Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/kernel/exit.c |
| 3 | * |
| 4 | * Copyright (C) 1991, 1992 Linus Torvalds |
| 5 | */ |
| 6 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | #include <linux/mm.h> |
| 8 | #include <linux/slab.h> |
| 9 | #include <linux/interrupt.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 10 | #include <linux/module.h> |
Randy.Dunlap | c59ede7 | 2006-01-11 12:17:46 -0800 | [diff] [blame] | 11 | #include <linux/capability.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 12 | #include <linux/completion.h> |
| 13 | #include <linux/personality.h> |
| 14 | #include <linux/tty.h> |
Jens Axboe | da9cbc8 | 2008-06-30 20:42:08 +0200 | [diff] [blame] | 15 | #include <linux/iocontext.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 16 | #include <linux/key.h> |
| 17 | #include <linux/security.h> |
| 18 | #include <linux/cpu.h> |
| 19 | #include <linux/acct.h> |
Jay Lan | 8f0ab51 | 2006-09-30 23:28:59 -0700 | [diff] [blame] | 20 | #include <linux/tsacct_kern.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 21 | #include <linux/file.h> |
Al Viro | 9f3acc3 | 2008-04-24 07:44:08 -0400 | [diff] [blame] | 22 | #include <linux/fdtable.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 23 | #include <linux/binfmts.h> |
Serge E. Hallyn | ab51601 | 2006-10-02 02:18:06 -0700 | [diff] [blame] | 24 | #include <linux/nsproxy.h> |
Sukadev Bhattiprolu | 84d7378 | 2006-12-08 02:38:01 -0800 | [diff] [blame] | 25 | #include <linux/pid_namespace.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 26 | #include <linux/ptrace.h> |
| 27 | #include <linux/profile.h> |
| 28 | #include <linux/mount.h> |
| 29 | #include <linux/proc_fs.h> |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 30 | #include <linux/kthread.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | #include <linux/mempolicy.h> |
Shailabh Nagar | c757249 | 2006-07-14 00:24:40 -0700 | [diff] [blame] | 32 | #include <linux/taskstats_kern.h> |
Shailabh Nagar | ca74e92 | 2006-07-14 00:24:36 -0700 | [diff] [blame] | 33 | #include <linux/delayacct.h> |
Rafael J. Wysocki | 8314418 | 2007-07-17 04:03:35 -0700 | [diff] [blame] | 34 | #include <linux/freezer.h> |
Paul Menage | b4f48b6 | 2007-10-18 23:39:33 -0700 | [diff] [blame] | 35 | #include <linux/cgroup.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 36 | #include <linux/syscalls.h> |
Jesper Juhl | 7ed20e1 | 2005-05-01 08:59:14 -0700 | [diff] [blame] | 37 | #include <linux/signal.h> |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 38 | #include <linux/posix-timers.h> |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 39 | #include <linux/cn_proc.h> |
Ingo Molnar | de5097c | 2006-01-09 15:59:21 -0800 | [diff] [blame] | 40 | #include <linux/mutex.h> |
Ingo Molnar | 0771dfe | 2006-03-27 01:16:22 -0800 | [diff] [blame] | 41 | #include <linux/futex.h> |
Jens Axboe | b92ce55 | 2006-04-11 13:52:07 +0200 | [diff] [blame] | 42 | #include <linux/pipe_fs_i.h> |
Al Viro | fa84cb9 | 2006-03-29 20:30:19 -0500 | [diff] [blame] | 43 | #include <linux/audit.h> /* for audit_free() */ |
Adrian Bunk | 83cc5ed | 2006-06-25 05:47:41 -0700 | [diff] [blame] | 44 | #include <linux/resource.h> |
David Howells | 0d67a46 | 2006-08-29 19:05:56 +0100 | [diff] [blame] | 45 | #include <linux/blkdev.h> |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 46 | #include <linux/task_io_accounting_ops.h> |
Roland McGrath | 30199f5 | 2008-07-25 19:45:46 -0700 | [diff] [blame] | 47 | #include <linux/tracehook.h> |
Al Viro | 5ad4e53 | 2009-03-29 19:50:06 -0400 | [diff] [blame] | 48 | #include <linux/fs_struct.h> |
David Howells | d84f4f9 | 2008-11-14 10:39:23 +1100 | [diff] [blame] | 49 | #include <linux/init_task.h> |
Ingo Molnar | cdd6c48 | 2009-09-21 12:02:48 +0200 | [diff] [blame] | 50 | #include <linux/perf_event.h> |
Steven Rostedt | ad8d75f | 2009-04-14 19:39:12 -0400 | [diff] [blame] | 51 | #include <trace/events/sched.h> |
Frederic Weisbecker | 24f1e32c | 2009-09-09 19:22:48 +0200 | [diff] [blame] | 52 | #include <linux/hw_breakpoint.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 53 | |
| 54 | #include <asm/uaccess.h> |
| 55 | #include <asm/unistd.h> |
| 56 | #include <asm/pgtable.h> |
| 57 | #include <asm/mmu_context.h> |
| 58 | |
Adrian Bunk | 408b664 | 2005-05-01 08:59:29 -0700 | [diff] [blame] | 59 | static void exit_mm(struct task_struct * tsk); |
| 60 | |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 61 | static void __unhash_process(struct task_struct *p, bool group_dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 62 | { |
| 63 | nr_threads--; |
| 64 | detach_pid(p, PIDTYPE_PID); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 65 | if (group_dead) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 66 | detach_pid(p, PIDTYPE_PGID); |
| 67 | detach_pid(p, PIDTYPE_SID); |
Oleg Nesterov | c97d989 | 2006-03-28 16:11:06 -0800 | [diff] [blame] | 68 | |
Eric W. Biederman | 5e85d4a | 2006-04-18 22:20:16 -0700 | [diff] [blame] | 69 | list_del_rcu(&p->tasks); |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 70 | list_del_init(&p->sibling); |
Oleg Nesterov | 73b9ebf | 2006-03-28 16:11:07 -0800 | [diff] [blame] | 71 | __get_cpu_var(process_counts)--; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 72 | } |
Oleg Nesterov | 47e6532 | 2006-03-28 16:11:25 -0800 | [diff] [blame] | 73 | list_del_rcu(&p->thread_group); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 74 | } |
| 75 | |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 76 | /* |
| 77 | * This function expects the tasklist_lock write-locked. |
| 78 | */ |
| 79 | static void __exit_signal(struct task_struct *tsk) |
| 80 | { |
| 81 | struct signal_struct *sig = tsk->signal; |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 82 | bool group_dead = thread_group_leader(tsk); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 83 | struct sighand_struct *sighand; |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 84 | struct tty_struct *uninitialized_var(tty); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 85 | |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 86 | sighand = rcu_dereference_check(tsk->sighand, |
| 87 | rcu_read_lock_held() || |
Paul E. McKenney | db1466b | 2010-03-03 07:46:56 -0800 | [diff] [blame] | 88 | lockdep_tasklist_lock_is_held()); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 89 | spin_lock(&sighand->siglock); |
| 90 | |
| 91 | posix_cpu_timers_exit(tsk); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 92 | if (group_dead) { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 93 | posix_cpu_timers_exit_group(tsk); |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 94 | tty = sig->tty; |
| 95 | sig->tty = NULL; |
Oleg Nesterov | 4a59994 | 2010-05-26 14:43:12 -0700 | [diff] [blame] | 96 | } else { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 97 | /* |
| 98 | * If there is any task waiting for the group exit |
| 99 | * then notify it: |
| 100 | */ |
Oleg Nesterov | d344193 | 2010-05-26 14:43:11 -0700 | [diff] [blame] | 101 | if (sig->notify_count > 0 && !--sig->notify_count) |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 102 | wake_up_process(sig->group_exit_task); |
Oleg Nesterov | 6db840f | 2007-10-16 23:27:23 -0700 | [diff] [blame] | 103 | |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 104 | if (tsk == sig->curr_target) |
| 105 | sig->curr_target = next_thread(tsk); |
| 106 | /* |
| 107 | * Accumulate here the counters for all threads but the |
| 108 | * group leader as they die, so they can be added into |
| 109 | * the process-wide totals when those are taken. |
| 110 | * The group leader stays around as a zombie as long |
| 111 | * as there are other threads. When it gets reaped, |
| 112 | * the exit.c code will add its counts into these totals. |
| 113 | * We won't ever get here for the group leader, since it |
| 114 | * will have been the last reference on the signal_struct. |
| 115 | */ |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 116 | sig->utime = cputime_add(sig->utime, tsk->utime); |
| 117 | sig->stime = cputime_add(sig->stime, tsk->stime); |
Hidetoshi Seto | d5b7c78 | 2009-11-26 14:49:05 +0900 | [diff] [blame] | 118 | sig->gtime = cputime_add(sig->gtime, tsk->gtime); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 119 | sig->min_flt += tsk->min_flt; |
| 120 | sig->maj_flt += tsk->maj_flt; |
| 121 | sig->nvcsw += tsk->nvcsw; |
| 122 | sig->nivcsw += tsk->nivcsw; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 123 | sig->inblock += task_io_get_inblock(tsk); |
| 124 | sig->oublock += task_io_get_oublock(tsk); |
Andrea Righi | 5995477 | 2008-07-27 17:29:15 +0200 | [diff] [blame] | 125 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
Peter Zijlstra | 32bd671 | 2009-02-05 12:24:15 +0100 | [diff] [blame] | 126 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 127 | } |
| 128 | |
Oleg Nesterov | b3ac022 | 2010-05-26 14:43:24 -0700 | [diff] [blame] | 129 | sig->nr_threads--; |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 130 | __unhash_process(tsk, group_dead); |
Oleg Nesterov | 5876700 | 2006-03-28 16:11:20 -0800 | [diff] [blame] | 131 | |
Oleg Nesterov | da7978b | 2008-05-23 13:04:41 -0700 | [diff] [blame] | 132 | /* |
| 133 | * Do this under ->siglock, we can race with another thread |
| 134 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. |
| 135 | */ |
| 136 | flush_sigqueue(&tsk->pending); |
Oleg Nesterov | a7e5328 | 2006-03-28 16:11:27 -0800 | [diff] [blame] | 137 | tsk->sighand = NULL; |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 138 | spin_unlock(&sighand->siglock); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 139 | |
Oleg Nesterov | a7e5328 | 2006-03-28 16:11:27 -0800 | [diff] [blame] | 140 | __cleanup_sighand(sighand); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 141 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 142 | if (group_dead) { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 143 | flush_sigqueue(&sig->shared_pending); |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 144 | tty_kref_put(tty); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 145 | } |
| 146 | } |
| 147 | |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 148 | static void delayed_put_task_struct(struct rcu_head *rhp) |
| 149 | { |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 150 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
| 151 | |
Ingo Molnar | cdd6c48 | 2009-09-21 12:02:48 +0200 | [diff] [blame] | 152 | #ifdef CONFIG_PERF_EVENTS |
| 153 | WARN_ON_ONCE(tsk->perf_event_ctxp); |
Ingo Molnar | eef6cbf | 2008-12-19 10:20:42 +0100 | [diff] [blame] | 154 | #endif |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 155 | trace_sched_process_free(tsk); |
| 156 | put_task_struct(tsk); |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 157 | } |
| 158 | |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 159 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 160 | void release_task(struct task_struct * p) |
| 161 | { |
Ingo Molnar | 36c8b58 | 2006-07-03 00:25:41 -0700 | [diff] [blame] | 162 | struct task_struct *leader; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 163 | int zap_leader; |
Oleg Nesterov | 1f09f97 | 2006-03-28 16:11:11 -0800 | [diff] [blame] | 164 | repeat: |
Roland McGrath | dae3357 | 2008-07-25 19:45:48 -0700 | [diff] [blame] | 165 | tracehook_prepare_release_task(p); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 166 | /* don't need to get the RCU readlock here - the process is dead and |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 167 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
| 168 | rcu_read_lock(); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 169 | atomic_dec(&__task_cred(p)->user->processes); |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 170 | rcu_read_unlock(); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 171 | |
Pavel Emelyanov | 60347f6 | 2007-10-18 23:40:03 -0700 | [diff] [blame] | 172 | proc_flush_task(p); |
Ingo Molnar | 0203026 | 2009-05-17 11:24:08 +0200 | [diff] [blame] | 173 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 174 | write_lock_irq(&tasklist_lock); |
