]>
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> |
34f192c6 | 43 | #include <linux/compat.h> |
b92ce558 | 44 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 45 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 46 | #include <linux/resource.h> |
0d67a46d | 47 | #include <linux/blkdev.h> |
6eaeeaba | 48 | #include <linux/task_io_accounting_ops.h> |
1da177e4 LT |
49 | |
50 | #include <asm/uaccess.h> | |
51 | #include <asm/unistd.h> | |
52 | #include <asm/pgtable.h> | |
53 | #include <asm/mmu_context.h> | |
54 | ||
408b664a AB |
55 | static void exit_mm(struct task_struct * tsk); |
56 | ||
d839fd4d ON |
57 | static inline int task_detached(struct task_struct *p) |
58 | { | |
59 | return p->exit_signal == -1; | |
60 | } | |
61 | ||
1da177e4 LT |
62 | static void __unhash_process(struct task_struct *p) |
63 | { | |
64 | nr_threads--; | |
65 | detach_pid(p, PIDTYPE_PID); | |
1da177e4 LT |
66 | if (thread_group_leader(p)) { |
67 | detach_pid(p, PIDTYPE_PGID); | |
68 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 69 | |
5e85d4ab | 70 | list_del_rcu(&p->tasks); |
73b9ebfe | 71 | __get_cpu_var(process_counts)--; |
1da177e4 | 72 | } |
47e65328 | 73 | list_del_rcu(&p->thread_group); |
c97d9893 | 74 | remove_parent(p); |
1da177e4 LT |
75 | } |
76 | ||
6a14c5c9 ON |
77 | /* |
78 | * This function expects the tasklist_lock write-locked. | |
79 | */ | |
80 | static void __exit_signal(struct task_struct *tsk) | |
81 | { | |
82 | struct signal_struct *sig = tsk->signal; | |
83 | struct sighand_struct *sighand; | |
84 | ||
85 | BUG_ON(!sig); | |
86 | BUG_ON(!atomic_read(&sig->count)); | |
87 | ||
88 | rcu_read_lock(); | |
89 | sighand = rcu_dereference(tsk->sighand); | |
90 | spin_lock(&sighand->siglock); | |
91 | ||
92 | posix_cpu_timers_exit(tsk); | |
93 | if (atomic_dec_and_test(&sig->count)) | |
94 | posix_cpu_timers_exit_group(tsk); | |
95 | else { | |
96 | /* | |
97 | * If there is any task waiting for the group exit | |
98 | * then notify it: | |
99 | */ | |
6db840fa | 100 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) |
6a14c5c9 | 101 | wake_up_process(sig->group_exit_task); |
6db840fa | 102 | |
6a14c5c9 ON |
103 | if (tsk == sig->curr_target) |
104 | sig->curr_target = next_thread(tsk); | |
105 | /* | |
106 | * Accumulate here the counters for all threads but the | |
107 | * group leader as they die, so they can be added into | |
108 | * the process-wide totals when those are taken. | |
109 | * The group leader stays around as a zombie as long | |
110 | * as there are other threads. When it gets reaped, | |
111 | * the exit.c code will add its counts into these totals. | |
112 | * We won't ever get here for the group leader, since it | |
113 | * will have been the last reference on the signal_struct. | |
114 | */ | |
115 | sig->utime = cputime_add(sig->utime, tsk->utime); | |
116 | sig->stime = cputime_add(sig->stime, tsk->stime); | |
9ac52315 | 117 | sig->gtime = cputime_add(sig->gtime, tsk->gtime); |
6a14c5c9 ON |
118 | sig->min_flt += tsk->min_flt; |
119 | sig->maj_flt += tsk->maj_flt; | |
120 | sig->nvcsw += tsk->nvcsw; | |
121 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
122 | sig->inblock += task_io_get_inblock(tsk); |
123 | sig->oublock += task_io_get_oublock(tsk); | |
172ba844 | 124 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
125 | sig = NULL; /* Marker for below. */ |
126 | } | |
127 | ||
5876700c ON |
128 | __unhash_process(tsk); |
129 | ||
da7978b0 ON |
130 | /* |
131 | * Do this under ->siglock, we can race with another thread | |
132 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
133 | */ | |
134 | flush_sigqueue(&tsk->pending); | |
135 | ||
6a14c5c9 | 136 | tsk->signal = NULL; |
a7e5328a | 137 | tsk->sighand = NULL; |
6a14c5c9 ON |
138 | spin_unlock(&sighand->siglock); |
139 | rcu_read_unlock(); | |
140 | ||
a7e5328a | 141 | __cleanup_sighand(sighand); |
6a14c5c9 | 142 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
6a14c5c9 ON |
143 | if (sig) { |
144 | flush_sigqueue(&sig->shared_pending); | |
093a8e8a | 145 | taskstats_tgid_free(sig); |
6a14c5c9 ON |
146 | __cleanup_signal(sig); |
147 | } | |
148 | } | |
149 | ||
8c7904a0 EB |
150 | static void delayed_put_task_struct(struct rcu_head *rhp) |
151 | { | |
152 | put_task_struct(container_of(rhp, struct task_struct, rcu)); | |
153 | } | |
154 | ||
1da177e4 LT |
155 | void release_task(struct task_struct * p) |
156 | { | |
36c8b586 | 157 | struct task_struct *leader; |
1da177e4 | 158 | int zap_leader; |
1f09f974 | 159 | repeat: |
1da177e4 | 160 | atomic_dec(&p->user->processes); |
60347f67 | 161 | proc_flush_task(p); |
1da177e4 | 162 | write_lock_irq(&tasklist_lock); |
1f09f974 | 163 | ptrace_unlink(p); |
1da177e4 LT |
164 | BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children)); |
165 | __exit_signal(p); | |
35f5cad8 | 166 | |
1da177e4 LT |
167 | /* |
168 | * If we are the last non-leader member of the thread | |
169 | * group, and the leader is zombie, then notify the | |
170 | * group leader's parent process. (if it wants notification.) | |
171 | */ | |
172 | zap_leader = 0; | |
173 | leader = p->group_leader; | |
174 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
d839fd4d | 175 | BUG_ON(task_detached(leader)); |
1da177e4 LT |
176 | do_notify_parent(leader, leader->exit_signal); |
177 | /* | |
178 | * If we were the last child thread and the leader has | |
179 | * exited already, and the leader's parent ignores SIGCHLD, | |
180 | * then we are the one who should release the leader. | |
181 | * | |
182 | * do_notify_parent() will have marked it self-reaping in | |
183 | * that case. | |
184 | */ | |
d839fd4d | 185 | zap_leader = task_detached(leader); |
1da177e4 LT |
186 | } |
187 | ||
1da177e4 | 188 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 189 | release_thread(p); |
8c7904a0 | 190 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
191 | |
192 | p = leader; | |
193 | if (unlikely(zap_leader)) | |
194 | goto repeat; | |
195 | } | |
196 | ||
1da177e4 LT |
197 | /* |
198 | * This checks not only the pgrp, but falls back on the pid if no | |
199 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
200 | * without this... | |
04a2e6a5 EB |
201 | * |
202 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 203 | */ |
04a2e6a5 | 204 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
205 | { |
206 | struct task_struct *p; | |
04a2e6a5 | 207 | struct pid *sid = NULL; |
62dfb554 | 208 | |
04a2e6a5 | 209 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 210 | if (p == NULL) |
04a2e6a5 | 211 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 212 | if (p != NULL) |
04a2e6a5 | 213 | sid = task_session(p); |
62dfb554 | 214 | |
1da177e4 LT |
215 | return sid; |
216 | } | |
217 | ||
218 | /* | |
219 | * Determine if a process group is "orphaned", according to the POSIX | |
220 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
221 | * by terminal-generated stop signals. Newly orphaned process groups are | |
222 | * to receive a SIGHUP and a SIGCONT. | |
223 | * | |
224 | * "I ask you, have you ever known what it is to be an orphan?" | |
225 | */ | |
0475ac08 | 226 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
227 | { |
228 | struct task_struct *p; | |
1da177e4 | 229 | |
0475ac08 | 230 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
231 | if ((p == ignored_task) || |
232 | (p->exit_state && thread_group_empty(p)) || | |
233 | is_global_init(p->real_parent)) | |
1da177e4 | 234 | continue; |
05e83df6 | 235 | |
0475ac08 | 236 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
237 | task_session(p->real_parent) == task_session(p)) |
238 | return 0; | |
0475ac08 | 239 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
240 | |
241 | return 1; | |
1da177e4 LT |
242 | } |
243 | ||
