]>
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> |
1da177e4 LT |
34 | #include <linux/cpuset.h> |
35 | #include <linux/syscalls.h> | |
7ed20e1a | 36 | #include <linux/signal.h> |
6a14c5c9 | 37 | #include <linux/posix-timers.h> |
9f46080c | 38 | #include <linux/cn_proc.h> |
de5097c2 | 39 | #include <linux/mutex.h> |
0771dfef | 40 | #include <linux/futex.h> |
34f192c6 | 41 | #include <linux/compat.h> |
b92ce558 | 42 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 43 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 44 | #include <linux/resource.h> |
0d67a46d | 45 | #include <linux/blkdev.h> |
6eaeeaba | 46 | #include <linux/task_io_accounting_ops.h> |
1da177e4 LT |
47 | |
48 | #include <asm/uaccess.h> | |
49 | #include <asm/unistd.h> | |
50 | #include <asm/pgtable.h> | |
51 | #include <asm/mmu_context.h> | |
52 | ||
53 | extern void sem_exit (void); | |
1da177e4 | 54 | |
408b664a AB |
55 | static void exit_mm(struct task_struct * tsk); |
56 | ||
1da177e4 LT |
57 | static void __unhash_process(struct task_struct *p) |
58 | { | |
59 | nr_threads--; | |
60 | detach_pid(p, PIDTYPE_PID); | |
1da177e4 LT |
61 | if (thread_group_leader(p)) { |
62 | detach_pid(p, PIDTYPE_PGID); | |
63 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 64 | |
5e85d4ab | 65 | list_del_rcu(&p->tasks); |
73b9ebfe | 66 | __get_cpu_var(process_counts)--; |
1da177e4 | 67 | } |
47e65328 | 68 | list_del_rcu(&p->thread_group); |
c97d9893 | 69 | remove_parent(p); |
1da177e4 LT |
70 | } |
71 | ||
6a14c5c9 ON |
72 | /* |
73 | * This function expects the tasklist_lock write-locked. | |
74 | */ | |
75 | static void __exit_signal(struct task_struct *tsk) | |
76 | { | |
77 | struct signal_struct *sig = tsk->signal; | |
78 | struct sighand_struct *sighand; | |
79 | ||
80 | BUG_ON(!sig); | |
81 | BUG_ON(!atomic_read(&sig->count)); | |
82 | ||
83 | rcu_read_lock(); | |
84 | sighand = rcu_dereference(tsk->sighand); | |
85 | spin_lock(&sighand->siglock); | |
86 | ||
fba2afaa DL |
87 | /* |
88 | * Notify that this sighand has been detached. This must | |
89 | * be called with the tsk->sighand lock held. Also, this | |
90 | * access tsk->sighand internally, so it must be called | |
91 | * before tsk->sighand is reset. | |
92 | */ | |
93 | signalfd_detach_locked(tsk); | |
94 | ||
6a14c5c9 ON |
95 | posix_cpu_timers_exit(tsk); |
96 | if (atomic_dec_and_test(&sig->count)) | |
97 | posix_cpu_timers_exit_group(tsk); | |
98 | else { | |
99 | /* | |
100 | * If there is any task waiting for the group exit | |
101 | * then notify it: | |
102 | */ | |
103 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) { | |
104 | wake_up_process(sig->group_exit_task); | |
105 | sig->group_exit_task = NULL; | |
106 | } | |
107 | if (tsk == sig->curr_target) | |
108 | sig->curr_target = next_thread(tsk); | |
109 | /* | |
110 | * Accumulate here the counters for all threads but the | |
111 | * group leader as they die, so they can be added into | |
112 | * the process-wide totals when those are taken. | |
113 | * The group leader stays around as a zombie as long | |
114 | * as there are other threads. When it gets reaped, | |
115 | * the exit.c code will add its counts into these totals. | |
116 | * We won't ever get here for the group leader, since it | |
117 | * will have been the last reference on the signal_struct. | |
118 | */ | |
119 | sig->utime = cputime_add(sig->utime, tsk->utime); | |
120 | sig->stime = cputime_add(sig->stime, tsk->stime); | |
121 | sig->min_flt += tsk->min_flt; | |
122 | sig->maj_flt += tsk->maj_flt; | |
123 | sig->nvcsw += tsk->nvcsw; | |
124 | sig->nivcsw += tsk->nivcsw; | |
125 | sig->sched_time += tsk->sched_time; | |
6eaeeaba ED |
126 | sig->inblock += task_io_get_inblock(tsk); |
127 | sig->oublock += task_io_get_oublock(tsk); | |
6a14c5c9 ON |
128 | sig = NULL; /* Marker for below. */ |
129 | } | |
130 | ||
5876700c ON |
131 | __unhash_process(tsk); |
132 | ||
6a14c5c9 | 133 | tsk->signal = NULL; |
a7e5328a | 134 | tsk->sighand = NULL; |
6a14c5c9 ON |
135 | spin_unlock(&sighand->siglock); |
136 | rcu_read_unlock(); | |
137 | ||
a7e5328a | 138 | __cleanup_sighand(sighand); |
6a14c5c9 ON |
139 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
140 | flush_sigqueue(&tsk->pending); | |
141 | if (sig) { | |
142 | flush_sigqueue(&sig->shared_pending); | |
093a8e8a | 143 | taskstats_tgid_free(sig); |
6a14c5c9 ON |
144 | __cleanup_signal(sig); |
145 | } | |
146 | } | |
147 | ||
8c7904a0 EB |
148 | static void delayed_put_task_struct(struct rcu_head *rhp) |
149 | { | |
150 | put_task_struct(container_of(rhp, struct task_struct, rcu)); | |
151 | } | |
152 | ||
1da177e4 LT |
153 | void release_task(struct task_struct * p) |
154 | { | |
36c8b586 | 155 | struct task_struct *leader; |
1da177e4 | 156 | int zap_leader; |
1f09f974 | 157 | repeat: |
1da177e4 | 158 | atomic_dec(&p->user->processes); |
1da177e4 | 159 | write_lock_irq(&tasklist_lock); |
1f09f974 | 160 | ptrace_unlink(p); |
1da177e4 LT |
161 | BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children)); |
162 | __exit_signal(p); | |
35f5cad8 | 163 | |
1da177e4 LT |
164 | /* |
165 | * If we are the last non-leader member of the thread | |
166 | * group, and the leader is zombie, then notify the | |
167 | * group leader's parent process. (if it wants notification.) | |
168 | */ | |
169 | zap_leader = 0; | |
170 | leader = p->group_leader; | |
171 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
172 | BUG_ON(leader->exit_signal == -1); | |
173 | do_notify_parent(leader, leader->exit_signal); | |
174 | /* | |
175 | * If we were the last child thread and the leader has | |
176 | * exited already, and the leader's parent ignores SIGCHLD, | |
177 | * then we are the one who should release the leader. | |
178 | * | |
179 | * do_notify_parent() will have marked it self-reaping in | |
180 | * that case. | |
181 | */ | |
182 | zap_leader = (leader->exit_signal == -1); | |
183 | } | |
184 | ||
185 | sched_exit(p); | |
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 | ||
1c573afe | 294 | if (!has_rt_policy(current) && (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); | |
391 | ||
392 | set_special_pids(1, 1); | |
24ec839c | 393 | proc_clear_tty(current); |
1da177e4 LT |
394 | |
395 | /* Block and flush all signals */ | |
396 | sigfillset(&blocked); | |
397 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
398 | flush_signals(current); | |
399 | ||
400 | /* Become as one with the init task */ | |
401 | ||
402 | exit_fs(current); /* current->fs->count--; */ | |
