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 | 166 | |
1da177e4 LT |
167 | void release_task(struct task_struct * p) |
168 | { | |
36c8b586 | 169 | struct task_struct *leader; |
1da177e4 | 170 | int zap_leader; |
1f09f974 | 171 | repeat: |
dae33574 | 172 | tracehook_prepare_release_task(p); |
1da177e4 | 173 | atomic_dec(&p->user->processes); |
60347f67 | 174 | proc_flush_task(p); |
1da177e4 | 175 | write_lock_irq(&tasklist_lock); |
dae33574 | 176 | tracehook_finish_release_task(p); |
1da177e4 | 177 | __exit_signal(p); |
35f5cad8 | 178 | |
1da177e4 LT |
179 | /* |
180 | * If we are the last non-leader member of the thread | |
181 | * group, and the leader is zombie, then notify the | |
182 | * group leader's parent process. (if it wants notification.) | |
183 | */ | |
184 | zap_leader = 0; | |
185 | leader = p->group_leader; | |
186 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
d839fd4d | 187 | BUG_ON(task_detached(leader)); |
1da177e4 LT |
188 | do_notify_parent(leader, leader->exit_signal); |
189 | /* | |
190 | * If we were the last child thread and the leader has | |
191 | * exited already, and the leader's parent ignores SIGCHLD, | |
192 | * then we are the one who should release the leader. | |
193 | * | |
194 | * do_notify_parent() will have marked it self-reaping in | |
195 | * that case. | |
196 | */ | |
d839fd4d | 197 | zap_leader = task_detached(leader); |
dae33574 RM |
198 | |
199 | /* | |
200 | * This maintains the invariant that release_task() | |
201 | * only runs on a task in EXIT_DEAD, just for sanity. | |
202 | */ | |
203 | if (zap_leader) | |
204 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
205 | } |
206 | ||
1da177e4 | 207 | write_unlock_irq(&tasklist_lock); |
1da177e4 | 208 | release_thread(p); |
8c7904a0 | 209 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
210 | |
211 | p = leader; | |
212 | if (unlikely(zap_leader)) | |
213 | goto repeat; | |
214 | } | |
215 | ||
1da177e4 LT |
216 | /* |
217 | * This checks not only the pgrp, but falls back on the pid if no | |
218 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
219 | * without this... | |
04a2e6a5 EB |
220 | * |
221 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 222 | */ |
04a2e6a5 | 223 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
224 | { |
225 | struct task_struct *p; | |
04a2e6a5 | 226 | struct pid *sid = NULL; |
62dfb554 | 227 | |
04a2e6a5 | 228 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 229 | if (p == NULL) |
04a2e6a5 | 230 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 231 | if (p != NULL) |
04a2e6a5 | 232 | sid = task_session(p); |
62dfb554 | 233 | |
1da177e4 LT |
234 | return sid; |
235 | } | |
236 | ||
237 | /* | |
238 | * Determine if a process group is "orphaned", according to the POSIX | |
239 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
240 | * by terminal-generated stop signals. Newly orphaned process groups are | |
241 | * to receive a SIGHUP and a SIGCONT. | |
242 | * | |
243 | * "I ask you, have you ever known what it is to be an orphan?" | |
244 | */ | |
0475ac08 | 245 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
246 | { |
247 | struct task_struct *p; | |
1da177e4 | 248 | |
0475ac08 | 249 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
05e83df6 ON |
250 | if ((p == ignored_task) || |
251 | (p->exit_state && thread_group_empty(p)) || | |
252 | is_global_init(p->real_parent)) | |
1da177e4 | 253 | continue; |
05e83df6 | 254 | |
0475ac08 | 255 | if (task_pgrp(p->real_parent) != pgrp && |
05e83df6 ON |
256 | task_session(p->real_parent) == task_session(p)) |
257 | return 0; | |
0475ac08 | 258 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
05e83df6 ON |
259 | |
260 | return 1; | |
1da177e4 LT |
261 | } |
262 | ||
3e7cd6c4 | 263 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
264 | { |
265 | int retval; | |
266 | ||
267 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 268 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
269 | read_unlock(&tasklist_lock); |
270 | ||
271 | return retval; | |
272 | } | |
273 | ||
0475ac08 | 274 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
275 | { |
276 | int retval = 0; | |
277 | struct task_struct *p; | |
278 | ||
0475ac08 | 279 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
338077e5 | 280 | if (!task_is_stopped(p)) |
1da177e4 | 281 | continue; |
1da177e4 LT |
282 | retval = 1; |
283 | break; | |
0475ac08 | 284 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
285 | return retval; |
286 | } | |
287 | ||
f49ee505 ON |
288 | /* |
289 | * Check to see if any process groups have become orphaned as | |
290 | * a result of our exiting, and if they have any stopped jobs, | |
291 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
292 | */ | |
293 | static void | |
294 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) | |
295 | { | |
296 | struct pid *pgrp = task_pgrp(tsk); | |
297 | struct task_struct *ignored_task = tsk; | |
298 | ||
299 | if (!parent) | |
300 | /* exit: our father is in a different pgrp than | |
301 | * we are and we were the only connection outside. | |
302 | */ | |
303 | parent = tsk->real_parent; | |
304 | else | |
305 | /* reparent: our child is in a different pgrp than | |
306 | * we are, and it was the only connection outside. | |
307 | */ | |
308 | ignored_task = NULL; | |
309 | ||
310 | if (task_pgrp(parent) != pgrp && | |
311 | task_session(parent) == task_session(tsk) && | |
312 | will_become_orphaned_pgrp(pgrp, ignored_task) && | |
313 | has_stopped_jobs(pgrp)) { | |
314 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
315 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
316 | } | |
317 | } | |
318 | ||
1da177e4 | 319 | /** |
49d769d5 | 320 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
321 | * |
322 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
323 | * it ever exits, it should generally reparent itself to kthreadd so it |
324 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
325 | * |
326 | * The various task state such as scheduling policy and priority may have | |
327 | * been inherited from a user process, so we reset them to sane values here. | |
328 | * | |
49d769d5 | 329 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 330 | */ |
49d769d5 | 331 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
332 | { |
333 | write_lock_irq(&tasklist_lock); | |
334 | ||
335 | ptrace_unlink(current); | |
336 | /* Reparent to init */ | |
49d769d5 | 337 | current->real_parent = current->parent = kthreadd_task; |
f470021a | 338 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
1da177e4 LT |
339 | |
340 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
341 | current->exit_signal = SIGCHLD; | |
342 | ||
e05606d3 | 343 | if (task_nice(current) < 0) |
1da177e4 LT |
344 | set_user_nice(current, 0); |
345 | /* cpus_allowed? */ | |
346 | /* rt_priority? */ | |
347 | /* signals? */ | |
348 | security_task_reparent_to_init(current); | |
349 | memcpy(current->signal->rlim, init_task.signal->rlim, | |
350 | sizeof(current->signal->rlim)); | |
351 | atomic_inc(&(INIT_USER->__count)); | |
352 | write_unlock_irq(&tasklist_lock); | |
353 | switch_uid(INIT_USER); | |
354 | } | |
355 | ||
8520d7c7 | 356 | void __set_special_pids(struct pid *pid) |
1da177e4 | 357 | { |
e19f247a | 358 | struct task_struct *curr = current->group_leader; |
8520d7c7 | 359 | pid_t nr = pid_nr(pid); |
1da177e4 | 360 | |
8520d7c7 | 361 | if (task_session(curr) != pid) { |
7d8da096 | 362 | change_pid(curr, PIDTYPE_SID, pid); |
8520d7c7 | 363 | set_task_session(curr, nr); |
1da177e4 | 364 | } |
8520d7c7 | 365 | if (task_pgrp(curr) != pid) { |
7d8da096 | 366 | change_pid(curr, PIDTYPE_PGID, pid); |
8520d7c7 | 367 | set_task_pgrp(curr, nr); |
1da177e4 LT |
368 | } |
369 | } | |
370 | ||
