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> |
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 LT |
21 | #include <linux/file.h> |
22 | #include <linux/binfmts.h> | |
ab516013 | 23 | #include <linux/nsproxy.h> |
84d73786 | 24 | #include <linux/pid_namespace.h> |
1da177e4 LT |
25 | #include <linux/ptrace.h> |
26 | #include <linux/profile.h> | |
27 | #include <linux/mount.h> | |
28 | #include <linux/proc_fs.h> | |
49d769d5 | 29 | #include <linux/kthread.h> |
1da177e4 | 30 | #include <linux/mempolicy.h> |
c757249a | 31 | #include <linux/taskstats_kern.h> |
ca74e92b | 32 | #include <linux/delayacct.h> |
83144186 | 33 | #include <linux/freezer.h> |
b4f48b63 | 34 | #include <linux/cgroup.h> |
1da177e4 | 35 | #include <linux/syscalls.h> |
7ed20e1a | 36 | #include <linux/signal.h> |
6a14c5c9 | 37 | #include <linux/posix-timers.h> |
9f46080c | 38 | #include <linux/cn_proc.h> |
de5097c2 | 39 | #include <linux/mutex.h> |
0771dfef | 40 | #include <linux/futex.h> |
34f192c6 | 41 | #include <linux/compat.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> |
1da177e4 LT |
47 | |
48 | #include <asm/uaccess.h> | |
49 | #include <asm/unistd.h> | |
50 | #include <asm/pgtable.h> | |
51 | #include <asm/mmu_context.h> | |
52 | ||
408b664a AB |
53 | static void exit_mm(struct task_struct * tsk); |
54 | ||
1da177e4 LT |
55 | static void __unhash_process(struct task_struct *p) |
56 | { | |
57 | nr_threads--; | |
58 | detach_pid(p, PIDTYPE_PID); | |
1da177e4 LT |
59 | if (thread_group_leader(p)) { |
60 | detach_pid(p, PIDTYPE_PGID); | |
61 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 62 | |
5e85d4ab | 63 | list_del_rcu(&p->tasks); |
73b9ebfe | 64 | __get_cpu_var(process_counts)--; |
1da177e4 | 65 | } |
47e65328 | 66 | list_del_rcu(&p->thread_group); |
c97d9893 | 67 | remove_parent(p); |
1da177e4 LT |
68 | } |
69 | ||
6a14c5c9 ON |
70 | /* |
71 | * This function expects the tasklist_lock write-locked. | |
72 | */ | |
73 | static void __exit_signal(struct task_struct *tsk) | |
74 | { | |
75 | struct signal_struct *sig = tsk->signal; | |
76 | struct sighand_struct *sighand; | |
77 | ||
78 | BUG_ON(!sig); | |
79 | BUG_ON(!atomic_read(&sig->count)); | |
80 | ||
81 | rcu_read_lock(); | |
82 | sighand = rcu_dereference(tsk->sighand); | |
83 | spin_lock(&sighand->siglock); | |
84 | ||
85 | posix_cpu_timers_exit(tsk); | |
86 | if (atomic_dec_and_test(&sig->count)) | |
87 | posix_cpu_timers_exit_group(tsk); | |
88 | else { | |
89 | /* | |
90 | * If there is any task waiting for the group exit | |
91 | * then notify it: | |
92 | */ | |
6db840fa | 93 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) |
6a14c5c9 | 94 | wake_up_process(sig->group_exit_task); |
6db840fa | 95 | |
6a14c5c9 ON |
96 | if (tsk == sig->curr_target) |
97 | sig->curr_target = next_thread(tsk); | |
98 | /* | |
99 | * Accumulate here the counters for all threads but the | |
100 | * group leader as they die, so they can be added into | |
101 | * the process-wide totals when those are taken. | |
102 | * The group leader stays around as a zombie as long | |
103 | * as there are other threads. When it gets reaped, | |
104 | * the exit.c code will add its counts into these totals. | |
105 | * We won't ever get here for the group leader, since it | |
106 | * will have been the last reference on the signal_struct. | |
107 | */ | |
108 | sig->utime = cputime_add(sig->utime, tsk->utime); | |
109 | sig->stime = cputime_add(sig->stime, tsk->stime); | |
9ac52315 | 110 | sig->gtime = cputime_add(sig->gtime, tsk->gtime); |
6a14c5c9 ON |
111 | sig->min_flt += tsk->min_flt; |
112 | sig->maj_flt += tsk->maj_flt; | |
113 | sig->nvcsw += tsk->nvcsw; | |
114 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
115 | sig->inblock += task_io_get_inblock(tsk); |
116 | sig->oublock += task_io_get_oublock(tsk); | |
172ba844 | 117 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
118 | sig = NULL; /* Marker for below. */ |
119 | } | |
120 | ||
5876700c ON |
121 | __unhash_process(tsk); |
122 | ||
6a14c5c9 | 123 | tsk->signal = NULL; |
a7e5328a | 124 | tsk->sighand = NULL; |
6a14c5c9 ON |
125 | spin_unlock(&sighand->siglock); |
126 | rcu_read_unlock(); | |
127 | ||
a7e5328a | 128 | __cleanup_sighand(sighand); |
6a14c5c9 ON |
129 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
130 | flush_sigqueue(&tsk->pending); | |
131 | if (sig) { | |
132 | flush_sigqueue(&sig->shared_pending); | |
093a8e8a | 133 | taskstats_tgid_free(sig); |
6a14c5c9 ON |
134 | __cleanup_signal(sig); |
135 | } | |
136 | } | |
137 | ||
8c7904a0 EB |
138 | static void delayed_put_task_struct(struct rcu_head *rhp) |
139 | { | |
140 | put_task_struct(container_of(rhp, struct task_struct, rcu)); | |
141 | } | |
142 | ||
1da177e4 LT |
143 | void release_task(struct task_struct * p) |
144 | { | |
36c8b586 | 145 | struct task_struct *leader; |
1da177e4 | 146 | int zap_leader; |
1f09f974 | 147 | repeat: |
1da177e4 | 148 | atomic_dec(&p->user->processes); |
60347f67 | 149 | proc_flush_task(p); |
1da177e4 | 150 | write_lock_irq(&tasklist_lock); |
1f09f974 | 151 | ptrace_unlink(p); |
1da177e4 LT |
152 | BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children)); |
153 | __exit_signal(p); | |
35f5cad8 | 154 | |
1da177e4 LT |
155 | /* |
156 | * If we are the last non-leader member of the thread | |
157 | * group, and the leader is zombie, then notify the | |
158 | * group leader's parent process. (if it wants notification.) | |
159 | */ | |
160 | zap_leader = 0; | |
161 | leader = p->group_leader; | |
162 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
163 | BUG_ON(leader->exit_signal == -1); | |
164 | do_notify_parent(leader, leader->exit_signal); | |
165 | /* | |
166 | * If we were the last child thread and the leader has | |
167 | * exited already, and the leader's parent ignores SIGCHLD, | |
168 | * then we are the one who should release the leader. | |
169 | * | |
170 | * do_notify_parent() will have marked it self-reaping in | |
171 | * that case. | |
172 | */ | |
173 | zap_leader = (leader->exit_signal == -1); | |
174 | } | |
175 | ||
1da177e4 | 176 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 177 | release_thread(p); |
8c7904a0 | 178 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
179 | |
180 | p = leader; | |
181 | if (unlikely(zap_leader)) | |
182 | goto repeat; | |
183 | } | |
184 | ||
1da177e4 LT |
185 | /* |
186 | * This checks not only the pgrp, but falls back on the pid if no | |
187 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
188 | * without this... | |
04a2e6a5 EB |
189 | * |
190 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 191 | */ |
04a2e6a5 | 192 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
193 | { |
194 | struct task_struct *p; | |
04a2e6a5 | 195 | struct pid *sid = NULL; |
62dfb554 | 196 | |
04a2e6a5 | 197 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 198 | if (p == NULL) |
04a2e6a5 | 199 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 200 | if (p != NULL) |
04a2e6a5 | 201 | sid = task_session(p); |
62dfb554 | 202 | |
1da177e4 LT |
203 | return sid; |
204 | } | |
205 | ||
206 | /* | |
207 | * Determine if a process group is "orphaned", according to the POSIX | |
208 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
209 | * by terminal-generated stop signals. Newly orphaned process groups are | |
210 | * to receive a SIGHUP and a SIGCONT. | |
211 | * | |
212 | * "I ask you, have you ever known what it is to be an orphan?" | |
213 | */ | |
0475ac08 | 214 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
215 | { |
216 | struct task_struct *p; | |
1da177e4 | 217 | |
0475ac08 | 218 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
219 | if ((p == ignored_task) || |
220 | (p->exit_state && thread_group_empty(p)) || | |
