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> |
Ying Han | 3d5992d | 2010-10-26 14:21:23 -0700 | [diff] [blame] | 53 | #include <linux/oom.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 54 | |
| 55 | #include <asm/uaccess.h> |
| 56 | #include <asm/unistd.h> |
| 57 | #include <asm/pgtable.h> |
| 58 | #include <asm/mmu_context.h> |
| 59 | |
Adrian Bunk | 408b664 | 2005-05-01 08:59:29 -0700 | [diff] [blame] | 60 | static void exit_mm(struct task_struct * tsk); |
| 61 | |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 62 | static void __unhash_process(struct task_struct *p, bool group_dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 63 | { |
| 64 | nr_threads--; |
| 65 | detach_pid(p, PIDTYPE_PID); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 66 | if (group_dead) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 67 | detach_pid(p, PIDTYPE_PGID); |
| 68 | detach_pid(p, PIDTYPE_SID); |
Oleg Nesterov | c97d989 | 2006-03-28 16:11:06 -0800 | [diff] [blame] | 69 | |
Eric W. Biederman | 5e85d4a | 2006-04-18 22:20:16 -0700 | [diff] [blame] | 70 | list_del_rcu(&p->tasks); |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 71 | list_del_init(&p->sibling); |
Oleg Nesterov | 73b9ebf | 2006-03-28 16:11:07 -0800 | [diff] [blame] | 72 | __get_cpu_var(process_counts)--; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 73 | } |
Oleg Nesterov | 47e6532 | 2006-03-28 16:11:25 -0800 | [diff] [blame] | 74 | list_del_rcu(&p->thread_group); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 75 | } |
| 76 | |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 77 | /* |
| 78 | * This function expects the tasklist_lock write-locked. |
| 79 | */ |
| 80 | static void __exit_signal(struct task_struct *tsk) |
| 81 | { |
| 82 | struct signal_struct *sig = tsk->signal; |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 83 | bool group_dead = thread_group_leader(tsk); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 84 | struct sighand_struct *sighand; |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 85 | struct tty_struct *uninitialized_var(tty); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 86 | |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 87 | sighand = rcu_dereference_check(tsk->sighand, |
| 88 | rcu_read_lock_held() || |
Paul E. McKenney | db1466b | 2010-03-03 07:46:56 -0800 | [diff] [blame] | 89 | lockdep_tasklist_lock_is_held()); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 90 | spin_lock(&sighand->siglock); |
| 91 | |
| 92 | posix_cpu_timers_exit(tsk); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 93 | if (group_dead) { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 94 | posix_cpu_timers_exit_group(tsk); |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 95 | tty = sig->tty; |
| 96 | sig->tty = NULL; |
Oleg Nesterov | 4a59994 | 2010-05-26 14:43:12 -0700 | [diff] [blame] | 97 | } else { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 98 | /* |
Oleg Nesterov | e0a7021 | 2010-11-05 16:53:42 +0100 | [diff] [blame] | 99 | * This can only happen if the caller is de_thread(). |
| 100 | * FIXME: this is the temporary hack, we should teach |
| 101 | * posix-cpu-timers to handle this case correctly. |
| 102 | */ |
| 103 | if (unlikely(has_group_leader_pid(tsk))) |
| 104 | posix_cpu_timers_exit_group(tsk); |
| 105 | |
| 106 | /* |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 107 | * If there is any task waiting for the group exit |
| 108 | * then notify it: |
| 109 | */ |
Oleg Nesterov | d344193 | 2010-05-26 14:43:11 -0700 | [diff] [blame] | 110 | if (sig->notify_count > 0 && !--sig->notify_count) |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 111 | wake_up_process(sig->group_exit_task); |
Oleg Nesterov | 6db840f | 2007-10-16 23:27:23 -0700 | [diff] [blame] | 112 | |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 113 | if (tsk == sig->curr_target) |
| 114 | sig->curr_target = next_thread(tsk); |
| 115 | /* |
| 116 | * Accumulate here the counters for all threads but the |
| 117 | * group leader as they die, so they can be added into |
| 118 | * the process-wide totals when those are taken. |
| 119 | * The group leader stays around as a zombie as long |
| 120 | * as there are other threads. When it gets reaped, |
| 121 | * the exit.c code will add its counts into these totals. |
| 122 | * We won't ever get here for the group leader, since it |
| 123 | * will have been the last reference on the signal_struct. |
| 124 | */ |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 125 | sig->utime = cputime_add(sig->utime, tsk->utime); |
| 126 | sig->stime = cputime_add(sig->stime, tsk->stime); |
Hidetoshi Seto | d5b7c78 | 2009-11-26 14:49:05 +0900 | [diff] [blame] | 127 | sig->gtime = cputime_add(sig->gtime, tsk->gtime); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 128 | sig->min_flt += tsk->min_flt; |
| 129 | sig->maj_flt += tsk->maj_flt; |
| 130 | sig->nvcsw += tsk->nvcsw; |
| 131 | sig->nivcsw += tsk->nivcsw; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 132 | sig->inblock += task_io_get_inblock(tsk); |
| 133 | sig->oublock += task_io_get_oublock(tsk); |
Andrea Righi | 5995477 | 2008-07-27 17:29:15 +0200 | [diff] [blame] | 134 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
Peter Zijlstra | 32bd671 | 2009-02-05 12:24:15 +0100 | [diff] [blame] | 135 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 136 | } |
| 137 | |
Oleg Nesterov | b3ac022 | 2010-05-26 14:43:24 -0700 | [diff] [blame] | 138 | sig->nr_threads--; |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 139 | __unhash_process(tsk, group_dead); |
Oleg Nesterov | 5876700 | 2006-03-28 16:11:20 -0800 | [diff] [blame] | 140 | |
Oleg Nesterov | da7978b | 2008-05-23 13:04:41 -0700 | [diff] [blame] | 141 | /* |
| 142 | * Do this under ->siglock, we can race with another thread |
| 143 | * doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals. |
| 144 | */ |
| 145 | flush_sigqueue(&tsk->pending); |
Oleg Nesterov | a7e5328 | 2006-03-28 16:11:27 -0800 | [diff] [blame] | 146 | tsk->sighand = NULL; |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 147 | spin_unlock(&sighand->siglock); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 148 | |
Oleg Nesterov | a7e5328 | 2006-03-28 16:11:27 -0800 | [diff] [blame] | 149 | __cleanup_sighand(sighand); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 150 | clear_tsk_thread_flag(tsk,TIF_SIGPENDING); |
Oleg Nesterov | d40e48e | 2010-05-26 14:43:19 -0700 | [diff] [blame] | 151 | if (group_dead) { |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 152 | flush_sigqueue(&sig->shared_pending); |
Oleg Nesterov | 4ada856 | 2010-05-26 14:43:17 -0700 | [diff] [blame] | 153 | tty_kref_put(tty); |
Oleg Nesterov | 6a14c5c | 2006-03-28 16:11:18 -0800 | [diff] [blame] | 154 | } |
| 155 | } |
| 156 | |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 157 | static void delayed_put_task_struct(struct rcu_head *rhp) |
| 158 | { |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 159 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); |
| 160 | |
Peter Zijlstra | 4e231c7 | 2010-09-09 21:01:59 +0200 | [diff] [blame] | 161 | perf_event_delayed_put(tsk); |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 162 | trace_sched_process_free(tsk); |
| 163 | put_task_struct(tsk); |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 164 | } |
| 165 | |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 166 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 167 | void release_task(struct task_struct * p) |
| 168 | { |
Ingo Molnar | 36c8b58 | 2006-07-03 00:25:41 -0700 | [diff] [blame] | 169 | struct task_struct *leader; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 170 | int zap_leader; |
Oleg Nesterov | 1f09f97 | 2006-03-28 16:11:11 -0800 | [diff] [blame] | 171 | repeat: |
Roland McGrath | dae3357 | 2008-07-25 19:45:48 -0700 | [diff] [blame] | 172 | tracehook_prepare_release_task(p); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 173 | /* 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] | 174 | * can't be modifying its own credentials. But shut RCU-lockdep up */ |
| 175 | rcu_read_lock(); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 176 | atomic_dec(&__task_cred(p)->user->processes); |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 177 | rcu_read_unlock(); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 178 | |
Pavel Emelyanov | 60347f6 | 2007-10-18 23:40:03 -0700 | [diff] [blame] | 179 | proc_flush_task(p); |
Ingo Molnar | 0203026 | 2009-05-17 11:24:08 +0200 | [diff] [blame] | 180 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 181 | write_lock_irq(&tasklist_lock); |
Roland McGrath | dae3357 | 2008-07-25 19:45:48 -0700 | [diff] [blame] | 182 | tracehook_finish_release_task(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 183 | __exit_signal(p); |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 184 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 185 | /* |
| 186 | * If we are the last non-leader member of the thread |
| 187 | * group, and the leader is zombie, then notify the |
| 188 | * group leader's parent process. (if it wants notification.) |
| 189 | */ |
| 190 | zap_leader = 0; |
| 191 | leader = p->group_leader; |
| 192 | if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) { |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 193 | BUG_ON(task_detached(leader)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 194 | do_notify_parent(leader, leader->exit_signal); |
| 195 | /* |
| 196 | * If we were the last child thread and the leader has |
| 197 | * exited already, and the leader's parent ignores SIGCHLD, |
| 198 | * then we are the one who should release the leader. |
| 199 | * |
| 200 | * do_notify_parent() will have marked it self-reaping in |
| 201 | * that case. |
| 202 | */ |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 203 | zap_leader = task_detached(leader); |
Roland McGrath | dae3357 | 2008-07-25 19:45:48 -0700 | [diff] [blame] | 204 | |
| 205 | /* |
| 206 | * This maintains the invariant that release_task() |
| 207 | * only runs on a task in EXIT_DEAD, just for sanity. |
| 208 | */ |
| 209 | if (zap_leader) |
| 210 | leader->exit_state = EXIT_DEAD; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 211 | } |
| 212 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 213 | write_unlock_irq(&tasklist_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 214 | release_thread(p); |
Eric W. Biederman | 8c7904a | 2006-03-31 02:31:37 -0800 | [diff] [blame] | 215 | call_rcu(&p->rcu, delayed_put_task_struct); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 216 | |
| 217 | p = leader; |
| 218 | if (unlikely(zap_leader)) |
| 219 | goto repeat; |
| 220 | } |
| 221 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 222 | /* |
| 223 | * This checks not only the pgrp, but falls back on the pid if no |
| 224 | * satisfactory pgrp is found. I dunno - gdb doesn't work correctly |
| 225 | * without this... |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 226 | * |
| 227 | * The caller must hold rcu lock or the tasklist lock. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 228 | */ |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 229 | struct pid *session_of_pgrp(struct pid *pgrp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 230 | { |
| 231 | struct task_struct *p; |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 232 | struct pid *sid = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 233 | |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 234 | p = pid_task(pgrp, PIDTYPE_PGID); |
Oleg Nesterov | 62dfb55 | 2006-12-08 02:38:03 -0800 | [diff] [blame] | 235 | if (p == NULL) |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 236 | p = pid_task(pgrp, PIDTYPE_PID); |
Oleg Nesterov | 62dfb55 | 2006-12-08 02:38:03 -0800 | [diff] [blame] | 237 | if (p != NULL) |
Eric W. Biederman | 04a2e6a | 2007-02-12 00:52:56 -0800 | [diff] [blame] | 238 | sid = task_session(p); |