3e7cd6c4 | 244 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
245 | { |
246 | int retval; | |
247 | ||
248 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 249 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
250 | read_unlock(&tasklist_lock); |
251 | ||
252 | return retval; | |
253 | } | |
254 | ||
0475ac08 | 255 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
256 | { |
257 | int retval = 0; | |
258 | struct task_struct *p; | |
259 | ||
0475ac08 | 260 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
338077e5 | 261 | if (!task_is_stopped(p)) |
1da177e4 | 262 | continue; |
1da177e4 LT |
263 | retval = 1; |
264 | break; | |
0475ac08 | 265 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
266 | return retval; |
267 | } | |
268 | ||
f49ee505 ON |
269 | /* |
270 | * Check to see if any process groups have become orphaned as | |
271 | * a result of our exiting, and if they have any stopped jobs, | |
272 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
273 | */ | |
274 | static void | |
275 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
276 | { | |
277 | struct pid *pgrp = task_pgrp(tsk); | |
278 | struct task_struct *ignored_task = tsk; | |
279 | ||
280 | if (!parent) | |
281 | /* exit: our father is in a different pgrp than | |
282 | * we are and we were the only connection outside. | |
283 | */ | |
284 | parent = tsk->real_parent; | |
285 | else | |
286 | /* reparent: our child is in a different pgrp than | |
287 | * we are, and it was the only connection outside. | |
288 | */ | |
289 | ignored_task = NULL; | |
290 | ||
291 | if (task_pgrp(parent) != pgrp && | |
292 | task_session(parent) == task_session(tsk) && | |
293 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
294 | has_stopped_jobs(pgrp)) { | |
295 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
296 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
297 | } | |
298 | } | |
299 | ||
1da177e4 | 300 | /** |
49d769d5 | 301 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
302 | * |
303 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
304 | * it ever exits, it should generally reparent itself to kthreadd so it |
305 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
306 | * |
307 | * The various task state such as scheduling policy and priority may have | |
308 | * been inherited from a user process, so we reset them to sane values here. | |
309 | * | |
49d769d5 | 310 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 311 | */ |
49d769d5 | 312 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
313 | { |
314 | write_lock_irq(&tasklist_lock); | |
315 | ||
316 | ptrace_unlink(current); | |
317 | /* Reparent to init */ | |
9b678ece | 318 | remove_parent(current); |
49d769d5 | 319 | current->real_parent = current->parent = kthreadd_task; |
9b678ece | 320 | add_parent(current); |
1da177e4 LT |
321 | |
322 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
323 | current->exit_signal = SIGCHLD; | |
324 | ||
e05606d3 | 325 | if (task_nice(current) < 0) |
1da177e4 LT |
326 | set_user_nice(current, 0); |
327 | /* cpus_allowed? */ | |
328 | /* rt_priority? */ | |
329 | /* signals? */ | |
330 | security_task_reparent_to_init(current); | |
331 | memcpy(current->signal->rlim, init_task.signal->rlim, | |
332 | sizeof(current->signal->rlim)); | |
333 | atomic_inc(&(INIT_USER->__count)); | |
334 | write_unlock_irq(&tasklist_lock); | |
335 | switch_uid(INIT_USER); | |
336 | } | |
337 | ||
8520d7c7 | 338 | void __set_special_pids(struct pid *pid) |
1da177e4 | 339 | { |
e19f247a | 340 | struct task_struct *curr = current->group_leader; |
8520d7c7 | 341 | pid_t nr = pid_nr(pid); |
1da177e4 | 342 | |
8520d7c7 | 343 | if (task_session(curr) != pid) { |
7d8da096 | 344 | change_pid(curr, PIDTYPE_SID, pid); |
8520d7c7 | 345 | set_task_session(curr, nr); |
1da177e4 | 346 | } |
8520d7c7 | 347 | if (task_pgrp(curr) != pid) { |
7d8da096 | 348 | change_pid(curr, PIDTYPE_PGID, pid); |
8520d7c7 | 349 | set_task_pgrp(curr, nr); |
1da177e4 LT |
350 | } |
351 | } | |
352 | ||
8520d7c7 | 353 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
354 | { |
355 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 356 | __set_special_pids(pid); |
1da177e4 LT |
357 | write_unlock_irq(&tasklist_lock); |
358 | } | |
359 | ||
360 | /* | |
361 | * Let kernel threads use this to say that they | |
362 | * allow a certain signal (since daemonize() will | |
363 | * have disabled all of them by default). | |
364 | */ | |
365 | int allow_signal(int sig) | |
366 | { | |
7ed20e1a | 367 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
368 | return -EINVAL; |
369 | ||
370 | spin_lock_irq(¤t->sighand->siglock); | |
371 | sigdelset(¤t->blocked, sig); | |
372 | if (!current->mm) { | |
373 | /* Kernel threads handle their own signals. | |
374 | Let the signal code know it'll be handled, so | |
375 | that they don't get converted to SIGKILL or | |
376 | just silently dropped */ | |
377 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
378 | } | |
379 | recalc_sigpending(); | |
380 | spin_unlock_irq(¤t->sighand->siglock); | |
381 | return 0; | |
382 | } | |
383 | ||
384 | EXPORT_SYMBOL(allow_signal); | |
385 | ||
386 | int disallow_signal(int sig) | |
387 | { | |
7ed20e1a | 388 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
389 | return -EINVAL; |
390 | ||
391 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 392 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
393 | recalc_sigpending(); |
394 | spin_unlock_irq(¤t->sighand->siglock); | |
395 | return 0; | |
396 | } | |
397 | ||
398 | EXPORT_SYMBOL(disallow_signal); | |
399 | ||
400 | /* | |
401 | * Put all the gunge required to become a kernel thread without | |
402 | * attached user resources in one place where it belongs. | |
403 | */ | |
404 | ||
405 | void daemonize(const char *name, ...) | |
406 | { | |
407 | va_list args; | |
408 | struct fs_struct *fs; | |
409 | sigset_t blocked; | |
410 | ||
411 | va_start(args, name); | |
412 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
413 | va_end(args); | |
414 | ||
415 | /* | |
416 | * If we were started as result of loading a module, close all of the | |
417 | * user space pages. We don't need them, and if we didn't close them | |
418 | * they would be locked into memory. | |
419 | */ | |
420 | exit_mm(current); | |
83144186 RW |
421 | /* |
422 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
423 | * or suspend transition begins right now. | |
424 | */ | |
425 | current->flags |= PF_NOFREEZE; | |
1da177e4 | 426 | |
8520d7c7 ON |
427 | if (current->nsproxy != &init_nsproxy) { |
428 | get_nsproxy(&init_nsproxy); | |
429 | switch_task_namespaces(current, &init_nsproxy); | |
430 | } | |
297bd42b | 431 | set_special_pids(&init_struct_pid); |
24ec839c | 432 | proc_clear_tty(current); |
1da177e4 LT |
433 | |
434 | /* Block and flush all signals */ | |
435 | sigfillset(&blocked); | |
436 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
437 | flush_signals(current); | |
438 | ||
439 | /* Become as one with the init task */ | |
440 | ||
441 | exit_fs(current); /* current->fs->count--; */ | |
442 | fs = init_task.fs; | |
443 | current->fs = fs; | |
444 | atomic_inc(&fs->count); | |
ab516013 | 445 | |
d4c5e41f | 446 | exit_files(current); |
1da177e4 LT |
447 | current->files = init_task.files; |
448 | atomic_inc(¤t->files->count); | |
449 | ||
49d769d5 | 450 | reparent_to_kthreadd(); |
1da177e4 LT |
451 | } |
452 | ||
453 | EXPORT_SYMBOL(daemonize); | |
454 | ||
858119e1 | 455 | static void close_files(struct files_struct * files) |
1da177e4 LT |
456 | { |
457 | int i, j; | |
badf1662 | 458 | struct fdtable *fdt; |
1da177e4 LT |
459 | |
460 | j = 0; | |
4fb3a538 DS |
461 | |
462 | /* | |
463 | * It is safe to dereference the fd table without RCU or | |
464 | * ->file_lock because this is the last reference to the | |