221 | is_global_init(p->real_parent)) | |
1da177e4 | 222 | continue; |
05e83df6 | 223 | |
0475ac08 | 224 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
225 | task_session(p->real_parent) == task_session(p)) |
226 | return 0; | |
0475ac08 | 227 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
228 | |
229 | return 1; | |
1da177e4 LT |
230 | } |
231 | ||
3e7cd6c4 | 232 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
233 | { |
234 | int retval; | |
235 | ||
236 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 237 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
238 | read_unlock(&tasklist_lock); |
239 | ||
240 | return retval; | |
241 | } | |
242 | ||
0475ac08 | 243 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
244 | { |
245 | int retval = 0; | |
246 | struct task_struct *p; | |
247 | ||
0475ac08 | 248 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
338077e5 | 249 | if (!task_is_stopped(p)) |
1da177e4 | 250 | continue; |
1da177e4 LT |
251 | retval = 1; |
252 | break; | |
0475ac08 | 253 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
254 | return retval; |
255 | } | |
256 | ||
f49ee505 ON |
257 | /* |
258 | * Check to see if any process groups have become orphaned as | |
259 | * a result of our exiting, and if they have any stopped jobs, | |
260 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
261 | */ | |
262 | static void | |
263 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
264 | { | |
265 | struct pid *pgrp = task_pgrp(tsk); | |
266 | struct task_struct *ignored_task = tsk; | |
267 | ||
268 | if (!parent) | |
269 | /* exit: our father is in a different pgrp than | |
270 | * we are and we were the only connection outside. | |
271 | */ | |
272 | parent = tsk->real_parent; | |
273 | else | |
274 | /* reparent: our child is in a different pgrp than | |
275 | * we are, and it was the only connection outside. | |
276 | */ | |
277 | ignored_task = NULL; | |
278 | ||
279 | if (task_pgrp(parent) != pgrp && | |
280 | task_session(parent) == task_session(tsk) && | |
281 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
282 | has_stopped_jobs(pgrp)) { | |
283 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
284 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
285 | } | |
286 | } | |
287 | ||
1da177e4 | 288 | /** |
49d769d5 | 289 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
290 | * |
291 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
292 | * it ever exits, it should generally reparent itself to kthreadd so it |
293 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
294 | * |
295 | * The various task state such as scheduling policy and priority may have | |
296 | * been inherited from a user process, so we reset them to sane values here. | |
297 | * | |
49d769d5 | 298 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 299 | */ |
49d769d5 | 300 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
301 | { |
302 | write_lock_irq(&tasklist_lock); | |
303 | ||
304 | ptrace_unlink(current); | |
305 | /* Reparent to init */ | |
9b678ece | 306 | remove_parent(current); |
49d769d5 | 307 | current->real_parent = current->parent = kthreadd_task; |
9b678ece | 308 | add_parent(current); |
1da177e4 LT |
309 | |
310 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
311 | current->exit_signal = SIGCHLD; | |
312 | ||
e05606d3 | 313 | if (task_nice(current) < 0) |
1da177e4 LT |
314 | set_user_nice(current, 0); |
315 | /* cpus_allowed? */ | |
316 | /* rt_priority? */ | |
317 | /* signals? */ | |
318 | security_task_reparent_to_init(current); | |
319 | memcpy(current->signal->rlim, init_task.signal->rlim, | |
320 | sizeof(current->signal->rlim)); | |
321 | atomic_inc(&(INIT_USER->__count)); | |
322 | write_unlock_irq(&tasklist_lock); | |
323 | switch_uid(INIT_USER); | |
324 | } | |
325 | ||
8520d7c7 | 326 | void __set_special_pids(struct pid *pid) |
1da177e4 | 327 | { |
e19f247a | 328 | struct task_struct *curr = current->group_leader; |
8520d7c7 | 329 | pid_t nr = pid_nr(pid); |
1da177e4 | 330 | |
8520d7c7 | 331 | if (task_session(curr) != pid) { |
1da177e4 | 332 | detach_pid(curr, PIDTYPE_SID); |
8520d7c7 ON |
333 | attach_pid(curr, PIDTYPE_SID, pid); |
334 | set_task_session(curr, nr); | |
1da177e4 | 335 | } |
8520d7c7 | 336 | if (task_pgrp(curr) != pid) { |
1da177e4 | 337 | detach_pid(curr, PIDTYPE_PGID); |
8520d7c7 ON |
338 | attach_pid(curr, PIDTYPE_PGID, pid); |
339 | set_task_pgrp(curr, nr); | |
1da177e4 LT |
340 | } |
341 | } | |
342 | ||
8520d7c7 | 343 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
344 | { |
345 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 346 | __set_special_pids(pid); |
1da177e4 LT |
347 | write_unlock_irq(&tasklist_lock); |
348 | } | |
349 | ||
350 | /* | |
351 | * Let kernel threads use this to say that they | |
352 | * allow a certain signal (since daemonize() will | |
353 | * have disabled all of them by default). | |
354 | */ | |
355 | int allow_signal(int sig) | |
356 | { | |
7ed20e1a | 357 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
358 | return -EINVAL; |
359 | ||
360 | spin_lock_irq(¤t->sighand->siglock); | |
361 | sigdelset(¤t->blocked, sig); | |
362 | if (!current->mm) { | |
363 | /* Kernel threads handle their own signals. | |
364 | Let the signal code know it'll be handled, so | |
365 | that they don't get converted to SIGKILL or | |
366 | just silently dropped */ | |
367 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
368 | } | |
369 | recalc_sigpending(); | |
370 | spin_unlock_irq(¤t->sighand->siglock); | |
371 | return 0; | |
372 | } | |
373 | ||
374 | EXPORT_SYMBOL(allow_signal); | |
375 | ||
376 | int disallow_signal(int sig) | |
377 | { | |
7ed20e1a | 378 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
379 | return -EINVAL; |
380 | ||
381 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 382 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
383 | recalc_sigpending(); |
384 | spin_unlock_irq(¤t->sighand->siglock); | |
385 | return 0; | |
386 | } | |
387 | ||
388 | EXPORT_SYMBOL(disallow_signal); | |
389 | ||
390 | /* | |
391 | * Put all the gunge required to become a kernel thread without | |
392 | * attached user resources in one place where it belongs. | |
393 | */ | |
394 | ||
395 | void daemonize(const char *name, ...) | |
396 | { | |
397 | va_list args; | |
398 | struct fs_struct *fs; | |
399 | sigset_t blocked; | |
400 | ||
401 | va_start(args, name); | |
402 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
403 | va_end(args); | |
404 | ||
405 | /* | |
406 | * If we were started as result of loading a module, close all of the | |
407 | * user space pages. We don't need them, and if we didn't close them | |
408 | * they would be locked into memory. | |
409 | */ | |
410 | exit_mm(current); | |
83144186 RW |
411 | /* |
412 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
413 | * or suspend transition begins right now. | |
414 | */ | |
415 | current->flags |= PF_NOFREEZE; | |
1da177e4 | 416 | |
8520d7c7 ON |
417 | if (current->nsproxy != &init_nsproxy) { |
418 | get_nsproxy(&init_nsproxy); | |
419 | switch_task_namespaces(current, &init_nsproxy); | |
420 | } | |
297bd42b | 421 | set_special_pids(&init_struct_pid); |
24ec839c | 422 | proc_clear_tty(current); |
1da177e4 LT |
423 | |
424 | /* Block and flush all signals */ | |
425 | sigfillset(&blocked); | |
426 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
427 | flush_signals(current); | |
428 | ||
429 | /* Become as one with the init task */ | |
430 | ||
431 | exit_fs(current); /* current->fs->count--; */ | |
432 | fs = init_task.fs; | |
433 | current->fs = fs; | |
434 | atomic_inc(&fs->count); | |
ab516013 | 435 | |
d4c5e41f | 436 | exit_files(current); |
1da177e4 LT |
437 | current->files = init_task.files; |