403 | fs = init_task.fs; | |
404 | current->fs = fs; | |
405 | atomic_inc(&fs->count); | |
ab516013 | 406 | |
444f378b | 407 | exit_task_namespaces(current); |
ab516013 | 408 | current->nsproxy = init_task.nsproxy; |
ab516013 SH |
409 | get_task_namespaces(current); |
410 | ||
1da177e4 LT |
411 | exit_files(current); |
412 | current->files = init_task.files; | |
413 | atomic_inc(¤t->files->count); | |
414 | ||
49d769d5 | 415 | reparent_to_kthreadd(); |
1da177e4 LT |
416 | } |
417 | ||
418 | EXPORT_SYMBOL(daemonize); | |
419 | ||
858119e1 | 420 | static void close_files(struct files_struct * files) |
1da177e4 LT |
421 | { |
422 | int i, j; | |
badf1662 | 423 | struct fdtable *fdt; |
1da177e4 LT |
424 | |
425 | j = 0; | |
4fb3a538 DS |
426 | |
427 | /* | |
428 | * It is safe to dereference the fd table without RCU or | |
429 | * ->file_lock because this is the last reference to the | |
430 | * files structure. | |
431 | */ | |
badf1662 | 432 | fdt = files_fdtable(files); |
1da177e4 LT |
433 | for (;;) { |
434 | unsigned long set; | |
435 | i = j * __NFDBITS; | |
bbea9f69 | 436 | if (i >= fdt->max_fds) |
1da177e4 | 437 | break; |
badf1662 | 438 | set = fdt->open_fds->fds_bits[j++]; |
1da177e4 LT |
439 | while (set) { |
440 | if (set & 1) { | |
badf1662 | 441 | struct file * file = xchg(&fdt->fd[i], NULL); |
944be0b2 | 442 | if (file) { |
1da177e4 | 443 | filp_close(file, files); |
944be0b2 IM |
444 | cond_resched(); |
445 | } | |
1da177e4 LT |
446 | } |
447 | i++; | |
448 | set >>= 1; | |
449 | } | |
450 | } | |
451 | } | |
452 | ||
453 | struct files_struct *get_files_struct(struct task_struct *task) | |
454 | { | |
455 | struct files_struct *files; | |
456 | ||
457 | task_lock(task); | |
458 | files = task->files; | |
459 | if (files) | |
460 | atomic_inc(&files->count); | |
461 | task_unlock(task); | |
462 | ||
463 | return files; | |
464 | } | |
465 | ||
466 | void fastcall put_files_struct(struct files_struct *files) | |
467 | { | |
badf1662 DS |
468 | struct fdtable *fdt; |
469 | ||
1da177e4 LT |
470 | if (atomic_dec_and_test(&files->count)) { |
471 | close_files(files); | |
472 | /* | |
473 | * Free the fd and fdset arrays if we expanded them. | |
ab2af1f5 DS |
474 | * If the fdtable was embedded, pass files for freeing |
475 | * at the end of the RCU grace period. Otherwise, | |
476 | * you can free files immediately. | |
1da177e4 | 477 | */ |
badf1662 | 478 | fdt = files_fdtable(files); |
4fd45812 | 479 | if (fdt != &files->fdtab) |
ab2af1f5 | 480 | kmem_cache_free(files_cachep, files); |
01b2d93c | 481 | free_fdtable(fdt); |
1da177e4 LT |
482 | } |
483 | } | |
484 | ||
485 | EXPORT_SYMBOL(put_files_struct); | |
486 | ||
3b9b8ab6 KK |
487 | void reset_files_struct(struct task_struct *tsk, struct files_struct *files) |
488 | { | |
489 | struct files_struct *old; | |
490 | ||
491 | old = tsk->files; | |
492 | task_lock(tsk); | |
493 | tsk->files = files; | |
494 | task_unlock(tsk); | |
495 | put_files_struct(old); | |
496 | } | |
497 | EXPORT_SYMBOL(reset_files_struct); | |
498 | ||
1da177e4 LT |
499 | static inline void __exit_files(struct task_struct *tsk) |
500 | { | |
501 | struct files_struct * files = tsk->files; | |
502 | ||
503 | if (files) { | |
504 | task_lock(tsk); | |
505 | tsk->files = NULL; | |
506 | task_unlock(tsk); | |
507 | put_files_struct(files); | |
508 | } | |
509 | } | |
510 | ||
511 | void exit_files(struct task_struct *tsk) | |
512 | { | |
513 | __exit_files(tsk); | |
514 | } | |
515 | ||
516 | static inline void __put_fs_struct(struct fs_struct *fs) | |
517 | { | |
518 | /* No need to hold fs->lock if we are killing it */ | |
519 | if (atomic_dec_and_test(&fs->count)) { | |
520 | dput(fs->root); | |
521 | mntput(fs->rootmnt); | |
522 | dput(fs->pwd); | |
523 | mntput(fs->pwdmnt); | |
524 | if (fs->altroot) { | |
525 | dput(fs->altroot); | |
526 | mntput(fs->altrootmnt); | |
527 | } | |
528 | kmem_cache_free(fs_cachep, fs); | |
529 | } | |
530 | } | |
531 | ||
532 | void put_fs_struct(struct fs_struct *fs) | |
533 | { | |
534 | __put_fs_struct(fs); | |
535 | } | |
536 | ||
537 | static inline void __exit_fs(struct task_struct *tsk) | |
538 | { | |
539 | struct fs_struct * fs = tsk->fs; | |
540 | ||
541 | if (fs) { | |
542 | task_lock(tsk); | |
543 | tsk->fs = NULL; | |
544 | task_unlock(tsk); | |
545 | __put_fs_struct(fs); | |
546 | } | |
547 | } | |
548 | ||
549 | void exit_fs(struct task_struct *tsk) | |
550 | { | |
551 | __exit_fs(tsk); | |
552 | } | |
553 | ||
554 | EXPORT_SYMBOL_GPL(exit_fs); | |
555 | ||
556 | /* | |
557 | * Turn us into a lazy TLB process if we | |
558 | * aren't already.. | |
559 | */ | |
408b664a | 560 | static void exit_mm(struct task_struct * tsk) |
1da177e4 LT |
561 | { |
562 | struct mm_struct *mm = tsk->mm; | |
563 | ||
564 | mm_release(tsk, mm); | |
565 | if (!mm) | |
566 | return; | |
567 | /* | |
568 | * Serialize with any possible pending coredump. | |
569 | * We must hold mmap_sem around checking core_waiters | |
570 | * and clearing tsk->mm. The core-inducing thread | |
571 | * will increment core_waiters for each thread in the | |
572 | * group with ->mm != NULL. | |
573 | */ | |
574 | down_read(&mm->mmap_sem); | |
575 | if (mm->core_waiters) { | |
576 | up_read(&mm->mmap_sem); | |
577 | down_write(&mm->mmap_sem); | |
578 | if (!--mm->core_waiters) | |
579 | complete(mm->core_startup_done); | |
580 | up_write(&mm->mmap_sem); | |
581 | ||
582 | wait_for_completion(&mm->core_done); | |
583 | down_read(&mm->mmap_sem); | |
584 | } | |
585 | atomic_inc(&mm->mm_count); | |
125e1874 | 586 | BUG_ON(mm != tsk->active_mm); |
1da177e4 LT |
587 | /* more a memory barrier than a real lock */ |
588 | task_lock(tsk); | |
589 | tsk->mm = NULL; | |
590 | up_read(&mm->mmap_sem); | |
591 | enter_lazy_tlb(mm, current); | |
592 | task_unlock(tsk); | |
593 | mmput(mm); | |
594 | } | |
595 | ||
36c8b586 IM |
596 | static inline void |
597 | choose_new_parent(struct task_struct *p, struct task_struct *reaper) | |
1da177e4 LT |
598 | { |
599 | /* | |
600 | * Make sure we're not reparenting to ourselves and that | |
601 | * the parent is not a zombie. | |
602 | */ | |
d799f035 | 603 | BUG_ON(p == reaper || reaper->exit_state); |
1da177e4 | 604 | p->real_parent = reaper; |
1da177e4 LT |
605 | } |
606 | ||
36c8b586 IM |
607 | static void |
608 | reparent_thread(struct task_struct *p, struct task_struct *father, int traced) | |
1da177e4 | 609 | { |
241ceee0 ON |
610 | if (p->pdeath_signal) |
611 | /* We already hold the tasklist_lock here. */ | |
