463 TH |
1413 | if (!task_stopped_code(p, ptrace)) |
1414 | return 0; | |
1415 | ||
ee7c82da ON |
1416 | exit_code = 0; |
1417 | spin_lock_irq(&p->sighand->siglock); | |
1418 | ||
90bc8d8b ON |
1419 | p_code = task_stopped_code(p, ptrace); |
1420 | if (unlikely(!p_code)) | |
ee7c82da ON |
1421 | goto unlock_sig; |
1422 | ||
90bc8d8b | 1423 | exit_code = *p_code; |
ee7c82da ON |
1424 | if (!exit_code) |
1425 | goto unlock_sig; | |
1426 | ||
9e8ae01d | 1427 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
90bc8d8b | 1428 | *p_code = 0; |
ee7c82da | 1429 | |
8ca937a6 | 1430 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
ee7c82da ON |
1431 | unlock_sig: |
1432 | spin_unlock_irq(&p->sighand->siglock); | |
1433 | if (!exit_code) | |
1da177e4 LT |
1434 | return 0; |
1435 | ||
1436 | /* | |
1437 | * Now we are pretty sure this task is interesting. | |
1438 | * Make sure it doesn't get reaped out from under us while we | |
1439 | * give up the lock and then examine it below. We don't want to | |
1440 | * keep holding onto the tasklist_lock while we call getrusage and | |
1441 | * possibly take page faults for user memory. | |
1442 | */ | |
1443 | get_task_struct(p); | |
6c5f3e7b | 1444 | pid = task_pid_vnr(p); |
f470021a | 1445 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1446 | read_unlock(&tasklist_lock); |
1447 | ||
9e8ae01d ON |
1448 | if (unlikely(wo->wo_flags & WNOWAIT)) |
1449 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); | |
1450 | ||
1451 | retval = wo->wo_rusage | |
1452 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1453 | if (!retval && wo->wo_stat) | |
1454 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); | |
1da177e4 | 1455 | |
9e8ae01d | 1456 | infop = wo->wo_info; |
1da177e4 LT |
1457 | if (!retval && infop) |
1458 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1459 | if (!retval && infop) | |
1460 | retval = put_user(0, &infop->si_errno); | |
1461 | if (!retval && infop) | |
6efcae46 | 1462 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1463 | if (!retval && infop) |
1464 | retval = put_user(exit_code, &infop->si_status); | |
1465 | if (!retval && infop) | |
c8950783 | 1466 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1467 | if (!retval && infop) |
ee7c82da | 1468 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1469 | if (!retval) |
c8950783 | 1470 | retval = pid; |
1da177e4 LT |
1471 | put_task_struct(p); |
1472 | ||
1473 | BUG_ON(!retval); | |
1474 | return retval; | |
1475 | } | |
1476 | ||
1477 | /* | |
1478 | * Handle do_wait work for one task in a live, non-stopped state. | |
1479 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1480 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1481 | * released the lock and the system call should return. | |
1482 | */ | |
9e8ae01d | 1483 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
1da177e4 LT |
1484 | { |
1485 | int retval; | |
1486 | pid_t pid; | |
1487 | uid_t uid; | |
1488 | ||
9e8ae01d | 1489 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
98abed02 RM |
1490 | return 0; |
1491 | ||
1da177e4 LT |
1492 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1493 | return 0; | |
1494 | ||
1495 | spin_lock_irq(&p->sighand->siglock); | |
1496 | /* Re-check with the lock held. */ | |
1497 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1498 | spin_unlock_irq(&p->sighand->siglock); | |
1499 | return 0; | |
1500 | } | |
9e8ae01d | 1501 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