Roland McGrath | dae3357 | 2008-07-25 19:45:48 -0700 | [diff] [blame] | 175 | tracehook_finish_release_task(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 176 | __exit_signal(p); |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 177 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 178 | /* |
| 179 | * If we are the last non-leader member of the thread |
| 180 | * group, and the leader is zombie, then notify the |
| 181 | * group leader's parent process. (if it wants notification.) |
| 182 | */ |
| 183 | zap_leader = 0; |
| 184 | leader = p->group_leader; |
| 185 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 186 | BUG_ON(task_detached(leader)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 187 | do_notify_parent(leader, leader->exit_signal); |
| 188 | /* |
| 189 | * If we were the last child thread and the leader has |
| 190 | * exited already, and the leader's parent ignores SIGCHLD, |
| 191 | * then we are the one who should release the leader. |
| 192 | * |
| 193 | * do_notify_parent() will have marked it self-reaping in |
| 194 | * that case. |
| 195 | */ |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 196 | zap_leader = task_detached(leader); |
Roland McGrath | dae3357 | 2008-07-25 19:45:48 -0700 | [diff] [blame] | 197 | |
| 198 | /* |
| 199 | * This maintains the invariant that release_task() |
| 200 | * only runs on a task in EXIT_DEAD, just for sanity. |
| 201 | */ |
| 202 | if (zap_leader) |
| 203 | leader->exit_state = EXIT_DEAD; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 204 | } |
| 205 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 206 | write_unlock_irq(&tasklist_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 207 | release_thread(p); |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 208 | call_rcu(&p->rcu, delayed_put_task_struct); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 209 | |
| 210 | p = leader; |
| 211 | if (unlikely(zap_leader)) |
| 212 | goto repeat; |
| 213 | } |
| 214 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 215 | /* |
| 216 | * This checks not only the pgrp, but falls back on the pid if no |
| 217 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly |
| 218 | * without this... |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 219 | * |
| 220 | * The caller must hold rcu lock or the tasklist lock. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 221 | */ |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 222 | struct pid *session_of_pgrp(struct pid *pgrp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 223 | { |
| 224 | struct task_struct *p; |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 225 | struct pid *sid = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 226 | |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 227 | p = pid_task(pgrp, PIDTYPE_PGID); |
Oleg Nesterov | 62dfb55 | 2006-12-08 02:38:03 -0800 | [diff] [blame] | 228 | if (p == NULL) |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 229 | p = pid_task(pgrp, PIDTYPE_PID); |
Oleg Nesterov | 62dfb55 | 2006-12-08 02:38:03 -0800 | [diff] [blame] | 230 | if (p != NULL) |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 231 | sid = task_session(p); |
Oleg Nesterov | 62dfb55 | 2006-12-08 02:38:03 -0800 | [diff] [blame] | 232 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 233 | return sid; |
| 234 | } |
| 235 | |
| 236 | /* |
| 237 | * Determine if a process group is "orphaned", according to the POSIX |
| 238 | * definition in 2.2.2.52. Orphaned process groups are not to be affected |
| 239 | * by terminal-generated stop signals. Newly orphaned process groups are |
| 240 | * to receive a SIGHUP and a SIGCONT. |
| 241 | * |
| 242 | * "I ask you, have you ever known what it is to be an orphan?" |
| 243 | */ |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 244 | static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 245 | { |
| 246 | struct task_struct *p; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 247 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 248 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 249 | if ((p == ignored_task) || |
| 250 | (p->exit_state && thread_group_empty(p)) || |
| 251 | is_global_init(p->real_parent)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 252 | continue; |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 253 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 254 | if (task_pgrp(p->real_parent) != pgrp && |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 255 | task_session(p->real_parent) == task_session(p)) |
| 256 | return 0; |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 257 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 258 | |
| 259 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 260 | } |
| 261 | |
Eric W. Biederman | 3e7cd6c | 2007-02-12 00:52:58 -0800 | [diff] [blame] | 262 | int is_current_pgrp_orphaned(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 263 | { |
| 264 | int retval; |
| 265 | |
| 266 | read_lock(&tasklist_lock); |
Eric W. Biederman | 3e7cd6c | 2007-02-12 00:52:58 -0800 | [diff] [blame] | 267 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 268 | read_unlock(&tasklist_lock); |
| 269 | |
| 270 | return retval; |
| 271 | } |
| 272 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 273 | static int has_stopped_jobs(struct pid *pgrp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 274 | { |
| 275 | int retval = 0; |
| 276 | struct task_struct *p; |
| 277 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 278 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Matthew Wilcox | 338077e | 2007-12-06 11:09:35 -0500 | [diff] [blame] | 279 | if (!task_is_stopped(p)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 280 | continue; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 281 | retval = 1; |
| 282 | break; |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 283 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 284 | return retval; |
| 285 | } |
| 286 | |
Oleg Nesterov | f49ee50 | 2008-03-02 21:44:40 +0300 | [diff] [blame] | 287 | /* |
| 288 | * Check to see if any process groups have become orphaned as |
| 289 | * a result of our exiting, and if they have any stopped jobs, |
| 290 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
| 291 | */ |
| 292 | static void |
| 293 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) |
| 294 | { |
| 295 | struct pid *pgrp = task_pgrp(tsk); |
| 296 | struct task_struct *ignored_task = tsk; |
| 297 | |
| 298 | if (!parent) |
| 299 | /* exit: our father is in a different pgrp than |
| 300 | * we are and we were the only connection outside. |
| 301 | */ |
| 302 | parent = tsk->real_parent; |
| 303 | else |
| 304 | /* reparent: our child is in a different pgrp than |
| 305 | * we are, and it was the only connection outside. |
| 306 | */ |
| 307 | ignored_task = NULL; |
| 308 | |
| 309 | if (task_pgrp(parent) != pgrp && |
| 310 | task_session(parent) == task_session(tsk) && |
| 311 | will_become_orphaned_pgrp(pgrp, ignored_task) && |
| 312 | has_stopped_jobs(pgrp)) { |
| 313 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); |
| 314 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); |
| 315 | } |
| 316 | } |
| 317 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 318 | /** |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 319 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 320 | * |
| 321 | * If a kernel thread is launched as a result of a system call, or if |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 322 | * it ever exits, it should generally reparent itself to kthreadd so it |
| 323 | * isn't in the way of other processes and is correctly cleaned up on exit. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 324 | * |
| 325 | * The various task state such as scheduling policy and priority may have |
| 326 | * been inherited from a user process, so we reset them to sane values here. |
| 327 | * |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 328 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 329 | */ |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 330 | static void reparent_to_kthreadd(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 331 | { |
| 332 | write_lock_irq(&tasklist_lock); |
| 333 | |
| 334 | ptrace_unlink(current); |
| 335 | /* Reparent to init */ |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 336 | current->real_parent = current->parent = kthreadd_task; |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 337 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 338 | |
| 339 | /* Set the exit signal to SIGCHLD so we signal init on exit */ |
| 340 | current->exit_signal = SIGCHLD; |
| 341 | |
Ingo Molnar | e05606d | 2007-07-09 18:51:59 +0200 | [diff] [blame] | 342 | if (task_nice(current) < 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 343 | set_user_nice(current, 0); |
| 344 | /* cpus_allowed? */ |
| 345 | /* rt_priority? */ |
| 346 | /* signals? */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 347 | memcpy(current->signal->rlim, init_task.signal->rlim, |
| 348 | sizeof(current->signal->rlim)); |
David Howells | d84f4f9 | 2008-11-14 10:39:23 +1100 | [diff] [blame] | 349 | |
| 350 | atomic_inc(&init_cred.usage); |
| 351 | commit_creds(&init_cred); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 352 | write_unlock_irq(&tasklist_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 353 | } |
| 354 | |
Oleg Nesterov | 8520d7c | 2008-02-08 04:19:09 -0800 | [diff] [blame] | 355 | void __set_special_pids(struct pid *pid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 356 | { |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 357 | struct task_struct *curr = current->group_leader; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 358 | |
Christian Borntraeger | 0d0df59 | 2009-10-26 16:49:34 -0700 | [diff] [blame] | 359 | if (task_session(curr) != pid) |
Oleg Nesterov | 7d8da09 | 2008-04-30 00:54:27 -0700 | [diff] [blame] | 360 | change_pid(curr, PIDTYPE_SID, pid); |
Oleg Nesterov | 1b0f7ff | 2009-04-02 16:58:39 -0700 | [diff] [blame] | 361 | |
| 362 | if (task_pgrp(curr) != pid) |
Oleg Nesterov | 7d8da09 | 2008-04-30 00:54:27 -0700 | [diff] [blame] | 363 | change_pid(curr, PIDTYPE_PGID, pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 364 | } |
| 365 | |
Oleg Nesterov | 8520d7c | 2008-02-08 04:19:09 -0800 | [diff] [blame] | 366 | static void set_special_pids(struct pid *pid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 367 | { |
| 368 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 8520d7c | 2008-02-08 04:19:09 -0800 | [diff] [blame] | 369 | __set_special_pids(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 370 | write_unlock_irq(&tasklist_lock); |
| 371 | } |
| 372 | |
| 373 | /* |
Oleg Nesterov | 8724513 | 2009-06-17 16:27:23 -0700 | [diff] [blame] | 374 | * Let kernel threads use this to say that they allow a certain signal. |
| 375 | * Must not be used if kthread was cloned with CLONE_SIGHAND. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 376 | */ |
| 377 | int allow_signal(int sig) |
| 378 | { |
Jesper Juhl | 7ed20e1 | 2005-05-01 08:59:14 -0700 | [diff] [blame] | 379 | if (!valid_signal(sig) || sig < 1) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 380 | return -EINVAL; |
| 381 | |
| 382 | spin_lock_irq(¤t->sighand->siglock); |
Oleg Nesterov | 8724513 | 2009-06-17 16:27:23 -0700 | [diff] [blame] | 383 | /* This is only needed for daemonize()'ed kthreads */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 384 | sigdelset(¤t->blocked, sig); |
Oleg Nesterov | 8724513 | 2009-06-17 16:27:23 -0700 | [diff] [blame] | 385 | /* |
| 386 | * Kernel threads handle their own signals. Let the signal code |
| 387 | * know it'll be handled, so that they don't get converted to |
| 388 | * SIGKILL or just silently dropped. |
| 389 | */ |
| 390 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 391 | recalc_sigpending(); |
| 392 | spin_unlock_irq(¤t->sighand->siglock); |
| 393 | return 0; |