438 | atomic_inc(¤t->files->count); | |
439 | ||
49d769d5 | 440 | reparent_to_kthreadd(); |
1da177e4 LT |
441 | } |
442 | ||
443 | EXPORT_SYMBOL(daemonize); | |
444 | ||
858119e1 | 445 | static void close_files(struct files_struct * files) |
1da177e4 LT |
446 | { |
447 | int i, j; | |
badf1662 | 448 | struct fdtable *fdt; |
1da177e4 LT |
449 | |
450 | j = 0; | |
4fb3a538 DS |
451 | |
452 | /* | |
453 | * It is safe to dereference the fd table without RCU or | |
454 | * ->file_lock because this is the last reference to the | |
455 | * files structure. | |
456 | */ | |
badf1662 | 457 | fdt = files_fdtable(files); |
1da177e4 LT |
458 | for (;;) { |
459 | unsigned long set; | |
460 | i = j * __NFDBITS; | |
bbea9f69 | 461 | if (i >= fdt->max_fds) |
1da177e4 | 462 | break; |
badf1662 | 463 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
464 | while (set) { |
465 | if (set & 1) { | |
badf1662 | 466 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 467 | if (file) { |
1da177e4 | 468 | filp_close(file, files); |
944be0b2 IM |
469 | cond_resched(); |
470 | } | |
1da177e4 LT |
471 | } |
472 | i++; | |
473 | set >>= 1; | |
474 | } | |
475 | } | |
476 | } | |
477 | ||
478 | struct files_struct *get_files_struct(struct task_struct *task) | |
479 | { | |
480 | struct files_struct *files; | |
481 | ||
482 | task_lock(task); | |
483 | files = task->files; | |
484 | if (files) | |
485 | atomic_inc(&files->count); | |
486 | task_unlock(task); | |
487 | ||
488 | return files; | |
489 | } | |
490 | ||
7ad5b3a5 | 491 | void put_files_struct(struct files_struct *files) |
1da177e4 | 492 | { |
badf1662 DS |
493 | struct fdtable *fdt; |
494 | ||
1da177e4 LT |
495 | if (atomic_dec_and_test(&files->count)) { |
496 | close_files(files); | |
497 | /* | |
498 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
499 | * If the fdtable was embedded, pass files for freeing |
500 | * at the end of the RCU grace period. Otherwise, | |
501 | * you can free files immediately. | |
1da177e4 | 502 | */ |
badf1662 | 503 | fdt = files_fdtable(files); |
4fd45812 | 504 | if (fdt != &files->fdtab) |
ab2af1f5 | 505 | kmem_cache_free(files_cachep, files); |
01b2d93c | 506 | free_fdtable(fdt); |
1da177e4 LT |
507 | } |
508 | } | |
509 | ||
3b125388 | 510 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 511 | { |
3b125388 | 512 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
513 | struct files_struct *old; |
514 | ||
515 | old = tsk->files; | |
516 | task_lock(tsk); | |
517 | tsk->files = files; | |
518 | task_unlock(tsk); | |
519 | put_files_struct(old); | |
520 | } | |
3b9b8ab6 | 521 | |
1ec7f1dd | 522 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
523 | { |
524 | struct files_struct * files = tsk->files; | |
525 | ||
526 | if (files) { | |
527 | task_lock(tsk); | |
528 | tsk->files = NULL; | |
529 | task_unlock(tsk); | |
530 | put_files_struct(files); | |
531 | } | |
532 | } | |
533 | ||
1ec7f1dd | 534 | void put_fs_struct(struct fs_struct *fs) |
1da177e4 LT |
535 | { |
536 | /* No need to hold fs->lock if we are killing it */ | |
537 | if (atomic_dec_and_test(&fs->count)) { | |
6ac08c39 JB |
538 | path_put(&fs->root); |
539 | path_put(&fs->pwd); | |
540 | if (fs->altroot.dentry) | |
541 | path_put(&fs->altroot); | |
1da177e4 LT |
542 | kmem_cache_free(fs_cachep, fs); |
543 | } | |
544 | } | |
545 | ||
1ec7f1dd | 546 | void exit_fs(struct task_struct *tsk) |
1da177e4 LT |
547 | { |
548 | struct fs_struct * fs = tsk->fs; | |
549 | ||
550 | if (fs) { | |
551 | task_lock(tsk); | |
552 | tsk->fs = NULL; | |
553 | task_unlock(tsk); | |
1ec7f1dd | 554 | put_fs_struct(fs); |
1da177e4 LT |
555 | } |
556 | } | |
557 | ||
1da177e4 LT |
558 | EXPORT_SYMBOL_GPL(exit_fs); |
559 | ||
cf475ad2 BS |
560 | #ifdef CONFIG_MM_OWNER |
561 | /* | |
562 | * Task p is exiting and it owned mm, lets find a new owner for it | |
563 | */ | |
564 | static inline int | |
565 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) | |
566 | { | |
567 | /* | |
568 | * If there are other users of the mm and the owner (us) is exiting | |
569 | * we need to find a new owner to take on the responsibility. | |
570 | */ | |
571 | if (!mm) | |
572 | return 0; | |
573 | if (atomic_read(&mm->mm_users) <= 1) | |
574 | return 0; | |
575 | if (mm->owner != p) | |
576 | return 0; | |
577 | return 1; | |
578 | } | |
579 | ||
580 | void mm_update_next_owner(struct mm_struct *mm) | |
581 | { | |
582 | struct task_struct *c, *g, *p = current; | |
583 | ||
584 | retry: | |
585 | if (!mm_need_new_owner(mm, p)) | |
586 | return; | |
587 | ||
588 | read_lock(&tasklist_lock); | |
589 | /* | |
590 | * Search in the children | |
591 | */ | |
592 | list_for_each_entry(c, &p->children, sibling) { | |
593 | if (c->mm == mm) | |
594 | goto assign_new_owner; | |
595 | } | |
596 | ||
597 | /* | |
598 | * Search in the siblings | |
599 | */ | |
600 | list_for_each_entry(c, &p->parent->children, sibling) { | |
601 | if (c->mm == mm) | |
602 | goto assign_new_owner; | |
603 | } | |
604 | ||
605 | /* | |
606 | * Search through everything else. We should not get | |
607 | * here often | |
608 | */ | |
609 | do_each_thread(g, c) { | |
610 | if (c->mm == mm) | |
611 | goto assign_new_owner; | |
612 | } while_each_thread(g, c); | |
613 | ||
614 | read_unlock(&tasklist_lock); | |
615 | return; | |
616 | ||
617 | assign_new_owner: | |
618 | BUG_ON(c == p); | |
619 | get_task_struct(c); | |
620 | /* | |
621 | * The task_lock protects c->mm from changing. | |
622 | * We always want mm->owner->mm == mm | |
623 | */ | |
624 | task_lock(c); | |
625 | /* | |
626 | * Delay read_unlock() till we have the task_lock() | |
627 | * to ensure that c does not slip away underneath us | |
628 | */ | |
629 | read_unlock(&tasklist_lock); | |
630 | if (c->mm != mm) { | |
631 | task_unlock(c); | |
632 | put_task_struct(c); | |
633 | goto retry; | |
634 | } | |
635 | cgroup_mm_owner_callbacks(mm->owner, c); | |
636 | mm->owner = c; | |
637 | task_unlock(c); | |
638 | put_task_struct(c); | |
639 | } | |
640 | #endif /* CONFIG_MM_OWNER */ | |
641 | ||
1da177e4 LT |
642 | /* |
643 | * Turn us into a lazy TLB process if we | |
644 | * aren't already.. | |
645 | */ | |
408b664a | 646 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
647 | { |
648 | struct mm_struct *mm = tsk->mm; | |
649 | ||
650 | mm_release(tsk, mm); | |
651 | if (!mm) | |
652 | return; | |
653 | /* | |
654 | * Serialize with any possible pending coredump. | |
655 | * We must hold mmap_sem around checking core_waiters | |
656 | * and clearing tsk->mm. The core-inducing thread | |
657 | * will increment core_waiters for each thread in the | |
658 | * group with ->mm != NULL. | |
659 | */ | |
660 | down_read(&mm->mmap_sem); | |
661 | if (mm->core_waiters) { | |
662 | up_read(&mm->mmap_sem); | |
663 | down_write(&mm->mmap_sem); | |
664 | if (!--mm->core_waiters) | |
665 | complete(mm->core_startup_done); | |
666 | up_write(&mm->mmap_sem); | |
667 | ||
668 | wait_for_completion(&mm->core_done); | |
669 | down_read(&mm->mmap_sem); | |
670 | } | |
671 | atomic_inc(&mm->mm_count); | |
125e1874 | 672 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
673 | /* more a memory barrier than a real lock */ |
674 | task_lock(tsk); | |
675 | tsk->mm = NULL; | |
676 | up_read(&mm->mmap_sem); | |
677 | enter_lazy_tlb(mm, current); | |
0c1eecfb RW |
678 | /* We don't want this task to be frozen prematurely */ |
679 | clear_freeze_flag(tsk); | |
1da177e4 | 680 | task_unlock(tsk); |
cf475ad2 | 681 | mm_update_next_owner(mm); |
1da177e4 LT |
682 | mmput(mm); |
683 | } | |
684 | ||
36c8b586 IM |
685 | static void |
686 | reparent_thread(struct task_struct *p, struct task_struct *father, int traced) | |
1da177e4 | 687 | { |
241ceee0 ON |
688 | if (p->pdeath_signal) |
689 | /* We already hold the tasklist_lock here. */ | |
