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> | |
6b3286ed | 15 | #include <linux/mnt_namespace.h> |
da9cbc87 | 16 | #include <linux/iocontext.h> |
1da177e4 LT |
17 | #include <linux/key.h> |
18 | #include <linux/security.h> | |
19 | #include <linux/cpu.h> | |
20 | #include <linux/acct.h> | |
8f0ab514 | 21 | #include <linux/tsacct_kern.h> |
1da177e4 | 22 | #include <linux/file.h> |
9f3acc31 | 23 | #include <linux/fdtable.h> |
1da177e4 | 24 | #include <linux/binfmts.h> |
ab516013 | 25 | #include <linux/nsproxy.h> |
84d73786 | 26 | #include <linux/pid_namespace.h> |
1da177e4 LT |
27 | #include <linux/ptrace.h> |
28 | #include <linux/profile.h> | |
29 | #include <linux/mount.h> | |
30 | #include <linux/proc_fs.h> | |
49d769d5 | 31 | #include <linux/kthread.h> |
1da177e4 | 32 | #include <linux/mempolicy.h> |
c757249a | 33 | #include <linux/taskstats_kern.h> |
ca74e92b | 34 | #include <linux/delayacct.h> |
83144186 | 35 | #include <linux/freezer.h> |
b4f48b63 | 36 | #include <linux/cgroup.h> |
1da177e4 | 37 | #include <linux/syscalls.h> |
7ed20e1a | 38 | #include <linux/signal.h> |
6a14c5c9 | 39 | #include <linux/posix-timers.h> |
9f46080c | 40 | #include <linux/cn_proc.h> |
de5097c2 | 41 | #include <linux/mutex.h> |
0771dfef | 42 | #include <linux/futex.h> |
b92ce558 | 43 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 44 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 45 | #include <linux/resource.h> |
0d67a46d | 46 | #include <linux/blkdev.h> |
6eaeeaba | 47 | #include <linux/task_io_accounting_ops.h> |
30199f5a | 48 | #include <linux/tracehook.h> |
5ad4e53b | 49 | #include <linux/fs_struct.h> |
d84f4f99 | 50 | #include <linux/init_task.h> |
0a16b607 | 51 | #include <trace/sched.h> |
1da177e4 LT |
52 | |
53 | #include <asm/uaccess.h> | |
54 | #include <asm/unistd.h> | |
55 | #include <asm/pgtable.h> | |
56 | #include <asm/mmu_context.h> | |
d84f4f99 | 57 | #include "cred-internals.h" |
1da177e4 | 58 | |
7e066fb8 MD |
59 | DEFINE_TRACE(sched_process_free); |
60 | DEFINE_TRACE(sched_process_exit); | |
61 | DEFINE_TRACE(sched_process_wait); | |
1da177e4 | 62 | |
408b664a AB |
63 | static void exit_mm(struct task_struct * tsk); |
64 | ||
d839fd4d ON |
65 | static inline int task_detached(struct task_struct *p) |
66 | { | |
67 | return p->exit_signal == -1; | |
68 | } | |
69 | ||
1da177e4 LT |
70 | static void __unhash_process(struct task_struct *p) |
71 | { | |
72 | nr_threads--; | |
73 | detach_pid(p, PIDTYPE_PID); | |
1da177e4 LT |
74 | if (thread_group_leader(p)) { |
75 | detach_pid(p, PIDTYPE_PGID); | |
76 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 77 | |
5e85d4ab | 78 | list_del_rcu(&p->tasks); |
73b9ebfe | 79 | __get_cpu_var(process_counts)--; |
1da177e4 | 80 | } |
47e65328 | 81 | list_del_rcu(&p->thread_group); |
f470021a | 82 | list_del_init(&p->sibling); |
1da177e4 LT |
83 | } |
84 | ||
6a14c5c9 ON |
85 | /* |
86 | * This function expects the tasklist_lock write-locked. | |
87 | */ | |
88 | static void __exit_signal(struct task_struct *tsk) | |
89 | { | |
90 | struct signal_struct *sig = tsk->signal; | |
91 | struct sighand_struct *sighand; | |
92 | ||
93 | BUG_ON(!sig); | |
94 | BUG_ON(!atomic_read(&sig->count)); | |
95 | ||
6a14c5c9 ON |
96 | sighand = rcu_dereference(tsk->sighand); |
97 | spin_lock(&sighand->siglock); | |
98 | ||
99 | posix_cpu_timers_exit(tsk); | |
100 | if (atomic_dec_and_test(&sig->count)) | |
101 | posix_cpu_timers_exit_group(tsk); | |
102 | else { | |
103 | /* | |
104 | * If there is any task waiting for the group exit | |
105 | * then notify it: | |
106 | */ | |
6db840fa | 107 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) |
6a14c5c9 | 108 | wake_up_process(sig->group_exit_task); |
6db840fa | 109 | |
6a14c5c9 ON |
110 | if (tsk == sig->curr_target) |
111 | sig->curr_target = next_thread(tsk); | |
112 | /* | |
113 | * Accumulate here the counters for all threads but the | |
114 | * group leader as they die, so they can be added into | |
115 | * the process-wide totals when those are taken. | |
116 | * The group leader stays around as a zombie as long | |
117 | * as there are other threads. When it gets reaped, | |
118 | * the exit.c code will add its counts into these totals. | |
119 | * We won't ever get here for the group leader, since it | |
120 | * will have been the last reference on the signal_struct. | |
121 | */ | |
32bd671d PZ |
122 | sig->utime = cputime_add(sig->utime, task_utime(tsk)); |
123 | sig->stime = cputime_add(sig->stime, task_stime(tsk)); | |
49048622 | 124 | sig->gtime = cputime_add(sig->gtime, task_gtime(tsk)); |
6a14c5c9 ON |
125 | sig->min_flt += tsk->min_flt; |
126 | sig->maj_flt += tsk->maj_flt; | |
127 | sig->nvcsw += tsk->nvcsw; | |
128 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
129 | sig->inblock += task_io_get_inblock(tsk); |
130 | sig->oublock += task_io_get_oublock(tsk); | |
5995477a | 131 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
32bd671d | 132 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
133 | sig = NULL; /* Marker for below. */ |
134 | } | |
135 | ||
5876700c ON |
136 | __unhash_process(tsk); |
137 | ||
da7978b0 ON |
138 | /* |
139 | * Do this under ->siglock, we can race with another thread | |
140 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
141 | */ | |
142 | flush_sigqueue(&tsk->pending); | |
143 | ||
6a14c5c9 | 144 | tsk->signal = NULL; |
a7e5328a | 145 | tsk->sighand = NULL; |
6a14c5c9 | 146 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 147 | |
a7e5328a | 148 | __cleanup_sighand(sighand); |
6a14c5c9 | 149 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
6a14c5c9 ON |
150 | if (sig) { |
151 | flush_sigqueue(&sig->shared_pending); | |
093a8e8a | 152 | taskstats_tgid_free(sig); |
ad474cac ON |
153 | /* |
154 | * Make sure ->signal can't go away under rq->lock, | |
155 | * see account_group_exec_runtime(). | |
156 | */ | |
157 | task_rq_unlock_wait(tsk); | |
6a14c5c9 ON |
158 | __cleanup_signal(sig); |
159 | } | |
160 | } | |
161 | ||
8c7904a0 EB |
162 | static void delayed_put_task_struct(struct rcu_head *rhp) |
163 | { | |
0a16b607 MD |
164 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
165 | ||
166 | trace_sched_process_free(tsk); | |
167 | put_task_struct(tsk); | |
8c7904a0 EB |
168 | } |
169 | ||
f470021a | 170 | |
1da177e4 LT |
171 | void release_task(struct task_struct * p) |
172 | { | |
36c8b586 | 173 | struct task_struct *leader; |
1da177e4 | 174 | int zap_leader; |
1f09f974 | 175 | repeat: |
dae33574 | 176 | tracehook_prepare_release_task(p); |
c69e8d9c DH |
177 | /* don't need to get the RCU readlock here - the process is dead and |
178 | * can't be modifying its own credentials */ | |
179 | atomic_dec(&__task_cred(p)->user->processes); | |
180 | ||
60347f67 | 181 | proc_flush_task(p); |
1da177e4 | 182 | write_lock_irq(&tasklist_lock); |
dae33574 | 183 | tracehook_finish_release_task(p); |
1da177e4 | 184 | __exit_signal(p); |
35f5cad8 | 185 | |
1da177e4 LT |
186 | /* |
187 | * If we are the last non-leader member of the thread | |
188 | * group, and the leader is zombie, then notify the | |
189 | * group leader's parent process. (if it wants notification.) | |
190 | */ | |
191 | zap_leader = 0; | |
192 | leader = p->group_leader; | |
193 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
d839fd4d | 194 | BUG_ON(task_detached(leader)); |
1da177e4 LT |
195 | do_notify_parent(leader, leader->exit_signal); |
196 | /* | |
197 | * If we were the last child thread and the leader has | |
198 | * exited already, and the leader's parent ignores SIGCHLD, | |
199 | * then we are the one who should release the leader. | |
200 | * | |
201 | * do_notify_parent() will have marked it self-reaping in | |
202 | * that case. | |
203 | */ | |
d839fd4d | 204 | zap_leader = task_detached(leader); |
dae33574 RM |
205 | |
206 | /* | |
207 | * This maintains the invariant that release_task() | |
208 | * only runs on a task in EXIT_DEAD, just for sanity. | |
209 | */ | |
210 | if (zap_leader) | |
211 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
212 | } |
213 | ||
1da177e4 | 214 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 215 | release_thread(p); |
8c7904a0 | 216 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
217 | |
218 | p = leader; | |