| 394 | } |
| 395 | |
| 396 | EXPORT_SYMBOL(allow_signal); |
| 397 | |
| 398 | int disallow_signal(int sig) |
| 399 | { |
Jesper Juhl | 7ed20e1 | 2005-05-01 08:59:14 -0700 | [diff] [blame] | 400 | if (!valid_signal(sig) || sig < 1) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 401 | return -EINVAL; |
| 402 | |
| 403 | spin_lock_irq(¤t->sighand->siglock); |
Oleg Nesterov | 10ab825 | 2007-05-09 02:34:37 -0700 | [diff] [blame] | 404 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 405 | recalc_sigpending(); |
| 406 | spin_unlock_irq(¤t->sighand->siglock); |
| 407 | return 0; |
| 408 | } |
| 409 | |
| 410 | EXPORT_SYMBOL(disallow_signal); |
| 411 | |
| 412 | /* |
| 413 | * Put all the gunge required to become a kernel thread without |
| 414 | * attached user resources in one place where it belongs. |
| 415 | */ |
| 416 | |
| 417 | void daemonize(const char *name, ...) |
| 418 | { |
| 419 | va_list args; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 420 | sigset_t blocked; |
| 421 | |
| 422 | va_start(args, name); |
| 423 | vsnprintf(current->comm, sizeof(current->comm), name, args); |
| 424 | va_end(args); |
| 425 | |
| 426 | /* |
| 427 | * If we were started as result of loading a module, close all of the |
| 428 | * user space pages. We don't need them, and if we didn't close them |
| 429 | * they would be locked into memory. |
| 430 | */ |
| 431 | exit_mm(current); |
Rafael J. Wysocki | 8314418 | 2007-07-17 04:03:35 -0700 | [diff] [blame] | 432 | /* |
| 433 | * We don't want to have TIF_FREEZE set if the system-wide hibernation |
| 434 | * or suspend transition begins right now. |
| 435 | */ |
Oleg Nesterov | 7b34e42 | 2008-07-25 01:47:37 -0700 | [diff] [blame] | 436 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 437 | |
Oleg Nesterov | 8520d7c | 2008-02-08 04:19:09 -0800 | [diff] [blame] | 438 | if (current->nsproxy != &init_nsproxy) { |
| 439 | get_nsproxy(&init_nsproxy); |
| 440 | switch_task_namespaces(current, &init_nsproxy); |
| 441 | } |
Oleg Nesterov | 297bd42 | 2008-02-08 04:19:10 -0800 | [diff] [blame] | 442 | set_special_pids(&init_struct_pid); |
Peter Zijlstra | 24ec839 | 2006-12-08 02:36:04 -0800 | [diff] [blame] | 443 | proc_clear_tty(current); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 444 | |
| 445 | /* Block and flush all signals */ |
| 446 | sigfillset(&blocked); |
| 447 | sigprocmask(SIG_BLOCK, &blocked, NULL); |
| 448 | flush_signals(current); |
| 449 | |
| 450 | /* Become as one with the init task */ |
| 451 | |
Al Viro | 3e93cd6 | 2009-03-29 19:00:13 -0400 | [diff] [blame] | 452 | daemonize_fs_struct(); |
Daniel Walker | d4c5e41 | 2007-10-18 23:39:59 -0700 | [diff] [blame] | 453 | exit_files(current); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 454 | current->files = init_task.files; |
| 455 | atomic_inc(¤t->files->count); |
| 456 | |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 457 | reparent_to_kthreadd(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 458 | } |
| 459 | |
| 460 | EXPORT_SYMBOL(daemonize); |
| 461 | |
Arjan van de Ven | 858119e | 2006-01-14 13:20:43 -0800 | [diff] [blame] | 462 | static void close_files(struct files_struct * files) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 463 | { |
| 464 | int i, j; |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 465 | struct fdtable *fdt; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 466 | |
| 467 | j = 0; |
Dipankar Sarma | 4fb3a53 | 2005-09-16 19:28:13 -0700 | [diff] [blame] | 468 | |
| 469 | /* |
| 470 | * It is safe to dereference the fd table without RCU or |
| 471 | * ->file_lock because this is the last reference to the |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 472 | * files structure. But use RCU to shut RCU-lockdep up. |
Dipankar Sarma | 4fb3a53 | 2005-09-16 19:28:13 -0700 | [diff] [blame] | 473 | */ |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 474 | rcu_read_lock(); |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 475 | fdt = files_fdtable(files); |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 476 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 477 | for (;;) { |
| 478 | unsigned long set; |
| 479 | i = j * __NFDBITS; |
Vadim Lobanov | bbea9f6 | 2006-12-10 02:21:12 -0800 | [diff] [blame] | 480 | if (i >= fdt->max_fds) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 481 | break; |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 482 | set = fdt->open_fds->fds_bits[j++]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 483 | while (set) { |
| 484 | if (set & 1) { |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 485 | struct file * file = xchg(&fdt->fd[i], NULL); |
Ingo Molnar | 944be0b | 2007-02-12 00:52:26 -0800 | [diff] [blame] | 486 | if (file) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 487 | filp_close(file, files); |
Ingo Molnar | 944be0b | 2007-02-12 00:52:26 -0800 | [diff] [blame] | 488 | cond_resched(); |
| 489 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 490 | } |
| 491 | i++; |
| 492 | set >>= 1; |
| 493 | } |
| 494 | } |
| 495 | } |
| 496 | |
| 497 | struct files_struct *get_files_struct(struct task_struct *task) |
| 498 | { |
| 499 | struct files_struct *files; |
| 500 | |
| 501 | task_lock(task); |
| 502 | files = task->files; |
| 503 | if (files) |
| 504 | atomic_inc(&files->count); |
| 505 | task_unlock(task); |
| 506 | |
| 507 | return files; |
| 508 | } |
| 509 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 510 | void put_files_struct(struct files_struct *files) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 511 | { |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 512 | struct fdtable *fdt; |
| 513 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 514 | if (atomic_dec_and_test(&files->count)) { |
| 515 | close_files(files); |
| 516 | /* |
| 517 | * Free the fd and fdset arrays if we expanded them. |
Dipankar Sarma | ab2af1f | 2005-09-09 13:04:13 -0700 | [diff] [blame] | 518 | * If the fdtable was embedded, pass files for freeing |
| 519 | * at the end of the RCU grace period. Otherwise, |
| 520 | * you can free files immediately. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 521 | */ |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 522 | rcu_read_lock(); |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 523 | fdt = files_fdtable(files); |
Vadim Lobanov | 4fd4581 | 2006-12-10 02:21:17 -0800 | [diff] [blame] | 524 | if (fdt != &files->fdtab) |
Dipankar Sarma | ab2af1f | 2005-09-09 13:04:13 -0700 | [diff] [blame] | 525 | kmem_cache_free(files_cachep, files); |
Vadim Lobanov | 01b2d93 | 2006-12-22 01:10:43 -0800 | [diff] [blame] | 526 | free_fdtable(fdt); |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 527 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 528 | } |
| 529 | } |
| 530 | |
Al Viro | 3b12538 | 2008-04-22 05:31:30 -0400 | [diff] [blame] | 531 | void reset_files_struct(struct files_struct *files) |
Kirill Korotaev | 3b9b8ab | 2006-09-29 02:00:05 -0700 | [diff] [blame] | 532 | { |
Al Viro | 3b12538 | 2008-04-22 05:31:30 -0400 | [diff] [blame] | 533 | struct task_struct *tsk = current; |
Kirill Korotaev | 3b9b8ab | 2006-09-29 02:00:05 -0700 | [diff] [blame] | 534 | struct files_struct *old; |
| 535 | |
| 536 | old = tsk->files; |
| 537 | task_lock(tsk); |
| 538 | tsk->files = files; |
| 539 | task_unlock(tsk); |
| 540 | put_files_struct(old); |
| 541 | } |
Kirill Korotaev | 3b9b8ab | 2006-09-29 02:00:05 -0700 | [diff] [blame] | 542 | |
Al Viro | 1ec7f1d | 2008-04-22 05:35:42 -0400 | [diff] [blame] | 543 | void exit_files(struct task_struct *tsk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 544 | { |
| 545 | struct files_struct * files = tsk->files; |
| 546 | |
| 547 | if (files) { |
| 548 | task_lock(tsk); |
| 549 | tsk->files = NULL; |
| 550 | task_unlock(tsk); |
| 551 | put_files_struct(files); |
| 552 | } |
| 553 | } |
| 554 | |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 555 | #ifdef CONFIG_MM_OWNER |
| 556 | /* |
| 557 | * Task p is exiting and it owned mm, lets find a new owner for it |
| 558 | */ |
| 559 | static inline int |
| 560 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) |
| 561 | { |
| 562 | /* |
| 563 | * If there are other users of the mm and the owner (us) is exiting |
| 564 | * we need to find a new owner to take on the responsibility. |
| 565 | */ |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 566 | if (atomic_read(&mm->mm_users) <= 1) |
| 567 | return 0; |
| 568 | if (mm->owner != p) |
| 569 | return 0; |
| 570 | return 1; |
| 571 | } |
| 572 | |
| 573 | void mm_update_next_owner(struct mm_struct *mm) |
| 574 | { |
| 575 | struct task_struct *c, *g, *p = current; |
| 576 | |
| 577 | retry: |
| 578 | if (!mm_need_new_owner(mm, p)) |
| 579 | return; |
| 580 | |
| 581 | read_lock(&tasklist_lock); |
| 582 | /* |
| 583 | * Search in the children |
| 584 | */ |
| 585 | list_for_each_entry(c, &p->children, sibling) { |
| 586 | if (c->mm == mm) |
| 587 | goto assign_new_owner; |
| 588 | } |
| 589 | |
| 590 | /* |
| 591 | * Search in the siblings |
| 592 | */ |
Oleg Nesterov | dea33cf | 2009-06-17 16:27:29 -0700 | [diff] [blame] | 593 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 594 | if (c->mm == mm) |
| 595 | goto assign_new_owner; |
| 596 | } |
| 597 | |
| 598 | /* |
| 599 | * Search through everything else. We should not get |
| 600 | * here often |
| 601 | */ |
| 602 | do_each_thread(g, c) { |
| 603 | if (c->mm == mm) |
| 604 | goto assign_new_owner; |
| 605 | } while_each_thread(g, c); |
| 606 | |
| 607 | read_unlock(&tasklist_lock); |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 608 | /* |
| 609 | * We found no owner yet mm_users > 1: this implies that we are |
| 610 | * most likely racing with swapoff (try_to_unuse()) or /proc or |
Hugh Dickins | e599137 | 2009-01-06 14:39:22 -0800 | [diff] [blame] | 611 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 612 | */ |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 613 | mm->owner = NULL; |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 614 | return; |
| 615 | |
| 616 | assign_new_owner: |
| 617 | BUG_ON(c == p); |
| 618 | get_task_struct(c); |
| 619 | /* |
| 620 | * The task_lock protects c->mm from changing. |
| 621 | * We always want mm->owner->mm == mm |
| 622 | */ |
| 623 | task_lock(c); |
Hugh Dickins | e599137 | 2009-01-06 14:39:22 -0800 | [diff] [blame] | 624 | /* |
| 625 | * Delay read_unlock() till we have the task_lock() |
| 626 | * to ensure that c does not slip away underneath us |
| 627 | */ |
| 628 | read_unlock(&tasklist_lock); |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 629 | if (c->mm != mm) { |
| 630 | task_unlock(c); |
| 631 | put_task_struct(c); |
| 632 | goto retry; |
| 633 | } |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 634 | mm->owner = c; |
| 635 | task_unlock(c); |
| 636 | put_task_struct(c); |
| 637 | } |
| 638 | #endif /* CONFIG_MM_OWNER */ |
| 639 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 640 | /* |
| 641 | * Turn us into a lazy TLB process if we |
| 642 | * aren't already.. |
| 643 | */ |
Adrian Bunk | 408b664 | 2005-05-01 08:59:29 -0700 | [diff] [blame] | 644 | static void exit_mm(struct task_struct * tsk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 645 | { |
| 646 | struct mm_struct *mm = tsk->mm; |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 647 | struct core_state *core_state; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 648 | |
| 649 | mm_release(tsk, mm); |
| 650 | if (!mm) |
| 651 | return; |
| 652 | /* |
| 653 | * Serialize with any possible pending coredump. |
Oleg Nesterov | 999d9fc | 2008-07-25 01:47:41 -0700 | [diff] [blame] | 654 | * We must hold mmap_sem around checking core_state |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 655 | * and clearing tsk->mm. The core-inducing thread |
Oleg Nesterov | 999d9fc | 2008-07-25 01:47:41 -0700 | [diff] [blame] | 656 | * will increment ->nr_threads for each thread in the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 657 | * group with ->mm != NULL. |
| 658 | */ |