690 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | |
691 | ||
1da177e4 LT |
692 | /* Move the child from its dying parent to the new one. */ |
693 | if (unlikely(traced)) { | |
694 | /* Preserve ptrace links if someone else is tracing this child. */ | |
695 | list_del_init(&p->ptrace_list); | |
696 | if (p->parent != p->real_parent) | |
697 | list_add(&p->ptrace_list, &p->real_parent->ptrace_children); | |
698 | } else { | |
699 | /* If this child is being traced, then we're the one tracing it | |
700 | * anyway, so let go of it. | |
701 | */ | |
702 | p->ptrace = 0; | |
6ac781b1 | 703 | remove_parent(p); |
1da177e4 | 704 | p->parent = p->real_parent; |
6ac781b1 | 705 | add_parent(p); |
1da177e4 | 706 | |
338077e5 | 707 | if (task_is_traced(p)) { |
1da177e4 LT |
708 | /* |
709 | * If it was at a trace stop, turn it into | |
710 | * a normal stop since it's no longer being | |
711 | * traced. | |
712 | */ | |
713 | ptrace_untrace(p); | |
714 | } | |
715 | } | |
716 | ||
b2b2cbc4 EB |
717 | /* If this is a threaded reparent there is no need to |
718 | * notify anyone anything has happened. | |
719 | */ | |
720 | if (p->real_parent->group_leader == father->group_leader) | |
721 | return; | |
722 | ||
723 | /* We don't want people slaying init. */ | |
724 | if (p->exit_signal != -1) | |
725 | p->exit_signal = SIGCHLD; | |
b2b2cbc4 EB |
726 | |
727 | /* If we'd notified the old parent about this child's death, | |
728 | * also notify the new parent. | |
729 | */ | |
730 | if (!traced && p->exit_state == EXIT_ZOMBIE && | |
731 | p->exit_signal != -1 && thread_group_empty(p)) | |
732 | do_notify_parent(p, p->exit_signal); | |
733 | ||
f49ee505 | 734 | kill_orphaned_pgrp(p, father); |
1da177e4 LT |
735 | } |
736 | ||
737 | /* | |
738 | * When we die, we re-parent all our children. | |
739 | * Try to give them to another thread in our thread | |
740 | * group, and if no such member exists, give it to | |
84d73786 SB |
741 | * the child reaper process (ie "init") in our pid |
742 | * space. | |
1da177e4 | 743 | */ |
762a24be | 744 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 745 | { |
03ff1797 | 746 | struct task_struct *p, *n, *reaper = father; |
762a24be ON |
747 | struct list_head ptrace_dead; |
748 | ||
749 | INIT_LIST_HEAD(&ptrace_dead); | |
750 | ||
751 | write_lock_irq(&tasklist_lock); | |
1da177e4 LT |
752 | |
753 | do { | |
754 | reaper = next_thread(reaper); | |
755 | if (reaper == father) { | |
88f21d81 | 756 | reaper = task_child_reaper(father); |
1da177e4 LT |
757 | break; |
758 | } | |
762a24be | 759 | } while (reaper->flags & PF_EXITING); |
1da177e4 LT |
760 | |
761 | /* | |
762 | * There are only two places where our children can be: | |
763 | * | |
764 | * - in our child list | |
765 | * - in our ptraced child list | |
766 | * | |
767 | * Search them and reparent children. | |
768 | */ | |
03ff1797 | 769 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
1da177e4 | 770 | int ptrace; |
1da177e4 LT |
771 | |
772 | ptrace = p->ptrace; | |
773 | ||
774 | /* if father isn't the real parent, then ptrace must be enabled */ | |
775 | BUG_ON(father != p->real_parent && !ptrace); | |
776 | ||
777 | if (father == p->real_parent) { | |
778 | /* reparent with a reaper, real father it's us */ | |
84eb646b | 779 | p->real_parent = reaper; |
1da177e4 LT |
780 | reparent_thread(p, father, 0); |
781 | } else { | |
782 | /* reparent ptraced task to its real parent */ | |
783 | __ptrace_unlink (p); | |
784 | if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 && | |
785 | thread_group_empty(p)) | |
786 | do_notify_parent(p, p->exit_signal); | |
787 | } | |
788 | ||
789 | /* | |
790 | * if the ptraced child is a zombie with exit_signal == -1 | |
791 | * we must collect it before we exit, or it will remain | |
792 | * zombie forever since we prevented it from self-reap itself | |
793 | * while it was being traced by us, to be able to see it in wait4. | |
794 | */ | |
795 | if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1)) | |
762a24be | 796 | list_add(&p->ptrace_list, &ptrace_dead); |
1da177e4 | 797 | } |
762a24be | 798 | |
03ff1797 | 799 | list_for_each_entry_safe(p, n, &father->ptrace_children, ptrace_list) { |
84eb646b | 800 | p->real_parent = reaper; |
1da177e4 LT |
801 | reparent_thread(p, father, 1); |
802 | } | |
762a24be ON |
803 | |
804 | write_unlock_irq(&tasklist_lock); | |
805 | BUG_ON(!list_empty(&father->children)); | |
806 | BUG_ON(!list_empty(&father->ptrace_children)); | |
807 | ||
808 | list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_list) { | |
809 | list_del_init(&p->ptrace_list); | |
810 | release_task(p); | |
811 | } | |
812 | ||
1da177e4 LT |
813 | } |
814 | ||
815 | /* | |
816 | * Send signals to all our closest relatives so that they know | |
817 | * to properly mourn us.. | |
818 | */ | |
821c7de7 | 819 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 LT |
820 | { |
821 | int state; | |
1da177e4 | 822 | |
1da177e4 LT |
823 | /* |
824 | * This does two things: | |
825 | * | |
826 | * A. Make init inherit all the child processes | |
827 | * B. Check to see if any process groups have become orphaned | |
828 | * as a result of our exiting, and if they have any stopped | |
829 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
830 | */ | |
762a24be | 831 | forget_original_parent(tsk); |
2e4a7072 | 832 | exit_task_namespaces(tsk); |
1da177e4 | 833 | |
762a24be | 834 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
835 | if (group_dead) |
836 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 837 | |
24728448 | 838 | /* Let father know we died |
1da177e4 LT |
839 | * |
840 | * Thread signals are configurable, but you aren't going to use | |
d4c5e41f | 841 | * that to send signals to arbitary processes. |
1da177e4 LT |
842 | * That stops right now. |
843 | * | |
844 | * If the parent exec id doesn't match the exec id we saved | |
845 | * when we started then we know the parent has changed security | |
846 | * domain. | |
847 | * | |
848 | * If our self_exec id doesn't match our parent_exec_id then | |
849 | * we have changed execution domain as these two values started | |
850 | * the same after a fork. | |
1da177e4 | 851 | */ |
1da177e4 | 852 | if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 && |
f49ee505 ON |
853 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
854 | tsk->self_exec_id != tsk->parent_exec_id) | |
1da177e4 LT |
855 | && !capable(CAP_KILL)) |
856 | tsk->exit_signal = SIGCHLD; | |
857 | ||
858 | ||
859 | /* If something other than our normal parent is ptracing us, then | |
860 | * send it a SIGCHLD instead of honoring exit_signal. exit_signal | |
861 | * only has special meaning to our real parent. | |
862 | */ | |
863 | if (tsk->exit_signal != -1 && thread_group_empty(tsk)) { | |
864 | int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD; | |
865 | do_notify_parent(tsk, signal); | |
866 | } else if (tsk->ptrace) { | |
867 | do_notify_parent(tsk, SIGCHLD); | |
868 | } | |
869 | ||
870 | state = EXIT_ZOMBIE; | |
24728448 | 871 | if (tsk->exit_signal == -1 && likely(!tsk->ptrace)) |
1da177e4 LT |
872 | state = EXIT_DEAD; |
873 | tsk->exit_state = state; | |
874 | ||
6db840fa ON |
875 | if (thread_group_leader(tsk) && |
876 | tsk->signal->notify_count < 0 && | |
877 | tsk->signal->group_exit_task) | |
878 | wake_up_process(tsk->signal->group_exit_task); | |
879 | ||
1da177e4 LT |
880 | write_unlock_irq(&tasklist_lock); |
881 | ||
1da177e4 LT |
882 | /* If the process is dead, release it - nobody will wait for it */ |
883 | if (state == EXIT_DEAD) | |
884 | release_task(tsk); | |
1da177e4 LT |
885 | } |
886 | ||
e18eecb8 JD |
887 | #ifdef CONFIG_DEBUG_STACK_USAGE |
888 | static void check_stack_usage(void) | |
889 | { | |
