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> |
30199f5a | 49 | #include <linux/tracehook.h> |
1da177e4 LT |
50 | |
51 | #include <asm/uaccess.h> | |
52 | #include <asm/unistd.h> | |
53 | #include <asm/pgtable.h> | |
54 | #include <asm/mmu_context.h> | |
55 | ||
408b664a AB |
56 | static void exit_mm(struct task_struct * tsk); |
57 | ||
d839fd4d ON |
58 | static inline int task_detached(struct task_struct *p) |
59 | { | |
60 | return p->exit_signal == -1; | |
61 | } | |
62 | ||
1da177e4 LT |
63 | static void __unhash_process(struct task_struct *p) |
64 | { | |
65 | nr_threads--; | |
66 | detach_pid(p, PIDTYPE_PID); | |
1da177e4 LT |
67 | if (thread_group_leader(p)) { |
68 | detach_pid(p, PIDTYPE_PGID); | |
69 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 70 | |
5e85d4ab | 71 | list_del_rcu(&p->tasks); |
73b9ebfe | 72 | __get_cpu_var(process_counts)--; |
1da177e4 | 73 | } |
47e65328 | 74 | list_del_rcu(&p->thread_group); |
f470021a | 75 | list_del_init(&p->sibling); |
1da177e4 LT |
76 | } |
77 | ||
6a14c5c9 ON |
78 | /* |
79 | * This function expects the tasklist_lock write-locked. | |
80 | */ | |
81 | static void __exit_signal(struct task_struct *tsk) | |
82 | { | |
83 | struct signal_struct *sig = tsk->signal; | |
84 | struct sighand_struct *sighand; | |
85 | ||
86 | BUG_ON(!sig); | |
87 | BUG_ON(!atomic_read(&sig->count)); | |
88 | ||
6a14c5c9 ON |
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); | |
297c5d92 AR |
124 | #ifdef CONFIG_TASK_XACCT |
125 | sig->rchar += tsk->rchar; | |
126 | sig->wchar += tsk->wchar; | |
127 | sig->syscr += tsk->syscr; | |
128 | sig->syscw += tsk->syscw; | |
129 | #endif /* CONFIG_TASK_XACCT */ | |
130 | #ifdef CONFIG_TASK_IO_ACCOUNTING | |
131 | sig->ioac.read_bytes += tsk->ioac.read_bytes; | |
132 | sig->ioac.write_bytes += tsk->ioac.write_bytes; | |
133 | sig->ioac.cancelled_write_bytes += | |
134 | tsk->ioac.cancelled_write_bytes; | |
135 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
172ba844 | 136 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
137 | sig = NULL; /* Marker for below. */ |
138 | } | |
139 | ||
5876700c ON |
140 | __unhash_process(tsk); |
141 | ||
da7978b0 ON |
142 | /* |
143 | * Do this under ->siglock, we can race with another thread | |
144 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. | |
145 | */ | |
146 | flush_sigqueue(&tsk->pending); | |
147 | ||
6a14c5c9 | 148 | tsk->signal = NULL; |
a7e5328a | 149 | tsk->sighand = NULL; |
6a14c5c9 | 150 | spin_unlock(&sighand->siglock); |
6a14c5c9 | 151 | |
a7e5328a | 152 | __cleanup_sighand(sighand); |
6a14c5c9 | 153 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
6a14c5c9 ON |
154 | if (sig) { |
155 | flush_sigqueue(&sig->shared_pending); | |
093a8e8a | 156 | taskstats_tgid_free(sig); |
6a14c5c9 ON |
157 | __cleanup_signal(sig); |
158 | } | |
159 | } | |
160 | ||
8c7904a0 EB |
161 | static void delayed_put_task_struct(struct rcu_head *rhp) |
162 | { | |
163 | put_task_struct(container_of(rhp, struct task_struct, rcu)); | |
164 | } | |
165 | ||
f470021a RM |
166 | /* |
167 | * Do final ptrace-related cleanup of a zombie being reaped. | |
168 | * | |
169 | * Called with write_lock(&tasklist_lock) held. | |
170 | */ | |
171 | static void ptrace_release_task(struct task_struct *p) | |
172 | { | |
173 | BUG_ON(!list_empty(&p->ptraced)); | |
174 | ptrace_unlink(p); | |
175 | BUG_ON(!list_empty(&p->ptrace_entry)); | |
176 | } | |
177 | ||
1da177e4 LT |
178 | void release_task(struct task_struct * p) |
179 | { | |
36c8b586 | 180 | struct task_struct *leader; |
1da177e4 | 181 | int zap_leader; |
1f09f974 | 182 | repeat: |
1da177e4 | 183 | atomic_dec(&p->user->processes); |
60347f67 | 184 | proc_flush_task(p); |
1da177e4 | 185 | write_lock_irq(&tasklist_lock); |
f470021a | 186 | ptrace_release_task(p); |
1da177e4 | 187 | __exit_signal(p); |
35f5cad8 | 188 | |
1da177e4 LT |
189 | /* |
190 | * If we are the last non-leader member of the thread | |
191 | * group, and the leader is zombie, then notify the | |
192 | * group leader's parent process. (if it wants notification.) | |
193 | */ | |
194 | zap_leader = 0; | |
195 | leader = p->group_leader; | |
196 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
d839fd4d | 197 | BUG_ON(task_detached(leader)); |
1da177e4 LT |
198 | do_notify_parent(leader, leader->exit_signal); |
199 | /* | |
200 | * If we were the last child thread and the leader has | |
201 | * exited already, and the leader's parent ignores SIGCHLD, | |
202 | * then we are the one who should release the leader. | |
203 | * | |
204 | * do_notify_parent() will have marked it self-reaping in | |
205 | * that case. | |
206 | */ | |
d839fd4d | 207 | zap_leader = task_detached(leader); |
1da177e4 LT |
208 | } |
209 | ||
1da177e4 | 210 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 211 | release_thread(p); |
8c7904a0 | 212 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
213 | |
214 | p = leader; | |
215 | if (unlikely(zap_leader)) | |
216 | goto repeat; | |
217 | } | |
218 | ||
1da177e4 LT |
219 | /* |
220 | * This checks not only the pgrp, but falls back on the pid if no | |
221 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
222 | * without this... | |
04a2e6a5 EB |
223 | * |
224 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 225 | */ |
04a2e6a5 | 226 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
227 | { |
228 | struct task_struct *p; | |
04a2e6a5 | 229 | struct pid *sid = NULL; |
62dfb554 | 230 | |
04a2e6a5 | 231 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 232 | if (p == NULL) |
04a2e6a5 | 233 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 234 | if (p != NULL) |
04a2e6a5 | 235 | sid = task_session(p); |
62dfb554 | 236 | |
1da177e4 LT |
237 | return sid; |
238 | } | |
239 | ||
240 | /* | |
241 | * Determine if a process group is "orphaned", according to the POSIX | |
242 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
243 | * by terminal-generated stop signals. Newly orphaned process groups are | |
244 | * to receive a SIGHUP and a SIGCONT. | |
245 | * | |
246 | * "I ask you, have you ever known what it is to be an orphan?" | |
247 | */ | |
0475ac08 | 248 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
249 | { |
250 | struct task_struct *p; | |
1da177e4 | 251 | |
0475ac08 | 252 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
253 | if ((p == ignored_task) || |
254 | (p->exit_state && thread_group_empty(p)) || | |
255 | is_global_init(p->real_parent)) | |
1da177e4 | 256 | continue; |
05e83df6 | 257 | |
0475ac08 | 258 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
259 | task_session(p->real_parent) == task_session(p)) |
260 | return 0; | |
0475ac08 | 261 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
262 | |
263 | return 1; | |
1da177e4 LT |
264 | } |
265 | ||
3e7cd6c4 | 266 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
267 | { |
268 | int retval; | |
269 | ||
270 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 271 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
272 | read_unlock(&tasklist_lock); |
273 | ||
274 | return retval; | |
275 | } | |
276 | ||
0475ac08 | 277 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
278 | { |
279 | int retval = 0; | |
280 | struct task_struct *p; | |
281 | ||
0475ac08 | 282 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
338077e5 | 283 | if (!task_is_stopped(p)) |
1da177e4 | 284 | continue; |
1da177e4 LT |
285 | retval = 1; |
286 | break; | |
0475ac08 | 287 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
288 | return retval; |
289 | } | |
290 | ||
f49ee505 ON |
291 | /* |
292 | * Check to see if any process groups have become orphaned as | |
293 | * a result of our exiting, and if they have any stopped jobs, | |
294 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
295 | */ | |
296 | static void | |
297 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
298 | { | |
299 | struct pid *pgrp = task_pgrp(tsk); | |
300 | struct task_struct *ignored_task = tsk; | |
301 | ||
302 | if (!parent) | |
303 | /* exit: our father is in a different pgrp than | |
304 | * we are and we were the only connection outside. | |
305 | */ | |
306 | parent = tsk->real_parent; | |
