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> | |
10 | #include <linux/smp_lock.h> | |
11 | #include <linux/module.h> | |
c59ede7b | 12 | #include <linux/capability.h> |
1da177e4 LT |
13 | #include <linux/completion.h> |
14 | #include <linux/personality.h> | |
15 | #include <linux/tty.h> | |
16 | #include <linux/namespace.h> | |
17 | #include <linux/key.h> | |
18 | #include <linux/security.h> | |
19 | #include <linux/cpu.h> | |
20 | #include <linux/acct.h> | |
8f0ab514 | 21 | #include <linux/tsacct_kern.h> |
1da177e4 LT |
22 | #include <linux/file.h> |
23 | #include <linux/binfmts.h> | |
24 | #include <linux/ptrace.h> | |
25 | #include <linux/profile.h> | |
26 | #include <linux/mount.h> | |
27 | #include <linux/proc_fs.h> | |
28 | #include <linux/mempolicy.h> | |
c757249a | 29 | #include <linux/taskstats_kern.h> |
ca74e92b | 30 | #include <linux/delayacct.h> |
1da177e4 LT |
31 | #include <linux/cpuset.h> |
32 | #include <linux/syscalls.h> | |
7ed20e1a | 33 | #include <linux/signal.h> |
6a14c5c9 | 34 | #include <linux/posix-timers.h> |
9f46080c | 35 | #include <linux/cn_proc.h> |
de5097c2 | 36 | #include <linux/mutex.h> |
0771dfef | 37 | #include <linux/futex.h> |
34f192c6 | 38 | #include <linux/compat.h> |
b92ce558 | 39 | #include <linux/pipe_fs_i.h> |
fa84cb93 | 40 | #include <linux/audit.h> /* for audit_free() */ |
83cc5ed3 | 41 | #include <linux/resource.h> |
0d67a46d | 42 | #include <linux/blkdev.h> |
1da177e4 LT |
43 | |
44 | #include <asm/uaccess.h> | |
45 | #include <asm/unistd.h> | |
46 | #include <asm/pgtable.h> | |
47 | #include <asm/mmu_context.h> | |
48 | ||
49 | extern void sem_exit (void); | |
50 | extern struct task_struct *child_reaper; | |
51 | ||
408b664a AB |
52 | static void exit_mm(struct task_struct * tsk); |
53 | ||
1da177e4 LT |
54 | static void __unhash_process(struct task_struct *p) |
55 | { | |
56 | nr_threads--; | |
57 | detach_pid(p, PIDTYPE_PID); | |
1da177e4 LT |
58 | if (thread_group_leader(p)) { |
59 | detach_pid(p, PIDTYPE_PGID); | |
60 | detach_pid(p, PIDTYPE_SID); | |
c97d9893 | 61 | |
5e85d4ab | 62 | list_del_rcu(&p->tasks); |
73b9ebfe | 63 | __get_cpu_var(process_counts)--; |
1da177e4 | 64 | } |
47e65328 | 65 | list_del_rcu(&p->thread_group); |
c97d9893 | 66 | remove_parent(p); |
1da177e4 LT |
67 | } |
68 | ||
6a14c5c9 ON |
69 | /* |
70 | * This function expects the tasklist_lock write-locked. | |
71 | */ | |
72 | static void __exit_signal(struct task_struct *tsk) | |
73 | { | |
74 | struct signal_struct *sig = tsk->signal; | |
75 | struct sighand_struct *sighand; | |
76 | ||
77 | BUG_ON(!sig); | |
78 | BUG_ON(!atomic_read(&sig->count)); | |
79 | ||
80 | rcu_read_lock(); | |
81 | sighand = rcu_dereference(tsk->sighand); | |
82 | spin_lock(&sighand->siglock); | |
83 | ||
84 | posix_cpu_timers_exit(tsk); | |
85 | if (atomic_dec_and_test(&sig->count)) | |
86 | posix_cpu_timers_exit_group(tsk); | |
87 | else { | |
88 | /* | |
89 | * If there is any task waiting for the group exit | |
90 | * then notify it: | |
91 | */ | |
92 | if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) { | |
93 | wake_up_process(sig->group_exit_task); | |
94 | sig->group_exit_task = NULL; | |
95 | } | |
96 | if (tsk == sig->curr_target) | |
97 | sig->curr_target = next_thread(tsk); | |
98 | /* | |
99 | * Accumulate here the counters for all threads but the | |
100 | * group leader as they die, so they can be added into | |
101 | * the process-wide totals when those are taken. | |
102 | * The group leader stays around as a zombie as long | |