219 | if (unlikely(zap_leader)) | |
220 | goto repeat; | |
221 | } | |
222 | ||
1da177e4 LT |
223 | /* |
224 | * This checks not only the pgrp, but falls back on the pid if no | |
225 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
226 | * without this... | |
04a2e6a5 EB |
227 | * |
228 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 229 | */ |
04a2e6a5 | 230 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
231 | { |
232 | struct task_struct *p; | |
04a2e6a5 | 233 | struct pid *sid = NULL; |
62dfb554 | 234 | |
04a2e6a5 | 235 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 236 | if (p == NULL) |
04a2e6a5 | 237 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 238 | if (p != NULL) |
04a2e6a5 | 239 | sid = task_session(p); |
62dfb554 | 240 | |
1da177e4 LT |
241 | return sid; |
242 | } | |
243 | ||
244 | /* | |
245 | * Determine if a process group is "orphaned", according to the POSIX | |
246 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
247 | * by terminal-generated stop signals. Newly orphaned process groups are | |
248 | * to receive a SIGHUP and a SIGCONT. | |
249 | * | |
250 | * "I ask you, have you ever known what it is to be an orphan?" | |
251 | */ | |
0475ac08 | 252 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
253 | { |
254 | struct task_struct *p; | |
1da177e4 | 255 | |
0475ac08 | 256 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
257 | if ((p == ignored_task) || |
258 | (p->exit_state && thread_group_empty(p)) || | |
259 | is_global_init(p->real_parent)) | |
1da177e4 | 260 | continue; |
05e83df6 | 261 | |
0475ac08 | 262 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
263 | task_session(p->real_parent) == task_session(p)) |
264 | return 0; | |
0475ac08 | 265 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
266 | |
267 | return 1; | |
1da177e4 LT |
268 | } |
269 | ||
3e7cd6c4 | 270 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
271 | { |
272 | int retval; | |
273 | ||
274 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 275 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
276 | read_unlock(&tasklist_lock); |
277 | ||
278 | return retval; | |
279 | } | |
280 | ||
0475ac08 | 281 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
282 | { |
283 | int retval = 0; | |
284 | struct task_struct *p; | |
285 | ||
0475ac08 | 286 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
338077e5 | 287 | if (!task_is_stopped(p)) |
1da177e4 | 288 | continue; |
1da177e4 LT |
289 | retval = 1; |
290 | break; | |
0475ac08 | 291 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
292 | return retval; |
293 | } | |
294 | ||
f49ee505 ON |
295 | /* |
296 | * Check to see if any process groups have become orphaned as | |
297 | * a result of our exiting, and if they have any stopped jobs, | |
298 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
299 | */ | |
300 | static void | |
301 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
302 | { | |
303 | struct pid *pgrp = task_pgrp(tsk); | |
304 | struct task_struct *ignored_task = tsk; | |
305 | ||
306 | if (!parent) | |
307 | /* exit: our father is in a different pgrp than | |
308 | * we are and we were the only connection outside. | |
309 | */ | |
310 | parent = tsk->real_parent; | |
311 | else | |
312 | /* reparent: our child is in a different pgrp than | |
313 | * we are, and it was the only connection outside. | |
314 | */ | |
315 | ignored_task = NULL; | |
316 | ||
317 | if (task_pgrp(parent) != pgrp && | |
318 | task_session(parent) == task_session(tsk) && | |
319 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
320 | has_stopped_jobs(pgrp)) { | |
321 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
322 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
323 | } | |
324 | } | |
325 | ||
1da177e4 | 326 | /** |
49d769d5 | 327 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
328 | * |
329 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
330 | * it ever exits, it should generally reparent itself to kthreadd so it |
331 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
332 | * |
333 | * The various task state such as scheduling policy and priority may have | |
334 | * been inherited from a user process, so we reset them to sane values here. | |
335 | * | |
49d769d5 | 336 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 337 | */ |
49d769d5 | 338 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
339 | { |
340 | write_lock_irq(&tasklist_lock); | |
341 | ||
342 | ptrace_unlink(current); | |
343 | /* Reparent to init */ | |
49d769d5 | 344 | current->real_parent = current->parent = kthreadd_task; |
f470021a | 345 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
1da177e4 LT |
346 | |
347 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
348 | current->exit_signal = SIGCHLD; | |
349 | ||
e05606d3 | 350 | if (task_nice(current) < 0) |
1da177e4 LT |
351 | set_user_nice(current, 0); |
352 | /* cpus_allowed? */ | |
353 | /* rt_priority? */ | |
354 | /* signals? */ | |
1da177e4 LT |
355 | memcpy(current->signal->rlim, init_task.signal->rlim, |
356 | sizeof(current->signal->rlim)); | |
d84f4f99 DH |
357 | |
358 | atomic_inc(&init_cred.usage); | |
359 | commit_creds(&init_cred); | |
1da177e4 | 360 | write_unlock_irq(&tasklist_lock); |
1da177e4 LT |
361 | } |
362 | ||
8520d7c7 | 363 | void __set_special_pids(struct pid *pid) |
1da177e4 | 364 | { |
e19f247a | 365 | struct task_struct *curr = current->group_leader; |
8520d7c7 | 366 | pid_t nr = pid_nr(pid); |
1da177e4 | 367 | |
8520d7c7 | 368 | if (task_session(curr) != pid) { |
7d8da096 | 369 | change_pid(curr, PIDTYPE_SID, pid); |
8520d7c7 | 370 | set_task_session(curr, nr); |
1da177e4 | 371 | } |
8520d7c7 | 372 | if (task_pgrp(curr) != pid) { |
7d8da096 | 373 | change_pid(curr, PIDTYPE_PGID, pid); |
8520d7c7 | 374 | set_task_pgrp(curr, nr); |
1da177e4 LT |
375 | } |
376 | } | |
377 | ||
8520d7c7 | 378 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
379 | { |
380 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 381 | __set_special_pids(pid); |
1da177e4 LT |
382 | write_unlock_irq(&tasklist_lock); |
383 | } | |
384 | ||
385 | /* | |
386 | * Let kernel threads use this to say that they | |
387 | * allow a certain signal (since daemonize() will | |
388 | * have disabled all of them by default). | |
389 | */ | |
390 | int allow_signal(int sig) | |
391 | { | |
7ed20e1a | 392 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
393 | return -EINVAL; |
394 | ||
395 | spin_lock_irq(¤t->sighand->siglock); | |
396 | sigdelset(¤t->blocked, sig); | |
397 | if (!current->mm) { | |
398 | /* Kernel threads handle their own signals. | |
399 | Let the signal code know it'll be handled, so | |
400 | that they don't get converted to SIGKILL or | |
401 | just silently dropped */ | |
402 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
403 | } | |
404 | recalc_sigpending(); | |
405 | spin_unlock_irq(¤t->sighand->siglock); | |
406 | return 0; | |
407 | } | |
408 | ||
409 | EXPORT_SYMBOL(allow_signal); | |
410 | ||
411 | int disallow_signal(int sig) | |
412 | { | |
7ed20e1a | 413 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
414 | return -EINVAL; |
415 | ||
416 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 417 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
418 | recalc_sigpending(); |
419 | spin_unlock_irq(¤t->sighand->siglock); | |
420 | return 0; | |
421 | } | |
422 | ||
423 | EXPORT_SYMBOL(disallow_signal); | |
424 | ||
425 | /* | |
426 | * Put all the gunge required to become a kernel thread without | |
427 | * attached user resources in one place where it belongs. | |
428 | */ | |
429 | ||
430 | void daemonize(const char *name, ...) | |
431 | { | |
432 | va_list args; | |
1da177e4 LT |
433 | sigset_t blocked; |
434 | ||
435 | va_start(args, name); | |
436 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
437 | va_end(args); | |
438 | ||
439 | /* | |
440 | * If we were started as result of loading a module, close all of the | |
441 | * user space pages. We don't need them, and if we didn't close them | |
442 | * they would be locked into memory. | |
443 | */ | |
444 | exit_mm(current); | |
83144186 RW |
445 | /* |
446 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
447 | * or suspend transition begins right now. | |