307 | else | |
308 | /* reparent: our child is in a different pgrp than | |
309 | * we are, and it was the only connection outside. | |
310 | */ | |
311 | ignored_task = NULL; | |
312 | ||
313 | if (task_pgrp(parent) != pgrp && | |
314 | task_session(parent) == task_session(tsk) && | |
315 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
316 | has_stopped_jobs(pgrp)) { | |
317 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
318 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
319 | } | |
320 | } | |
321 | ||
1da177e4 | 322 | /** |
49d769d5 | 323 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
324 | * |
325 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
326 | * it ever exits, it should generally reparent itself to kthreadd so it |
327 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
328 | * |
329 | * The various task state such as scheduling policy and priority may have | |
330 | * been inherited from a user process, so we reset them to sane values here. | |
331 | * | |
49d769d5 | 332 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 333 | */ |
49d769d5 | 334 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
335 | { |
336 | write_lock_irq(&tasklist_lock); | |
337 | ||
338 | ptrace_unlink(current); | |
339 | /* Reparent to init */ | |
49d769d5 | 340 | current->real_parent = current->parent = kthreadd_task; |
f470021a | 341 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
1da177e4 LT |
342 | |
343 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
344 | current->exit_signal = SIGCHLD; | |
345 | ||
e05606d3 | 346 | if (task_nice(current) < 0) |
1da177e4 LT |
347 | set_user_nice(current, 0); |
348 | /* cpus_allowed? */ | |
349 | /* rt_priority? */ | |
350 | /* signals? */ | |
351 | security_task_reparent_to_init(current); | |
352 | memcpy(current->signal->rlim, init_task.signal->rlim, | |
353 | sizeof(current->signal->rlim)); | |
354 | atomic_inc(&(INIT_USER->__count)); | |
355 | write_unlock_irq(&tasklist_lock); | |
356 | switch_uid(INIT_USER); | |
357 | } | |
358 | ||
8520d7c7 | 359 | void __set_special_pids(struct pid *pid) |
1da177e4 | 360 | { |
e19f247a | 361 | struct task_struct *curr = current->group_leader; |
8520d7c7 | 362 | pid_t nr = pid_nr(pid); |
1da177e4 | 363 | |
8520d7c7 | 364 | if (task_session(curr) != pid) { |
7d8da096 | 365 | change_pid(curr, PIDTYPE_SID, pid); |
8520d7c7 | 366 | set_task_session(curr, nr); |
1da177e4 | 367 | } |
8520d7c7 | 368 | if (task_pgrp(curr) != pid) { |
7d8da096 | 369 | change_pid(curr, PIDTYPE_PGID, pid); |
8520d7c7 | 370 | set_task_pgrp(curr, nr); |
1da177e4 LT |
371 | } |
372 | } | |
373 | ||
8520d7c7 | 374 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
375 | { |
376 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 377 | __set_special_pids(pid); |
1da177e4 LT |
378 | write_unlock_irq(&tasklist_lock); |
379 | } | |
380 | ||
381 | /* | |
382 | * Let kernel threads use this to say that they | |
383 | * allow a certain signal (since daemonize() will | |
384 | * have disabled all of them by default). | |
385 | */ | |
386 | int allow_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); | |
392 | sigdelset(¤t->blocked, sig); | |
393 | if (!current->mm) { | |
394 | /* Kernel threads handle their own signals. | |
395 | Let the signal code know it'll be handled, so | |
396 | that they don't get converted to SIGKILL or | |
397 | just silently dropped */ | |
398 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
399 | } | |
400 | recalc_sigpending(); | |
401 | spin_unlock_irq(¤t->sighand->siglock); | |
402 | return 0; | |
403 | } | |
404 | ||
405 | EXPORT_SYMBOL(allow_signal); | |
406 | ||
407 | int disallow_signal(int sig) | |
408 | { | |
7ed20e1a | 409 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
410 | return -EINVAL; |
411 | ||
412 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 413 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
414 | recalc_sigpending(); |
415 | spin_unlock_irq(¤t->sighand->siglock); | |
416 | return 0; | |
417 | } | |
418 | ||
419 | EXPORT_SYMBOL(disallow_signal); | |
420 | ||
421 | /* | |
422 | * Put all the gunge required to become a kernel thread without | |
423 | * attached user resources in one place where it belongs. | |
424 | */ | |
425 | ||
426 | void daemonize(const char *name, ...) | |
427 | { | |
428 | va_list args; | |
429 | struct fs_struct *fs; | |
430 | sigset_t blocked; | |
431 | ||
432 | va_start(args, name); | |
433 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
434 | va_end(args); | |
435 | ||
436 | /* | |
437 | * If we were started as result of loading a module, close all of the | |
438 | * user space pages. We don't need them, and if we didn't close them | |
439 | * they would be locked into memory. | |
440 | */ | |
441 | exit_mm(current); | |
83144186 RW |
442 | /* |
443 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
444 | * or suspend transition begins right now. | |
445 | */ | |
7b34e428 | 446 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
1da177e4 | 447 | |
8520d7c7 ON |
448 | if (current->nsproxy != &init_nsproxy) { |
449 | get_nsproxy(&init_nsproxy); | |
450 | switch_task_namespaces(current, &init_nsproxy); | |
451 | } | |
297bd42b | 452 | set_special_pids(&init_struct_pid); |
24ec839c | 453 | proc_clear_tty(current); |
1da177e4 LT |
454 | |
455 | /* Block and flush all signals */ | |
456 | sigfillset(&blocked); | |
457 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
458 | flush_signals(current); | |
459 | ||
460 | /* Become as one with the init task */ | |
461 | ||
462 | exit_fs(current); /* current->fs->count--; */ | |
463 | fs = init_task.fs; | |
464 | current->fs = fs; | |
465 | atomic_inc(&fs->count); | |
ab516013 | 466 | |
d4c5e41f | 467 | exit_files(current); |
1da177e4 LT |
468 | current->files = init_task.files; |
469 | atomic_inc(¤t->files->count); | |
470 | ||
49d769d5 | 471 | reparent_to_kthreadd(); |
1da177e4 LT |
472 | } |
473 | ||
474 | EXPORT_SYMBOL(daemonize); | |
475 | ||
858119e1 | 476 | static void close_files(struct files_struct * files) |
1da177e4 LT |
477 | { |
478 | int i, j; | |
badf1662 | 479 | struct fdtable *fdt; |
1da177e4 LT |
480 | |
481 | j = 0; | |
4fb3a538 DS |
482 | |
483 | /* | |
484 | * It is safe to dereference the fd table without RCU or | |
485 | * ->file_lock because this is the last reference to the | |
486 | * files structure. | |
487 | */ | |
badf1662 | 488 | fdt = files_fdtable(files); |
1da177e4 LT |
489 | for (;;) { |
490 | unsigned long set; | |
491 | i = j * __NFDBITS; | |
bbea9f69 | 492 | if (i >= fdt->max_fds) |
1da177e4 | 493 | break; |
badf1662 | 494 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
495 | while (set) { |
496 | if (set & 1) { | |
badf1662 | 497 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 498 | if (file) { |
1da177e4 | 499 | filp_close(file, files); |
944be0b2 IM |
500 | cond_resched(); |
501 | } | |
1da177e4 LT |
502 | } |
503 | i++; | |
504 | set >>= 1; | |
505 | } | |
506 | } | |
507 | } | |
508 | ||
509 | struct files_struct *get_files_struct(struct task_struct *task) | |
510 | { | |
511 | struct files_struct *files; | |
512 | ||
513 | task_lock(task); | |
514 | files = task->files; | |
515 | if (files) | |
516 | atomic_inc(&files->count); | |
517 | task_unlock(task); | |
518 | ||
519 | return files; | |
520 | } | |
521 | ||
7ad5b3a5 | 522 | void put_files_struct(struct files_struct *files) |
1da177e4 | 523 | { |
badf1662 DS |
524 | struct fdtable *fdt; |
525 | ||
1da177e4 LT |
526 | if (atomic_dec_and_test(&files->count)) { |
527 | close_files(files); | |
528 | /* | |
529 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
530 | * If the fdtable was embedded, pass files for freeing |
531 | * at the end of the RCU grace period. Otherwise, | |
532 | * you can free files immediately. | |
1da177e4 | 533 | */ |
badf1662 | 534 | fdt = files_fdtable(files); |
4fd45812 | 535 | if (fdt != &files->fdtab) |
ab2af1f5 | 536 | kmem_cache_free(files_cachep, files); |
01b2d93c | 537 | free_fdtable(fdt); |
1da177e4 LT |
538 | } |
539 | } | |
540 | ||
3b125388 | 541 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 542 | { |
3b125388 | 543 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