Oleg Nesterov | 62dfb55 | 2006-12-08 02:38:03 -0800 | [diff] [blame] | 239 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 240 | return sid; |
| 241 | } |
| 242 | |
| 243 | /* |
| 244 | * Determine if a process group is "orphaned", according to the POSIX |
| 245 | * definition in 2.2.2.52. Orphaned process groups are not to be affected |
| 246 | * by terminal-generated stop signals. Newly orphaned process groups are |
| 247 | * to receive a SIGHUP and a SIGCONT. |
| 248 | * |
| 249 | * "I ask you, have you ever known what it is to be an orphan?" |
| 250 | */ |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 251 | 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] | 252 | { |
| 253 | struct task_struct *p; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 255 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 256 | if ((p == ignored_task) || |
| 257 | (p->exit_state && thread_group_empty(p)) || |
| 258 | is_global_init(p->real_parent)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 259 | continue; |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 260 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 261 | if (task_pgrp(p->real_parent) != pgrp && |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 262 | task_session(p->real_parent) == task_session(p)) |
| 263 | return 0; |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 264 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Oleg Nesterov | 05e83df | 2008-03-02 21:44:42 +0300 | [diff] [blame] | 265 | |
| 266 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 267 | } |
| 268 | |
Eric W. Biederman | 3e7cd6c | 2007-02-12 00:52:58 -0800 | [diff] [blame] | 269 | int is_current_pgrp_orphaned(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 270 | { |
| 271 | int retval; |
| 272 | |
| 273 | read_lock(&tasklist_lock); |
Eric W. Biederman | 3e7cd6c | 2007-02-12 00:52:58 -0800 | [diff] [blame] | 274 | retval = will_become_orphaned_pgrp(task_pgrp(current), NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 275 | read_unlock(&tasklist_lock); |
| 276 | |
| 277 | return retval; |
| 278 | } |
| 279 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 280 | static int has_stopped_jobs(struct pid *pgrp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 281 | { |
| 282 | int retval = 0; |
| 283 | struct task_struct *p; |
| 284 | |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 285 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Matthew Wilcox | 338077e | 2007-12-06 11:09:35 -0500 | [diff] [blame] | 286 | if (!task_is_stopped(p)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 287 | continue; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 288 | retval = 1; |
| 289 | break; |
Eric W. Biederman | 0475ac0 | 2007-02-12 00:52:57 -0800 | [diff] [blame] | 290 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 291 | return retval; |
| 292 | } |
| 293 | |
Oleg Nesterov | f49ee50 | 2008-03-02 21:44:40 +0300 | [diff] [blame] | 294 | /* |
| 295 | * Check to see if any process groups have become orphaned as |
| 296 | * a result of our exiting, and if they have any stopped jobs, |
| 297 | * send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
| 298 | */ |
| 299 | static void |
| 300 | kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent) |
| 301 | { |
| 302 | struct pid *pgrp = task_pgrp(tsk); |
| 303 | struct task_struct *ignored_task = tsk; |
| 304 | |
| 305 | if (!parent) |
| 306 | /* exit: our father is in a different pgrp than |
| 307 | * we are and we were the only connection outside. |
| 308 | */ |
| 309 | parent = tsk->real_parent; |
| 310 | else |
| 311 | /* reparent: our child is in a different pgrp than |
| 312 | * we are, and it was the only connection outside. |
| 313 | */ |
| 314 | ignored_task = NULL; |
| 315 | |
| 316 | if (task_pgrp(parent) != pgrp && |
| 317 | task_session(parent) == task_session(tsk) && |
| 318 | will_become_orphaned_pgrp(pgrp, ignored_task) && |
| 319 | has_stopped_jobs(pgrp)) { |
| 320 | __kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp); |
| 321 | __kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp); |
| 322 | } |
| 323 | } |
| 324 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 325 | /** |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 326 | * reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 327 | * |
| 328 | * 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] | 329 | * it ever exits, it should generally reparent itself to kthreadd so it |
| 330 | * 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] | 331 | * |
| 332 | * The various task state such as scheduling policy and priority may have |
| 333 | * been inherited from a user process, so we reset them to sane values here. |
| 334 | * |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 335 | * NOTE that reparent_to_kthreadd() gives the caller full capabilities. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 336 | */ |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 337 | static void reparent_to_kthreadd(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 338 | { |
| 339 | write_lock_irq(&tasklist_lock); |
| 340 | |
| 341 | ptrace_unlink(current); |
| 342 | /* Reparent to init */ |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 343 | current->real_parent = current->parent = kthreadd_task; |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 344 | list_move_tail(¤t->sibling, ¤t->real_parent->children); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 345 | |
| 346 | /* Set the exit signal to SIGCHLD so we signal init on exit */ |
| 347 | current->exit_signal = SIGCHLD; |
| 348 | |
Ingo Molnar | e05606d | 2007-07-09 18:51:59 +0200 | [diff] [blame] | 349 | if (task_nice(current) < 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 350 | set_user_nice(current, 0); |
| 351 | /* cpus_allowed? */ |
| 352 | /* rt_priority? */ |
| 353 | /* signals? */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 354 | memcpy(current->signal->rlim, init_task.signal->rlim, |
| 355 | sizeof(current->signal->rlim)); |
David Howells | d84f4f9 | 2008-11-14 10:39:23 +1100 | [diff] [blame] | 356 | |
| 357 | atomic_inc(&init_cred.usage); |
| 358 | commit_creds(&init_cred); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 359 | write_unlock_irq(&tasklist_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 360 | } |
| 361 | |
Oleg Nesterov | 8520d7c | 2008-02-08 04:19:09 -0800 | [diff] [blame] | 362 | void __set_special_pids(struct pid *pid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 363 | { |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 364 | struct task_struct *curr = current->group_leader; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 365 | |
Christian Borntraeger | 0d0df59 | 2009-10-26 16:49:34 -0700 | [diff] [blame] | 366 | if (task_session(curr) != pid) |
Oleg Nesterov | 7d8da09 | 2008-04-30 00:54:27 -0700 | [diff] [blame] | 367 | change_pid(curr, PIDTYPE_SID, pid); |
Oleg Nesterov | 1b0f7ff | 2009-04-02 16:58:39 -0700 | [diff] [blame] | 368 | |
| 369 | if (task_pgrp(curr) != pid) |
Oleg Nesterov | 7d8da09 | 2008-04-30 00:54:27 -0700 | [diff] [blame] | 370 | change_pid(curr, PIDTYPE_PGID, pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 371 | } |
| 372 | |
Oleg Nesterov | 8520d7c | 2008-02-08 04:19:09 -0800 | [diff] [blame] | 373 | static void set_special_pids(struct pid *pid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 374 | { |
| 375 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 8520d7c | 2008-02-08 04:19:09 -0800 | [diff] [blame] | 376 | __set_special_pids(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 377 | write_unlock_irq(&tasklist_lock); |
| 378 | } |
| 379 | |
| 380 | /* |
Oleg Nesterov | 8724513 | 2009-06-17 16:27:23 -0700 | [diff] [blame] | 381 | * Let kernel threads use this to say that they allow a certain signal. |
| 382 | * Must not be used if kthread was cloned with CLONE_SIGHAND. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 383 | */ |
| 384 | int allow_signal(int sig) |
| 385 | { |
Jesper Juhl | 7ed20e1 | 2005-05-01 08:59:14 -0700 | [diff] [blame] | 386 | if (!valid_signal(sig) || sig < 1) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 387 | return -EINVAL; |
| 388 | |
| 389 | spin_lock_irq(¤t->sighand->siglock); |
Oleg Nesterov | 8724513 | 2009-06-17 16:27:23 -0700 | [diff] [blame] | 390 | /* This is only needed for daemonize()'ed kthreads */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 391 | sigdelset(¤t->blocked, sig); |
Oleg Nesterov | 8724513 | 2009-06-17 16:27:23 -0700 | [diff] [blame] | 392 | /* |
| 393 | * Kernel threads handle their own signals. Let the signal code |
| 394 | * know it'll be handled, so that they don't get converted to |
| 395 | * SIGKILL or just silently dropped. |
| 396 | */ |
| 397 | current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 398 | recalc_sigpending(); |
| 399 | spin_unlock_irq(¤t->sighand->siglock); |
| 400 | return 0; |
| 401 | } |
| 402 | |
| 403 | EXPORT_SYMBOL(allow_signal); |
| 404 | |
| 405 | int disallow_signal(int sig) |
| 406 | { |
Jesper Juhl | 7ed20e1 | 2005-05-01 08:59:14 -0700 | [diff] [blame] | 407 | if (!valid_signal(sig) || sig < 1) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 408 | return -EINVAL; |
| 409 | |
| 410 | spin_lock_irq(¤t->sighand->siglock); |
Oleg Nesterov | 10ab825 | 2007-05-09 02:34:37 -0700 | [diff] [blame] | 411 | current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 412 | recalc_sigpending(); |
| 413 | spin_unlock_irq(¤t->sighand->siglock); |
| 414 | return 0; |
| 415 | } |
| 416 | |
| 417 | EXPORT_SYMBOL(disallow_signal); |
| 418 | |
| 419 | /* |
| 420 | * Put all the gunge required to become a kernel thread without |
| 421 | * attached user resources in one place where it belongs. |
| 422 | */ |
| 423 | |
| 424 | void daemonize(const char *name, ...) |
| 425 | { |
| 426 | va_list args; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 427 | sigset_t blocked; |
| 428 | |
| 429 | va_start(args, name); |
| 430 | vsnprintf(current->comm, sizeof(current->comm), name, args); |
| 431 | va_end(args); |
| 432 | |
| 433 | /* |
| 434 | * If we were started as result of loading a module, close all of the |
| 435 | * user space pages. We don't need them, and if we didn't close them |
| 436 | * they would be locked into memory. |
| 437 | */ |
| 438 | exit_mm(current); |
Rafael J. Wysocki | 8314418 | 2007-07-17 04:03:35 -0700 | [diff] [blame] | 439 | /* |
| 440 | * We don't want to have TIF_FREEZE set if the system-wide hibernation |
| 441 | * or suspend transition begins right now. |
| 442 | */ |
Oleg Nesterov | 7b34e42 | 2008-07-25 01:47:37 -0700 | [diff] [blame] | 443 | current->flags |= (PF_NOFREEZE | PF_KTHREAD); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 444 | |
Oleg Nesterov | 8520d7c | 2008-02-08 04:19:09 -0800 | [diff] [blame] | 445 | if (current->nsproxy != &init_nsproxy) { |
| 446 | get_nsproxy(&init_nsproxy); |
| 447 | switch_task_namespaces(current, &init_nsproxy); |
| 448 | } |
Oleg Nesterov | 297bd42 | 2008-02-08 04:19:10 -0800 | [diff] [blame] | 449 | set_special_pids(&init_struct_pid); |
Peter Zijlstra | 24ec839 | 2006-12-08 02:36:04 -0800 | [diff] [blame] | 450 | proc_clear_tty(current); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 451 | |
| 452 | /* Block and flush all signals */ |
| 453 | sigfillset(&blocked); |
| 454 | sigprocmask(SIG_BLOCK, &blocked, NULL); |
| 455 | flush_signals(current); |
| 456 | |
| 457 | /* Become as one with the init task */ |
| 458 | |
Al Viro | 3e93cd6 | 2009-03-29 19:00:13 -0400 | [diff] [blame] | 459 | daemonize_fs_struct(); |
Daniel Walker | d4c5e41 | 2007-10-18 23:39:59 -0700 | [diff] [blame] | 460 | exit_files(current); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 461 | current->files = init_task.files; |
| 462 | atomic_inc(¤t->files->count); |
| 463 | |
Eric W. Biederman | 49d769d | 2007-05-09 02:34:33 -0700 | [diff] [blame] | 464 | reparent_to_kthreadd(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 465 | } |
| 466 | |
| 467 | EXPORT_SYMBOL(daemonize); |
| 468 | |