8520d7c7 | 371 | static void set_special_pids(struct pid *pid) |
1da177e4 LT |
372 | { |
373 | write_lock_irq(&tasklist_lock); | |
8520d7c7 | 374 | __set_special_pids(pid); |
1da177e4 LT |
375 | write_unlock_irq(&tasklist_lock); |
376 | } | |
377 | ||
378 | /* | |
379 | * Let kernel threads use this to say that they | |
380 | * allow a certain signal (since daemonize() will | |
381 | * have disabled all of them by default). | |
382 | */ | |
383 | int allow_signal(int sig) | |
384 | { | |
7ed20e1a | 385 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
386 | return -EINVAL; |
387 | ||
388 | spin_lock_irq(¤t->sighand->siglock); | |
389 | sigdelset(¤t->blocked, sig); | |
390 | if (!current->mm) { | |
391 | /* Kernel threads handle their own signals. | |
392 | Let the signal code know it'll be handled, so | |
393 | that they don't get converted to SIGKILL or | |
394 | just silently dropped */ | |
395 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
396 | } | |
397 | recalc_sigpending(); | |
398 | spin_unlock_irq(¤t->sighand->siglock); | |
399 | return 0; | |
400 | } | |
401 | ||
402 | EXPORT_SYMBOL(allow_signal); | |
403 | ||
404 | int disallow_signal(int sig) | |
405 | { | |
7ed20e1a | 406 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
407 | return -EINVAL; |
408 | ||
409 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 410 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
411 | recalc_sigpending(); |
412 | spin_unlock_irq(¤t->sighand->siglock); | |
413 | return 0; | |
414 | } | |
415 | ||
416 | EXPORT_SYMBOL(disallow_signal); | |
417 | ||
418 | /* | |
419 | * Put all the gunge required to become a kernel thread without | |
420 | * attached user resources in one place where it belongs. | |
421 | */ | |
422 | ||
423 | void daemonize(const char *name, ...) | |
424 | { | |
425 | va_list args; | |
426 | struct fs_struct *fs; | |
427 | sigset_t blocked; | |
428 | ||
429 | va_start(args, name); | |
430 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
431 | va_end(args); | |
432 | ||
433 | /* | |
434 | * If we were started as result of loading a module, close all of the | |
435 | * user space pages. We don't need them, and if we didn't close them | |
436 | * they would be locked into memory. | |
437 | */ | |
438 | exit_mm(current); | |
83144186 RW |
439 | /* |
440 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
441 | * or suspend transition begins right now. | |
442 | */ | |
7b34e428 | 443 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
1da177e4 | 444 | |
8520d7c7 ON |
445 | if (current->nsproxy != &init_nsproxy) { |
446 | get_nsproxy(&init_nsproxy); | |
447 | switch_task_namespaces(current, &init_nsproxy); | |
448 | } | |
297bd42b | 449 | set_special_pids(&init_struct_pid); |
24ec839c | 450 | proc_clear_tty(current); |
1da177e4 LT |
451 | |
452 | /* Block and flush all signals */ | |
453 | sigfillset(&blocked); | |
454 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
455 | flush_signals(current); | |
456 | ||
457 | /* Become as one with the init task */ | |
458 | ||
459 | exit_fs(current); /* current->fs->count--; */ | |
460 | fs = init_task.fs; | |
461 | current->fs = fs; | |
462 | atomic_inc(&fs->count); | |
ab516013 | 463 | |
d4c5e41f | 464 | exit_files(current); |
1da177e4 LT |
465 | current->files = init_task.files; |
466 | atomic_inc(¤t->files->count); | |
467 | ||
49d769d5 | 468 | reparent_to_kthreadd(); |
1da177e4 LT |
469 | } |
470 | ||
471 | EXPORT_SYMBOL(daemonize); | |
472 | ||
858119e1 | 473 | static void close_files(struct files_struct * files) |
1da177e4 LT |
474 | { |
475 | int i, j; | |
badf1662 | 476 | struct fdtable *fdt; |
1da177e4 LT |
477 | |
478 | j = 0; | |
4fb3a538 DS |
479 | |
480 | /* | |
481 | * It is safe to dereference the fd table without RCU or | |
482 | * ->file_lock because this is the last reference to the | |
483 | * files structure. | |
484 | */ | |
badf1662 | 485 | fdt = files_fdtable(files); |
1da177e4 LT |
486 | for (;;) { |
487 | unsigned long set; | |
488 | i = j * __NFDBITS; | |
bbea9f69 | 489 | if (i >= fdt->max_fds) |
1da177e4 | 490 | break; |
badf1662 | 491 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
492 | while (set) { |
493 | if (set & 1) { | |
badf1662 | 494 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 495 | if (file) { |
1da177e4 | 496 | filp_close(file, files); |
944be0b2 IM |
497 | cond_resched(); |
498 | } | |
1da177e4 LT |
499 | } |
500 | i++; | |
501 | set >>= 1; | |
502 | } | |
503 | } | |
504 | } | |
505 | ||
506 | struct files_struct *get_files_struct(struct task_struct *task) | |
507 | { | |
508 | struct files_struct *files; | |
509 | ||
510 | task_lock(task); | |
511 | files = task->files; | |
512 | if (files) | |
513 | atomic_inc(&files->count); | |
514 | task_unlock(task); | |
515 | ||
516 | return files; | |
517 | } | |
518 | ||
7ad5b3a5 | 519 | void put_files_struct(struct files_struct *files) |
1da177e4 | 520 | { |
badf1662 DS |
521 | struct fdtable *fdt; |
522 | ||
1da177e4 LT |
523 | if (atomic_dec_and_test(&files->count)) { |
524 | close_files(files); | |
525 | /* | |
526 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
527 | * If the fdtable was embedded, pass files for freeing |
528 | * at the end of the RCU grace period. Otherwise, | |
529 | * you can free files immediately. | |
1da177e4 | 530 | */ |
badf1662 | 531 | fdt = files_fdtable(files); |
4fd45812 | 532 | if (fdt != &files->fdtab) |
ab2af1f5 | 533 | kmem_cache_free(files_cachep, files); |
01b2d93c | 534 | free_fdtable(fdt); |
1da177e4 LT |
535 | } |
536 | } | |
537 | ||
3b125388 | 538 | void reset_files_struct(struct files_struct *files) |
3b9b8ab6 | 539 | { |
3b125388 | 540 | struct task_struct *tsk = current; |
3b9b8ab6 KK |
541 | struct files_struct *old; |
542 | ||
543 | old = tsk->files; | |
544 | task_lock(tsk); | |
545 | tsk->files = files; | |
546 | task_unlock(tsk); | |
547 | put_files_struct(old); | |
548 | } | |
3b9b8ab6 | 549 | |
1ec7f1dd | 550 | void exit_files(struct task_struct *tsk) |
1da177e4 LT |
551 | { |
552 | struct files_struct * files = tsk->files; | |
553 | ||
554 | if (files) { | |
555 | task_lock(tsk); | |
556 | tsk->files = NULL; | |
557 | task_unlock(tsk); | |
558 | put_files_struct(files); | |
559 | } | |
560 | } | |
561 | ||
1ec7f1dd | 562 | void put_fs_struct(struct fs_struct *fs) |
1da177e4 LT |
563 | { |
564 | /* No need to hold fs->lock if we are killing it */ | |
565 | if (atomic_dec_and_test(&fs->count)) { | |
6ac08c39 JB |
566 | path_put(&fs->root); |
567 | path_put(&fs->pwd); | |
568 | if (fs->altroot.dentry) | |
569 | path_put(&fs->altroot); | |
1da177e4 LT |
570 | kmem_cache_free(fs_cachep, fs); |
571 | } | |
572 | } | |
573 | ||
1ec7f1dd | 574 | void exit_fs(struct task_struct *tsk) |
1da177e4 LT |
575 | { |
576 | struct fs_struct * fs = tsk->fs; | |
577 | ||
578 | if (fs) { | |
579 | task_lock(tsk); | |
580 | tsk->fs = NULL; | |
581 | task_unlock(tsk); | |
1ec7f1dd | 582 | put_fs_struct(fs); |
1da177e4 LT |
583 | } |
584 | } | |
585 | ||
1da177e4 LT |
586 | EXPORT_SYMBOL_GPL(exit_fs); |
587 | ||
cf475ad2 BS |
588 | #ifdef CONFIG_MM_OWNER |
589 | /* | |
590 | * Task p is exiting and it owned mm, lets find a new owner for it | |
591 | */ | |
592 | static inline int | |
593 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) | |
594 | { | |
595 | /* | |
596 | * If there are other users of the mm and the owner (us) is exiting | |
597 | * we need to find a new owner to take on the responsibility. | |
598 | */ | |
599 | if (!mm) | |
600 | return 0; | |
601 | if (atomic_read(&mm->mm_users) <= 1) | |
602 | return 0; | |
603 | if (mm->owner != p) | |
604 | return 0; | |
605 | return 1; | |
606 | } | |
607 | ||
608 | void mm_update_next_owner(struct mm_struct *mm) | |
609 | { | |
610 | struct task_struct *c, *g, *p = current; | |
611 | ||
612 | retry: | |