465 | * files structure. | |
466 | */ | |
badf1662 | 467 | fdt = files_fdtable(files); |
1da177e4 LT |
468 | for (;;) { |
469 | unsigned long set; | |
470 | i = j * __NFDBITS; | |
bbea9f69 | 471 | if (i >= fdt->max_fds) |
1da177e4 | 472 | break; |
badf1662 | 473 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
474 | while (set) { |
475 | if (set & 1) { | |
badf1662 | 476 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 477 | if (file) { |
1da177e4 | 478 | filp_close(file, files); |
944be0b2 IM |
479 | cond_resched(); |
480 | } | |
1da177e4 LT |
481 | } |
482 | i++; | |
483 | set >>= 1; | |
484 | } | |
485 | } | |
486 | } | |
487 | ||
488 | struct files_struct *get_files_struct(struct task_struct *task) | |
489 | { | |
490 | struct files_struct *files; | |
491 | ||
492 | task_lock(task); | |
493 | files = task->files; | |
494 | if (files) | |
495 | atomic_inc(&files->count); | |
496 | task_unlock(task); | |
497 | ||
498 | return files; | |
499 | } | |
500 | ||
7ad5b3a5 | 501 | void put_files_struct(struct files_struct *files) |
1da177e4 | 502 | { |
badf1662 DS |
503 | struct fdtable *fdt; |
504 | ||
1da177e4 LT |
505 | if (atomic_dec_and_test(&files->count)) { |
506 | close_files(files); | |
507 | /* | |
508 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
509 | * If the fdtable was embedded, pass files for freeing |
510 | * at the end of the RCU grace period. Otherwise, | |
511 | * you can free files immediately. | |
1da177e4 | 512 | */ |
badf1662 | 513 | fdt = files_fdtable(files); |
4fd45812 | 514 | if (fdt != &files->fdtab) |
ab2af1f5 | 515 | kmem_cache_free(files_cachep, files); |
01b2d93c | 516 | free_fdtable(fdt); |
1da177e4 LT |
517 | } |
518 | } | |
519 | ||
3b125388 | 520 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 521 | { |
3b125388 | 522 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
523 | struct files_struct *old; |
524 | ||
525 | old = tsk->files; | |
526 | task_lock(tsk); | |
527 | tsk->files = files; | |
528 | task_unlock(tsk); | |
529 | put_files_struct(old); | |
530 | } | |
3b9b8ab6 | 531 | |
1ec7f1dd | 532 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
533 | { |
534 | struct files_struct * files = tsk->files; | |
535 | ||
536 | if (files) { | |
537 | task_lock(tsk); | |
538 | tsk->files = NULL; | |
539 | task_unlock(tsk); | |
540 | put_files_struct(files); | |
541 | } | |
542 | } | |
543 | ||
1ec7f1dd | 544 | void put_fs_struct(struct fs_struct *fs) |
1da177e4 LT |
545 | { |
546 | /* No need to hold fs->lock if we are killing it */ | |
547 | if (atomic_dec_and_test(&fs->count)) { | |
6ac08c39 JB |
548 | path_put(&fs->root); |
549 | path_put(&fs->pwd); | |
550 | if (fs->altroot.dentry) | |
551 | path_put(&fs->altroot); | |
1da177e4 LT |
552 | kmem_cache_free(fs_cachep, fs); |
553 | } | |
554 | } | |
555 | ||
1ec7f1dd | 556 | void exit_fs(struct task_struct *tsk) |
1da177e4 LT |
557 | { |
558 | struct fs_struct * fs = tsk->fs; | |
559 | ||
560 | if (fs) { | |
561 | task_lock(tsk); | |
562 | tsk->fs = NULL; | |
563 | task_unlock(tsk); | |
1ec7f1dd | 564 | put_fs_struct(fs); |
1da177e4 LT |
565 | } |
566 | } | |
567 | ||
1da177e4 LT |
568 | EXPORT_SYMBOL_GPL(exit_fs); |
569 | ||
cf475ad2 BS |
570 | #ifdef CONFIG_MM_OWNER |
571 | /* | |
572 | * Task p is exiting and it owned mm, lets find a new owner for it | |
573 | */ | |
574 | static inline int | |
575 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) | |
576 | { | |
577 | /* | |
578 | * If there are other users of the mm and the owner (us) is exiting | |
579 | * we need to find a new owner to take on the responsibility. | |
580 | */ | |
581 | if (!mm) | |
582 | return 0; | |
583 | if (atomic_read(&mm->mm_users) <= 1) | |
584 | return 0; | |
585 | if (mm->owner != p) | |
586 | return 0; | |
587 | return 1; | |
588 | } | |
589 | ||
590 | void mm_update_next_owner(struct mm_struct *mm) | |
591 | { | |
592 | struct task_struct *c, *g, *p = current; | |
593 | ||
594 | retry: | |
595 | if (!mm_need_new_owner(mm, p)) | |
596 | return; | |
597 | ||
598 | read_lock(&tasklist_lock); | |
599 | /* | |
600 | * Search in the children | |
601 | */ | |
602 | list_for_each_entry(c, &p->children, sibling) { | |
603 | if (c->mm == mm) | |
604 | goto assign_new_owner; | |
605 | } | |
606 | ||
607 | /* | |
608 | * Search in the siblings | |
609 | */ | |
610 | list_for_each_entry(c, &p->parent->children, sibling) { | |
611 | if (c->mm == mm) | |
612 | goto assign_new_owner; | |
613 | } | |
614 | ||
615 | /* | |
616 | * Search through everything else. We should not get | |
617 | * here often | |
618 | */ | |
619 | do_each_thread(g, c) { | |
620 | if (c->mm == mm) | |
621 | goto assign_new_owner; | |
622 | } while_each_thread(g, c); | |
623 | ||
624 | read_unlock(&tasklist_lock); | |
625 | return; | |
626 | ||
627 | assign_new_owner: | |
628 | BUG_ON(c == p); | |
629 | get_task_struct(c); | |
630 | /* | |
631 | * The task_lock protects c->mm from changing. | |
632 | * We always want mm->owner->mm == mm | |
633 | */ | |
634 | task_lock(c); | |
635 | /* | |
636 | * Delay read_unlock() till we have the task_lock() | |
637 | * to ensure that c does not slip away underneath us | |
638 | */ | |
639 | read_unlock(&tasklist_lock); | |
640 | if (c->mm != mm) { | |
641 | task_unlock(c); | |
642 | put_task_struct(c); | |
643 | goto retry; | |
644 | } | |
645 | cgroup_mm_owner_callbacks(mm->owner, c); | |
646 | mm->owner = c; | |
647 | task_unlock(c); | |
648 | put_task_struct(c); | |
649 | } | |
650 | #endif /* CONFIG_MM_OWNER */ | |
651 | ||
1da177e4 LT |
652 | /* |
653 | * Turn us into a lazy TLB process if we | |
654 | * aren't already.. | |
655 | */ | |
408b664a | 656 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
657 | { |
658 | struct mm_struct *mm = tsk->mm; | |
659 | ||
660 | mm_release(tsk, mm); | |
661 | if (!mm) | |
662 | return; | |
663 | /* | |
664 | * Serialize with any possible pending coredump. | |
665 | * We must hold mmap_sem around checking core_waiters | |
666 | * and clearing tsk->mm. The core-inducing thread | |
667 | * will increment core_waiters for each thread in the | |
668 | * group with ->mm != NULL. | |
669 | */ | |
670 | down_read(&mm->mmap_sem); | |
671 | if (mm->core_waiters) { | |
672 | up_read(&mm->mmap_sem); | |
673 | down_write(&mm->mmap_sem); | |
674 | if (!--mm->core_waiters) | |
675 | complete(mm->core_startup_done); | |
676 | up_write(&mm->mmap_sem); | |
677 | ||
678 | wait_for_completion(&mm->core_done); | |
679 | down_read(&mm->mmap_sem); | |
680 | } | |
681 | atomic_inc(&mm->mm_count); | |
125e1874 | 682 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
683 | /* more a memory barrier than a real lock */ |
684 | task_lock(tsk); | |
685 | tsk->mm = NULL; | |
686 | up_read(&mm->mmap_sem); | |
687 | enter_lazy_tlb(mm, current); | |
0c1eecfb RW |
688 | /* We don't want this task to be frozen prematurely */ |
689 | clear_freeze_flag(tsk); | |
1da177e4 | 690 | task_unlock(tsk); |
cf475ad2 | 691 | mm_update_next_owner(mm); |
1da177e4 LT |
692 | mmput(mm); |
693 | } | |
694 | ||
36c8b586 IM |
695 | static void |
696 | reparent_thread(struct task_struct *p, struct task_struct *father, int traced) | |
1da177e4 | 697 | { |
241ceee0 ON |
698 | if (p->pdeath_signal) |
699 | /* We already hold the tasklist_lock here. */ | |
700 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | |
701 | ||
1da177e4 LT |
702 | /* Move the child from its dying parent to the new one. */ |
703 | if (unlikely(traced)) { | |
704 | /* Preserve ptrace links if someone else is tracing this child. */ | |
705 | list_del_init(&p->ptrace_list); | |
53b6f9fb | 706 | if (ptrace_reparented(p)) |
1da177e4 LT |
707 | list_add(&p->ptrace_list, &p->real_parent->ptrace_children); |
708 | } else { | |
709 | /* If this child is being traced, then we're the one tracing it | |