544 | struct files_struct *old; |
545 | ||
546 | old = tsk->files; | |
547 | task_lock(tsk); | |
548 | tsk->files = files; | |
549 | task_unlock(tsk); | |
550 | put_files_struct(old); | |
551 | } | |
3b9b8ab6 | 552 | |
1ec7f1dd | 553 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
554 | { |
555 | struct files_struct * files = tsk->files; | |
556 | ||
557 | if (files) { | |
558 | task_lock(tsk); | |
559 | tsk->files = NULL; | |
560 | task_unlock(tsk); | |
561 | put_files_struct(files); | |
562 | } | |
563 | } | |
564 | ||
1ec7f1dd | 565 | void put_fs_struct(struct fs_struct *fs) |
1da177e4 LT |
566 | { |
567 | /* No need to hold fs->lock if we are killing it */ | |
568 | if (atomic_dec_and_test(&fs->count)) { | |
6ac08c39 JB |
569 | path_put(&fs->root); |
570 | path_put(&fs->pwd); | |
571 | if (fs->altroot.dentry) | |
572 | path_put(&fs->altroot); | |
1da177e4 LT |
573 | kmem_cache_free(fs_cachep, fs); |
574 | } | |
575 | } | |
576 | ||
1ec7f1dd | 577 | void exit_fs(struct task_struct *tsk) |
1da177e4 LT |
578 | { |
579 | struct fs_struct * fs = tsk->fs; | |
580 | ||
581 | if (fs) { | |
582 | task_lock(tsk); | |
583 | tsk->fs = NULL; | |
584 | task_unlock(tsk); | |
1ec7f1dd | 585 | put_fs_struct(fs); |
1da177e4 LT |
586 | } |
587 | } | |
588 | ||
1da177e4 LT |
589 | EXPORT_SYMBOL_GPL(exit_fs); |
590 | ||
cf475ad2 BS |
591 | #ifdef CONFIG_MM_OWNER |
592 | /* | |
593 | * Task p is exiting and it owned mm, lets find a new owner for it | |
594 | */ | |
595 | static inline int | |
596 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) | |
597 | { | |
598 | /* | |
599 | * If there are other users of the mm and the owner (us) is exiting | |
600 | * we need to find a new owner to take on the responsibility. | |
601 | */ | |
602 | if (!mm) | |
603 | return 0; | |
604 | if (atomic_read(&mm->mm_users) <= 1) | |
605 | return 0; | |
606 | if (mm->owner != p) | |
607 | return 0; | |
608 | return 1; | |
609 | } | |
610 | ||
611 | void mm_update_next_owner(struct mm_struct *mm) | |
612 | { | |
613 | struct task_struct *c, *g, *p = current; | |
614 | ||
615 | retry: | |
616 | if (!mm_need_new_owner(mm, p)) | |
617 | return; | |
618 | ||
619 | read_lock(&tasklist_lock); | |
620 | /* | |
621 | * Search in the children | |
622 | */ | |
623 | list_for_each_entry(c, &p->children, sibling) { | |
624 | if (c->mm == mm) | |
625 | goto assign_new_owner; | |
626 | } | |
627 | ||
628 | /* | |
629 | * Search in the siblings | |
630 | */ | |
631 | list_for_each_entry(c, &p->parent->children, sibling) { | |
632 | if (c->mm == mm) | |
633 | goto assign_new_owner; | |
634 | } | |
635 | ||
636 | /* | |
637 | * Search through everything else. We should not get | |
638 | * here often | |
639 | */ | |
640 | do_each_thread(g, c) { | |
641 | if (c->mm == mm) | |
642 | goto assign_new_owner; | |
643 | } while_each_thread(g, c); | |
644 | ||
645 | read_unlock(&tasklist_lock); | |
646 | return; | |
647 | ||
648 | assign_new_owner: | |
649 | BUG_ON(c == p); | |
650 | get_task_struct(c); | |
651 | /* | |
652 | * The task_lock protects c->mm from changing. | |
653 | * We always want mm->owner->mm == mm | |
654 | */ | |
655 | task_lock(c); | |
656 | /* | |
657 | * Delay read_unlock() till we have the task_lock() | |
658 | * to ensure that c does not slip away underneath us | |
659 | */ | |
660 | read_unlock(&tasklist_lock); | |
661 | if (c->mm != mm) { | |
662 | task_unlock(c); | |
663 | put_task_struct(c); | |
664 | goto retry; | |
665 | } | |
666 | cgroup_mm_owner_callbacks(mm->owner, c); | |
667 | mm->owner = c; | |
668 | task_unlock(c); | |
669 | put_task_struct(c); | |
670 | } | |
671 | #endif /* CONFIG_MM_OWNER */ | |
672 | ||
1da177e4 LT |
673 | /* |
674 | * Turn us into a lazy TLB process if we | |
675 | * aren't already.. | |
676 | */ | |
408b664a | 677 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
678 | { |
679 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 680 | struct core_state *core_state; |
1da177e4 LT |
681 | |
682 | mm_release(tsk, mm); | |
683 | if (!mm) | |
684 | return; | |
685 | /* | |
686 | * Serialize with any possible pending coredump. | |
999d9fc1 | 687 | * We must hold mmap_sem around checking core_state |
1da177e4 | 688 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 689 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
690 | * group with ->mm != NULL. |
691 | */ | |
692 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
693 | core_state = mm->core_state; |
694 | if (core_state) { | |
695 | struct core_thread self; | |
1da177e4 | 696 | up_read(&mm->mmap_sem); |
c5f1cc8c | 697 | |
b564daf8 ON |
698 | self.task = tsk; |
699 | self.next = xchg(&core_state->dumper.next, &self); | |
700 | /* | |
701 | * Implies mb(), the result of xchg() must be visible | |
702 | * to core_state->dumper. | |
703 | */ | |
704 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
705 | complete(&core_state->startup); | |
1da177e4 | 706 | |
a94e2d40 ON |
707 | for (;;) { |
708 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
709 | if (!self.task) /* see coredump_finish() */ | |
710 | break; | |
711 | schedule(); | |
712 | } | |
713 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
714 | down_read(&mm->mmap_sem); |
715 | } | |
716 | atomic_inc(&mm->mm_count); | |
125e1874 | 717 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
718 | /* more a memory barrier than a real lock */ |
719 | task_lock(tsk); | |
720 | tsk->mm = NULL; | |
721 | up_read(&mm->mmap_sem); | |
722 | enter_lazy_tlb(mm, current); | |
0c1eecfb RW |
723 | /* We don't want this task to be frozen prematurely */ |
724 | clear_freeze_flag(tsk); | |
1da177e4 | 725 | task_unlock(tsk); |
cf475ad2 | 726 | mm_update_next_owner(mm); |
1da177e4 LT |
727 | mmput(mm); |
728 | } | |
729 | ||
666f164f RM |
730 | /* |
731 | * Return nonzero if @parent's children should reap themselves. | |
732 | * | |
733 | * Called with write_lock_irq(&tasklist_lock) held. | |
734 | */ | |
735 | static int ignoring_children(struct task_struct *parent) | |
736 | { | |
737 | int ret; | |
738 | struct sighand_struct *psig = parent->sighand; | |
739 | unsigned long flags; | |
740 | spin_lock_irqsave(&psig->siglock, flags); | |
741 | ret = (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN || | |
742 | (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT)); | |
743 | spin_unlock_irqrestore(&psig->siglock, flags); | |
744 | return ret; | |
745 | } | |
746 | ||
f470021a RM |
747 | /* |
748 | * Detach all tasks we were using ptrace on. | |
749 | * Any that need to be release_task'd are put on the @dead list. | |
750 | * | |
751 | * Called with write_lock(&tasklist_lock) held. | |
752 | */ | |
753 | static void ptrace_exit(struct task_struct *parent, struct list_head *dead) | |
1da177e4 | 754 | { |
f470021a | 755 | struct task_struct *p, *n; |
666f164f | 756 | int ign = -1; |
241ceee0 | 757 | |
f470021a RM |
758 | list_for_each_entry_safe(p, n, &parent->ptraced, ptrace_entry) { |
759 | __ptrace_unlink(p); | |
760 | ||
761 | if (p->exit_state != EXIT_ZOMBIE) | |
762 | continue; | |
763 | ||
764 | /* | |
765 | * If it's a zombie, our attachedness prevented normal | |
766 | * parent notification or self-reaping. Do notification | |
767 | * now if it would have happened earlier. If it should | |
768 | * reap itself, add it to the @dead list. We can't call | |
769 | * release_task() here because we already hold tasklist_lock. | |
770 | * | |
771 | * If it's our own child, there is no notification to do. | |
666f164f RM |
772 | * But if our normal children self-reap, then this child |
773 | * was prevented by ptrace and we must reap it now. | |
1da177e4 | 774 | */ |
f470021a RM |
775 | if (!task_detached(p) && thread_group_empty(p)) { |
776 | if (!same_thread_group(p->real_parent, parent)) | |
777 | do_notify_parent(p, p->exit_signal); | |
666f164f RM |
778 | else { |
779 | if (ign < 0) | |
780 | ign = ignoring_children(parent); | |
781 | if (ign) | |
782 | p->exit_signal = -1; | |
783 | } | |
f470021a | 784 | } |
1da177e4 | 785 | |
f470021a | 786 | if (task_detached(p)) { |
1da177e4 | 787 | /* |
f470021a | 788 | * Mark it as in the process of being reaped. |
1da177e4 | 789 | */ |
f470021a RM |