613 | if (!mm_need_new_owner(mm, p)) | |
614 | return; | |
615 | ||
616 | read_lock(&tasklist_lock); | |
617 | /* | |
618 | * Search in the children | |
619 | */ | |
620 | list_for_each_entry(c, &p->children, sibling) { | |
621 | if (c->mm == mm) | |
622 | goto assign_new_owner; | |
623 | } | |
624 | ||
625 | /* | |
626 | * Search in the siblings | |
627 | */ | |
628 | list_for_each_entry(c, &p->parent->children, sibling) { | |
629 | if (c->mm == mm) | |
630 | goto assign_new_owner; | |
631 | } | |
632 | ||
633 | /* | |
634 | * Search through everything else. We should not get | |
635 | * here often | |
636 | */ | |
637 | do_each_thread(g, c) { | |
638 | if (c->mm == mm) | |
639 | goto assign_new_owner; | |
640 | } while_each_thread(g, c); | |
641 | ||
642 | read_unlock(&tasklist_lock); | |
643 | return; | |
644 | ||
645 | assign_new_owner: | |
646 | BUG_ON(c == p); | |
647 | get_task_struct(c); | |
648 | /* | |
649 | * The task_lock protects c->mm from changing. | |
650 | * We always want mm->owner->mm == mm | |
651 | */ | |
652 | task_lock(c); | |
653 | /* | |
654 | * Delay read_unlock() till we have the task_lock() | |
655 | * to ensure that c does not slip away underneath us | |
656 | */ | |
657 | read_unlock(&tasklist_lock); | |
658 | if (c->mm != mm) { | |
659 | task_unlock(c); | |
660 | put_task_struct(c); | |
661 | goto retry; | |
662 | } | |
663 | cgroup_mm_owner_callbacks(mm->owner, c); | |
664 | mm->owner = c; | |
665 | task_unlock(c); | |
666 | put_task_struct(c); | |
667 | } | |
668 | #endif /* CONFIG_MM_OWNER */ | |
669 | ||
1da177e4 LT |
670 | /* |
671 | * Turn us into a lazy TLB process if we | |
672 | * aren't already.. | |
673 | */ | |
408b664a | 674 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
675 | { |
676 | struct mm_struct *mm = tsk->mm; | |
b564daf8 | 677 | struct core_state *core_state; |
1da177e4 LT |
678 | |
679 | mm_release(tsk, mm); | |
680 | if (!mm) | |
681 | return; | |
682 | /* | |
683 | * Serialize with any possible pending coredump. | |
999d9fc1 | 684 | * We must hold mmap_sem around checking core_state |
1da177e4 | 685 | * and clearing tsk->mm. The core-inducing thread |
999d9fc1 | 686 | * will increment ->nr_threads for each thread in the |
1da177e4 LT |
687 | * group with ->mm != NULL. |
688 | */ | |
689 | down_read(&mm->mmap_sem); | |
b564daf8 ON |
690 | core_state = mm->core_state; |
691 | if (core_state) { | |
692 | struct core_thread self; | |
1da177e4 | 693 | up_read(&mm->mmap_sem); |
c5f1cc8c | 694 | |
b564daf8 ON |
695 | self.task = tsk; |
696 | self.next = xchg(&core_state->dumper.next, &self); | |
697 | /* | |
698 | * Implies mb(), the result of xchg() must be visible | |
699 | * to core_state->dumper. | |
700 | */ | |
701 | if (atomic_dec_and_test(&core_state->nr_threads)) | |
702 | complete(&core_state->startup); | |
1da177e4 | 703 | |
a94e2d40 ON |
704 | for (;;) { |
705 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); | |
706 | if (!self.task) /* see coredump_finish() */ | |
707 | break; | |
708 | schedule(); | |
709 | } | |
710 | __set_task_state(tsk, TASK_RUNNING); | |
1da177e4 LT |
711 | down_read(&mm->mmap_sem); |
712 | } | |
713 | atomic_inc(&mm->mm_count); | |
125e1874 | 714 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
715 | /* more a memory barrier than a real lock */ |
716 | task_lock(tsk); | |
717 | tsk->mm = NULL; | |
718 | up_read(&mm->mmap_sem); | |
719 | enter_lazy_tlb(mm, current); | |
0c1eecfb RW |
720 | /* We don't want this task to be frozen prematurely */ |
721 | clear_freeze_flag(tsk); | |
1da177e4 | 722 | task_unlock(tsk); |
cf475ad2 | 723 | mm_update_next_owner(mm); |
1da177e4 LT |
724 | mmput(mm); |
725 | } | |
726 | ||
666f164f RM |
727 | /* |
728 | * Return nonzero if @parent's children should reap themselves. | |
729 | * | |
730 | * Called with write_lock_irq(&tasklist_lock) held. | |
731 | */ | |
732 | static int ignoring_children(struct task_struct *parent) | |
733 | { | |
734 | int ret; | |
735 | struct sighand_struct *psig = parent->sighand; | |
736 | unsigned long flags; | |
737 | spin_lock_irqsave(&psig->siglock, flags); | |
738 | ret = (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN || | |
739 | (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT)); | |
740 | spin_unlock_irqrestore(&psig->siglock, flags); | |
741 | return ret; | |
742 | } | |
743 | ||
f470021a RM |
744 | /* |
745 | * Detach all tasks we were using ptrace on. | |
746 | * Any that need to be release_task'd are put on the @dead list. | |
747 | * | |
748 | * Called with write_lock(&tasklist_lock) held. | |
749 | */ | |
750 | static void ptrace_exit(struct task_struct *parent, struct list_head *dead) | |
1da177e4 | 751 | { |
f470021a | 752 | struct task_struct *p, *n; |
666f164f | 753 | int ign = -1; |
241ceee0 | 754 | |
f470021a RM |
755 | list_for_each_entry_safe(p, n, &parent->ptraced, ptrace_entry) { |
756 | __ptrace_unlink(p); | |
757 | ||
758 | if (p->exit_state != EXIT_ZOMBIE) | |
759 | continue; | |
760 | ||
761 | /* | |
762 | * If it's a zombie, our attachedness prevented normal | |
763 | * parent notification or self-reaping. Do notification | |
764 | * now if it would have happened earlier. If it should | |
765 | * reap itself, add it to the @dead list. We can't call | |
766 | * release_task() here because we already hold tasklist_lock. | |
767 | * | |
768 | * If it's our own child, there is no notification to do. | |
666f164f RM |
769 | * But if our normal children self-reap, then this child |
770 | * was prevented by ptrace and we must reap it now. | |
1da177e4 | 771 | */ |
f470021a RM |
772 | if (!task_detached(p) && thread_group_empty(p)) { |
773 | if (!same_thread_group(p->real_parent, parent)) | |
774 | do_notify_parent(p, p->exit_signal); | |
666f164f RM |
775 | else { |
776 | if (ign < 0) | |
777 | ign = ignoring_children(parent); | |
778 | if (ign) | |
779 | p->exit_signal = -1; | |
780 | } | |
f470021a | 781 | } |
1da177e4 | 782 | |
f470021a | 783 | if (task_detached(p)) { |
1da177e4 | 784 | /* |
f470021a | 785 | * Mark it as in the process of being reaped. |
1da177e4 | 786 | */ |
f470021a RM |
787 | p->exit_state = EXIT_DEAD; |
788 | list_add(&p->ptrace_entry, dead); | |
1da177e4 LT |
789 | } |
790 | } | |
f470021a RM |
791 | } |
792 | ||
793 | /* | |
794 | * Finish up exit-time ptrace cleanup. | |
795 | * | |
796 | * Called without locks. | |
797 | */ | |
798 | static void ptrace_exit_finish(struct task_struct *parent, | |
799 | struct list_head *dead) | |
800 | { | |
801 | struct task_struct *p, *n; | |
802 | ||
803 | BUG_ON(!list_empty(&parent->ptraced)); | |
804 | ||
805 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | |
806 | list_del_init(&p->ptrace_entry); | |
807 | release_task(p); | |
808 | } | |
809 | } | |
810 | ||
811 | static void reparent_thread(struct task_struct *p, struct task_struct *father) | |
812 | { | |
813 | if (p->pdeath_signal) | |
814 | /* We already hold the tasklist_lock here. */ | |
815 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | |
816 | ||
817 | list_move_tail(&p->sibling, &p->real_parent->children); | |
1da177e4 | 818 | |
b2b2cbc4 EB |
819 | /* If this is a threaded reparent there is no need to |
820 | * notify anyone anything has happened. | |
821 | */ | |
376e1d25 | 822 | if (same_thread_group(p->real_parent, father)) |
b2b2cbc4 EB |
823 | return; |
824 | ||
825 | /* We don't want people slaying init. */ | |
d839fd4d | 826 | if (!task_detached(p)) |
b2b2cbc4 | 827 | p->exit_signal = SIGCHLD; |
b2b2cbc4 EB |
828 | |
829 | /* If we'd notified the old parent about this child's death, | |
830 | * also notify the new parent. | |
831 | */ | |
f470021a RM |
832 | if (!ptrace_reparented(p) && |
833 | p->exit_state == EXIT_ZOMBIE && | |
d839fd4d | 834 | !task_detached(p) && thread_group_empty(p)) |
b2b2cbc4 EB |
835 | do_notify_parent(p, p->exit_signal); |