448 | */ | |
7b34e428 | 449 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
1da177e4 | 450 | |
8520d7c7 ON |
451 | if (current->nsproxy != &init_nsproxy) { |
452 | get_nsproxy(&init_nsproxy); | |
453 | switch_task_namespaces(current, &init_nsproxy); | |
454 | } | |
297bd42b | 455 | set_special_pids(&init_struct_pid); |
24ec839c | 456 | proc_clear_tty(current); |
1da177e4 LT |
457 | |
458 | /* Block and flush all signals */ | |
459 | sigfillset(&blocked); | |
460 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
461 | flush_signals(current); | |
462 | ||
463 | /* Become as one with the init task */ | |
464 | ||
3e93cd67 | 465 | daemonize_fs_struct(); |
d4c5e41f | 466 | exit_files(current); |
1da177e4 LT |
467 | current->files = init_task.files; |
468 | atomic_inc(¤t->files->count); | |
469 | ||
49d769d5 | 470 | reparent_to_kthreadd(); |
1da177e4 LT |
471 | } |
472 | ||
473 | EXPORT_SYMBOL(daemonize); | |
474 | ||
858119e1 | 475 | static void close_files(struct files_struct * files) |
1da177e4 LT |
476 | { |
477 | int i, j; | |
badf1662 | 478 | struct fdtable *fdt; |
1da177e4 LT |
479 | |
480 | j = 0; | |
4fb3a538 DS |
481 | |
482 | /* | |
483 | * It is safe to dereference the fd table without RCU or | |
484 | * ->file_lock because this is the last reference to the | |
485 | * files structure. | |
486 | */ | |
badf1662 | 487 | fdt = files_fdtable(files); |
1da177e4 LT |
488 | for (;;) { |
489 | unsigned long set; | |
490 | i = j * __NFDBITS; | |
bbea9f69 | 491 | if (i >= fdt->max_fds) |
1da177e4 | 492 | break; |
badf1662 | 493 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
494 | while (set) { |
495 | if (set & 1) { | |
badf1662 | 496 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 497 | if (file) { |
1da177e4 | 498 | filp_close(file, files); |
944be0b2 IM |
499 | cond_resched(); |
500 | } | |
1da177e4 LT |
501 | } |
502 | i++; | |
503 | set >>= 1; | |
504 | } | |
505 | } | |
506 | } | |
507 | ||
508 | struct files_struct *get_files_struct(struct task_struct *task) | |
509 | { | |
510 | struct files_struct *files; | |
511 | ||
512 | task_lock(task); | |
513 | files = task->files; | |
514 | if (files) | |
515 | atomic_inc(&files->count); | |
516 | task_unlock(task); | |
517 | ||
518 | return files; | |
519 | } | |
520 | ||
7ad5b3a5 | 521 | void put_files_struct(struct files_struct *files) |
1da177e4 | 522 | { |
badf1662 DS |
523 | struct fdtable *fdt; |
524 | ||
1da177e4 LT |
525 | if (atomic_dec_and_test(&files->count)) { |
526 | close_files(files); | |
527 | /* | |
528 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
529 | * If the fdtable was embedded, pass files for freeing |
530 | * at the end of the RCU grace period. Otherwise, | |
531 | * you can free files immediately. | |
1da177e4 | 532 | */ |
badf1662 | 533 | fdt = files_fdtable(files); |
4fd45812 | 534 | if (fdt != &files->fdtab) |
ab2af1f5 | 535 | kmem_cache_free(files_cachep, files); |
01b2d93c | 536 | free_fdtable(fdt); |
1da177e4 LT |
537 | } |
538 | } | |
539 | ||
3b125388 | 540 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 541 | { |
3b125388 | 542 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
543 | struct files_struct *old; |
544 | ||
545 | old = tsk->files; | |
546 | task_lock(tsk); | |
547 | tsk->files = files; | |
548 | task_unlock(tsk); | |
549 | put_files_struct(old); | |
550 | } | |
3b9b8ab6 | 551 | |
1ec7f1dd | 552 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
553 | { |
554 | struct files_struct * files = tsk->files; | |
555 | ||
556 | if (files) { | |
557 | task_lock(tsk); | |
558 | tsk->files = NULL; | |
559 | task_unlock(tsk); | |
560 | put_files_struct(files); | |
561 | } | |
562 | } | |
563 | ||
cf475ad2 BS |
564 | #ifdef CONFIG_MM_OWNER |
565 | /* | |
566 | * Task p is exiting and it owned mm, lets find a new owner for it | |
567 | */ | |
568 | static inline int | |
569 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) | |
570 | { | |
571 | /* | |
572 | * If there are other users of the mm and the owner (us) is exiting | |
573 | * we need to find a new owner to take on the responsibility. | |
574 | */ | |
cf475ad2 BS |
575 | if (atomic_read(&mm->mm_users) <= 1) |
576 | return 0; | |
577 | if (mm->owner != p) | |
578 | return 0; | |
579 | return 1; | |
580 | } | |
581 | ||
582 | void mm_update_next_owner(struct mm_struct *mm) | |
583 | { | |
584 | struct task_struct *c, *g, *p = current; | |
585 | ||
586 | retry: | |
587 | if (!mm_need_new_owner(mm, p)) | |
588 | return; | |
589 | ||
590 | read_lock(&tasklist_lock); | |
591 | /* | |
592 | * Search in the children | |
593 | */ | |
594 | list_for_each_entry(c, &p->children, sibling) { | |
595 | if (c->mm == mm) | |
596 | goto assign_new_owner; | |
597 | } | |
598 | ||
599 | /* | |
600 | * Search in the siblings | |
601 | */ | |
602 | list_for_each_entry(c, &p->parent->children, sibling) { | |
603 | if (c->mm == mm) | |
604 | goto assign_new_owner; | |
605 | } | |
606 | ||
607 | /* | |
608 | * Search through everything else. We should not get | |
609 | * here often | |
610 | */ | |
611 | do_each_thread(g, c) { | |
612 | if (c->mm == mm) | |
613 | goto assign_new_owner; | |
614 | } while_each_thread(g, c); | |
615 | ||
616 | read_unlock(&tasklist_lock); | |
31a78f23 BS |
617 | /* |
618 | * We found no owner yet mm_users > 1: this implies that we are | |
619 | * most likely racing with swapoff (try_to_unuse()) or /proc or | |
e5991371 | 620 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
31a78f23 | 621 | */ |
31a78f23 | 622 | mm->owner = NULL; |
cf475ad2 BS |
623 | return; |
624 | ||
625 | assign_new_owner: | |
626 | BUG_ON(c == p); | |
627 | get_task_struct(c); | |
628 | /* | |
629 | * The task_lock protects c->mm from changing. | |
630 | * We always want mm->owner->mm == mm | |
631 | */ | |
632 | task_lock(c); | |
e5991371 HD |
633 | /* |
634 | * Delay read_unlock() till we have the task_lock() | |
635 | * to ensure that c does not slip away underneath us | |
636 | */ | |
637 | read_unlock(&tasklist_lock); | |
cf475ad2 BS |
638 | if (c->mm != mm) { |
639 | task_unlock(c); | |
640 | put_task_struct(c); | |
641 | goto retry; | |
642 | } | |
cf475ad2 BS |
643 | mm->owner = c; |
644 | task_unlock(c); | |
645 | put_task_struct(c); | |
646 | } | |
647 | #endif /* CONFIG_MM_OWNER */ | |
648 | ||
1da177e4 LT |
649 | /* |
650 | * Turn us into a lazy TLB process if we | |
651 | * aren't already.. | |
652 | */ | |
408b664a | 653 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
654 | { |
655 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 656 | struct core_state *core_state; |
1da177e4 LT |
657 | |
658 | mm_release(tsk, mm); | |
659 | if (!mm) | |
660 | return; | |
661 | /* | |
662 | * Serialize with any possible pending coredump. | |
999d9fc1 | 663 | * We must hold mmap_sem around checking core_state |
1da177e4 | 664 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 665 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
666 | * group with ->mm != NULL. |
667 | */ | |
668 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
669 | core_state = mm->core_state; |
670 | if (core_state) { | |
671 | struct core_thread self; | |
1da177e4 | 672 | up_read(&mm->mmap_sem); |
1da177e4 | 673 | |
b564daf8 ON |
674 | self.task = tsk; |
675 | self.next = xchg(&core_state->dumper.next, &self); | |
676 | /* | |
677 | * Implies mb(), the result of xchg() must be visible | |
678 | * to core_state->dumper. | |
679 | */ | |
680 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
681 | complete(&core_state->startup); | |
1da177e4 | 682 | |
a94e2d40 ON |
683 | for (;;) { |
684 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
685 | if (!self.task) /* see coredump_finish() */ | |
686 | break; | |
687 | schedule(); | |
688 | } | |
689 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
690 | down_read(&mm->mmap_sem); |
691 | } | |
692 | atomic_inc(&mm->mm_count); | |
125e1874 | 693 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
694 | /* more a memory barrier than a real lock */ |
695 | task_lock(tsk); | |
696 | tsk->mm = NULL; | |
697 | up_read(&mm->mmap_sem); | |
698 | enter_lazy_tlb(mm, current); | |
0c1eecfb RW |
699 | /* We don't want this task to be frozen prematurely */ |
700 | clear_freeze_flag(tsk); | |
1da177e4 | 701 | task_unlock(tsk); |