| 659 | down_read(&mm->mmap_sem); |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 660 | core_state = mm->core_state; |
| 661 | if (core_state) { |
| 662 | struct core_thread self; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 663 | up_read(&mm->mmap_sem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 664 | |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 665 | self.task = tsk; |
| 666 | self.next = xchg(&core_state->dumper.next, &self); |
| 667 | /* |
| 668 | * Implies mb(), the result of xchg() must be visible |
| 669 | * to core_state->dumper. |
| 670 | */ |
| 671 | if (atomic_dec_and_test(&core_state->nr_threads)) |
| 672 | complete(&core_state->startup); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 673 | |
Oleg Nesterov | a94e2d4 | 2008-07-25 01:47:46 -0700 | [diff] [blame] | 674 | for (;;) { |
| 675 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
| 676 | if (!self.task) /* see coredump_finish() */ |
| 677 | break; |
| 678 | schedule(); |
| 679 | } |
| 680 | __set_task_state(tsk, TASK_RUNNING); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 681 | down_read(&mm->mmap_sem); |
| 682 | } |
| 683 | atomic_inc(&mm->mm_count); |
Eric Sesterhenn | 125e187 | 2006-06-23 02:06:06 -0700 | [diff] [blame] | 684 | BUG_ON(mm != tsk->active_mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 685 | /* more a memory barrier than a real lock */ |
| 686 | task_lock(tsk); |
| 687 | tsk->mm = NULL; |
| 688 | up_read(&mm->mmap_sem); |
| 689 | enter_lazy_tlb(mm, current); |
Rafael J. Wysocki | 0c1eecf | 2007-07-19 01:47:33 -0700 | [diff] [blame] | 690 | /* We don't want this task to be frozen prematurely */ |
| 691 | clear_freeze_flag(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 692 | task_unlock(tsk); |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 693 | mm_update_next_owner(mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 694 | mmput(mm); |
| 695 | } |
| 696 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 697 | /* |
| 698 | * When we die, we re-parent all our children. |
| 699 | * Try to give them to another thread in our thread |
| 700 | * group, and if no such member exists, give it to |
Sukadev Bhattiprolu | 84d7378 | 2006-12-08 02:38:01 -0800 | [diff] [blame] | 701 | * the child reaper process (ie "init") in our pid |
| 702 | * space. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 703 | */ |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 704 | static struct task_struct *find_new_reaper(struct task_struct *father) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 705 | { |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 706 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
| 707 | struct task_struct *thread; |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 708 | |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 709 | thread = father; |
| 710 | while_each_thread(father, thread) { |
| 711 | if (thread->flags & PF_EXITING) |
| 712 | continue; |
| 713 | if (unlikely(pid_ns->child_reaper == father)) |
| 714 | pid_ns->child_reaper = thread; |
| 715 | return thread; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 716 | } |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 717 | |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 718 | if (unlikely(pid_ns->child_reaper == father)) { |
| 719 | write_unlock_irq(&tasklist_lock); |
| 720 | if (unlikely(pid_ns == &init_pid_ns)) |
| 721 | panic("Attempted to kill init!"); |
| 722 | |
| 723 | zap_pid_ns_processes(pid_ns); |
| 724 | write_lock_irq(&tasklist_lock); |
| 725 | /* |
| 726 | * We can not clear ->child_reaper or leave it alone. |
| 727 | * There may by stealth EXIT_DEAD tasks on ->children, |
| 728 | * forget_original_parent() must move them somewhere. |
| 729 | */ |
| 730 | pid_ns->child_reaper = init_pid_ns.child_reaper; |
| 731 | } |
| 732 | |
| 733 | return pid_ns->child_reaper; |
| 734 | } |
| 735 | |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 736 | /* |
| 737 | * Any that need to be release_task'd are put on the @dead list. |
| 738 | */ |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 739 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 740 | struct list_head *dead) |
| 741 | { |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 742 | list_move_tail(&p->sibling, &p->real_parent->children); |
| 743 | |
| 744 | if (task_detached(p)) |
| 745 | return; |
| 746 | /* |
| 747 | * If this is a threaded reparent there is no need to |
| 748 | * notify anyone anything has happened. |
| 749 | */ |
| 750 | if (same_thread_group(p->real_parent, father)) |
| 751 | return; |
| 752 | |
| 753 | /* We don't want people slaying init. */ |
| 754 | p->exit_signal = SIGCHLD; |
| 755 | |
| 756 | /* If it has exited notify the new parent about this child's death. */ |
Oleg Nesterov | 5cb1144 | 2009-06-17 16:27:30 -0700 | [diff] [blame] | 757 | if (!task_ptrace(p) && |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 758 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
| 759 | do_notify_parent(p, p->exit_signal); |
| 760 | if (task_detached(p)) { |
| 761 | p->exit_state = EXIT_DEAD; |
| 762 | list_move_tail(&p->sibling, dead); |
| 763 | } |
| 764 | } |
| 765 | |
| 766 | kill_orphaned_pgrp(p, father); |
| 767 | } |
| 768 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 769 | static void forget_original_parent(struct task_struct *father) |
| 770 | { |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 771 | struct task_struct *p, *n, *reaper; |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 772 | LIST_HEAD(dead_children); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 773 | |
Oleg Nesterov | 39c626a | 2009-04-02 16:58:18 -0700 | [diff] [blame] | 774 | exit_ptrace(father); |
| 775 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 776 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 777 | reaper = find_new_reaper(father); |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 778 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 779 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 780 | struct task_struct *t = p; |
| 781 | do { |
| 782 | t->real_parent = reaper; |
| 783 | if (t->parent == father) { |
| 784 | BUG_ON(task_ptrace(t)); |
| 785 | t->parent = t->real_parent; |
| 786 | } |
| 787 | if (t->pdeath_signal) |
| 788 | group_send_sig_info(t->pdeath_signal, |
| 789 | SEND_SIG_NOINFO, t); |
| 790 | } while_each_thread(p, t); |
| 791 | reparent_leader(father, p, &dead_children); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 792 | } |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 793 | write_unlock_irq(&tasklist_lock); |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 794 | |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 795 | BUG_ON(!list_empty(&father->children)); |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 796 | |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 797 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
| 798 | list_del_init(&p->sibling); |
Oleg Nesterov | 39c626a | 2009-04-02 16:58:18 -0700 | [diff] [blame] | 799 | release_task(p); |
| 800 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 801 | } |
| 802 | |
| 803 | /* |
| 804 | * Send signals to all our closest relatives so that they know |
| 805 | * to properly mourn us.. |
| 806 | */ |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 807 | static void exit_notify(struct task_struct *tsk, int group_dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 808 | { |
Roland McGrath | 2b2a1ff | 2008-07-25 19:45:54 -0700 | [diff] [blame] | 809 | int signal; |
| 810 | void *cookie; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 811 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 812 | /* |
| 813 | * This does two things: |
| 814 | * |
| 815 | * A. Make init inherit all the child processes |
| 816 | * B. Check to see if any process groups have become orphaned |
| 817 | * as a result of our exiting, and if they have any stopped |
| 818 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
| 819 | */ |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 820 | forget_original_parent(tsk); |
Pavel Emelyanov | 2e4a707 | 2007-10-18 23:40:01 -0700 | [diff] [blame] | 821 | exit_task_namespaces(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 822 | |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 823 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 824 | if (group_dead) |
| 825 | kill_orphaned_pgrp(tsk->group_leader, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 826 | |
Oleg Nesterov | 2472844 | 2007-08-04 01:04:41 +0400 | [diff] [blame] | 827 | /* Let father know we died |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 828 | * |
| 829 | * Thread signals are configurable, but you aren't going to use |
Daniel Walker | d4c5e41 | 2007-10-18 23:39:59 -0700 | [diff] [blame] | 830 | * that to send signals to arbitary processes. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 831 | * That stops right now. |
| 832 | * |
| 833 | * If the parent exec id doesn't match the exec id we saved |
| 834 | * when we started then we know the parent has changed security |
| 835 | * domain. |
| 836 | * |
| 837 | * If our self_exec id doesn't match our parent_exec_id then |
| 838 | * we have changed execution domain as these two values started |
| 839 | * the same after a fork. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 840 | */ |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 841 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
Oleg Nesterov | f49ee50 | 2008-03-02 21:44:40 +0300 | [diff] [blame] | 842 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
Oleg Nesterov | 432870d | 2009-04-06 16:16:02 +0200 | [diff] [blame] | 843 | tsk->self_exec_id != tsk->parent_exec_id)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 844 | tsk->exit_signal = SIGCHLD; |
| 845 | |
Roland McGrath | 2b2a1ff | 2008-07-25 19:45:54 -0700 | [diff] [blame] | 846 | signal = tracehook_notify_death(tsk, &cookie, group_dead); |
Roland McGrath | 5c7edcd | 2008-07-31 02:04:09 -0700 | [diff] [blame] | 847 | if (signal >= 0) |
Roland McGrath | 2b2a1ff | 2008-07-25 19:45:54 -0700 | [diff] [blame] | 848 | signal = do_notify_parent(tsk, signal); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 849 | |
Roland McGrath | 5c7edcd | 2008-07-31 02:04:09 -0700 | [diff] [blame] | 850 | tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 851 | |
Oleg Nesterov | 9c33916 | 2010-05-26 14:43:10 -0700 | [diff] [blame] | 852 | /* mt-exec, de_thread() is waiting for group leader */ |
| 853 | if (unlikely(tsk->signal->notify_count < 0)) |
Oleg Nesterov | 6db840f | 2007-10-16 23:27:23 -0700 | [diff] [blame] | 854 | wake_up_process(tsk->signal->group_exit_task); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 855 | write_unlock_irq(&tasklist_lock); |
| 856 | |
Roland McGrath | 2b2a1ff | 2008-07-25 19:45:54 -0700 | [diff] [blame] | 857 | tracehook_report_death(tsk, signal, cookie, group_dead); |
| 858 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 859 | /* If the process is dead, release it - nobody will wait for it */ |
Roland McGrath | 5c7edcd | 2008-07-31 02:04:09 -0700 | [diff] [blame] | 860 | if (signal == DEATH_REAP) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 861 | release_task(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 862 | } |
| 863 | |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 864 | #ifdef CONFIG_DEBUG_STACK_USAGE |
| 865 | static void check_stack_usage(void) |
| 866 | { |
| 867 | static DEFINE_SPINLOCK(low_water_lock); |
| 868 | static int lowest_to_date = THREAD_SIZE; |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 869 | unsigned long free; |
| 870 | |
Eric Sandeen | 7c9f886 | 2008-04-22 16:38:23 -0500 | [diff] [blame] | 871 | free = stack_not_used(current); |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 872 | |
| 873 | if (free >= lowest_to_date) |
| 874 | return; |
| 875 | |
| 876 | spin_lock(&low_water_lock); |
| 877 | if (free < lowest_to_date) { |
| 878 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " |
| 879 | "left\n", |
| 880 | current->comm, free); |
| 881 | lowest_to_date = free; |