790 | p->exit_state = EXIT_DEAD; |
791 | list_add(&p->ptrace_entry, dead); | |
1da177e4 LT |
792 | } |
793 | } | |
f470021a RM |
794 | } |
795 | ||
796 | /* | |
797 | * Finish up exit-time ptrace cleanup. | |
798 | * | |
799 | * Called without locks. | |
800 | */ | |
801 | static void ptrace_exit_finish(struct task_struct *parent, | |
802 | struct list_head *dead) | |
803 | { | |
804 | struct task_struct *p, *n; | |
805 | ||
806 | BUG_ON(!list_empty(&parent->ptraced)); | |
807 | ||
808 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
809 | list_del_init(&p->ptrace_entry); | |
810 | release_task(p); | |
811 | } | |
812 | } | |
813 | ||
814 | static void reparent_thread(struct task_struct *p, struct task_struct *father) | |
815 | { | |
816 | if (p->pdeath_signal) | |
817 | /* We already hold the tasklist_lock here. */ | |
818 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | |
819 | ||
820 | list_move_tail(&p->sibling, &p->real_parent->children); | |
1da177e4 | 821 | |
b2b2cbc4 EB |
822 | /* If this is a threaded reparent there is no need to |
823 | * notify anyone anything has happened. | |
824 | */ | |
376e1d25 | 825 | if (same_thread_group(p->real_parent, father)) |
b2b2cbc4 EB |
826 | return; |
827 | ||
828 | /* We don't want people slaying init. */ | |
d839fd4d | 829 | if (!task_detached(p)) |
b2b2cbc4 | 830 | p->exit_signal = SIGCHLD; |
b2b2cbc4 EB |
831 | |
832 | /* If we'd notified the old parent about this child's death, | |
833 | * also notify the new parent. | |
834 | */ | |
f470021a RM |
835 | if (!ptrace_reparented(p) && |
836 | p->exit_state == EXIT_ZOMBIE && | |
d839fd4d | 837 | !task_detached(p) && thread_group_empty(p)) |
b2b2cbc4 EB |
838 | do_notify_parent(p, p->exit_signal); |
839 | ||
f49ee505 | 840 | kill_orphaned_pgrp(p, father); |
1da177e4 LT |
841 | } |
842 | ||
843 | /* | |
844 | * When we die, we re-parent all our children. | |
845 | * Try to give them to another thread in our thread | |
846 | * group, and if no such member exists, give it to | |
84d73786 SB |
847 | * the child reaper process (ie "init") in our pid |
848 | * space. | |
1da177e4 | 849 | */ |
762a24be | 850 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 851 | { |
03ff1797 | 852 | struct task_struct *p, *n, *reaper = father; |
f470021a | 853 | LIST_HEAD(ptrace_dead); |
762a24be ON |
854 | |
855 | write_lock_irq(&tasklist_lock); | |
1da177e4 | 856 | |
f470021a RM |
857 | /* |
858 | * First clean up ptrace if we were using it. | |
859 | */ | |
860 | ptrace_exit(father, &ptrace_dead); | |
861 | ||
1da177e4 LT |
862 | do { |
863 | reaper = next_thread(reaper); | |
864 | if (reaper == father) { | |
88f21d81 | 865 | reaper = task_child_reaper(father); |
1da177e4 LT |
866 | break; |
867 | } | |
762a24be | 868 | } while (reaper->flags & PF_EXITING); |
1da177e4 | 869 | |
03ff1797 | 870 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
84eb646b | 871 | p->real_parent = reaper; |
f470021a RM |
872 | if (p->parent == father) { |
873 | BUG_ON(p->ptrace); | |
874 | p->parent = p->real_parent; | |
875 | } | |
876 | reparent_thread(p, father); | |
1da177e4 | 877 | } |
762a24be ON |
878 | |
879 | write_unlock_irq(&tasklist_lock); | |
880 | BUG_ON(!list_empty(&father->children)); | |
762a24be | 881 | |
f470021a | 882 | ptrace_exit_finish(father, &ptrace_dead); |
1da177e4 LT |
883 | } |
884 | ||
885 | /* | |
886 | * Send signals to all our closest relatives so that they know | |
887 | * to properly mourn us.. | |
888 | */ | |
821c7de7 | 889 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 LT |
890 | { |
891 | int state; | |
1da177e4 | 892 | |
1da177e4 LT |
893 | /* |
894 | * This does two things: | |
895 | * | |
896 | * A. Make init inherit all the child processes | |
897 | * B. Check to see if any process groups have become orphaned | |
898 | * as a result of our exiting, and if they have any stopped | |
899 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
900 | */ | |
762a24be | 901 | forget_original_parent(tsk); |
2e4a7072 | 902 | exit_task_namespaces(tsk); |
1da177e4 | 903 | |
762a24be | 904 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
905 | if (group_dead) |
906 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 907 | |
24728448 | 908 | /* Let father know we died |
1da177e4 LT |
909 | * |
910 | * Thread signals are configurable, but you aren't going to use | |
d4c5e41f | 911 | * that to send signals to arbitary processes. |
1da177e4 LT |
912 | * That stops right now. |
913 | * | |
914 | * If the parent exec id doesn't match the exec id we saved | |
915 | * when we started then we know the parent has changed security | |
916 | * domain. | |
917 | * | |
918 | * If our self_exec id doesn't match our parent_exec_id then | |
919 | * we have changed execution domain as these two values started | |
920 | * the same after a fork. | |
1da177e4 | 921 | */ |
d839fd4d | 922 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
f49ee505 | 923 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
d839fd4d ON |
924 | tsk->self_exec_id != tsk->parent_exec_id) && |
925 | !capable(CAP_KILL)) | |
1da177e4 LT |
926 | tsk->exit_signal = SIGCHLD; |
927 | ||
1da177e4 LT |
928 | /* If something other than our normal parent is ptracing us, then |
929 | * send it a SIGCHLD instead of honoring exit_signal. exit_signal | |
930 | * only has special meaning to our real parent. | |
931 | */ | |
d839fd4d | 932 | if (!task_detached(tsk) && thread_group_empty(tsk)) { |
53b6f9fb ON |
933 | int signal = ptrace_reparented(tsk) ? |
934 | SIGCHLD : tsk->exit_signal; | |
1da177e4 LT |
935 | do_notify_parent(tsk, signal); |
936 | } else if (tsk->ptrace) { | |
937 | do_notify_parent(tsk, SIGCHLD); | |
938 | } | |
939 | ||
940 | state = EXIT_ZOMBIE; | |
d839fd4d | 941 | if (task_detached(tsk) && likely(!tsk->ptrace)) |
1da177e4 LT |
942 | state = EXIT_DEAD; |
943 | tsk->exit_state = state; | |
944 | ||
2800d8d1 | 945 | /* mt-exec, de_thread() is waiting for us */ |
6db840fa ON |
946 | if (thread_group_leader(tsk) && |
947 | tsk->signal->notify_count < 0 && | |
948 | tsk->signal->group_exit_task) | |
949 | wake_up_process(tsk->signal->group_exit_task); | |
950 | ||
1da177e4 LT |
951 | write_unlock_irq(&tasklist_lock); |
952 | ||
1da177e4 LT |
953 | /* If the process is dead, release it - nobody will wait for it */ |
954 | if (state == EXIT_DEAD) | |
955 | release_task(tsk); | |
1da177e4 LT |
956 | } |
957 | ||
e18eecb8 JD |
958 | #ifdef CONFIG_DEBUG_STACK_USAGE |
959 | static void check_stack_usage(void) | |
960 | { | |
961 | static DEFINE_SPINLOCK(low_water_lock); | |
962 | static int lowest_to_date = THREAD_SIZE; | |
963 | unsigned long *n = end_of_stack(current); | |
964 | unsigned long free; | |
965 | ||
966 | while (*n == 0) | |
967 | n++; | |
968 | free = (unsigned long)n - (unsigned long)end_of_stack(current); | |
969 | ||
970 | if (free >= lowest_to_date) | |
971 | return; | |
972 | ||
973 | spin_lock(&low_water_lock); | |
974 | if (free < lowest_to_date) { | |
975 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
976 | "left\n", | |
977 | current->comm, free); | |
978 | lowest_to_date = free; | |
979 | } | |
980 | spin_unlock(&low_water_lock); | |
981 | } | |
982 | #else | |
983 | static inline void check_stack_usage(void) {} | |
984 | #endif | |
985 | ||
84eb646b ON |
986 | static inline void exit_child_reaper(struct task_struct *tsk) |
987 | { | |
88f21d81 | 988 | if (likely(tsk->group_leader != task_child_reaper(tsk))) |
84eb646b ON |
989 | return; |
990 | ||
3eb07c8c SB |
991 | if (tsk->nsproxy->pid_ns == &init_pid_ns) |
992 | panic("Attempted to kill init!"); | |
993 | ||
994 | /* | |
995 | * @tsk is the last thread in the 'cgroup-init' and is exiting. | |
996 | * Terminate all remaining processes in the namespace and reap them | |
997 | * before exiting @tsk. | |
998 | * | |
999 | * Note that @tsk (last thread of cgroup-init) may not necessarily | |
1000 | * be the child-reaper (i.e main thread of cgroup-init) of the | |
1001 | * namespace i.e the child_reaper may have already exited. | |
1002 | * | |
1003 | * Even after a child_reaper exits, we let it inherit orphaned children, | |
1004 | * because, pid_ns->child_reaper remains valid as long as there is | |