1da177e4 | 1502 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
8ca937a6 | 1503 | uid = from_kuid_munged(current_user_ns(), task_uid(p)); |
1da177e4 LT |
1504 | spin_unlock_irq(&p->sighand->siglock); |
1505 | ||
6c5f3e7b | 1506 | pid = task_pid_vnr(p); |
1da177e4 LT |
1507 | get_task_struct(p); |
1508 | read_unlock(&tasklist_lock); | |
1509 | ||
9e8ae01d ON |
1510 | if (!wo->wo_info) { |
1511 | retval = wo->wo_rusage | |
1512 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; | |
1da177e4 | 1513 | put_task_struct(p); |
9e8ae01d ON |
1514 | if (!retval && wo->wo_stat) |
1515 | retval = put_user(0xffff, wo->wo_stat); | |
1da177e4 | 1516 | if (!retval) |
3a515e4a | 1517 | retval = pid; |
1da177e4 | 1518 | } else { |
9e8ae01d ON |
1519 | retval = wait_noreap_copyout(wo, p, pid, uid, |
1520 | CLD_CONTINUED, SIGCONT); | |
1da177e4 LT |
1521 | BUG_ON(retval == 0); |
1522 | } | |
1523 | ||
1524 | return retval; | |
1525 | } | |
1526 | ||
98abed02 RM |
1527 | /* |
1528 | * Consider @p for a wait by @parent. | |
1529 | * | |
9e8ae01d | 1530 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1531 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1532 | * Returns zero if the search for a child should continue; | |
9e8ae01d | 1533 | * then ->notask_error is 0 if @p is an eligible child, |
14dd0b81 | 1534 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1535 | */ |
b6e763f0 ON |
1536 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
1537 | struct task_struct *p) | |
98abed02 | 1538 | { |
9e8ae01d | 1539 | int ret = eligible_child(wo, p); |
14dd0b81 | 1540 | if (!ret) |
98abed02 RM |
1541 | return ret; |
1542 | ||
a2322e1d | 1543 | ret = security_task_wait(p); |
14dd0b81 RM |
1544 | if (unlikely(ret < 0)) { |
1545 | /* | |
1546 | * If we have not yet seen any eligible child, | |
1547 | * then let this error code replace -ECHILD. | |
1548 | * A permission error will give the user a clue | |
1549 | * to look for security policy problems, rather | |
1550 | * than for mysterious wait bugs. | |
1551 | */ | |
9e8ae01d ON |
1552 | if (wo->notask_error) |
1553 | wo->notask_error = ret; | |
78a3d9d5 | 1554 | return 0; |
14dd0b81 RM |
1555 | } |
1556 | ||
823b018e | 1557 | /* dead body doesn't have much to contribute */ |
50b8d257 ON |
1558 | if (unlikely(p->exit_state == EXIT_DEAD)) { |
1559 | /* | |
1560 | * But do not ignore this task until the tracer does | |
1561 | * wait_task_zombie()->do_notify_parent(). | |
1562 | */ | |
1563 | if (likely(!ptrace) && unlikely(ptrace_reparented(p))) | |
1564 | wo->notask_error = 0; | |
823b018e | 1565 | return 0; |
50b8d257 | 1566 | } |
823b018e | 1567 | |
45cb24a1 TH |
1568 | /* slay zombie? */ |
1569 | if (p->exit_state == EXIT_ZOMBIE) { | |
f470021a | 1570 | /* |
45cb24a1 TH |
1571 | * A zombie ptracee is only visible to its ptracer. |
1572 | * Notification and reaping will be cascaded to the real | |
1573 | * parent when the ptracer detaches. | |
f470021a | 1574 | */ |
d21142ec | 1575 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
45cb24a1 TH |
1576 | /* it will become visible, clear notask_error */ |
1577 | wo->notask_error = 0; | |
1578 | return 0; | |
1579 | } | |
f470021a | 1580 | |
9b84cca2 TH |
1581 | /* we don't reap group leaders with subthreads */ |
1582 | if (!delay_group_leader(p)) | |
1583 | return wait_task_zombie(wo, p); | |
98abed02 | 1584 | |
f470021a | 1585 | /* |
9b84cca2 TH |
1586 | * Allow access to stopped/continued state via zombie by |
1587 | * falling through. Clearing of notask_error is complex. | |
1588 | * | |
