Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/exit.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
1da177e4 LT |
7 | #include <linux/mm.h> |
8 | #include <linux/slab.h> | |
9 | #include <linux/interrupt.h> | |
1da177e4 | 10 | #include <linux/module.h> |
c59ede7b | 11 | #include <linux/capability.h> |
1da177e4 LT |
12 | #include <linux/completion.h> |
13 | #include <linux/personality.h> | |
14 | #include <linux/tty.h> | |
6b3286ed | 15 | #include <linux/mnt_namespace.h> |
1da177e4 LT |
16 | #include <linux/key.h> |
17 | #include <linux/security.h> | |
18 | #include <linux/cpu.h> | |
19 | #include <linux/acct.h> | |
8f0ab514 | 20 | #include <linux/tsacct_kern.h> |
1da177e4 LT |
21 | #include <linux/file.h> |
22 | #include <linux/binfmts.h> | |
ab516013 | 23 | #include <linux/nsproxy.h> |
84d73786 | 24 | #include <linux/pid_namespace.h> |
1da177e4 LT |
25 | #include <linux/ptrace.h> |
26 | #include <linux/profile.h> | |
fba2afaa | 27 | #include <linux/signalfd.h> |
1da177e4 LT |
28 | #include <linux/mount.h> |
29 | #include <linux/proc_fs.h> | |
49d769d5 | 30 | #include <linux/kthread.h> |
1da177e4 | 31 | #include <linux/mempolicy.h> |
c757249a | 32 | #include <linux/taskstats_kern.h> |
ca74e92b | 33 | #include <linux/delayacct.h> |
83144186 | 34 | #include <linux/freezer.h> |
1da177e4 LT |
35 | #include <linux/cpuset.h> |
36 | #include <linux/syscalls.h> | |
7ed20e1a | 37 | #include <linux/signal.h> |
6a14c5c9 | 38 | #include <linux/posix-timers.h> |
9f46080c | 39 | #include <linux/cn_proc.h> |
de5097c2 | 40 | #include <linux/mutex.h> |
0771dfef | 41 | #include <linux/futex.h> |
34f192c6 | 42 | #include <linux/compat.h> |
b92ce558 | 43 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 44 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 45 | #include <linux/resource.h> |
0d67a46d | 46 | #include <linux/blkdev.h> |
6eaeeaba | 47 | #include <linux/task_io_accounting_ops.h> |
1da177e4 LT |
48 | |
49 | #include <asm/uaccess.h> | |
50 | #include <asm/unistd.h> | |
51 | #include <asm/pgtable.h> | |
52 | #include <asm/mmu_context.h> | |
53 | ||
54 | extern void sem_exit (void); | |
1da177e4 | 55 | |
408b664a AB |
56 | static void exit_mm(struct task_struct * tsk); |
57 | ||
1da177e4 LT |
58 | static void __unhash_process(struct task_struct *p) |
59 | { | |
60 | nr_threads--; | |
61 | detach_pid(p, PIDTYPE_PID); | |
1da177e4 LT |
62 | if (thread_group_leader(p)) { |
63 | detach_pid(p, PIDTYPE_PGID); | |
64 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 65 | |
5e85d4ab | 66 | list_del_rcu(&p->tasks); |
73b9ebfe | 67 | __get_cpu_var(process_counts)--; |
1da177e4 | 68 | } |
47e65328 | 69 | list_del_rcu(&p->thread_group); |
c97d9893 | 70 | remove_parent(p); |
1da177e4 LT |
71 | } |
72 | ||
6a14c5c9 ON |
73 | /* |
74 | * This function expects the tasklist_lock write-locked. | |
75 | */ | |
76 | static void __exit_signal(struct task_struct *tsk) | |
77 | { | |
78 | struct signal_struct *sig = tsk->signal; | |
79 | struct sighand_struct *sighand; | |
80 | ||
81 | BUG_ON(!sig); | |
82 | BUG_ON(!atomic_read(&sig->count)); | |
83 | ||
84 | rcu_read_lock(); | |
85 | sighand = rcu_dereference(tsk->sighand); | |
86 | spin_lock(&sighand->siglock); | |
87 | ||
fba2afaa DL |
88 | /* |
89 | * Notify that this sighand has been detached. This must | |
90 | * be called with the tsk->sighand lock held. Also, this | |
91 | * access tsk->sighand internally, so it must be called | |
92 | * before tsk->sighand is reset. | |
93 | */ | |
94 | signalfd_detach_locked(tsk); | |
95 | ||
6a14c5c9 ON |
96 | posix_cpu_timers_exit(tsk); |
97 | if (atomic_dec_and_test(&sig->count)) | |
98 | posix_cpu_timers_exit_group(tsk); | |
99 | else { | |
100 | /* | |
101 | * If there is any task waiting for the group exit | |
102 | * then notify it: | |
103 | */ | |
104 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) { | |
105 | wake_up_process(sig->group_exit_task); | |
106 | sig->group_exit_task = NULL; | |
107 | } | |
108 | if (tsk == sig->curr_target) | |
109 | sig->curr_target = next_thread(tsk); | |
110 | /* | |
111 | * Accumulate here the counters for all threads but the | |
112 | * group leader as they die, so they can be added into | |
113 | * the process-wide totals when those are taken. | |
114 | * The group leader stays around as a zombie as long | |
115 | * as there are other threads. When it gets reaped, | |
116 | * the exit.c code will add its counts into these totals. | |
117 | * We won't ever get here for the group leader, since it | |
118 | * will have been the last reference on the signal_struct. | |
119 | */ | |
120 | sig->utime = cputime_add(sig->utime, tsk->utime); | |
121 | sig->stime = cputime_add(sig->stime, tsk->stime); | |
122 | sig->min_flt += tsk->min_flt; | |
123 | sig->maj_flt += tsk->maj_flt; | |
124 | sig->nvcsw += tsk->nvcsw; | |
125 | sig->nivcsw += tsk->nivcsw; | |
6eaeeaba ED |
126 | sig->inblock += task_io_get_inblock(tsk); |
127 | sig->oublock += task_io_get_oublock(tsk); | |
172ba844 | 128 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
6a14c5c9 ON |
129 | sig = NULL; /* Marker for below. */ |
130 | } | |
131 | ||
5876700c ON |
132 | __unhash_process(tsk); |
133 | ||
6a14c5c9 | 134 | tsk->signal = NULL; |
a7e5328a | 135 | tsk->sighand = NULL; |
6a14c5c9 ON |
136 | spin_unlock(&sighand->siglock); |
137 | rcu_read_unlock(); | |
138 | ||
a7e5328a | 139 | __cleanup_sighand(sighand); |
6a14c5c9 ON |
140 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
141 | flush_sigqueue(&tsk->pending); | |
142 | if (sig) { | |
143 | flush_sigqueue(&sig->shared_pending); | |
093a8e8a | 144 | taskstats_tgid_free(sig); |
6a14c5c9 ON |
145 | __cleanup_signal(sig); |
146 | } | |
147 | } | |
148 | ||
8c7904a0 EB |
149 | static void delayed_put_task_struct(struct rcu_head *rhp) |
150 | { | |
151 | put_task_struct(container_of(rhp, struct task_struct, rcu)); | |
152 | } | |
153 | ||
1da177e4 LT |
154 | void release_task(struct task_struct * p) |
155 | { | |
36c8b586 | 156 | struct task_struct *leader; |
1da177e4 | 157 | int zap_leader; |
1f09f974 | 158 | repeat: |
1da177e4 | 159 | atomic_dec(&p->user->processes); |
1da177e4 | 160 | write_lock_irq(&tasklist_lock); |
1f09f974 | 161 | ptrace_unlink(p); |
1da177e4 LT |
162 | BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children)); |
163 | __exit_signal(p); | |
35f5cad8 | 164 | |
1da177e4 LT |
165 | /* |
166 | * If we are the last non-leader member of the thread | |
167 | * group, and the leader is zombie, then notify the | |
168 | * group leader's parent process. (if it wants notification.) | |
169 | */ | |
170 | zap_leader = 0; | |
171 | leader = p->group_leader; | |
172 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
173 | BUG_ON(leader->exit_signal == -1); | |
174 | do_notify_parent(leader, leader->exit_signal); | |
175 | /* | |
176 | * If we were the last child thread and the leader has | |
177 | * exited already, and the leader's parent ignores SIGCHLD, | |
178 | * then we are the one who should release the leader. | |
179 | * | |
180 | * do_notify_parent() will have marked it self-reaping in | |
181 | * that case. | |
182 | */ | |
183 | zap_leader = (leader->exit_signal == -1); | |
184 | } | |
185 | ||
1da177e4 | 186 | write_unlock_irq(&tasklist_lock); |
48e6484d | 187 | proc_flush_task(p); |
1da177e4 | 188 | release_thread(p); |
8c7904a0 | 189 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
190 | |
191 | p = leader; | |
192 | if (unlikely(zap_leader)) | |
193 | goto repeat; | |
194 | } | |
195 | ||
1da177e4 LT |
196 | /* |
197 | * This checks not only the pgrp, but falls back on the pid if no | |
198 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
199 | * without this... | |
04a2e6a5 EB |
200 | * |
201 | * The caller must hold rcu lock or the tasklist lock. | |
1da177e4 | 202 | */ |
04a2e6a5 | 203 | struct pid *session_of_pgrp(struct pid *pgrp) |