1005 | * at least one living sub-thread in the cgroup init. | |
1006 | ||
1007 | * This living sub-thread of the cgroup-init will be notified when | |
1008 | * a child inherited by the 'child-reaper' exits (do_notify_parent() | |
1009 | * uses __group_send_sig_info()). Further, when reaping child processes, | |
1010 | * do_wait() iterates over children of all living sub threads. | |
1011 | ||
1012 | * i.e even though 'child_reaper' thread is listed as the parent of the | |
1013 | * orphaned children, any living sub-thread in the cgroup-init can | |
1014 | * perform the role of the child_reaper. | |
1015 | */ | |
1016 | zap_pid_ns_processes(tsk->nsproxy->pid_ns); | |
84eb646b ON |
1017 | } |
1018 | ||
7ad5b3a5 | 1019 | NORET_TYPE void do_exit(long code) |
1da177e4 LT |
1020 | { |
1021 | struct task_struct *tsk = current; | |
1022 | int group_dead; | |
1023 | ||
1024 | profile_task_exit(tsk); | |
1025 | ||
22e2c507 JA |
1026 | WARN_ON(atomic_read(&tsk->fs_excl)); |
1027 | ||
1da177e4 LT |
1028 | if (unlikely(in_interrupt())) |
1029 | panic("Aiee, killing interrupt handler!"); | |
1030 | if (unlikely(!tsk->pid)) | |
1031 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 1032 | |
30199f5a | 1033 | tracehook_report_exit(&code); |
1da177e4 | 1034 | |
df164db5 AN |
1035 | /* |
1036 | * We're taking recursive faults here in do_exit. Safest is to just | |
1037 | * leave this task alone and wait for reboot. | |
1038 | */ | |
1039 | if (unlikely(tsk->flags & PF_EXITING)) { | |
1040 | printk(KERN_ALERT | |
1041 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
1042 | /* |
1043 | * We can do this unlocked here. The futex code uses | |
1044 | * this flag just to verify whether the pi state | |
1045 | * cleanup has been done or not. In the worst case it | |
1046 | * loops once more. We pretend that the cleanup was | |
1047 | * done as there is no way to return. Either the | |
1048 | * OWNER_DIED bit is set by now or we push the blocked | |
1049 | * task into the wait for ever nirwana as well. | |
1050 | */ | |
1051 | tsk->flags |= PF_EXITPIDONE; | |
afc847b7 AV |
1052 | if (tsk->io_context) |
1053 | exit_io_context(); | |
df164db5 AN |
1054 | set_current_state(TASK_UNINTERRUPTIBLE); |
1055 | schedule(); | |
1056 | } | |
1057 | ||
d12619b5 | 1058 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
1059 | /* |
1060 | * tsk->flags are checked in the futex code to protect against | |
1061 | * an exiting task cleaning up the robust pi futexes. | |
1062 | */ | |
d2ee7198 ON |
1063 | smp_mb(); |
1064 | spin_unlock_wait(&tsk->pi_lock); | |
1da177e4 | 1065 | |
1da177e4 LT |
1066 | if (unlikely(in_atomic())) |
1067 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 1068 | current->comm, task_pid_nr(current), |
1da177e4 LT |
1069 | preempt_count()); |
1070 | ||
1071 | acct_update_integrals(tsk); | |
365e9c87 HD |
1072 | if (tsk->mm) { |
1073 | update_hiwater_rss(tsk->mm); | |
1074 | update_hiwater_vm(tsk->mm); | |
1075 | } | |
1da177e4 | 1076 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 1077 | if (group_dead) { |
84eb646b | 1078 | exit_child_reaper(tsk); |
778e9a9c | 1079 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 1080 | exit_itimers(tsk->signal); |
c3068951 | 1081 | } |
f6ec29a4 | 1082 | acct_collect(code, group_dead); |
42b2dd0a | 1083 | #ifdef CONFIG_FUTEX |
0771dfef IM |
1084 | if (unlikely(tsk->robust_list)) |
1085 | exit_robust_list(tsk); | |
42b2dd0a | 1086 | #ifdef CONFIG_COMPAT |
34f192c6 IM |
1087 | if (unlikely(tsk->compat_robust_list)) |
1088 | compat_exit_robust_list(tsk); | |
42b2dd0a | 1089 | #endif |
34f192c6 | 1090 | #endif |
522ed776 MT |
1091 | if (group_dead) |
1092 | tty_audit_exit(); | |
fa84cb93 AV |
1093 | if (unlikely(tsk->audit_context)) |
1094 | audit_free(tsk); | |
115085ea | 1095 | |
f2ab6d88 | 1096 | tsk->exit_code = code; |
115085ea | 1097 | taskstats_exit(tsk, group_dead); |
c757249a | 1098 | |
1da177e4 LT |
1099 | exit_mm(tsk); |
1100 | ||
0e464814 | 1101 | if (group_dead) |
f6ec29a4 | 1102 | acct_process(); |
1da177e4 | 1103 | exit_sem(tsk); |
1ec7f1dd AV |
1104 | exit_files(tsk); |
1105 | exit_fs(tsk); | |
e18eecb8 | 1106 | check_stack_usage(); |
1da177e4 | 1107 | exit_thread(); |
b4f48b63 | 1108 | cgroup_exit(tsk, 1); |
1da177e4 LT |
1109 | exit_keys(tsk); |
1110 | ||
1111 | if (group_dead && tsk->signal->leader) | |
1112 | disassociate_ctty(1); | |
1113 | ||
a1261f54 | 1114 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 LT |
1115 | if (tsk->binfmt) |
1116 | module_put(tsk->binfmt->module); | |
1117 | ||
9f46080c | 1118 | proc_exit_connector(tsk); |
821c7de7 | 1119 | exit_notify(tsk, group_dead); |
1da177e4 | 1120 | #ifdef CONFIG_NUMA |
f0be3d32 | 1121 | mpol_put(tsk->mempolicy); |
1da177e4 LT |
1122 | tsk->mempolicy = NULL; |
1123 | #endif | |
42b2dd0a | 1124 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
1125 | /* |
1126 | * This must happen late, after the PID is not | |
1127 | * hashed anymore: | |
1128 | */ | |
1129 | if (unlikely(!list_empty(&tsk->pi_state_list))) | |
1130 | exit_pi_state_list(tsk); | |
1131 | if (unlikely(current->pi_state_cache)) | |
1132 | kfree(current->pi_state_cache); | |
42b2dd0a | 1133 | #endif |
de5097c2 | 1134 | /* |
9a11b49a | 1135 | * Make sure we are holding no locks: |
de5097c2 | 1136 | */ |
9a11b49a | 1137 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1138 | /* |
1139 | * We can do this unlocked here. The futex code uses this flag | |
1140 | * just to verify whether the pi state cleanup has been done | |
1141 | * or not. In the worst case it loops once more. | |
1142 | */ | |
1143 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1144 | |
afc847b7 AV |
1145 | if (tsk->io_context) |
1146 | exit_io_context(); | |
1147 | ||
b92ce558 JA |
1148 | if (tsk->splice_pipe) |
1149 | __free_pipe_info(tsk->splice_pipe); | |
1150 | ||
7407251a | 1151 | preempt_disable(); |
55a101f8 | 1152 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1153 | tsk->state = TASK_DEAD; |
7407251a | 1154 | |
1da177e4 LT |
1155 | schedule(); |
1156 | BUG(); | |
1157 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1158 | for (;;) |
1159 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1160 | } |
1161 | ||
012914da RA |
1162 | EXPORT_SYMBOL_GPL(do_exit); |
1163 | ||
1da177e4 LT |
1164 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1165 | { | |
1166 | if (comp) | |
1167 | complete(comp); | |
55a101f8 | 1168 | |
1da177e4 LT |
1169 | do_exit(code); |
1170 | } | |
1171 | ||
1172 | EXPORT_SYMBOL(complete_and_exit); | |
1173 | ||
1174 | asmlinkage long sys_exit(int error_code) | |
1175 | { | |
1176 | do_exit((error_code&0xff)<<8); | |
1177 | } | |
1178 | ||
1da177e4 LT |
1179 | /* |
1180 | * Take down every thread in the group. This is called by fatal signals | |
1181 | * as well as by sys_exit_group (below). | |
1182 | */ | |
1183 | NORET_TYPE void | |
1184 | do_group_exit(int exit_code) | |
1185 | { | |
bfc4b089 ON |
1186 | struct signal_struct *sig = current->signal; |
1187 | ||
1da177e4 LT |
1188 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1189 | ||
bfc4b089 ON |
1190 | if (signal_group_exit(sig)) |
1191 | exit_code = sig->group_exit_code; | |
1da177e4 | 1192 | else if (!thread_group_empty(current)) { |
1da177e4 | 1193 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1194 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1195 | if (signal_group_exit(sig)) |
1da177e4 LT |
1196 | /* Another thread got here before we took the lock. */ |
1197 | exit_code = sig->group_exit_code; | |
1198 | else { | |
1da177e4 | 1199 | sig->group_exit_code = exit_code; |
ed5d2cac | 1200 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1201 | zap_other_threads(current); |
1202 | } | |
1203 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1204 | } |
1205 | ||
1206 | do_exit(exit_code); | |
1207 | /* NOTREACHED */ | |
1208 | } | |
1209 | ||
1210 | /* | |
1211 | * this kills every thread in the thread group. Note that any externally | |
1212 | * wait4()-ing process will get the correct exit code - even if this | |
1213 | * thread is not the thread group leader. | |
1214 | */ | |
1215 | asmlinkage void sys_exit_group(int error_code) | |
1216 | { | |
1217 | do_group_exit((error_code & 0xff) << 8); | |