1589 | * When !@ptrace: | |
1590 | * | |
1591 | * If WEXITED is set, notask_error should naturally be | |
1592 | * cleared. If not, subset of WSTOPPED|WCONTINUED is set, | |
1593 | * so, if there are live subthreads, there are events to | |
1594 | * wait for. If all subthreads are dead, it's still safe | |
1595 | * to clear - this function will be called again in finite | |
1596 | * amount time once all the subthreads are released and | |
1597 | * will then return without clearing. | |
1598 | * | |
1599 | * When @ptrace: | |
1600 | * | |
1601 | * Stopped state is per-task and thus can't change once the | |
1602 | * target task dies. Only continued and exited can happen. | |
1603 | * Clear notask_error if WCONTINUED | WEXITED. | |
1604 | */ | |
1605 | if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED))) | |
1606 | wo->notask_error = 0; | |
1607 | } else { | |
45cb24a1 TH |
1608 | /* |
1609 | * If @p is ptraced by a task in its real parent's group, | |
1610 | * hide group stop/continued state when looking at @p as | |
1611 | * the real parent; otherwise, a single stop can be | |
1612 | * reported twice as group and ptrace stops. | |
1613 | * | |
1614 | * If a ptracer wants to distinguish the two events for its | |
1615 | * own children, it should create a separate process which | |
1616 | * takes the role of real parent. | |
1617 | */ | |
479bf98c | 1618 | if (likely(!ptrace) && p->ptrace && !ptrace_reparented(p)) |
45cb24a1 TH |
1619 | return 0; |
1620 | ||
9b84cca2 TH |
1621 | /* |
1622 | * @p is alive and it's gonna stop, continue or exit, so | |
1623 | * there always is something to wait for. | |
f470021a | 1624 | */ |
9e8ae01d | 1625 | wo->notask_error = 0; |
f470021a RM |
1626 | } |
1627 | ||
98abed02 | 1628 | /* |
45cb24a1 TH |
1629 | * Wait for stopped. Depending on @ptrace, different stopped state |
1630 | * is used and the two don't interact with each other. | |
98abed02 | 1631 | */ |
19e27463 TH |
1632 | ret = wait_task_stopped(wo, ptrace, p); |
1633 | if (ret) | |
1634 | return ret; | |
98abed02 RM |
1635 | |
1636 | /* | |
45cb24a1 TH |
1637 | * Wait for continued. There's only one continued state and the |
1638 | * ptracer can consume it which can confuse the real parent. Don't | |
1639 | * use WCONTINUED from ptracer. You don't need or want it. | |
98abed02 | 1640 | */ |
9e8ae01d | 1641 | return wait_task_continued(wo, p); |
98abed02 RM |
1642 | } |
1643 | ||
1644 | /* | |
1645 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1646 | * | |
9e8ae01d | 1647 | * -ECHILD should be in ->notask_error before the first call. |
98abed02 RM |
1648 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
1649 | * Returns zero if the search for a child should continue; then | |
9e8ae01d | 1650 | * ->notask_error is 0 if there were any eligible children, |
14dd0b81 | 1651 | * or another error from security_task_wait(), or still -ECHILD. |
98abed02 | 1652 | */ |
9e8ae01d | 1653 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1654 | { |
1655 | struct task_struct *p; | |
1656 | ||
1657 | list_for_each_entry(p, &tsk->children, sibling) { | |
9cd80bbb ON |
1658 | int ret = wait_consider_task(wo, 0, p); |
1659 | if (ret) | |
1660 | return ret; | |
98abed02 RM |
1661 | } |
1662 | ||
1663 | return 0; | |
1664 | } | |
1665 | ||
9e8ae01d | 1666 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
98abed02 RM |
1667 | { |
1668 | struct task_struct *p; | |
1669 | ||
f470021a | 1670 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
b6e763f0 | 1671 | int ret = wait_consider_task(wo, 1, p); |