1da177e4 LT |
204 | { |
205 | struct task_struct *p; | |
04a2e6a5 | 206 | struct pid *sid = NULL; |
62dfb554 | 207 | |
04a2e6a5 | 208 | p = pid_task(pgrp, PIDTYPE_PGID); |
62dfb554 | 209 | if (p == NULL) |
04a2e6a5 | 210 | p = pid_task(pgrp, PIDTYPE_PID); |
62dfb554 | 211 | if (p != NULL) |
04a2e6a5 | 212 | sid = task_session(p); |
62dfb554 | 213 | |
1da177e4 LT |
214 | return sid; |
215 | } | |
216 | ||
217 | /* | |
218 | * Determine if a process group is "orphaned", according to the POSIX | |
219 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
220 | * by terminal-generated stop signals. Newly orphaned process groups are | |
221 | * to receive a SIGHUP and a SIGCONT. | |
222 | * | |
223 | * "I ask you, have you ever known what it is to be an orphan?" | |
224 | */ | |
0475ac08 | 225 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
226 | { |
227 | struct task_struct *p; | |
228 | int ret = 1; | |
229 | ||
0475ac08 | 230 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
1da177e4 LT |
231 | if (p == ignored_task |
232 | || p->exit_state | |
f400e198 | 233 | || is_init(p->real_parent)) |
1da177e4 | 234 | continue; |
0475ac08 EB |
235 | if (task_pgrp(p->real_parent) != pgrp && |
236 | task_session(p->real_parent) == task_session(p)) { | |
1da177e4 LT |
237 | ret = 0; |
238 | break; | |
239 | } | |
0475ac08 | 240 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
241 | return ret; /* (sighing) "Often!" */ |
242 | } | |
243 | ||
3e7cd6c4 | 244 | int is_current_pgrp_orphaned(void) |
1da177e4 LT |
245 | { |
246 | int retval; | |
247 | ||
248 | read_lock(&tasklist_lock); | |
3e7cd6c4 | 249 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
1da177e4 LT |
250 | read_unlock(&tasklist_lock); |
251 | ||
252 | return retval; | |
253 | } | |
254 | ||
0475ac08 | 255 | static int has_stopped_jobs(struct pid *pgrp) |
1da177e4 LT |
256 | { |
257 | int retval = 0; | |
258 | struct task_struct *p; | |
259 | ||
0475ac08 | 260 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
1da177e4 LT |
261 | if (p->state != TASK_STOPPED) |
262 | continue; | |
1da177e4 LT |
263 | retval = 1; |
264 | break; | |
0475ac08 | 265 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
266 | return retval; |
267 | } | |
268 | ||
269 | /** | |
49d769d5 | 270 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
1da177e4 LT |
271 | * |
272 | * If a kernel thread is launched as a result of a system call, or if | |
49d769d5 EB |
273 | * it ever exits, it should generally reparent itself to kthreadd so it |
274 | * isn't in the way of other processes and is correctly cleaned up on exit. | |
1da177e4 LT |
275 | * |
276 | * The various task state such as scheduling policy and priority may have | |
277 | * been inherited from a user process, so we reset them to sane values here. | |
278 | * | |
49d769d5 | 279 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
1da177e4 | 280 | */ |
49d769d5 | 281 | static void reparent_to_kthreadd(void) |
1da177e4 LT |
282 | { |
283 | write_lock_irq(&tasklist_lock); | |
284 | ||
285 | ptrace_unlink(current); | |
286 | /* Reparent to init */ | |
9b678ece | 287 | remove_parent(current); |
49d769d5 | 288 | current->real_parent = current->parent = kthreadd_task; |
9b678ece | 289 | add_parent(current); |
1da177e4 LT |
290 | |
291 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
292 | current->exit_signal = SIGCHLD; | |
293 | ||
e05606d3 | 294 | if (task_nice(current) < 0) |
1da177e4 LT |
295 | set_user_nice(current, 0); |
296 | /* cpus_allowed? */ | |
297 | /* rt_priority? */ | |
298 | /* signals? */ | |
299 | security_task_reparent_to_init(current); | |
300 | memcpy(current->signal->rlim, init_task.signal->rlim, | |
301 | sizeof(current->signal->rlim)); | |
302 | atomic_inc(&(INIT_USER->__count)); | |
303 | write_unlock_irq(&tasklist_lock); | |
304 | switch_uid(INIT_USER); | |
305 | } | |
306 | ||
307 | void __set_special_pids(pid_t session, pid_t pgrp) | |
308 | { | |
e19f247a | 309 | struct task_struct *curr = current->group_leader; |
1da177e4 | 310 | |
937949d9 | 311 | if (process_session(curr) != session) { |
1da177e4 | 312 | detach_pid(curr, PIDTYPE_SID); |
1ec320af | 313 | set_signal_session(curr->signal, session); |
e713d0da | 314 | attach_pid(curr, PIDTYPE_SID, find_pid(session)); |
1da177e4 LT |
315 | } |
316 | if (process_group(curr) != pgrp) { | |
317 | detach_pid(curr, PIDTYPE_PGID); | |
318 | curr->signal->pgrp = pgrp; | |
e713d0da | 319 | attach_pid(curr, PIDTYPE_PGID, find_pid(pgrp)); |
1da177e4 LT |
320 | } |
321 | } | |
322 | ||
ae424ae4 | 323 | static void set_special_pids(pid_t session, pid_t pgrp) |
1da177e4 LT |
324 | { |
325 | write_lock_irq(&tasklist_lock); | |
326 | __set_special_pids(session, pgrp); | |
327 | write_unlock_irq(&tasklist_lock); | |
328 | } | |
329 | ||
330 | /* | |
331 | * Let kernel threads use this to say that they | |
332 | * allow a certain signal (since daemonize() will | |
333 | * have disabled all of them by default). | |
334 | */ | |
335 | int allow_signal(int sig) | |
336 | { | |
7ed20e1a | 337 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
338 | return -EINVAL; |
339 | ||
340 | spin_lock_irq(¤t->sighand->siglock); | |
341 | sigdelset(¤t->blocked, sig); | |
342 | if (!current->mm) { | |
343 | /* Kernel threads handle their own signals. | |
344 | Let the signal code know it'll be handled, so | |
345 | that they don't get converted to SIGKILL or | |
346 | just silently dropped */ | |
347 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
348 | } | |
349 | recalc_sigpending(); | |
350 | spin_unlock_irq(¤t->sighand->siglock); | |
351 | return 0; | |
352 | } | |
353 | ||
354 | EXPORT_SYMBOL(allow_signal); | |
355 | ||
356 | int disallow_signal(int sig) | |
357 | { | |
7ed20e1a | 358 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
359 | return -EINVAL; |
360 | ||
361 | spin_lock_irq(¤t->sighand->siglock); | |
10ab825b | 362 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
1da177e4 LT |
363 | recalc_sigpending(); |
364 | spin_unlock_irq(¤t->sighand->siglock); | |
365 | return 0; | |
366 | } | |
367 | ||
368 | EXPORT_SYMBOL(disallow_signal); | |
369 | ||
370 | /* | |
371 | * Put all the gunge required to become a kernel thread without | |
372 | * attached user resources in one place where it belongs. | |
373 | */ | |
374 | ||
375 | void daemonize(const char *name, ...) | |
376 | { | |
377 | va_list args; | |
378 | struct fs_struct *fs; | |
379 | sigset_t blocked; | |
380 | ||
381 | va_start(args, name); | |
382 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
383 | va_end(args); | |
384 | ||
385 | /* | |
386 | * If we were started as result of loading a module, close all of the | |
387 | * user space pages. We don't need them, and if we didn't close them | |
388 | * they would be locked into memory. | |
389 | */ | |
390 | exit_mm(current); | |
83144186 RW |
391 | /* |
392 | * We don't want to have TIF_FREEZE set if the system-wide hibernation | |
393 | * or suspend transition begins right now. | |
394 | */ | |
395 | current->flags |= PF_NOFREEZE; | |
1da177e4 LT |
396 | |
397 | set_special_pids(1, 1); | |
24ec839c | 398 | proc_clear_tty(current); |
1da177e4 LT |
399 | |
400 | /* Block and flush all signals */ | |
401 | sigfillset(&blocked); | |
402 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
403 | flush_signals(current); | |
404 | ||
405 | /* Become as one with the init task */ | |
406 | ||
407 | exit_fs(current); /* current->fs->count--; */ | |
408 | fs = init_task.fs; | |
409 | current->fs = fs; | |
410 | atomic_inc(&fs->count); | |
ab516013 | 411 | |
444f378b | 412 | exit_task_namespaces(current); |
ab516013 | 413 | current->nsproxy = init_task.nsproxy; |
ab516013 SH |
414 | get_task_namespaces(current); |
415 | ||
1da177e4 LT |
416 | exit_files(current); |
417 | current->files = init_task.files; | |