836 | ||
f49ee505 | 837 | kill_orphaned_pgrp(p, father); |
1da177e4 LT |
838 | } |
839 | ||
840 | /* | |
841 | * When we die, we re-parent all our children. | |
842 | * Try to give them to another thread in our thread | |
843 | * group, and if no such member exists, give it to | |
84d73786 SB |
844 | * the child reaper process (ie "init") in our pid |
845 | * space. | |
1da177e4 | 846 | */ |
762a24be | 847 | static void forget_original_parent(struct task_struct *father) |
1da177e4 | 848 | { |
03ff1797 | 849 | struct task_struct *p, *n, *reaper = father; |
f470021a | 850 | LIST_HEAD(ptrace_dead); |
762a24be ON |
851 | |
852 | write_lock_irq(&tasklist_lock); | |
1da177e4 | 853 | |
f470021a RM |
854 | /* |
855 | * First clean up ptrace if we were using it. | |
856 | */ | |
857 | ptrace_exit(father, &ptrace_dead); | |
858 | ||
1da177e4 LT |
859 | do { |
860 | reaper = next_thread(reaper); | |
861 | if (reaper == father) { | |
88f21d81 | 862 | reaper = task_child_reaper(father); |
1da177e4 LT |
863 | break; |
864 | } | |
762a24be | 865 | } while (reaper->flags & PF_EXITING); |
1da177e4 | 866 | |
03ff1797 | 867 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
84eb646b | 868 | p->real_parent = reaper; |
f470021a RM |
869 | if (p->parent == father) { |
870 | BUG_ON(p->ptrace); | |
871 | p->parent = p->real_parent; | |
872 | } | |
873 | reparent_thread(p, father); | |
1da177e4 | 874 | } |
762a24be ON |
875 | |
876 | write_unlock_irq(&tasklist_lock); | |
877 | BUG_ON(!list_empty(&father->children)); | |
762a24be | 878 | |
f470021a | 879 | ptrace_exit_finish(father, &ptrace_dead); |
1da177e4 LT |
880 | } |
881 | ||
882 | /* | |
883 | * Send signals to all our closest relatives so that they know | |
884 | * to properly mourn us.. | |
885 | */ | |
821c7de7 | 886 | static void exit_notify(struct task_struct *tsk, int group_dead) |
1da177e4 | 887 | { |
2b2a1ff6 RM |
888 | int signal; |
889 | void *cookie; | |
1da177e4 | 890 | |
1da177e4 LT |
891 | /* |
892 | * This does two things: | |
893 | * | |
894 | * A. Make init inherit all the child processes | |
895 | * B. Check to see if any process groups have become orphaned | |
896 | * as a result of our exiting, and if they have any stopped | |
897 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
898 | */ | |
762a24be | 899 | forget_original_parent(tsk); |
2e4a7072 | 900 | exit_task_namespaces(tsk); |
1da177e4 | 901 | |
762a24be | 902 | write_lock_irq(&tasklist_lock); |
821c7de7 ON |
903 | if (group_dead) |
904 | kill_orphaned_pgrp(tsk->group_leader, NULL); | |
1da177e4 | 905 | |
24728448 | 906 | /* Let father know we died |
1da177e4 LT |
907 | * |
908 | * Thread signals are configurable, but you aren't going to use | |
d4c5e41f | 909 | * that to send signals to arbitary processes. |
1da177e4 LT |
910 | * That stops right now. |
911 | * | |
912 | * If the parent exec id doesn't match the exec id we saved | |
913 | * when we started then we know the parent has changed security | |
914 | * domain. | |
915 | * | |
916 | * If our self_exec id doesn't match our parent_exec_id then | |
917 | * we have changed execution domain as these two values started | |
918 | * the same after a fork. | |
1da177e4 | 919 | */ |
d839fd4d | 920 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
f49ee505 | 921 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
d839fd4d ON |
922 | tsk->self_exec_id != tsk->parent_exec_id) && |
923 | !capable(CAP_KILL)) | |
1da177e4 LT |
924 | tsk->exit_signal = SIGCHLD; |
925 | ||
2b2a1ff6 RM |
926 | signal = tracehook_notify_death(tsk, &cookie, group_dead); |
927 | if (signal > 0) | |
928 | signal = do_notify_parent(tsk, signal); | |
1da177e4 | 929 | |
2b2a1ff6 | 930 | tsk->exit_state = signal < 0 ? EXIT_DEAD : EXIT_ZOMBIE; |
1da177e4 | 931 | |
2800d8d1 | 932 | /* mt-exec, de_thread() is waiting for us */ |
6db840fa ON |
933 | if (thread_group_leader(tsk) && |
934 | tsk->signal->notify_count < 0 && | |
935 | tsk->signal->group_exit_task) | |
936 | wake_up_process(tsk->signal->group_exit_task); | |
937 | ||
1da177e4 LT |
938 | write_unlock_irq(&tasklist_lock); |
939 | ||
2b2a1ff6 RM |
940 | tracehook_report_death(tsk, signal, cookie, group_dead); |
941 | ||
1da177e4 | 942 | /* If the process is dead, release it - nobody will wait for it */ |
2b2a1ff6 | 943 | if (signal < 0) |
1da177e4 | 944 | release_task(tsk); |
1da177e4 LT |
945 | } |
946 | ||
e18eecb8 JD |
947 | #ifdef CONFIG_DEBUG_STACK_USAGE |
948 | static void check_stack_usage(void) | |
949 | { | |
950 | static DEFINE_SPINLOCK(low_water_lock); | |
951 | static int lowest_to_date = THREAD_SIZE; | |
952 | unsigned long *n = end_of_stack(current); | |
953 | unsigned long free; | |
954 | ||
955 | while (*n == 0) | |
956 | n++; | |
957 | free = (unsigned long)n - (unsigned long)end_of_stack(current); | |
958 | ||
959 | if (free >= lowest_to_date) | |
960 | return; | |
961 | ||
962 | spin_lock(&low_water_lock); | |
963 | if (free < lowest_to_date) { | |
964 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
965 | "left\n", | |
966 | current->comm, free); | |
967 | lowest_to_date = free; | |
968 | } | |
969 | spin_unlock(&low_water_lock); | |
970 | } | |
971 | #else | |
972 | static inline void check_stack_usage(void) {} | |
973 | #endif | |
974 | ||
84eb646b ON |
975 | static inline void exit_child_reaper(struct task_struct *tsk) |
976 | { | |
88f21d81 | 977 | if (likely(tsk->group_leader != task_child_reaper(tsk))) |
84eb646b ON |
978 | return; |
979 | ||
3eb07c8c SB |
980 | if (tsk->nsproxy->pid_ns == &init_pid_ns) |
981 | panic("Attempted to kill init!"); | |
982 | ||
983 | /* | |
984 | * @tsk is the last thread in the 'cgroup-init' and is exiting. | |
985 | * Terminate all remaining processes in the namespace and reap them | |
986 | * before exiting @tsk. | |
987 | * | |
988 | * Note that @tsk (last thread of cgroup-init) may not necessarily | |
989 | * be the child-reaper (i.e main thread of cgroup-init) of the | |
990 | * namespace i.e the child_reaper may have already exited. | |
991 | * | |
992 | * Even after a child_reaper exits, we let it inherit orphaned children, | |
993 | * because, pid_ns->child_reaper remains valid as long as there is | |
994 | * at least one living sub-thread in the cgroup init. | |
995 | ||
996 | * This living sub-thread of the cgroup-init will be notified when | |
997 | * a child inherited by the 'child-reaper' exits (do_notify_parent() | |
998 | * uses __group_send_sig_info()). Further, when reaping child processes, | |
999 | * do_wait() iterates over children of all living sub threads. | |
1000 | ||
1001 | * i.e even though 'child_reaper' thread is listed as the parent of the | |
1002 | * orphaned children, any living sub-thread in the cgroup-init can | |
1003 | * perform the role of the child_reaper. | |
1004 | */ | |
1005 | zap_pid_ns_processes(tsk->nsproxy->pid_ns); | |
84eb646b ON |
1006 | } |
1007 | ||
7ad5b3a5 | 1008 | NORET_TYPE void do_exit(long code) |
1da177e4 LT |
1009 | { |
1010 | struct task_struct *tsk = current; | |
1011 | int group_dead; | |
1012 | ||
1013 | profile_task_exit(tsk); | |
1014 | ||
22e2c507 JA |
1015 | WARN_ON(atomic_read(&tsk->fs_excl)); |
1016 | ||
1da177e4 LT |
1017 | if (unlikely(in_interrupt())) |
1018 | panic("Aiee, killing interrupt handler!"); | |
1019 | if (unlikely(!tsk->pid)) | |
1020 | panic("Attempted to kill the idle task!"); | |
1da177e4 | 1021 | |
30199f5a | 1022 | tracehook_report_exit(&code); |
1da177e4 | 1023 | |
df164db5 AN |
1024 | /* |
1025 | * We're taking recursive faults here in do_exit. Safest is to just | |
1026 | * leave this task alone and wait for reboot. | |
1027 | */ | |
1028 | if (unlikely(tsk->flags & PF_EXITING)) { | |
1029 | printk(KERN_ALERT | |
1030 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
1031 | /* |
1032 | * We can do this unlocked here. The futex code uses | |