103 | * as there are other threads. When it gets reaped, | |
104 | * the exit.c code will add its counts into these totals. | |
105 | * We won't ever get here for the group leader, since it | |
106 | * will have been the last reference on the signal_struct. | |
107 | */ | |
108 | sig->utime = cputime_add(sig->utime, tsk->utime); | |
109 | sig->stime = cputime_add(sig->stime, tsk->stime); | |
110 | sig->min_flt += tsk->min_flt; | |
111 | sig->maj_flt += tsk->maj_flt; | |
112 | sig->nvcsw += tsk->nvcsw; | |
113 | sig->nivcsw += tsk->nivcsw; | |
114 | sig->sched_time += tsk->sched_time; | |
115 | sig = NULL; /* Marker for below. */ | |
116 | } | |
117 | ||
5876700c ON |
118 | __unhash_process(tsk); |
119 | ||
6a14c5c9 | 120 | tsk->signal = NULL; |
a7e5328a | 121 | tsk->sighand = NULL; |
6a14c5c9 ON |
122 | spin_unlock(&sighand->siglock); |
123 | rcu_read_unlock(); | |
124 | ||
a7e5328a | 125 | __cleanup_sighand(sighand); |
6a14c5c9 ON |
126 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
127 | flush_sigqueue(&tsk->pending); | |
128 | if (sig) { | |
129 | flush_sigqueue(&sig->shared_pending); | |
130 | __cleanup_signal(sig); | |
131 | } | |
132 | } | |
133 | ||
8c7904a0 EB |
134 | static void delayed_put_task_struct(struct rcu_head *rhp) |
135 | { | |
136 | put_task_struct(container_of(rhp, struct task_struct, rcu)); | |
137 | } | |
138 | ||
1da177e4 LT |
139 | void release_task(struct task_struct * p) |
140 | { | |
36c8b586 | 141 | struct task_struct *leader; |
1da177e4 | 142 | int zap_leader; |
1f09f974 | 143 | repeat: |
1da177e4 | 144 | atomic_dec(&p->user->processes); |
1da177e4 | 145 | write_lock_irq(&tasklist_lock); |
1f09f974 | 146 | ptrace_unlink(p); |
1da177e4 LT |
147 | BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children)); |
148 | __exit_signal(p); | |
35f5cad8 | 149 | |
1da177e4 LT |
150 | /* |
151 | * If we are the last non-leader member of the thread | |
152 | * group, and the leader is zombie, then notify the | |
153 | * group leader's parent process. (if it wants notification.) | |
154 | */ | |
155 | zap_leader = 0; | |
156 | leader = p->group_leader; | |
157 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { | |
158 | BUG_ON(leader->exit_signal == -1); | |
159 | do_notify_parent(leader, leader->exit_signal); | |
160 | /* | |
161 | * If we were the last child thread and the leader has | |
162 | * exited already, and the leader's parent ignores SIGCHLD, | |
163 | * then we are the one who should release the leader. | |
164 | * | |
165 | * do_notify_parent() will have marked it self-reaping in | |
166 | * that case. | |
167 | */ | |
168 | zap_leader = (leader->exit_signal == -1); | |
169 | } | |
170 | ||
171 | sched_exit(p); | |
172 | write_unlock_irq(&tasklist_lock); | |
48e6484d | 173 | proc_flush_task(p); |
1da177e4 | 174 | release_thread(p); |
8c7904a0 | 175 | call_rcu(&p->rcu, delayed_put_task_struct); |
1da177e4 LT |
176 | |
177 | p = leader; | |
178 | if (unlikely(zap_leader)) | |
179 | goto repeat; | |
180 | } | |
181 | ||
1da177e4 LT |
182 | /* |
183 | * This checks not only the pgrp, but falls back on the pid if no | |
184 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly | |
185 | * without this... | |
186 | */ | |
187 | int session_of_pgrp(int pgrp) | |
188 | { | |
189 | struct task_struct *p; | |
190 | int sid = -1; | |
191 | ||
192 | read_lock(&tasklist_lock); | |
193 | do_each_task_pid(pgrp, PIDTYPE_PGID, p) { | |
194 | if (p->signal->session > 0) { | |
195 | sid = p->signal->session; | |
196 | goto out; | |
197 | } | |