f470021a | 1672 | if (ret) |
98abed02 | 1673 | return ret; |
98abed02 RM |
1674 | } |
1675 | ||
1676 | return 0; | |
1677 | } | |
1678 | ||
0b7570e7 ON |
1679 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
1680 | int sync, void *key) | |
1681 | { | |
1682 | struct wait_opts *wo = container_of(wait, struct wait_opts, | |
1683 | child_wait); | |
1684 | struct task_struct *p = key; | |
1685 | ||
5c01ba49 | 1686 | if (!eligible_pid(wo, p)) |
0b7570e7 ON |
1687 | return 0; |
1688 | ||
b4fe5182 ON |
1689 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
1690 | return 0; | |
1691 | ||
0b7570e7 ON |
1692 | return default_wake_function(wait, mode, sync, key); |
1693 | } | |
1694 | ||
a7f0765e ON |
1695 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
1696 | { | |
0b7570e7 ON |
1697 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
1698 | TASK_INTERRUPTIBLE, 1, p); | |
a7f0765e ON |
1699 | } |
1700 | ||
9e8ae01d | 1701 | static long do_wait(struct wait_opts *wo) |
1da177e4 | 1702 | { |
1da177e4 | 1703 | struct task_struct *tsk; |
98abed02 | 1704 | int retval; |
1da177e4 | 1705 | |
9e8ae01d | 1706 | trace_sched_process_wait(wo->wo_pid); |
0a16b607 | 1707 | |
0b7570e7 ON |
1708 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
1709 | wo->child_wait.private = current; | |
1710 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); | |
1da177e4 | 1711 | repeat: |
98abed02 RM |
1712 | /* |
1713 | * If there is nothing that can match our critiera just get out. | |
9e8ae01d ON |
1714 | * We will clear ->notask_error to zero if we see any child that |
1715 | * might later match our criteria, even if we are not able to reap | |
1716 | * it yet. | |
98abed02 | 1717 | */ |
64a16caf | 1718 | wo->notask_error = -ECHILD; |
9e8ae01d ON |
1719 | if ((wo->wo_type < PIDTYPE_MAX) && |
1720 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) | |
64a16caf | 1721 | goto notask; |
161550d7 | 1722 | |
f95d39d1 | 1723 | set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
1724 | read_lock(&tasklist_lock); |
1725 | tsk = current; | |
1726 | do { | |
64a16caf ON |
1727 | retval = do_wait_thread(wo, tsk); |
1728 | if (retval) | |
1729 | goto end; | |
9e8ae01d | 1730 | |
64a16caf ON |
1731 | retval = ptrace_do_wait(wo, tsk); |
1732 | if (retval) | |
98abed02 | 1733 | goto end; |
98abed02 | 1734 | |
9e8ae01d | 1735 | if (wo->wo_flags & __WNOTHREAD) |
1da177e4 | 1736 | break; |
a3f6dfb7 | 1737 | } while_each_thread(current, tsk); |
1da177e4 | 1738 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1739 | |
64a16caf | 1740 | notask: |
9e8ae01d ON |
1741 | retval = wo->notask_error; |
1742 | if (!retval && !(wo->wo_flags & WNOHANG)) { | |
1da177e4 | 1743 | retval = -ERESTARTSYS; |
98abed02 RM |
1744 | if (!signal_pending(current)) { |
1745 | schedule(); | |
1746 | goto repeat; | |
1747 | } | |
1da177e4 | 1748 | } |
1da177e4 | 1749 | end: |
f95d39d1 | 1750 | __set_current_state(TASK_RUNNING); |
0b7570e7 | 1751 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
1da177e4 LT |
1752 | return retval; |
1753 | } | |
1754 | ||
17da2bd9 HC |
1755 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1756 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1757 | { |
9e8ae01d | 1758 | struct wait_opts wo; |
161550d7 EB |
1759 | struct pid *pid = NULL; |
1760 | enum pid_type type; | |
1da177e4 LT |
1761 | long ret; |
1762 | ||
1763 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1764 | return -EINVAL; | |