612 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | |
613 | ||
1da177e4 LT |
614 | /* Move the child from its dying parent to the new one. */ |
615 | if (unlikely(traced)) { | |
616 | /* Preserve ptrace links if someone else is tracing this child. */ | |
617 | list_del_init(&p->ptrace_list); | |
618 | if (p->parent != p->real_parent) | |
619 | list_add(&p->ptrace_list, &p->real_parent->ptrace_children); | |
620 | } else { | |
621 | /* If this child is being traced, then we're the one tracing it | |
622 | * anyway, so let go of it. | |
623 | */ | |
624 | p->ptrace = 0; | |
6ac781b1 | 625 | remove_parent(p); |
1da177e4 | 626 | p->parent = p->real_parent; |
6ac781b1 | 627 | add_parent(p); |
1da177e4 | 628 | |
b2b2cbc4 | 629 | if (p->state == TASK_TRACED) { |
1da177e4 LT |
630 | /* |
631 | * If it was at a trace stop, turn it into | |
632 | * a normal stop since it's no longer being | |
633 | * traced. | |
634 | */ | |
635 | ptrace_untrace(p); | |
636 | } | |
637 | } | |
638 | ||
b2b2cbc4 EB |
639 | /* If this is a threaded reparent there is no need to |
640 | * notify anyone anything has happened. | |
641 | */ | |
642 | if (p->real_parent->group_leader == father->group_leader) | |
643 | return; | |
644 | ||
645 | /* We don't want people slaying init. */ | |
646 | if (p->exit_signal != -1) | |
647 | p->exit_signal = SIGCHLD; | |
b2b2cbc4 EB |
648 | |
649 | /* If we'd notified the old parent about this child's death, | |
650 | * also notify the new parent. | |
651 | */ | |
652 | if (!traced && p->exit_state == EXIT_ZOMBIE && | |
653 | p->exit_signal != -1 && thread_group_empty(p)) | |
654 | do_notify_parent(p, p->exit_signal); | |
655 | ||
1da177e4 LT |
656 | /* |
657 | * process group orphan check | |
658 | * Case ii: Our child is in a different pgrp | |
659 | * than we are, and it was the only connection | |
660 | * outside, so the child pgrp is now orphaned. | |
661 | */ | |
0475ac08 EB |
662 | if ((task_pgrp(p) != task_pgrp(father)) && |
663 | (task_session(p) == task_session(father))) { | |
664 | struct pid *pgrp = task_pgrp(p); | |
1da177e4 | 665 | |
937949d9 CLG |
666 | if (will_become_orphaned_pgrp(pgrp, NULL) && |
667 | has_stopped_jobs(pgrp)) { | |
0475ac08 EB |
668 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); |
669 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
1da177e4 LT |
670 | } |
671 | } | |
672 | } | |
673 | ||
674 | /* | |
675 | * When we die, we re-parent all our children. | |
676 | * Try to give them to another thread in our thread | |
677 | * group, and if no such member exists, give it to | |
84d73786 SB |
678 | * the child reaper process (ie "init") in our pid |
679 | * space. | |
1da177e4 | 680 | */ |
36c8b586 IM |
681 | static void |
682 | forget_original_parent(struct task_struct *father, struct list_head *to_release) | |
1da177e4 LT |
683 | { |
684 | struct task_struct *p, *reaper = father; | |
685 | struct list_head *_p, *_n; | |
686 | ||
687 | do { | |
688 | reaper = next_thread(reaper); | |
689 | if (reaper == father) { | |
84d73786 | 690 | reaper = child_reaper(father); |
1da177e4 LT |
691 | break; |
692 | } | |
693 | } while (reaper->exit_state); | |
694 | ||
695 | /* | |
696 | * There are only two places where our children can be: | |
697 | * | |
698 | * - in our child list | |
699 | * - in our ptraced child list | |
700 | * | |
701 | * Search them and reparent children. | |
702 | */ | |
703 | list_for_each_safe(_p, _n, &father->children) { | |
704 | int ptrace; | |
36c8b586 | 705 | p = list_entry(_p, struct task_struct, sibling); |
1da177e4 LT |
706 | |
707 | ptrace = p->ptrace; | |
708 | ||
709 | /* if father isn't the real parent, then ptrace must be enabled */ | |
710 | BUG_ON(father != p->real_parent && !ptrace); | |
711 | ||
712 | if (father == p->real_parent) { | |
713 | /* reparent with a reaper, real father it's us */ | |
d799f035 | 714 | choose_new_parent(p, reaper); |
1da177e4 LT |
715 | reparent_thread(p, father, 0); |
716 | } else { | |
717 | /* reparent ptraced task to its real parent */ | |
718 | __ptrace_unlink (p); | |
719 | if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 && | |
720 | thread_group_empty(p)) | |
721 | do_notify_parent(p, p->exit_signal); | |
722 | } | |
723 | ||
724 | /* | |
725 | * if the ptraced child is a zombie with exit_signal == -1 | |
726 | * we must collect it before we exit, or it will remain | |
727 | * zombie forever since we prevented it from self-reap itself | |
728 | * while it was being traced by us, to be able to see it in wait4. | |
729 | */ | |
730 | if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1)) | |
731 | list_add(&p->ptrace_list, to_release); | |
732 | } | |
733 | list_for_each_safe(_p, _n, &father->ptrace_children) { | |
36c8b586 | 734 | p = list_entry(_p, struct task_struct, ptrace_list); |
d799f035 | 735 | choose_new_parent(p, reaper); |
1da177e4 LT |
736 | reparent_thread(p, father, 1); |
737 | } | |
738 | } | |
739 | ||
740 | /* | |
741 | * Send signals to all our closest relatives so that they know | |
742 | * to properly mourn us.. | |
743 | */ | |
744 | static void exit_notify(struct task_struct *tsk) | |
745 | { | |
746 | int state; | |
747 | struct task_struct *t; | |
748 | struct list_head ptrace_dead, *_p, *_n; | |
0475ac08 | 749 | struct pid *pgrp; |
1da177e4 LT |
750 | |
751 | if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT) | |
752 | && !thread_group_empty(tsk)) { | |
753 | /* | |
754 | * This occurs when there was a race between our exit | |
755 | * syscall and a group signal choosing us as the one to | |
756 | * wake up. It could be that we are the only thread | |
757 | * alerted to check for pending signals, but another thread | |
758 | * should be woken now to take the signal since we will not. | |
759 | * Now we'll wake all the threads in the group just to make | |
760 | * sure someone gets all the pending signals. | |
761 | */ | |
762 | read_lock(&tasklist_lock); | |
763 | spin_lock_irq(&tsk->sighand->siglock); | |
764 | for (t = next_thread(tsk); t != tsk; t = next_thread(t)) | |
7bb44ade RM |
765 | if (!signal_pending(t) && !(t->flags & PF_EXITING)) |
766 | recalc_sigpending_and_wake(t); | |
1da177e4 LT |
767 | spin_unlock_irq(&tsk->sighand->siglock); |
768 | read_unlock(&tasklist_lock); | |
769 | } | |
770 | ||
771 | write_lock_irq(&tasklist_lock); | |
772 | ||
773 | /* | |
774 | * This does two things: | |
775 | * | |
776 | * A. Make init inherit all the child processes | |
777 | * B. Check to see if any process groups have become orphaned | |