710 | * anyway, so let go of it. | |
711 | */ | |
712 | p->ptrace = 0; | |
6ac781b1 | 713 | remove_parent(p); |
1da177e4 | 714 | p->parent = p->real_parent; |
6ac781b1 | 715 | add_parent(p); |
1da177e4 | 716 | |
338077e5 | 717 | if (task_is_traced(p)) { |
1da177e4 LT |
718 | /* |
719 | * If it was at a trace stop, turn it into | |
720 | * a normal stop since it's no longer being | |
721 | * traced. | |
722 | */ | |
723 | ptrace_untrace(p); | |
724 | } | |
725 | } | |
726 | ||
b2b2cbc4 EB |
727 | /* If this is a threaded reparent there is no need to |
728 | * notify anyone anything has happened. | |
729 | */ | |
376e1d25 | 730 | if (same_thread_group(p->real_parent, father)) |
b2b2cbc4 EB |
731 | return; |
732 | ||
733 | /* We don't want people slaying init. */ | |
d839fd4d | 734 | if (!task_detached(p)) |
b2b2cbc4 | 735 | p->exit_signal = SIGCHLD; |
b2b2cbc4 EB |
736 | |
737 | /* If we'd notified the old parent about this child's death, | |
738 | * also notify the new parent. | |
739 | */ | |
740 | if (!traced && p->exit_state == EXIT_ZOMBIE && | |
d839fd4d | 741 | !task_detached(p) && thread_group_empty(p)) |
b2b2cbc4 EB |
742 | do_notify_parent(p, p->exit_signal); |
743 | ||
f49ee505 | 744 | kill_orphaned_pgrp(p, father); |
1da177e4 LT |
745 | } |
746 | ||
747 | /* | |
748 | * When we die, we re-parent all our children. | |
749 | * Try to give them to another thread in our thread | |
750 | * group, and if no such member exists, give it to | |
84d73786 SB |
751 | * the child reaper process (ie "init") in our pid |
752 | * space. | |
1da177e4 | 753 | */ |
762a24be | 754 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 755 | { |
03ff1797 | 756 | struct task_struct *p, *n, *reaper = father; |
762a24be ON |
757 | struct list_head ptrace_dead; |
758 | ||
759 | INIT_LIST_HEAD(&ptrace_dead); | |
760 | ||
761 | write_lock_irq(&tasklist_lock); | |
1da177e4 LT |
762 | |
763 | do { | |
764 | reaper = next_thread(reaper); | |
765 | if (reaper == father) { | |
88f21d81 | 766 | reaper = task_child_reaper(father); |
1da177e4 LT |
767 | break; |
768 | } | |
762a24be | 769 | } while (reaper->flags & PF_EXITING); |
1da177e4 LT |
770 | |
771 | /* | |
772 | * There are only two places where our children can be: | |
773 | * | |
774 | * - in our child list | |
775 | * - in our ptraced child list | |
776 | * | |
777 | * Search them and reparent children. | |
778 | */ | |
03ff1797 | 779 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
1da177e4 | 780 | int ptrace; |
1da177e4 LT |
781 | |
782 | ptrace = p->ptrace; | |
783 | ||
784 | /* if father isn't the real parent, then ptrace must be enabled */ | |
785 | BUG_ON(father != p->real_parent && !ptrace); | |
786 | ||
787 | if (father == p->real_parent) { | |
788 | /* reparent with a reaper, real father it's us */ | |
84eb646b | 789 | p->real_parent = reaper; |
1da177e4 LT |
790 | reparent_thread(p, father, 0); |
791 | } else { | |
792 | /* reparent ptraced task to its real parent */ | |
793 | __ptrace_unlink (p); | |
d839fd4d | 794 | if (p->exit_state == EXIT_ZOMBIE && !task_detached(p) && |
1da177e4 LT |
795 | thread_group_empty(p)) |
796 | do_notify_parent(p, p->exit_signal); | |
797 | } | |
798 | ||
799 | /* | |
d839fd4d ON |
800 | * if the ptraced child is a detached zombie we must collect |
801 | * it before we exit, or it will remain zombie forever since | |
802 | * we prevented it from self-reap itself while it was being | |
803 | * traced by us, to be able to see it in wait4. | |
1da177e4 | 804 | */ |
d839fd4d | 805 | if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && task_detached(p))) |
762a24be | 806 | list_add(&p->ptrace_list, &ptrace_dead); |
1da177e4 | 807 | } |
762a24be | 808 | |
03ff1797 | 809 | list_for_each_entry_safe(p, n, &father->ptrace_children, ptrace_list) { |
84eb646b | 810 | p->real_parent = reaper; |
1da177e4 LT |
811 | reparent_thread(p, father, 1); |
812 | } | |
762a24be ON |
813 | |
814 | write_unlock_irq(&tasklist_lock); | |
815 | BUG_ON(!list_empty(&father->children)); | |
816 | BUG_ON(!list_empty(&father->ptrace_children)); | |
817 | ||
818 | list_for_each_entry_safe(p, n, &ptrace_dead, ptrace_list) { | |
819 | list_del_init(&p->ptrace_list); | |
820 | release_task(p); | |
821 | } | |
822 | ||
1da177e4 LT |
823 | } |
824 | ||
825 | /* | |
826 | * Send signals to all our closest relatives so that they know | |
827 | * to properly mourn us.. | |
828 | */ | |
821c7de7 | 829 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 LT |
830 | { |
831 | int state; | |
1da177e4 | 832 | |
1da177e4 LT |
833 | /* |
834 | * This does two things: | |
835 | * | |
836 | * A. Make init inherit all the child processes | |
837 | * B. Check to see if any process groups have become orphaned | |
838 | * as a result of our exiting, and if they have any stopped | |
839 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
840 | */ | |
762a24be | 841 | forget_original_parent(tsk); |
2e4a7072 | 842 | exit_task_namespaces(tsk); |
1da177e4 | 843 | |
762a24be | 844 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
845 | if (group_dead) |
846 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 847 | |
24728448 | 848 | /* Let father know we died |
1da177e4 LT |
849 | * |
850 | * Thread signals are configurable, but you aren't going to use | |
d4c5e41f | 851 | * that to send signals to arbitary processes. |
1da177e4 LT |
852 | * That stops right now. |
853 | * | |
854 | * If the parent exec id doesn't match the exec id we saved | |
855 | * when we started then we know the parent has changed security | |
856 | * domain. | |
857 | * | |
858 | * If our self_exec id doesn't match our parent_exec_id then | |
859 | * we have changed execution domain as these two values started | |
860 | * the same after a fork. | |
1da177e4 | 861 | */ |
d839fd4d | 862 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
f49ee505 | 863 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
d839fd4d ON |
864 | tsk->self_exec_id != tsk->parent_exec_id) && |
865 | !capable(CAP_KILL)) | |
1da177e4 LT |
866 | tsk->exit_signal = SIGCHLD; |
867 | ||
1da177e4 LT |
868 | /* If something other than our normal parent is ptracing us, then |
869 | * send it a SIGCHLD instead of honoring exit_signal. exit_signal | |
870 | * only has special meaning to our real parent. | |
871 | */ | |
d839fd4d | 872 | if (!task_detached(tsk) && thread_group_empty(tsk)) { |
53b6f9fb ON |
873 | int signal = ptrace_reparented(tsk) ? |
874 | SIGCHLD : tsk->exit_signal; | |
1da177e4 LT |
875 | do_notify_parent(tsk, signal); |
876 | } else if (tsk->ptrace) { | |
877 | do_notify_parent(tsk, SIGCHLD); | |
878 | } | |
879 | ||
880 | state = EXIT_ZOMBIE; | |
d839fd4d | 881 | if (task_detached(tsk) && likely(!tsk->ptrace)) |
1da177e4 LT |
882 | state = EXIT_DEAD; |
883 | tsk->exit_state = state; | |
884 | ||
2800d8d1 | 885 | /* mt-exec, de_thread() is waiting for us */ |
6db840fa ON |
886 | if (thread_group_leader(tsk) && |
887 | tsk->signal->notify_count < 0 && | |
888 | tsk->signal->group_exit_task) | |
889 | wake_up_process(tsk->signal->group_exit_task); | |
890 | ||
1da177e4 LT |
891 | write_unlock_irq(&tasklist_lock); |
892 | ||
1da177e4 LT |
893 | /* If the process is dead, release it - nobody will wait for it */ |
894 | if (state == EXIT_DEAD) | |
895 | release_task(tsk); | |
1da177e4 LT |
896 | } |
897 | ||
e18eecb8 JD |
898 | #ifdef CONFIG_DEBUG_STACK_USAGE |
899 | static void check_stack_usage(void) | |
900 | { | |
901 | static DEFINE_SPINLOCK(low_water_lock); | |
902 | static int lowest_to_date = THREAD_SIZE; | |
903 | unsigned long *n = end_of_stack(current); | |
904 | unsigned long free; | |
905 | ||
906 | while (*n == 0) | |
907 | n++; | |
908 | free = (unsigned long)n - (unsigned long)end_of_stack(current); | |
909 | ||
910 | if (free >= lowest_to_date) | |
911 | return; | |
912 | ||