Arjan van de Ven | 858119e | 2006-01-14 13:20:43 -0800 | [diff] [blame] | 469 | static void close_files(struct files_struct * files) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 470 | { |
| 471 | int i, j; |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 472 | struct fdtable *fdt; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 473 | |
| 474 | j = 0; |
Dipankar Sarma | 4fb3a53 | 2005-09-16 19:28:13 -0700 | [diff] [blame] | 475 | |
| 476 | /* |
| 477 | * It is safe to dereference the fd table without RCU or |
| 478 | * ->file_lock because this is the last reference to the |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 479 | * files structure. But use RCU to shut RCU-lockdep up. |
Dipankar Sarma | 4fb3a53 | 2005-09-16 19:28:13 -0700 | [diff] [blame] | 480 | */ |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 481 | rcu_read_lock(); |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 482 | fdt = files_fdtable(files); |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 483 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 484 | for (;;) { |
| 485 | unsigned long set; |
| 486 | i = j * __NFDBITS; |
Vadim Lobanov | bbea9f6 | 2006-12-10 02:21:12 -0800 | [diff] [blame] | 487 | if (i >= fdt->max_fds) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 488 | break; |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 489 | set = fdt->open_fds->fds_bits[j++]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 490 | while (set) { |
| 491 | if (set & 1) { |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 492 | struct file * file = xchg(&fdt->fd[i], NULL); |
Ingo Molnar | 944be0b | 2007-02-12 00:52:26 -0800 | [diff] [blame] | 493 | if (file) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 494 | filp_close(file, files); |
Ingo Molnar | 944be0b | 2007-02-12 00:52:26 -0800 | [diff] [blame] | 495 | cond_resched(); |
| 496 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 497 | } |
| 498 | i++; |
| 499 | set >>= 1; |
| 500 | } |
| 501 | } |
| 502 | } |
| 503 | |
| 504 | struct files_struct *get_files_struct(struct task_struct *task) |
| 505 | { |
| 506 | struct files_struct *files; |
| 507 | |
| 508 | task_lock(task); |
| 509 | files = task->files; |
| 510 | if (files) |
| 511 | atomic_inc(&files->count); |
| 512 | task_unlock(task); |
| 513 | |
| 514 | return files; |
| 515 | } |
| 516 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 517 | void put_files_struct(struct files_struct *files) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 518 | { |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 519 | struct fdtable *fdt; |
| 520 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 521 | if (atomic_dec_and_test(&files->count)) { |
| 522 | close_files(files); |
| 523 | /* |
| 524 | * Free the fd and fdset arrays if we expanded them. |
Dipankar Sarma | ab2af1f | 2005-09-09 13:04:13 -0700 | [diff] [blame] | 525 | * If the fdtable was embedded, pass files for freeing |
| 526 | * at the end of the RCU grace period. Otherwise, |
| 527 | * you can free files immediately. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 528 | */ |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 529 | rcu_read_lock(); |
Dipankar Sarma | badf166 | 2005-09-09 13:04:10 -0700 | [diff] [blame] | 530 | fdt = files_fdtable(files); |
Vadim Lobanov | 4fd4581 | 2006-12-10 02:21:17 -0800 | [diff] [blame] | 531 | if (fdt != &files->fdtab) |
Dipankar Sarma | ab2af1f | 2005-09-09 13:04:13 -0700 | [diff] [blame] | 532 | kmem_cache_free(files_cachep, files); |
Vadim Lobanov | 01b2d93 | 2006-12-22 01:10:43 -0800 | [diff] [blame] | 533 | free_fdtable(fdt); |
Paul E. McKenney | d11c563 | 2010-02-22 17:04:50 -0800 | [diff] [blame] | 534 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 535 | } |
| 536 | } |
| 537 | |
Al Viro | 3b12538 | 2008-04-22 05:31:30 -0400 | [diff] [blame] | 538 | void reset_files_struct(struct files_struct *files) |
Kirill Korotaev | 3b9b8ab | 2006-09-29 02:00:05 -0700 | [diff] [blame] | 539 | { |
Al Viro | 3b12538 | 2008-04-22 05:31:30 -0400 | [diff] [blame] | 540 | struct task_struct *tsk = current; |
Kirill Korotaev | 3b9b8ab | 2006-09-29 02:00:05 -0700 | [diff] [blame] | 541 | struct files_struct *old; |
| 542 | |
| 543 | old = tsk->files; |
| 544 | task_lock(tsk); |
| 545 | tsk->files = files; |
| 546 | task_unlock(tsk); |
| 547 | put_files_struct(old); |
| 548 | } |
Kirill Korotaev | 3b9b8ab | 2006-09-29 02:00:05 -0700 | [diff] [blame] | 549 | |
Al Viro | 1ec7f1d | 2008-04-22 05:35:42 -0400 | [diff] [blame] | 550 | void exit_files(struct task_struct *tsk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 551 | { |
| 552 | struct files_struct * files = tsk->files; |
| 553 | |
| 554 | if (files) { |
| 555 | task_lock(tsk); |
| 556 | tsk->files = NULL; |
| 557 | task_unlock(tsk); |
| 558 | put_files_struct(files); |
| 559 | } |
| 560 | } |
| 561 | |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 562 | #ifdef CONFIG_MM_OWNER |
| 563 | /* |
| 564 | * Task p is exiting and it owned mm, lets find a new owner for it |
| 565 | */ |
| 566 | static inline int |
| 567 | mm_need_new_owner(struct mm_struct *mm, struct task_struct *p) |
| 568 | { |
| 569 | /* |
| 570 | * If there are other users of the mm and the owner (us) is exiting |
| 571 | * we need to find a new owner to take on the responsibility. |
| 572 | */ |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 573 | if (atomic_read(&mm->mm_users) <= 1) |
| 574 | return 0; |
| 575 | if (mm->owner != p) |
| 576 | return 0; |
| 577 | return 1; |
| 578 | } |
| 579 | |
| 580 | void mm_update_next_owner(struct mm_struct *mm) |
| 581 | { |
| 582 | struct task_struct *c, *g, *p = current; |
| 583 | |
| 584 | retry: |
| 585 | if (!mm_need_new_owner(mm, p)) |
| 586 | return; |
| 587 | |
| 588 | read_lock(&tasklist_lock); |
| 589 | /* |
| 590 | * Search in the children |
| 591 | */ |
| 592 | list_for_each_entry(c, &p->children, sibling) { |
| 593 | if (c->mm == mm) |
| 594 | goto assign_new_owner; |
| 595 | } |
| 596 | |
| 597 | /* |
| 598 | * Search in the siblings |
| 599 | */ |
Oleg Nesterov | dea33cf | 2009-06-17 16:27:29 -0700 | [diff] [blame] | 600 | list_for_each_entry(c, &p->real_parent->children, sibling) { |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 601 | if (c->mm == mm) |
| 602 | goto assign_new_owner; |
| 603 | } |
| 604 | |
| 605 | /* |
| 606 | * Search through everything else. We should not get |
| 607 | * here often |
| 608 | */ |
| 609 | do_each_thread(g, c) { |
| 610 | if (c->mm == mm) |
| 611 | goto assign_new_owner; |
| 612 | } while_each_thread(g, c); |
| 613 | |
| 614 | read_unlock(&tasklist_lock); |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 615 | /* |
| 616 | * We found no owner yet mm_users > 1: this implies that we are |
| 617 | * most likely racing with swapoff (try_to_unuse()) or /proc or |
Hugh Dickins | e599137 | 2009-01-06 14:39:22 -0800 | [diff] [blame] | 618 | * ptrace or page migration (get_task_mm()). Mark owner as NULL. |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 619 | */ |
Balbir Singh | 31a78f2 | 2008-09-28 23:09:31 +0100 | [diff] [blame] | 620 | mm->owner = NULL; |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 621 | return; |
| 622 | |
| 623 | assign_new_owner: |
| 624 | BUG_ON(c == p); |
| 625 | get_task_struct(c); |
| 626 | /* |
| 627 | * The task_lock protects c->mm from changing. |
| 628 | * We always want mm->owner->mm == mm |
| 629 | */ |
| 630 | task_lock(c); |
Hugh Dickins | e599137 | 2009-01-06 14:39:22 -0800 | [diff] [blame] | 631 | /* |
| 632 | * Delay read_unlock() till we have the task_lock() |
| 633 | * to ensure that c does not slip away underneath us |
| 634 | */ |
| 635 | read_unlock(&tasklist_lock); |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 636 | if (c->mm != mm) { |
| 637 | task_unlock(c); |
| 638 | put_task_struct(c); |
| 639 | goto retry; |
| 640 | } |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 641 | mm->owner = c; |
| 642 | task_unlock(c); |
| 643 | put_task_struct(c); |
| 644 | } |
| 645 | #endif /* CONFIG_MM_OWNER */ |
| 646 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 647 | /* |
| 648 | * Turn us into a lazy TLB process if we |
| 649 | * aren't already.. |
| 650 | */ |
Adrian Bunk | 408b664 | 2005-05-01 08:59:29 -0700 | [diff] [blame] | 651 | static void exit_mm(struct task_struct * tsk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 652 | { |
| 653 | struct mm_struct *mm = tsk->mm; |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 654 | struct core_state *core_state; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 655 | |
| 656 | mm_release(tsk, mm); |
| 657 | if (!mm) |
| 658 | return; |
| 659 | /* |
| 660 | * Serialize with any possible pending coredump. |
Oleg Nesterov | 999d9fc | 2008-07-25 01:47:41 -0700 | [diff] [blame] | 661 | * We must hold mmap_sem around checking core_state |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 662 | * and clearing tsk->mm. The core-inducing thread |
Oleg Nesterov | 999d9fc | 2008-07-25 01:47:41 -0700 | [diff] [blame] | 663 | * will increment ->nr_threads for each thread in the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 664 | * group with ->mm != NULL. |
| 665 | */ |
| 666 | down_read(&mm->mmap_sem); |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 667 | core_state = mm->core_state; |
| 668 | if (core_state) { |
| 669 | struct core_thread self; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 670 | up_read(&mm->mmap_sem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 671 | |
Oleg Nesterov | b564daf | 2008-07-25 01:47:44 -0700 | [diff] [blame] | 672 | self.task = tsk; |
| 673 | self.next = xchg(&core_state->dumper.next, &self); |
| 674 | /* |
| 675 | * Implies mb(), the result of xchg() must be visible |
| 676 | * to core_state->dumper. |
| 677 | */ |
| 678 | if (atomic_dec_and_test(&core_state->nr_threads)) |
| 679 | complete(&core_state->startup); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 680 | |
Oleg Nesterov | a94e2d4 | 2008-07-25 01:47:46 -0700 | [diff] [blame] | 681 | for (;;) { |
| 682 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
| 683 | if (!self.task) /* see coredump_finish() */ |
| 684 | break; |
| 685 | schedule(); |
| 686 | } |
| 687 | __set_task_state(tsk, TASK_RUNNING); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 688 | down_read(&mm->mmap_sem); |
| 689 | } |
| 690 | atomic_inc(&mm->mm_count); |
Eric Sesterhenn | 125e187 | 2006-06-23 02:06:06 -0700 | [diff] [blame] | 691 | BUG_ON(mm != tsk->active_mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 692 | /* more a memory barrier than a real lock */ |
| 693 | task_lock(tsk); |
| 694 | tsk->mm = NULL; |
| 695 | up_read(&mm->mmap_sem); |
| 696 | enter_lazy_tlb(mm, current); |
Rafael J. Wysocki | 0c1eecf | 2007-07-19 01:47:33 -0700 | [diff] [blame] | 697 | /* We don't want this task to be frozen prematurely */ |
| 698 | clear_freeze_flag(tsk); |
Ying Han | 3d5992d | 2010-10-26 14:21:23 -0700 | [diff] [blame] | 699 | if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) |
| 700 | atomic_dec(&mm->oom_disable_count); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 701 | task_unlock(tsk); |
Balbir Singh | cf475ad | 2008-04-29 01:00:16 -0700 | [diff] [blame] | 702 | mm_update_next_owner(mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 703 | mmput(mm); |
| 704 | } |
| 705 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 706 | /* |
| 707 | * When we die, we re-parent all our children. |
| 708 | * Try to give them to another thread in our thread |
| 709 | * group, and if no such member exists, give it to |
Sukadev Bhattiprolu | 84d7378 | 2006-12-08 02:38:01 -0800 | [diff] [blame] | 710 | * the child reaper process (ie "init") in our pid |
| 711 | * space. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 712 | */ |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 713 | static struct task_struct *find_new_reaper(struct task_struct *father) |
Namhyung Kim | d16e15f | 2010-10-27 15:34:10 -0700 | [diff] [blame] | 714 | __releases(&tasklist_lock) |