cf475ad2 | 702 | mm_update_next_owner(mm); |
1da177e4 LT |
703 | mmput(mm); |
704 | } | |
705 | ||
666f164f RM |
706 | /* |
707 | * Return nonzero if @parent's children should reap themselves. | |
708 | * | |
709 | * Called with write_lock_irq(&tasklist_lock) held. | |
710 | */ | |
711 | static int ignoring_children(struct task_struct *parent) | |
1da177e4 | 712 | { |
666f164f RM |
713 | int ret; |
714 | struct sighand_struct *psig = parent->sighand; | |
715 | unsigned long flags; | |
716 | spin_lock_irqsave(&psig->siglock, flags); | |
717 | ret = (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN || | |
718 | (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT)); | |
719 | spin_unlock_irqrestore(&psig->siglock, flags); | |
720 | return ret; | |
721 | } | |
241ceee0 | 722 | |
f470021a RM |
723 | /* |
724 | * Detach all tasks we were using ptrace on. | |
725 | * Any that need to be release_task'd are put on the @dead list. | |
726 | * | |
727 | * Called with write_lock(&tasklist_lock) held. | |
728 | */ | |
729 | static void ptrace_exit(struct task_struct *parent, struct list_head *dead) | |
1da177e4 | 730 | { |
f470021a | 731 | struct task_struct *p, *n; |
666f164f | 732 | int ign = -1; |
241ceee0 | 733 | |
f470021a RM |
734 | list_for_each_entry_safe(p, n, &parent->ptraced, ptrace_entry) { |
735 | __ptrace_unlink(p); | |
736 | ||
737 | if (p->exit_state != EXIT_ZOMBIE) | |
738 | continue; | |
739 | ||
740 | /* | |
741 | * If it's a zombie, our attachedness prevented normal | |
742 | * parent notification or self-reaping. Do notification | |
743 | * now if it would have happened earlier. If it should | |
744 | * reap itself, add it to the @dead list. We can't call | |
745 | * release_task() here because we already hold tasklist_lock. | |
746 | * | |
747 | * If it's our own child, there is no notification to do. | |
666f164f RM |
748 | * But if our normal children self-reap, then this child |
749 | * was prevented by ptrace and we must reap it now. | |
1da177e4 | 750 | */ |
f470021a RM |
751 | if (!task_detached(p) && thread_group_empty(p)) { |
752 | if (!same_thread_group(p->real_parent, parent)) | |
753 | do_notify_parent(p, p->exit_signal); | |
666f164f RM |
754 | else { |
755 | if (ign < 0) | |
756 | ign = ignoring_children(parent); | |
757 | if (ign) | |
758 | p->exit_signal = -1; | |
759 | } | |
f470021a | 760 | } |
1da177e4 | 761 | |
f470021a | 762 | if (task_detached(p)) { |
1da177e4 | 763 | /* |
f470021a | 764 | * Mark it as in the process of being reaped. |
1da177e4 | 765 | */ |
f470021a RM |
766 | p->exit_state = EXIT_DEAD; |
767 | list_add(&p->ptrace_entry, dead); | |
1da177e4 LT |
768 | } |
769 | } | |
f470021a RM |
770 | } |
771 | ||
772 | /* | |
773 | * Finish up exit-time ptrace cleanup. | |
774 | * | |
775 | * Called without locks. | |
776 | */ | |
777 | static void ptrace_exit_finish(struct task_struct *parent, | |
778 | struct list_head *dead) | |
779 | { | |
780 | struct task_struct *p, *n; | |
781 | ||
782 | BUG_ON(!list_empty(&parent->ptraced)); | |
783 | ||
784 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
785 | list_del_init(&p->ptrace_entry); | |
786 | release_task(p); | |
787 | } | |
788 | } | |
789 | ||
790 | static void reparent_thread(struct task_struct *p, struct task_struct *father) | |
791 | { | |
792 | if (p->pdeath_signal) | |
793 | /* We already hold the tasklist_lock here. */ | |
794 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | |
795 | ||
796 | list_move_tail(&p->sibling, &p->real_parent->children); | |
1da177e4 | 797 | |
b2b2cbc4 EB |
798 | /* If this is a threaded reparent there is no need to |
799 | * notify anyone anything has happened. | |
800 | */ | |
376e1d25 | 801 | if (same_thread_group(p->real_parent, father)) |
b2b2cbc4 EB |
802 | return; |
803 | ||
804 | /* We don't want people slaying init. */ | |
d839fd4d | 805 | if (!task_detached(p)) |
b2b2cbc4 | 806 | p->exit_signal = SIGCHLD; |
b2b2cbc4 EB |
807 | |
808 | /* If we'd notified the old parent about this child's death, | |
809 | * also notify the new parent. | |
810 | */ | |
f470021a RM |
811 | if (!ptrace_reparented(p) && |
812 | p->exit_state == EXIT_ZOMBIE && | |
d839fd4d | 813 | !task_detached(p) && thread_group_empty(p)) |
b2b2cbc4 EB |
814 | do_notify_parent(p, p->exit_signal); |
815 | ||
f49ee505 | 816 | kill_orphaned_pgrp(p, father); |
1da177e4 LT |
817 | } |
818 | ||
819 | /* | |
820 | * When we die, we re-parent all our children. | |
821 | * Try to give them to another thread in our thread | |
822 | * group, and if no such member exists, give it to | |
84d73786 SB |
823 | * the child reaper process (ie "init") in our pid |
824 | * space. | |
1da177e4 | 825 | */ |
950bbabb | 826 | static struct task_struct *find_new_reaper(struct task_struct *father) |
1da177e4 | 827 | { |
950bbabb ON |
828 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
829 | struct task_struct *thread; | |
1da177e4 | 830 | |
950bbabb ON |
831 | thread = father; |
832 | while_each_thread(father, thread) { | |
833 | if (thread->flags & PF_EXITING) | |
834 | continue; | |
835 | if (unlikely(pid_ns->child_reaper == father)) | |
836 | pid_ns->child_reaper = thread; | |
837 | return thread; | |
838 | } | |
1da177e4 | 839 | |
950bbabb ON |
840 | if (unlikely(pid_ns->child_reaper == father)) { |
841 | write_unlock_irq(&tasklist_lock); | |
842 | if (unlikely(pid_ns == &init_pid_ns)) | |
843 | panic("Attempted to kill init!"); | |
1da177e4 | 844 | |
950bbabb ON |
845 | zap_pid_ns_processes(pid_ns); |
846 | write_lock_irq(&tasklist_lock); | |
1da177e4 | 847 | /* |
950bbabb ON |
848 | * We can not clear ->child_reaper or leave it alone. |
849 | * There may by stealth EXIT_DEAD tasks on ->children, | |
850 | * forget_original_parent() must move them somewhere. | |
1da177e4 | 851 | */ |
950bbabb | 852 | pid_ns->child_reaper = init_pid_ns.child_reaper; |
1da177e4 | 853 | } |
762a24be | 854 | |
950bbabb ON |
855 | return pid_ns->child_reaper; |
856 | } | |
857 | ||
762a24be | 858 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 859 | { |
950bbabb | 860 | struct task_struct *p, *n, *reaper; |
f470021a | 861 | LIST_HEAD(ptrace_dead); |
762a24be ON |
862 | |
863 | write_lock_irq(&tasklist_lock); | |
950bbabb | 864 | reaper = find_new_reaper(father); |
f470021a RM |
865 | /* |
866 | * First clean up ptrace if we were using it. | |
867 | */ | |
868 | ptrace_exit(father, &ptrace_dead); | |
869 | ||
03ff1797 | 870 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
84eb646b | 871 | p->real_parent = reaper; |
f470021a RM |
872 | if (p->parent == father) { |
873 | BUG_ON(p->ptrace); | |
874 | p->parent = p->real_parent; | |
875 | } | |
876 | reparent_thread(p, father); | |
1da177e4 | 877 | } |
762a24be ON |
878 | |
879 | write_unlock_irq(&tasklist_lock); | |
880 | BUG_ON(!list_empty(&father->children)); | |
762a24be | 881 | |
f470021a | 882 | ptrace_exit_finish(father, &ptrace_dead); |
1da177e4 LT |
883 | } |
884 | ||
885 | /* | |
886 | * Send signals to all our closest relatives so that they know | |
887 | * to properly mourn us.. | |
888 | */ | |
821c7de7 | 889 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 890 | { |
2b2a1ff6 RM |
891 | int signal; |
892 | void *cookie; | |
1da177e4 | 893 | |
1da177e4 LT |
894 | /* |
895 | * This does two things: | |
896 | * | |
897 | * A. Make init inherit all the child processes | |
898 | * B. Check to see if any process groups have become orphaned | |
899 | * as a result of our exiting, and if they have any stopped | |
900 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
901 | */ | |
762a24be | 902 | forget_original_parent(tsk); |
2e4a7072 | 903 | exit_task_namespaces(tsk); |
1da177e4 | 904 | |
762a24be | 905 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
906 | if (group_dead) |
907 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 908 | |
24728448 | 909 | /* Let father know we died |
1da177e4 LT |
910 | * |
911 | * Thread signals are configurable, but you aren't going to use | |
d4c5e41f | 912 | * that to send signals to arbitary processes. |
1da177e4 LT |
913 | * That stops right now. |
914 | * | |
915 | * If the parent exec id doesn't match the exec id we saved | |
916 | * when we started then we know the parent has changed security | |
917 | * domain. | |
918 | * | |