913 | spin_lock(&low_water_lock); | |
914 | if (free < lowest_to_date) { | |
915 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
916 | "left\n", | |
917 | current->comm, free); | |
918 | lowest_to_date = free; | |
919 | } | |
920 | spin_unlock(&low_water_lock); | |
921 | } | |
922 | #else | |
923 | static inline void check_stack_usage(void) {} | |
924 | #endif | |
925 | ||
84eb646b ON |
926 | static inline void exit_child_reaper(struct task_struct *tsk) |
927 | { | |
88f21d81 | 928 | if (likely(tsk->group_leader != task_child_reaper(tsk))) |
84eb646b ON |
929 | return; |
930 | ||
3eb07c8c SB |
931 | if (tsk->nsproxy->pid_ns == &init_pid_ns) |
932 | panic("Attempted to kill init!"); | |
933 | ||
934 | /* | |
935 | * @tsk is the last thread in the 'cgroup-init' and is exiting. | |
936 | * Terminate all remaining processes in the namespace and reap them | |
937 | * before exiting @tsk. | |
938 | * | |
939 | * Note that @tsk (last thread of cgroup-init) may not necessarily | |
940 | * be the child-reaper (i.e main thread of cgroup-init) of the | |
941 | * namespace i.e the child_reaper may have already exited. | |
942 | * | |
943 | * Even after a child_reaper exits, we let it inherit orphaned children, | |
944 | * because, pid_ns->child_reaper remains valid as long as there is | |
945 | * at least one living sub-thread in the cgroup init. | |
946 | ||
947 | * This living sub-thread of the cgroup-init will be notified when | |
948 | * a child inherited by the 'child-reaper' exits (do_notify_parent() | |
949 | * uses __group_send_sig_info()). Further, when reaping child processes, | |
950 | * do_wait() iterates over children of all living sub threads. | |
951 | ||
952 | * i.e even though 'child_reaper' thread is listed as the parent of the | |
953 | * orphaned children, any living sub-thread in the cgroup-init can | |
954 | * perform the role of the child_reaper. | |
955 | */ | |
956 | zap_pid_ns_processes(tsk->nsproxy->pid_ns); | |
84eb646b ON |
957 | } |
958 | ||
7ad5b3a5 | 959 | NORET_TYPE void do_exit(long code) |
1da177e4 LT |
960 | { |
961 | struct task_struct *tsk = current; | |
962 | int group_dead; | |
963 | ||
964 | profile_task_exit(tsk); | |
965 | ||
22e2c507 JA |
966 | WARN_ON(atomic_read(&tsk->fs_excl)); |
967 | ||
1da177e4 LT |
968 | if (unlikely(in_interrupt())) |
969 | panic("Aiee, killing interrupt handler!"); | |
970 | if (unlikely(!tsk->pid)) | |
971 | panic("Attempted to kill the idle task!"); | |
1da177e4 LT |
972 | |
973 | if (unlikely(current->ptrace & PT_TRACE_EXIT)) { | |
974 | current->ptrace_message = code; | |
975 | ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP); | |
976 | } | |
977 | ||
df164db5 AN |
978 | /* |
979 | * We're taking recursive faults here in do_exit. Safest is to just | |
980 | * leave this task alone and wait for reboot. | |
981 | */ | |
982 | if (unlikely(tsk->flags & PF_EXITING)) { | |
983 | printk(KERN_ALERT | |
984 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
985 | /* |
986 | * We can do this unlocked here. The futex code uses | |
987 | * this flag just to verify whether the pi state | |
988 | * cleanup has been done or not. In the worst case it | |
989 | * loops once more. We pretend that the cleanup was | |
990 | * done as there is no way to return. Either the | |
991 | * OWNER_DIED bit is set by now or we push the blocked | |
992 | * task into the wait for ever nirwana as well. | |
993 | */ | |
994 | tsk->flags |= PF_EXITPIDONE; | |
afc847b7 AV |
995 | if (tsk->io_context) |
996 | exit_io_context(); | |
df164db5 AN |
997 | set_current_state(TASK_UNINTERRUPTIBLE); |
998 | schedule(); | |
999 | } | |
1000 | ||
d12619b5 | 1001 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
1002 | /* |
1003 | * tsk->flags are checked in the futex code to protect against | |
1004 | * an exiting task cleaning up the robust pi futexes. | |
1005 | */ | |
d2ee7198 ON |
1006 | smp_mb(); |
1007 | spin_unlock_wait(&tsk->pi_lock); | |
1da177e4 | 1008 | |
1da177e4 LT |
1009 | if (unlikely(in_atomic())) |
1010 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 1011 | current->comm, task_pid_nr(current), |
1da177e4 LT |
1012 | preempt_count()); |
1013 | ||
1014 | acct_update_integrals(tsk); | |
365e9c87 HD |
1015 | if (tsk->mm) { |
1016 | update_hiwater_rss(tsk->mm); | |
1017 | update_hiwater_vm(tsk->mm); | |
1018 | } | |
1da177e4 | 1019 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 1020 | if (group_dead) { |
84eb646b | 1021 | exit_child_reaper(tsk); |
778e9a9c | 1022 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 1023 | exit_itimers(tsk->signal); |
c3068951 | 1024 | } |
f6ec29a4 | 1025 | acct_collect(code, group_dead); |
42b2dd0a | 1026 | #ifdef CONFIG_FUTEX |
0771dfef IM |
1027 | if (unlikely(tsk->robust_list)) |
1028 | exit_robust_list(tsk); | |
42b2dd0a | 1029 | #ifdef CONFIG_COMPAT |
34f192c6 IM |
1030 | if (unlikely(tsk->compat_robust_list)) |
1031 | compat_exit_robust_list(tsk); | |
42b2dd0a | 1032 | #endif |
34f192c6 | 1033 | #endif |
522ed776 MT |
1034 | if (group_dead) |
1035 | tty_audit_exit(); | |
fa84cb93 AV |
1036 | if (unlikely(tsk->audit_context)) |
1037 | audit_free(tsk); | |
115085ea | 1038 | |
f2ab6d88 | 1039 | tsk->exit_code = code; |
115085ea | 1040 | taskstats_exit(tsk, group_dead); |
c757249a | 1041 | |
1da177e4 LT |
1042 | exit_mm(tsk); |
1043 | ||
0e464814 | 1044 | if (group_dead) |
f6ec29a4 | 1045 | acct_process(); |
1da177e4 | 1046 | exit_sem(tsk); |
1ec7f1dd AV |
1047 | exit_files(tsk); |
1048 | exit_fs(tsk); | |
e18eecb8 | 1049 | check_stack_usage(); |
1da177e4 | 1050 | exit_thread(); |
b4f48b63 | 1051 | cgroup_exit(tsk, 1); |
1da177e4 LT |
1052 | exit_keys(tsk); |
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; |
7407251a | 1097 | |
1da177e4 LT |
1098 | schedule(); |
1099 | BUG(); | |
1100 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1101 | for (;;) |
1102 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1103 | } |
1104 | ||
012914da RA |
1105 | EXPORT_SYMBOL_GPL(do_exit); |
1106 | ||
1da177e4 LT |
1107 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1108 | { | |
1109 | if (comp) | |
1110 | complete(comp); | |
55a101f8 | 1111 | |
1da177e4 LT |
1112 | do_exit(code); |
1113 | } | |
1114 | ||
1115 | EXPORT_SYMBOL(complete_and_exit); | |
1116 | ||
1117 | asmlinkage long sys_exit(int error_code) | |
1118 | { | |
1119 | do_exit((error_code&0xff)<<8); | |
1120 | } | |
1121 | ||
1da177e4 LT |
1122 | /* |
1123 | * Take down every thread in the group. This is called by fatal signals | |
1124 | * as well as by sys_exit_group (below). | |
1125 | */ | |
1126 | NORET_TYPE void | |
1127 | do_group_exit(int exit_code) | |
1128 | { | |
bfc4b089 ON |
1129 | struct signal_struct *sig = current->signal; |
1130 | ||
1da177e4 LT |
1131 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1132 | ||
bfc4b089 ON |
1133 | if (signal_group_exit(sig)) |
1134 | exit_code = sig->group_exit_code; | |
1da177e4 | 1135 | else if (!thread_group_empty(current)) { |
1da177e4 | 1136 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1137 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1138 | if (signal_group_exit(sig)) |
1da177e4 LT |
1139 | /* Another thread got here before we took the lock. */ |
1140 | exit_code = sig->group_exit_code; | |
1141 | else { | |
1da177e4 | 1142 | sig->group_exit_code = exit_code; |
ed5d2cac | 1143 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1144 | zap_other_threads(current); |
1145 | } | |
1146 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1147 | } |
1148 | ||
1149 | do_exit(exit_code); | |
1150 | /* NOTREACHED */ | |
1151 | } | |
1152 | ||
1153 | /* | |
1154 | * this kills every thread in the thread group. Note that any externally | |
1155 | * wait4()-ing process will get the correct exit code - even if this | |
1156 | * thread is not the thread group leader. | |