| 715 | __acquires(&tasklist_lock) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 716 | { |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 717 | struct pid_namespace *pid_ns = task_active_pid_ns(father); |
| 718 | struct task_struct *thread; |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 719 | |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 720 | thread = father; |
| 721 | while_each_thread(father, thread) { |
| 722 | if (thread->flags & PF_EXITING) |
| 723 | continue; |
| 724 | if (unlikely(pid_ns->child_reaper == father)) |
| 725 | pid_ns->child_reaper = thread; |
| 726 | return thread; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 727 | } |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 728 | |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 729 | if (unlikely(pid_ns->child_reaper == father)) { |
| 730 | write_unlock_irq(&tasklist_lock); |
| 731 | if (unlikely(pid_ns == &init_pid_ns)) |
| 732 | panic("Attempted to kill init!"); |
| 733 | |
| 734 | zap_pid_ns_processes(pid_ns); |
| 735 | write_lock_irq(&tasklist_lock); |
| 736 | /* |
| 737 | * We can not clear ->child_reaper or leave it alone. |
| 738 | * There may by stealth EXIT_DEAD tasks on ->children, |
| 739 | * forget_original_parent() must move them somewhere. |
| 740 | */ |
| 741 | pid_ns->child_reaper = init_pid_ns.child_reaper; |
| 742 | } |
| 743 | |
| 744 | return pid_ns->child_reaper; |
| 745 | } |
| 746 | |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 747 | /* |
| 748 | * Any that need to be release_task'd are put on the @dead list. |
| 749 | */ |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 750 | static void reparent_leader(struct task_struct *father, struct task_struct *p, |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 751 | struct list_head *dead) |
| 752 | { |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 753 | list_move_tail(&p->sibling, &p->real_parent->children); |
| 754 | |
| 755 | if (task_detached(p)) |
| 756 | return; |
| 757 | /* |
| 758 | * If this is a threaded reparent there is no need to |
| 759 | * notify anyone anything has happened. |
| 760 | */ |
| 761 | if (same_thread_group(p->real_parent, father)) |
| 762 | return; |
| 763 | |
| 764 | /* We don't want people slaying init. */ |
| 765 | p->exit_signal = SIGCHLD; |
| 766 | |
| 767 | /* 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] | 768 | if (!task_ptrace(p) && |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 769 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { |
| 770 | do_notify_parent(p, p->exit_signal); |
| 771 | if (task_detached(p)) { |
| 772 | p->exit_state = EXIT_DEAD; |
| 773 | list_move_tail(&p->sibling, dead); |
| 774 | } |
| 775 | } |
| 776 | |
| 777 | kill_orphaned_pgrp(p, father); |
| 778 | } |
| 779 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 780 | static void forget_original_parent(struct task_struct *father) |
| 781 | { |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 782 | struct task_struct *p, *n, *reaper; |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 783 | LIST_HEAD(dead_children); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 784 | |
| 785 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | c7e49c1 | 2010-08-10 18:03:07 -0700 | [diff] [blame] | 786 | /* |
| 787 | * Note that exit_ptrace() and find_new_reaper() might |
| 788 | * drop tasklist_lock and reacquire it. |
| 789 | */ |
| 790 | exit_ptrace(father); |
Oleg Nesterov | 950bbab | 2008-09-02 14:35:49 -0700 | [diff] [blame] | 791 | reaper = find_new_reaper(father); |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 792 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 793 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 794 | struct task_struct *t = p; |
| 795 | do { |
| 796 | t->real_parent = reaper; |
| 797 | if (t->parent == father) { |
| 798 | BUG_ON(task_ptrace(t)); |
| 799 | t->parent = t->real_parent; |
| 800 | } |
| 801 | if (t->pdeath_signal) |
| 802 | group_send_sig_info(t->pdeath_signal, |
| 803 | SEND_SIG_NOINFO, t); |
| 804 | } while_each_thread(p, t); |
| 805 | reparent_leader(father, p, &dead_children); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 806 | } |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 807 | write_unlock_irq(&tasklist_lock); |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 808 | |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 809 | BUG_ON(!list_empty(&father->children)); |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 810 | |
Oleg Nesterov | 5dfc80b | 2009-04-02 16:58:19 -0700 | [diff] [blame] | 811 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
| 812 | list_del_init(&p->sibling); |
Oleg Nesterov | 39c626a | 2009-04-02 16:58:18 -0700 | [diff] [blame] | 813 | release_task(p); |
| 814 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 815 | } |
| 816 | |
| 817 | /* |
| 818 | * Send signals to all our closest relatives so that they know |
| 819 | * to properly mourn us.. |
| 820 | */ |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 821 | static void exit_notify(struct task_struct *tsk, int group_dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 822 | { |
Roland McGrath | 2b2a1ff | 2008-07-25 19:45:54 -0700 | [diff] [blame] | 823 | int signal; |
| 824 | void *cookie; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 825 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 826 | /* |
| 827 | * This does two things: |
| 828 | * |
| 829 | * A. Make init inherit all the child processes |
| 830 | * B. Check to see if any process groups have become orphaned |
| 831 | * as a result of our exiting, and if they have any stopped |
| 832 | * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2) |
| 833 | */ |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 834 | forget_original_parent(tsk); |
Pavel Emelyanov | 2e4a707 | 2007-10-18 23:40:01 -0700 | [diff] [blame] | 835 | exit_task_namespaces(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 836 | |
Oleg Nesterov | 762a24b | 2007-10-18 23:40:00 -0700 | [diff] [blame] | 837 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 838 | if (group_dead) |
| 839 | kill_orphaned_pgrp(tsk->group_leader, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 840 | |
Oleg Nesterov | 2472844 | 2007-08-04 01:04:41 +0400 | [diff] [blame] | 841 | /* Let father know we died |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 842 | * |
| 843 | * Thread signals are configurable, but you aren't going to use |
Daniel Walker | d4c5e41 | 2007-10-18 23:39:59 -0700 | [diff] [blame] | 844 | * that to send signals to arbitary processes. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 845 | * That stops right now. |
| 846 | * |
| 847 | * If the parent exec id doesn't match the exec id we saved |
| 848 | * when we started then we know the parent has changed security |
| 849 | * domain. |
| 850 | * |
| 851 | * If our self_exec id doesn't match our parent_exec_id then |
| 852 | * we have changed execution domain as these two values started |
| 853 | * the same after a fork. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 854 | */ |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 855 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
Oleg Nesterov | f49ee50 | 2008-03-02 21:44:40 +0300 | [diff] [blame] | 856 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
Oleg Nesterov | 432870d | 2009-04-06 16:16:02 +0200 | [diff] [blame] | 857 | tsk->self_exec_id != tsk->parent_exec_id)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 858 | tsk->exit_signal = SIGCHLD; |
| 859 | |
Roland McGrath | 2b2a1ff | 2008-07-25 19:45:54 -0700 | [diff] [blame] | 860 | signal = tracehook_notify_death(tsk, &cookie, group_dead); |
Roland McGrath | 5c7edcd | 2008-07-31 02:04:09 -0700 | [diff] [blame] | 861 | if (signal >= 0) |
Roland McGrath | 2b2a1ff | 2008-07-25 19:45:54 -0700 | [diff] [blame] | 862 | signal = do_notify_parent(tsk, signal); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 863 | |
Roland McGrath | 5c7edcd | 2008-07-31 02:04:09 -0700 | [diff] [blame] | 864 | tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 865 | |
Oleg Nesterov | 9c33916 | 2010-05-26 14:43:10 -0700 | [diff] [blame] | 866 | /* mt-exec, de_thread() is waiting for group leader */ |
| 867 | if (unlikely(tsk->signal->notify_count < 0)) |
Oleg Nesterov | 6db840f | 2007-10-16 23:27:23 -0700 | [diff] [blame] | 868 | wake_up_process(tsk->signal->group_exit_task); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 869 | write_unlock_irq(&tasklist_lock); |
| 870 | |
Roland McGrath | 2b2a1ff | 2008-07-25 19:45:54 -0700 | [diff] [blame] | 871 | tracehook_report_death(tsk, signal, cookie, group_dead); |
| 872 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 873 | /* If the process is dead, release it - nobody will wait for it */ |
Roland McGrath | 5c7edcd | 2008-07-31 02:04:09 -0700 | [diff] [blame] | 874 | if (signal == DEATH_REAP) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 875 | release_task(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 876 | } |
| 877 | |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 878 | #ifdef CONFIG_DEBUG_STACK_USAGE |
| 879 | static void check_stack_usage(void) |
| 880 | { |
| 881 | static DEFINE_SPINLOCK(low_water_lock); |
| 882 | static int lowest_to_date = THREAD_SIZE; |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 883 | unsigned long free; |
| 884 | |
Eric Sandeen | 7c9f886 | 2008-04-22 16:38:23 -0500 | [diff] [blame] | 885 | free = stack_not_used(current); |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 886 | |
| 887 | if (free >= lowest_to_date) |
| 888 | return; |
| 889 | |
| 890 | spin_lock(&low_water_lock); |
| 891 | if (free < lowest_to_date) { |
| 892 | printk(KERN_WARNING "%s used greatest stack depth: %lu bytes " |
| 893 | "left\n", |
| 894 | current->comm, free); |
| 895 | lowest_to_date = free; |
| 896 | } |
| 897 | spin_unlock(&low_water_lock); |
| 898 | } |
| 899 | #else |
| 900 | static inline void check_stack_usage(void) {} |
| 901 | #endif |
| 902 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 903 | NORET_TYPE void do_exit(long code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 904 | { |
| 905 | struct task_struct *tsk = current; |
| 906 | int group_dead; |
| 907 | |
| 908 | profile_task_exit(tsk); |
| 909 | |
Jens Axboe | 22e2c50 | 2005-06-27 10:55:12 +0200 | [diff] [blame] | 910 | WARN_ON(atomic_read(&tsk->fs_excl)); |
| 911 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 912 | if (unlikely(in_interrupt())) |
| 913 | panic("Aiee, killing interrupt handler!"); |
| 914 | if (unlikely(!tsk->pid)) |
| 915 | panic("Attempted to kill the idle task!"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 916 | |
Roland McGrath | 30199f5 | 2008-07-25 19:45:46 -0700 | [diff] [blame] | 917 | tracehook_report_exit(&code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 918 | |
David Howells | e0e8173 | 2009-09-02 09:13:40 +0100 | [diff] [blame] | 919 | validate_creds_for_do_exit(tsk); |
| 920 | |
Alexander Nyberg | df164db | 2005-06-23 00:09:13 -0700 | [diff] [blame] | 921 | /* |
| 922 | * We're taking recursive faults here in do_exit. Safest is to just |
| 923 | * leave this task alone and wait for reboot. |
| 924 | */ |
| 925 | if (unlikely(tsk->flags & PF_EXITING)) { |
| 926 | printk(KERN_ALERT |
| 927 | "Fixing recursive fault but reboot is needed!\n"); |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 928 | /* |
| 929 | * We can do this unlocked here. The futex code uses |
| 930 | * this flag just to verify whether the pi state |
| 931 | * cleanup has been done or not. In the worst case it |
| 932 | * loops once more. We pretend that the cleanup was |
| 933 | * done as there is no way to return. Either the |
| 934 | * OWNER_DIED bit is set by now or we push the blocked |
| 935 | * task into the wait for ever nirwana as well. |
| 936 | */ |
| 937 | tsk->flags |= PF_EXITPIDONE; |
Alexander Nyberg | df164db | 2005-06-23 00:09:13 -0700 | [diff] [blame] | 938 | set_current_state(TASK_UNINTERRUPTIBLE); |
| 939 | schedule(); |