1218 | } | |
1219 | ||
161550d7 EB |
1220 | static struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
1221 | { | |
1222 | struct pid *pid = NULL; | |
1223 | if (type == PIDTYPE_PID) | |
1224 | pid = task->pids[type].pid; | |
1225 | else if (type < PIDTYPE_MAX) | |
1226 | pid = task->group_leader->pids[type].pid; | |
1227 | return pid; | |
1228 | } | |
1229 | ||
1230 | static int eligible_child(enum pid_type type, struct pid *pid, int options, | |
1231 | struct task_struct *p) | |
1da177e4 | 1232 | { |
73243284 RM |
1233 | int err; |
1234 | ||
161550d7 EB |
1235 | if (type < PIDTYPE_MAX) { |
1236 | if (task_pid_type(p, type) != pid) | |
1da177e4 LT |
1237 | return 0; |
1238 | } | |
1239 | ||
1da177e4 LT |
1240 | /* Wait for all children (clone and not) if __WALL is set; |
1241 | * otherwise, wait for clone children *only* if __WCLONE is | |
1242 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1243 | * A "clone" child here is one that reports to its parent | |
1244 | * using a signal other than SIGCHLD.) */ | |
1245 | if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0)) | |
1246 | && !(options & __WALL)) | |
1247 | return 0; | |
1da177e4 | 1248 | |
73243284 | 1249 | err = security_task_wait(p); |
14dd0b81 RM |
1250 | if (err) |
1251 | return err; | |
1da177e4 | 1252 | |
14dd0b81 | 1253 | return 1; |
1da177e4 LT |
1254 | } |
1255 | ||
36c8b586 | 1256 | static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid, |
1da177e4 LT |
1257 | int why, int status, |
1258 | struct siginfo __user *infop, | |
1259 | struct rusage __user *rusagep) | |
1260 | { | |
1261 | int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0; | |
36c8b586 | 1262 | |
1da177e4 LT |
1263 | put_task_struct(p); |
1264 | if (!retval) | |
1265 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1266 | if (!retval) | |
1267 | retval = put_user(0, &infop->si_errno); | |
1268 | if (!retval) | |
1269 | retval = put_user((short)why, &infop->si_code); | |
1270 | if (!retval) | |
1271 | retval = put_user(pid, &infop->si_pid); | |
1272 | if (!retval) | |
1273 | retval = put_user(uid, &infop->si_uid); | |
1274 | if (!retval) | |
1275 | retval = put_user(status, &infop->si_status); | |
1276 | if (!retval) | |
1277 | retval = pid; | |
1278 | return retval; | |
1279 | } | |
1280 | ||
1281 | /* | |
1282 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1283 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1284 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1285 | * released the lock and the system call should return. | |
1286 | */ | |
98abed02 | 1287 | static int wait_task_zombie(struct task_struct *p, int options, |
1da177e4 LT |
1288 | struct siginfo __user *infop, |
1289 | int __user *stat_addr, struct rusage __user *ru) | |
1290 | { | |
1291 | unsigned long state; | |
2f4e6e2a | 1292 | int retval, status, traced; |
6c5f3e7b | 1293 | pid_t pid = task_pid_vnr(p); |
1da177e4 | 1294 | |
98abed02 RM |
1295 | if (!likely(options & WEXITED)) |
1296 | return 0; | |
1297 | ||
1298 | if (unlikely(options & WNOWAIT)) { | |
1da177e4 LT |
1299 | uid_t uid = p->uid; |
1300 | int exit_code = p->exit_code; | |
1301 | int why, status; | |
1302 | ||
1da177e4 LT |
1303 | get_task_struct(p); |
1304 | read_unlock(&tasklist_lock); | |
1305 | if ((exit_code & 0x7f) == 0) { | |
1306 | why = CLD_EXITED; | |
1307 | status = exit_code >> 8; | |
1308 | } else { | |
1309 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1310 | status = exit_code & 0x7f; | |
1311 | } | |
1312 | return wait_noreap_copyout(p, pid, uid, why, | |
1313 | status, infop, ru); | |
1314 | } | |
1315 | ||
1316 | /* | |
1317 | * Try to move the task's state to DEAD | |
1318 | * only one thread is allowed to do this: | |
1319 | */ | |
1320 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1321 | if (state != EXIT_ZOMBIE) { | |
1322 | BUG_ON(state != EXIT_DEAD); | |
1323 | return 0; | |
1324 | } | |
1da177e4 | 1325 | |
53b6f9fb | 1326 | traced = ptrace_reparented(p); |
2f4e6e2a ON |
1327 | |
1328 | if (likely(!traced)) { | |
3795e161 JJ |
1329 | struct signal_struct *psig; |
1330 | struct signal_struct *sig; | |
1331 | ||
1da177e4 LT |
1332 | /* |
1333 | * The resource counters for the group leader are in its | |
1334 | * own task_struct. Those for dead threads in the group | |
1335 | * are in its signal_struct, as are those for the child | |
1336 | * processes it has previously reaped. All these | |
1337 | * accumulate in the parent's signal_struct c* fields. | |
1338 | * | |
1339 | * We don't bother to take a lock here to protect these | |
1340 | * p->signal fields, because they are only touched by | |
1341 | * __exit_signal, which runs with tasklist_lock | |
1342 | * write-locked anyway, and so is excluded here. We do | |
1343 | * need to protect the access to p->parent->signal fields, | |
1344 | * as other threads in the parent group can be right | |
1345 | * here reaping other children at the same time. | |
1346 | */ | |
1347 | spin_lock_irq(&p->parent->sighand->siglock); | |
3795e161 JJ |
1348 | psig = p->parent->signal; |
1349 | sig = p->signal; | |
1350 | psig->cutime = | |
1351 | cputime_add(psig->cutime, | |
1da177e4 | 1352 | cputime_add(p->utime, |
3795e161 JJ |
1353 | cputime_add(sig->utime, |
1354 | sig->cutime))); | |
1355 | psig->cstime = | |
1356 | cputime_add(psig->cstime, | |
1da177e4 | 1357 | cputime_add(p->stime, |
3795e161 JJ |
1358 | cputime_add(sig->stime, |
1359 | sig->cstime))); | |
9ac52315 LV |
1360 | psig->cgtime = |
1361 | cputime_add(psig->cgtime, | |
1362 | cputime_add(p->gtime, | |
1363 | cputime_add(sig->gtime, | |
1364 | sig->cgtime))); | |
3795e161 JJ |
1365 | psig->cmin_flt += |
1366 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1367 | psig->cmaj_flt += | |
1368 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1369 | psig->cnvcsw += | |
1370 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1371 | psig->cnivcsw += | |
1372 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1373 | psig->cinblock += |
1374 | task_io_get_inblock(p) + | |
1375 | sig->inblock + sig->cinblock; | |
1376 | psig->coublock += | |
1377 | task_io_get_oublock(p) + | |
1378 | sig->oublock + sig->coublock; | |
297c5d92 AR |
1379 | #ifdef CONFIG_TASK_XACCT |
1380 | psig->rchar += p->rchar + sig->rchar; | |
1381 | psig->wchar += p->wchar + sig->wchar; | |
1382 | psig->syscr += p->syscr + sig->syscr; | |
1383 | psig->syscw += p->syscw + sig->syscw; | |
1384 | #endif /* CONFIG_TASK_XACCT */ | |
1385 | #ifdef CONFIG_TASK_IO_ACCOUNTING | |
1386 | psig->ioac.read_bytes += | |
1387 | p->ioac.read_bytes + sig->ioac.read_bytes; | |
1388 | psig->ioac.write_bytes += | |
1389 | p->ioac.write_bytes + sig->ioac.write_bytes; | |
1390 | psig->ioac.cancelled_write_bytes += | |
1391 | p->ioac.cancelled_write_bytes + | |
1392 | sig->ioac.cancelled_write_bytes; | |
1393 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
1da177e4 LT |
1394 | spin_unlock_irq(&p->parent->sighand->siglock); |
1395 | } | |
1396 | ||
1397 | /* | |
1398 | * Now we are sure this task is interesting, and no other | |
1399 | * thread can reap it because we set its state to EXIT_DEAD. | |
1400 | */ | |
1401 | read_unlock(&tasklist_lock); | |
1402 | ||
1403 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1404 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) | |
1405 | ? p->signal->group_exit_code : p->exit_code; | |
1406 | if (!retval && stat_addr) | |
1407 | retval = put_user(status, stat_addr); | |
1408 | if (!retval && infop) | |
1409 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1410 | if (!retval && infop) | |
1411 | retval = put_user(0, &infop->si_errno); | |
1412 | if (!retval && infop) { | |
1413 | int why; | |
1414 | ||
1415 | if ((status & 0x7f) == 0) { | |
1416 | why = CLD_EXITED; | |
1417 | status >>= 8; | |
1418 | } else { | |
1419 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1420 | status &= 0x7f; | |
1421 | } | |
1422 | retval = put_user((short)why, &infop->si_code); | |
1423 | if (!retval) | |
1424 | retval = put_user(status, &infop->si_status); | |
1425 | } | |
1426 | if (!retval && infop) | |
3a515e4a | 1427 | retval = put_user(pid, &infop->si_pid); |
1da177e4 LT |