| 882 | } |
| 883 | spin_unlock(&low_water_lock); |
| 884 | } |
| 885 | #else |
| 886 | static inline void check_stack_usage(void) {} |
| 887 | #endif |
| 888 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 889 | NORET_TYPE void do_exit(long code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 890 | { |
| 891 | struct task_struct *tsk = current; |
| 892 | int group_dead; |
| 893 | |
| 894 | profile_task_exit(tsk); |
| 895 | |
Jens Axboe | 22e2c50 | 2005-06-27 10:55:12 +0200 | [diff] [blame] | 896 | WARN_ON(atomic_read(&tsk->fs_excl)); |
| 897 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 898 | if (unlikely(in_interrupt())) |
| 899 | panic("Aiee, killing interrupt handler!"); |
| 900 | if (unlikely(!tsk->pid)) |
| 901 | panic("Attempted to kill the idle task!"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 902 | |
Roland McGrath | 30199f5 | 2008-07-25 19:45:46 -0700 | [diff] [blame] | 903 | tracehook_report_exit(&code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 904 | |
David Howells | e0e8173 | 2009-09-02 09:13:40 +0100 | [diff] [blame] | 905 | validate_creds_for_do_exit(tsk); |
| 906 | |
Alexander Nyberg | df164db | 2005-06-23 00:09:13 -0700 | [diff] [blame] | 907 | /* |
| 908 | * We're taking recursive faults here in do_exit. Safest is to just |
| 909 | * leave this task alone and wait for reboot. |
| 910 | */ |
| 911 | if (unlikely(tsk->flags & PF_EXITING)) { |
| 912 | printk(KERN_ALERT |
| 913 | "Fixing recursive fault but reboot is needed!\n"); |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 914 | /* |
| 915 | * We can do this unlocked here. The futex code uses |
| 916 | * this flag just to verify whether the pi state |
| 917 | * cleanup has been done or not. In the worst case it |
| 918 | * loops once more. We pretend that the cleanup was |
| 919 | * done as there is no way to return. Either the |
| 920 | * OWNER_DIED bit is set by now or we push the blocked |
| 921 | * task into the wait for ever nirwana as well. |
| 922 | */ |
| 923 | tsk->flags |= PF_EXITPIDONE; |
Alexander Nyberg | df164db | 2005-06-23 00:09:13 -0700 | [diff] [blame] | 924 | set_current_state(TASK_UNINTERRUPTIBLE); |
| 925 | schedule(); |
| 926 | } |
| 927 | |
Thomas Gleixner | 3aa551c | 2009-03-23 18:28:15 +0100 | [diff] [blame] | 928 | exit_irq_thread(); |
| 929 | |
Oleg Nesterov | d12619b | 2008-02-08 04:19:12 -0800 | [diff] [blame] | 930 | exit_signals(tsk); /* sets PF_EXITING */ |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 931 | /* |
| 932 | * tsk->flags are checked in the futex code to protect against |
| 933 | * an exiting task cleaning up the robust pi futexes. |
| 934 | */ |
Oleg Nesterov | d2ee719 | 2007-10-16 23:26:47 -0700 | [diff] [blame] | 935 | smp_mb(); |
Thomas Gleixner | 1d61548 | 2009-11-17 14:54:03 +0100 | [diff] [blame] | 936 | raw_spin_unlock_wait(&tsk->pi_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 937 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 938 | if (unlikely(in_atomic())) |
| 939 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", |
Pavel Emelyanov | ba25f9d | 2007-10-18 23:40:40 -0700 | [diff] [blame] | 940 | current->comm, task_pid_nr(current), |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 941 | preempt_count()); |
| 942 | |
| 943 | acct_update_integrals(tsk); |
KAMEZAWA Hiroyuki | 34e5523 | 2010-03-05 13:41:40 -0800 | [diff] [blame] | 944 | /* sync mm's RSS info before statistics gathering */ |
KAMEZAWA Hiroyuki | a3a2e76 | 2010-04-06 14:34:42 -0700 | [diff] [blame] | 945 | if (tsk->mm) |
| 946 | sync_mm_rss(tsk, tsk->mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 947 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
Andrew Morton | c306895 | 2005-08-04 16:49:32 -0700 | [diff] [blame] | 948 | if (group_dead) { |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 949 | hrtimer_cancel(&tsk->signal->real_timer); |
Roland McGrath | 25f407f | 2005-10-21 15:03:29 -0700 | [diff] [blame] | 950 | exit_itimers(tsk->signal); |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 951 | if (tsk->mm) |
| 952 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); |
Andrew Morton | c306895 | 2005-08-04 16:49:32 -0700 | [diff] [blame] | 953 | } |
KaiGai Kohei | f6ec29a | 2006-06-25 05:49:25 -0700 | [diff] [blame] | 954 | acct_collect(code, group_dead); |
Miloslav Trmac | 522ed77 | 2007-07-15 23:40:56 -0700 | [diff] [blame] | 955 | if (group_dead) |
| 956 | tty_audit_exit(); |
Al Viro | fa84cb9 | 2006-03-29 20:30:19 -0500 | [diff] [blame] | 957 | if (unlikely(tsk->audit_context)) |
| 958 | audit_free(tsk); |
Oleg Nesterov | 115085e | 2006-12-06 20:36:51 -0800 | [diff] [blame] | 959 | |
Jonathan Lim | f2ab6d8 | 2007-08-30 23:56:23 -0700 | [diff] [blame] | 960 | tsk->exit_code = code; |
Oleg Nesterov | 115085e | 2006-12-06 20:36:51 -0800 | [diff] [blame] | 961 | taskstats_exit(tsk, group_dead); |
Shailabh Nagar | c757249 | 2006-07-14 00:24:40 -0700 | [diff] [blame] | 962 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 963 | exit_mm(tsk); |
| 964 | |
KaiGai Kohei | 0e46481 | 2006-06-25 05:49:24 -0700 | [diff] [blame] | 965 | if (group_dead) |
KaiGai Kohei | f6ec29a | 2006-06-25 05:49:25 -0700 | [diff] [blame] | 966 | acct_process(); |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 967 | trace_sched_process_exit(tsk); |
| 968 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 969 | exit_sem(tsk); |
Al Viro | 1ec7f1d | 2008-04-22 05:35:42 -0400 | [diff] [blame] | 970 | exit_files(tsk); |
| 971 | exit_fs(tsk); |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 972 | check_stack_usage(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 973 | exit_thread(); |
Paul Menage | b4f48b6 | 2007-10-18 23:39:33 -0700 | [diff] [blame] | 974 | cgroup_exit(tsk, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 975 | |
Alan Cox | 5ec93d1 | 2009-11-30 13:18:45 +0000 | [diff] [blame] | 976 | if (group_dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 977 | disassociate_ctty(1); |
| 978 | |
Al Viro | a1261f5 | 2005-11-13 16:06:55 -0800 | [diff] [blame] | 979 | module_put(task_thread_info(tsk)->exec_domain->module); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 980 | |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 981 | proc_exit_connector(tsk); |
Ingo Molnar | 33b2fb3 | 2009-05-17 11:08:41 +0200 | [diff] [blame] | 982 | |
| 983 | /* |
Frederic Weisbecker | 24f1e32c | 2009-09-09 19:22:48 +0200 | [diff] [blame] | 984 | * FIXME: do that only when needed, using sched_exit tracepoint |
| 985 | */ |
| 986 | flush_ptrace_hw_breakpoint(tsk); |
| 987 | /* |
Ingo Molnar | 33b2fb3 | 2009-05-17 11:08:41 +0200 | [diff] [blame] | 988 | * Flush inherited counters to the parent - before the parent |
| 989 | * gets woken up by child-exit notifications. |
| 990 | */ |
Ingo Molnar | cdd6c48 | 2009-09-21 12:02:48 +0200 | [diff] [blame] | 991 | perf_event_exit_task(tsk); |
Ingo Molnar | 33b2fb3 | 2009-05-17 11:08:41 +0200 | [diff] [blame] | 992 | |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 993 | exit_notify(tsk, group_dead); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 994 | #ifdef CONFIG_NUMA |
Miao Xie | c0ff745 | 2010-05-24 14:32:08 -0700 | [diff] [blame] | 995 | task_lock(tsk); |
Lee Schermerhorn | f0be3d3 | 2008-04-28 02:13:08 -0700 | [diff] [blame] | 996 | mpol_put(tsk->mempolicy); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 997 | tsk->mempolicy = NULL; |
Miao Xie | c0ff745 | 2010-05-24 14:32:08 -0700 | [diff] [blame] | 998 | task_unlock(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 999 | #endif |
Alexey Dobriyan | 42b2dd0 | 2007-10-16 23:27:30 -0700 | [diff] [blame] | 1000 | #ifdef CONFIG_FUTEX |
Ingo Molnar | c87e283 | 2006-06-27 02:54:58 -0700 | [diff] [blame] | 1001 | if (unlikely(current->pi_state_cache)) |
| 1002 | kfree(current->pi_state_cache); |
Alexey Dobriyan | 42b2dd0 | 2007-10-16 23:27:30 -0700 | [diff] [blame] | 1003 | #endif |
Ingo Molnar | c87e283 | 2006-06-27 02:54:58 -0700 | [diff] [blame] | 1004 | /* |
Ingo Molnar | 9a11b49a | 2006-07-03 00:24:33 -0700 | [diff] [blame] | 1005 | * Make sure we are holding no locks: |
Ingo Molnar | de5097c | 2006-01-09 15:59:21 -0800 | [diff] [blame] | 1006 | */ |
Ingo Molnar | 9a11b49a | 2006-07-03 00:24:33 -0700 | [diff] [blame] | 1007 | debug_check_no_locks_held(tsk); |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 1008 | /* |
| 1009 | * We can do this unlocked here. The futex code uses this flag |
| 1010 | * just to verify whether the pi state cleanup has been done |
| 1011 | * or not. In the worst case it loops once more. |
| 1012 | */ |
| 1013 | tsk->flags |= PF_EXITPIDONE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1014 | |
Al Viro | afc847b | 2006-02-28 12:51:55 -0500 | [diff] [blame] | 1015 | if (tsk->io_context) |
Louis Rilling | b69f229 | 2009-12-04 14:52:42 +0100 | [diff] [blame] | 1016 | exit_io_context(tsk); |
Al Viro | afc847b | 2006-02-28 12:51:55 -0500 | [diff] [blame] | 1017 | |
Jens Axboe | b92ce55 | 2006-04-11 13:52:07 +0200 | [diff] [blame] | 1018 | if (tsk->splice_pipe) |
| 1019 | __free_pipe_info(tsk->splice_pipe); |
| 1020 | |
David Howells | e0e8173 | 2009-09-02 09:13:40 +0100 | [diff] [blame] | 1021 | validate_creds_for_do_exit(tsk); |
| 1022 | |
Coywolf Qi Hunt | 7407251 | 2005-10-30 15:02:47 -0800 | [diff] [blame] | 1023 | preempt_disable(); |
Paul E. McKenney | f41d911 | 2009-08-22 13:56:52 -0700 | [diff] [blame] | 1024 | exit_rcu(); |
Oleg Nesterov | 55a101f | 2006-09-29 02:01:10 -0700 | [diff] [blame] | 1025 | /* causes final put_task_struct in finish_task_switch(). */ |
Oleg Nesterov | c394cc9 | 2006-09-29 02:01:11 -0700 | [diff] [blame] | 1026 | tsk->state = TASK_DEAD; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1027 | schedule(); |
| 1028 | BUG(); |
| 1029 | /* Avoid "noreturn function does return". */ |
Alan Cox | 54306cf | 2006-09-29 02:00:42 -0700 | [diff] [blame] | 1030 | for (;;) |
| 1031 | cpu_relax(); /* For when BUG is null */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1032 | } |
| 1033 | |
Russ Anderson | 012914d | 2005-04-23 00:08:00 -0700 | [diff] [blame] | 1034 | EXPORT_SYMBOL_GPL(do_exit); |
| 1035 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1036 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
| 1037 | { |
| 1038 | if (comp) |
| 1039 | complete(comp); |
Oleg Nesterov | 55a101f | 2006-09-29 02:01:10 -0700 | [diff] [blame] | 1040 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1041 | do_exit(code); |
| 1042 | } |
| 1043 | |
| 1044 | EXPORT_SYMBOL(complete_and_exit); |
| 1045 | |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 1046 | SYSCALL_DEFINE1(exit, int, error_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1047 | { |
| 1048 | do_exit((error_code&0xff)<<8); |
| 1049 | } |
| 1050 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1051 | /* |
| 1052 | * Take down every thread in the group. This is called by fatal signals |
| 1053 | * as well as by sys_exit_group (below). |
| 1054 | */ |
| 1055 | NORET_TYPE void |
| 1056 | do_group_exit(int exit_code) |
| 1057 | { |
Oleg Nesterov | bfc4b08 | 2008-04-30 00:52:36 -0700 | [diff] [blame] | 1058 | struct signal_struct *sig = current->signal; |
| 1059 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1060 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
| 1061 | |
Oleg Nesterov | bfc4b08 | 2008-04-30 00:52:36 -0700 | [diff] [blame] | 1062 | if (signal_group_exit(sig)) |
| 1063 | exit_code = sig->group_exit_code; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1064 | else if (!thread_group_empty(current)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1065 | struct sighand_struct *const sighand = current->sighand; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1066 | spin_lock_irq(&sighand->siglock); |
Oleg Nesterov | ed5d2ca | 2008-02-04 22:27:24 -0800 | [diff] [blame] | 1067 | if (signal_group_exit(sig)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1068 | /* Another thread got here before we took the lock. */ |
| 1069 | exit_code = sig->group_exit_code; |
| 1070 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1071 | sig->group_exit_code = exit_code; |