919 | * If our self_exec id doesn't match our parent_exec_id then | |
920 | * we have changed execution domain as these two values started | |
921 | * the same after a fork. | |
1da177e4 | 922 | */ |
d839fd4d | 923 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
f49ee505 | 924 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
d839fd4d ON |
925 | tsk->self_exec_id != tsk->parent_exec_id) && |
926 | !capable(CAP_KILL)) | |
1da177e4 LT |
927 | tsk->exit_signal = SIGCHLD; |
928 | ||
2b2a1ff6 | 929 | signal = tracehook_notify_death(tsk, &cookie, group_dead); |
5c7edcd7 | 930 | if (signal >= 0) |
2b2a1ff6 | 931 | signal = do_notify_parent(tsk, signal); |
1da177e4 | 932 | |
5c7edcd7 | 933 | tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE; |
1da177e4 | 934 | |
2800d8d1 | 935 | /* mt-exec, de_thread() is waiting for us */ |
6db840fa | 936 | if (thread_group_leader(tsk) && |
2633f0e5 SV |
937 | tsk->signal->group_exit_task && |
938 | tsk->signal->notify_count < 0) | |
6db840fa ON |
939 | wake_up_process(tsk->signal->group_exit_task); |
940 | ||
1da177e4 LT |
941 | write_unlock_irq(&tasklist_lock); |
942 | ||
2b2a1ff6 RM |
943 | tracehook_report_death(tsk, signal, cookie, group_dead); |
944 | ||
1da177e4 | 945 | /* If the process is dead, release it - nobody will wait for it */ |
5c7edcd7 | 946 | if (signal == DEATH_REAP) |
1da177e4 | 947 | release_task(tsk); |
1da177e4 LT |
948 | } |
949 | ||
e18eecb8 JD |
950 | #ifdef CONFIG_DEBUG_STACK_USAGE |
951 | static void check_stack_usage(void) | |
952 | { | |
953 | static DEFINE_SPINLOCK(low_water_lock); | |
954 | static int lowest_to_date = THREAD_SIZE; | |
e18eecb8 JD |
955 | unsigned long free; |
956 | ||
7c9f8861 | 957 | free = stack_not_used(current); |
e18eecb8 JD |
958 | |
959 | if (free >= lowest_to_date) | |
960 | return; | |
961 | ||
962 | spin_lock(&low_water_lock); | |
963 | if (free < lowest_to_date) { | |
964 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
965 | "left\n", | |
966 | current->comm, free); | |
967 | lowest_to_date = free; | |
968 | } | |
969 | spin_unlock(&low_water_lock); | |
970 | } | |
971 | #else | |
972 | static inline void check_stack_usage(void) {} | |
973 | #endif | |
974 | ||
7ad5b3a5 | 975 | NORET_TYPE void do_exit(long code) |
1da177e4 LT |
976 | { |
977 | struct task_struct *tsk = current; | |
978 | int group_dead; | |
979 | ||
980 | profile_task_exit(tsk); | |
981 | ||
22e2c507 JA |
982 | WARN_ON(atomic_read(&tsk->fs_excl)); |
983 | ||
1da177e4 LT |
984 | if (unlikely(in_interrupt())) |
985 | panic("Aiee, killing interrupt handler!"); | |
986 | if (unlikely(!tsk->pid)) | |
987 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 988 | |
30199f5a | 989 | tracehook_report_exit(&code); |
1da177e4 | 990 | |
df164db5 AN |
991 | /* |
992 | * We're taking recursive faults here in do_exit. Safest is to just | |
993 | * leave this task alone and wait for reboot. | |
994 | */ | |
995 | if (unlikely(tsk->flags & PF_EXITING)) { | |
996 | printk(KERN_ALERT | |
997 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
998 | /* |
999 | * We can do this unlocked here. The futex code uses | |
1000 | * this flag just to verify whether the pi state | |
1001 | * cleanup has been done or not. In the worst case it | |
1002 | * loops once more. We pretend that the cleanup was | |
1003 | * done as there is no way to return. Either the | |
1004 | * OWNER_DIED bit is set by now or we push the blocked | |
1005 | * task into the wait for ever nirwana as well. | |
1006 | */ | |
1007 | tsk->flags |= PF_EXITPIDONE; | |
df164db5 AN |
1008 | set_current_state(TASK_UNINTERRUPTIBLE); |
1009 | schedule(); | |
1010 | } | |
1011 | ||
d12619b5 | 1012 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
1013 | /* |
1014 | * tsk->flags are checked in the futex code to protect against | |
1015 | * an exiting task cleaning up the robust pi futexes. | |
1016 | */ | |
d2ee7198 ON |
1017 | smp_mb(); |
1018 | spin_unlock_wait(&tsk->pi_lock); | |
1da177e4 | 1019 | |
1da177e4 LT |
1020 | if (unlikely(in_atomic())) |
1021 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 1022 | current->comm, task_pid_nr(current), |
1da177e4 LT |
1023 | preempt_count()); |
1024 | ||
1025 | acct_update_integrals(tsk); | |
901608d9 | 1026 | |
1da177e4 | 1027 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 1028 | if (group_dead) { |
778e9a9c | 1029 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 1030 | exit_itimers(tsk->signal); |
c3068951 | 1031 | } |
f6ec29a4 | 1032 | acct_collect(code, group_dead); |
522ed776 MT |
1033 | if (group_dead) |
1034 | tty_audit_exit(); | |
fa84cb93 AV |
1035 | if (unlikely(tsk->audit_context)) |
1036 | audit_free(tsk); | |
115085ea | 1037 | |
f2ab6d88 | 1038 | tsk->exit_code = code; |
115085ea | 1039 | taskstats_exit(tsk, group_dead); |
c757249a | 1040 | |
1da177e4 LT |
1041 | exit_mm(tsk); |
1042 | ||
0e464814 | 1043 | if (group_dead) |
f6ec29a4 | 1044 | acct_process(); |
0a16b607 MD |
1045 | trace_sched_process_exit(tsk); |
1046 | ||
1da177e4 | 1047 | exit_sem(tsk); |
1ec7f1dd AV |
1048 | exit_files(tsk); |
1049 | exit_fs(tsk); | |
e18eecb8 | 1050 | check_stack_usage(); |
1da177e4 | 1051 | exit_thread(); |
b4f48b63 | 1052 | cgroup_exit(tsk, 1); |
1da177e4 LT |
1053 | |
1054 | if (group_dead && tsk->signal->leader) | |
1055 | disassociate_ctty(1); | |
1056 | ||
a1261f54 | 1057 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 LT |
1058 | if (tsk->binfmt) |
1059 | module_put(tsk->binfmt->module); | |
1060 | ||
9f46080c | 1061 | proc_exit_connector(tsk); |
821c7de7 | 1062 | exit_notify(tsk, group_dead); |
1da177e4 | 1063 | #ifdef CONFIG_NUMA |
f0be3d32 | 1064 | mpol_put(tsk->mempolicy); |
1da177e4 LT |
1065 | tsk->mempolicy = NULL; |
1066 | #endif | |
42b2dd0a | 1067 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
1068 | /* |
1069 | * This must happen late, after the PID is not | |
1070 | * hashed anymore: | |
1071 | */ | |
1072 | if (unlikely(!list_empty(&tsk->pi_state_list))) | |
1073 | exit_pi_state_list(tsk); | |
1074 | if (unlikely(current->pi_state_cache)) | |
1075 | kfree(current->pi_state_cache); | |
42b2dd0a | 1076 | #endif |
de5097c2 | 1077 | /* |
9a11b49a | 1078 | * Make sure we are holding no locks: |
de5097c2 | 1079 | */ |
9a11b49a | 1080 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1081 | /* |
1082 | * We can do this unlocked here. The futex code uses this flag | |
1083 | * just to verify whether the pi state cleanup has been done | |
1084 | * or not. In the worst case it loops once more. | |
1085 | */ | |
1086 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1087 | |
afc847b7 AV |
1088 | if (tsk->io_context) |
1089 | exit_io_context(); | |
1090 | ||
b92ce558 JA |
1091 | if (tsk->splice_pipe) |
1092 | __free_pipe_info(tsk->splice_pipe); | |
1093 | ||
7407251a | 1094 | preempt_disable(); |
55a101f8 | 1095 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1096 | tsk->state = TASK_DEAD; |
1da177e4 LT |
1097 | schedule(); |
1098 | BUG(); | |
1099 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1100 | for (;;) |
1101 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1102 | } |
1103 | ||
012914da RA |
1104 | EXPORT_SYMBOL_GPL(do_exit); |
1105 | ||
1da177e4 LT |
1106 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1107 | { | |
1108 | if (comp) | |
1109 | complete(comp); | |
55a101f8 | 1110 | |
1da177e4 LT |
1111 | do_exit(code); |
1112 | } | |
1113 | ||
1114 | EXPORT_SYMBOL(complete_and_exit); | |
1115 | ||
754fe8d2 | 1116 | SYSCALL_DEFINE1(exit, int, error_code) |
1da177e4 LT |
1117 | { |
1118 | do_exit((error_code&0xff)<<8); | |
1119 | } | |
1120 | ||
1da177e4 LT |
1121 | /* |
1122 | * Take down every thread in the group. This is called by fatal signals | |
1123 | * as well as by sys_exit_group (below). | |
1124 | */ | |
1125 | NORET_TYPE void | |
1126 | do_group_exit(int exit_code) | |
1127 | { | |
bfc4b089 ON |
1128 | struct signal_struct *sig = current->signal; |
1129 | ||
1da177e4 LT |
1130 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1131 | ||
bfc4b089 ON |
1132 | if (signal_group_exit(sig)) |
1133 | exit_code = sig->group_exit_code; | |
1da177e4 | 1134 | else if (!thread_group_empty(current)) { |
1da177e4 | 1135 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1136 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1137 | if (signal_group_exit(sig)) |
