[PATCH 3/3] reparent kernel threads to swapper

[PATCH 3/3] reparent kernel threads to swapper

[Oleg Nesterov]

A lot of kernel threads parented to /sbin/init slow down do_wait() when
a non-detached (user-space) process exits. Change reparent_kthread() to
use init_task as a parent. Since init_task can't go away, we don't need
to put the caller on init_task.children list.

NOTE: pstree doesn't show kernel threads with this patch.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>

--- 2.6.21-rc5/kernel/exit.c~3_SWAPPER	2007-04-10 21:59:41.000000000 +0400
+++ 2.6.21-rc5/kernel/exit.c	2007-04-10 22:25:25.000000000 +0400
@@ -255,11 +255,11 @@ static int has_stopped_jobs(struct pid *
 }
 
 /**
- * reparent_kthread - Reparent the calling kernel thread to the init task of the pid space that the thread belongs to.
+ * reparent_kthread - Reparent the calling kernel thread to swapper.
  *
  * If a kernel thread is launched as a result of a system call, or if
- * it ever exits, it should generally reparent itself to init so that
- * it is correctly cleaned up on exit.
+ * it ever exits, it should generally reparent itself so that it is
+ * correctly cleaned up on exit.
  *
  * The various task state such as scheduling policy and priority may have
  * been inherited from a user process, so we reset them to sane values here.
@@ -272,9 +272,8 @@ void reparent_kthread(void)
 
 	ptrace_unlink(current);
 	remove_parent(current);
-	current->parent = init_pid_ns.child_reaper;
+	current->parent = &init_task;
 	current->real_parent = current->parent;
-	add_parent(current);
 
 	/* make the task auto-reap */
 	current->exit_signal = -1;

Re: [PATCH 3/3] reparent kernel threads to swapper

[Eric W. Biederman]

Oleg Nesterov <oleg@tv-sign.ru> writes:

> A lot of kernel threads parented to /sbin/init slow down do_wait() when
> a non-detached (user-space) process exits. Change reparent_kthread() to
> use init_task as a parent. Since init_task can't go away, we don't need
> to put the caller on init_task.children list.

Is there any downside of putting tasks on init_task.children?

If not it is probably a good idea to put processes there just for good measure.

> NOTE: pstree doesn't show kernel threads with this patch.
>
> Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
>
> --- 2.6.21-rc5/kernel/exit.c~3_SWAPPER	2007-04-10 21:59:41.000000000 +0400
> +++ 2.6.21-rc5/kernel/exit.c	2007-04-10 22:25:25.000000000 +0400
> @@ -255,11 +255,11 @@ static int has_stopped_jobs(struct pid *
>  }
>  
>  /**
> - * reparent_kthread - Reparent the calling kernel thread to the init task of
> the pid space that the thread belongs to.
> + * reparent_kthread - Reparent the calling kernel thread to swapper.
>   *
>   * If a kernel thread is launched as a result of a system call, or if
> - * it ever exits, it should generally reparent itself to init so that
> - * it is correctly cleaned up on exit.
> + * it ever exits, it should generally reparent itself so that it is
> + * correctly cleaned up on exit.
>   *
>   * The various task state such as scheduling policy and priority may have
>   * been inherited from a user process, so we reset them to sane values here.
> @@ -272,9 +272,8 @@ void reparent_kthread(void)
>  
>  	ptrace_unlink(current);
>  	remove_parent(current);
> -	current->parent = init_pid_ns.child_reaper;
> +	current->parent = &init_task;
>  	current->real_parent = current->parent;
> -	add_parent(current);
>  
>  	/* make the task auto-reap */
>  	current->exit_signal = -1;