198 | } while_each_task_pid(pgrp, PIDTYPE_PGID, p); | |
199 | p = find_task_by_pid(pgrp); | |
200 | if (p) | |
201 | sid = p->signal->session; | |
202 | out: | |
203 | read_unlock(&tasklist_lock); | |
204 | ||
205 | return sid; | |
206 | } | |
207 | ||
208 | /* | |
209 | * Determine if a process group is "orphaned", according to the POSIX | |
210 | * definition in 2.2.2.52. Orphaned process groups are not to be affected | |
211 | * by terminal-generated stop signals. Newly orphaned process groups are | |
212 | * to receive a SIGHUP and a SIGCONT. | |
213 | * | |
214 | * "I ask you, have you ever known what it is to be an orphan?" | |
215 | */ | |
36c8b586 | 216 | static int will_become_orphaned_pgrp(int pgrp, struct task_struct *ignored_task) |
1da177e4 LT |
217 | { |
218 | struct task_struct *p; | |
219 | int ret = 1; | |
220 | ||
221 | do_each_task_pid(pgrp, PIDTYPE_PGID, p) { | |
222 | if (p == ignored_task | |
223 | || p->exit_state | |
f400e198 | 224 | || is_init(p->real_parent)) |
1da177e4 LT |
225 | continue; |
226 | if (process_group(p->real_parent) != pgrp | |
227 | && p->real_parent->signal->session == p->signal->session) { | |
228 | ret = 0; | |
229 | break; | |
230 | } | |
231 | } while_each_task_pid(pgrp, PIDTYPE_PGID, p); | |
232 | return ret; /* (sighing) "Often!" */ | |
233 | } | |
234 | ||
235 | int is_orphaned_pgrp(int pgrp) | |
236 | { | |
237 | int retval; | |
238 | ||
239 | read_lock(&tasklist_lock); | |
240 | retval = will_become_orphaned_pgrp(pgrp, NULL); | |
241 | read_unlock(&tasklist_lock); | |
242 | ||
243 | return retval; | |
244 | } | |
245 | ||
858119e1 | 246 | static int has_stopped_jobs(int pgrp) |
1da177e4 LT |
247 | { |
248 | int retval = 0; | |
249 | struct task_struct *p; | |
250 | ||
251 | do_each_task_pid(pgrp, PIDTYPE_PGID, p) { | |
252 | if (p->state != TASK_STOPPED) | |
253 | continue; | |
1da177e4 LT |
254 | retval = 1; |
255 | break; | |
256 | } while_each_task_pid(pgrp, PIDTYPE_PGID, p); | |
257 | return retval; | |
258 | } | |
259 | ||
260 | /** | |
4dc3b16b | 261 | * reparent_to_init - Reparent the calling kernel thread to the init task. |
1da177e4 LT |
262 | * |
263 | * If a kernel thread is launched as a result of a system call, or if | |
264 | * it ever exits, it should generally reparent itself to init so that | |
265 | * it is correctly cleaned up on exit. | |
266 | * | |
267 | * The various task state such as scheduling policy and priority may have | |
268 | * been inherited from a user process, so we reset them to sane values here. | |
269 | * | |
270 | * NOTE that reparent_to_init() gives the caller full capabilities. | |
271 | */ | |
858119e1 | 272 | static void reparent_to_init(void) |
1da177e4 LT |
273 | { |
274 | write_lock_irq(&tasklist_lock); | |
275 | ||
276 | ptrace_unlink(current); | |
277 | /* Reparent to init */ | |
9b678ece | 278 | remove_parent(current); |
1da177e4 LT |
279 | current->parent = child_reaper; |
280 | current->real_parent = child_reaper; | |
9b678ece | 281 | add_parent(current); |
1da177e4 LT |
282 | |
283 | /* Set the exit signal to SIGCHLD so we signal init on exit */ | |
284 | current->exit_signal = SIGCHLD; | |
285 | ||
1c573afe | 286 | if (!has_rt_policy(current) && (task_nice(current) < 0)) |
1da177e4 LT |
287 | set_user_nice(current, 0); |
288 | /* cpus_allowed? */ | |
289 | /* rt_priority? */ | |
290 | /* signals? */ | |
291 | security_task_reparent_to_init(current); | |
292 | memcpy(current->signal->rlim, init_task.signal->rlim, | |
293 | sizeof(current->signal->rlim)); | |