1da177e4 LT |
1138 | /* Another thread got here before we took the lock. */ |
1139 | exit_code = sig->group_exit_code; | |
1140 | else { | |
1da177e4 | 1141 | sig->group_exit_code = exit_code; |
ed5d2cac | 1142 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1143 | zap_other_threads(current); |
1144 | } | |
1145 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1146 | } |
1147 | ||
1148 | do_exit(exit_code); | |
1149 | /* NOTREACHED */ | |
1150 | } | |
1151 | ||
1152 | /* | |
1153 | * this kills every thread in the thread group. Note that any externally | |
1154 | * wait4()-ing process will get the correct exit code - even if this | |
1155 | * thread is not the thread group leader. | |
1156 | */ | |
754fe8d2 | 1157 | SYSCALL_DEFINE1(exit_group, int, error_code) |
1da177e4 LT |
1158 | { |
1159 | do_group_exit((error_code & 0xff) << 8); | |
2ed7c03e HC |
1160 | /* NOTREACHED */ |
1161 | return 0; | |
1da177e4 LT |
1162 | } |
1163 | ||
161550d7 EB |
1164 | static struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
1165 | { | |
1166 | struct pid *pid = NULL; | |
1167 | if (type == PIDTYPE_PID) | |
1168 | pid = task->pids[type].pid; | |
1169 | else if (type < PIDTYPE_MAX) | |
1170 | pid = task->group_leader->pids[type].pid; | |
1171 | return pid; | |
1172 | } | |
1173 | ||
1174 | static int eligible_child(enum pid_type type, struct pid *pid, int options, | |
1175 | struct task_struct *p) | |
1da177e4 | 1176 | { |
73243284 RM |
1177 | int err; |
1178 | ||
161550d7 EB |
1179 | if (type < PIDTYPE_MAX) { |
1180 | if (task_pid_type(p, type) != pid) | |
1da177e4 LT |
1181 | return 0; |
1182 | } | |
1183 | ||
1da177e4 LT |
1184 | /* Wait for all children (clone and not) if __WALL is set; |
1185 | * otherwise, wait for clone children *only* if __WCLONE is | |
1186 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1187 | * A "clone" child here is one that reports to its parent | |
1188 | * using a signal other than SIGCHLD.) */ | |
1189 | if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0)) | |
1190 | && !(options & __WALL)) | |
1191 | return 0; | |
1da177e4 | 1192 | |
73243284 | 1193 | err = security_task_wait(p); |
14dd0b81 RM |
1194 | if (err) |
1195 | return err; | |
1da177e4 | 1196 | |
14dd0b81 | 1197 | return 1; |
1da177e4 LT |
1198 | } |
1199 | ||
36c8b586 | 1200 | static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid, |
1da177e4 LT |
1201 | int why, int status, |
1202 | struct siginfo __user *infop, | |
1203 | struct rusage __user *rusagep) | |
1204 | { | |
1205 | int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0; | |
36c8b586 | 1206 | |
1da177e4 LT |
1207 | put_task_struct(p); |
1208 | if (!retval) | |
1209 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1210 | if (!retval) | |
1211 | retval = put_user(0, &infop->si_errno); | |
1212 | if (!retval) | |
1213 | retval = put_user((short)why, &infop->si_code); | |
1214 | if (!retval) | |
1215 | retval = put_user(pid, &infop->si_pid); | |
1216 | if (!retval) | |
1217 | retval = put_user(uid, &infop->si_uid); | |
1218 | if (!retval) | |
1219 | retval = put_user(status, &infop->si_status); | |
1220 | if (!retval) | |
1221 | retval = pid; | |
1222 | return retval; | |
1223 | } | |
1224 | ||
1225 | /* | |
1226 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1227 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1228 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1229 | * released the lock and the system call should return. | |
1230 | */ | |
98abed02 | 1231 | static int wait_task_zombie(struct task_struct *p, int options, |
1da177e4 LT |
1232 | struct siginfo __user *infop, |
1233 | int __user *stat_addr, struct rusage __user *ru) | |
1234 | { | |
1235 | unsigned long state; | |
2f4e6e2a | 1236 | int retval, status, traced; |
6c5f3e7b | 1237 | pid_t pid = task_pid_vnr(p); |
c69e8d9c | 1238 | uid_t uid = __task_cred(p)->uid; |
1da177e4 | 1239 | |
98abed02 RM |
1240 | if (!likely(options & WEXITED)) |
1241 | return 0; | |
1242 | ||
1243 | if (unlikely(options & WNOWAIT)) { | |
1da177e4 LT |
1244 | int exit_code = p->exit_code; |
1245 | int why, status; | |
1246 | ||
1da177e4 LT |
1247 | get_task_struct(p); |
1248 | read_unlock(&tasklist_lock); | |
1249 | if ((exit_code & 0x7f) == 0) { | |
1250 | why = CLD_EXITED; | |
1251 | status = exit_code >> 8; | |
1252 | } else { | |
1253 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1254 | status = exit_code & 0x7f; | |
1255 | } | |
1256 | return wait_noreap_copyout(p, pid, uid, why, | |
1257 | status, infop, ru); | |
1258 | } | |
1259 | ||
1260 | /* | |
1261 | * Try to move the task's state to DEAD | |
1262 | * only one thread is allowed to do this: | |
1263 | */ | |
1264 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1265 | if (state != EXIT_ZOMBIE) { | |
1266 | BUG_ON(state != EXIT_DEAD); | |
1267 | return 0; | |
1268 | } | |
1da177e4 | 1269 | |
53b6f9fb | 1270 | traced = ptrace_reparented(p); |
2f4e6e2a ON |
1271 | |
1272 | if (likely(!traced)) { | |
3795e161 JJ |
1273 | struct signal_struct *psig; |
1274 | struct signal_struct *sig; | |
f06febc9 | 1275 | struct task_cputime cputime; |
3795e161 | 1276 | |
1da177e4 LT |
1277 | /* |
1278 | * The resource counters for the group leader are in its | |
1279 | * own task_struct. Those for dead threads in the group | |
1280 | * are in its signal_struct, as are those for the child | |
1281 | * processes it has previously reaped. All these | |
1282 | * accumulate in the parent's signal_struct c* fields. | |
1283 | * | |
1284 | * We don't bother to take a lock here to protect these | |
1285 | * p->signal fields, because they are only touched by | |
1286 | * __exit_signal, which runs with tasklist_lock | |
1287 | * write-locked anyway, and so is excluded here. We do | |
1288 | * need to protect the access to p->parent->signal fields, | |
1289 | * as other threads in the parent group can be right | |
1290 | * here reaping other children at the same time. | |
f06febc9 FM |
1291 | * |
1292 | * We use thread_group_cputime() to get times for the thread | |
1293 | * group, which consolidates times for all threads in the | |
1294 | * group including the group leader. | |
1da177e4 | 1295 | */ |
2b5fe6de | 1296 | thread_group_cputime(p, &cputime); |
1da177e4 | 1297 | spin_lock_irq(&p->parent->sighand->siglock); |
3795e161 JJ |
1298 | psig = p->parent->signal; |
1299 | sig = p->signal; | |
1300 | psig->cutime = | |
1301 | cputime_add(psig->cutime, | |
f06febc9 FM |
1302 | cputime_add(cputime.utime, |
1303 | sig->cutime)); | |
3795e161 JJ |
1304 | psig->cstime = |
1305 | cputime_add(psig->cstime, | |
f06febc9 FM |
1306 | cputime_add(cputime.stime, |
1307 | sig->cstime)); | |
9ac52315 LV |
1308 | psig->cgtime = |
1309 | cputime_add(psig->cgtime, | |
1310 | cputime_add(p->gtime, | |
1311 | cputime_add(sig->gtime, | |
1312 | sig->cgtime))); | |
3795e161 JJ |
1313 | psig->cmin_flt += |
1314 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1315 | psig->cmaj_flt += | |
1316 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1317 | psig->cnvcsw += | |
1318 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1319 | psig->cnivcsw += | |
1320 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1321 | psig->cinblock += |
1322 | task_io_get_inblock(p) + | |
1323 | sig->inblock + sig->cinblock; | |
1324 | psig->coublock += | |
1325 | task_io_get_oublock(p) + | |
1326 | sig->oublock + sig->coublock; | |
5995477a AR |
1327 | task_io_accounting_add(&psig->ioac, &p->ioac); |
1328 | task_io_accounting_add(&psig->ioac, &sig->ioac); | |
1da177e4 LT |
1329 | spin_unlock_irq(&p->parent->sighand->siglock); |
1330 | } | |
1331 | ||
1332 | /* | |
1333 | * Now we are sure this task is interesting, and no other | |
1334 | * thread can reap it because we set its state to EXIT_DEAD. | |
1335 | */ | |
1336 | read_unlock(&tasklist_lock); | |
1337 | ||
1338 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1339 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) | |
1340 | ? p->signal->group_exit_code : p->exit_code; | |
1341 | if (!retval && stat_addr) | |
1342 | retval = put_user(status, stat_addr); | |
1343 | if (!retval && infop) | |
1344 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1345 | if (!retval && infop) | |