| 940 | } |
| 941 | |
Thomas Gleixner | 3aa551c | 2009-03-23 18:28:15 +0100 | [diff] [blame] | 942 | exit_irq_thread(); |
| 943 | |
Oleg Nesterov | d12619b | 2008-02-08 04:19:12 -0800 | [diff] [blame] | 944 | exit_signals(tsk); /* sets PF_EXITING */ |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 945 | /* |
| 946 | * tsk->flags are checked in the futex code to protect against |
| 947 | * an exiting task cleaning up the robust pi futexes. |
| 948 | */ |
Oleg Nesterov | d2ee719 | 2007-10-16 23:26:47 -0700 | [diff] [blame] | 949 | smp_mb(); |
Thomas Gleixner | 1d61548 | 2009-11-17 14:54:03 +0100 | [diff] [blame] | 950 | raw_spin_unlock_wait(&tsk->pi_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 951 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 952 | if (unlikely(in_atomic())) |
| 953 | printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n", |
Pavel Emelyanov | ba25f9d | 2007-10-18 23:40:40 -0700 | [diff] [blame] | 954 | current->comm, task_pid_nr(current), |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 955 | preempt_count()); |
| 956 | |
| 957 | acct_update_integrals(tsk); |
KAMEZAWA Hiroyuki | 34e5523 | 2010-03-05 13:41:40 -0800 | [diff] [blame] | 958 | /* sync mm's RSS info before statistics gathering */ |
KAMEZAWA Hiroyuki | a3a2e76 | 2010-04-06 14:34:42 -0700 | [diff] [blame] | 959 | if (tsk->mm) |
| 960 | sync_mm_rss(tsk, tsk->mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 961 | group_dead = atomic_dec_and_test(&tsk->signal->live); |
Andrew Morton | c306895 | 2005-08-04 16:49:32 -0700 | [diff] [blame] | 962 | if (group_dead) { |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 963 | hrtimer_cancel(&tsk->signal->real_timer); |
Roland McGrath | 25f407f | 2005-10-21 15:03:29 -0700 | [diff] [blame] | 964 | exit_itimers(tsk->signal); |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 965 | if (tsk->mm) |
| 966 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, tsk->mm); |
Andrew Morton | c306895 | 2005-08-04 16:49:32 -0700 | [diff] [blame] | 967 | } |
KaiGai Kohei | f6ec29a | 2006-06-25 05:49:25 -0700 | [diff] [blame] | 968 | acct_collect(code, group_dead); |
Miloslav Trmac | 522ed77 | 2007-07-15 23:40:56 -0700 | [diff] [blame] | 969 | if (group_dead) |
| 970 | tty_audit_exit(); |
Al Viro | fa84cb9 | 2006-03-29 20:30:19 -0500 | [diff] [blame] | 971 | if (unlikely(tsk->audit_context)) |
| 972 | audit_free(tsk); |
Oleg Nesterov | 115085e | 2006-12-06 20:36:51 -0800 | [diff] [blame] | 973 | |
Jonathan Lim | f2ab6d8 | 2007-08-30 23:56:23 -0700 | [diff] [blame] | 974 | tsk->exit_code = code; |
Oleg Nesterov | 115085e | 2006-12-06 20:36:51 -0800 | [diff] [blame] | 975 | taskstats_exit(tsk, group_dead); |
Shailabh Nagar | c757249 | 2006-07-14 00:24:40 -0700 | [diff] [blame] | 976 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 977 | exit_mm(tsk); |
| 978 | |
KaiGai Kohei | 0e46481 | 2006-06-25 05:49:24 -0700 | [diff] [blame] | 979 | if (group_dead) |
KaiGai Kohei | f6ec29a | 2006-06-25 05:49:25 -0700 | [diff] [blame] | 980 | acct_process(); |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 981 | trace_sched_process_exit(tsk); |
| 982 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 983 | exit_sem(tsk); |
Al Viro | 1ec7f1d | 2008-04-22 05:35:42 -0400 | [diff] [blame] | 984 | exit_files(tsk); |
| 985 | exit_fs(tsk); |
Jeff Dike | e18eecb | 2007-07-15 23:38:48 -0700 | [diff] [blame] | 986 | check_stack_usage(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 987 | exit_thread(); |
Paul Menage | b4f48b6 | 2007-10-18 23:39:33 -0700 | [diff] [blame] | 988 | cgroup_exit(tsk, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 989 | |
Alan Cox | 5ec93d1 | 2009-11-30 13:18:45 +0000 | [diff] [blame] | 990 | if (group_dead) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 991 | disassociate_ctty(1); |
| 992 | |
Al Viro | a1261f5 | 2005-11-13 16:06:55 -0800 | [diff] [blame] | 993 | module_put(task_thread_info(tsk)->exec_domain->module); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 994 | |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 995 | proc_exit_connector(tsk); |
Ingo Molnar | 33b2fb3 | 2009-05-17 11:08:41 +0200 | [diff] [blame] | 996 | |
| 997 | /* |
Frederic Weisbecker | 24f1e32c | 2009-09-09 19:22:48 +0200 | [diff] [blame] | 998 | * FIXME: do that only when needed, using sched_exit tracepoint |
| 999 | */ |
| 1000 | flush_ptrace_hw_breakpoint(tsk); |
| 1001 | /* |
Ingo Molnar | 33b2fb3 | 2009-05-17 11:08:41 +0200 | [diff] [blame] | 1002 | * Flush inherited counters to the parent - before the parent |
| 1003 | * gets woken up by child-exit notifications. |
| 1004 | */ |
Ingo Molnar | cdd6c48 | 2009-09-21 12:02:48 +0200 | [diff] [blame] | 1005 | perf_event_exit_task(tsk); |
Ingo Molnar | 33b2fb3 | 2009-05-17 11:08:41 +0200 | [diff] [blame] | 1006 | |
Oleg Nesterov | 821c7de | 2008-03-02 21:44:44 +0300 | [diff] [blame] | 1007 | exit_notify(tsk, group_dead); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1008 | #ifdef CONFIG_NUMA |
Miao Xie | c0ff745 | 2010-05-24 14:32:08 -0700 | [diff] [blame] | 1009 | task_lock(tsk); |
Lee Schermerhorn | f0be3d3 | 2008-04-28 02:13:08 -0700 | [diff] [blame] | 1010 | mpol_put(tsk->mempolicy); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1011 | tsk->mempolicy = NULL; |
Miao Xie | c0ff745 | 2010-05-24 14:32:08 -0700 | [diff] [blame] | 1012 | task_unlock(tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1013 | #endif |
Alexey Dobriyan | 42b2dd0 | 2007-10-16 23:27:30 -0700 | [diff] [blame] | 1014 | #ifdef CONFIG_FUTEX |
Ingo Molnar | c87e283 | 2006-06-27 02:54:58 -0700 | [diff] [blame] | 1015 | if (unlikely(current->pi_state_cache)) |
| 1016 | kfree(current->pi_state_cache); |
Alexey Dobriyan | 42b2dd0 | 2007-10-16 23:27:30 -0700 | [diff] [blame] | 1017 | #endif |
Ingo Molnar | c87e283 | 2006-06-27 02:54:58 -0700 | [diff] [blame] | 1018 | /* |
Ingo Molnar | 9a11b49a | 2006-07-03 00:24:33 -0700 | [diff] [blame] | 1019 | * Make sure we are holding no locks: |
Ingo Molnar | de5097c | 2006-01-09 15:59:21 -0800 | [diff] [blame] | 1020 | */ |
Ingo Molnar | 9a11b49a | 2006-07-03 00:24:33 -0700 | [diff] [blame] | 1021 | debug_check_no_locks_held(tsk); |
Alexey Kuznetsov | 778e9a9 | 2007-06-08 13:47:00 -0700 | [diff] [blame] | 1022 | /* |
| 1023 | * We can do this unlocked here. The futex code uses this flag |
| 1024 | * just to verify whether the pi state cleanup has been done |
| 1025 | * or not. In the worst case it loops once more. |
| 1026 | */ |
| 1027 | tsk->flags |= PF_EXITPIDONE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1028 | |
Al Viro | afc847b | 2006-02-28 12:51:55 -0500 | [diff] [blame] | 1029 | if (tsk->io_context) |
Louis Rilling | b69f229 | 2009-12-04 14:52:42 +0100 | [diff] [blame] | 1030 | exit_io_context(tsk); |
Al Viro | afc847b | 2006-02-28 12:51:55 -0500 | [diff] [blame] | 1031 | |
Jens Axboe | b92ce55 | 2006-04-11 13:52:07 +0200 | [diff] [blame] | 1032 | if (tsk->splice_pipe) |
| 1033 | __free_pipe_info(tsk->splice_pipe); |
| 1034 | |
David Howells | e0e8173 | 2009-09-02 09:13:40 +0100 | [diff] [blame] | 1035 | validate_creds_for_do_exit(tsk); |
| 1036 | |
Coywolf Qi Hunt | 7407251 | 2005-10-30 15:02:47 -0800 | [diff] [blame] | 1037 | preempt_disable(); |
Paul E. McKenney | f41d911 | 2009-08-22 13:56:52 -0700 | [diff] [blame] | 1038 | exit_rcu(); |
Oleg Nesterov | 55a101f | 2006-09-29 02:01:10 -0700 | [diff] [blame] | 1039 | /* causes final put_task_struct in finish_task_switch(). */ |
Oleg Nesterov | c394cc9 | 2006-09-29 02:01:11 -0700 | [diff] [blame] | 1040 | tsk->state = TASK_DEAD; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1041 | schedule(); |
| 1042 | BUG(); |
| 1043 | /* Avoid "noreturn function does return". */ |
Alan Cox | 54306cf | 2006-09-29 02:00:42 -0700 | [diff] [blame] | 1044 | for (;;) |
| 1045 | cpu_relax(); /* For when BUG is null */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1046 | } |
| 1047 | |
Russ Anderson | 012914d | 2005-04-23 00:08:00 -0700 | [diff] [blame] | 1048 | EXPORT_SYMBOL_GPL(do_exit); |
| 1049 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1050 | NORET_TYPE void complete_and_exit(struct completion *comp, long code) |
| 1051 | { |
| 1052 | if (comp) |
| 1053 | complete(comp); |
Oleg Nesterov | 55a101f | 2006-09-29 02:01:10 -0700 | [diff] [blame] | 1054 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1055 | do_exit(code); |
| 1056 | } |
| 1057 | |
| 1058 | EXPORT_SYMBOL(complete_and_exit); |
| 1059 | |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 1060 | SYSCALL_DEFINE1(exit, int, error_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1061 | { |
| 1062 | do_exit((error_code&0xff)<<8); |
| 1063 | } |
| 1064 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1065 | /* |
| 1066 | * Take down every thread in the group. This is called by fatal signals |
| 1067 | * as well as by sys_exit_group (below). |
| 1068 | */ |
| 1069 | NORET_TYPE void |
| 1070 | do_group_exit(int exit_code) |
| 1071 | { |
Oleg Nesterov | bfc4b08 | 2008-04-30 00:52:36 -0700 | [diff] [blame] | 1072 | struct signal_struct *sig = current->signal; |
| 1073 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1074 | BUG_ON(exit_code & 0x80); /* core dumps don't get here */ |
| 1075 | |
Oleg Nesterov | bfc4b08 | 2008-04-30 00:52:36 -0700 | [diff] [blame] | 1076 | if (signal_group_exit(sig)) |
| 1077 | exit_code = sig->group_exit_code; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1078 | else if (!thread_group_empty(current)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1079 | struct sighand_struct *const sighand = current->sighand; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1080 | spin_lock_irq(&sighand->siglock); |
Oleg Nesterov | ed5d2ca | 2008-02-04 22:27:24 -0800 | [diff] [blame] | 1081 | if (signal_group_exit(sig)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1082 | /* Another thread got here before we took the lock. */ |
| 1083 | exit_code = sig->group_exit_code; |
| 1084 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1085 | sig->group_exit_code = exit_code; |
Oleg Nesterov | ed5d2ca | 2008-02-04 22:27:24 -0800 | [diff] [blame] | 1086 | sig->flags = SIGNAL_GROUP_EXIT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1087 | zap_other_threads(current); |
| 1088 | } |
| 1089 | spin_unlock_irq(&sighand->siglock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1090 | } |
| 1091 | |
| 1092 | do_exit(exit_code); |
| 1093 | /* NOTREACHED */ |
| 1094 | } |
| 1095 | |
| 1096 | /* |
| 1097 | * this kills every thread in the thread group. Note that any externally |
| 1098 | * wait4()-ing process will get the correct exit code - even if this |
| 1099 | * thread is not the thread group leader. |
| 1100 | */ |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 1101 | SYSCALL_DEFINE1(exit_group, int, error_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1102 | { |
| 1103 | do_group_exit((error_code & 0xff) << 8); |
Heiko Carstens | 2ed7c03 | 2009-01-14 14:13:54 +0100 | [diff] [blame] | 1104 | /* NOTREACHED */ |
| 1105 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1106 | } |
| 1107 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1108 | struct wait_opts { |
| 1109 | enum pid_type wo_type; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1110 | int wo_flags; |
Richard Kennedy | e1eb1eb | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 1111 | struct pid *wo_pid; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1112 | |
| 1113 | struct siginfo __user *wo_info; |
| 1114 | int __user *wo_stat; |
| 1115 | struct rusage __user *wo_rusage; |
| 1116 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1117 | wait_queue_t child_wait; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1118 | int notask_error; |
| 1119 | }; |
| 1120 | |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 1121 | static inline |
| 1122 | 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] | 1123 | { |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 1124 | if (type != PIDTYPE_PID) |
| 1125 | task = task->group_leader; |