1033 | * this flag just to verify whether the pi state | |
1034 | * cleanup has been done or not. In the worst case it | |
1035 | * loops once more. We pretend that the cleanup was | |
1036 | * done as there is no way to return. Either the | |
1037 | * OWNER_DIED bit is set by now or we push the blocked | |
1038 | * task into the wait for ever nirwana as well. | |
1039 | */ | |
1040 | tsk->flags |= PF_EXITPIDONE; | |
afc847b7 AV |
1041 | if (tsk->io_context) |
1042 | exit_io_context(); | |
df164db5 AN |
1043 | set_current_state(TASK_UNINTERRUPTIBLE); |
1044 | schedule(); | |
1045 | } | |
1046 | ||
d12619b5 | 1047 | exit_signals(tsk); /* sets PF_EXITING */ |
778e9a9c AK |
1048 | /* |
1049 | * tsk->flags are checked in the futex code to protect against | |
1050 | * an exiting task cleaning up the robust pi futexes. | |
1051 | */ | |
d2ee7198 ON |
1052 | smp_mb(); |
1053 | spin_unlock_wait(&tsk->pi_lock); | |
1da177e4 | 1054 | |
1da177e4 LT |
1055 | if (unlikely(in_atomic())) |
1056 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
ba25f9dc | 1057 | current->comm, task_pid_nr(current), |
1da177e4 LT |
1058 | preempt_count()); |
1059 | ||
1060 | acct_update_integrals(tsk); | |
365e9c87 HD |
1061 | if (tsk->mm) { |
1062 | update_hiwater_rss(tsk->mm); | |
1063 | update_hiwater_vm(tsk->mm); | |
1064 | } | |
1da177e4 | 1065 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 1066 | if (group_dead) { |
84eb646b | 1067 | exit_child_reaper(tsk); |
778e9a9c | 1068 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 1069 | exit_itimers(tsk->signal); |
c3068951 | 1070 | } |
f6ec29a4 | 1071 | acct_collect(code, group_dead); |
42b2dd0a | 1072 | #ifdef CONFIG_FUTEX |
0771dfef IM |
1073 | if (unlikely(tsk->robust_list)) |
1074 | exit_robust_list(tsk); | |
42b2dd0a | 1075 | #ifdef CONFIG_COMPAT |
34f192c6 IM |
1076 | if (unlikely(tsk->compat_robust_list)) |
1077 | compat_exit_robust_list(tsk); | |
42b2dd0a | 1078 | #endif |
34f192c6 | 1079 | #endif |
522ed776 MT |
1080 | if (group_dead) |
1081 | tty_audit_exit(); | |
fa84cb93 AV |
1082 | if (unlikely(tsk->audit_context)) |
1083 | audit_free(tsk); | |
115085ea | 1084 | |
f2ab6d88 | 1085 | tsk->exit_code = code; |
115085ea | 1086 | taskstats_exit(tsk, group_dead); |
c757249a | 1087 | |
1da177e4 LT |
1088 | exit_mm(tsk); |
1089 | ||
0e464814 | 1090 | if (group_dead) |
f6ec29a4 | 1091 | acct_process(); |
1da177e4 | 1092 | exit_sem(tsk); |
1ec7f1dd AV |
1093 | exit_files(tsk); |
1094 | exit_fs(tsk); | |
e18eecb8 | 1095 | check_stack_usage(); |
1da177e4 | 1096 | exit_thread(); |
b4f48b63 | 1097 | cgroup_exit(tsk, 1); |
1da177e4 LT |
1098 | exit_keys(tsk); |
1099 | ||
1100 | if (group_dead && tsk->signal->leader) | |
1101 | disassociate_ctty(1); | |
1102 | ||
a1261f54 | 1103 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 LT |
1104 | if (tsk->binfmt) |
1105 | module_put(tsk->binfmt->module); | |
1106 | ||
9f46080c | 1107 | proc_exit_connector(tsk); |
821c7de7 | 1108 | exit_notify(tsk, group_dead); |
1da177e4 | 1109 | #ifdef CONFIG_NUMA |
f0be3d32 | 1110 | mpol_put(tsk->mempolicy); |
1da177e4 LT |
1111 | tsk->mempolicy = NULL; |
1112 | #endif | |
42b2dd0a | 1113 | #ifdef CONFIG_FUTEX |
c87e2837 IM |
1114 | /* |
1115 | * This must happen late, after the PID is not | |
1116 | * hashed anymore: | |
1117 | */ | |
1118 | if (unlikely(!list_empty(&tsk->pi_state_list))) | |
1119 | exit_pi_state_list(tsk); | |
1120 | if (unlikely(current->pi_state_cache)) | |
1121 | kfree(current->pi_state_cache); | |
42b2dd0a | 1122 | #endif |
de5097c2 | 1123 | /* |
9a11b49a | 1124 | * Make sure we are holding no locks: |
de5097c2 | 1125 | */ |
9a11b49a | 1126 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1127 | /* |
1128 | * We can do this unlocked here. The futex code uses this flag | |
1129 | * just to verify whether the pi state cleanup has been done | |
1130 | * or not. In the worst case it loops once more. | |
1131 | */ | |
1132 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1133 | |
afc847b7 AV |
1134 | if (tsk->io_context) |
1135 | exit_io_context(); | |
1136 | ||
b92ce558 JA |
1137 | if (tsk->splice_pipe) |
1138 | __free_pipe_info(tsk->splice_pipe); | |
1139 | ||
7407251a | 1140 | preempt_disable(); |
55a101f8 | 1141 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1142 | tsk->state = TASK_DEAD; |
7407251a | 1143 | |
1da177e4 LT |
1144 | schedule(); |
1145 | BUG(); | |
1146 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1147 | for (;;) |
1148 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1149 | } |
1150 | ||
012914da RA |
1151 | EXPORT_SYMBOL_GPL(do_exit); |
1152 | ||
1da177e4 LT |
1153 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1154 | { | |
1155 | if (comp) | |
1156 | complete(comp); | |
55a101f8 | 1157 | |
1da177e4 LT |
1158 | do_exit(code); |
1159 | } | |
1160 | ||
1161 | EXPORT_SYMBOL(complete_and_exit); | |
1162 | ||
1163 | asmlinkage long sys_exit(int error_code) | |
1164 | { | |
1165 | do_exit((error_code&0xff)<<8); | |
1166 | } | |
1167 | ||
1da177e4 LT |
1168 | /* |
1169 | * Take down every thread in the group. This is called by fatal signals | |
1170 | * as well as by sys_exit_group (below). | |
1171 | */ | |
1172 | NORET_TYPE void | |
1173 | do_group_exit(int exit_code) | |
1174 | { | |
bfc4b089 ON |
1175 | struct signal_struct *sig = current->signal; |
1176 | ||
1da177e4 LT |
1177 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
1178 | ||
bfc4b089 ON |
1179 | if (signal_group_exit(sig)) |
1180 | exit_code = sig->group_exit_code; | |
1da177e4 | 1181 | else if (!thread_group_empty(current)) { |
1da177e4 | 1182 | struct sighand_struct *const sighand = current->sighand; |
1da177e4 | 1183 | spin_lock_irq(&sighand->siglock); |
ed5d2cac | 1184 | if (signal_group_exit(sig)) |
1da177e4 LT |
1185 | /* Another thread got here before we took the lock. */ |
1186 | exit_code = sig->group_exit_code; | |
1187 | else { | |
1da177e4 | 1188 | sig->group_exit_code = exit_code; |
ed5d2cac | 1189 | sig->flags = SIGNAL_GROUP_EXIT; |
1da177e4 LT |
1190 | zap_other_threads(current); |
1191 | } | |
1192 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1193 | } |
1194 | ||
1195 | do_exit(exit_code); | |
1196 | /* NOTREACHED */ | |
1197 | } | |
1198 | ||
1199 | /* | |
1200 | * this kills every thread in the thread group. Note that any externally | |
1201 | * wait4()-ing process will get the correct exit code - even if this | |
1202 | * thread is not the thread group leader. | |
1203 | */ | |
1204 | asmlinkage void sys_exit_group(int error_code) | |
1205 | { | |
1206 | do_group_exit((error_code & 0xff) << 8); | |
1207 | } | |
1208 | ||
161550d7 EB |
1209 | static struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
1210 | { | |
1211 | struct pid *pid = NULL; | |
1212 | if (type == PIDTYPE_PID) | |
1213 | pid = task->pids[type].pid; | |
1214 | else if (type < PIDTYPE_MAX) | |
1215 | pid = task->group_leader->pids[type].pid; | |
1216 | return pid; | |
1217 | } | |
1218 | ||
1219 | static int eligible_child(enum pid_type type, struct pid *pid, int options, | |
1220 | struct task_struct *p) | |
1da177e4 | 1221 | { |
73243284 RM |
1222 | int err; |
1223 | ||
161550d7 EB |
1224 | if (type < PIDTYPE_MAX) { |
1225 | if (task_pid_type(p, type) != pid) | |
1da177e4 LT |
1226 | return 0; |
1227 | } | |
1228 | ||
1da177e4 LT |
1229 | /* Wait for all children (clone and not) if __WALL is set; |
1230 | * otherwise, wait for clone children *only* if __WCLONE is | |
1231 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1232 | * A "clone" child here is one that reports to its parent | |
1233 | * using a signal other than SIGCHLD.) */ | |
1234 | if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0)) | |
1235 | && !(options & __WALL)) | |
1236 | return 0; | |
1da177e4 | 1237 | |