778 | * as a result of our exiting, and if they have any stopped | |
779 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
780 | */ | |
781 | ||
782 | INIT_LIST_HEAD(&ptrace_dead); | |
783 | forget_original_parent(tsk, &ptrace_dead); | |
784 | BUG_ON(!list_empty(&tsk->children)); | |
785 | BUG_ON(!list_empty(&tsk->ptrace_children)); | |
786 | ||
787 | /* | |
788 | * Check to see if any process groups have become orphaned | |
789 | * as a result of our exiting, and if they have any stopped | |
790 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) | |
791 | * | |
792 | * Case i: Our father is in a different pgrp than we are | |
793 | * and we were the only connection outside, so our pgrp | |
794 | * is about to become orphaned. | |
795 | */ | |
796 | ||
797 | t = tsk->real_parent; | |
798 | ||
0475ac08 EB |
799 | pgrp = task_pgrp(tsk); |
800 | if ((task_pgrp(t) != pgrp) && | |
14e9d573 | 801 | (task_session(t) == task_session(tsk)) && |
0475ac08 EB |
802 | will_become_orphaned_pgrp(pgrp, tsk) && |
803 | has_stopped_jobs(pgrp)) { | |
804 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); | |
805 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); | |
1da177e4 LT |
806 | } |
807 | ||
808 | /* Let father know we died | |
809 | * | |
810 | * Thread signals are configurable, but you aren't going to use | |
811 | * that to send signals to arbitary processes. | |
812 | * That stops right now. | |
813 | * | |
814 | * If the parent exec id doesn't match the exec id we saved | |
815 | * when we started then we know the parent has changed security | |
816 | * domain. | |
817 | * | |
818 | * If our self_exec id doesn't match our parent_exec_id then | |
819 | * we have changed execution domain as these two values started | |
820 | * the same after a fork. | |
821 | * | |
822 | */ | |
823 | ||
824 | if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 && | |
825 | ( tsk->parent_exec_id != t->self_exec_id || | |
826 | tsk->self_exec_id != tsk->parent_exec_id) | |
827 | && !capable(CAP_KILL)) | |
828 | tsk->exit_signal = SIGCHLD; | |
829 | ||
830 | ||
831 | /* If something other than our normal parent is ptracing us, then | |
832 | * send it a SIGCHLD instead of honoring exit_signal. exit_signal | |
833 | * only has special meaning to our real parent. | |
834 | */ | |
835 | if (tsk->exit_signal != -1 && thread_group_empty(tsk)) { | |
836 | int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD; | |
837 | do_notify_parent(tsk, signal); | |
838 | } else if (tsk->ptrace) { | |
839 | do_notify_parent(tsk, SIGCHLD); | |
840 | } | |
841 | ||
842 | state = EXIT_ZOMBIE; | |
843 | if (tsk->exit_signal == -1 && | |
844 | (likely(tsk->ptrace == 0) || | |
845 | unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT))) | |
846 | state = EXIT_DEAD; | |
847 | tsk->exit_state = state; | |
848 | ||
849 | write_unlock_irq(&tasklist_lock); | |
850 | ||
851 | list_for_each_safe(_p, _n, &ptrace_dead) { | |
852 | list_del_init(_p); | |
36c8b586 | 853 | t = list_entry(_p, struct task_struct, ptrace_list); |
1da177e4 LT |
854 | release_task(t); |
855 | } | |
856 | ||
857 | /* If the process is dead, release it - nobody will wait for it */ | |
858 | if (state == EXIT_DEAD) | |
859 | release_task(tsk); | |
1da177e4 LT |
860 | } |
861 | ||
862 | fastcall NORET_TYPE void do_exit(long code) | |
863 | { | |
864 | struct task_struct *tsk = current; | |
865 | int group_dead; | |
866 | ||
867 | profile_task_exit(tsk); | |
868 | ||
22e2c507 JA |
869 | WARN_ON(atomic_read(&tsk->fs_excl)); |
870 | ||
1da177e4 LT |
871 | if (unlikely(in_interrupt())) |
872 | panic("Aiee, killing interrupt handler!"); | |
873 | if (unlikely(!tsk->pid)) | |
874 | panic("Attempted to kill the idle task!"); | |
84d73786 SB |
875 | if (unlikely(tsk == child_reaper(tsk))) { |
876 | if (tsk->nsproxy->pid_ns != &init_pid_ns) | |
877 | tsk->nsproxy->pid_ns->child_reaper = init_pid_ns.child_reaper; | |
878 | else | |
879 | panic("Attempted to kill init!"); | |
880 | } | |
881 | ||
1da177e4 LT |
882 | |
883 | if (unlikely(current->ptrace & PT_TRACE_EXIT)) { | |
884 | current->ptrace_message = code; | |
885 | ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP); | |
886 | } | |
887 | ||
df164db5 AN |
888 | /* |
889 | * We're taking recursive faults here in do_exit. Safest is to just | |
890 | * leave this task alone and wait for reboot. | |
891 | */ | |
892 | if (unlikely(tsk->flags & PF_EXITING)) { | |
893 | printk(KERN_ALERT | |
894 | "Fixing recursive fault but reboot is needed!\n"); | |
afc847b7 AV |
895 | if (tsk->io_context) |
896 | exit_io_context(); | |
df164db5 AN |
897 | set_current_state(TASK_UNINTERRUPTIBLE); |
898 | schedule(); | |
899 | } | |
900 | ||
1da177e4 LT |
901 | tsk->flags |= PF_EXITING; |
902 | ||
1da177e4 LT |
903 | if (unlikely(in_atomic())) |
904 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", | |
905 | current->comm, current->pid, | |
906 | preempt_count()); | |
907 | ||
908 | acct_update_integrals(tsk); | |
365e9c87 HD |
909 | if (tsk->mm) { |
910 | update_hiwater_rss(tsk->mm); | |
911 | update_hiwater_vm(tsk->mm); | |
912 | } | |
1da177e4 | 913 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
c3068951 | 914 | if (group_dead) { |
2ff678b8 | 915 | hrtimer_cancel(&tsk->signal->real_timer); |
25f407f0 | 916 | exit_itimers(tsk->signal); |
c3068951 | 917 | } |
f6ec29a4 | 918 | acct_collect(code, group_dead); |
0771dfef IM |
919 | if (unlikely(tsk->robust_list)) |
920 | exit_robust_list(tsk); | |
2aa92581 | 921 | #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT) |
34f192c6 IM |
922 | if (unlikely(tsk->compat_robust_list)) |
923 | compat_exit_robust_list(tsk); | |
924 | #endif | |
fa84cb93 AV |
925 | if (unlikely(tsk->audit_context)) |
926 | audit_free(tsk); | |
115085ea ON |
927 | |
928 | taskstats_exit(tsk, group_dead); | |
c757249a | 929 | |
1da177e4 LT |
930 | exit_mm(tsk); |
931 | ||
0e464814 | 932 | if (group_dead) |
f6ec29a4 | 933 | acct_process(); |
1da177e4 LT |
934 | exit_sem(tsk); |
935 | __exit_files(tsk); | |
936 | __exit_fs(tsk); | |
1da177e4 LT |
937 | exit_thread(); |
938 | cpuset_exit(tsk); | |
939 | exit_keys(tsk); | |
940 | ||
941 | if (group_dead && tsk->signal->leader) | |
942 | disassociate_ctty(1); | |
943 | ||
a1261f54 | 944 | module_put(task_thread_info(tsk)->exec_domain->module); |
1da177e4 LT |
945 | if (tsk->binfmt) |
946 | module_put(tsk->binfmt->module); | |
947 | ||
948 | tsk->exit_code = code; | |
9f46080c | 949 | proc_exit_connector(tsk); |