Oleg Nesterov | ed5d2ca | 2008-02-04 22:27:24 -0800 | [diff] [blame] | 1072 | sig->flags = SIGNAL_GROUP_EXIT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1073 | zap_other_threads(current); |
| 1074 | } |
| 1075 | spin_unlock_irq(&sighand->siglock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1076 | } |
| 1077 | |
| 1078 | do_exit(exit_code); |
| 1079 | /* NOTREACHED */ |
| 1080 | } |
| 1081 | |
| 1082 | /* |
| 1083 | * this kills every thread in the thread group. Note that any externally |
| 1084 | * wait4()-ing process will get the correct exit code - even if this |
| 1085 | * thread is not the thread group leader. |
| 1086 | */ |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 1087 | SYSCALL_DEFINE1(exit_group, int, error_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1088 | { |
| 1089 | do_group_exit((error_code & 0xff) << 8); |
Heiko Carstens | 2ed7c03 | 2009-01-14 14:13:54 +0100 | [diff] [blame] | 1090 | /* NOTREACHED */ |
| 1091 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1092 | } |
| 1093 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1094 | struct wait_opts { |
| 1095 | enum pid_type wo_type; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1096 | int wo_flags; |
Richard Kennedy | e1eb1eb | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 1097 | struct pid *wo_pid; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1098 | |
| 1099 | struct siginfo __user *wo_info; |
| 1100 | int __user *wo_stat; |
| 1101 | struct rusage __user *wo_rusage; |
| 1102 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1103 | wait_queue_t child_wait; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1104 | int notask_error; |
| 1105 | }; |
| 1106 | |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 1107 | static inline |
| 1108 | struct pid *task_pid_type(struct task_struct *task, enum pid_type type) |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1109 | { |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 1110 | if (type != PIDTYPE_PID) |
| 1111 | task = task->group_leader; |
| 1112 | return task->pids[type].pid; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1113 | } |
| 1114 | |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 1115 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 1116 | { |
| 1117 | return wo->wo_type == PIDTYPE_MAX || |
| 1118 | task_pid_type(p, wo->wo_type) == wo->wo_pid; |
| 1119 | } |
| 1120 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1121 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1122 | { |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 1123 | if (!eligible_pid(wo, p)) |
| 1124 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1125 | /* Wait for all children (clone and not) if __WALL is set; |
| 1126 | * otherwise, wait for clone children *only* if __WCLONE is |
| 1127 | * set; otherwise, wait for non-clone children *only*. (Note: |
| 1128 | * A "clone" child here is one that reports to its parent |
| 1129 | * using a signal other than SIGCHLD.) */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1130 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
| 1131 | && !(wo->wo_flags & __WALL)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1132 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1133 | |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1134 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1135 | } |
| 1136 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1137 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
| 1138 | pid_t pid, uid_t uid, int why, int status) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1139 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1140 | struct siginfo __user *infop; |
| 1141 | int retval = wo->wo_rusage |
| 1142 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Ingo Molnar | 36c8b58 | 2006-07-03 00:25:41 -0700 | [diff] [blame] | 1143 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1144 | put_task_struct(p); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1145 | infop = wo->wo_info; |
Vitaly Mayatskikh | b6fe2d1 | 2009-09-23 15:56:52 -0700 | [diff] [blame] | 1146 | if (infop) { |
| 1147 | if (!retval) |
| 1148 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1149 | if (!retval) |
| 1150 | retval = put_user(0, &infop->si_errno); |
| 1151 | if (!retval) |
| 1152 | retval = put_user((short)why, &infop->si_code); |
| 1153 | if (!retval) |
| 1154 | retval = put_user(pid, &infop->si_pid); |
| 1155 | if (!retval) |
| 1156 | retval = put_user(uid, &infop->si_uid); |
| 1157 | if (!retval) |
| 1158 | retval = put_user(status, &infop->si_status); |
| 1159 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1160 | if (!retval) |
| 1161 | retval = pid; |
| 1162 | return retval; |
| 1163 | } |
| 1164 | |
| 1165 | /* |
| 1166 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold |
| 1167 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
| 1168 | * the lock and this task is uninteresting. If we return nonzero, we have |
| 1169 | * released the lock and the system call should return. |
| 1170 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1171 | static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1172 | { |
| 1173 | unsigned long state; |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1174 | int retval, status, traced; |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1175 | pid_t pid = task_pid_vnr(p); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 1176 | uid_t uid = __task_cred(p)->uid; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1177 | struct siginfo __user *infop; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1178 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1179 | if (!likely(wo->wo_flags & WEXITED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1180 | return 0; |
| 1181 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1182 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1183 | int exit_code = p->exit_code; |
Thiago Farina | f3abd4f | 2010-03-05 13:42:52 -0800 | [diff] [blame] | 1184 | int why; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1185 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1186 | get_task_struct(p); |
| 1187 | read_unlock(&tasklist_lock); |
| 1188 | if ((exit_code & 0x7f) == 0) { |
| 1189 | why = CLD_EXITED; |
| 1190 | status = exit_code >> 8; |
| 1191 | } else { |
| 1192 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; |
| 1193 | status = exit_code & 0x7f; |
| 1194 | } |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1195 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1196 | } |
| 1197 | |
| 1198 | /* |
| 1199 | * Try to move the task's state to DEAD |
| 1200 | * only one thread is allowed to do this: |
| 1201 | */ |
| 1202 | state = xchg(&p->exit_state, EXIT_DEAD); |
| 1203 | if (state != EXIT_ZOMBIE) { |
| 1204 | BUG_ON(state != EXIT_DEAD); |
| 1205 | return 0; |
| 1206 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1207 | |
Oleg Nesterov | 53b6f9f | 2008-04-30 00:53:13 -0700 | [diff] [blame] | 1208 | traced = ptrace_reparented(p); |
Oleg Nesterov | befca96 | 2009-06-18 16:49:11 -0700 | [diff] [blame] | 1209 | /* |
| 1210 | * It can be ptraced but not reparented, check |
| 1211 | * !task_detached() to filter out sub-threads. |
| 1212 | */ |
| 1213 | if (likely(!traced) && likely(!task_detached(p))) { |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1214 | struct signal_struct *psig; |
| 1215 | struct signal_struct *sig; |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 1216 | unsigned long maxrss; |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1217 | cputime_t tgutime, tgstime; |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1218 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1219 | /* |
| 1220 | * The resource counters for the group leader are in its |
| 1221 | * own task_struct. Those for dead threads in the group |
| 1222 | * are in its signal_struct, as are those for the child |
| 1223 | * processes it has previously reaped. All these |
| 1224 | * accumulate in the parent's signal_struct c* fields. |
| 1225 | * |
| 1226 | * We don't bother to take a lock here to protect these |
| 1227 | * p->signal fields, because they are only touched by |
| 1228 | * __exit_signal, which runs with tasklist_lock |
| 1229 | * write-locked anyway, and so is excluded here. We do |
Oleg Nesterov | d1e98f4 | 2009-06-17 16:27:34 -0700 | [diff] [blame] | 1230 | * need to protect the access to parent->signal fields, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1231 | * as other threads in the parent group can be right |
| 1232 | * here reaping other children at the same time. |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1233 | * |
| 1234 | * We use thread_group_times() to get times for the thread |
| 1235 | * group, which consolidates times for all threads in the |
| 1236 | * group including the group leader. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1237 | */ |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1238 | thread_group_times(p, &tgutime, &tgstime); |
Oleg Nesterov | d1e98f4 | 2009-06-17 16:27:34 -0700 | [diff] [blame] | 1239 | spin_lock_irq(&p->real_parent->sighand->siglock); |
| 1240 | psig = p->real_parent->signal; |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1241 | sig = p->signal; |
| 1242 | psig->cutime = |
| 1243 | cputime_add(psig->cutime, |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1244 | cputime_add(tgutime, |
| 1245 | sig->cutime)); |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1246 | psig->cstime = |
| 1247 | cputime_add(psig->cstime, |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1248 | cputime_add(tgstime, |
| 1249 | sig->cstime)); |
Laurent Vivier | 9ac5231 | 2007-10-15 17:00:19 +0200 | [diff] [blame] | 1250 | psig->cgtime = |
| 1251 | cputime_add(psig->cgtime, |
| 1252 | cputime_add(p->gtime, |
| 1253 | cputime_add(sig->gtime, |
| 1254 | sig->cgtime))); |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 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; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 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; |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 1269 | maxrss = max(sig->maxrss, sig->cmaxrss); |
| 1270 | if (psig->cmaxrss < maxrss) |
| 1271 | psig->cmaxrss = maxrss; |
Andrea Righi | 5995477 | 2008-07-27 17:29:15 +0200 | [diff] [blame] | 1272 | task_io_accounting_add(&psig->ioac, &p->ioac); |
| 1273 | task_io_accounting_add(&psig->ioac, &sig->ioac); |
Oleg Nesterov | d1e98f4 | 2009-06-17 16:27:34 -0700 | [diff] [blame] | 1274 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1275 | } |
| 1276 | |
| 1277 | /* |
| 1278 | * Now we are sure this task is interesting, and no other |
| 1279 | * thread can reap it because we set its state to EXIT_DEAD. |
| 1280 | */ |
| 1281 | read_unlock(&tasklist_lock); |
| 1282 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1283 | retval = wo->wo_rusage |
| 1284 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1285 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
| 1286 | ? p->signal->group_exit_code : p->exit_code; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1287 | if (!retval && wo->wo_stat) |
| 1288 | retval = put_user(status, wo->wo_stat); |
| 1289 | |
| 1290 | infop = wo->wo_info; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1291 | if (!retval && infop) |
| 1292 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1293 | if (!retval && infop) |
| 1294 | retval = put_user(0, &infop->si_errno); |
| 1295 | if (!retval && infop) { |
| 1296 | int why; |
| 1297 | |
| 1298 | if ((status & 0x7f) == 0) { |
| 1299 | why = CLD_EXITED; |
| 1300 | status >>= 8; |
| 1301 | } else { |
| 1302 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; |
| 1303 | status &= 0x7f; |
| 1304 | } |
| 1305 | retval = put_user((short)why, &infop->si_code); |
| 1306 | if (!retval) |
| 1307 | retval = put_user(status, &infop->si_status); |
| 1308 | } |