73243284 | 1238 | err = security_task_wait(p); |
14dd0b81 RM |
1239 | if (err) |
1240 | return err; | |
1da177e4 | 1241 | |
14dd0b81 | 1242 | return 1; |
1da177e4 LT |
1243 | } |
1244 | ||
36c8b586 | 1245 | static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid, |
1da177e4 LT |
1246 | int why, int status, |
1247 | struct siginfo __user *infop, | |
1248 | struct rusage __user *rusagep) | |
1249 | { | |
1250 | int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0; | |
36c8b586 | 1251 | |
1da177e4 LT |
1252 | put_task_struct(p); |
1253 | if (!retval) | |
1254 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1255 | if (!retval) | |
1256 | retval = put_user(0, &infop->si_errno); | |
1257 | if (!retval) | |
1258 | retval = put_user((short)why, &infop->si_code); | |
1259 | if (!retval) | |
1260 | retval = put_user(pid, &infop->si_pid); | |
1261 | if (!retval) | |
1262 | retval = put_user(uid, &infop->si_uid); | |
1263 | if (!retval) | |
1264 | retval = put_user(status, &infop->si_status); | |
1265 | if (!retval) | |
1266 | retval = pid; | |
1267 | return retval; | |
1268 | } | |
1269 | ||
1270 | /* | |
1271 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1272 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1273 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1274 | * released the lock and the system call should return. | |
1275 | */ | |
98abed02 | 1276 | static int wait_task_zombie(struct task_struct *p, int options, |
1da177e4 LT |
1277 | struct siginfo __user *infop, |
1278 | int __user *stat_addr, struct rusage __user *ru) | |
1279 | { | |
1280 | unsigned long state; | |
2f4e6e2a | 1281 | int retval, status, traced; |
6c5f3e7b | 1282 | pid_t pid = task_pid_vnr(p); |
1da177e4 | 1283 | |
98abed02 RM |
1284 | if (!likely(options & WEXITED)) |
1285 | return 0; | |
1286 | ||
1287 | if (unlikely(options & WNOWAIT)) { | |
1da177e4 LT |
1288 | uid_t uid = p->uid; |
1289 | int exit_code = p->exit_code; | |
1290 | int why, status; | |
1291 | ||
1da177e4 LT |
1292 | get_task_struct(p); |
1293 | read_unlock(&tasklist_lock); | |
1294 | if ((exit_code & 0x7f) == 0) { | |
1295 | why = CLD_EXITED; | |
1296 | status = exit_code >> 8; | |
1297 | } else { | |
1298 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1299 | status = exit_code & 0x7f; | |
1300 | } | |
1301 | return wait_noreap_copyout(p, pid, uid, why, | |
1302 | status, infop, ru); | |
1303 | } | |
1304 | ||
1305 | /* | |
1306 | * Try to move the task's state to DEAD | |
1307 | * only one thread is allowed to do this: | |
1308 | */ | |
1309 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1310 | if (state != EXIT_ZOMBIE) { | |
1311 | BUG_ON(state != EXIT_DEAD); | |
1312 | return 0; | |
1313 | } | |
1da177e4 | 1314 | |
53b6f9fb | 1315 | traced = ptrace_reparented(p); |
2f4e6e2a ON |
1316 | |
1317 | if (likely(!traced)) { | |
3795e161 JJ |
1318 | struct signal_struct *psig; |
1319 | struct signal_struct *sig; | |
1320 | ||
1da177e4 LT |
1321 | /* |
1322 | * The resource counters for the group leader are in its | |
1323 | * own task_struct. Those for dead threads in the group | |
1324 | * are in its signal_struct, as are those for the child | |
1325 | * processes it has previously reaped. All these | |
1326 | * accumulate in the parent's signal_struct c* fields. | |
1327 | * | |
1328 | * We don't bother to take a lock here to protect these | |
1329 | * p->signal fields, because they are only touched by | |
1330 | * __exit_signal, which runs with tasklist_lock | |
1331 | * write-locked anyway, and so is excluded here. We do | |
1332 | * need to protect the access to p->parent->signal fields, | |
1333 | * as other threads in the parent group can be right | |
1334 | * here reaping other children at the same time. | |
1335 | */ | |
1336 | spin_lock_irq(&p->parent->sighand->siglock); | |
3795e161 JJ |
1337 | psig = p->parent->signal; |
1338 | sig = p->signal; | |
1339 | psig->cutime = | |
1340 | cputime_add(psig->cutime, | |
1da177e4 | 1341 | cputime_add(p->utime, |
3795e161 JJ |
1342 | cputime_add(sig->utime, |
1343 | sig->cutime))); | |
1344 | psig->cstime = | |
1345 | cputime_add(psig->cstime, | |
1da177e4 | 1346 | cputime_add(p->stime, |
3795e161 JJ |
1347 | cputime_add(sig->stime, |
1348 | sig->cstime))); | |
9ac52315 LV |
1349 | psig->cgtime = |
1350 | cputime_add(psig->cgtime, | |
1351 | cputime_add(p->gtime, | |
1352 | cputime_add(sig->gtime, | |
1353 | sig->cgtime))); | |
3795e161 JJ |
1354 | psig->cmin_flt += |
1355 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1356 | psig->cmaj_flt += | |
1357 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1358 | psig->cnvcsw += | |
1359 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1360 | psig->cnivcsw += | |
1361 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1362 | psig->cinblock += |
1363 | task_io_get_inblock(p) + | |
1364 | sig->inblock + sig->cinblock; | |
1365 | psig->coublock += | |
1366 | task_io_get_oublock(p) + | |
1367 | sig->oublock + sig->coublock; | |
297c5d92 AR |
1368 | #ifdef CONFIG_TASK_XACCT |
1369 | psig->rchar += p->rchar + sig->rchar; | |
1370 | psig->wchar += p->wchar + sig->wchar; | |
1371 | psig->syscr += p->syscr + sig->syscr; | |
1372 | psig->syscw += p->syscw + sig->syscw; | |
1373 | #endif /* CONFIG_TASK_XACCT */ | |
1374 | #ifdef CONFIG_TASK_IO_ACCOUNTING | |
1375 | psig->ioac.read_bytes += | |
1376 | p->ioac.read_bytes + sig->ioac.read_bytes; | |
1377 | psig->ioac.write_bytes += | |
1378 | p->ioac.write_bytes + sig->ioac.write_bytes; | |
1379 | psig->ioac.cancelled_write_bytes += | |
1380 | p->ioac.cancelled_write_bytes + | |
1381 | sig->ioac.cancelled_write_bytes; | |
1382 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
1da177e4 LT |
1383 | spin_unlock_irq(&p->parent->sighand->siglock); |
1384 | } | |
1385 | ||
1386 | /* | |
1387 | * Now we are sure this task is interesting, and no other | |
1388 | * thread can reap it because we set its state to EXIT_DEAD. | |
1389 | */ | |
1390 | read_unlock(&tasklist_lock); | |
1391 | ||
1392 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1393 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) | |
1394 | ? p->signal->group_exit_code : p->exit_code; | |
1395 | if (!retval && stat_addr) | |
1396 | retval = put_user(status, stat_addr); | |
1397 | if (!retval && infop) | |
1398 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1399 | if (!retval && infop) | |
1400 | retval = put_user(0, &infop->si_errno); | |
1401 | if (!retval && infop) { | |
1402 | int why; | |
1403 | ||
1404 | if ((status & 0x7f) == 0) { | |
1405 | why = CLD_EXITED; | |
1406 | status >>= 8; | |
1407 | } else { | |
1408 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1409 | status &= 0x7f; | |
1410 | } | |
1411 | retval = put_user((short)why, &infop->si_code); | |
1412 | if (!retval) | |
1413 | retval = put_user(status, &infop->si_status); | |
1414 | } | |
1415 | if (!retval && infop) | |
3a515e4a | 1416 | retval = put_user(pid, &infop->si_pid); |
1da177e4 LT |
1417 | if (!retval && infop) |
1418 | retval = put_user(p->uid, &infop->si_uid); | |
2f4e6e2a | 1419 | if (!retval) |
3a515e4a | 1420 | retval = pid; |
2f4e6e2a ON |
1421 | |
1422 | if (traced) { | |
1da177e4 | 1423 | write_lock_irq(&tasklist_lock); |
2f4e6e2a ON |
1424 | /* We dropped tasklist, ptracer could die and untrace */ |
1425 | ptrace_unlink(p); | |
1426 | /* | |
1427 | * If this is not a detached task, notify the parent. | |
1428 | * If it's still not detached after that, don't release | |
1429 | * it now. | |
1430 | */ | |
d839fd4d | 1431 | if (!task_detached(p)) { |
2f4e6e2a | 1432 | do_notify_parent(p, p->exit_signal); |
d839fd4d | 1433 | if (!task_detached(p)) { |
2f4e6e2a ON |
1434 | p->exit_state = EXIT_ZOMBIE; |
1435 | p = NULL; | |
1da177e4 LT |
1436 | } |
1437 | } | |
1438 | write_unlock_irq(&tasklist_lock); | |