418 | atomic_inc(¤t->files->count); | |
419 | ||
49d769d5 | 420 | reparent_to_kthreadd(); |
1da177e4 LT |
421 | } |
422 | ||
423 | EXPORT_SYMBOL(daemonize); | |
424 | ||
858119e1 | 425 | static void close_files(struct files_struct * files) |
1da177e4 LT |
426 | { |
427 | int i, j; | |
badf1662 | 428 | struct fdtable *fdt; |
1da177e4 LT |
429 | |
430 | j = 0; | |
4fb3a538 DS |
431 | |
432 | /* | |
433 | * It is safe to dereference the fd table without RCU or | |
434 | * ->file_lock because this is the last reference to the | |
435 | * files structure. | |
436 | */ | |
badf1662 | 437 | fdt = files_fdtable(files); |
1da177e4 LT |
438 | for (;;) { |
439 | unsigned long set; | |
440 | i = j * __NFDBITS; | |
bbea9f69 | 441 | if (i >= fdt->max_fds) |
1da177e4 | 442 | break; |
badf1662 | 443 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
444 | while (set) { |
445 | if (set & 1) { | |
badf1662 | 446 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 447 | if (file) { |
1da177e4 | 448 | filp_close(file, files); |
944be0b2 IM |
449 | cond_resched(); |
450 | } | |
1da177e4 LT |
451 | } |
452 | i++; | |
453 | set >>= 1; | |
454 | } | |
455 | } | |
456 | } | |
457 | ||
458 | struct files_struct *get_files_struct(struct task_struct *task) | |
459 | { | |
460 | struct files_struct *files; | |
461 | ||
462 | task_lock(task); | |
463 | files = task->files; | |
464 | if (files) | |
465 | atomic_inc(&files->count); | |
466 | task_unlock(task); | |
467 | ||
468 | return files; | |
469 | } | |
470 | ||
471 | void fastcall put_files_struct(struct files_struct *files) | |
472 | { | |
badf1662 DS |
473 | struct fdtable *fdt; |
474 | ||
1da177e4 LT |
475 | if (atomic_dec_and_test(&files->count)) { |
476 | close_files(files); | |
477 | /* | |
478 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
479 | * If the fdtable was embedded, pass files for freeing |
480 | * at the end of the RCU grace period. Otherwise, | |
481 | * you can free files immediately. | |
1da177e4 | 482 | */ |
badf1662 | 483 | fdt = files_fdtable(files); |
4fd45812 | 484 | if (fdt != &files->fdtab) |
ab2af1f5 | 485 | kmem_cache_free(files_cachep, files); |
01b2d93c | 486 | free_fdtable(fdt); |
1da177e4 LT |
487 | } |
488 | } | |
489 | ||
490 | EXPORT_SYMBOL(put_files_struct); | |
491 | ||
3b9b8ab6 KK |
492 | void reset_files_struct(struct task_struct *tsk, struct files_struct *files) |
493 | { | |
494 | struct files_struct *old; | |
495 | ||
496 | old = tsk->files; | |
497 | task_lock(tsk); | |
498 | tsk->files = files; | |
499 | task_unlock(tsk); | |
500 | put_files_struct(old); | |
501 | } | |
502 | EXPORT_SYMBOL(reset_files_struct); | |
503 | ||
1da177e4 LT |
504 | static inline void __exit_files(struct task_struct *tsk) |
505 | { | |
506 | struct files_struct * files = tsk->files; | |
507 | ||
508 | if (files) { | |
509 | task_lock(tsk); | |
510 | tsk->files = NULL; | |
511 | task_unlock(tsk); | |
512 | put_files_struct(files); | |
513 | } | |
514 | } | |
515 | ||
516 | void exit_files(struct task_struct *tsk) | |
517 | { | |
518 | __exit_files(tsk); | |
519 | } | |
520 | ||
521 | static inline void __put_fs_struct(struct fs_struct *fs) | |
522 | { | |
523 | /* No need to hold fs->lock if we are killing it */ | |
524 | if (atomic_dec_and_test(&fs->count)) { | |
525 | dput(fs->root); | |
526 | mntput(fs->rootmnt); | |
527 | dput(fs->pwd); | |
528 | mntput(fs->pwdmnt); | |
529 | if (fs->altroot) { | |
530 | dput(fs->altroot); | |
531 | mntput(fs->altrootmnt); | |
532 | } | |
533 | kmem_cache_free(fs_cachep, fs); | |
534 | } | |
535 | } | |
536 | ||
537 | void put_fs_struct(struct fs_struct *fs) | |
538 | { | |
539 | __put_fs_struct(fs); | |
540 | } | |
541 | ||
542 | static inline void __exit_fs(struct task_struct *tsk) | |
543 | { | |
544 | struct fs_struct * fs = tsk->fs; | |
545 | ||
546 | if (fs) { | |
547 | task_lock(tsk); | |
548 | tsk->fs = NULL; | |
549 | task_unlock(tsk); | |
550 | __put_fs_struct(fs); | |
551 | } | |
552 | } | |
553 | ||
554 | void exit_fs(struct task_struct *tsk) | |
555 | { | |
556 | __exit_fs(tsk); | |
557 | } | |
558 | ||
559 | EXPORT_SYMBOL_GPL(exit_fs); | |
560 | ||
561 | /* | |
562 | * Turn us into a lazy TLB process if we | |
563 | * aren't already.. | |
564 | */ | |
408b664a | 565 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
566 | { |
567 | struct mm_struct *mm = tsk->mm; | |
568 | ||
569 | mm_release(tsk, mm); | |
570 | if (!mm) | |
571 | return; | |
572 | /* | |
573 | * Serialize with any possible pending coredump. | |
574 | * We must hold mmap_sem around checking core_waiters | |
575 | * and clearing tsk->mm. The core-inducing thread | |
576 | * will increment core_waiters for each thread in the | |
577 | * group with ->mm != NULL. | |
578 | */ | |
579 | down_read(&mm->mmap_sem); | |
580 | if (mm->core_waiters) { | |
581 | up_read(&mm->mmap_sem); | |
582 | down_write(&mm->mmap_sem); | |
583 | if (!--mm->core_waiters) | |
584 | complete(mm->core_startup_done); | |
585 | up_write(&mm->mmap_sem); | |
586 | ||
587 | wait_for_completion(&mm->core_done); | |
588 | down_read(&mm->mmap_sem); | |
589 | } | |
590 | atomic_inc(&mm->mm_count); | |
125e1874 | 591 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
592 | /* more a memory barrier than a real lock */ |
593 | task_lock(tsk); | |
594 | tsk->mm = NULL; | |
595 | up_read(&mm->mmap_sem); | |
596 | enter_lazy_tlb(mm, current); | |
597 | task_unlock(tsk); | |
598 | mmput(mm); | |
599 | } | |
600 | ||
36c8b586 IM |
601 | static inline void |
602 | choose_new_parent(struct task_struct *p, struct task_struct *reaper) | |
1da177e4 LT |
603 | { |
604 | /* | |
605 | * Make sure we're not reparenting to ourselves and that | |
606 | * the parent is not a zombie. | |
607 | */ | |
d799f035 | 608 | BUG_ON(p == reaper || reaper->exit_state); |
1da177e4 | 609 | p->real_parent = reaper; |
1da177e4 LT |
610 | } |
611 | ||
36c8b586 IM |
612 | static void |
613 | reparent_thread(struct task_struct *p, struct task_struct *father, int traced) | |
1da177e4 | 614 | { |
241ceee0 ON |
615 | if (p->pdeath_signal) |
616 | /* We already hold the tasklist_lock here. */ | |
617 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | |
618 | ||
1da177e4 LT |
619 | /* Move the child from its dying parent to the new one. */ |
620 | if (unlikely(traced)) { | |
621 | /* Preserve ptrace links if someone else is tracing this child. */ | |
622 | list_del_init(&p->ptrace_list); | |
623 | if (p->parent != p->real_parent) | |
624 | list_add(&p->ptrace_list, &p->real_parent->ptrace_children); | |
625 | } else { | |
626 | /* If this child is being traced, then we're the one tracing it | |
627 | * anyway, so let go of it. | |
628 | */ | |
629 | p->ptrace = 0; | |
6ac781b1 | 630 | remove_parent(p); |
1da177e4 | 631 | p->parent = p->real_parent; |
6ac781b1 | 632 | add_parent(p); |
1da177e4 | 633 | |
b2b2cbc4 | 634 | if (p->state == TASK_TRACED) { |
1da177e4 LT |
635 | /* |
636 | * If it was at a trace stop, turn it into | |
637 | * a normal stop since it's no longer being | |
638 | * traced. | |
639 | */ | |
640 | ptrace_untrace(p); | |
641 | } | |
642 | } | |
643 | ||
b2b2cbc4 EB |
644 | /* If this is a threaded reparent there is no need to |
645 | * notify anyone anything has happened. | |
646 | */ | |
647 | if (p->real_parent->group_leader == father->group_leader) | |
648 | return; | |
649 | ||
650 | /* We don't want people slaying init. */ | |
651 | if (p->exit_signal != -1) | |
652 | p->exit_signal = SIGCHLD; | |
b2b2cbc4 EB |
653 | |
654 | /* If we'd notified the old parent about this child's death, | |
655 | * also notify the new parent. | |
656 | */ | |
657 | if (!traced && p->exit_state == EXIT_ZOMBIE && | |
658 | p->exit_signal != -1 && thread_group_empty(p)) | |
659 | do_notify_parent(p, p->exit_signal); | |
660 | ||
1da177e4 LT |
661 | /* |
662 | * process group orphan check | |