1346 | retval = put_user(0, &infop->si_errno); | |
1347 | if (!retval && infop) { | |
1348 | int why; | |
1349 | ||
1350 | if ((status & 0x7f) == 0) { | |
1351 | why = CLD_EXITED; | |
1352 | status >>= 8; | |
1353 | } else { | |
1354 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1355 | status &= 0x7f; | |
1356 | } | |
1357 | retval = put_user((short)why, &infop->si_code); | |
1358 | if (!retval) | |
1359 | retval = put_user(status, &infop->si_status); | |
1360 | } | |
1361 | if (!retval && infop) | |
3a515e4a | 1362 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1363 | if (!retval && infop) |
c69e8d9c | 1364 | retval = put_user(uid, &infop->si_uid); |
2f4e6e2a | 1365 | if (!retval) |
3a515e4a | 1366 | retval = pid; |
2f4e6e2a ON |
1367 | |
1368 | if (traced) { | |
1da177e4 | 1369 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1370 | /* We dropped tasklist, ptracer could die and untrace */ |
1371 | ptrace_unlink(p); | |
1372 | /* | |
1373 | * If this is not a detached task, notify the parent. | |
1374 | * If it's still not detached after that, don't release | |
1375 | * it now. | |
1376 | */ | |
d839fd4d | 1377 | if (!task_detached(p)) { |
2f4e6e2a | 1378 | do_notify_parent(p, p->exit_signal); |
d839fd4d | 1379 | if (!task_detached(p)) { |
2f4e6e2a ON |
1380 | p->exit_state = EXIT_ZOMBIE; |
1381 | p = NULL; | |
1da177e4 LT |
1382 | } |
1383 | } | |
1384 | write_unlock_irq(&tasklist_lock); | |
1385 | } | |
1386 | if (p != NULL) | |
1387 | release_task(p); | |
2f4e6e2a | 1388 | |
1da177e4 LT |
1389 | return retval; |
1390 | } | |
1391 | ||
1392 | /* | |
1393 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1394 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1395 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1396 | * released the lock and the system call should return. | |
1397 | */ | |
f470021a | 1398 | static int wait_task_stopped(int ptrace, struct task_struct *p, |
98abed02 | 1399 | int options, struct siginfo __user *infop, |
1da177e4 LT |
1400 | int __user *stat_addr, struct rusage __user *ru) |
1401 | { | |
ee7c82da ON |
1402 | int retval, exit_code, why; |
1403 | uid_t uid = 0; /* unneeded, required by compiler */ | |
c8950783 | 1404 | pid_t pid; |
1da177e4 | 1405 | |
f470021a | 1406 | if (!(options & WUNTRACED)) |
98abed02 RM |
1407 | return 0; |
1408 | ||
ee7c82da ON |
1409 | exit_code = 0; |
1410 | spin_lock_irq(&p->sighand->siglock); | |
1411 | ||
1412 | if (unlikely(!task_is_stopped_or_traced(p))) | |
1413 | goto unlock_sig; | |
1414 | ||
f470021a | 1415 | if (!ptrace && p->signal->group_stop_count > 0) |
1da177e4 LT |
1416 | /* |
1417 | * A group stop is in progress and this is the group leader. | |
1418 | * We won't report until all threads have stopped. | |
1419 | */ | |
ee7c82da ON |
1420 | goto unlock_sig; |
1421 | ||
1422 | exit_code = p->exit_code; | |
1423 | if (!exit_code) | |
1424 | goto unlock_sig; | |
1425 | ||
98abed02 | 1426 | if (!unlikely(options & WNOWAIT)) |
ee7c82da ON |
1427 | p->exit_code = 0; |
1428 | ||
c69e8d9c DH |
1429 | /* don't need the RCU readlock here as we're holding a spinlock */ |
1430 | uid = __task_cred(p)->uid; | |
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 | ||
98abed02 | 1448 | if (unlikely(options & WNOWAIT)) |
1da177e4 | 1449 | return wait_noreap_copyout(p, pid, uid, |
e6ceb32a | 1450 | why, exit_code, |
1da177e4 | 1451 | infop, ru); |
1da177e4 LT |
1452 | |
1453 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1454 | if (!retval && stat_addr) | |
1455 | retval = put_user((exit_code << 8) | 0x7f, stat_addr); | |
1456 | if (!retval && infop) | |
1457 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1458 | if (!retval && infop) | |
1459 | retval = put_user(0, &infop->si_errno); | |
1460 | if (!retval && infop) | |
6efcae46 | 1461 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1462 | if (!retval && infop) |
1463 | retval = put_user(exit_code, &infop->si_status); | |
1464 | if (!retval && infop) | |
c8950783 | 1465 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1466 | if (!retval && infop) |
ee7c82da | 1467 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1468 | if (!retval) |
c8950783 | 1469 | retval = pid; |
1da177e4 LT |
1470 | put_task_struct(p); |
1471 | ||
1472 | BUG_ON(!retval); | |
1473 | return retval; | |
1474 | } | |
1475 | ||
1476 | /* | |
1477 | * Handle do_wait work for one task in a live, non-stopped state. | |
1478 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1479 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1480 | * released the lock and the system call should return. | |
1481 | */ | |
98abed02 | 1482 | static int wait_task_continued(struct task_struct *p, int options, |
1da177e4 LT |
1483 | struct siginfo __user *infop, |
1484 | int __user *stat_addr, struct rusage __user *ru) | |
1485 | { | |
1486 | int retval; | |
1487 | pid_t pid; | |
1488 | uid_t uid; | |
1489 | ||
98abed02 RM |
1490 | if (!unlikely(options & WCONTINUED)) |
1491 | return 0; | |
1492 | ||
1da177e4 LT |
1493 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1494 | return 0; | |
1495 | ||
1496 | spin_lock_irq(&p->sighand->siglock); | |
1497 | /* Re-check with the lock held. */ | |
1498 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1499 | spin_unlock_irq(&p->sighand->siglock); | |
1500 | return 0; | |
1501 | } | |
98abed02 | 1502 | if (!unlikely(options & WNOWAIT)) |
1da177e4 | 1503 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
c69e8d9c | 1504 | uid = __task_cred(p)->uid; |
1da177e4 LT |
1505 | spin_unlock_irq(&p->sighand->siglock); |
1506 | ||
6c5f3e7b | 1507 | pid = task_pid_vnr(p); |
1da177e4 LT |
1508 | get_task_struct(p); |
1509 | read_unlock(&tasklist_lock); | |
1510 | ||
1511 | if (!infop) { | |
1512 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1513 | put_task_struct(p); | |
1514 | if (!retval && stat_addr) | |
1515 | retval = put_user(0xffff, stat_addr); | |
1516 | if (!retval) | |
3a515e4a | 1517 | retval = pid; |
1da177e4 LT |
1518 | } else { |
1519 | retval = wait_noreap_copyout(p, pid, uid, | |
1520 | CLD_CONTINUED, SIGCONT, | |
1521 | infop, ru); | |
1522 | BUG_ON(retval == 0); | |
1523 | } | |
1524 | ||
1525 | return retval; | |
1526 | } | |
1527 | ||
98abed02 RM |
1528 | /* |
1529 | * Consider @p for a wait by @parent. | |
1530 | * | |
1531 | * -ECHILD should be in *@notask_error before the first call. | |
1532 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | |
1533 | * Returns zero if the search for a child should continue; | |
14dd0b81 RM |
1534 | * then *@notask_error is 0 if @p is an eligible child, |
1535 | * or another error from security_task_wait(), or still -ECHILD. | |
98abed02 | 1536 | */ |
f470021a | 1537 | static int wait_consider_task(struct task_struct *parent, int ptrace, |
98abed02 RM |
1538 | struct task_struct *p, int *notask_error, |
1539 | enum pid_type type, struct pid *pid, int options, | |
1540 | struct siginfo __user *infop, | |
1541 | int __user *stat_addr, struct rusage __user *ru) | |
1542 | { | |
1543 | int ret = eligible_child(type, pid, options, p); | |
14dd0b81 | 1544 | if (!ret) |
98abed02 RM |
1545 | return ret; |
1546 | ||
14dd0b81 RM |
1547 | if (unlikely(ret < 0)) { |
1548 | /* | |
1549 | * If we have not yet seen any eligible child, | |
1550 | * then let this error code replace -ECHILD. | |
1551 | * A permission error will give the user a clue | |
1552 | * to look for security policy problems, rather | |
1553 | * than for mysterious wait bugs. | |
1554 | */ | |
1555 | if (*notask_error) | |
1556 | *notask_error = ret; | |
1557 | } | |
1558 | ||
f470021a RM |
1559 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1560 | /* | |
1561 | * This child is hidden by ptrace. | |
1562 | * We aren't allowed to see it now, but eventually we will. | |
1563 | */ | |
1564 | *notask_error = 0; | |
1565 | return 0; | |
1566 | } | |
1567 | ||
98abed02 RM |
1568 | if (p->exit_state == EXIT_DEAD) |
1569 | return 0; | |
1570 | ||
1571 | /* | |
1572 | * We don't reap group leaders with subthreads. | |
1573 | */ | |
1574 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) | |
1575 | return wait_task_zombie(p, options, infop, stat_addr, ru); | |
1576 | ||
1577 | /* | |
1578 | * It's stopped or running now, so it might | |