1428 | if (!retval && infop) |
1429 | retval = put_user(p->uid, &infop->si_uid); | |
2f4e6e2a | 1430 | if (!retval) |
3a515e4a | 1431 | retval = pid; |
2f4e6e2a ON |
1432 | |
1433 | if (traced) { | |
1da177e4 | 1434 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1435 | /* We dropped tasklist, ptracer could die and untrace */ |
1436 | ptrace_unlink(p); | |
1437 | /* | |
1438 | * If this is not a detached task, notify the parent. | |
1439 | * If it's still not detached after that, don't release | |
1440 | * it now. | |
1441 | */ | |
d839fd4d | 1442 | if (!task_detached(p)) { |
2f4e6e2a | 1443 | do_notify_parent(p, p->exit_signal); |
d839fd4d | 1444 | if (!task_detached(p)) { |
2f4e6e2a ON |
1445 | p->exit_state = EXIT_ZOMBIE; |
1446 | p = NULL; | |
1da177e4 LT |
1447 | } |
1448 | } | |
1449 | write_unlock_irq(&tasklist_lock); | |
1450 | } | |
1451 | if (p != NULL) | |
1452 | release_task(p); | |
2f4e6e2a | 1453 | |
1da177e4 LT |
1454 | return retval; |
1455 | } | |
1456 | ||
1457 | /* | |
1458 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1459 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1460 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1461 | * released the lock and the system call should return. | |
1462 | */ | |
f470021a | 1463 | static int wait_task_stopped(int ptrace, struct task_struct *p, |
98abed02 | 1464 | int options, struct siginfo __user *infop, |
1da177e4 LT |
1465 | int __user *stat_addr, struct rusage __user *ru) |
1466 | { | |
ee7c82da ON |
1467 | int retval, exit_code, why; |
1468 | uid_t uid = 0; /* unneeded, required by compiler */ | |
c8950783 | 1469 | pid_t pid; |
1da177e4 | 1470 | |
f470021a | 1471 | if (!(options & WUNTRACED)) |
98abed02 RM |
1472 | return 0; |
1473 | ||
ee7c82da ON |
1474 | exit_code = 0; |
1475 | spin_lock_irq(&p->sighand->siglock); | |
1476 | ||
1477 | if (unlikely(!task_is_stopped_or_traced(p))) | |
1478 | goto unlock_sig; | |
1479 | ||
f470021a | 1480 | if (!ptrace && p->signal->group_stop_count > 0) |
1da177e4 LT |
1481 | /* |
1482 | * A group stop is in progress and this is the group leader. | |
1483 | * We won't report until all threads have stopped. | |
1484 | */ | |
ee7c82da ON |
1485 | goto unlock_sig; |
1486 | ||
1487 | exit_code = p->exit_code; | |
1488 | if (!exit_code) | |
1489 | goto unlock_sig; | |
1490 | ||
98abed02 | 1491 | if (!unlikely(options & WNOWAIT)) |
ee7c82da ON |
1492 | p->exit_code = 0; |
1493 | ||
1494 | uid = p->uid; | |
1495 | unlock_sig: | |
1496 | spin_unlock_irq(&p->sighand->siglock); | |
1497 | if (!exit_code) | |
1da177e4 LT |
1498 | return 0; |
1499 | ||
1500 | /* | |
1501 | * Now we are pretty sure this task is interesting. | |
1502 | * Make sure it doesn't get reaped out from under us while we | |
1503 | * give up the lock and then examine it below. We don't want to | |
1504 | * keep holding onto the tasklist_lock while we call getrusage and | |
1505 | * possibly take page faults for user memory. | |
1506 | */ | |
1507 | get_task_struct(p); | |
6c5f3e7b | 1508 | pid = task_pid_vnr(p); |
f470021a | 1509 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1510 | read_unlock(&tasklist_lock); |
1511 | ||
98abed02 | 1512 | if (unlikely(options & WNOWAIT)) |
1da177e4 | 1513 | return wait_noreap_copyout(p, pid, uid, |
e6ceb32a | 1514 | why, exit_code, |
1da177e4 | 1515 | infop, ru); |
1da177e4 LT |
1516 | |
1517 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1518 | if (!retval && stat_addr) | |
1519 | retval = put_user((exit_code << 8) | 0x7f, stat_addr); | |
1520 | if (!retval && infop) | |
1521 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1522 | if (!retval && infop) | |
1523 | retval = put_user(0, &infop->si_errno); | |
1524 | if (!retval && infop) | |
6efcae46 | 1525 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1526 | if (!retval && infop) |
1527 | retval = put_user(exit_code, &infop->si_status); | |
1528 | if (!retval && infop) | |
c8950783 | 1529 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1530 | if (!retval && infop) |
ee7c82da | 1531 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1532 | if (!retval) |
c8950783 | 1533 | retval = pid; |
1da177e4 LT |
1534 | put_task_struct(p); |
1535 | ||
1536 | BUG_ON(!retval); | |
1537 | return retval; | |
1538 | } | |
1539 | ||
1540 | /* | |
1541 | * Handle do_wait work for one task in a live, non-stopped state. | |
1542 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1543 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1544 | * released the lock and the system call should return. | |
1545 | */ | |
98abed02 | 1546 | static int wait_task_continued(struct task_struct *p, int options, |
1da177e4 LT |
1547 | struct siginfo __user *infop, |
1548 | int __user *stat_addr, struct rusage __user *ru) | |
1549 | { | |
1550 | int retval; | |
1551 | pid_t pid; | |
1552 | uid_t uid; | |
1553 | ||
98abed02 RM |
1554 | if (!unlikely(options & WCONTINUED)) |
1555 | return 0; | |
1556 | ||
1da177e4 LT |
1557 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1558 | return 0; | |
1559 | ||
1560 | spin_lock_irq(&p->sighand->siglock); | |
1561 | /* Re-check with the lock held. */ | |
1562 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1563 | spin_unlock_irq(&p->sighand->siglock); | |
1564 | return 0; | |
1565 | } | |
98abed02 | 1566 | if (!unlikely(options & WNOWAIT)) |
1da177e4 LT |
1567 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
1568 | spin_unlock_irq(&p->sighand->siglock); | |
1569 | ||
6c5f3e7b | 1570 | pid = task_pid_vnr(p); |
1da177e4 LT |
1571 | uid = p->uid; |
1572 | get_task_struct(p); | |
1573 | read_unlock(&tasklist_lock); | |
1574 | ||
1575 | if (!infop) { | |
1576 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1577 | put_task_struct(p); | |
1578 | if (!retval && stat_addr) | |
1579 | retval = put_user(0xffff, stat_addr); | |
1580 | if (!retval) | |
3a515e4a | 1581 | retval = pid; |
1da177e4 LT |
1582 | } else { |
1583 | retval = wait_noreap_copyout(p, pid, uid, | |
1584 | CLD_CONTINUED, SIGCONT, | |
1585 | infop, ru); | |
1586 | BUG_ON(retval == 0); | |
1587 | } | |
1588 | ||
1589 | return retval; | |
1590 | } | |
1591 | ||
98abed02 RM |
1592 | /* |
1593 | * Consider @p for a wait by @parent. | |
1594 | * | |
1595 | * -ECHILD should be in *@notask_error before the first call. | |
1596 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | |
1597 | * Returns zero if the search for a child should continue; | |
14dd0b81 RM |
1598 | * then *@notask_error is 0 if @p is an eligible child, |
1599 | * or another error from security_task_wait(), or still -ECHILD. | |
98abed02 | 1600 | */ |
f470021a | 1601 | static int wait_consider_task(struct task_struct *parent, int ptrace, |
98abed02 RM |
1602 | struct task_struct *p, int *notask_error, |
1603 | enum pid_type type, struct pid *pid, int options, | |
1604 | struct siginfo __user *infop, | |
1605 | int __user *stat_addr, struct rusage __user *ru) | |
1606 | { | |
1607 | int ret = eligible_child(type, pid, options, p); | |
14dd0b81 | 1608 | if (!ret) |
98abed02 RM |
1609 | return ret; |
1610 | ||
14dd0b81 RM |
1611 | if (unlikely(ret < 0)) { |
1612 | /* | |
1613 | * If we have not yet seen any eligible child, | |
1614 | * then let this error code replace -ECHILD. | |
1615 | * A permission error will give the user a clue | |
1616 | * to look for security policy problems, rather | |
1617 | * than for mysterious wait bugs. | |
1618 | */ | |
1619 | if (*notask_error) | |
1620 | *notask_error = ret; | |
1621 | } | |
1622 | ||
f470021a RM |
1623 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1624 | /* | |
1625 | * This child is hidden by ptrace. | |
1626 | * We aren't allowed to see it now, but eventually we will. | |
1627 | */ | |
1628 | *notask_error = 0; | |
1629 | return 0; | |
1630 | } | |
1631 | ||
98abed02 RM |
1632 | if (p->exit_state == EXIT_DEAD) |
1633 | return 0; | |
1634 | ||
1635 | /* | |
1636 | * We don't reap group leaders with subthreads. | |
1637 | */ | |
1638 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) | |
1639 | return wait_task_zombie(p, options, infop, stat_addr, ru); | |
1640 | ||
1641 | /* | |
1642 | * It's stopped or running now, so it might | |
1643 | * later continue, exit, or stop again. | |