663 | * Case ii: Our child is in a different pgrp | |
664 | * than we are, and it was the only connection | |
665 | * outside, so the child pgrp is now orphaned. | |
666 | */ | |
0475ac08 EB |
667 | if ((task_pgrp(p) != task_pgrp(father)) && |
668 | (task_session(p) == task_session(father))) { | |
669 | struct pid *pgrp = task_pgrp(p); | |
1da177e4 | 670 | |
937949d9 CLG |
671 | if (will_become_orphaned_pgrp(pgrp, NULL) && |
672 | has_stopped_jobs(pgrp)) { | |
0475ac08 EB |
673 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); |
674 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
1da177e4 LT |
675 | } |
676 | } | |
677 | } | |
678 | ||
679 | /* | |
680 | * When we die, we re-parent all our children. | |
681 | * Try to give them to another thread in our thread | |
682 | * group, and if no such member exists, give it to | |
84d73786 SB |
683 | * the child reaper process (ie "init") in our pid |
684 | * space. | |
1da177e4 | 685 | */ |
36c8b586 IM |
686 | static void |
687 | forget_original_parent(struct task_struct *father, struct list_head *to_release) | |
1da177e4 LT |
688 | { |
689 | struct task_struct *p, *reaper = father; | |
690 | struct list_head *_p, *_n; | |
691 | ||
692 | do { | |
693 | reaper = next_thread(reaper); | |
694 | if (reaper == father) { | |
84d73786 | 695 | reaper = child_reaper(father); |
1da177e4 LT |
696 | break; |
697 | } | |
698 | } while (reaper->exit_state); | |
699 | ||
700 | /* | |
701 | * There are only two places where our children can be: | |
702 | * | |
703 | * - in our child list | |
704 | * - in our ptraced child list | |
705 | * | |
706 | * Search them and reparent children. | |
707 | */ | |
708 | list_for_each_safe(_p, _n, &father->children) { | |
709 | int ptrace; | |
36c8b586 | 710 | p = list_entry(_p, struct task_struct, sibling); |
1da177e4 LT |
711 | |
712 | ptrace = p->ptrace; | |
713 | ||
714 | /* if father isn't the real parent, then ptrace must be enabled */ | |
715 | BUG_ON(father != p->real_parent && !ptrace); | |
716 | ||
717 | if (father == p->real_parent) { | |
718 | /* reparent with a reaper, real father it's us */ | |
d799f035 | 719 | choose_new_parent(p, reaper); |
1da177e4 LT |
720 | reparent_thread(p, father, 0); |
721 | } else { | |
722 | /* reparent ptraced task to its real parent */ | |
723 | __ptrace_unlink (p); | |
724 | if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 && | |
725 | thread_group_empty(p)) | |
726 | do_notify_parent(p, p->exit_signal); | |
727 | } | |
728 | ||
729 | /* | |
730 | * if the ptraced child is a zombie with exit_signal == -1 | |
731 | * we must collect it before we exit, or it will remain | |
732 | * zombie forever since we prevented it from self-reap itself | |
733 | * while it was being traced by us, to be able to see it in wait4. | |
734 | */ | |
735 | if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1)) | |
736 | list_add(&p->ptrace_list, to_release); | |
737 | } | |
738 | list_for_each_safe(_p, _n, &father->ptrace_children) { | |
36c8b586 | 739 | p = list_entry(_p, struct task_struct, ptrace_list); |
d799f035 | 740 | choose_new_parent(p, reaper); |
1da177e4 LT |
741 | reparent_thread(p, father, 1); |
742 | } | |
743 | } | |
744 | ||
745 | /* | |
746 | * Send signals to all our closest relatives so that they know | |
747 | * to properly mourn us.. | |
748 | */ | |
749 | static void exit_notify(struct task_struct *tsk) | |
750 | { | |
751 | int state; | |
752 | struct task_struct *t; | |
753 | struct list_head ptrace_dead, *_p, *_n; | |
0475ac08 | 754 | struct pid *pgrp; |
1da177e4 LT |
755 | |
756 | if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT) | |
757 | && !thread_group_empty(tsk)) { | |
758 | /* | |
759 | * This occurs when there was a race between our exit | |
760 | * syscall and a group signal choosing us as the one to | |
761 | * wake up. It could be that we are the only thread | |
762 | * alerted to check for pending signals, but another thread | |
763 | * should be woken now to take the signal since we will not. | |
764 | * Now we'll wake all the threads in the group just to make | |
765 | * sure someone gets all the pending signals. | |
766 | */ | |
767 | read_lock(&tasklist_lock); | |
768 | spin_lock_irq(&tsk->sighand->siglock); | |
769 | for (t = next_thread(tsk); t != tsk; t = next_thread(t)) | |
7bb44ade RM |
770 | if (!signal_pending(t) && !(t->flags & PF_EXITING)) |
771 | recalc_sigpending_and_wake(t); | |
1da177e4 LT |
772 | spin_unlock_irq(&tsk->sighand->siglock); |
773 | read_unlock(&tasklist_lock); | |
774 | } | |
775 | ||
776 | write_lock_irq(&tasklist_lock); | |
777 | ||
778 | /* | |
779 | * This does two things: | |
780 | * | |
781 | * A. Make init inherit all the child processes | |
782 | * B. Check to see if any process groups have become orphaned | |
783 | * as a result of our exiting, and if they have any stopped | |
784 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
785 | */ | |
786 | ||
787 | INIT_LIST_HEAD(&ptrace_dead); | |
788 | forget_original_parent(tsk, &ptrace_dead); | |
789 | BUG_ON(!list_empty(&tsk->children)); | |
790 | BUG_ON(!list_empty(&tsk->ptrace_children)); | |
791 | ||
792 | /* | |
793 | * Check to see if any process groups have become orphaned | |
794 | * as a result of our exiting, and if they have any stopped | |
795 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
796 | * | |
797 | * Case i: Our father is in a different pgrp than we are | |
798 | * and we were the only connection outside, so our pgrp | |
799 | * is about to become orphaned. | |
800 | */ | |
801 | ||
802 | t = tsk->real_parent; | |
803 | ||
0475ac08 EB |
804 | pgrp = task_pgrp(tsk); |
805 | if ((task_pgrp(t) != pgrp) && | |
14e9d573 | 806 | (task_session(t) == task_session(tsk)) && |
0475ac08 EB |
807 | will_become_orphaned_pgrp(pgrp, tsk) && |
808 | has_stopped_jobs(pgrp)) { | |
809 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
810 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
1da177e4 LT |
811 | } |
812 | ||
813 | /* Let father know we died | |
814 | * | |
815 | * Thread signals are configurable, but you aren't going to use | |
816 | * that to send signals to arbitary processes. | |
817 | * That stops right now. | |
818 | * | |
819 | * If the parent exec id doesn't match the exec id we saved | |
820 | * when we started then we know the parent has changed security | |
821 | * domain. | |
822 | * | |
823 | * If our self_exec id doesn't match our parent_exec_id then | |
824 | * we have changed execution domain as these two values started | |
825 | * the same after a fork. | |
826 | * | |
827 | */ | |
828 | ||
829 | if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 && | |
830 | ( tsk->parent_exec_id != t->self_exec_id || | |
831 | tsk->self_exec_id != tsk->parent_exec_id) | |
832 | && !capable(CAP_KILL)) | |
833 | tsk->exit_signal = SIGCHLD; | |
834 | ||
835 | ||
836 | /* If something other than our normal parent is ptracing us, then | |
837 | * send it a SIGCHLD instead of honoring exit_signal. exit_signal | |
838 | * only has special meaning to our real parent. | |
839 | */ | |
840 | if (tsk->exit_signal != -1 && thread_group_empty(tsk)) { | |
841 | int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD; | |
842 | do_notify_parent(tsk, signal); | |
843 | } else if (tsk->ptrace) { | |
844 | do_notify_parent(tsk, SIGCHLD); | |
845 | } | |
846 | ||
847 | state = EXIT_ZOMBIE; | |
848 | if (tsk->exit_signal == -1 && | |
849 | (likely(tsk->ptrace == 0) || | |
850 | unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT))) | |
851 | state = EXIT_DEAD; | |
852 | tsk->exit_state = state; | |
853 | ||
854 | write_unlock_irq(&tasklist_lock); | |
855 | ||
856 | list_for_each_safe(_p, _n, &ptrace_dead) { | |
857 | list_del_init(_p); | |
36c8b586 | 858 | t = list_entry(_p, struct task_struct, ptrace_list); |
1da177e4 LT |
859 | release_task(t); |