444f378b | 950 | exit_task_namespaces(tsk); |
0f245285 | 951 | exit_notify(tsk); |
1da177e4 LT |
952 | #ifdef CONFIG_NUMA |
953 | mpol_free(tsk->mempolicy); | |
954 | tsk->mempolicy = NULL; | |
955 | #endif | |
c87e2837 IM |
956 | /* |
957 | * This must happen late, after the PID is not | |
958 | * hashed anymore: | |
959 | */ | |
960 | if (unlikely(!list_empty(&tsk->pi_state_list))) | |
961 | exit_pi_state_list(tsk); | |
962 | if (unlikely(current->pi_state_cache)) | |
963 | kfree(current->pi_state_cache); | |
de5097c2 | 964 | /* |
9a11b49a | 965 | * Make sure we are holding no locks: |
de5097c2 | 966 | */ |
9a11b49a | 967 | debug_check_no_locks_held(tsk); |
1da177e4 | 968 | |
afc847b7 AV |
969 | if (tsk->io_context) |
970 | exit_io_context(); | |
971 | ||
b92ce558 JA |
972 | if (tsk->splice_pipe) |
973 | __free_pipe_info(tsk->splice_pipe); | |
974 | ||
7407251a | 975 | preempt_disable(); |
55a101f8 | 976 | /* causes final put_task_struct in finish_task_switch(). */ |
c394cc9f | 977 | tsk->state = TASK_DEAD; |
7407251a | 978 | |
1da177e4 LT |
979 | schedule(); |
980 | BUG(); | |
981 | /* Avoid "noreturn function does return". */ | |
54306cf0 AC |
982 | for (;;) |
983 | cpu_relax(); /* For when BUG is null */ | |
1da177e4 LT |
984 | } |
985 | ||
012914da RA |
986 | EXPORT_SYMBOL_GPL(do_exit); |
987 | ||
1da177e4 LT |
988 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
989 | { | |
990 | if (comp) | |
991 | complete(comp); | |
55a101f8 | 992 | |
1da177e4 LT |
993 | do_exit(code); |
994 | } | |
995 | ||
996 | EXPORT_SYMBOL(complete_and_exit); | |
997 | ||
998 | asmlinkage long sys_exit(int error_code) | |
999 | { | |
1000 | do_exit((error_code&0xff)<<8); | |
1001 | } | |
1002 | ||
1da177e4 LT |
1003 | /* |
1004 | * Take down every thread in the group. This is called by fatal signals | |
1005 | * as well as by sys_exit_group (below). | |
1006 | */ | |
1007 | NORET_TYPE void | |
1008 | do_group_exit(int exit_code) | |
1009 | { | |
1010 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ | |
1011 | ||
1012 | if (current->signal->flags & SIGNAL_GROUP_EXIT) | |
1013 | exit_code = current->signal->group_exit_code; | |
1014 | else if (!thread_group_empty(current)) { | |
1015 | struct signal_struct *const sig = current->signal; | |
1016 | struct sighand_struct *const sighand = current->sighand; | |
1da177e4 LT |
1017 | spin_lock_irq(&sighand->siglock); |
1018 | if (sig->flags & SIGNAL_GROUP_EXIT) | |
1019 | /* Another thread got here before we took the lock. */ | |
1020 | exit_code = sig->group_exit_code; | |
1021 | else { | |
1da177e4 LT |
1022 | sig->group_exit_code = exit_code; |
1023 | zap_other_threads(current); | |
1024 | } | |
1025 | spin_unlock_irq(&sighand->siglock); | |
1da177e4 LT |
1026 | } |
1027 | ||
1028 | do_exit(exit_code); | |
1029 | /* NOTREACHED */ | |
1030 | } | |
1031 | ||
1032 | /* | |
1033 | * this kills every thread in the thread group. Note that any externally | |
1034 | * wait4()-ing process will get the correct exit code - even if this | |
1035 | * thread is not the thread group leader. | |
1036 | */ | |
1037 | asmlinkage void sys_exit_group(int error_code) | |
1038 | { | |
1039 | do_group_exit((error_code & 0xff) << 8); | |
1040 | } | |
1041 | ||
36c8b586 | 1042 | static int eligible_child(pid_t pid, int options, struct task_struct *p) |
1da177e4 | 1043 | { |
73243284 RM |
1044 | int err; |
1045 | ||
1da177e4 LT |
1046 | if (pid > 0) { |
1047 | if (p->pid != pid) | |
1048 | return 0; | |
1049 | } else if (!pid) { | |
1050 | if (process_group(p) != process_group(current)) | |
1051 | return 0; | |
1052 | } else if (pid != -1) { | |
1053 | if (process_group(p) != -pid) | |
1054 | return 0; | |
1055 | } | |
1056 | ||
1057 | /* | |
1058 | * Do not consider detached threads that are | |
1059 | * not ptraced: | |
1060 | */ | |
1061 | if (p->exit_signal == -1 && !p->ptrace) | |
1062 | return 0; | |
1063 | ||
1064 | /* Wait for all children (clone and not) if __WALL is set; | |
1065 | * otherwise, wait for clone children *only* if __WCLONE is | |
1066 | * set; otherwise, wait for non-clone children *only*. (Note: | |
1067 | * A "clone" child here is one that reports to its parent | |
1068 | * using a signal other than SIGCHLD.) */ | |
1069 | if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0)) | |
1070 | && !(options & __WALL)) | |
1071 | return 0; | |
1072 | /* | |
1073 | * Do not consider thread group leaders that are | |
1074 | * in a non-empty thread group: | |
1075 | */ | |
3b6362b8 | 1076 | if (delay_group_leader(p)) |
1da177e4 LT |
1077 | return 2; |
1078 | ||
73243284 RM |
1079 | err = security_task_wait(p); |
1080 | if (err) | |
1081 | return err; | |
1da177e4 LT |
1082 | |
1083 | return 1; | |
1084 | } | |
1085 | ||
36c8b586 | 1086 | static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid, |
1da177e4 LT |
1087 | int why, int status, |
1088 | struct siginfo __user *infop, | |
1089 | struct rusage __user *rusagep) | |
1090 | { | |
1091 | int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0; | |
36c8b586 | 1092 | |
1da177e4 LT |
1093 | put_task_struct(p); |
1094 | if (!retval) | |
1095 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1096 | if (!retval) | |
1097 | retval = put_user(0, &infop->si_errno); | |
1098 | if (!retval) | |
1099 | retval = put_user((short)why, &infop->si_code); | |
1100 | if (!retval) | |
1101 | retval = put_user(pid, &infop->si_pid); | |
1102 | if (!retval) | |
1103 | retval = put_user(uid, &infop->si_uid); | |
1104 | if (!retval) | |
1105 | retval = put_user(status, &infop->si_status); | |
1106 | if (!retval) | |
1107 | retval = pid; | |
1108 | return retval; | |
1109 | } | |
1110 | ||
1111 | /* | |
1112 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold | |
1113 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1114 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1115 | * released the lock and the system call should return. | |
1116 | */ | |
36c8b586 | 1117 | static int wait_task_zombie(struct task_struct *p, int noreap, |
1da177e4 LT |
1118 | struct siginfo __user *infop, |
1119 | int __user *stat_addr, struct rusage __user *ru) | |
1120 | { | |
1121 | unsigned long state; | |
1122 | int retval; | |
1123 | int status; | |
1124 | ||
1125 | if (unlikely(noreap)) { | |
1126 | pid_t pid = p->pid; | |
1127 | uid_t uid = p->uid; | |
1128 | int exit_code = p->exit_code; | |
1129 | int why, status; | |
1130 | ||
1131 | if (unlikely(p->exit_state != EXIT_ZOMBIE)) | |
1132 | return 0; | |
1133 | if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) | |