890 | static DEFINE_SPINLOCK(low_water_lock); | |
891 | static int lowest_to_date = THREAD_SIZE; | |
892 | unsigned long *n = end_of_stack(current); | |
893 | unsigned long free; | |
894 | ||
895 | while (*n == 0) | |
896 | n++; | |
897 | free = (unsigned long)n - (unsigned long)end_of_stack(current); | |
898 | ||
899 | if (free >= lowest_to_date) | |
900 | return; | |
901 | ||
902 | spin_lock(&low_water_lock); | |
903 | if (free < lowest_to_date) { | |
904 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
905 | "left\n", | |
906 | current->comm, free); | |
907 | lowest_to_date = free; | |
908 | } | |
909 | spin_unlock(&low_water_lock); | |
910 | } | |
911 | #else | |
912 | static inline void check_stack_usage(void) {} | |
913 | #endif | |
914 | ||
84eb646b ON |
915 | static inline void exit_child_reaper(struct task_struct *tsk) |
916 | { | |
88f21d81 | 917 | if (likely(tsk->group_leader != task_child_reaper(tsk))) |
84eb646b ON |
918 | return; |
919 | ||
3eb07c8c SB |
920 | if (tsk->nsproxy->pid_ns == &init_pid_ns) |
921 | panic("Attempted to kill init!"); | |
922 | ||
923 | /* | |
924 | * @tsk is the last thread in the 'cgroup-init' and is exiting. | |
925 | * Terminate all remaining processes in the namespace and reap them | |
926 | * before exiting @tsk. | |
927 | * | |
928 | * Note that @tsk (last thread of cgroup-init) may not necessarily | |
929 | * be the child-reaper (i.e main thread of cgroup-init) of the | |
930 | * namespace i.e the child_reaper may have already exited. | |
931 | * | |
932 | * Even after a child_reaper exits, we let it inherit orphaned children, | |
933 | * because, pid_ns->child_reaper remains valid as long as there is | |
934 | * at least one living sub-thread in the cgroup init. | |
935 | ||
936 | * This living sub-thread of the cgroup-init will be notified when | |
937 | * a child inherited by the 'child-reaper' exits (do_notify_parent() | |
938 | * uses __group_send_sig_info()). Further, when reaping child processes, | |
939 | * do_wait() iterates over children of all living sub threads. | |
940 | ||
941 | * i.e even though 'child_reaper' thread is listed as the parent of the | |
942 | * orphaned children, any living sub-thread in the cgroup-init can | |
943 | * perform the role of the child_reaper. | |
944 | */ | |
945 | zap_pid_ns_processes(tsk->nsproxy->pid_ns); | |
84eb646b ON |
946 | } |
947 | ||
7ad5b3a5 | 948 | NORET_TYPE void do_exit(long code) |
1da177e4 LT |
949 | { |
950 | struct task_struct *tsk = current; | |
951 | int group_dead; | |
952 | ||
953 | profile_task_exit(tsk); | |
954 | ||
22e2c507 JA |
955 | WARN_ON(atomic_read(&tsk->fs_excl)); |
956 | ||
1da177e4 LT |
957 | if (unlikely(in_interrupt())) |
958 | panic("Aiee, killing interrupt handler!"); | |
959 | if (unlikely(!tsk->pid)) | |
960 | panic("Attempted to kill the idle task!"); | |
1da177e4 LT |
961 | |
962 | if (unlikely(current->ptrace & PT_TRACE_EXIT)) { | |
963 | current->ptrace_message = code; | |
964 | ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP); | |
965 | } | |
966 | ||
df164db5 AN |
967 | /* |
968 | * We're taking recursive faults here in do_exit. Safest is to just | |
969 | * leave this task alone and wait for reboot. | |
970 | */ | |
971 | if (unlikely(tsk->flags & PF_EXITING)) { | |
972 | printk(KERN_ALERT | |
973 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
974 | /* |
975 | * We can do this unlocked here. The futex code uses | |
976 | * this flag just to verify whether the pi state | |
977 | * cleanup has been done or not. In the worst case it | |
978 | * loops once more. We pretend that the cleanup was | |
979 | * done as there is no way to return. Either the | |
980 | * OWNER_DIED bit is set by now or we push the blocked | |
981 | * task into the wait for ever nirwana as well. | |
982 | */ | |
983 | tsk->flags |= PF_EXITPIDONE; | |
afc847b7 AV |
984 | if (tsk->io_context) |
985 | exit_io_context(); | |
df164db5 AN |
986 | set_current_state(TASK_UNINTERRUPTIBLE); |
987 | schedule(); | |
988 | } | |
989 | ||
d12619b5 | 990 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
991 | /* |
992 | * tsk->flags are checked in the futex code to protect against | |
993 | * an exiting task cleaning up the robust pi futexes. | |
994 | */ | |
d2ee7198 ON |
995 | smp_mb(); |
996 | spin_unlock_wait(&tsk->pi_lock); | |
1da177e4 | 997 | |
1da177e4 LT |
998 | if (unlikely(in_atomic())) |
999 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 1000 | current->comm, task_pid_nr(current), |
1da177e4 LT |
1001 | preempt_count()); |
1002 | ||
1003 | acct_update_integrals(tsk); | |
365e9c87 HD |
1004 | if (tsk->mm) { |
1005 | update_hiwater_rss(tsk->mm); | |
1006 | update_hiwater_vm(tsk->mm); | |
1007 | } | |
1da177e4 | 1008 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 1009 | if (group_dead) { |
84eb646b | 1010 | exit_child_reaper(tsk); |
778e9a9c | 1011 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 1012 | exit_itimers(tsk->signal); |
c3068951 | 1013 | } |
f6ec29a4 | 1014 | acct_collect(code, group_dead); |
42b2dd0a | 1015 | #ifdef CONFIG_FUTEX |
0771dfef IM |
1016 | if (unlikely(tsk->robust_list)) |
1017 | exit_robust_list(tsk); | |
42b2dd0a | 1018 | #ifdef CONFIG_COMPAT |
34f192c6 IM |
1019 | if (unlikely(tsk->compat_robust_list)) |
1020 | compat_exit_robust_list(tsk); | |
42b2dd0a | 1021 | #endif |
34f192c6 | 1022 | #endif |
522ed776 MT |
1023 | if (group_dead) |
1024 | tty_audit_exit(); | |
fa84cb93 AV |
1025 | if (unlikely(tsk->audit_context)) |
1026 | audit_free(tsk); | |
115085ea | 1027 | |
f2ab6d88 | 1028 | tsk->exit_code = code; |
115085ea | 1029 | taskstats_exit(tsk, group_dead); |
c757249a | 1030 | |
1da177e4 LT |
1031 | exit_mm(tsk); |
1032 | ||
0e464814 | 1033 | if (group_dead) |
f6ec29a4 | 1034 | acct_process(); |
1da177e4 | 1035 | exit_sem(tsk); |
1ec7f1dd AV |
1036 | exit_files(tsk); |
1037 | exit_fs(tsk); | |
e18eecb8 | 1038 | check_stack_usage(); |
1da177e4 | 1039 | exit_thread(); |
b4f48b63 | 1040 | cgroup_exit(tsk, 1); |
1da177e4 LT |
1041 | exit_keys(tsk); |
1042 | ||
1043 | if (group_dead && tsk->signal->leader) | |
1044 | disassociate_ctty(1); | |
1045 | ||
a1261f54 | 1046 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 LT |
1047 | if (tsk->binfmt) |
1048 | module_put(tsk->binfmt->module); | |
1049 | ||
9f46080c | 1050 | proc_exit_connector(tsk); |
821c7de7 | 1051 | exit_notify(tsk, group_dead); |
1da177e4 | 1052 | #ifdef CONFIG_NUMA |
f0be3d32 | 1053 | mpol_put(tsk->mempolicy); |
1da177e4 LT |
1054 | tsk->mempolicy = NULL; |
1055 | #endif | |
42b2dd0a | 1056 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
1057 | /* |
1058 | * This must happen late, after the PID is not | |
1059 | * hashed anymore: | |
1060 | */ | |
1061 | if (unlikely(!list_empty(&tsk->pi_state_list))) | |
1062 | exit_pi_state_list(tsk); | |
1063 | if (unlikely(current->pi_state_cache)) | |
1064 | kfree(current->pi_state_cache); | |
42b2dd0a | 1065 | #endif |
de5097c2 | 1066 | /* |
9a11b49a | 1067 | * Make sure we are holding no locks: |
de5097c2 | 1068 | */ |
9a11b49a | 1069 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1070 | /* |
1071 | * We can do this unlocked here. The futex code uses this flag | |
1072 | * just to verify whether the pi state cleanup has been done | |
1073 | * or not. In the worst case it loops once more. | |
1074 | */ | |
1075 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1076 | |
afc847b7 AV |
1077 | if (tsk->io_context) |
1078 | exit_io_context(); | |
1079 | ||
b92ce558 JA |
1080 | if (tsk->splice_pipe) |
1081 | __free_pipe_info(tsk->splice_pipe); | |
1082 | ||
7407251a | 1083 | preempt_disable(); |
55a101f8 | 1084 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1085 | tsk->state = TASK_DEAD; |
7407251a | 1086 | |
1da177e4 LT |
1087 | schedule(); |
1088 | BUG(); | |