| 1126 | return task->pids[type].pid; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1127 | } |
| 1128 | |
Oleg Nesterov | 989264f | 2009-09-23 15:56:49 -0700 | [diff] [blame] | 1129 | static int eligible_pid(struct wait_opts *wo, struct task_struct *p) |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 1130 | { |
| 1131 | return wo->wo_type == PIDTYPE_MAX || |
| 1132 | task_pid_type(p, wo->wo_type) == wo->wo_pid; |
| 1133 | } |
| 1134 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1135 | static int eligible_child(struct wait_opts *wo, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1136 | { |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 1137 | if (!eligible_pid(wo, p)) |
| 1138 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1139 | /* Wait for all children (clone and not) if __WALL is set; |
| 1140 | * otherwise, wait for clone children *only* if __WCLONE is |
| 1141 | * set; otherwise, wait for non-clone children *only*. (Note: |
| 1142 | * A "clone" child here is one that reports to its parent |
| 1143 | * using a signal other than SIGCHLD.) */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1144 | if (((p->exit_signal != SIGCHLD) ^ !!(wo->wo_flags & __WCLONE)) |
| 1145 | && !(wo->wo_flags & __WALL)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1146 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1147 | |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1148 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1149 | } |
| 1150 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1151 | static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p, |
| 1152 | pid_t pid, uid_t uid, int why, int status) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1153 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1154 | struct siginfo __user *infop; |
| 1155 | int retval = wo->wo_rusage |
| 1156 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Ingo Molnar | 36c8b58 | 2006-07-03 00:25:41 -0700 | [diff] [blame] | 1157 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1158 | put_task_struct(p); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1159 | infop = wo->wo_info; |
Vitaly Mayatskikh | b6fe2d1 | 2009-09-23 15:56:52 -0700 | [diff] [blame] | 1160 | if (infop) { |
| 1161 | if (!retval) |
| 1162 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1163 | if (!retval) |
| 1164 | retval = put_user(0, &infop->si_errno); |
| 1165 | if (!retval) |
| 1166 | retval = put_user((short)why, &infop->si_code); |
| 1167 | if (!retval) |
| 1168 | retval = put_user(pid, &infop->si_pid); |
| 1169 | if (!retval) |
| 1170 | retval = put_user(uid, &infop->si_uid); |
| 1171 | if (!retval) |
| 1172 | retval = put_user(status, &infop->si_status); |
| 1173 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1174 | if (!retval) |
| 1175 | retval = pid; |
| 1176 | return retval; |
| 1177 | } |
| 1178 | |
| 1179 | /* |
| 1180 | * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold |
| 1181 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
| 1182 | * the lock and this task is uninteresting. If we return nonzero, we have |
| 1183 | * released the lock and the system call should return. |
| 1184 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1185 | 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] | 1186 | { |
| 1187 | unsigned long state; |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1188 | int retval, status, traced; |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1189 | pid_t pid = task_pid_vnr(p); |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 1190 | uid_t uid = __task_cred(p)->uid; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1191 | struct siginfo __user *infop; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1192 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1193 | if (!likely(wo->wo_flags & WEXITED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1194 | return 0; |
| 1195 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1196 | if (unlikely(wo->wo_flags & WNOWAIT)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1197 | int exit_code = p->exit_code; |
Thiago Farina | f3abd4f | 2010-03-05 13:42:52 -0800 | [diff] [blame] | 1198 | int why; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1199 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1200 | get_task_struct(p); |
| 1201 | read_unlock(&tasklist_lock); |
| 1202 | if ((exit_code & 0x7f) == 0) { |
| 1203 | why = CLD_EXITED; |
| 1204 | status = exit_code >> 8; |
| 1205 | } else { |
| 1206 | why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED; |
| 1207 | status = exit_code & 0x7f; |
| 1208 | } |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1209 | return wait_noreap_copyout(wo, p, pid, uid, why, status); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1210 | } |
| 1211 | |
| 1212 | /* |
| 1213 | * Try to move the task's state to DEAD |
| 1214 | * only one thread is allowed to do this: |
| 1215 | */ |
| 1216 | state = xchg(&p->exit_state, EXIT_DEAD); |
| 1217 | if (state != EXIT_ZOMBIE) { |
| 1218 | BUG_ON(state != EXIT_DEAD); |
| 1219 | return 0; |
| 1220 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1221 | |
Oleg Nesterov | 53b6f9f | 2008-04-30 00:53:13 -0700 | [diff] [blame] | 1222 | traced = ptrace_reparented(p); |
Oleg Nesterov | befca96 | 2009-06-18 16:49:11 -0700 | [diff] [blame] | 1223 | /* |
| 1224 | * It can be ptraced but not reparented, check |
| 1225 | * !task_detached() to filter out sub-threads. |
| 1226 | */ |
| 1227 | if (likely(!traced) && likely(!task_detached(p))) { |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1228 | struct signal_struct *psig; |
| 1229 | struct signal_struct *sig; |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 1230 | unsigned long maxrss; |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1231 | cputime_t tgutime, tgstime; |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1232 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1233 | /* |
| 1234 | * The resource counters for the group leader are in its |
| 1235 | * own task_struct. Those for dead threads in the group |
| 1236 | * are in its signal_struct, as are those for the child |
| 1237 | * processes it has previously reaped. All these |
| 1238 | * accumulate in the parent's signal_struct c* fields. |
| 1239 | * |
| 1240 | * We don't bother to take a lock here to protect these |
| 1241 | * p->signal fields, because they are only touched by |
| 1242 | * __exit_signal, which runs with tasklist_lock |
| 1243 | * write-locked anyway, and so is excluded here. We do |
Oleg Nesterov | d1e98f4 | 2009-06-17 16:27:34 -0700 | [diff] [blame] | 1244 | * need to protect the access to parent->signal fields, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1245 | * as other threads in the parent group can be right |
| 1246 | * here reaping other children at the same time. |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1247 | * |
| 1248 | * We use thread_group_times() to get times for the thread |
| 1249 | * group, which consolidates times for all threads in the |
| 1250 | * group including the group leader. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1251 | */ |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1252 | thread_group_times(p, &tgutime, &tgstime); |
Oleg Nesterov | d1e98f4 | 2009-06-17 16:27:34 -0700 | [diff] [blame] | 1253 | spin_lock_irq(&p->real_parent->sighand->siglock); |
| 1254 | psig = p->real_parent->signal; |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1255 | sig = p->signal; |
| 1256 | psig->cutime = |
| 1257 | cputime_add(psig->cutime, |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1258 | cputime_add(tgutime, |
| 1259 | sig->cutime)); |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1260 | psig->cstime = |
| 1261 | cputime_add(psig->cstime, |
Hidetoshi Seto | 0cf55e1 | 2009-12-02 17:28:07 +0900 | [diff] [blame] | 1262 | cputime_add(tgstime, |
| 1263 | sig->cstime)); |
Laurent Vivier | 9ac5231 | 2007-10-15 17:00:19 +0200 | [diff] [blame] | 1264 | psig->cgtime = |
| 1265 | cputime_add(psig->cgtime, |
| 1266 | cputime_add(p->gtime, |
| 1267 | cputime_add(sig->gtime, |
| 1268 | sig->cgtime))); |
Jesper Juhl | 3795e16 | 2006-01-09 20:54:39 -0800 | [diff] [blame] | 1269 | psig->cmin_flt += |
| 1270 | p->min_flt + sig->min_flt + sig->cmin_flt; |
| 1271 | psig->cmaj_flt += |
| 1272 | p->maj_flt + sig->maj_flt + sig->cmaj_flt; |
| 1273 | psig->cnvcsw += |
| 1274 | p->nvcsw + sig->nvcsw + sig->cnvcsw; |
| 1275 | psig->cnivcsw += |
| 1276 | p->nivcsw + sig->nivcsw + sig->cnivcsw; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 1277 | psig->cinblock += |
| 1278 | task_io_get_inblock(p) + |
| 1279 | sig->inblock + sig->cinblock; |
| 1280 | psig->coublock += |
| 1281 | task_io_get_oublock(p) + |
| 1282 | sig->oublock + sig->coublock; |
Jiri Pirko | 1f10206 | 2009-09-22 16:44:10 -0700 | [diff] [blame] | 1283 | maxrss = max(sig->maxrss, sig->cmaxrss); |
| 1284 | if (psig->cmaxrss < maxrss) |
| 1285 | psig->cmaxrss = maxrss; |
Andrea Righi | 5995477 | 2008-07-27 17:29:15 +0200 | [diff] [blame] | 1286 | task_io_accounting_add(&psig->ioac, &p->ioac); |
| 1287 | task_io_accounting_add(&psig->ioac, &sig->ioac); |
Oleg Nesterov | d1e98f4 | 2009-06-17 16:27:34 -0700 | [diff] [blame] | 1288 | spin_unlock_irq(&p->real_parent->sighand->siglock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1289 | } |
| 1290 | |
| 1291 | /* |
| 1292 | * Now we are sure this task is interesting, and no other |
| 1293 | * thread can reap it because we set its state to EXIT_DEAD. |
| 1294 | */ |
| 1295 | read_unlock(&tasklist_lock); |
| 1296 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1297 | retval = wo->wo_rusage |
| 1298 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1299 | status = (p->signal->flags & SIGNAL_GROUP_EXIT) |
| 1300 | ? p->signal->group_exit_code : p->exit_code; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1301 | if (!retval && wo->wo_stat) |
| 1302 | retval = put_user(status, wo->wo_stat); |
| 1303 | |
| 1304 | infop = wo->wo_info; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1305 | if (!retval && infop) |
| 1306 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1307 | if (!retval && infop) |
| 1308 | retval = put_user(0, &infop->si_errno); |
| 1309 | if (!retval && infop) { |
| 1310 | int why; |
| 1311 | |
| 1312 | if ((status & 0x7f) == 0) { |
| 1313 | why = CLD_EXITED; |
| 1314 | status >>= 8; |
| 1315 | } else { |
| 1316 | why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; |
| 1317 | status &= 0x7f; |
| 1318 | } |
| 1319 | retval = put_user((short)why, &infop->si_code); |
| 1320 | if (!retval) |
| 1321 | retval = put_user(status, &infop->si_status); |
| 1322 | } |
| 1323 | if (!retval && infop) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1324 | retval = put_user(pid, &infop->si_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1325 | if (!retval && infop) |
David Howells | c69e8d9 | 2008-11-14 10:39:19 +1100 | [diff] [blame] | 1326 | retval = put_user(uid, &infop->si_uid); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1327 | if (!retval) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1328 | retval = pid; |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1329 | |
| 1330 | if (traced) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1331 | write_lock_irq(&tasklist_lock); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1332 | /* We dropped tasklist, ptracer could die and untrace */ |
| 1333 | ptrace_unlink(p); |
| 1334 | /* |
| 1335 | * If this is not a detached task, notify the parent. |
| 1336 | * If it's still not detached after that, don't release |
| 1337 | * it now. |
| 1338 | */ |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 1339 | if (!task_detached(p)) { |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1340 | do_notify_parent(p, p->exit_signal); |
Oleg Nesterov | d839fd4 | 2008-04-30 00:53:11 -0700 | [diff] [blame] | 1341 | if (!task_detached(p)) { |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1342 | p->exit_state = EXIT_ZOMBIE; |