1134 | return 0; | |
1135 | get_task_struct(p); | |
1136 | read_unlock(&tasklist_lock); | |
1137 | if ((exit_code & 0x7f) == 0) { | |
1138 | why = CLD_EXITED; | |
1139 | status = exit_code >> 8; | |
1140 | } else { | |
1141 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1142 | status = exit_code & 0x7f; | |
1143 | } | |
1144 | return wait_noreap_copyout(p, pid, uid, why, | |
1145 | status, infop, ru); | |
1146 | } | |
1147 | ||
1148 | /* | |
1149 | * Try to move the task's state to DEAD | |
1150 | * only one thread is allowed to do this: | |
1151 | */ | |
1152 | state = xchg(&p->exit_state, EXIT_DEAD); | |
1153 | if (state != EXIT_ZOMBIE) { | |
1154 | BUG_ON(state != EXIT_DEAD); | |
1155 | return 0; | |
1156 | } | |
1157 | if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) { | |
1158 | /* | |
1159 | * This can only happen in a race with a ptraced thread | |
1160 | * dying on another processor. | |
1161 | */ | |
1162 | return 0; | |
1163 | } | |
1164 | ||
1165 | if (likely(p->real_parent == p->parent) && likely(p->signal)) { | |
3795e161 JJ |
1166 | struct signal_struct *psig; |
1167 | struct signal_struct *sig; | |
1168 | ||
1da177e4 LT |
1169 | /* |
1170 | * The resource counters for the group leader are in its | |
1171 | * own task_struct. Those for dead threads in the group | |
1172 | * are in its signal_struct, as are those for the child | |
1173 | * processes it has previously reaped. All these | |
1174 | * accumulate in the parent's signal_struct c* fields. | |
1175 | * | |
1176 | * We don't bother to take a lock here to protect these | |
1177 | * p->signal fields, because they are only touched by | |
1178 | * __exit_signal, which runs with tasklist_lock | |
1179 | * write-locked anyway, and so is excluded here. We do | |
1180 | * need to protect the access to p->parent->signal fields, | |
1181 | * as other threads in the parent group can be right | |
1182 | * here reaping other children at the same time. | |
1183 | */ | |
1184 | spin_lock_irq(&p->parent->sighand->siglock); | |
3795e161 JJ |
1185 | psig = p->parent->signal; |
1186 | sig = p->signal; | |
1187 | psig->cutime = | |
1188 | cputime_add(psig->cutime, | |
1da177e4 | 1189 | cputime_add(p->utime, |
3795e161 JJ |
1190 | cputime_add(sig->utime, |
1191 | sig->cutime))); | |
1192 | psig->cstime = | |
1193 | cputime_add(psig->cstime, | |
1da177e4 | 1194 | cputime_add(p->stime, |
3795e161 JJ |
1195 | cputime_add(sig->stime, |
1196 | sig->cstime))); | |
1197 | psig->cmin_flt += | |
1198 | p->min_flt + sig->min_flt + sig->cmin_flt; | |
1199 | psig->cmaj_flt += | |
1200 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; | |
1201 | psig->cnvcsw += | |
1202 | p->nvcsw + sig->nvcsw + sig->cnvcsw; | |
1203 | psig->cnivcsw += | |
1204 | p->nivcsw + sig->nivcsw + sig->cnivcsw; | |
6eaeeaba ED |
1205 | psig->cinblock += |
1206 | task_io_get_inblock(p) + | |
1207 | sig->inblock + sig->cinblock; | |
1208 | psig->coublock += | |
1209 | task_io_get_oublock(p) + | |
1210 | sig->oublock + sig->coublock; | |
1da177e4 LT |
1211 | spin_unlock_irq(&p->parent->sighand->siglock); |
1212 | } | |
1213 | ||
1214 | /* | |
1215 | * Now we are sure this task is interesting, and no other | |
1216 | * thread can reap it because we set its state to EXIT_DEAD. | |
1217 | */ | |
1218 | read_unlock(&tasklist_lock); | |
1219 | ||
1220 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1221 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) | |
1222 | ? p->signal->group_exit_code : p->exit_code; | |
1223 | if (!retval && stat_addr) | |
1224 | retval = put_user(status, stat_addr); | |
1225 | if (!retval && infop) | |
1226 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1227 | if (!retval && infop) | |
1228 | retval = put_user(0, &infop->si_errno); | |
1229 | if (!retval && infop) { | |
1230 | int why; | |
1231 | ||
1232 | if ((status & 0x7f) == 0) { | |
1233 | why = CLD_EXITED; | |
1234 | status >>= 8; | |
1235 | } else { | |
1236 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; | |
1237 | status &= 0x7f; | |
1238 | } | |
1239 | retval = put_user((short)why, &infop->si_code); | |
1240 | if (!retval) | |
1241 | retval = put_user(status, &infop->si_status); | |
1242 | } | |
1243 | if (!retval && infop) | |
1244 | retval = put_user(p->pid, &infop->si_pid); | |
1245 | if (!retval && infop) | |
1246 | retval = put_user(p->uid, &infop->si_uid); | |
1247 | if (retval) { | |
1248 | // TODO: is this safe? | |
1249 | p->exit_state = EXIT_ZOMBIE; | |
1250 | return retval; | |
1251 | } | |
1252 | retval = p->pid; | |
1253 | if (p->real_parent != p->parent) { | |
1254 | write_lock_irq(&tasklist_lock); | |
1255 | /* Double-check with lock held. */ | |
1256 | if (p->real_parent != p->parent) { | |
1257 | __ptrace_unlink(p); | |
1258 | // TODO: is this safe? | |
1259 | p->exit_state = EXIT_ZOMBIE; | |
1260 | /* | |
1261 | * If this is not a detached task, notify the parent. | |
1262 | * If it's still not detached after that, don't release | |
1263 | * it now. | |
1264 | */ | |
1265 | if (p->exit_signal != -1) { | |
1266 | do_notify_parent(p, p->exit_signal); | |
1267 | if (p->exit_signal != -1) | |
1268 | p = NULL; | |
1269 | } | |
1270 | } | |
1271 | write_unlock_irq(&tasklist_lock); | |
1272 | } | |
1273 | if (p != NULL) | |
1274 | release_task(p); | |
1275 | BUG_ON(!retval); | |
1276 | return retval; | |
1277 | } | |
1278 | ||
1279 | /* | |
1280 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | |
1281 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1282 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1283 | * released the lock and the system call should return. | |
1284 | */ | |
36c8b586 IM |
1285 | static int wait_task_stopped(struct task_struct *p, int delayed_group_leader, |
1286 | int noreap, struct siginfo __user *infop, | |
1da177e4 LT |
1287 | int __user *stat_addr, struct rusage __user *ru) |
1288 | { | |
1289 | int retval, exit_code; | |
1290 | ||
1291 | if (!p->exit_code) | |
1292 | return 0; | |
1293 | if (delayed_group_leader && !(p->ptrace & PT_PTRACED) && | |
1294 | p->signal && p->signal->group_stop_count > 0) | |
1295 | /* | |
1296 | * A group stop is in progress and this is the group leader. | |
1297 | * We won't report until all threads have stopped. | |
1298 | */ | |
1299 | return 0; | |
1300 | ||
1301 | /* | |
1302 | * Now we are pretty sure this task is interesting. | |
1303 | * Make sure it doesn't get reaped out from under us while we | |
1304 | * give up the lock and then examine it below. We don't want to | |