1439 | } | |
1440 | if (p != NULL) | |
1441 | release_task(p); | |
2f4e6e2a | 1442 | |
1da177e4 LT |
1443 | return retval; |
1444 | } | |
1445 | ||
1446 | /* | |
1447 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1448 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1449 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1450 | * released the lock and the system call should return. | |
1451 | */ | |
f470021a | 1452 | static int wait_task_stopped(int ptrace, struct task_struct *p, |
98abed02 | 1453 | int options, struct siginfo __user *infop, |
1da177e4 LT |
1454 | int __user *stat_addr, struct rusage __user *ru) |
1455 | { | |
ee7c82da ON |
1456 | int retval, exit_code, why; |
1457 | uid_t uid = 0; /* unneeded, required by compiler */ | |
c8950783 | 1458 | pid_t pid; |
1da177e4 | 1459 | |
f470021a | 1460 | if (!(options & WUNTRACED)) |
98abed02 RM |
1461 | return 0; |
1462 | ||
ee7c82da ON |
1463 | exit_code = 0; |
1464 | spin_lock_irq(&p->sighand->siglock); | |
1465 | ||
1466 | if (unlikely(!task_is_stopped_or_traced(p))) | |
1467 | goto unlock_sig; | |
1468 | ||
f470021a | 1469 | if (!ptrace && p->signal->group_stop_count > 0) |
1da177e4 LT |
1470 | /* |
1471 | * A group stop is in progress and this is the group leader. | |
1472 | * We won't report until all threads have stopped. | |
1473 | */ | |
ee7c82da ON |
1474 | goto unlock_sig; |
1475 | ||
1476 | exit_code = p->exit_code; | |
1477 | if (!exit_code) | |
1478 | goto unlock_sig; | |
1479 | ||
98abed02 | 1480 | if (!unlikely(options & WNOWAIT)) |
ee7c82da ON |
1481 | p->exit_code = 0; |
1482 | ||
1483 | uid = p->uid; | |
1484 | unlock_sig: | |
1485 | spin_unlock_irq(&p->sighand->siglock); | |
1486 | if (!exit_code) | |
1da177e4 LT |
1487 | return 0; |
1488 | ||
1489 | /* | |
1490 | * Now we are pretty sure this task is interesting. | |
1491 | * Make sure it doesn't get reaped out from under us while we | |
1492 | * give up the lock and then examine it below. We don't want to | |
1493 | * keep holding onto the tasklist_lock while we call getrusage and | |
1494 | * possibly take page faults for user memory. | |
1495 | */ | |
1496 | get_task_struct(p); | |
6c5f3e7b | 1497 | pid = task_pid_vnr(p); |
f470021a | 1498 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
1da177e4 LT |
1499 | read_unlock(&tasklist_lock); |
1500 | ||
98abed02 | 1501 | if (unlikely(options & WNOWAIT)) |
1da177e4 | 1502 | return wait_noreap_copyout(p, pid, uid, |
e6ceb32a | 1503 | why, exit_code, |
1da177e4 | 1504 | infop, ru); |
1da177e4 LT |
1505 | |
1506 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1507 | if (!retval && stat_addr) | |
1508 | retval = put_user((exit_code << 8) | 0x7f, stat_addr); | |
1509 | if (!retval && infop) | |
1510 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1511 | if (!retval && infop) | |
1512 | retval = put_user(0, &infop->si_errno); | |
1513 | if (!retval && infop) | |
6efcae46 | 1514 | retval = put_user((short)why, &infop->si_code); |
1da177e4 LT |
1515 | if (!retval && infop) |
1516 | retval = put_user(exit_code, &infop->si_status); | |
1517 | if (!retval && infop) | |
c8950783 | 1518 | retval = put_user(pid, &infop->si_pid); |
1da177e4 | 1519 | if (!retval && infop) |
ee7c82da | 1520 | retval = put_user(uid, &infop->si_uid); |
1da177e4 | 1521 | if (!retval) |
c8950783 | 1522 | retval = pid; |
1da177e4 LT |
1523 | put_task_struct(p); |
1524 | ||
1525 | BUG_ON(!retval); | |
1526 | return retval; | |
1527 | } | |
1528 | ||
1529 | /* | |
1530 | * Handle do_wait work for one task in a live, non-stopped state. | |
1531 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1532 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1533 | * released the lock and the system call should return. | |
1534 | */ | |
98abed02 | 1535 | static int wait_task_continued(struct task_struct *p, int options, |
1da177e4 LT |
1536 | struct siginfo __user *infop, |
1537 | int __user *stat_addr, struct rusage __user *ru) | |
1538 | { | |
1539 | int retval; | |
1540 | pid_t pid; | |
1541 | uid_t uid; | |
1542 | ||
98abed02 RM |
1543 | if (!unlikely(options & WCONTINUED)) |
1544 | return 0; | |
1545 | ||
1da177e4 LT |
1546 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
1547 | return 0; | |
1548 | ||
1549 | spin_lock_irq(&p->sighand->siglock); | |
1550 | /* Re-check with the lock held. */ | |
1551 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1552 | spin_unlock_irq(&p->sighand->siglock); | |
1553 | return 0; | |
1554 | } | |
98abed02 | 1555 | if (!unlikely(options & WNOWAIT)) |
1da177e4 LT |
1556 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
1557 | spin_unlock_irq(&p->sighand->siglock); | |
1558 | ||
6c5f3e7b | 1559 | pid = task_pid_vnr(p); |
1da177e4 LT |
1560 | uid = p->uid; |
1561 | get_task_struct(p); | |
1562 | read_unlock(&tasklist_lock); | |
1563 | ||
1564 | if (!infop) { | |
1565 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1566 | put_task_struct(p); | |
1567 | if (!retval && stat_addr) | |
1568 | retval = put_user(0xffff, stat_addr); | |
1569 | if (!retval) | |
3a515e4a | 1570 | retval = pid; |
1da177e4 LT |
1571 | } else { |
1572 | retval = wait_noreap_copyout(p, pid, uid, | |
1573 | CLD_CONTINUED, SIGCONT, | |
1574 | infop, ru); | |
1575 | BUG_ON(retval == 0); | |
1576 | } | |
1577 | ||
1578 | return retval; | |
1579 | } | |
1580 | ||
98abed02 RM |
1581 | /* |
1582 | * Consider @p for a wait by @parent. | |
1583 | * | |
1584 | * -ECHILD should be in *@notask_error before the first call. | |
1585 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | |
1586 | * Returns zero if the search for a child should continue; | |
14dd0b81 RM |
1587 | * then *@notask_error is 0 if @p is an eligible child, |
1588 | * or another error from security_task_wait(), or still -ECHILD. | |
98abed02 | 1589 | */ |
f470021a | 1590 | static int wait_consider_task(struct task_struct *parent, int ptrace, |
98abed02 RM |
1591 | struct task_struct *p, int *notask_error, |
1592 | enum pid_type type, struct pid *pid, int options, | |
1593 | struct siginfo __user *infop, | |
1594 | int __user *stat_addr, struct rusage __user *ru) | |
1595 | { | |
1596 | int ret = eligible_child(type, pid, options, p); | |
14dd0b81 | 1597 | if (!ret) |
98abed02 RM |
1598 | return ret; |
1599 | ||
14dd0b81 RM |
1600 | if (unlikely(ret < 0)) { |
1601 | /* | |
1602 | * If we have not yet seen any eligible child, | |
1603 | * then let this error code replace -ECHILD. | |
1604 | * A permission error will give the user a clue | |
1605 | * to look for security policy problems, rather | |
1606 | * than for mysterious wait bugs. | |
1607 | */ | |
1608 | if (*notask_error) | |
1609 | *notask_error = ret; | |
1610 | } | |
1611 | ||
f470021a RM |
1612 | if (likely(!ptrace) && unlikely(p->ptrace)) { |
1613 | /* | |
1614 | * This child is hidden by ptrace. | |
1615 | * We aren't allowed to see it now, but eventually we will. | |
1616 | */ | |
1617 | *notask_error = 0; | |
1618 | return 0; | |
1619 | } | |
1620 | ||
98abed02 RM |
1621 | if (p->exit_state == EXIT_DEAD) |
1622 | return 0; | |
1623 | ||
1624 | /* | |
1625 | * We don't reap group leaders with subthreads. | |
1626 | */ | |
1627 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) | |
1628 | return wait_task_zombie(p, options, infop, stat_addr, ru); | |
1629 | ||
1630 | /* | |
1631 | * It's stopped or running now, so it might | |
1632 | * later continue, exit, or stop again. | |
1633 | */ | |
1634 | *notask_error = 0; | |
1635 | ||
1636 | if (task_is_stopped_or_traced(p)) | |
f470021a RM |
1637 | return wait_task_stopped(ptrace, p, options, |
1638 | infop, stat_addr, ru); | |
98abed02 RM |
1639 | |
1640 | return wait_task_continued(p, options, infop, stat_addr, ru); | |
1641 | } | |
1642 | ||
1643 | /* | |
1644 | * Do the work of do_wait() for one thread in the group, @tsk. | |