| 1343 | p = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1344 | } |
| 1345 | } |
| 1346 | write_unlock_irq(&tasklist_lock); |
| 1347 | } |
| 1348 | if (p != NULL) |
| 1349 | release_task(p); |
Oleg Nesterov | 2f4e6e2 | 2007-10-16 23:26:58 -0700 | [diff] [blame] | 1350 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1351 | return retval; |
| 1352 | } |
| 1353 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1354 | static int *task_stopped_code(struct task_struct *p, bool ptrace) |
| 1355 | { |
| 1356 | if (ptrace) { |
| 1357 | if (task_is_stopped_or_traced(p)) |
| 1358 | return &p->exit_code; |
| 1359 | } else { |
| 1360 | if (p->signal->flags & SIGNAL_STOP_STOPPED) |
| 1361 | return &p->signal->group_exit_code; |
| 1362 | } |
| 1363 | return NULL; |
| 1364 | } |
| 1365 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1366 | /* |
| 1367 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold |
| 1368 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
| 1369 | * the lock and this task is uninteresting. If we return nonzero, we have |
| 1370 | * released the lock and the system call should return. |
| 1371 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1372 | static int wait_task_stopped(struct wait_opts *wo, |
| 1373 | int ptrace, struct task_struct *p) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1374 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1375 | struct siginfo __user *infop; |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1376 | int retval, exit_code, *p_code, why; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1377 | uid_t uid = 0; /* unneeded, required by compiler */ |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1378 | pid_t pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1379 | |
Oleg Nesterov | 4791802 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1380 | /* |
| 1381 | * Traditionally we see ptrace'd stopped tasks regardless of options. |
| 1382 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1383 | if (!ptrace && !(wo->wo_flags & WUNTRACED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1384 | return 0; |
| 1385 | |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1386 | exit_code = 0; |
| 1387 | spin_lock_irq(&p->sighand->siglock); |
| 1388 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1389 | p_code = task_stopped_code(p, ptrace); |
| 1390 | if (unlikely(!p_code)) |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1391 | goto unlock_sig; |
| 1392 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1393 | exit_code = *p_code; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1394 | if (!exit_code) |
| 1395 | goto unlock_sig; |
| 1396 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1397 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1398 | *p_code = 0; |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1399 | |
Daniel J Blueman | f362b73 | 2010-08-17 23:56:55 +0100 | [diff] [blame] | 1400 | uid = task_uid(p); |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1401 | unlock_sig: |
| 1402 | spin_unlock_irq(&p->sighand->siglock); |
| 1403 | if (!exit_code) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1404 | return 0; |
| 1405 | |
| 1406 | /* |
| 1407 | * Now we are pretty sure this task is interesting. |
| 1408 | * Make sure it doesn't get reaped out from under us while we |
| 1409 | * give up the lock and then examine it below. We don't want to |
| 1410 | * keep holding onto the tasklist_lock while we call getrusage and |
| 1411 | * possibly take page faults for user memory. |
| 1412 | */ |
| 1413 | get_task_struct(p); |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1414 | pid = task_pid_vnr(p); |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1415 | why = ptrace ? CLD_TRAPPED : CLD_STOPPED; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1416 | read_unlock(&tasklist_lock); |
| 1417 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1418 | if (unlikely(wo->wo_flags & WNOWAIT)) |
| 1419 | return wait_noreap_copyout(wo, p, pid, uid, why, exit_code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1420 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1421 | retval = wo->wo_rusage |
| 1422 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
| 1423 | if (!retval && wo->wo_stat) |
| 1424 | retval = put_user((exit_code << 8) | 0x7f, wo->wo_stat); |
| 1425 | |
| 1426 | infop = wo->wo_info; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1427 | if (!retval && infop) |
| 1428 | retval = put_user(SIGCHLD, &infop->si_signo); |
| 1429 | if (!retval && infop) |
| 1430 | retval = put_user(0, &infop->si_errno); |
| 1431 | if (!retval && infop) |
Roland McGrath | 6efcae4 | 2008-03-08 11:41:22 -0800 | [diff] [blame] | 1432 | retval = put_user((short)why, &infop->si_code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1433 | if (!retval && infop) |
| 1434 | retval = put_user(exit_code, &infop->si_status); |
| 1435 | if (!retval && infop) |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1436 | retval = put_user(pid, &infop->si_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1437 | if (!retval && infop) |
Oleg Nesterov | ee7c82d | 2008-02-08 04:19:01 -0800 | [diff] [blame] | 1438 | retval = put_user(uid, &infop->si_uid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1439 | if (!retval) |
Oleg Nesterov | c895078 | 2007-11-28 16:21:24 -0800 | [diff] [blame] | 1440 | retval = pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1441 | put_task_struct(p); |
| 1442 | |
| 1443 | BUG_ON(!retval); |
| 1444 | return retval; |
| 1445 | } |
| 1446 | |
| 1447 | /* |
| 1448 | * Handle do_wait work for one task in a live, non-stopped state. |
| 1449 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
| 1450 | * the lock and this task is uninteresting. If we return nonzero, we have |
| 1451 | * released the lock and the system call should return. |
| 1452 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1453 | 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] | 1454 | { |
| 1455 | int retval; |
| 1456 | pid_t pid; |
| 1457 | uid_t uid; |
| 1458 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1459 | if (!unlikely(wo->wo_flags & WCONTINUED)) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1460 | return 0; |
| 1461 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1462 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) |
| 1463 | return 0; |
| 1464 | |
| 1465 | spin_lock_irq(&p->sighand->siglock); |
| 1466 | /* Re-check with the lock held. */ |
| 1467 | if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) { |
| 1468 | spin_unlock_irq(&p->sighand->siglock); |
| 1469 | return 0; |
| 1470 | } |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1471 | if (!unlikely(wo->wo_flags & WNOWAIT)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1472 | p->signal->flags &= ~SIGNAL_STOP_CONTINUED; |
Daniel J Blueman | f362b73 | 2010-08-17 23:56:55 +0100 | [diff] [blame] | 1473 | uid = task_uid(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1474 | spin_unlock_irq(&p->sighand->siglock); |
| 1475 | |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 1476 | pid = task_pid_vnr(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1477 | get_task_struct(p); |
| 1478 | read_unlock(&tasklist_lock); |
| 1479 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1480 | if (!wo->wo_info) { |
| 1481 | retval = wo->wo_rusage |
| 1482 | ? getrusage(p, RUSAGE_BOTH, wo->wo_rusage) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1483 | put_task_struct(p); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1484 | if (!retval && wo->wo_stat) |
| 1485 | retval = put_user(0xffff, wo->wo_stat); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1486 | if (!retval) |
Oleg Nesterov | 3a515e4 | 2008-02-08 04:19:07 -0800 | [diff] [blame] | 1487 | retval = pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1488 | } else { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1489 | retval = wait_noreap_copyout(wo, p, pid, uid, |
| 1490 | CLD_CONTINUED, SIGCONT); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1491 | BUG_ON(retval == 0); |
| 1492 | } |
| 1493 | |
| 1494 | return retval; |
| 1495 | } |
| 1496 | |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1497 | /* |
| 1498 | * Consider @p for a wait by @parent. |
| 1499 | * |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1500 | * -ECHILD should be in ->notask_error before the first call. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1501 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
| 1502 | * Returns zero if the search for a child should continue; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1503 | * then ->notask_error is 0 if @p is an eligible child, |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1504 | * or another error from security_task_wait(), or still -ECHILD. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1505 | */ |
Oleg Nesterov | b6e763f | 2009-09-23 15:56:50 -0700 | [diff] [blame] | 1506 | static int wait_consider_task(struct wait_opts *wo, int ptrace, |
| 1507 | struct task_struct *p) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1508 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1509 | int ret = eligible_child(wo, p); |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1510 | if (!ret) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1511 | return ret; |
| 1512 | |
Oleg Nesterov | a2322e1 | 2009-09-23 15:56:45 -0700 | [diff] [blame] | 1513 | ret = security_task_wait(p); |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1514 | if (unlikely(ret < 0)) { |
| 1515 | /* |
| 1516 | * If we have not yet seen any eligible child, |
| 1517 | * then let this error code replace -ECHILD. |
| 1518 | * A permission error will give the user a clue |
| 1519 | * to look for security policy problems, rather |
| 1520 | * than for mysterious wait bugs. |
| 1521 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1522 | if (wo->notask_error) |
| 1523 | wo->notask_error = ret; |
Oleg Nesterov | 78a3d9d | 2009-04-29 18:01:23 +0200 | [diff] [blame] | 1524 | return 0; |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1525 | } |
| 1526 | |
Oleg Nesterov | 5cb1144 | 2009-06-17 16:27:30 -0700 | [diff] [blame] | 1527 | if (likely(!ptrace) && unlikely(task_ptrace(p))) { |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1528 | /* |
| 1529 | * This child is hidden by ptrace. |
| 1530 | * We aren't allowed to see it now, but eventually we will. |
| 1531 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1532 | wo->notask_error = 0; |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1533 | return 0; |
| 1534 | } |
| 1535 | |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1536 | if (p->exit_state == EXIT_DEAD) |
| 1537 | return 0; |
| 1538 | |
| 1539 | /* |
| 1540 | * We don't reap group leaders with subthreads. |
| 1541 | */ |
| 1542 | if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p)) |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1543 | return wait_task_zombie(wo, p); |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1544 | |
| 1545 | /* |
| 1546 | * It's stopped or running now, so it might |
| 1547 | * later continue, exit, or stop again. |
| 1548 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1549 | wo->notask_error = 0; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1550 | |
Oleg Nesterov | 90bc8d8 | 2009-04-02 16:57:58 -0700 | [diff] [blame] | 1551 | if (task_stopped_code(p, ptrace)) |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1552 | return wait_task_stopped(wo, ptrace, p); |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1553 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1554 | return wait_task_continued(wo, p); |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1555 | } |
| 1556 | |
| 1557 | /* |
| 1558 | * Do the work of do_wait() for one thread in the group, @tsk. |
| 1559 | * |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1560 | * -ECHILD should be in ->notask_error before the first call. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1561 | * Returns nonzero for a final return, when we have unlocked tasklist_lock. |