1644 | */ | |
1645 | *notask_error = 0; | |
1646 | ||
1647 | if (task_is_stopped_or_traced(p)) | |
f470021a RM |
1648 | return wait_task_stopped(ptrace, p, options, |
1649 | infop, stat_addr, ru); | |
98abed02 RM |
1650 | |
1651 | return wait_task_continued(p, options, infop, stat_addr, ru); | |
1652 | } | |
1653 | ||
1654 | /* | |
1655 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1656 | * | |
1657 | * -ECHILD should be in *@notask_error before the first call. | |
1658 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | |
1659 | * Returns zero if the search for a child should continue; then | |
14dd0b81 RM |
1660 | * *@notask_error is 0 if there were any eligible children, |
1661 | * or another error from security_task_wait(), or still -ECHILD. | |
98abed02 RM |
1662 | */ |
1663 | static int do_wait_thread(struct task_struct *tsk, int *notask_error, | |
1664 | enum pid_type type, struct pid *pid, int options, | |
1665 | struct siginfo __user *infop, int __user *stat_addr, | |
1666 | struct rusage __user *ru) | |
1667 | { | |
1668 | struct task_struct *p; | |
1669 | ||
1670 | list_for_each_entry(p, &tsk->children, sibling) { | |
f470021a RM |
1671 | /* |
1672 | * Do not consider detached threads. | |
1673 | */ | |
1674 | if (!task_detached(p)) { | |
1675 | int ret = wait_consider_task(tsk, 0, p, notask_error, | |
1676 | type, pid, options, | |
1677 | infop, stat_addr, ru); | |
1678 | if (ret) | |
1679 | return ret; | |
1680 | } | |
98abed02 RM |
1681 | } |
1682 | ||
1683 | return 0; | |
1684 | } | |
1685 | ||
1686 | static int ptrace_do_wait(struct task_struct *tsk, int *notask_error, | |
1687 | enum pid_type type, struct pid *pid, int options, | |
1688 | struct siginfo __user *infop, int __user *stat_addr, | |
1689 | struct rusage __user *ru) | |
1690 | { | |
1691 | struct task_struct *p; | |
1692 | ||
1693 | /* | |
f470021a | 1694 | * Traditionally we see ptrace'd stopped tasks regardless of options. |
98abed02 | 1695 | */ |
f470021a | 1696 | options |= WUNTRACED; |
98abed02 | 1697 | |
f470021a RM |
1698 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
1699 | int ret = wait_consider_task(tsk, 1, p, notask_error, | |
1700 | type, pid, options, | |
1701 | infop, stat_addr, ru); | |
1702 | if (ret) | |
98abed02 | 1703 | return ret; |
98abed02 RM |
1704 | } |
1705 | ||
1706 | return 0; | |
1707 | } | |
1708 | ||
161550d7 EB |
1709 | static long do_wait(enum pid_type type, struct pid *pid, int options, |
1710 | struct siginfo __user *infop, int __user *stat_addr, | |
1711 | struct rusage __user *ru) | |
1da177e4 LT |
1712 | { |
1713 | DECLARE_WAITQUEUE(wait, current); | |
1714 | struct task_struct *tsk; | |
98abed02 | 1715 | int retval; |
1da177e4 LT |
1716 | |
1717 | add_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1718 | repeat: | |
98abed02 RM |
1719 | /* |
1720 | * If there is nothing that can match our critiera just get out. | |
1721 | * We will clear @retval to zero if we see any child that might later | |
1722 | * match our criteria, even if we are not able to reap it yet. | |
1723 | */ | |
161550d7 EB |
1724 | retval = -ECHILD; |
1725 | if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type]))) | |
1726 | goto end; | |
1727 | ||
1da177e4 LT |
1728 | current->state = TASK_INTERRUPTIBLE; |
1729 | read_lock(&tasklist_lock); | |
1730 | tsk = current; | |
1731 | do { | |
98abed02 RM |
1732 | int tsk_result = do_wait_thread(tsk, &retval, |
1733 | type, pid, options, | |
1734 | infop, stat_addr, ru); | |
1735 | if (!tsk_result) | |
1736 | tsk_result = ptrace_do_wait(tsk, &retval, | |
1737 | type, pid, options, | |
1738 | infop, stat_addr, ru); | |
1739 | if (tsk_result) { | |
1740 | /* | |
1741 | * tasklist_lock is unlocked and we have a final result. | |
1742 | */ | |
1743 | retval = tsk_result; | |
1744 | goto end; | |
1da177e4 | 1745 | } |
98abed02 | 1746 | |
1da177e4 LT |
1747 | if (options & __WNOTHREAD) |
1748 | break; | |
1749 | tsk = next_thread(tsk); | |
125e1874 | 1750 | BUG_ON(tsk->signal != current->signal); |
1da177e4 | 1751 | } while (tsk != current); |
1da177e4 | 1752 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1753 | |
98abed02 | 1754 | if (!retval && !(options & WNOHANG)) { |
1da177e4 | 1755 | retval = -ERESTARTSYS; |
98abed02 RM |
1756 | if (!signal_pending(current)) { |
1757 | schedule(); | |
1758 | goto repeat; | |
1759 | } | |
1da177e4 | 1760 | } |
98abed02 | 1761 | |
1da177e4 LT |
1762 | end: |
1763 | current->state = TASK_RUNNING; | |
1764 | remove_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1765 | if (infop) { | |
1766 | if (retval > 0) | |
9cbab810 | 1767 | retval = 0; |
1da177e4 LT |
1768 | else { |
1769 | /* | |
1770 | * For a WNOHANG return, clear out all the fields | |
1771 | * we would set so the user can easily tell the | |
1772 | * difference. | |
1773 | */ | |
1774 | if (!retval) | |
1775 | retval = put_user(0, &infop->si_signo); | |
1776 | if (!retval) | |
1777 | retval = put_user(0, &infop->si_errno); | |
1778 | if (!retval) | |
1779 | retval = put_user(0, &infop->si_code); | |
1780 | if (!retval) | |
1781 | retval = put_user(0, &infop->si_pid); | |
1782 | if (!retval) | |
1783 | retval = put_user(0, &infop->si_uid); | |
1784 | if (!retval) | |
1785 | retval = put_user(0, &infop->si_status); | |
1786 | } | |
1787 | } | |
1788 | return retval; | |
1789 | } | |
1790 | ||
161550d7 | 1791 | asmlinkage long sys_waitid(int which, pid_t upid, |
1da177e4 LT |
1792 | struct siginfo __user *infop, int options, |
1793 | struct rusage __user *ru) | |
1794 | { | |
161550d7 EB |
1795 | struct pid *pid = NULL; |
1796 | enum pid_type type; | |
1da177e4 LT |
1797 | long ret; |
1798 | ||
1799 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1800 | return -EINVAL; | |
1801 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1802 | return -EINVAL; | |
1803 | ||
1804 | switch (which) { | |
1805 | case P_ALL: | |
161550d7 | 1806 | type = PIDTYPE_MAX; |
1da177e4 LT |
1807 | break; |
1808 | case P_PID: | |
161550d7 EB |
1809 | type = PIDTYPE_PID; |
1810 | if (upid <= 0) | |
1da177e4 LT |
1811 | return -EINVAL; |
1812 | break; | |
1813 | case P_PGID: | |
161550d7 EB |
1814 | type = PIDTYPE_PGID; |
1815 | if (upid <= 0) | |
1da177e4 | 1816 | return -EINVAL; |
1da177e4 LT |
1817 | break; |
1818 | default: | |
1819 | return -EINVAL; | |
1820 | } | |
1821 | ||
161550d7 EB |
1822 | if (type < PIDTYPE_MAX) |
1823 | pid = find_get_pid(upid); | |
1824 | ret = do_wait(type, pid, options, infop, NULL, ru); | |
1825 | put_pid(pid); | |
1da177e4 LT |
1826 | |
1827 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1828 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1829 | return ret; |
1830 | } | |
1831 | ||
161550d7 | 1832 | asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr, |
1da177e4 LT |
1833 | int options, struct rusage __user *ru) |
1834 | { | |
161550d7 EB |
1835 | struct pid *pid = NULL; |
1836 | enum pid_type type; | |
1da177e4 LT |
1837 | long ret; |
1838 | ||
1839 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1840 | __WNOTHREAD|__WCLONE|__WALL)) | |
1841 | return -EINVAL; | |
161550d7 EB |
1842 | |
1843 | if (upid == -1) | |
1844 | type = PIDTYPE_MAX; | |
1845 | else if (upid < 0) { | |
1846 | type = PIDTYPE_PGID; | |
1847 | pid = find_get_pid(-upid); | |
1848 | } else if (upid == 0) { | |
1849 | type = PIDTYPE_PGID; | |
1850 | pid = get_pid(task_pgrp(current)); | |
1851 | } else /* upid > 0 */ { | |
1852 | type = PIDTYPE_PID; | |
1853 | pid = find_get_pid(upid); | |
1854 | } | |
1855 | ||
1856 | ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru); | |
1857 | put_pid(pid); | |
1da177e4 LT |
1858 | |
1859 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1860 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1861 | return ret; |
1862 | } | |
1863 | ||
1864 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1865 | ||
1866 | /* | |
1867 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1868 | * implemented by calling sys_wait4() from libc.a. | |
1869 | */ | |
1870 | asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options) | |
1871 | { | |
1872 | return sys_wait4(pid, stat_addr, options, NULL); | |
1873 | } | |
1874 | ||
1875 | #endif |