860 | } | |
861 | ||
862 | /* If the process is dead, release it - nobody will wait for it */ | |
863 | if (state == EXIT_DEAD) | |
864 | release_task(tsk); | |
1da177e4 LT |
865 | } |
866 | ||
e18eecb8 JD |
867 | #ifdef CONFIG_DEBUG_STACK_USAGE |
868 | static void check_stack_usage(void) | |
869 | { | |
870 | static DEFINE_SPINLOCK(low_water_lock); | |
871 | static int lowest_to_date = THREAD_SIZE; | |
872 | unsigned long *n = end_of_stack(current); | |
873 | unsigned long free; | |
874 | ||
875 | while (*n == 0) | |
876 | n++; | |
877 | free = (unsigned long)n - (unsigned long)end_of_stack(current); | |
878 | ||
879 | if (free >= lowest_to_date) | |
880 | return; | |
881 | ||
882 | spin_lock(&low_water_lock); | |
883 | if (free < lowest_to_date) { | |
884 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " | |
885 | "left\n", | |
886 | current->comm, free); | |
887 | lowest_to_date = free; | |
888 | } | |
889 | spin_unlock(&low_water_lock); | |
890 | } | |
891 | #else | |
892 | static inline void check_stack_usage(void) {} | |
893 | #endif | |
894 | ||
1da177e4 LT |
895 | fastcall NORET_TYPE void do_exit(long code) |
896 | { | |
897 | struct task_struct *tsk = current; | |
898 | int group_dead; | |
899 | ||
900 | profile_task_exit(tsk); | |
901 | ||
22e2c507 JA |
902 | WARN_ON(atomic_read(&tsk->fs_excl)); |
903 | ||
1da177e4 LT |
904 | if (unlikely(in_interrupt())) |
905 | panic("Aiee, killing interrupt handler!"); | |
906 | if (unlikely(!tsk->pid)) | |
907 | panic("Attempted to kill the idle task!"); | |
84d73786 SB |
908 | if (unlikely(tsk == child_reaper(tsk))) { |
909 | if (tsk->nsproxy->pid_ns != &init_pid_ns) | |
910 | tsk->nsproxy->pid_ns->child_reaper = init_pid_ns.child_reaper; | |
911 | else | |
912 | panic("Attempted to kill init!"); | |
913 | } | |
914 | ||
1da177e4 LT |
915 | |
916 | if (unlikely(current->ptrace & PT_TRACE_EXIT)) { | |
917 | current->ptrace_message = code; | |
918 | ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP); | |
919 | } | |
920 | ||
df164db5 AN |
921 | /* |
922 | * We're taking recursive faults here in do_exit. Safest is to just | |
923 | * leave this task alone and wait for reboot. | |
924 | */ | |
925 | if (unlikely(tsk->flags & PF_EXITING)) { | |
926 | printk(KERN_ALERT | |
927 | "Fixing recursive fault but reboot is needed!\n"); | |
778e9a9c AK |
928 | /* |
929 | * We can do this unlocked here. The futex code uses | |
930 | * this flag just to verify whether the pi state | |
931 | * cleanup has been done or not. In the worst case it | |
932 | * loops once more. We pretend that the cleanup was | |
933 | * done as there is no way to return. Either the | |
934 | * OWNER_DIED bit is set by now or we push the blocked | |
935 | * task into the wait for ever nirwana as well. | |
936 | */ | |
937 | tsk->flags |= PF_EXITPIDONE; | |
afc847b7 AV |
938 | if (tsk->io_context) |
939 | exit_io_context(); | |
df164db5 AN |
940 | set_current_state(TASK_UNINTERRUPTIBLE); |
941 | schedule(); | |
942 | } | |
943 | ||
778e9a9c AK |
944 | /* |
945 | * tsk->flags are checked in the futex code to protect against | |
946 | * an exiting task cleaning up the robust pi futexes. | |
947 | */ | |
948 | spin_lock_irq(&tsk->pi_lock); | |
1da177e4 | 949 | tsk->flags |= PF_EXITING; |
778e9a9c | 950 | spin_unlock_irq(&tsk->pi_lock); |
1da177e4 | 951 | |
1da177e4 LT |
952 | if (unlikely(in_atomic())) |
953 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
954 | current->comm, current->pid, | |
955 | preempt_count()); | |
956 | ||
957 | acct_update_integrals(tsk); | |
365e9c87 HD |
958 | if (tsk->mm) { |
959 | update_hiwater_rss(tsk->mm); | |
960 | update_hiwater_vm(tsk->mm); | |
961 | } | |
1da177e4 | 962 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 963 | if (group_dead) { |
778e9a9c | 964 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 965 | exit_itimers(tsk->signal); |
c3068951 | 966 | } |
f6ec29a4 | 967 | acct_collect(code, group_dead); |
0771dfef IM |
968 | if (unlikely(tsk->robust_list)) |
969 | exit_robust_list(tsk); | |
2aa92581 | 970 | #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT) |
34f192c6 IM |
971 | if (unlikely(tsk->compat_robust_list)) |
972 | compat_exit_robust_list(tsk); | |
973 | #endif | |
522ed776 MT |
974 | if (group_dead) |
975 | tty_audit_exit(); | |
fa84cb93 AV |
976 | if (unlikely(tsk->audit_context)) |
977 | audit_free(tsk); | |
115085ea ON |
978 | |
979 | taskstats_exit(tsk, group_dead); | |
c757249a | 980 | |
1da177e4 LT |
981 | exit_mm(tsk); |
982 | ||
0e464814 | 983 | if (group_dead) |
f6ec29a4 | 984 | acct_process(); |
1da177e4 LT |
985 | exit_sem(tsk); |
986 | __exit_files(tsk); | |
987 | __exit_fs(tsk); | |
e18eecb8 | 988 | check_stack_usage(); |
1da177e4 LT |
989 | exit_thread(); |
990 | cpuset_exit(tsk); | |
991 | exit_keys(tsk); | |
992 | ||
993 | if (group_dead && tsk->signal->leader) | |
994 | disassociate_ctty(1); | |
995 | ||
a1261f54 | 996 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 LT |
997 | if (tsk->binfmt) |
998 | module_put(tsk->binfmt->module); | |
999 | ||
1000 | tsk->exit_code = code; | |
9f46080c | 1001 | proc_exit_connector(tsk); |
444f378b | 1002 | exit_task_namespaces(tsk); |
0f245285 | 1003 | exit_notify(tsk); |
1da177e4 LT |
1004 | #ifdef CONFIG_NUMA |
1005 | mpol_free(tsk->mempolicy); | |
1006 | tsk->mempolicy = NULL; | |
1007 | #endif | |
c87e2837 IM |
1008 | /* |
1009 | * This must happen late, after the PID is not | |
1010 | * hashed anymore: | |
1011 | */ | |
1012 | if (unlikely(!list_empty(&tsk->pi_state_list))) | |
1013 | exit_pi_state_list(tsk); | |
1014 | if (unlikely(current->pi_state_cache)) | |
1015 | kfree(current->pi_state_cache); | |
de5097c2 | 1016 | /* |
9a11b49a | 1017 | * Make sure we are holding no locks: |
de5097c2 | 1018 | */ |
9a11b49a | 1019 | debug_check_no_locks_held(tsk); |
778e9a9c AK |
1020 | /* |
1021 | * We can do this unlocked here. The futex code uses this flag | |
1022 | * just to verify whether the pi state cleanup has been done | |
1023 | * or not. In the worst case it loops once more. | |
1024 | */ | |
1025 | tsk->flags |= PF_EXITPIDONE; | |
1da177e4 | 1026 | |
afc847b7 AV |
1027 | if (tsk->io_context) |
1028 | exit_io_context(); | |
1029 | ||
b92ce558 JA |
1030 | if (tsk->splice_pipe) |
1031 | __free_pipe_info(tsk->splice_pipe); | |
1032 | ||
7407251a | 1033 | preempt_disable(); |
55a101f8 | 1034 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 1035 | tsk->state = TASK_DEAD; |
7407251a | 1036 | |
1da177e4 LT |
1037 | schedule(); |
1038 | BUG(); | |
1039 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
1040 | for (;;) |
1041 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
1042 | } |
1043 | ||
012914da RA |
1044 | EXPORT_SYMBOL_GPL(do_exit); |
1045 | ||
1da177e4 LT |
1046 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
1047 | { | |
1048 | if (comp) | |
1049 | complete(comp); | |
55a101f8 | 1050 | |
1da177e4 LT |
1051 | do_exit(code); |
1052 | } | |
1053 | ||
1054 | EXPORT_SYMBOL(complete_and_exit); | |
1055 | ||
1056 | asmlinkage long sys_exit(int error_code) | |
1057 | { | |
1058 | do_exit((error_code&0xff)<<8); | |
1059 | } | |
1060 | ||
1da177e4 LT |
1061 | /* |
1062 | * Take down every thread in the group. This is called by fatal signals | |
1063 | * as well as by sys_exit_group (below). | |
1064 | */ | |
1065 | NORET_TYPE void | |
1066 | do_group_exit(int exit_code) | |
1067 | { | |
1068 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ | |
1069 | ||
1070 | if (current->signal->flags & SIGNAL_GROUP_EXIT) | |
1071 | exit_code = current->signal->group_exit_code; | |
1072 | else if (!thread_group_empty(current)) { | |
1073 | struct signal_struct *const sig = current->signal; | |
1074 | struct sighand_struct *const sighand = current->sighand; | |
1da177e4 LT |
1075 | spin_lock_irq(&sighand->siglock); |