| 1562 | * Returns zero if the search for a child should continue; then |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1563 | * ->notask_error is 0 if there were any eligible children, |
Roland McGrath | 14dd0b8 | 2008-03-30 18:41:25 -0700 | [diff] [blame] | 1564 | * or another error from security_task_wait(), or still -ECHILD. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1565 | */ |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1566 | 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] | 1567 | { |
| 1568 | struct task_struct *p; |
| 1569 | |
| 1570 | list_for_each_entry(p, &tsk->children, sibling) { |
Oleg Nesterov | 9cd80bb | 2009-12-17 15:27:15 -0800 | [diff] [blame] | 1571 | int ret = wait_consider_task(wo, 0, p); |
| 1572 | if (ret) |
| 1573 | return ret; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1574 | } |
| 1575 | |
| 1576 | return 0; |
| 1577 | } |
| 1578 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1579 | 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] | 1580 | { |
| 1581 | struct task_struct *p; |
| 1582 | |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1583 | list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { |
Oleg Nesterov | b6e763f | 2009-09-23 15:56:50 -0700 | [diff] [blame] | 1584 | int ret = wait_consider_task(wo, 1, p); |
Roland McGrath | f470021 | 2008-03-24 18:36:23 -0700 | [diff] [blame] | 1585 | if (ret) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1586 | return ret; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1587 | } |
| 1588 | |
| 1589 | return 0; |
| 1590 | } |
| 1591 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1592 | static int child_wait_callback(wait_queue_t *wait, unsigned mode, |
| 1593 | int sync, void *key) |
| 1594 | { |
| 1595 | struct wait_opts *wo = container_of(wait, struct wait_opts, |
| 1596 | child_wait); |
| 1597 | struct task_struct *p = key; |
| 1598 | |
Oleg Nesterov | 5c01ba4 | 2009-09-23 15:56:48 -0700 | [diff] [blame] | 1599 | if (!eligible_pid(wo, p)) |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1600 | return 0; |
| 1601 | |
Oleg Nesterov | b4fe518 | 2009-09-23 15:56:47 -0700 | [diff] [blame] | 1602 | if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent) |
| 1603 | return 0; |
| 1604 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1605 | return default_wake_function(wait, mode, sync, key); |
| 1606 | } |
| 1607 | |
Oleg Nesterov | a7f0765 | 2009-09-23 15:56:44 -0700 | [diff] [blame] | 1608 | void __wake_up_parent(struct task_struct *p, struct task_struct *parent) |
| 1609 | { |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1610 | __wake_up_sync_key(&parent->signal->wait_chldexit, |
| 1611 | TASK_INTERRUPTIBLE, 1, p); |
Oleg Nesterov | a7f0765 | 2009-09-23 15:56:44 -0700 | [diff] [blame] | 1612 | } |
| 1613 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1614 | static long do_wait(struct wait_opts *wo) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1615 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1616 | struct task_struct *tsk; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1617 | int retval; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1618 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1619 | trace_sched_process_wait(wo->wo_pid); |
Mathieu Desnoyers | 0a16b60 | 2008-07-18 12:16:17 -0400 | [diff] [blame] | 1620 | |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1621 | init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); |
| 1622 | wo->child_wait.private = current; |
| 1623 | add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1624 | repeat: |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1625 | /* |
| 1626 | * If there is nothing that can match our critiera just get out. |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1627 | * We will clear ->notask_error to zero if we see any child that |
| 1628 | * might later match our criteria, even if we are not able to reap |
| 1629 | * it yet. |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1630 | */ |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1631 | wo->notask_error = -ECHILD; |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1632 | if ((wo->wo_type < PIDTYPE_MAX) && |
| 1633 | (!wo->wo_pid || hlist_empty(&wo->wo_pid->tasks[wo->wo_type]))) |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1634 | goto notask; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1635 | |
Oleg Nesterov | f95d39d | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 1636 | set_current_state(TASK_INTERRUPTIBLE); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1637 | read_lock(&tasklist_lock); |
| 1638 | tsk = current; |
| 1639 | do { |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1640 | retval = do_wait_thread(wo, tsk); |
| 1641 | if (retval) |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1642 | goto end; |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1643 | |
| 1644 | retval = ptrace_do_wait(wo, tsk); |
| 1645 | if (retval) |
| 1646 | goto end; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1647 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1648 | if (wo->wo_flags & __WNOTHREAD) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1649 | break; |
Oleg Nesterov | a3f6dfb | 2009-06-17 16:27:41 -0700 | [diff] [blame] | 1650 | } while_each_thread(current, tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1651 | read_unlock(&tasklist_lock); |
Oleg Nesterov | f2cc3eb | 2008-02-08 04:19:06 -0800 | [diff] [blame] | 1652 | |
Oleg Nesterov | 64a16ca | 2009-06-17 16:27:40 -0700 | [diff] [blame] | 1653 | notask: |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1654 | retval = wo->notask_error; |
| 1655 | if (!retval && !(wo->wo_flags & WNOHANG)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1656 | retval = -ERESTARTSYS; |
Roland McGrath | 98abed0 | 2008-03-19 19:24:59 -0700 | [diff] [blame] | 1657 | if (!signal_pending(current)) { |
| 1658 | schedule(); |
| 1659 | goto repeat; |
| 1660 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1661 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1662 | end: |
Oleg Nesterov | f95d39d | 2009-06-17 16:27:42 -0700 | [diff] [blame] | 1663 | __set_current_state(TASK_RUNNING); |
Oleg Nesterov | 0b7570e | 2009-09-23 15:56:46 -0700 | [diff] [blame] | 1664 | remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1665 | return retval; |
| 1666 | } |
| 1667 | |
Heiko Carstens | 17da2bd | 2009-01-14 14:14:10 +0100 | [diff] [blame] | 1668 | SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *, |
| 1669 | infop, int, options, struct rusage __user *, ru) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1670 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1671 | struct wait_opts wo; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1672 | struct pid *pid = NULL; |
| 1673 | enum pid_type type; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1674 | long ret; |
| 1675 | |
| 1676 | if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED)) |
| 1677 | return -EINVAL; |
| 1678 | if (!(options & (WEXITED|WSTOPPED|WCONTINUED))) |
| 1679 | return -EINVAL; |
| 1680 | |
| 1681 | switch (which) { |
| 1682 | case P_ALL: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1683 | type = PIDTYPE_MAX; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1684 | break; |
| 1685 | case P_PID: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1686 | type = PIDTYPE_PID; |
| 1687 | if (upid <= 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1688 | return -EINVAL; |
| 1689 | break; |
| 1690 | case P_PGID: |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1691 | type = PIDTYPE_PGID; |
| 1692 | if (upid <= 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1693 | return -EINVAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1694 | break; |
| 1695 | default: |
| 1696 | return -EINVAL; |
| 1697 | } |
| 1698 | |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1699 | if (type < PIDTYPE_MAX) |
| 1700 | pid = find_get_pid(upid); |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1701 | |
| 1702 | wo.wo_type = type; |
| 1703 | wo.wo_pid = pid; |
| 1704 | wo.wo_flags = options; |
| 1705 | wo.wo_info = infop; |
| 1706 | wo.wo_stat = NULL; |
| 1707 | wo.wo_rusage = ru; |
| 1708 | ret = do_wait(&wo); |
Vitaly Mayatskikh | dfe16df | 2009-09-23 15:56:51 -0700 | [diff] [blame] | 1709 | |
| 1710 | if (ret > 0) { |
| 1711 | ret = 0; |
| 1712 | } else if (infop) { |
| 1713 | /* |
| 1714 | * For a WNOHANG return, clear out all the fields |
| 1715 | * we would set so the user can easily tell the |
| 1716 | * difference. |
| 1717 | */ |
| 1718 | if (!ret) |
| 1719 | ret = put_user(0, &infop->si_signo); |
| 1720 | if (!ret) |
| 1721 | ret = put_user(0, &infop->si_errno); |
| 1722 | if (!ret) |
| 1723 | ret = put_user(0, &infop->si_code); |
| 1724 | if (!ret) |
| 1725 | ret = put_user(0, &infop->si_pid); |
| 1726 | if (!ret) |
| 1727 | ret = put_user(0, &infop->si_uid); |
| 1728 | if (!ret) |
| 1729 | ret = put_user(0, &infop->si_status); |
| 1730 | } |
| 1731 | |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1732 | put_pid(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1733 | |
| 1734 | /* avoid REGPARM breakage on x86: */ |
Roland McGrath | 54a0151 | 2008-04-10 15:37:38 -0700 | [diff] [blame] | 1735 | asmlinkage_protect(5, ret, which, upid, infop, options, ru); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1736 | return ret; |
| 1737 | } |
| 1738 | |
Heiko Carstens | 754fe8d | 2009-01-14 14:14:09 +0100 | [diff] [blame] | 1739 | SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, |
| 1740 | int, options, struct rusage __user *, ru) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1741 | { |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1742 | struct wait_opts wo; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1743 | struct pid *pid = NULL; |
| 1744 | enum pid_type type; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1745 | long ret; |
| 1746 | |
| 1747 | if (options & ~(WNOHANG|WUNTRACED|WCONTINUED| |
| 1748 | __WNOTHREAD|__WCLONE|__WALL)) |
| 1749 | return -EINVAL; |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1750 | |
| 1751 | if (upid == -1) |
| 1752 | type = PIDTYPE_MAX; |
| 1753 | else if (upid < 0) { |
| 1754 | type = PIDTYPE_PGID; |
| 1755 | pid = find_get_pid(-upid); |
| 1756 | } else if (upid == 0) { |
| 1757 | type = PIDTYPE_PGID; |
Oleg Nesterov | 2ae448e | 2009-04-02 16:58:36 -0700 | [diff] [blame] | 1758 | pid = get_task_pid(current, PIDTYPE_PGID); |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1759 | } else /* upid > 0 */ { |
| 1760 | type = PIDTYPE_PID; |
| 1761 | pid = find_get_pid(upid); |
| 1762 | } |
| 1763 | |
Oleg Nesterov | 9e8ae01 | 2009-06-17 16:27:39 -0700 | [diff] [blame] | 1764 | wo.wo_type = type; |
| 1765 | wo.wo_pid = pid; |
| 1766 | wo.wo_flags = options | WEXITED; |
| 1767 | wo.wo_info = NULL; |
| 1768 | wo.wo_stat = stat_addr; |
| 1769 | wo.wo_rusage = ru; |
| 1770 | ret = do_wait(&wo); |
Eric W. Biederman | 161550d | 2008-02-08 04:19:14 -0800 | [diff] [blame] | 1771 | put_pid(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1772 | |
| 1773 | /* avoid REGPARM breakage on x86: */ |
Roland McGrath | 54a0151 | 2008-04-10 15:37:38 -0700 | [diff] [blame] | 1774 | asmlinkage_protect(4, ret, upid, stat_addr, options, ru); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1775 | return ret; |
| 1776 | } |
| 1777 | |
| 1778 | #ifdef __ARCH_WANT_SYS_WAITPID |
| 1779 | |
| 1780 | /* |
| 1781 | * sys_waitpid() remains for compatibility. waitpid() should be |
| 1782 | * implemented by calling sys_wait4() from libc.a. |
| 1783 | */ |
Heiko Carstens | 17da2bd | 2009-01-14 14:14:10 +0100 | [diff] [blame] | 1784 | SYSCALL_DEFINE3(waitpid, pid_t, pid, int __user *, stat_addr, int, options) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1785 | { |
| 1786 | return sys_wait4(pid, stat_addr, options, NULL); |
| 1787 | } |
| 1788 | |
| 1789 | #endif |