| 1309 | if (!retval && infop) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1310 | retval = put_user(pid, &infop->si_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1311 | if (!retval && infop) |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 1312 | retval = put_user(uid, &infop->si_uid); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1313 | if (!retval) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1314 | retval = pid; |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1315 | |
| 1316 | if (traced) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1317 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1318 | /* We dropped tasklist, ptracer could die and untrace */ |
| 1319 | ptrace_unlink(p); |
| 1320 | /* |
| 1321 | * If this is not a detached task, notify the parent. |
| 1322 | * If it's still not detached after that, don't release |
| 1323 | * it now. |
| 1324 | */ |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 1325 | if (!task_detached(p)) { |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1326 | do_notify_parent(p, p->exit_signal); |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 1327 | if (!task_detached(p)) { |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1328 | p->exit_state = EXIT_ZOMBIE; |
| 1329 | p = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1330 | } |
| 1331 | } |
| 1332 | write_unlock_irq(&tasklist_lock); |
| 1333 | } |
| 1334 | if (p != NULL) |
| 1335 | release_task(p); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1336 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1337 | return retval; |
| 1338 | } |
| 1339 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1340 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
| 1341 | { |
| 1342 | if (ptrace) { |
| 1343 | if (task_is_stopped_or_traced(p)) |
| 1344 | return &p->exit_code; |
| 1345 | } else { |
| 1346 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
| 1347 | return &p->signal->group_exit_code; |
| 1348 | } |
| 1349 | return NULL; |
| 1350 | } |
| 1351 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1352 | /* |
| 1353 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold |
| 1354 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
| 1355 | * the lock and this task is uninteresting. If we return nonzero, we have |
| 1356 | * released the lock and the system call should return. |
| 1357 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1358 | static int wait_task_stopped(struct wait_opts *wo, |
| 1359 | int ptrace, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1360 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1361 | struct siginfo __user *infop; |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1362 | int retval, exit_code, *p_code, why; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1363 | uid_t uid = 0; /* unneeded, required by compiler */ |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1364 | pid_t pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1365 | |
Oleg Nesterov | 4791802 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1366 | /* |
| 1367 | * Traditionally we see ptrace'd stopped tasks regardless of options. |
| 1368 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1369 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1370 | return 0; |
| 1371 | |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1372 | exit_code = 0; |
| 1373 | spin_lock_irq(&p->sighand->siglock); |
| 1374 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1375 | p_code = task_stopped_code(p, ptrace); |
| 1376 | if (unlikely(!p_code)) |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1377 | goto unlock_sig; |
| 1378 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1379 | exit_code = *p_code; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1380 | if (!exit_code) |
| 1381 | goto unlock_sig; |
| 1382 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1383 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1384 | *p_code = 0; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1385 | |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 1386 | /* don't need the RCU readlock here as we're holding a spinlock */ |
| 1387 | uid = __task_cred(p)->uid; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1388 | unlock_sig: |
| 1389 | spin_unlock_irq(&p->sighand->siglock); |
| 1390 | if (!exit_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1391 | return 0; |
| 1392 | |
| 1393 | /* |
| 1394 | * Now we are pretty sure this task is interesting. |
| 1395 | * Make sure it doesn't get reaped out from under us while we |
| 1396 | * give up the lock and then examine it below. We don't want to |
| 1397 | * keep holding onto the tasklist_lock while we call getrusage and |
| 1398 | * possibly take page faults for user memory. |
| 1399 | */ |
| 1400 | get_task_struct(p); |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1401 | pid = task_pid_vnr(p); |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1402 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1403 | read_unlock(&tasklist_lock); |
| 1404 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1405 | if (unlikely(wo->wo_flags & WNOWAIT)) |
| 1406 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1407 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1408 | retval = wo->wo_rusage |
| 1409 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
| 1410 | if (!retval && wo->wo_stat) |
| 1411 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); |
| 1412 | |
| 1413 | infop = wo->wo_info; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1414 | if (!retval && infop) |
| 1415 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1416 | if (!retval && infop) |
| 1417 | retval = put_user(0, &infop->si_errno); |
| 1418 | if (!retval && infop) |
Roland McGrath | 6efcae4 | 2008-03-08 11:41:22 -0800 | [diff] [blame] | 1419 | retval = put_user((short)why, &infop->si_code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1420 | if (!retval && infop) |
| 1421 | retval = put_user(exit_code, &infop->si_status); |
| 1422 | if (!retval && infop) |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1423 | retval = put_user(pid, &infop->si_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1424 | if (!retval && infop) |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1425 | retval = put_user(uid, &infop->si_uid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1426 | if (!retval) |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1427 | retval = pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1428 | put_task_struct(p); |
| 1429 | |
| 1430 | BUG_ON(!retval); |
| 1431 | return retval; |
| 1432 | } |
| 1433 | |
| 1434 | /* |
| 1435 | * Handle do_wait work for one task in a live, non-stopped state. |
| 1436 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
| 1437 | * the lock and this task is uninteresting. If we return nonzero, we have |
| 1438 | * released the lock and the system call should return. |
| 1439 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1440 | static int wait_task_continued(struct wait_opts *wo, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1441 | { |
| 1442 | int retval; |
| 1443 | pid_t pid; |
| 1444 | uid_t uid; |
| 1445 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1446 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1447 | return 0; |
| 1448 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1449 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
| 1450 | return 0; |
| 1451 | |
| 1452 | spin_lock_irq(&p->sighand->siglock); |
| 1453 | /* Re-check with the lock held. */ |
| 1454 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { |
| 1455 | spin_unlock_irq(&p->sighand->siglock); |
| 1456 | return 0; |
| 1457 | } |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1458 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1459 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 1460 | uid = __task_cred(p)->uid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1461 | spin_unlock_irq(&p->sighand->siglock); |
| 1462 | |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1463 | pid = task_pid_vnr(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1464 | get_task_struct(p); |
| 1465 | read_unlock(&tasklist_lock); |
| 1466 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1467 | if (!wo->wo_info) { |
| 1468 | retval = wo->wo_rusage |
| 1469 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1470 | put_task_struct(p); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1471 | if (!retval && wo->wo_stat) |
| 1472 | retval = put_user(0xffff, wo->wo_stat); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1473 | if (!retval) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1474 | retval = pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1475 | } else { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1476 | retval = wait_noreap_copyout(wo, p, pid, uid, |
| 1477 | CLD_CONTINUED, SIGCONT); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1478 | BUG_ON(retval == 0); |
| 1479 | } |
| 1480 | |
| 1481 | return retval; |
| 1482 | } |
| 1483 | |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1484 | /* |
| 1485 | * Consider @p for a wait by @parent. |
| 1486 | * |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1487 | * -ECHILD should be in ->notask_error before the first call. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1488 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
| 1489 | * Returns zero if the search for a child should continue; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1490 | * then ->notask_error is 0 if @p is an eligible child, |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1491 | * or another error from security_task_wait(), or still -ECHILD. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1492 | */ |
Oleg Nesterov | b6e763f | 2009-09-23 15:56:50 -0700 | [diff] [blame] | 1493 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
| 1494 | struct task_struct *p) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1495 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1496 | int ret = eligible_child(wo, p); |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1497 | if (!ret) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1498 | return ret; |
| 1499 | |
Oleg Nesterov | a2322e1 | 2009-09-23 15:56:45 -0700 | [diff] [blame] | 1500 | ret = security_task_wait(p); |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1501 | if (unlikely(ret < 0)) { |
| 1502 | /* |
| 1503 | * If we have not yet seen any eligible child, |
| 1504 | * then let this error code replace -ECHILD. |
| 1505 | * A permission error will give the user a clue |
| 1506 | * to look for security policy problems, rather |
| 1507 | * than for mysterious wait bugs. |
| 1508 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1509 | if (wo->notask_error) |
| 1510 | wo->notask_error = ret; |
Oleg Nesterov | 78a3d9d | 2009-04-29 18:01:23 +0200 | [diff] [blame] | 1511 | return 0; |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1512 | } |
| 1513 | |
Oleg Nesterov | 5cb1144 | 2009-06-17 16:27:30 -0700 | [diff] [blame] | 1514 | if (likely(!ptrace) && unlikely(task_ptrace(p))) { |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1515 | /* |
| 1516 | * This child is hidden by ptrace. |
| 1517 | * We aren't allowed to see it now, but eventually we will. |
| 1518 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1519 | wo->notask_error = 0; |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1520 | return 0; |
| 1521 | } |
| 1522 | |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1523 | if (p->exit_state == EXIT_DEAD) |
| 1524 | return 0; |
| 1525 | |
| 1526 | /* |
| 1527 | * We don't reap group leaders with subthreads. |
| 1528 | */ |
| 1529 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1530 | return wait_task_zombie(wo, p); |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1531 | |
| 1532 | /* |
| 1533 | * It's stopped or running now, so it might |
| 1534 | * later continue, exit, or stop again. |
