From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner+w=401wt.eu-S1030224AbXDKTmx@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1030224AbXDKTmx (ORCPT <rfc822;w@1wt.eu>);
	Wed, 11 Apr 2007 15:42:53 -0400
Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1030241AbXDKTmx
	(ORCPT <rfc822;linux-kernel-outgoing>);
	Wed, 11 Apr 2007 15:42:53 -0400
Received: from ebiederm.dsl.xmission.com ([166.70.28.69]:32945 "EHLO
	ebiederm.dsl.xmission.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1030224AbXDKTmw (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Wed, 11 Apr 2007 15:42:52 -0400
From: ebiederm@xmission.com (Eric W. Biederman)
To: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Andrew Morton <akpm@linux-foundation.org>,
       Davide Libenzi <davidel@xmailserver.org>,
       Jan Engelhardt <jengelh@linux01.gwdg.de>, Ingo Molnar <mingo@elte.hu>,
       Linus Torvalds <torvalds@linux-foundation.org>,
       Robin Holt <holt@sgi.com>, Roland McGrath <roland@redhat.com>,
       "Serge E. Hallyn" <serge@hallyn.com>, linux-kernel@vger.kernel.org,
       "Rafael J. Wysocki" <rjw@sisk.pl>, Gautham R Shenoy <ego@in.ibm.com>
Subject: Re: [PATCH] kthread: Don't depend on work queues (take 2)
References: <20070410185133.GA104@tv-sign.ru>
	<m13b37v517.fsf@ebiederm.dsl.xmission.com>
	<20070411000316.52f2551e.akpm@linux-foundation.org>
	<m1hcrmsr4g.fsf_-_@ebiederm.dsl.xmission.com>
	<20070411192809.GA106@tv-sign.ru>
Date: Wed, 11 Apr 2007 13:40:23 -0600
In-Reply-To: <20070411192809.GA106@tv-sign.ru> (Oleg Nesterov's message of
	"Wed, 11 Apr 2007 23:28:09 +0400")
Message-ID: <m14pnmsnrs.fsf@ebiederm.dsl.xmission.com>
User-Agent: Gnus/5.110006 (No Gnus v0.6) Emacs/21.4 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Sender: linux-kernel-owner@vger.kernel.org
X-Mailing-List: linux-kernel@vger.kernel.org

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

> On 04/11, Eric W. Biederman wrote:
>> 
>> @@ -435,8 +436,12 @@ static void __init setup_command_line(char *command_line)
>>  static void noinline rest_init(void)
>>  	__releases(kernel_lock)
>>  {
>> +	int pid;
>>  	kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
>>  	numa_default_policy();
>> +
>> +	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
>> +	kthreadd_task = find_task_by_pid(pid);
>>  	unlock_kernel();
>
> Just curious. What if kernel/kthread.c declares
>
> 	static struct task_struct *kthreadd_task = &init_task;
>
> an then kthreadd_setup() does kthreadd_task = current. I assume it is always
> safe to try_to_wake_up(idle_thread), because it always TASK_RUNNING. This
> way we don't need to export kthreadd_task.

I did it this way largely so I could use the export in reparent_to_XXX in
daemonize.  This way I don't have races in finding kthreadd.  Plus
I didn't think of the trick of using the idle_thread...

>> +	spin_lock(&kthread_create_lock);
>> +	list_add_tail(&create.list, &kthread_create_list);
>> +	wake_up_process(kthreadd_task);
>> +	spin_unlock(&kthread_create_lock);
>
> Very minor nit, but we don't need to do wake_up under spin_unlock().

I guess that is true.  However it doesn't hurt either.  I guess
I was keeping the form that I used with wait queues where it may
have mattered.  Either that or I just copied a bad example.

Eric