1157 | */ | |
1158 | asmlinkage void sys_exit_group(int error_code) | |
1159 | { | |
1160 | do_group_exit((error_code & 0xff) << 8); | |
1161 | } | |
1162 | ||
161550d7 EB |
1163 | static struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
1164 | { | |
1165 | struct pid *pid = NULL; | |
1166 | if (type == PIDTYPE_PID) | |
1167 | pid = task->pids[type].pid; | |
1168 | else if (type < PIDTYPE_MAX) | |
1169 | pid = task->group_leader->pids[type].pid; | |
1170 | return pid; | |
1171 | } | |
1172 | ||
1173 | static int eligible_child(enum pid_type type, struct pid *pid, int options, | |
1174 | struct task_struct *p) | |
1da177e4 | 1175 | { |
73243284 RM |
1176 | int err; |
1177 | ||
161550d7 EB |
1178 | if (type < PIDTYPE_MAX) { |
1179 | if (task_pid_type(p, type) != pid) | |
1da177e4 LT |
1180 | return 0; |
1181 | } | |
1182 | ||
1183 | /* | |
1184 | * Do not consider detached threads that are | |
1185 | * not ptraced: | |
1186 | */ | |
d839fd4d | 1187 | if (task_detached(p) && !p->ptrace) |
1da177e4 LT |
1188 | return 0; |
1189 | ||
1190 | /* Wait for all children (clone and not) if __WALL is set; | |
1191 | * otherwise, wait for clone children *only* if __WCLONE is | |
1192 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1193 | * A "clone" child here is one that reports to its parent | |
1194 | * using a signal other than SIGCHLD.) */ | |
1195 | if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0)) | |
1196 | && !(options & __WALL)) | |
1197 | return 0; | |
1da177e4 | 1198 | |
73243284 | 1199 | err = security_task_wait(p); |
f2cc3eb1 ON |
1200 | if (likely(!err)) |
1201 | return 1; | |
1da177e4 | 1202 | |
161550d7 | 1203 | if (type != PIDTYPE_PID) |
f2cc3eb1 ON |
1204 | return 0; |
1205 | /* This child was explicitly requested, abort */ | |
1206 | read_unlock(&tasklist_lock); | |
1207 | return err; | |
1da177e4 LT |
1208 | } |
1209 | ||
36c8b586 | 1210 | static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid, |
1da177e4 LT |
1211 | int why, int status, |
1212 | struct siginfo __user *infop, | |
1213 | struct rusage __user *rusagep) | |
1214 | { | |
1215 | int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0; | |
36c8b586 | 1216 | |
1da177e4 LT |
1217 | put_task_struct(p); |
1218 | if (!retval) | |
1219 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1220 | if (!retval) | |
1221 | retval = put_user(0, &infop->si_errno); | |
1222 | if (!retval) | |
1223 | retval = put_user((short)why, &infop->si_code); | |
1224 | if (!retval) | |
1225 | retval = put_user(pid, &infop->si_pid); | |
1226 | if (!retval) | |
1227 | retval = put_user(uid, &infop->si_uid); | |
1228 | if (!retval) | |
1229 | retval = put_user(status, &infop->si_status); | |
1230 | if (!retval) | |
1231 | retval = pid; | |
1232 | return retval; | |
1233 | } | |
1234 | ||
1235 | /* | |
1236 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1237 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1238 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1239 | * released the lock and the system call should return. | |
1240 | */ | |
98abed02 | 1241 | static int wait_task_zombie(struct task_struct *p, int options, |
1da177e4 LT |
1242 | struct siginfo __user *infop, |
1243 | int __user *stat_addr, struct rusage __user *ru) | |
1244 | { | |
1245 | unsigned long state; | |
2f4e6e2a | 1246 | int retval, status, traced; |
6c5f3e7b | 1247 | pid_t pid = task_pid_vnr(p); |
1da177e4 | 1248 | |
98abed02 RM |
1249 | if (!likely(options & WEXITED)) |
1250 | return 0; | |
1251 | ||
1252 | if (unlikely(options & WNOWAIT)) { | |
1da177e4 LT |
1253 | uid_t uid = p->uid; |
1254 | int exit_code = p->exit_code; | |
1255 | int why, status; | |
1256 | ||
1da177e4 LT |
1257 | get_task_struct(p); |
1258 | read_unlock(&tasklist_lock); | |
1259 | if ((exit_code & 0x7f) == 0) { | |
1260 | why = CLD_EXITED; | |
1261 | status = exit_code >> 8; | |
1262 | } else { | |
1263 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1264 | status = exit_code & 0x7f; | |
1265 | } | |
1266 | return wait_noreap_copyout(p, pid, uid, why, | |
1267 | status, infop, ru); | |
1268 | } | |
1269 | ||
1270 | /* | |
1271 | * Try to move the task's state to DEAD | |
1272 | * only one thread is allowed to do this: | |
1273 | */ | |
1274 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1275 | if (state != EXIT_ZOMBIE) { | |
1276 | BUG_ON(state != EXIT_DEAD); | |
1277 | return 0; | |
1278 | } | |
1da177e4 | 1279 | |
53b6f9fb | 1280 | traced = ptrace_reparented(p); |
2f4e6e2a ON |
1281 | |
1282 | if (likely(!traced)) { | |
3795e161 JJ |
1283 | struct signal_struct *psig; |
1284 | struct signal_struct *sig; | |
1285 | ||
1da177e4 LT |
1286 | /* |
1287 | * The resource counters for the group leader are in its | |
1288 | * own task_struct. Those for dead threads in the group | |
1289 | * are in its signal_struct, as are those for the child | |
1290 | * processes it has previously reaped. All these | |
1291 | * accumulate in the parent's signal_struct c* fields. | |
1292 | * | |
1293 | * We don't bother to take a lock here to protect these | |
1294 | * p->signal fields, because they are only touched by | |
1295 | * __exit_signal, which runs with tasklist_lock | |
1296 | * write-locked anyway, and so is excluded here. We do | |
1297 | * need to protect the access to p->parent->signal fields, | |
1298 | * as other threads in the parent group can be right | |
1299 | * here reaping other children at the same time. | |
1300 | */ | |
1301 | spin_lock_irq(&p->parent->sighand->siglock); | |
3795e161 JJ |
1302 | psig = p->parent->signal; |
1303 | sig = p->signal; | |
1304 | psig->cutime = | |
1305 | cputime_add(psig->cutime, | |
1da177e4 | 1306 | cputime_add(p->utime, |
3795e161 JJ |
1307 | cputime_add(sig->utime, |
1308 | sig->cutime))); | |
1309 | psig->cstime = | |
1310 | cputime_add(psig->cstime, | |
1da177e4 | 1311 | cputime_add(p->stime, |
3795e161 JJ |
1312 | cputime_add(sig->stime, |
1313 | sig->cstime))); | |
9ac52315 LV |
1314 | psig->cgtime = |
1315 | cputime_add(psig->cgtime, | |
1316 | cputime_add(p->gtime, | |
1317 | cputime_add(sig->gtime, | |
1318 | sig->cgtime))); | |
3795e161 JJ |
1319 | psig->cmin_flt += |
1320 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1321 | psig->cmaj_flt += | |
1322 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1323 | psig->cnvcsw += | |
1324 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1325 | psig->cnivcsw += | |
1326 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1327 | psig->cinblock += |
1328 | task_io_get_inblock(p) + | |
1329 | sig->inblock + sig->cinblock; | |
1330 | psig->coublock += | |
1331 | task_io_get_oublock(p) + | |
1332 | sig->oublock + sig->coublock; | |
1da177e4 LT |
1333 | spin_unlock_irq(&p->parent->sighand->siglock); |
1334 | } | |
1335 | ||
1336 | /* | |
1337 | * Now we are sure this task is interesting, and no other | |
1338 | * thread can reap it because we set its state to EXIT_DEAD. | |
1339 | */ | |
1340 | read_unlock(&tasklist_lock); | |
1341 | ||
1342 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1343 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) | |
1344 | ? p->signal->group_exit_code : p->exit_code; | |
1345 | if (!retval && stat_addr) | |
1346 | retval = put_user(status, stat_addr); | |
1347 | if (!retval && infop) | |
1348 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1349 | if (!retval && infop) | |
1350 | retval = put_user(0, &infop->si_errno); | |
1351 | if (!retval && infop) { | |
1352 | int why; | |
1353 | ||
1354 | if ((status & 0x7f) == 0) { | |
1355 | why = CLD_EXITED; | |
1356 | status >>= 8; | |
1357 | } else { | |
1358 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1359 | status &= 0x7f; | |
1360 | } | |
1361 | retval = put_user((short)why, &infop->si_code); | |
1362 | if (!retval) | |