1089 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1090 | for (;;) |
1091 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1092 | } |
1093 | ||
012914da RA |
1094 | EXPORT_SYMBOL_GPL(do_exit); |
1095 | ||
1da177e4 LT |
1096 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1097 | { | |
1098 | if (comp) | |
1099 | complete(comp); | |
55a101f8 | 1100 | |
1da177e4 LT |
1101 | do_exit(code); |
1102 | } | |
1103 | ||
1104 | EXPORT_SYMBOL(complete_and_exit); | |
1105 | ||
1106 | asmlinkage long sys_exit(int error_code) | |
1107 | { | |
1108 | do_exit((error_code&0xff)<<8); | |
1109 | } | |
1110 | ||
1da177e4 LT |
1111 | /* |
1112 | * Take down every thread in the group. This is called by fatal signals | |
1113 | * as well as by sys_exit_group (below). | |
1114 | */ | |
1115 | NORET_TYPE void | |
1116 | do_group_exit(int exit_code) | |
1117 | { | |
bfc4b089 ON |
1118 | struct signal_struct *sig = current->signal; |
1119 | ||
1da177e4 LT |
1120 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1121 | ||
bfc4b089 ON |
1122 | if (signal_group_exit(sig)) |
1123 | exit_code = sig->group_exit_code; | |
1da177e4 | 1124 | else if (!thread_group_empty(current)) { |
1da177e4 | 1125 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1126 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1127 | if (signal_group_exit(sig)) |
1da177e4 LT |
1128 | /* Another thread got here before we took the lock. */ |
1129 | exit_code = sig->group_exit_code; | |
1130 | else { | |
1da177e4 | 1131 | sig->group_exit_code = exit_code; |
ed5d2cac | 1132 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1133 | zap_other_threads(current); |
1134 | } | |
1135 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1136 | } |
1137 | ||
1138 | do_exit(exit_code); | |
1139 | /* NOTREACHED */ | |
1140 | } | |
1141 | ||
1142 | /* | |
1143 | * this kills every thread in the thread group. Note that any externally | |
1144 | * wait4()-ing process will get the correct exit code - even if this | |
1145 | * thread is not the thread group leader. | |
1146 | */ | |
1147 | asmlinkage void sys_exit_group(int error_code) | |
1148 | { | |
1149 | do_group_exit((error_code & 0xff) << 8); | |
1150 | } | |
1151 | ||
161550d7 EB |
1152 | static struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
1153 | { | |
1154 | struct pid *pid = NULL; | |
1155 | if (type == PIDTYPE_PID) | |
1156 | pid = task->pids[type].pid; | |
1157 | else if (type < PIDTYPE_MAX) | |
1158 | pid = task->group_leader->pids[type].pid; | |
1159 | return pid; | |
1160 | } | |
1161 | ||
1162 | static int eligible_child(enum pid_type type, struct pid *pid, int options, | |
1163 | struct task_struct *p) | |
1da177e4 | 1164 | { |
73243284 RM |
1165 | int err; |
1166 | ||
161550d7 EB |
1167 | if (type < PIDTYPE_MAX) { |
1168 | if (task_pid_type(p, type) != pid) | |
1da177e4 LT |
1169 | return 0; |
1170 | } | |
1171 | ||
1172 | /* | |
1173 | * Do not consider detached threads that are | |
1174 | * not ptraced: | |
1175 | */ | |
1176 | if (p->exit_signal == -1 && !p->ptrace) | |
1177 | return 0; | |
1178 | ||
1179 | /* Wait for all children (clone and not) if __WALL is set; | |
1180 | * otherwise, wait for clone children *only* if __WCLONE is | |
1181 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1182 | * A "clone" child here is one that reports to its parent | |
1183 | * using a signal other than SIGCHLD.) */ | |
1184 | if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0)) | |
1185 | && !(options & __WALL)) | |
1186 | return 0; | |
1da177e4 | 1187 | |
73243284 | 1188 | err = security_task_wait(p); |
f2cc3eb1 ON |
1189 | if (likely(!err)) |
1190 | return 1; | |
1da177e4 | 1191 | |
161550d7 | 1192 | if (type != PIDTYPE_PID) |
f2cc3eb1 ON |
1193 | return 0; |
1194 | /* This child was explicitly requested, abort */ | |
1195 | read_unlock(&tasklist_lock); | |
1196 | return err; | |
1da177e4 LT |
1197 | } |
1198 | ||
36c8b586 | 1199 | static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid, |
1da177e4 LT |
1200 | int why, int status, |
1201 | struct siginfo __user *infop, | |
1202 | struct rusage __user *rusagep) | |
1203 | { | |
1204 | int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0; | |
36c8b586 | 1205 | |
1da177e4 LT |
1206 | put_task_struct(p); |
1207 | if (!retval) | |
1208 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1209 | if (!retval) | |
1210 | retval = put_user(0, &infop->si_errno); | |
1211 | if (!retval) | |
1212 | retval = put_user((short)why, &infop->si_code); | |
1213 | if (!retval) | |
1214 | retval = put_user(pid, &infop->si_pid); | |
1215 | if (!retval) | |
1216 | retval = put_user(uid, &infop->si_uid); | |
1217 | if (!retval) | |
1218 | retval = put_user(status, &infop->si_status); | |
1219 | if (!retval) | |
1220 | retval = pid; | |
1221 | return retval; | |
1222 | } | |
1223 | ||
1224 | /* | |
1225 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1226 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1227 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1228 | * released the lock and the system call should return. | |
1229 | */ | |
36c8b586 | 1230 | static int wait_task_zombie(struct task_struct *p, int noreap, |
1da177e4 LT |
1231 | struct siginfo __user *infop, |
1232 | int __user *stat_addr, struct rusage __user *ru) | |
1233 | { | |
1234 | unsigned long state; | |
2f4e6e2a | 1235 | int retval, status, traced; |
6c5f3e7b | 1236 | pid_t pid = task_pid_vnr(p); |
1da177e4 LT |
1237 | |
1238 | if (unlikely(noreap)) { | |
1da177e4 LT |
1239 | uid_t uid = p->uid; |
1240 | int exit_code = p->exit_code; | |
1241 | int why, status; | |
1242 | ||
1da177e4 LT |
1243 | get_task_struct(p); |
1244 | read_unlock(&tasklist_lock); | |
1245 | if ((exit_code & 0x7f) == 0) { | |
1246 | why = CLD_EXITED; | |
1247 | status = exit_code >> 8; | |
1248 | } else { | |
1249 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1250 | status = exit_code & 0x7f; | |
1251 | } | |
1252 | return wait_noreap_copyout(p, pid, uid, why, | |
1253 | status, infop, ru); | |
1254 | } | |
1255 | ||
1256 | /* | |
1257 | * Try to move the task's state to DEAD | |
1258 | * only one thread is allowed to do this: | |
1259 | */ | |
1260 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1261 | if (state != EXIT_ZOMBIE) { | |
1262 | BUG_ON(state != EXIT_DEAD); | |
1263 | return 0; | |
1264 | } | |
1da177e4 | 1265 | |
2f4e6e2a ON |
1266 | /* traced means p->ptrace, but not vice versa */ |
1267 | traced = (p->real_parent != p->parent); | |
1268 | ||
1269 | if (likely(!traced)) { | |
3795e161 JJ |
1270 | struct signal_struct *psig; |
1271 | struct signal_struct *sig; | |
1272 | ||
1da177e4 LT |
1273 | /* |
1274 | * The resource counters for the group leader are in its | |
1275 | * own task_struct. Those for dead threads in the group | |
1276 | * are in its signal_struct, as are those for the child | |
1277 | * processes it has previously reaped. All these | |
1278 | * accumulate in the parent's signal_struct c* fields. | |
1279 | * | |
1280 | * We don't bother to take a lock here to protect these | |
1281 | * p->signal fields, because they are only touched by | |
1282 | * __exit_signal, which runs with tasklist_lock | |
1283 | * write-locked anyway, and so is excluded here. We do | |
1284 | * need to protect the access to p->parent->signal fields, | |
1285 | * as other threads in the parent group can be right | |
1286 | * here reaping other children at the same time. | |
1287 | */ | |
1288 | spin_lock_irq(&p->parent->sighand->siglock); | |
3795e161 JJ |
1289 | psig = p->parent->signal; |
1290 | sig = p->signal; | |
1291 | psig->cutime = | |
1292 | cputime_add(psig->cutime, | |
1da177e4 | 1293 | cputime_add(p->utime, |
3795e161 JJ |
1294 | cputime_add(sig->utime, |
1295 | sig->cutime))); | |