1765 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1766 | return -EINVAL; | |
1767 | ||
1768 | switch (which) { | |
1769 | case P_ALL: | |
161550d7 | 1770 | type = PIDTYPE_MAX; |
1da177e4 LT |
1771 | break; |
1772 | case P_PID: | |
161550d7 EB |
1773 | type = PIDTYPE_PID; |
1774 | if (upid <= 0) | |
1da177e4 LT |
1775 | return -EINVAL; |
1776 | break; | |
1777 | case P_PGID: | |
161550d7 EB |
1778 | type = PIDTYPE_PGID; |
1779 | if (upid <= 0) | |
1da177e4 | 1780 | return -EINVAL; |
1da177e4 LT |
1781 | break; |
1782 | default: | |
1783 | return -EINVAL; | |
1784 | } | |
1785 | ||
161550d7 EB |
1786 | if (type < PIDTYPE_MAX) |
1787 | pid = find_get_pid(upid); | |
9e8ae01d ON |
1788 | |
1789 | wo.wo_type = type; | |
1790 | wo.wo_pid = pid; | |
1791 | wo.wo_flags = options; | |
1792 | wo.wo_info = infop; | |
1793 | wo.wo_stat = NULL; | |
1794 | wo.wo_rusage = ru; | |
1795 | ret = do_wait(&wo); | |
dfe16dfa VM |
1796 | |
1797 | if (ret > 0) { | |
1798 | ret = 0; | |
1799 | } else if (infop) { | |
1800 | /* | |
1801 | * For a WNOHANG return, clear out all the fields | |
1802 | * we would set so the user can easily tell the | |
1803 | * difference. | |
1804 | */ | |
1805 | if (!ret) | |
1806 | ret = put_user(0, &infop->si_signo); | |
1807 | if (!ret) | |
1808 | ret = put_user(0, &infop->si_errno); | |
1809 | if (!ret) | |
1810 | ret = put_user(0, &infop->si_code); | |
1811 | if (!ret) | |
1812 | ret = put_user(0, &infop->si_pid); | |
1813 | if (!ret) | |
1814 | ret = put_user(0, &infop->si_uid); | |
1815 | if (!ret) | |
1816 | ret = put_user(0, &infop->si_status); | |
1817 | } | |
1818 | ||
161550d7 | 1819 | put_pid(pid); |
1da177e4 LT |
1820 | |
1821 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1822 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1823 | return ret; |
1824 | } | |
1825 | ||
754fe8d2 HC |
1826 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1827 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1828 | { |
9e8ae01d | 1829 | struct wait_opts wo; |
161550d7 EB |
1830 | struct pid *pid = NULL; |
1831 | enum pid_type type; | |
1da177e4 LT |
1832 | long ret; |
1833 | ||
1834 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1835 | __WNOTHREAD|__WCLONE|__WALL)) | |
1836 | return -EINVAL; | |
161550d7 EB |
1837 | |
1838 | if (upid == -1) | |
1839 | type = PIDTYPE_MAX; | |
1840 | else if (upid < 0) { | |
1841 | type = PIDTYPE_PGID; | |
1842 | pid = find_get_pid(-upid); | |
1843 | } else if (upid == 0) { | |
1844 | type = PIDTYPE_PGID; | |
2ae448ef | 1845 | pid = get_task_pid(current, PIDTYPE_PGID); |
161550d7 EB |
1846 | } else /* upid > 0 */ { |
1847 | type = PIDTYPE_PID; | |
1848 | pid = find_get_pid(upid); | |
1849 | } | |
1850 | ||
9e8ae01d ON |
1851 | wo.wo_type = type; |
1852 | wo.wo_pid = pid; | |
1853 | wo.wo_flags = options | WEXITED; | |
1854 | wo.wo_info = NULL; | |
1855 | wo.wo_stat = stat_addr; | |
1856 | wo.wo_rusage = ru; | |
1857 | ret = do_wait(&wo); | |
161550d7 | 1858 | put_pid(pid); |
1da177e4 LT |
1859 | |
1860 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1861 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1862 | return ret; |
1863 | } | |
1864 | ||
1865 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1866 | ||
1867 | /* | |
1868 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1869 | * implemented by calling sys_wait4() from libc.a. | |
1870 | */ | |
17da2bd9 | 1871 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1872 | { |
1873 | return sys_wait4(pid, stat_addr, options, NULL); | |
1874 | } | |
1875 | ||
1876 | #endif |