294 | atomic_inc(&(INIT_USER->__count)); | |
295 | write_unlock_irq(&tasklist_lock); | |
296 | switch_uid(INIT_USER); | |
297 | } | |
298 | ||
299 | void __set_special_pids(pid_t session, pid_t pgrp) | |
300 | { | |
e19f247a | 301 | struct task_struct *curr = current->group_leader; |
1da177e4 LT |
302 | |
303 | if (curr->signal->session != session) { | |
304 | detach_pid(curr, PIDTYPE_SID); | |
305 | curr->signal->session = session; | |
306 | attach_pid(curr, PIDTYPE_SID, session); | |
307 | } | |
308 | if (process_group(curr) != pgrp) { | |
309 | detach_pid(curr, PIDTYPE_PGID); | |
310 | curr->signal->pgrp = pgrp; | |
311 | attach_pid(curr, PIDTYPE_PGID, pgrp); | |
312 | } | |
313 | } | |
314 | ||
315 | void set_special_pids(pid_t session, pid_t pgrp) | |
316 | { | |
317 | write_lock_irq(&tasklist_lock); | |
318 | __set_special_pids(session, pgrp); | |
319 | write_unlock_irq(&tasklist_lock); | |
320 | } | |
321 | ||
322 | /* | |
323 | * Let kernel threads use this to say that they | |
324 | * allow a certain signal (since daemonize() will | |
325 | * have disabled all of them by default). | |
326 | */ | |
327 | int allow_signal(int sig) | |
328 | { | |
7ed20e1a | 329 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
330 | return -EINVAL; |
331 | ||
332 | spin_lock_irq(¤t->sighand->siglock); | |
333 | sigdelset(¤t->blocked, sig); | |
334 | if (!current->mm) { | |
335 | /* Kernel threads handle their own signals. | |
336 | Let the signal code know it'll be handled, so | |
337 | that they don't get converted to SIGKILL or | |
338 | just silently dropped */ | |
339 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; | |
340 | } | |
341 | recalc_sigpending(); | |
342 | spin_unlock_irq(¤t->sighand->siglock); | |
343 | return 0; | |
344 | } | |
345 | ||
346 | EXPORT_SYMBOL(allow_signal); | |
347 | ||
348 | int disallow_signal(int sig) | |
349 | { | |
7ed20e1a | 350 | if (!valid_signal(sig) || sig < 1) |
1da177e4 LT |
351 | return -EINVAL; |
352 | ||
353 | spin_lock_irq(¤t->sighand->siglock); | |
354 | sigaddset(¤t->blocked, sig); | |
355 | recalc_sigpending(); | |
356 | spin_unlock_irq(¤t->sighand->siglock); | |
357 | return 0; | |
358 | } | |
359 | ||
360 | EXPORT_SYMBOL(disallow_signal); | |
361 | ||
362 | /* | |
363 | * Put all the gunge required to become a kernel thread without | |
364 | * attached user resources in one place where it belongs. | |
365 | */ | |
366 | ||
367 | void daemonize(const char *name, ...) | |
368 | { | |
369 | va_list args; | |
370 | struct fs_struct *fs; | |
371 | sigset_t blocked; | |
372 | ||
373 | va_start(args, name); | |
374 | vsnprintf(current->comm, sizeof(current->comm), name, args); | |
375 | va_end(args); | |
376 | ||
377 | /* | |
378 | * If we were started as result of loading a module, close all of the | |
379 | * user space pages. We don't need them, and if we didn't close them | |
380 | * they would be locked into memory. | |
381 | */ | |
382 | exit_mm(current); | |
383 | ||
384 | set_special_pids(1, 1); | |
70522e12 | 385 | mutex_lock(&tty_mutex); |
1da177e4 | 386 | current->signal->tty = NULL; |
70522e12 | 387 | mutex_unlock(&tty_mutex); |
1da177e4 LT |
388 | |
389 | /* Block and flush all signals */ | |
390 | sigfillset(&blocked); | |
391 | sigprocmask(SIG_BLOCK, &blocked, NULL); | |
392 | flush_signals(current); | |
393 | ||
394 | /* Become as one with the init task */ | |
395 | ||
396 | exit_fs(current); /* current->fs->count--; */ | |
397 | fs = init_task.fs; | |
398 | current->fs = fs; | |
399 | atomic_inc(&fs->count); |