1363 | retval = put_user(status, &infop->si_status); | |
1364 | } | |
1365 | if (!retval && infop) | |
3a515e4a | 1366 | retval = put_user(pid, &infop->si_pid); |
1da177e4 LT |
1367 | if (!retval && infop) |
1368 | retval = put_user(p->uid, &infop->si_uid); | |
2f4e6e2a | 1369 | if (!retval) |
3a515e4a | 1370 | retval = pid; |
2f4e6e2a ON |
1371 | |
1372 | if (traced) { | |
1da177e4 | 1373 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1374 | /* We dropped tasklist, ptracer could die and untrace */ |
1375 | ptrace_unlink(p); | |
1376 | /* | |
1377 | * If this is not a detached task, notify the parent. | |
1378 | * If it's still not detached after that, don't release | |
1379 | * it now. | |
1380 | */ | |
d839fd4d | 1381 | if (!task_detached(p)) { |
2f4e6e2a | 1382 | do_notify_parent(p, p->exit_signal); |
d839fd4d | 1383 | if (!task_detached(p)) { |
2f4e6e2a ON |
1384 | p->exit_state = EXIT_ZOMBIE; |
1385 | p = NULL; | |
1da177e4 LT |
1386 | } |
1387 | } | |
1388 | write_unlock_irq(&tasklist_lock); | |
1389 | } | |
1390 | if (p != NULL) | |
1391 | release_task(p); | |
2f4e6e2a | 1392 | |
1da177e4 LT |
1393 | return retval; |
1394 | } | |
1395 | ||
1396 | /* | |
1397 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1398 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1399 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1400 | * released the lock and the system call should return. | |
1401 | */ | |
1bad95c3 | 1402 | static int wait_task_stopped(struct task_struct *p, |
98abed02 | 1403 | int options, struct siginfo __user *infop, |
1da177e4 LT |
1404 | int __user *stat_addr, struct rusage __user *ru) |
1405 | { | |
ee7c82da ON |
1406 | int retval, exit_code, why; |
1407 | uid_t uid = 0; /* unneeded, required by compiler */ | |
c8950783 | 1408 | pid_t pid; |
1da177e4 | 1409 | |
98abed02 RM |
1410 | if (!(p->ptrace & PT_PTRACED) && !(options & WUNTRACED)) |
1411 | return 0; | |
1412 | ||
ee7c82da ON |
1413 | exit_code = 0; |
1414 | spin_lock_irq(&p->sighand->siglock); | |
1415 | ||
1416 | if (unlikely(!task_is_stopped_or_traced(p))) | |
1417 | goto unlock_sig; | |
1418 | ||
1bad95c3 | 1419 | if (!(p->ptrace & PT_PTRACED) && p->signal->group_stop_count > 0) |
1da177e4 LT |
1420 | /* |
1421 | * A group stop is in progress and this is the group leader. | |
1422 | * We won't report until all threads have stopped. | |
1423 | */ | |
ee7c82da ON |
1424 | goto unlock_sig; |
1425 | ||
1426 | exit_code = p->exit_code; | |
1427 | if (!exit_code) | |
1428 | goto unlock_sig; | |
1429 | ||
98abed02 | 1430 | if (!unlikely(options & WNOWAIT)) |
ee7c82da ON |
1431 | p->exit_code = 0; |
1432 | ||
1433 | uid = p->uid; | |
1434 | unlock_sig: | |
1435 | spin_unlock_irq(&p->sighand->siglock); | |
1436 | if (!exit_code) | |
1da177e4 LT |
1437 | return 0; |
1438 | ||
1439 | /* | |
1440 | * Now we are pretty sure this task is interesting. | |
1441 | * Make sure it doesn't get reaped out from under us while we | |
1442 | * give up the lock and then examine it below. We don't want to | |
1443 | * keep holding onto the tasklist_lock while we call getrusage and | |
1444 | * possibly take page faults for user memory. | |
1445 | */ | |
1446 | get_task_struct(p); | |
6c5f3e7b | 1447 | pid = task_pid_vnr(p); |
ee7c82da | 1448 | why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1449 | read_unlock(&tasklist_lock); |
1450 | ||
98abed02 | 1451 | if (unlikely(options & WNOWAIT)) |
1da177e4 | 1452 | return wait_noreap_copyout(p, pid, uid, |
e6ceb32a | 1453 | why, exit_code, |
1da177e4 | 1454 | infop, ru); |
1da177e4 LT |
1455 | |
1456 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1457 | if (!retval && stat_addr) | |
1458 | retval = put_user((exit_code << 8) | 0x7f, stat_addr); | |
1459 | if (!retval && infop) | |
1460 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1461 | if (!retval && infop) | |
1462 | retval = put_user(0, &infop->si_errno); | |
1463 | if (!retval && infop) | |
6efcae46 | 1464 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1465 | if (!retval && infop) |
1466 | retval = put_user(exit_code, &infop->si_status); | |
1467 | if (!retval && infop) | |
c8950783 | 1468 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1469 | if (!retval && infop) |
ee7c82da | 1470 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1471 | if (!retval) |
c8950783 | 1472 | retval = pid; |
1da177e4 LT |
1473 | put_task_struct(p); |
1474 | ||
1475 | BUG_ON(!retval); | |
1476 | return retval; | |
1477 | } | |
1478 | ||
1479 | /* | |
1480 | * Handle do_wait work for one task in a live, non-stopped state. | |
1481 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1482 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1483 | * released the lock and the system call should return. | |
1484 | */ | |
98abed02 | 1485 | static int wait_task_continued(struct task_struct *p, int options, |
1da177e4 LT |
1486 | struct siginfo __user *infop, |
1487 | int __user *stat_addr, struct rusage __user *ru) | |
1488 | { | |
1489 | int retval; | |
1490 | pid_t pid; | |
1491 | uid_t uid; | |
1492 | ||
98abed02 RM |
1493 | if (!unlikely(options & WCONTINUED)) |
1494 | return 0; | |
1495 | ||
1da177e4 LT |
1496 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1497 | return 0; | |
1498 | ||
1499 | spin_lock_irq(&p->sighand->siglock); | |
1500 | /* Re-check with the lock held. */ | |
1501 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1502 | spin_unlock_irq(&p->sighand->siglock); | |
1503 | return 0; | |
1504 | } | |
98abed02 | 1505 | if (!unlikely(options & WNOWAIT)) |
1da177e4 LT |
1506 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
1507 | spin_unlock_irq(&p->sighand->siglock); | |
1508 | ||
6c5f3e7b | 1509 | pid = task_pid_vnr(p); |
1da177e4 LT |
1510 | uid = p->uid; |
1511 | get_task_struct(p); | |
1512 | read_unlock(&tasklist_lock); | |
1513 | ||
1514 | if (!infop) { | |
1515 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1516 | put_task_struct(p); | |
1517 | if (!retval && stat_addr) | |
1518 | retval = put_user(0xffff, stat_addr); | |
1519 | if (!retval) | |
3a515e4a | 1520 | retval = pid; |
1da177e4 LT |
1521 | } else { |
1522 | retval = wait_noreap_copyout(p, pid, uid, | |
1523 | CLD_CONTINUED, SIGCONT, | |
1524 | infop, ru); | |
1525 | BUG_ON(retval == 0); | |
1526 | } | |
1527 | ||
1528 | return retval; | |
1529 | } | |
1530 | ||
98abed02 RM |
1531 | /* |
1532 | * Consider @p for a wait by @parent. | |
1533 | * | |
1534 | * -ECHILD should be in *@notask_error before the first call. | |
1535 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | |
1536 | * Returns zero if the search for a child should continue; | |
1537 | * then *@notask_error is 0 if @p is an eligible child, or still -ECHILD. | |
1538 | */ | |
1539 | static int wait_consider_task(struct task_struct *parent, | |
1540 | struct task_struct *p, int *notask_error, | |
1541 | enum pid_type type, struct pid *pid, int options, | |
1542 | struct siginfo __user *infop, | |
1543 | int __user *stat_addr, struct rusage __user *ru) | |
1544 | { | |
1545 | int ret = eligible_child(type, pid, options, p); | |
1546 | if (ret <= 0) | |
1547 | return ret; | |
1548 | ||
1549 | if (p->exit_state == EXIT_DEAD) | |
1550 | return 0; | |
1551 | ||
1552 | /* | |
1553 | * We don't reap group leaders with subthreads. | |
1554 | */ | |
1555 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) | |
1556 | return wait_task_zombie(p, options, infop, stat_addr, ru); | |
1557 | ||
1558 | /* | |
1559 | * It's stopped or running now, so it might | |
1560 | * later continue, exit, or stop again. | |
1561 | */ | |
1562 | *notask_error = 0; | |
1563 | ||
1564 | if (task_is_stopped_or_traced(p)) | |
1565 | return wait_task_stopped(p, options, infop, stat_addr, ru); | |
1566 | ||
1567 | return wait_task_continued(p, options, infop, stat_addr, ru); | |