1296 | psig->cstime = | |
1297 | cputime_add(psig->cstime, | |
1da177e4 | 1298 | cputime_add(p->stime, |
3795e161 JJ |
1299 | cputime_add(sig->stime, |
1300 | sig->cstime))); | |
9ac52315 LV |
1301 | psig->cgtime = |
1302 | cputime_add(psig->cgtime, | |
1303 | cputime_add(p->gtime, | |
1304 | cputime_add(sig->gtime, | |
1305 | sig->cgtime))); | |
3795e161 JJ |
1306 | psig->cmin_flt += |
1307 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1308 | psig->cmaj_flt += | |
1309 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1310 | psig->cnvcsw += | |
1311 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1312 | psig->cnivcsw += | |
1313 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1314 | psig->cinblock += |
1315 | task_io_get_inblock(p) + | |
1316 | sig->inblock + sig->cinblock; | |
1317 | psig->coublock += | |
1318 | task_io_get_oublock(p) + | |
1319 | sig->oublock + sig->coublock; | |
1da177e4 LT |
1320 | spin_unlock_irq(&p->parent->sighand->siglock); |
1321 | } | |
1322 | ||
1323 | /* | |
1324 | * Now we are sure this task is interesting, and no other | |
1325 | * thread can reap it because we set its state to EXIT_DEAD. | |
1326 | */ | |
1327 | read_unlock(&tasklist_lock); | |
1328 | ||
1329 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1330 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) | |
1331 | ? p->signal->group_exit_code : p->exit_code; | |
1332 | if (!retval && stat_addr) | |
1333 | retval = put_user(status, stat_addr); | |
1334 | if (!retval && infop) | |
1335 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1336 | if (!retval && infop) | |
1337 | retval = put_user(0, &infop->si_errno); | |
1338 | if (!retval && infop) { | |
1339 | int why; | |
1340 | ||
1341 | if ((status & 0x7f) == 0) { | |
1342 | why = CLD_EXITED; | |
1343 | status >>= 8; | |
1344 | } else { | |
1345 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1346 | status &= 0x7f; | |
1347 | } | |
1348 | retval = put_user((short)why, &infop->si_code); | |
1349 | if (!retval) | |
1350 | retval = put_user(status, &infop->si_status); | |
1351 | } | |
1352 | if (!retval && infop) | |
3a515e4a | 1353 | retval = put_user(pid, &infop->si_pid); |
1da177e4 LT |
1354 | if (!retval && infop) |
1355 | retval = put_user(p->uid, &infop->si_uid); | |
2f4e6e2a | 1356 | if (!retval) |
3a515e4a | 1357 | retval = pid; |
2f4e6e2a ON |
1358 | |
1359 | if (traced) { | |
1da177e4 | 1360 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1361 | /* We dropped tasklist, ptracer could die and untrace */ |
1362 | ptrace_unlink(p); | |
1363 | /* | |
1364 | * If this is not a detached task, notify the parent. | |
1365 | * If it's still not detached after that, don't release | |
1366 | * it now. | |
1367 | */ | |
1368 | if (p->exit_signal != -1) { | |
1369 | do_notify_parent(p, p->exit_signal); | |
1da177e4 | 1370 | if (p->exit_signal != -1) { |
2f4e6e2a ON |
1371 | p->exit_state = EXIT_ZOMBIE; |
1372 | p = NULL; | |
1da177e4 LT |
1373 | } |
1374 | } | |
1375 | write_unlock_irq(&tasklist_lock); | |
1376 | } | |
1377 | if (p != NULL) | |
1378 | release_task(p); | |
2f4e6e2a | 1379 | |
1da177e4 LT |
1380 | return retval; |
1381 | } | |
1382 | ||
1383 | /* | |
1384 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1385 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1386 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1387 | * released the lock and the system call should return. | |
1388 | */ | |
1bad95c3 | 1389 | static int wait_task_stopped(struct task_struct *p, |
36c8b586 | 1390 | int noreap, struct siginfo __user *infop, |
1da177e4 LT |
1391 | int __user *stat_addr, struct rusage __user *ru) |
1392 | { | |
ee7c82da ON |
1393 | int retval, exit_code, why; |
1394 | uid_t uid = 0; /* unneeded, required by compiler */ | |
c8950783 | 1395 | pid_t pid; |
1da177e4 | 1396 | |
ee7c82da ON |
1397 | exit_code = 0; |
1398 | spin_lock_irq(&p->sighand->siglock); | |
1399 | ||
1400 | if (unlikely(!task_is_stopped_or_traced(p))) | |
1401 | goto unlock_sig; | |
1402 | ||
1bad95c3 | 1403 | if (!(p->ptrace & PT_PTRACED) && p->signal->group_stop_count > 0) |
1da177e4 LT |
1404 | /* |
1405 | * A group stop is in progress and this is the group leader. | |
1406 | * We won't report until all threads have stopped. | |
1407 | */ | |
ee7c82da ON |
1408 | goto unlock_sig; |
1409 | ||
1410 | exit_code = p->exit_code; | |
1411 | if (!exit_code) | |
1412 | goto unlock_sig; | |
1413 | ||
1414 | if (!noreap) | |
1415 | p->exit_code = 0; | |
1416 | ||
1417 | uid = p->uid; | |
1418 | unlock_sig: | |
1419 | spin_unlock_irq(&p->sighand->siglock); | |
1420 | if (!exit_code) | |
1da177e4 LT |
1421 | return 0; |
1422 | ||
1423 | /* | |
1424 | * Now we are pretty sure this task is interesting. | |
1425 | * Make sure it doesn't get reaped out from under us while we | |
1426 | * give up the lock and then examine it below. We don't want to | |
1427 | * keep holding onto the tasklist_lock while we call getrusage and | |
1428 | * possibly take page faults for user memory. | |
1429 | */ | |
1430 | get_task_struct(p); | |
6c5f3e7b | 1431 | pid = task_pid_vnr(p); |
ee7c82da | 1432 | why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1433 | read_unlock(&tasklist_lock); |
1434 | ||
ee7c82da | 1435 | if (unlikely(noreap)) |
1da177e4 | 1436 | return wait_noreap_copyout(p, pid, uid, |
e6ceb32a | 1437 | why, exit_code, |
1da177e4 | 1438 | infop, ru); |
1da177e4 LT |
1439 | |
1440 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1441 | if (!retval && stat_addr) | |
1442 | retval = put_user((exit_code << 8) | 0x7f, stat_addr); | |
1443 | if (!retval && infop) | |
1444 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1445 | if (!retval && infop) | |
1446 | retval = put_user(0, &infop->si_errno); | |
1447 | if (!retval && infop) | |
6efcae46 | 1448 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1449 | if (!retval && infop) |
1450 | retval = put_user(exit_code, &infop->si_status); | |
1451 | if (!retval && infop) | |
c8950783 | 1452 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1453 | if (!retval && infop) |
ee7c82da | 1454 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1455 | if (!retval) |
c8950783 | 1456 | retval = pid; |
1da177e4 LT |
1457 | put_task_struct(p); |
1458 | ||
1459 | BUG_ON(!retval); | |
1460 | return retval; | |
1461 | } | |
1462 | ||
1463 | /* | |
1464 | * Handle do_wait work for one task in a live, non-stopped state. | |
1465 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1466 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1467 | * released the lock and the system call should return. | |
1468 | */ | |
36c8b586 | 1469 | static int wait_task_continued(struct task_struct *p, int noreap, |
1da177e4 LT |
1470 | struct siginfo __user *infop, |
1471 | int __user *stat_addr, struct rusage __user *ru) | |
1472 | { | |
1473 | int retval; | |
1474 | pid_t pid; | |
1475 | uid_t uid; | |
1476 | ||
1da177e4 LT |
1477 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1478 | return 0; | |
1479 | ||
1480 | spin_lock_irq(&p->sighand->siglock); | |
1481 | /* Re-check with the lock held. */ | |
1482 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1483 | spin_unlock_irq(&p->sighand->siglock); | |
1484 | return 0; | |
1485 | } | |
1486 | if (!noreap) | |
1487 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; | |
1488 | spin_unlock_irq(&p->sighand->siglock); | |
1489 | ||
6c5f3e7b | 1490 | pid = task_pid_vnr(p); |
1da177e4 LT |
1491 | uid = p->uid; |
1492 | get_task_struct(p); | |
1493 | read_unlock(&tasklist_lock); | |
1494 | ||
1495 | if (!infop) { | |
1496 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1497 | put_task_struct(p); | |
1498 | if (!retval && stat_addr) | |
1499 | retval = put_user(0xffff, stat_addr); | |
1500 | if (!retval) | |