1579 | * later continue, exit, or stop again. | |
1580 | */ | |
1581 | *notask_error = 0; | |
1582 | ||
1583 | if (task_is_stopped_or_traced(p)) | |
f470021a RM |
1584 | return wait_task_stopped(ptrace, p, options, |
1585 | infop, stat_addr, ru); | |
98abed02 RM |
1586 | |
1587 | return wait_task_continued(p, options, infop, stat_addr, ru); | |
1588 | } | |
1589 | ||
1590 | /* | |
1591 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1592 | * | |
1593 | * -ECHILD should be in *@notask_error before the first call. | |
1594 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | |
1595 | * Returns zero if the search for a child should continue; then | |
14dd0b81 RM |
1596 | * *@notask_error is 0 if there were any eligible children, |
1597 | * or another error from security_task_wait(), or still -ECHILD. | |
98abed02 RM |
1598 | */ |
1599 | static int do_wait_thread(struct task_struct *tsk, int *notask_error, | |
1600 | enum pid_type type, struct pid *pid, int options, | |
1601 | struct siginfo __user *infop, int __user *stat_addr, | |
1602 | struct rusage __user *ru) | |
1603 | { | |
1604 | struct task_struct *p; | |
1605 | ||
1606 | list_for_each_entry(p, &tsk->children, sibling) { | |
f470021a RM |
1607 | /* |
1608 | * Do not consider detached threads. | |
1609 | */ | |
1610 | if (!task_detached(p)) { | |
1611 | int ret = wait_consider_task(tsk, 0, p, notask_error, | |
1612 | type, pid, options, | |
1613 | infop, stat_addr, ru); | |
1614 | if (ret) | |
1615 | return ret; | |
1616 | } | |
98abed02 RM |
1617 | } |
1618 | ||
1619 | return 0; | |
1620 | } | |
1621 | ||
1622 | static int ptrace_do_wait(struct task_struct *tsk, int *notask_error, | |
1623 | enum pid_type type, struct pid *pid, int options, | |
1624 | struct siginfo __user *infop, int __user *stat_addr, | |
1625 | struct rusage __user *ru) | |
1626 | { | |
1627 | struct task_struct *p; | |
1628 | ||
1629 | /* | |
f470021a | 1630 | * Traditionally we see ptrace'd stopped tasks regardless of options. |
98abed02 | 1631 | */ |
f470021a | 1632 | options |= WUNTRACED; |
98abed02 | 1633 | |
f470021a RM |
1634 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
1635 | int ret = wait_consider_task(tsk, 1, p, notask_error, | |
1636 | type, pid, options, | |
1637 | infop, stat_addr, ru); | |
1638 | if (ret) | |
98abed02 | 1639 | return ret; |
98abed02 RM |
1640 | } |
1641 | ||
1642 | return 0; | |
1643 | } | |
1644 | ||
161550d7 EB |
1645 | static long do_wait(enum pid_type type, struct pid *pid, int options, |
1646 | struct siginfo __user *infop, int __user *stat_addr, | |
1647 | struct rusage __user *ru) | |
1da177e4 LT |
1648 | { |
1649 | DECLARE_WAITQUEUE(wait, current); | |
1650 | struct task_struct *tsk; | |
98abed02 | 1651 | int retval; |
1da177e4 | 1652 | |
0a16b607 MD |
1653 | trace_sched_process_wait(pid); |
1654 | ||
1da177e4 LT |
1655 | add_wait_queue(¤t->signal->wait_chldexit,&wait); |
1656 | repeat: | |
98abed02 RM |
1657 | /* |
1658 | * If there is nothing that can match our critiera just get out. | |
1659 | * We will clear @retval to zero if we see any child that might later | |
1660 | * match our criteria, even if we are not able to reap it yet. | |
1661 | */ | |
161550d7 EB |
1662 | retval = -ECHILD; |
1663 | if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type]))) | |
1664 | goto end; | |
1665 | ||
1da177e4 LT |
1666 | current->state = TASK_INTERRUPTIBLE; |
1667 | read_lock(&tasklist_lock); | |
1668 | tsk = current; | |
1669 | do { | |
98abed02 RM |
1670 | int tsk_result = do_wait_thread(tsk, &retval, |
1671 | type, pid, options, | |
1672 | infop, stat_addr, ru); | |
1673 | if (!tsk_result) | |
1674 | tsk_result = ptrace_do_wait(tsk, &retval, | |
1675 | type, pid, options, | |
1676 | infop, stat_addr, ru); | |
1677 | if (tsk_result) { | |
1678 | /* | |
1679 | * tasklist_lock is unlocked and we have a final result. | |
1680 | */ | |
1681 | retval = tsk_result; | |
1682 | goto end; | |
1da177e4 | 1683 | } |
98abed02 | 1684 | |
1da177e4 LT |
1685 | if (options & __WNOTHREAD) |
1686 | break; | |
1687 | tsk = next_thread(tsk); | |
125e1874 | 1688 | BUG_ON(tsk->signal != current->signal); |
1da177e4 | 1689 | } while (tsk != current); |
1da177e4 | 1690 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1691 | |
98abed02 | 1692 | if (!retval && !(options & WNOHANG)) { |
1da177e4 | 1693 | retval = -ERESTARTSYS; |
98abed02 RM |
1694 | if (!signal_pending(current)) { |
1695 | schedule(); | |
1696 | goto repeat; | |
1697 | } | |
1da177e4 | 1698 | } |
98abed02 | 1699 | |
1da177e4 LT |
1700 | end: |
1701 | current->state = TASK_RUNNING; | |
1702 | remove_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1703 | if (infop) { | |
1704 | if (retval > 0) | |
9cbab810 | 1705 | retval = 0; |
1da177e4 LT |
1706 | else { |
1707 | /* | |
1708 | * For a WNOHANG return, clear out all the fields | |
1709 | * we would set so the user can easily tell the | |
1710 | * difference. | |
1711 | */ | |
1712 | if (!retval) | |
1713 | retval = put_user(0, &infop->si_signo); | |
1714 | if (!retval) | |
1715 | retval = put_user(0, &infop->si_errno); | |
1716 | if (!retval) | |
1717 | retval = put_user(0, &infop->si_code); | |
1718 | if (!retval) | |
1719 | retval = put_user(0, &infop->si_pid); | |
1720 | if (!retval) | |
1721 | retval = put_user(0, &infop->si_uid); | |
1722 | if (!retval) | |
1723 | retval = put_user(0, &infop->si_status); | |
1724 | } | |
1725 | } | |
1726 | return retval; | |
1727 | } | |
1728 | ||
17da2bd9 HC |
1729 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
1730 | infop, int, options, struct rusage __user *, ru) | |
1da177e4 | 1731 | { |
161550d7 EB |
1732 | struct pid *pid = NULL; |
1733 | enum pid_type type; | |
1da177e4 LT |
1734 | long ret; |
1735 | ||
1736 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1737 | return -EINVAL; | |
1738 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1739 | return -EINVAL; | |
1740 | ||
1741 | switch (which) { | |
1742 | case P_ALL: | |
161550d7 | 1743 | type = PIDTYPE_MAX; |
1da177e4 LT |
1744 | break; |
1745 | case P_PID: | |
161550d7 EB |
1746 | type = PIDTYPE_PID; |
1747 | if (upid <= 0) | |
1da177e4 LT |
1748 | return -EINVAL; |
1749 | break; | |
1750 | case P_PGID: | |
161550d7 EB |
1751 | type = PIDTYPE_PGID; |
1752 | if (upid <= 0) | |
1da177e4 | 1753 | return -EINVAL; |
1da177e4 LT |
1754 | break; |
1755 | default: | |
1756 | return -EINVAL; | |
1757 | } | |
1758 | ||
161550d7 EB |
1759 | if (type < PIDTYPE_MAX) |
1760 | pid = find_get_pid(upid); | |
1761 | ret = do_wait(type, pid, options, infop, NULL, ru); | |
1762 | put_pid(pid); | |
1da177e4 LT |
1763 | |
1764 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1765 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1766 | return ret; |
1767 | } | |
1768 | ||
754fe8d2 HC |
1769 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
1770 | int, options, struct rusage __user *, ru) | |
1da177e4 | 1771 | { |
161550d7 EB |
1772 | struct pid *pid = NULL; |
1773 | enum pid_type type; | |
1da177e4 LT |
1774 | long ret; |
1775 | ||
1776 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1777 | __WNOTHREAD|__WCLONE|__WALL)) | |
1778 | return -EINVAL; | |
161550d7 EB |
1779 | |
1780 | if (upid == -1) | |
1781 | type = PIDTYPE_MAX; | |
1782 | else if (upid < 0) { | |
1783 | type = PIDTYPE_PGID; | |
1784 | pid = find_get_pid(-upid); | |
1785 | } else if (upid == 0) { | |
1786 | type = PIDTYPE_PGID; | |
1787 | pid = get_pid(task_pgrp(current)); | |
1788 | } else /* upid > 0 */ { | |
1789 | type = PIDTYPE_PID; | |
1790 | pid = find_get_pid(upid); | |
1791 | } | |
1792 | ||
1793 | ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru); | |
1794 | put_pid(pid); | |
1da177e4 LT |
1795 | |
1796 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1797 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1798 | return ret; |
1799 | } | |
1800 | ||
1801 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1802 | ||
1803 | /* | |
1804 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1805 | * implemented by calling sys_wait4() from libc.a. | |
1806 | */ | |
17da2bd9 | 1807 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
1da177e4 LT |
1808 | { |
1809 | return sys_wait4(pid, stat_addr, options, NULL); | |
1810 | } | |
1811 | ||
1812 | #endif |