1568 | } | |
1569 | ||
1570 | /* | |
1571 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1572 | * | |
1573 | * -ECHILD should be in *@notask_error before the first call. | |
1574 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | |
1575 | * Returns zero if the search for a child should continue; then | |
1576 | * *@notask_error is 0 if there were any eligible children, or still -ECHILD. | |
1577 | */ | |
1578 | static int do_wait_thread(struct task_struct *tsk, int *notask_error, | |
1579 | enum pid_type type, struct pid *pid, int options, | |
1580 | struct siginfo __user *infop, int __user *stat_addr, | |
1581 | struct rusage __user *ru) | |
1582 | { | |
1583 | struct task_struct *p; | |
1584 | ||
1585 | list_for_each_entry(p, &tsk->children, sibling) { | |
1586 | int ret = wait_consider_task(tsk, p, notask_error, | |
1587 | type, pid, options, | |
1588 | infop, stat_addr, ru); | |
1589 | if (ret) | |
1590 | return ret; | |
1591 | } | |
1592 | ||
1593 | return 0; | |
1594 | } | |
1595 | ||
1596 | static int ptrace_do_wait(struct task_struct *tsk, int *notask_error, | |
1597 | enum pid_type type, struct pid *pid, int options, | |
1598 | struct siginfo __user *infop, int __user *stat_addr, | |
1599 | struct rusage __user *ru) | |
1600 | { | |
1601 | struct task_struct *p; | |
1602 | ||
1603 | /* | |
1604 | * If we never saw an eligile child, check for children stolen by | |
1605 | * ptrace. We don't leave -ECHILD in *@notask_error if there are any, | |
1606 | * because we will eventually be allowed to wait for them again. | |
1607 | */ | |
1608 | if (!*notask_error) | |
1609 | return 0; | |
1610 | ||
1611 | list_for_each_entry(p, &tsk->ptrace_children, ptrace_list) { | |
1612 | int ret = eligible_child(type, pid, options, p); | |
1613 | if (unlikely(ret < 0)) | |
1614 | return ret; | |
1615 | if (ret) { | |
1616 | *notask_error = 0; | |
1617 | return 0; | |
1618 | } | |
1619 | } | |
1620 | ||
1621 | return 0; | |
1622 | } | |
1623 | ||
161550d7 EB |
1624 | static long do_wait(enum pid_type type, struct pid *pid, int options, |
1625 | struct siginfo __user *infop, int __user *stat_addr, | |
1626 | struct rusage __user *ru) | |
1da177e4 LT |
1627 | { |
1628 | DECLARE_WAITQUEUE(wait, current); | |
1629 | struct task_struct *tsk; | |
98abed02 | 1630 | int retval; |
1da177e4 LT |
1631 | |
1632 | add_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1633 | repeat: | |
98abed02 RM |
1634 | /* |
1635 | * If there is nothing that can match our critiera just get out. | |
1636 | * We will clear @retval to zero if we see any child that might later | |
1637 | * match our criteria, even if we are not able to reap it yet. | |
1638 | */ | |
161550d7 EB |
1639 | retval = -ECHILD; |
1640 | if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type]))) | |
1641 | goto end; | |
1642 | ||
1da177e4 LT |
1643 | current->state = TASK_INTERRUPTIBLE; |
1644 | read_lock(&tasklist_lock); | |
1645 | tsk = current; | |
1646 | do { | |
98abed02 RM |
1647 | int tsk_result = do_wait_thread(tsk, &retval, |
1648 | type, pid, options, | |
1649 | infop, stat_addr, ru); | |
1650 | if (!tsk_result) | |
1651 | tsk_result = ptrace_do_wait(tsk, &retval, | |
1652 | type, pid, options, | |
1653 | infop, stat_addr, ru); | |
1654 | if (tsk_result) { | |
1655 | /* | |
1656 | * tasklist_lock is unlocked and we have a final result. | |
1657 | */ | |
1658 | retval = tsk_result; | |
1659 | goto end; | |
1da177e4 | 1660 | } |
98abed02 | 1661 | |
1da177e4 LT |
1662 | if (options & __WNOTHREAD) |
1663 | break; | |
1664 | tsk = next_thread(tsk); | |
125e1874 | 1665 | BUG_ON(tsk->signal != current->signal); |
1da177e4 | 1666 | } while (tsk != current); |
1da177e4 | 1667 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1668 | |
98abed02 | 1669 | if (!retval && !(options & WNOHANG)) { |
1da177e4 | 1670 | retval = -ERESTARTSYS; |
98abed02 RM |
1671 | if (!signal_pending(current)) { |
1672 | schedule(); | |
1673 | goto repeat; | |
1674 | } | |
1da177e4 | 1675 | } |
98abed02 | 1676 | |
1da177e4 LT |
1677 | end: |
1678 | current->state = TASK_RUNNING; | |
1679 | remove_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1680 | if (infop) { | |
1681 | if (retval > 0) | |
9cbab810 | 1682 | retval = 0; |
1da177e4 LT |
1683 | else { |
1684 | /* | |
1685 | * For a WNOHANG return, clear out all the fields | |
1686 | * we would set so the user can easily tell the | |
1687 | * difference. | |
1688 | */ | |
1689 | if (!retval) | |
1690 | retval = put_user(0, &infop->si_signo); | |
1691 | if (!retval) | |
1692 | retval = put_user(0, &infop->si_errno); | |
1693 | if (!retval) | |
1694 | retval = put_user(0, &infop->si_code); | |
1695 | if (!retval) | |
1696 | retval = put_user(0, &infop->si_pid); | |
1697 | if (!retval) | |
1698 | retval = put_user(0, &infop->si_uid); | |
1699 | if (!retval) | |
1700 | retval = put_user(0, &infop->si_status); | |
1701 | } | |
1702 | } | |
1703 | return retval; | |
1704 | } | |
1705 | ||
161550d7 | 1706 | asmlinkage long sys_waitid(int which, pid_t upid, |
1da177e4 LT |
1707 | struct siginfo __user *infop, int options, |
1708 | struct rusage __user *ru) | |
1709 | { | |
161550d7 EB |
1710 | struct pid *pid = NULL; |
1711 | enum pid_type type; | |
1da177e4 LT |
1712 | long ret; |
1713 | ||
1714 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1715 | return -EINVAL; | |
1716 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1717 | return -EINVAL; | |
1718 | ||
1719 | switch (which) { | |
1720 | case P_ALL: | |
161550d7 | 1721 | type = PIDTYPE_MAX; |
1da177e4 LT |
1722 | break; |
1723 | case P_PID: | |
161550d7 EB |
1724 | type = PIDTYPE_PID; |
1725 | if (upid <= 0) | |
1da177e4 LT |
1726 | return -EINVAL; |
1727 | break; | |
1728 | case P_PGID: | |
161550d7 EB |
1729 | type = PIDTYPE_PGID; |
1730 | if (upid <= 0) | |
1da177e4 | 1731 | return -EINVAL; |
1da177e4 LT |
1732 | break; |
1733 | default: | |
1734 | return -EINVAL; | |
1735 | } | |
1736 | ||
161550d7 EB |
1737 | if (type < PIDTYPE_MAX) |
1738 | pid = find_get_pid(upid); | |
1739 | ret = do_wait(type, pid, options, infop, NULL, ru); | |
1740 | put_pid(pid); | |
1da177e4 LT |
1741 | |
1742 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1743 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1744 | return ret; |
1745 | } | |
1746 | ||
161550d7 | 1747 | asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr, |
1da177e4 LT |
1748 | int options, struct rusage __user *ru) |
1749 | { | |
161550d7 EB |
1750 | struct pid *pid = NULL; |
1751 | enum pid_type type; | |
1da177e4 LT |
1752 | long ret; |
1753 | ||
1754 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1755 | __WNOTHREAD|__WCLONE|__WALL)) | |
1756 | return -EINVAL; | |
161550d7 EB |
1757 | |
1758 | if (upid == -1) | |
1759 | type = PIDTYPE_MAX; | |
1760 | else if (upid < 0) { | |
1761 | type = PIDTYPE_PGID; | |
1762 | pid = find_get_pid(-upid); | |
1763 | } else if (upid == 0) { | |
1764 | type = PIDTYPE_PGID; | |
1765 | pid = get_pid(task_pgrp(current)); | |
1766 | } else /* upid > 0 */ { | |
1767 | type = PIDTYPE_PID; | |
1768 | pid = find_get_pid(upid); | |
1769 | } | |
1770 | ||
1771 | ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru); | |
1772 | put_pid(pid); | |
1da177e4 LT |
1773 | |
1774 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1775 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1776 | return ret; |
1777 | } | |
1778 | ||
1779 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1780 | ||
1781 | /* | |
1782 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1783 | * implemented by calling sys_wait4() from libc.a. | |
1784 | */ | |
1785 | asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options) | |
1786 | { | |
1787 | return sys_wait4(pid, stat_addr, options, NULL); | |
1788 | } | |
1789 | ||
1790 | #endif |