3a515e4a | 1501 | retval = pid; |
1da177e4 LT |
1502 | } else { |
1503 | retval = wait_noreap_copyout(p, pid, uid, | |
1504 | CLD_CONTINUED, SIGCONT, | |
1505 | infop, ru); | |
1506 | BUG_ON(retval == 0); | |
1507 | } | |
1508 | ||
1509 | return retval; | |
1510 | } | |
1511 | ||
161550d7 EB |
1512 | static long do_wait(enum pid_type type, struct pid *pid, int options, |
1513 | struct siginfo __user *infop, int __user *stat_addr, | |
1514 | struct rusage __user *ru) | |
1da177e4 LT |
1515 | { |
1516 | DECLARE_WAITQUEUE(wait, current); | |
1517 | struct task_struct *tsk; | |
1518 | int flag, retval; | |
1519 | ||
1520 | add_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1521 | repeat: | |
161550d7 EB |
1522 | /* If there is nothing that can match our critier just get out */ |
1523 | retval = -ECHILD; | |
1524 | if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type]))) | |
1525 | goto end; | |
1526 | ||
1da177e4 LT |
1527 | /* |
1528 | * We will set this flag if we see any child that might later | |
1529 | * match our criteria, even if we are not able to reap it yet. | |
1530 | */ | |
f2cc3eb1 | 1531 | flag = retval = 0; |
1da177e4 LT |
1532 | current->state = TASK_INTERRUPTIBLE; |
1533 | read_lock(&tasklist_lock); | |
1534 | tsk = current; | |
1535 | do { | |
1536 | struct task_struct *p; | |
1da177e4 | 1537 | |
03ff1797 | 1538 | list_for_each_entry(p, &tsk->children, sibling) { |
161550d7 | 1539 | int ret = eligible_child(type, pid, options, p); |
1da177e4 LT |
1540 | if (!ret) |
1541 | continue; | |
1542 | ||
73243284 | 1543 | if (unlikely(ret < 0)) { |
f2cc3eb1 ON |
1544 | retval = ret; |
1545 | } else if (task_is_stopped_or_traced(p)) { | |
1da177e4 LT |
1546 | /* |
1547 | * It's stopped now, so it might later | |
1548 | * continue, exit, or stop again. | |
1549 | */ | |
1550 | flag = 1; | |
34a1738f ON |
1551 | if (!(p->ptrace & PT_PTRACED) && |
1552 | !(options & WUNTRACED)) | |
1553 | continue; | |
338077e5 | 1554 | |
1bad95c3 | 1555 | retval = wait_task_stopped(p, |
338077e5 MW |
1556 | (options & WNOWAIT), infop, |
1557 | stat_addr, ru); | |
96fabbf5 ON |
1558 | } else if (p->exit_state == EXIT_ZOMBIE && |
1559 | !delay_group_leader(p)) { | |
338077e5 | 1560 | /* |
96fabbf5 | 1561 | * We don't reap group leaders with subthreads. |
338077e5 | 1562 | */ |
338077e5 | 1563 | if (!likely(options & WEXITED)) |
1da177e4 | 1564 | continue; |
338077e5 MW |
1565 | retval = wait_task_zombie(p, |
1566 | (options & WNOWAIT), infop, | |
1567 | stat_addr, ru); | |
0a76fe8e | 1568 | } else if (p->exit_state != EXIT_DEAD) { |
1da177e4 LT |
1569 | /* |
1570 | * It's running now, so it might later | |
1571 | * exit, stop, or stop and then continue. | |
1572 | */ | |
1573 | flag = 1; | |
1574 | if (!unlikely(options & WCONTINUED)) | |
1575 | continue; | |
338077e5 MW |
1576 | retval = wait_task_continued(p, |
1577 | (options & WNOWAIT), infop, | |
1578 | stat_addr, ru); | |
1da177e4 | 1579 | } |
9cbab810 ON |
1580 | if (retval != 0) /* tasklist_lock released */ |
1581 | goto end; | |
1da177e4 LT |
1582 | } |
1583 | if (!flag) { | |
03ff1797 | 1584 | list_for_each_entry(p, &tsk->ptrace_children, |
f2cc3eb1 | 1585 | ptrace_list) { |
161550d7 | 1586 | flag = eligible_child(type, pid, options, p); |
f2cc3eb1 | 1587 | if (!flag) |
1da177e4 | 1588 | continue; |
f2cc3eb1 ON |
1589 | if (likely(flag > 0)) |
1590 | break; | |
1591 | retval = flag; | |
1592 | goto end; | |
1da177e4 LT |
1593 | } |
1594 | } | |
1595 | if (options & __WNOTHREAD) | |
1596 | break; | |
1597 | tsk = next_thread(tsk); | |
125e1874 | 1598 | BUG_ON(tsk->signal != current->signal); |
1da177e4 | 1599 | } while (tsk != current); |
1da177e4 | 1600 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1601 | |
1da177e4 | 1602 | if (flag) { |
1da177e4 LT |
1603 | if (options & WNOHANG) |
1604 | goto end; | |
1605 | retval = -ERESTARTSYS; | |
1606 | if (signal_pending(current)) | |
1607 | goto end; | |
1608 | schedule(); | |
1609 | goto repeat; | |
1610 | } | |
1611 | retval = -ECHILD; | |
1612 | end: | |
1613 | current->state = TASK_RUNNING; | |
1614 | remove_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1615 | if (infop) { | |
1616 | if (retval > 0) | |
9cbab810 | 1617 | retval = 0; |
1da177e4 LT |
1618 | else { |
1619 | /* | |
1620 | * For a WNOHANG return, clear out all the fields | |
1621 | * we would set so the user can easily tell the | |
1622 | * difference. | |
1623 | */ | |
1624 | if (!retval) | |
1625 | retval = put_user(0, &infop->si_signo); | |
1626 | if (!retval) | |
1627 | retval = put_user(0, &infop->si_errno); | |
1628 | if (!retval) | |
1629 | retval = put_user(0, &infop->si_code); | |
1630 | if (!retval) | |
1631 | retval = put_user(0, &infop->si_pid); | |
1632 | if (!retval) | |
1633 | retval = put_user(0, &infop->si_uid); | |
1634 | if (!retval) | |
1635 | retval = put_user(0, &infop->si_status); | |
1636 | } | |
1637 | } | |
1638 | return retval; | |
1639 | } | |
1640 | ||
161550d7 | 1641 | asmlinkage long sys_waitid(int which, pid_t upid, |
1da177e4 LT |
1642 | struct siginfo __user *infop, int options, |
1643 | struct rusage __user *ru) | |
1644 | { | |
161550d7 EB |
1645 | struct pid *pid = NULL; |
1646 | enum pid_type type; | |
1da177e4 LT |
1647 | long ret; |
1648 | ||
1649 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1650 | return -EINVAL; | |
1651 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1652 | return -EINVAL; | |
1653 | ||
1654 | switch (which) { | |
1655 | case P_ALL: | |
161550d7 | 1656 | type = PIDTYPE_MAX; |
1da177e4 LT |
1657 | break; |
1658 | case P_PID: | |
161550d7 EB |
1659 | type = PIDTYPE_PID; |
1660 | if (upid <= 0) | |
1da177e4 LT |
1661 | return -EINVAL; |
1662 | break; | |
1663 | case P_PGID: | |
161550d7 EB |
1664 | type = PIDTYPE_PGID; |
1665 | if (upid <= 0) | |
1da177e4 | 1666 | return -EINVAL; |
1da177e4 LT |
1667 | break; |
1668 | default: | |
1669 | return -EINVAL; | |
1670 | } | |
1671 | ||
161550d7 EB |
1672 | if (type < PIDTYPE_MAX) |
1673 | pid = find_get_pid(upid); | |
1674 | ret = do_wait(type, pid, options, infop, NULL, ru); | |
1675 | put_pid(pid); | |
1da177e4 LT |
1676 | |
1677 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1678 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1679 | return ret; |
1680 | } | |
1681 | ||
161550d7 | 1682 | asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr, |
1da177e4 LT |
1683 | int options, struct rusage __user *ru) |
1684 | { | |
161550d7 EB |
1685 | struct pid *pid = NULL; |
1686 | enum pid_type type; | |
1da177e4 LT |
1687 | long ret; |
1688 | ||
1689 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1690 | __WNOTHREAD|__WCLONE|__WALL)) | |
1691 | return -EINVAL; | |
161550d7 EB |
1692 | |
1693 | if (upid == -1) | |
1694 | type = PIDTYPE_MAX; | |
1695 | else if (upid < 0) { | |
1696 | type = PIDTYPE_PGID; | |
1697 | pid = find_get_pid(-upid); | |
1698 | } else if (upid == 0) { | |
1699 | type = PIDTYPE_PGID; | |
1700 | pid = get_pid(task_pgrp(current)); | |
1701 | } else /* upid > 0 */ { | |
1702 | type = PIDTYPE_PID; | |
1703 | pid = find_get_pid(upid); | |
1704 | } | |
1705 | ||
1706 | ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru); | |
1707 | put_pid(pid); | |
1da177e4 LT |
1708 | |
1709 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1710 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1711 | return ret; |
1712 | } | |
1713 | ||
1714 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1715 | ||
1716 | /* | |
1717 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1718 | * implemented by calling sys_wait4() from libc.a. | |
1719 | */ | |
1720 | asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options) | |
1721 | { | |
1722 | return sys_wait4(pid, stat_addr, options, NULL); | |
1723 | } | |
1724 | ||
1725 | #endif |