1076 | if (sig->flags & SIGNAL_GROUP_EXIT) | |
1077 | /* Another thread got here before we took the lock. */ | |
1078 | exit_code = sig->group_exit_code; | |
1079 | else { | |
1da177e4 LT |
1080 | sig->group_exit_code = exit_code; |
1081 | zap_other_threads(current); | |
1082 | } | |
1083 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1084 | } |
1085 | ||
1086 | do_exit(exit_code); | |
1087 | /* NOTREACHED */ | |
1088 | } | |
1089 | ||
1090 | /* | |
1091 | * this kills every thread in the thread group. Note that any externally | |
1092 | * wait4()-ing process will get the correct exit code - even if this | |
1093 | * thread is not the thread group leader. | |
1094 | */ | |
1095 | asmlinkage void sys_exit_group(int error_code) | |
1096 | { | |
1097 | do_group_exit((error_code & 0xff) << 8); | |
1098 | } | |
1099 | ||
36c8b586 | 1100 | static int eligible_child(pid_t pid, int options, struct task_struct *p) |
1da177e4 | 1101 | { |
73243284 RM |
1102 | int err; |
1103 | ||
1da177e4 LT |
1104 | if (pid > 0) { |
1105 | if (p->pid != pid) | |
1106 | return 0; | |
1107 | } else if (!pid) { | |
1108 | if (process_group(p) != process_group(current)) | |
1109 | return 0; | |
1110 | } else if (pid != -1) { | |
1111 | if (process_group(p) != -pid) | |
1112 | return 0; | |
1113 | } | |
1114 | ||
1115 | /* | |
1116 | * Do not consider detached threads that are | |
1117 | * not ptraced: | |
1118 | */ | |
1119 | if (p->exit_signal == -1 && !p->ptrace) | |
1120 | return 0; | |
1121 | ||
1122 | /* Wait for all children (clone and not) if __WALL is set; | |
1123 | * otherwise, wait for clone children *only* if __WCLONE is | |
1124 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1125 | * A "clone" child here is one that reports to its parent | |
1126 | * using a signal other than SIGCHLD.) */ | |
1127 | if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0)) | |
1128 | && !(options & __WALL)) | |
1129 | return 0; | |
1130 | /* | |
1131 | * Do not consider thread group leaders that are | |
1132 | * in a non-empty thread group: | |
1133 | */ | |
3b6362b8 | 1134 | if (delay_group_leader(p)) |
1da177e4 LT |
1135 | return 2; |
1136 | ||
73243284 RM |
1137 | err = security_task_wait(p); |
1138 | if (err) | |
1139 | return err; | |
1da177e4 LT |
1140 | |
1141 | return 1; | |
1142 | } | |
1143 | ||
36c8b586 | 1144 | static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid, |
1da177e4 LT |
1145 | int why, int status, |
1146 | struct siginfo __user *infop, | |
1147 | struct rusage __user *rusagep) | |
1148 | { | |
1149 | int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0; | |
36c8b586 | 1150 | |
1da177e4 LT |
1151 | put_task_struct(p); |
1152 | if (!retval) | |
1153 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1154 | if (!retval) | |
1155 | retval = put_user(0, &infop->si_errno); | |
1156 | if (!retval) | |
1157 | retval = put_user((short)why, &infop->si_code); | |
1158 | if (!retval) | |
1159 | retval = put_user(pid, &infop->si_pid); | |
1160 | if (!retval) | |
1161 | retval = put_user(uid, &infop->si_uid); | |
1162 | if (!retval) | |
1163 | retval = put_user(status, &infop->si_status); | |
1164 | if (!retval) | |
1165 | retval = pid; | |
1166 | return retval; | |
1167 | } | |
1168 | ||
1169 | /* | |
1170 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1171 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1172 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1173 | * released the lock and the system call should return. | |
1174 | */ | |
36c8b586 | 1175 | static int wait_task_zombie(struct task_struct *p, int noreap, |
1da177e4 LT |
1176 | struct siginfo __user *infop, |
1177 | int __user *stat_addr, struct rusage __user *ru) | |
1178 | { | |
1179 | unsigned long state; | |
1180 | int retval; | |
1181 | int status; | |
1182 | ||
1183 | if (unlikely(noreap)) { | |
1184 | pid_t pid = p->pid; | |
1185 | uid_t uid = p->uid; | |
1186 | int exit_code = p->exit_code; | |
1187 | int why, status; | |
1188 | ||
1189 | if (unlikely(p->exit_state != EXIT_ZOMBIE)) | |
1190 | return 0; | |
1191 | if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) | |
1192 | return 0; | |
1193 | get_task_struct(p); | |
1194 | read_unlock(&tasklist_lock); | |
1195 | if ((exit_code & 0x7f) == 0) { | |
1196 | why = CLD_EXITED; | |
1197 | status = exit_code >> 8; | |
1198 | } else { | |
1199 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1200 | status = exit_code & 0x7f; | |
1201 | } | |
1202 | return wait_noreap_copyout(p, pid, uid, why, | |
1203 | status, infop, ru); | |
1204 | } | |
1205 | ||
1206 | /* | |
1207 | * Try to move the task's state to DEAD | |
1208 | * only one thread is allowed to do this: | |
1209 | */ | |
1210 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1211 | if (state != EXIT_ZOMBIE) { | |
1212 | BUG_ON(state != EXIT_DEAD); | |
1213 | return 0; | |
1214 | } | |
1215 | if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) { | |
1216 | /* | |
1217 | * This can only happen in a race with a ptraced thread | |
1218 | * dying on another processor. | |
1219 | */ | |
1220 | return 0; | |
1221 | } | |
1222 | ||
1223 | if (likely(p->real_parent == p->parent) && likely(p->signal)) { | |
3795e161 JJ |
1224 | struct signal_struct *psig; |
1225 | struct signal_struct *sig; | |
1226 | ||
1da177e4 LT |
1227 | /* |
1228 | * The resource counters for the group leader are in its | |
1229 | * own task_struct. Those for dead threads in the group | |
1230 | * are in its signal_struct, as are those for the child | |
1231 | * processes it has previously reaped. All these | |
1232 | * accumulate in the parent's signal_struct c* fields. | |
1233 | * | |
1234 | * We don't bother to take a lock here to protect these | |
1235 | * p->signal fields, because they are only touched by | |
1236 | * __exit_signal, which runs with tasklist_lock | |
1237 | * write-locked anyway, and so is excluded here. We do | |
1238 | * need to protect the access to p->parent->signal fields, | |
1239 | * as other threads in the parent group can be right | |
1240 | * here reaping other children at the same time. | |
1241 | */ | |
1242 | spin_lock_irq(&p->parent->sighand->siglock); | |
3795e161 JJ |
1243 | psig = p->parent->signal; |
1244 | sig = p->signal; | |
1245 | psig->cutime = | |
1246 | cputime_add(psig->cutime, | |
1da177e4 | 1247 | cputime_add(p->utime, |
3795e161 JJ |
1248 | cputime_add(sig->utime, |
1249 | sig->cutime))); | |
1250 | psig->cstime = | |
1251 | cputime_add(psig->cstime, | |
1da177e4 | 1252 | cputime_add(p->stime, |
3795e161 JJ |
1253 | cputime_add(sig->stime, |
1254 | sig->cstime))); | |
1255 | psig->cmin_flt += | |
1256 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1257 | psig->cmaj_flt += | |
1258 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1259 | psig->cnvcsw += | |
1260 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1261 | psig->cnivcsw += | |
1262 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1263 | psig->cinblock += |
1264 | task_io_get_inblock(p) + | |
1265 | sig->inblock + sig->cinblock; | |
1266 | psig->coublock += | |
1267 | task_io_get_oublock(p) + | |
1268 | sig->oublock + sig->coublock; | |
1da177e4 LT |
1269 | spin_unlock_irq(&p->parent->sighand->siglock); |
1270 | } | |
1271 | ||
1272 | /* | |
1273 | * Now we are sure this task is interesting, and no other | |
1274 | * thread can reap it because we set its state to EXIT_DEAD. | |
1275 | */ | |
1276 | read_unlock(&tasklist_lock); | |
1277 | ||
1278 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1279 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) | |
1280 | ? p->signal->group_exit_code : p->exit_code; | |
1281 | if (!retval && stat_addr) | |
1282 | retval = put_user(status, stat_addr); | |
1283 | if (!retval && infop) | |
1284 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1285 | if (!retval && infop) | |