1305 | * keep holding onto the tasklist_lock while we call getrusage and | |
1306 | * possibly take page faults for user memory. | |
1307 | */ | |
1308 | get_task_struct(p); | |
1309 | read_unlock(&tasklist_lock); | |
1310 | ||
1311 | if (unlikely(noreap)) { | |
1312 | pid_t pid = p->pid; | |
1313 | uid_t uid = p->uid; | |
1314 | int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED; | |
1315 | ||
1316 | exit_code = p->exit_code; | |
1317 | if (unlikely(!exit_code) || | |
14bf01bb | 1318 | unlikely(p->state & TASK_TRACED)) |
1da177e4 LT |
1319 | goto bail_ref; |
1320 | return wait_noreap_copyout(p, pid, uid, | |
1321 | why, (exit_code << 8) | 0x7f, | |
1322 | infop, ru); | |
1323 | } | |
1324 | ||
1325 | write_lock_irq(&tasklist_lock); | |
1326 | ||
1327 | /* | |
1328 | * This uses xchg to be atomic with the thread resuming and setting | |
1329 | * it. It must also be done with the write lock held to prevent a | |
1330 | * race with the EXIT_ZOMBIE case. | |
1331 | */ | |
1332 | exit_code = xchg(&p->exit_code, 0); | |
1333 | if (unlikely(p->exit_state)) { | |
1334 | /* | |
1335 | * The task resumed and then died. Let the next iteration | |
1336 | * catch it in EXIT_ZOMBIE. Note that exit_code might | |
1337 | * already be zero here if it resumed and did _exit(0). | |
1338 | * The task itself is dead and won't touch exit_code again; | |
1339 | * other processors in this function are locked out. | |
1340 | */ | |
1341 | p->exit_code = exit_code; | |
1342 | exit_code = 0; | |
1343 | } | |
1344 | if (unlikely(exit_code == 0)) { | |
1345 | /* | |
1346 | * Another thread in this function got to it first, or it | |
1347 | * resumed, or it resumed and then died. | |
1348 | */ | |
1349 | write_unlock_irq(&tasklist_lock); | |
1350 | bail_ref: | |
1351 | put_task_struct(p); | |
1352 | /* | |
1353 | * We are returning to the wait loop without having successfully | |
1354 | * removed the process and having released the lock. We cannot | |
1355 | * continue, since the "p" task pointer is potentially stale. | |
1356 | * | |
1357 | * Return -EAGAIN, and do_wait() will restart the loop from the | |
1358 | * beginning. Do _not_ re-acquire the lock. | |
1359 | */ | |
1360 | return -EAGAIN; | |
1361 | } | |
1362 | ||
1363 | /* move to end of parent's list to avoid starvation */ | |
1364 | remove_parent(p); | |
8fafabd8 | 1365 | add_parent(p); |
1da177e4 LT |
1366 | |
1367 | write_unlock_irq(&tasklist_lock); | |
1368 | ||
1369 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1370 | if (!retval && stat_addr) | |
1371 | retval = put_user((exit_code << 8) | 0x7f, stat_addr); | |
1372 | if (!retval && infop) | |
1373 | retval = put_user(SIGCHLD, &infop->si_signo); | |
1374 | if (!retval && infop) | |
1375 | retval = put_user(0, &infop->si_errno); | |
1376 | if (!retval && infop) | |
1377 | retval = put_user((short)((p->ptrace & PT_PTRACED) | |
1378 | ? CLD_TRAPPED : CLD_STOPPED), | |
1379 | &infop->si_code); | |
1380 | if (!retval && infop) | |
1381 | retval = put_user(exit_code, &infop->si_status); | |
1382 | if (!retval && infop) | |
1383 | retval = put_user(p->pid, &infop->si_pid); | |
1384 | if (!retval && infop) | |
1385 | retval = put_user(p->uid, &infop->si_uid); | |
1386 | if (!retval) | |
1387 | retval = p->pid; | |
1388 | put_task_struct(p); | |
1389 | ||
1390 | BUG_ON(!retval); | |
1391 | return retval; | |
1392 | } | |
1393 | ||
1394 | /* | |
1395 | * Handle do_wait work for one task in a live, non-stopped state. | |
1396 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | |
1397 | * the lock and this task is uninteresting. If we return nonzero, we have | |
1398 | * released the lock and the system call should return. | |
1399 | */ | |
36c8b586 | 1400 | static int wait_task_continued(struct task_struct *p, int noreap, |
1da177e4 LT |
1401 | struct siginfo __user *infop, |
1402 | int __user *stat_addr, struct rusage __user *ru) | |
1403 | { | |
1404 | int retval; | |
1405 | pid_t pid; | |
1406 | uid_t uid; | |
1407 | ||
1408 | if (unlikely(!p->signal)) | |
1409 | return 0; | |
1410 | ||
1411 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) | |
1412 | return 0; | |
1413 | ||
1414 | spin_lock_irq(&p->sighand->siglock); | |
1415 | /* Re-check with the lock held. */ | |
1416 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { | |
1417 | spin_unlock_irq(&p->sighand->siglock); | |
1418 | return 0; | |
1419 | } | |
1420 | if (!noreap) | |
1421 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; | |
1422 | spin_unlock_irq(&p->sighand->siglock); | |
1423 | ||
1424 | pid = p->pid; | |
1425 | uid = p->uid; | |
1426 | get_task_struct(p); | |
1427 | read_unlock(&tasklist_lock); | |
1428 | ||
1429 | if (!infop) { | |
1430 | retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0; | |
1431 | put_task_struct(p); | |
1432 | if (!retval && stat_addr) | |
1433 | retval = put_user(0xffff, stat_addr); | |
1434 | if (!retval) | |
1435 | retval = p->pid; | |
1436 | } else { | |
1437 | retval = wait_noreap_copyout(p, pid, uid, | |
1438 | CLD_CONTINUED, SIGCONT, | |
1439 | infop, ru); | |
1440 | BUG_ON(retval == 0); | |
1441 | } | |
1442 | ||
1443 | return retval; | |
1444 | } | |
1445 | ||
1446 | ||
1447 | static inline int my_ptrace_child(struct task_struct *p) | |
1448 | { | |
1449 | if (!(p->ptrace & PT_PTRACED)) | |
1450 | return 0; | |
1451 | if (!(p->ptrace & PT_ATTACHED)) | |
1452 | return 1; | |
1453 | /* | |
1454 | * This child was PTRACE_ATTACH'd. We should be seeing it only if | |
1455 | * we are the attacher. If we are the real parent, this is a race | |
1456 | * inside ptrace_attach. It is waiting for the tasklist_lock, | |
1457 | * which we have to switch the parent links, but has already set | |
1458 | * the flags in p->ptrace. | |
1459 | */ | |
1460 | return (p->parent != p->real_parent); | |
1461 | } | |
1462 | ||
1463 | static long do_wait(pid_t pid, int options, struct siginfo __user *infop, | |
1464 | int __user *stat_addr, struct rusage __user *ru) | |
1465 | { | |
1466 | DECLARE_WAITQUEUE(wait, current); | |
1467 | struct task_struct *tsk; | |
1468 | int flag, retval; | |
73243284 | 1469 | int allowed, denied; |
1da177e4 LT |
1470 | |
1471 | add_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1472 | repeat: | |
1473 | /* | |
1474 | * We will set this flag if we see any child that might later | |
1475 | * match our criteria, even if we are not able to reap it yet. | |
1476 | */ | |
1477 | flag = 0; | |
73243284 | 1478 | allowed = denied = 0; |