1645 | * | |
1646 | * -ECHILD should be in *@notask_error before the first call. | |
1647 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. | |
1648 | * Returns zero if the search for a child should continue; then | |
14dd0b81 RM |
1649 | * *@notask_error is 0 if there were any eligible children, |
1650 | * or another error from security_task_wait(), or still -ECHILD. | |
98abed02 RM |
1651 | */ |
1652 | static int do_wait_thread(struct task_struct *tsk, int *notask_error, | |
1653 | enum pid_type type, struct pid *pid, int options, | |
1654 | struct siginfo __user *infop, int __user *stat_addr, | |
1655 | struct rusage __user *ru) | |
1656 | { | |
1657 | struct task_struct *p; | |
1658 | ||
1659 | list_for_each_entry(p, &tsk->children, sibling) { | |
f470021a RM |
1660 | /* |
1661 | * Do not consider detached threads. | |
1662 | */ | |
1663 | if (!task_detached(p)) { | |
1664 | int ret = wait_consider_task(tsk, 0, p, notask_error, | |
1665 | type, pid, options, | |
1666 | infop, stat_addr, ru); | |
1667 | if (ret) | |
1668 | return ret; | |
1669 | } | |
98abed02 RM |
1670 | } |
1671 | ||
1672 | return 0; | |
1673 | } | |
1674 | ||
1675 | static int ptrace_do_wait(struct task_struct *tsk, int *notask_error, | |
1676 | enum pid_type type, struct pid *pid, int options, | |
1677 | struct siginfo __user *infop, int __user *stat_addr, | |
1678 | struct rusage __user *ru) | |
1679 | { | |
1680 | struct task_struct *p; | |
1681 | ||
1682 | /* | |
f470021a | 1683 | * Traditionally we see ptrace'd stopped tasks regardless of options. |
98abed02 | 1684 | */ |
f470021a | 1685 | options |= WUNTRACED; |
98abed02 | 1686 | |
f470021a RM |
1687 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
1688 | int ret = wait_consider_task(tsk, 1, p, notask_error, | |
1689 | type, pid, options, | |
1690 | infop, stat_addr, ru); | |
1691 | if (ret) | |
98abed02 | 1692 | return ret; |
98abed02 RM |
1693 | } |
1694 | ||
1695 | return 0; | |
1696 | } | |
1697 | ||
161550d7 EB |
1698 | static long do_wait(enum pid_type type, struct pid *pid, int options, |
1699 | struct siginfo __user *infop, int __user *stat_addr, | |
1700 | struct rusage __user *ru) | |
1da177e4 LT |
1701 | { |
1702 | DECLARE_WAITQUEUE(wait, current); | |
1703 | struct task_struct *tsk; | |
98abed02 | 1704 | int retval; |
1da177e4 LT |
1705 | |
1706 | add_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1707 | repeat: | |
98abed02 RM |
1708 | /* |
1709 | * If there is nothing that can match our critiera just get out. | |
1710 | * We will clear @retval to zero if we see any child that might later | |
1711 | * match our criteria, even if we are not able to reap it yet. | |
1712 | */ | |
161550d7 EB |
1713 | retval = -ECHILD; |
1714 | if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type]))) | |
1715 | goto end; | |
1716 | ||
1da177e4 LT |
1717 | current->state = TASK_INTERRUPTIBLE; |
1718 | read_lock(&tasklist_lock); | |
1719 | tsk = current; | |
1720 | do { | |
98abed02 RM |
1721 | int tsk_result = do_wait_thread(tsk, &retval, |
1722 | type, pid, options, | |
1723 | infop, stat_addr, ru); | |
1724 | if (!tsk_result) | |
1725 | tsk_result = ptrace_do_wait(tsk, &retval, | |
1726 | type, pid, options, | |
1727 | infop, stat_addr, ru); | |
1728 | if (tsk_result) { | |
1729 | /* | |
1730 | * tasklist_lock is unlocked and we have a final result. | |
1731 | */ | |
1732 | retval = tsk_result; | |
1733 | goto end; | |
1da177e4 | 1734 | } |
98abed02 | 1735 | |
1da177e4 LT |
1736 | if (options & __WNOTHREAD) |
1737 | break; | |
1738 | tsk = next_thread(tsk); | |
125e1874 | 1739 | BUG_ON(tsk->signal != current->signal); |
1da177e4 | 1740 | } while (tsk != current); |
1da177e4 | 1741 | read_unlock(&tasklist_lock); |
f2cc3eb1 | 1742 | |
98abed02 | 1743 | if (!retval && !(options & WNOHANG)) { |
1da177e4 | 1744 | retval = -ERESTARTSYS; |
98abed02 RM |
1745 | if (!signal_pending(current)) { |
1746 | schedule(); | |
1747 | goto repeat; | |
1748 | } | |
1da177e4 | 1749 | } |
98abed02 | 1750 | |
1da177e4 LT |
1751 | end: |
1752 | current->state = TASK_RUNNING; | |
1753 | remove_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1754 | if (infop) { | |
1755 | if (retval > 0) | |
9cbab810 | 1756 | retval = 0; |
1da177e4 LT |
1757 | else { |
1758 | /* | |
1759 | * For a WNOHANG return, clear out all the fields | |
1760 | * we would set so the user can easily tell the | |
1761 | * difference. | |
1762 | */ | |
1763 | if (!retval) | |
1764 | retval = put_user(0, &infop->si_signo); | |
1765 | if (!retval) | |
1766 | retval = put_user(0, &infop->si_errno); | |
1767 | if (!retval) | |
1768 | retval = put_user(0, &infop->si_code); | |
1769 | if (!retval) | |
1770 | retval = put_user(0, &infop->si_pid); | |
1771 | if (!retval) | |
1772 | retval = put_user(0, &infop->si_uid); | |
1773 | if (!retval) | |
1774 | retval = put_user(0, &infop->si_status); | |
1775 | } | |
1776 | } | |
1777 | return retval; | |
1778 | } | |
1779 | ||
161550d7 | 1780 | asmlinkage long sys_waitid(int which, pid_t upid, |
1da177e4 LT |
1781 | struct siginfo __user *infop, int options, |
1782 | struct rusage __user *ru) | |
1783 | { | |
161550d7 EB |
1784 | struct pid *pid = NULL; |
1785 | enum pid_type type; | |
1da177e4 LT |
1786 | long ret; |
1787 | ||
1788 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1789 | return -EINVAL; | |
1790 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1791 | return -EINVAL; | |
1792 | ||
1793 | switch (which) { | |
1794 | case P_ALL: | |
161550d7 | 1795 | type = PIDTYPE_MAX; |
1da177e4 LT |
1796 | break; |
1797 | case P_PID: | |
161550d7 EB |
1798 | type = PIDTYPE_PID; |
1799 | if (upid <= 0) | |
1da177e4 LT |
1800 | return -EINVAL; |
1801 | break; | |
1802 | case P_PGID: | |
161550d7 EB |
1803 | type = PIDTYPE_PGID; |
1804 | if (upid <= 0) | |
1da177e4 | 1805 | return -EINVAL; |
1da177e4 LT |
1806 | break; |
1807 | default: | |
1808 | return -EINVAL; | |
1809 | } | |
1810 | ||
161550d7 EB |
1811 | if (type < PIDTYPE_MAX) |
1812 | pid = find_get_pid(upid); | |
1813 | ret = do_wait(type, pid, options, infop, NULL, ru); | |
1814 | put_pid(pid); | |
1da177e4 LT |
1815 | |
1816 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1817 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
1da177e4 LT |
1818 | return ret; |
1819 | } | |
1820 | ||
161550d7 | 1821 | asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr, |
1da177e4 LT |
1822 | int options, struct rusage __user *ru) |
1823 | { | |
161550d7 EB |
1824 | struct pid *pid = NULL; |
1825 | enum pid_type type; | |
1da177e4 LT |
1826 | long ret; |
1827 | ||
1828 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1829 | __WNOTHREAD|__WCLONE|__WALL)) | |
1830 | return -EINVAL; | |
161550d7 EB |
1831 | |
1832 | if (upid == -1) | |
1833 | type = PIDTYPE_MAX; | |
1834 | else if (upid < 0) { | |
1835 | type = PIDTYPE_PGID; | |
1836 | pid = find_get_pid(-upid); | |
1837 | } else if (upid == 0) { | |
1838 | type = PIDTYPE_PGID; | |
1839 | pid = get_pid(task_pgrp(current)); | |
1840 | } else /* upid > 0 */ { | |
1841 | type = PIDTYPE_PID; | |
1842 | pid = find_get_pid(upid); | |
1843 | } | |
1844 | ||
1845 | ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru); | |
1846 | put_pid(pid); | |
1da177e4 LT |
1847 | |
1848 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1849 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
1da177e4 LT |
1850 | return ret; |
1851 | } | |
1852 | ||
1853 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1854 | ||
1855 | /* | |
1856 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1857 | * implemented by calling sys_wait4() from libc.a. | |
1858 | */ | |
1859 | asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options) | |
1860 | { | |
1861 | return sys_wait4(pid, stat_addr, options, NULL); | |
1862 | } | |
1863 | ||
1864 | #endif |