1286 | retval = put_user(0, &infop->si_errno); | |
1287 | if (!retval && infop) { | |
1288 | int why; | |
1289 | ||
1290 | if ((status & 0x7f) == 0) { | |
1291 | why = CLD_EXITED; | |
1292 | status >>= 8; | |
1293 | } else { | |
1294 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1295 | status &= 0x7f; | |
1296 | } | |
1297 | retval = put_user((short)why, &infop->si_code); | |
1298 | if (!retval) | |
1299 | retval = put_user(status, &infop->si_status); | |
1300 | } | |
1301 | if (!retval && infop) | |
1302 | retval = put_user(p->pid, &infop->si_pid); | |
1303 | if (!retval && infop) | |
1304 | retval = put_user(p->uid, &infop->si_uid); | |
1305 | if (retval) { | |
1306 | // TODO: is this safe? | |
1307 | p->exit_state = EXIT_ZOMBIE; | |
1308 | return retval; | |
1309 | } | |
1310 | retval = p->pid; | |
1311 | if (p->real_parent != p->parent) { | |
1312 | write_lock_irq(&tasklist_lock); | |
1313 | /* Double-check with lock held. */ | |
1314 | if (p->real_parent != p->parent) { | |
1315 | __ptrace_unlink(p); | |
1316 | // TODO: is this safe? | |
1317 | p->exit_state = EXIT_ZOMBIE; | |
1318 | /* | |
1319 | * If this is not a detached task, notify the parent. | |
1320 | * If it's still not detached after that, don't release | |
1321 | * it now. | |
1322 | */ | |
1323 | if (p->exit_signal != -1) { | |
1324 | do_notify_parent(p, p->exit_signal); | |
1325 | if (p->exit_signal != -1) | |
1326 | p = NULL; | |
1327 | } | |
1328 | } | |
1329 | write_unlock_irq(&tasklist_lock); | |
1330 | } | |
1331 | if (p != NULL) | |
1332 | release_task(p); | |
1333 | BUG_ON(!retval); | |
1334 | return retval; | |
1335 | } | |
1336 | ||
1337 | /* | |
1338 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1339 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1340 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1341 | * released the lock and the system call should return. | |
1342 | */ | |
36c8b586 IM |
1343 | static int wait_task_stopped(struct task_struct *p, int delayed_group_leader, |
1344 | int noreap, struct siginfo __user *infop, | |
1da177e4 LT |
1345 | int __user *stat_addr, struct rusage __user *ru) |
1346 | { | |
1347 | int retval, exit_code; | |
1348 | ||
1349 | if (!p->exit_code) | |
1350 | return 0; | |
1351 | if (delayed_group_leader && !(p->ptrace & PT_PTRACED) && | |
1352 | p->signal && p->signal->group_stop_count > 0) | |
1353 | /* | |
1354 | * A group stop is in progress and this is the group leader. | |
1355 | * We won't report until all threads have stopped. | |
1356 | */ | |
1357 | return 0; | |
1358 | ||
1359 | /* | |
1360 | * Now we are pretty sure this task is interesting. | |
1361 | * Make sure it doesn't get reaped out from under us while we | |
1362 | * give up the lock and then examine it below. We don't want to | |
1363 | * keep holding onto the tasklist_lock while we call getrusage and | |
1364 | * possibly take page faults for user memory. | |
1365 | */ | |
1366 | get_task_struct(p); | |
1367 | read_unlock(&tasklist_lock); | |
1368 | ||
1369 | if (unlikely(noreap)) { | |
1370 | pid_t pid = p->pid; | |
1371 | uid_t uid = p->uid; | |
1372 | int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED; | |
1373 | ||
1374 | exit_code = p->exit_code; | |
1375 | if (unlikely(!exit_code) || | |
14bf01bb | 1376 | unlikely(p->state & TASK_TRACED)) |
1da177e4 LT |
1377 | goto bail_ref; |
1378 | return wait_noreap_copyout(p, pid, uid, | |
1379 | why, (exit_code << 8) | 0x7f, | |
1380 | infop, ru); | |
1381 | } | |
1382 | ||
1383 | write_lock_irq(&tasklist_lock); | |
1384 | ||
1385 | /* | |
1386 | * This uses xchg to be atomic with the thread resuming and setting | |
1387 | * it. It must also be done with the write lock held to prevent a | |
1388 | * race with the EXIT_ZOMBIE case. | |
1389 | */ | |
1390 | exit_code = xchg(&p->exit_code, 0); | |
1391 | if (unlikely(p->exit_state)) { | |
1392 | /* | |
1393 | * The task resumed and then died. Let the next iteration | |
1394 | * catch it in EXIT_ZOMBIE. Note that exit_code might | |
1395 | * already be zero here if it resumed and did _exit(0). | |
1396 | * The task itself is dead and won't touch exit_code again; | |
1397 | * other processors in this function are locked out. | |
1398 | */ | |
1399 | p->exit_code = exit_code; | |
1400 | exit_code = 0; | |
1401 | } | |
1402 | if (unlikely(exit_code == 0)) { | |
1403 | /* | |
1404 | * Another thread in this function got to it first, or it | |
1405 | * resumed, or it resumed and then died. | |
1406 | */ | |
1407 | write_unlock_irq(&tasklist_lock); | |
1408 | bail_ref: | |
1409 | put_task_struct(p); | |
1410 | /* | |
1411 | * We are returning to the wait loop without having successfully | |
1412 | * removed the process and having released the lock. We cannot | |
1413 | * continue, since the "p" task pointer is potentially stale. | |
1414 | * | |
1415 | * Return -EAGAIN, and do_wait() will restart the loop from the | |
1416 | * beginning. Do _not_ re-acquire the lock. | |
1417 | */ | |
1418 | return -EAGAIN; | |
1419 | } | |
1420 | ||
1421 | /* move to end of parent's list to avoid starvation */ | |
1422 | remove_parent(p); | |
8fafabd8 | 1423 | add_parent(p); |
1da177e4 LT |
1424 | |
1425 | write_unlock_irq(&tasklist_lock); | |
1426 | ||
1427 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1428 | if (!retval && stat_addr) | |
1429 | retval = put_user((exit_code << 8) | 0x7f, stat_addr); | |
1430 | if (!retval && infop) | |
1431 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1432 | if (!retval && infop) | |
1433 | retval = put_user(0, &infop->si_errno); | |
1434 | if (!retval && infop) | |
1435 | retval = put_user((short)((p->ptrace & PT_PTRACED) | |
1436 | ? CLD_TRAPPED : CLD_STOPPED), | |
1437 | &infop->si_code); | |
1438 | if (!retval && infop) | |
1439 | retval = put_user(exit_code, &infop->si_status); | |
1440 | if (!retval && infop) | |
1441 | retval = put_user(p->pid, &infop->si_pid); | |
1442 | if (!retval && infop) | |
1443 | retval = put_user(p->uid, &infop->si_uid); | |
1444 | if (!retval) | |
1445 | retval = p->pid; | |
1446 | put_task_struct(p); | |
1447 | ||
1448 | BUG_ON(!retval); | |
1449 | return retval; | |
1450 | } | |
1451 | ||
1452 | /* | |
1453 | * Handle do_wait work for one task in a live, non-stopped state. | |
1454 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1455 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1456 | * released the lock and the system call should return. | |
1457 | */ | |
36c8b586 | 1458 | static int wait_task_continued(struct task_struct *p, int noreap, |
1da177e4 LT |
1459 | struct siginfo __user *infop, |
1460 | int __user *stat_addr, struct rusage __user *ru) | |
1461 | { | |
1462 | int retval; | |
1463 | pid_t pid; | |
1464 | uid_t uid; | |
1465 | ||
1466 | if (unlikely(!p->signal)) | |
1467 | return 0; | |
1468 | ||
1469 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) | |
1470 | return 0; | |
1471 | ||
1472 | spin_lock_irq(&p->sighand->siglock); | |
1473 | /* Re-check with the lock held. */ | |
1474 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1475 | spin_unlock_irq(&p->sighand->siglock); | |
1476 | return 0; | |
1477 | } | |
1478 | if (!noreap) | |
1479 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; | |
1480 | spin_unlock_irq(&p->sighand->siglock); | |
1481 | ||
1482 | pid = p->pid; | |
1483 | uid = p->uid; | |
1484 | get_task_struct(p); | |
1485 | read_unlock(&tasklist_lock); | |
1486 | ||
1487 | if (!infop) { | |
1488 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1489 | put_task_struct(p); | |
1490 | if (!retval && stat_addr) | |
1491 | retval = put_user(0xffff, stat_addr); | |
1492 | if (!retval) | |
1493 | retval = p->pid; | |
1494 | } else { | |
1495 | retval = wait_noreap_copyout(p, pid, uid, | |
1496 | CLD_CONTINUED, SIGCONT, | |
1497 | infop, ru); | |
1498 | BUG_ON(retval == 0); | |
1499 | } | |
1500 | ||