1da177e4 LT |
1479 | current->state = TASK_INTERRUPTIBLE; |
1480 | read_lock(&tasklist_lock); | |
1481 | tsk = current; | |
1482 | do { | |
1483 | struct task_struct *p; | |
1484 | struct list_head *_p; | |
1485 | int ret; | |
1486 | ||
1487 | list_for_each(_p,&tsk->children) { | |
36c8b586 | 1488 | p = list_entry(_p, struct task_struct, sibling); |
1da177e4 LT |
1489 | |
1490 | ret = eligible_child(pid, options, p); | |
1491 | if (!ret) | |
1492 | continue; | |
1493 | ||
73243284 RM |
1494 | if (unlikely(ret < 0)) { |
1495 | denied = ret; | |
1496 | continue; | |
1497 | } | |
1498 | allowed = 1; | |
1499 | ||
1da177e4 LT |
1500 | switch (p->state) { |
1501 | case TASK_TRACED: | |
7f2a5255 RM |
1502 | /* |
1503 | * When we hit the race with PTRACE_ATTACH, | |
1504 | * we will not report this child. But the | |
1505 | * race means it has not yet been moved to | |
1506 | * our ptrace_children list, so we need to | |
1507 | * set the flag here to avoid a spurious ECHILD | |
1508 | * when the race happens with the only child. | |
1509 | */ | |
1510 | flag = 1; | |
1da177e4 LT |
1511 | if (!my_ptrace_child(p)) |
1512 | continue; | |
1513 | /*FALLTHROUGH*/ | |
1514 | case TASK_STOPPED: | |
1515 | /* | |
1516 | * It's stopped now, so it might later | |
1517 | * continue, exit, or stop again. | |
1518 | */ | |
1519 | flag = 1; | |
1520 | if (!(options & WUNTRACED) && | |
1521 | !my_ptrace_child(p)) | |
1522 | continue; | |
1523 | retval = wait_task_stopped(p, ret == 2, | |
1524 | (options & WNOWAIT), | |
1525 | infop, | |
1526 | stat_addr, ru); | |
1527 | if (retval == -EAGAIN) | |
1528 | goto repeat; | |
1529 | if (retval != 0) /* He released the lock. */ | |
1530 | goto end; | |
1531 | break; | |
1532 | default: | |
1533 | // case EXIT_DEAD: | |
1534 | if (p->exit_state == EXIT_DEAD) | |
1535 | continue; | |
1536 | // case EXIT_ZOMBIE: | |
1537 | if (p->exit_state == EXIT_ZOMBIE) { | |
1538 | /* | |
1539 | * Eligible but we cannot release | |
1540 | * it yet: | |
1541 | */ | |
1542 | if (ret == 2) | |
1543 | goto check_continued; | |
1544 | if (!likely(options & WEXITED)) | |
1545 | continue; | |
1546 | retval = wait_task_zombie( | |
1547 | p, (options & WNOWAIT), | |
1548 | infop, stat_addr, ru); | |
1549 | /* He released the lock. */ | |
1550 | if (retval != 0) | |
1551 | goto end; | |
1552 | break; | |
1553 | } | |
1554 | check_continued: | |
1555 | /* | |
1556 | * It's running now, so it might later | |
1557 | * exit, stop, or stop and then continue. | |
1558 | */ | |
1559 | flag = 1; | |
1560 | if (!unlikely(options & WCONTINUED)) | |
1561 | continue; | |
1562 | retval = wait_task_continued( | |
1563 | p, (options & WNOWAIT), | |
1564 | infop, stat_addr, ru); | |
1565 | if (retval != 0) /* He released the lock. */ | |
1566 | goto end; | |
1567 | break; | |
1568 | } | |
1569 | } | |
1570 | if (!flag) { | |
1571 | list_for_each(_p, &tsk->ptrace_children) { | |
1572 | p = list_entry(_p, struct task_struct, | |
1573 | ptrace_list); | |
1574 | if (!eligible_child(pid, options, p)) | |
1575 | continue; | |
1576 | flag = 1; | |
1577 | break; | |
1578 | } | |
1579 | } | |
1580 | if (options & __WNOTHREAD) | |
1581 | break; | |
1582 | tsk = next_thread(tsk); | |
125e1874 | 1583 | BUG_ON(tsk->signal != current->signal); |
1da177e4 LT |
1584 | } while (tsk != current); |
1585 | ||
1586 | read_unlock(&tasklist_lock); | |
1587 | if (flag) { | |
1588 | retval = 0; | |
1589 | if (options & WNOHANG) | |
1590 | goto end; | |
1591 | retval = -ERESTARTSYS; | |
1592 | if (signal_pending(current)) | |
1593 | goto end; | |
1594 | schedule(); | |
1595 | goto repeat; | |
1596 | } | |
1597 | retval = -ECHILD; | |
73243284 RM |
1598 | if (unlikely(denied) && !allowed) |
1599 | retval = denied; | |
1da177e4 LT |
1600 | end: |
1601 | current->state = TASK_RUNNING; | |
1602 | remove_wait_queue(¤t->signal->wait_chldexit,&wait); | |
1603 | if (infop) { | |
1604 | if (retval > 0) | |
1605 | retval = 0; | |
1606 | else { | |
1607 | /* | |
1608 | * For a WNOHANG return, clear out all the fields | |
1609 | * we would set so the user can easily tell the | |
1610 | * difference. | |
1611 | */ | |
1612 | if (!retval) | |
1613 | retval = put_user(0, &infop->si_signo); | |
1614 | if (!retval) | |
1615 | retval = put_user(0, &infop->si_errno); | |
1616 | if (!retval) | |
1617 | retval = put_user(0, &infop->si_code); | |
1618 | if (!retval) | |
1619 | retval = put_user(0, &infop->si_pid); | |
1620 | if (!retval) | |
1621 | retval = put_user(0, &infop->si_uid); | |
1622 | if (!retval) | |
1623 | retval = put_user(0, &infop->si_status); | |
1624 | } | |
1625 | } | |
1626 | return retval; | |
1627 | } | |
1628 | ||
1629 | asmlinkage long sys_waitid(int which, pid_t pid, | |
1630 | struct siginfo __user *infop, int options, | |
1631 | struct rusage __user *ru) | |
1632 | { | |
1633 | long ret; | |
1634 | ||
1635 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) | |
1636 | return -EINVAL; | |
1637 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) | |
1638 | return -EINVAL; | |
1639 | ||
1640 | switch (which) { | |
1641 | case P_ALL: | |
1642 | pid = -1; | |
1643 | break; | |
1644 | case P_PID: | |
1645 | if (pid <= 0) | |
1646 | return -EINVAL; | |
1647 | break; | |
1648 | case P_PGID: | |
1649 | if (pid <= 0) | |
1650 | return -EINVAL; | |
1651 | pid = -pid; | |
1652 | break; | |
1653 | default: | |
1654 | return -EINVAL; | |
1655 | } | |
1656 | ||
1657 | ret = do_wait(pid, options, infop, NULL, ru); | |
1658 | ||
1659 | /* avoid REGPARM breakage on x86: */ | |
1660 | prevent_tail_call(ret); | |
1661 | return ret; | |
1662 | } | |
1663 | ||
1664 | asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr, | |
1665 | int options, struct rusage __user *ru) | |
1666 | { | |
1667 | long ret; | |
1668 | ||
1669 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| | |
1670 | __WNOTHREAD|__WCLONE|__WALL)) | |
1671 | return -EINVAL; | |
1672 | ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru); | |
1673 | ||
1674 | /* avoid REGPARM breakage on x86: */ | |
1675 | prevent_tail_call(ret); | |
1676 | return ret; | |
1677 | } | |
1678 | ||
1679 | #ifdef __ARCH_WANT_SYS_WAITPID | |
1680 | ||
1681 | /* | |
1682 | * sys_waitpid() remains for compatibility. waitpid() should be | |
1683 | * implemented by calling sys_wait4() from libc.a. | |
1684 | */ | |
1685 | asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options) | |
1686 | { | |
1687 | return sys_wait4(pid, stat_addr, options, NULL); | |
1688 | } | |
1689 | ||
1690 | #endif |