| 1535 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1536 | wo->notask_error = 0; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1537 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1538 | if (task_stopped_code(p, ptrace)) |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1539 | return wait_task_stopped(wo, ptrace, p); |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1540 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1541 | return wait_task_continued(wo, p); |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1542 | } |
| 1543 | |
| 1544 | /* |
| 1545 | * Do the work of do_wait() for one thread in the group, @tsk. |
| 1546 | * |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1547 | * -ECHILD should be in ->notask_error before the first call. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1548 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
| 1549 | * Returns zero if the search for a child should continue; then |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1550 | * ->notask_error is 0 if there were any eligible children, |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1551 | * or another error from security_task_wait(), or still -ECHILD. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1552 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1553 | static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1554 | { |
| 1555 | struct task_struct *p; |
| 1556 | |
| 1557 | list_for_each_entry(p, &tsk->children, sibling) { |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 1558 | int ret = wait_consider_task(wo, 0, p); |
| 1559 | if (ret) |
| 1560 | return ret; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1561 | } |
| 1562 | |
| 1563 | return 0; |
| 1564 | } |
| 1565 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1566 | static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1567 | { |
| 1568 | struct task_struct *p; |
| 1569 | |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1570 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
Oleg Nesterov | b6e763f | 2009-09-23 15:56:50 -0700 | [diff] [blame] | 1571 | int ret = wait_consider_task(wo, 1, p); |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1572 | if (ret) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1573 | return ret; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1574 | } |
| 1575 | |
| 1576 | return 0; |
| 1577 | } |
| 1578 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1579 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
| 1580 | int sync, void *key) |
| 1581 | { |
| 1582 | struct wait_opts *wo = container_of(wait, struct wait_opts, |
| 1583 | child_wait); |
| 1584 | struct task_struct *p = key; |
| 1585 | |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 1586 | if (!eligible_pid(wo, p)) |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1587 | return 0; |
| 1588 | |
Oleg Nesterov | b4fe518 | 2009-09-23 15:56:47 -0700 | [diff] [blame] | 1589 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
| 1590 | return 0; |
| 1591 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1592 | return default_wake_function(wait, mode, sync, key); |
| 1593 | } |
| 1594 | |
Oleg Nesterov | a7f0765 | 2009-09-23 15:56:44 -0700 | [diff] [blame] | 1595 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
| 1596 | { |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1597 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
| 1598 | TASK_INTERRUPTIBLE, 1, p); |
Oleg Nesterov | a7f0765 | 2009-09-23 15:56:44 -0700 | [diff] [blame] | 1599 | } |
| 1600 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1601 | static long do_wait(struct wait_opts *wo) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1602 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1603 | struct task_struct *tsk; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1604 | int retval; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1605 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1606 | trace_sched_process_wait(wo->wo_pid); |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 1607 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1608 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
| 1609 | wo->child_wait.private = current; |
| 1610 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1611 | repeat: |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1612 | /* |
| 1613 | * If there is nothing that can match our critiera just get out. |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1614 | * We will clear ->notask_error to zero if we see any child that |
| 1615 | * might later match our criteria, even if we are not able to reap |
| 1616 | * it yet. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1617 | */ |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1618 | wo->notask_error = -ECHILD; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1619 | if ((wo->wo_type < PIDTYPE_MAX) && |
| 1620 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1621 | goto notask; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1622 | |
Oleg Nesterov | f95d39d | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 1623 | set_current_state(TASK_INTERRUPTIBLE); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1624 | read_lock(&tasklist_lock); |
| 1625 | tsk = current; |
| 1626 | do { |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1627 | retval = do_wait_thread(wo, tsk); |
| 1628 | if (retval) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1629 | goto end; |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1630 | |
| 1631 | retval = ptrace_do_wait(wo, tsk); |
| 1632 | if (retval) |
| 1633 | goto end; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1634 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1635 | if (wo->wo_flags & __WNOTHREAD) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1636 | break; |
Oleg Nesterov | a3f6dfb | 2009-06-17 16:27:41 -0700 | [diff] [blame] | 1637 | } while_each_thread(current, tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1638 | read_unlock(&tasklist_lock); |
Oleg Nesterov | f2cc3eb | 2008-02-08 04:19:06 -0800 | [diff] [blame] | 1639 | |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1640 | notask: |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1641 | retval = wo->notask_error; |
| 1642 | if (!retval && !(wo->wo_flags & WNOHANG)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1643 | retval = -ERESTARTSYS; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1644 | if (!signal_pending(current)) { |
| 1645 | schedule(); |
| 1646 | goto repeat; |
| 1647 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1648 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1649 | end: |
Oleg Nesterov | f95d39d | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 1650 | __set_current_state(TASK_RUNNING); |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1651 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1652 | return retval; |
| 1653 | } |
| 1654 | |
Heiko Carstens | 17da2bd | 2009-01-14 14:14:10 +0100 | [diff] [blame] | 1655 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
| 1656 | infop, int, options, struct rusage __user *, ru) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1657 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1658 | struct wait_opts wo; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1659 | struct pid *pid = NULL; |
| 1660 | enum pid_type type; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1661 | long ret; |
| 1662 | |
| 1663 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) |
| 1664 | return -EINVAL; |
| 1665 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) |
| 1666 | return -EINVAL; |
| 1667 | |
| 1668 | switch (which) { |
| 1669 | case P_ALL: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1670 | type = PIDTYPE_MAX; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1671 | break; |
| 1672 | case P_PID: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1673 | type = PIDTYPE_PID; |
| 1674 | if (upid <= 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1675 | return -EINVAL; |
| 1676 | break; |
| 1677 | case P_PGID: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1678 | type = PIDTYPE_PGID; |
| 1679 | if (upid <= 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1680 | return -EINVAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1681 | break; |
| 1682 | default: |
| 1683 | return -EINVAL; |
| 1684 | } |
| 1685 | |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1686 | if (type < PIDTYPE_MAX) |
| 1687 | pid = find_get_pid(upid); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1688 | |
| 1689 | wo.wo_type = type; |
| 1690 | wo.wo_pid = pid; |
| 1691 | wo.wo_flags = options; |
| 1692 | wo.wo_info = infop; |
| 1693 | wo.wo_stat = NULL; |
| 1694 | wo.wo_rusage = ru; |
| 1695 | ret = do_wait(&wo); |
Vitaly Mayatskikh | dfe16df | 2009-09-23 15:56:51 -0700 | [diff] [blame] | 1696 | |
| 1697 | if (ret > 0) { |
| 1698 | ret = 0; |
| 1699 | } else if (infop) { |
| 1700 | /* |
| 1701 | * For a WNOHANG return, clear out all the fields |
| 1702 | * we would set so the user can easily tell the |
| 1703 | * difference. |
| 1704 | */ |
| 1705 | if (!ret) |
| 1706 | ret = put_user(0, &infop->si_signo); |
| 1707 | if (!ret) |
| 1708 | ret = put_user(0, &infop->si_errno); |
| 1709 | if (!ret) |
| 1710 | ret = put_user(0, &infop->si_code); |
| 1711 | if (!ret) |
| 1712 | ret = put_user(0, &infop->si_pid); |
| 1713 | if (!ret) |
| 1714 | ret = put_user(0, &infop->si_uid); |
| 1715 | if (!ret) |
| 1716 | ret = put_user(0, &infop->si_status); |
| 1717 | } |
| 1718 | |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1719 | put_pid(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1720 | |
| 1721 | /* avoid REGPARM breakage on x86: */ |
Roland McGrath | 54a0151 | 2008-04-10 15:37:38 -0700 | [diff] [blame] | 1722 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1723 | return ret; |
| 1724 | } |
| 1725 | |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 1726 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
| 1727 | int, options, struct rusage __user *, ru) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1728 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1729 | struct wait_opts wo; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1730 | struct pid *pid = NULL; |
| 1731 | enum pid_type type; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1732 | long ret; |
| 1733 | |
| 1734 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| |
| 1735 | __WNOTHREAD|__WCLONE|__WALL)) |
| 1736 | return -EINVAL; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1737 | |
| 1738 | if (upid == -1) |
| 1739 | type = PIDTYPE_MAX; |
| 1740 | else if (upid < 0) { |
| 1741 | type = PIDTYPE_PGID; |
| 1742 | pid = find_get_pid(-upid); |
| 1743 | } else if (upid == 0) { |
| 1744 | type = PIDTYPE_PGID; |
Oleg Nesterov | 2ae448e | 2009-04-02 16:58:36 -0700 | [diff] [blame] | 1745 | pid = get_task_pid(current, PIDTYPE_PGID); |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1746 | } else /* upid > 0 */ { |
| 1747 | type = PIDTYPE_PID; |
| 1748 | pid = find_get_pid(upid); |
| 1749 | } |
| 1750 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1751 | wo.wo_type = type; |
| 1752 | wo.wo_pid = pid; |
| 1753 | wo.wo_flags = options | WEXITED; |
| 1754 | wo.wo_info = NULL; |
| 1755 | wo.wo_stat = stat_addr; |
| 1756 | wo.wo_rusage = ru; |
| 1757 | ret = do_wait(&wo); |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1758 | put_pid(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1759 | |
| 1760 | /* avoid REGPARM breakage on x86: */ |
Roland McGrath | 54a0151 | 2008-04-10 15:37:38 -0700 | [diff] [blame] | 1761 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1762 | return ret; |
| 1763 | } |
| 1764 | |
| 1765 | #ifdef __ARCH_WANT_SYS_WAITPID |
| 1766 | |
| 1767 | /* |
| 1768 | * sys_waitpid() remains for compatibility. waitpid() should be |
| 1769 | * implemented by calling sys_wait4() from libc.a. |
| 1770 | */ |
Heiko Carstens | 17da2bd | 2009-01-14 14:14:10 +0100 | [diff] [blame] | 1771 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1772 | { |
| 1773 | return sys_wait4(pid, stat_addr, options, NULL); |
| 1774 | } |
| 1775 | |
| 1776 | #endif |