1501 | return retval; | |
1502 | } | |
1503 | ||
1504 | ||
1505 | static inline int my_ptrace_child(struct task_struct *p) | |
1506 | { | |
1507 | if (!(p->ptrace & PT_PTRACED)) | |
1508 | return 0; | |
1509 | if (!(p->ptrace & PT_ATTACHED)) | |
1510 | return 1; | |
1511 | /* | |
1512 | * This child was PTRACE_ATTACH'd. We should be seeing it only if | |
1513 | * we are the attacher. If we are the real parent, this is a race | |
1514 | * inside ptrace_attach. It is waiting for the tasklist_lock, | |
1515 | * which we have to switch the parent links, but has already set | |
1516 | * the flags in p->ptrace. | |
1517 | */ | |
1518 | return (p->parent != p->real_parent); | |
1519 | } | |
1520 | ||
1521 | static long do_wait(pid_t pid, int options, struct siginfo __user *infop, | |
1522 | int __user *stat_addr, struct rusage __user *ru) | |
1523 | { | |
1524 | DECLARE_WAITQUEUE(wait, current); | |
1525 | struct task_struct *tsk; | |
1526 | int flag, retval; | |
73243284 | 1527 | int allowed, denied; |
1da177e4 LT |
1528 | |
1529 | add_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1530 | repeat: | |
1531 | /* | |
1532 | * We will set this flag if we see any child that might later | |
1533 | * match our criteria, even if we are not able to reap it yet. | |
1534 | */ | |
1535 | flag = 0; | |
73243284 | 1536 | allowed = denied = 0; |
1da177e4 LT |
1537 | current->state = TASK_INTERRUPTIBLE; |
1538 | read_lock(&tasklist_lock); | |
1539 | tsk = current; | |
1540 | do { | |
1541 | struct task_struct *p; | |
1542 | struct list_head *_p; | |
1543 | int ret; | |
1544 | ||
1545 | list_for_each(_p,&tsk->children) { | |
36c8b586 | 1546 | p = list_entry(_p, struct task_struct, sibling); |
1da177e4 LT |
1547 | |
1548 | ret = eligible_child(pid, options, p); | |
1549 | if (!ret) | |
1550 | continue; | |
1551 | ||
73243284 RM |
1552 | if (unlikely(ret < 0)) { |
1553 | denied = ret; | |
1554 | continue; | |
1555 | } | |
1556 | allowed = 1; | |
1557 | ||
1da177e4 LT |
1558 | switch (p->state) { |
1559 | case TASK_TRACED: | |
7f2a5255 RM |
1560 | /* |
1561 | * When we hit the race with PTRACE_ATTACH, | |
1562 | * we will not report this child. But the | |
1563 | * race means it has not yet been moved to | |
1564 | * our ptrace_children list, so we need to | |
1565 | * set the flag here to avoid a spurious ECHILD | |
1566 | * when the race happens with the only child. | |
1567 | */ | |
1568 | flag = 1; | |
1da177e4 LT |
1569 | if (!my_ptrace_child(p)) |
1570 | continue; | |
1571 | /*FALLTHROUGH*/ | |
1572 | case TASK_STOPPED: | |
1573 | /* | |
1574 | * It's stopped now, so it might later | |
1575 | * continue, exit, or stop again. | |
1576 | */ | |
1577 | flag = 1; | |
1578 | if (!(options & WUNTRACED) && | |
1579 | !my_ptrace_child(p)) | |
1580 | continue; | |
1581 | retval = wait_task_stopped(p, ret == 2, | |
1582 | (options & WNOWAIT), | |
1583 | infop, | |
1584 | stat_addr, ru); | |
1585 | if (retval == -EAGAIN) | |
1586 | goto repeat; | |
1587 | if (retval != 0) /* He released the lock. */ | |
1588 | goto end; | |
1589 | break; | |
1590 | default: | |
1591 | // case EXIT_DEAD: | |
1592 | if (p->exit_state == EXIT_DEAD) | |
1593 | continue; | |
1594 | // case EXIT_ZOMBIE: | |
1595 | if (p->exit_state == EXIT_ZOMBIE) { | |
1596 | /* | |
1597 | * Eligible but we cannot release | |
1598 | * it yet: | |
1599 | */ | |
1600 | if (ret == 2) | |
1601 | goto check_continued; | |
1602 | if (!likely(options & WEXITED)) | |
1603 | continue; | |
1604 | retval = wait_task_zombie( | |
1605 | p, (options & WNOWAIT), | |
1606 | infop, stat_addr, ru); | |
1607 | /* He released the lock. */ | |
1608 | if (retval != 0) | |
1609 | goto end; | |
1610 | break; | |
1611 | } | |
1612 | check_continued: | |
1613 | /* | |
1614 | * It's running now, so it might later | |
1615 | * exit, stop, or stop and then continue. | |
1616 | */ | |
1617 | flag = 1; | |
1618 | if (!unlikely(options & WCONTINUED)) | |
1619 | continue; | |
1620 | retval = wait_task_continued( | |
1621 | p, (options & WNOWAIT), | |
1622 | infop, stat_addr, ru); | |
1623 | if (retval != 0) /* He released the lock. */ | |
1624 | goto end; | |
1625 | break; | |
1626 | } | |
1627 | } | |
1628 | if (!flag) { | |
1629 | list_for_each(_p, &tsk->ptrace_children) { | |
1630 | p = list_entry(_p, struct task_struct, | |
1631 | ptrace_list); | |
1632 | if (!eligible_child(pid, options, p)) | |
1633 | continue; | |
1634 | flag = 1; | |
1635 | break; | |
1636 | } | |
1637 | } | |
1638 | if (options & __WNOTHREAD) | |
1639 | break; | |
1640 | tsk = next_thread(tsk); | |
125e1874 | 1641 | BUG_ON(tsk->signal != current->signal); |
1da177e4 LT |
1642 | } while (tsk != current); |
1643 | ||
1644 | read_unlock(&tasklist_lock); | |
1645 | if (flag) { | |
1646 | retval = 0; | |
1647 | if (options & WNOHANG) | |
1648 | goto end; | |
1649 | retval = -ERESTARTSYS; | |
1650 | if (signal_pending(current)) | |
1651 | goto end; | |
1652 | schedule(); | |
1653 | goto repeat; | |
1654 | } | |
1655 | retval = -ECHILD; | |
73243284 RM |
1656 | if (unlikely(denied) && !allowed) |
1657 | retval = denied; | |
1da177e4 LT |
1658 | end: |
1659 | current->state = TASK_RUNNING; | |
1660 | remove_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1661 | if (infop) { | |
1662 | if (retval > 0) | |
1663 | retval = 0; | |
1664 | else { | |
1665 | /* | |
1666 | * For a WNOHANG return, clear out all the fields | |
1667 | * we would set so the user can easily tell the | |
1668 | * difference. | |
1669 | */ | |
1670 | if (!retval) | |
1671 | retval = put_user(0, &infop->si_signo); | |
1672 | if (!retval) | |
1673 | retval = put_user(0, &infop->si_errno); | |
1674 | if (!retval) | |
1675 | retval = put_user(0, &infop->si_code); | |
1676 | if (!retval) | |
1677 | retval = put_user(0, &infop->si_pid); | |
1678 | if (!retval) | |
1679 | retval = put_user(0, &infop->si_uid); | |
1680 | if (!retval) | |
1681 | retval = put_user(0, &infop->si_status); | |
1682 | } | |
1683 | } | |
1684 | return retval; | |
1685 | } | |
1686 | ||
1687 | asmlinkage long sys_waitid(int which, pid_t pid, | |
1688 | struct siginfo __user *infop, int options, | |
1689 | struct rusage __user *ru) | |
1690 | { | |
1691 | long ret; | |
1692 | ||
1693 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1694 | return -EINVAL; | |
1695 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1696 | return -EINVAL; | |
1697 | ||
1698 | switch (which) { | |
1699 | case P_ALL: | |
1700 | pid = -1; | |
1701 | break; | |
1702 | case P_PID: | |
1703 | if (pid <= 0) | |
1704 | return -EINVAL; | |
1705 | break; | |
1706 | case P_PGID: | |
1707 | if (pid <= 0) | |
1708 | return -EINVAL; | |
1709 | pid = -pid; | |
1710 | break; | |
1711 | default: | |
1712 | return -EINVAL; | |
1713 | } | |
1714 | ||
1715 | ret = do_wait(pid, options, infop, NULL, ru); | |
1716 | ||
1717 | /* avoid REGPARM breakage on x86: */ | |
1718 | prevent_tail_call(ret); | |
1719 | return ret; | |
1720 | } | |
1721 | ||
1722 | asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr, | |
1723 | int options, struct rusage __user *ru) | |
1724 | { | |
1725 | long ret; | |
1726 | ||
1727 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1728 | __WNOTHREAD|__WCLONE|__WALL)) | |
1729 | return -EINVAL; | |
1730 | ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru); | |
1731 | ||
1732 | /* avoid REGPARM breakage on x86: */ | |
1733 | prevent_tail_call(ret); | |
1734 | return ret; | |
1735 | } | |
1736 | ||
1737 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1738 | ||
1739 | /* | |
1740 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1741 | * implemented by calling sys_wait4() from libc.a. | |
1742 | */ | |
1743 | asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options) | |
1744 | { | |
1745 | return sys_wait4(pid, stat_addr, options, NULL); | |
1746 | } | |
1747 | ||
1748 | #endif |