Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1752723Ab3FZQPN (ORCPT <rfc822;w@1wt.eu>);
	Wed, 26 Jun 2013 12:15:13 -0400
Received: from mail-ve0-f176.google.com ([209.85.128.176]:41449 "EHLO
	mail-ve0-f176.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1752492Ab3FZQPL (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Wed, 26 Jun 2013 12:15:11 -0400
MIME-Version: 1.0
In-Reply-To: <20130626155728.GA2141@redhat.com>
References: <0e00e9073855c02a382d49ba1ede9c4fda3451b7.1372189875.git.luto@amacapital.net>
 <20130625201602.GB1726@moon> <20130625201705.GA15966@redhat.com>
 <CALCETrVxm2wHvLgexG6fZO+h1rNHC_qDCv6B+4XuANHgSQU4iw@mail.gmail.com>
 <20130625215247.GC1726@moon> <CALCETrV2WZ3m82D4THfxvmASW8AOf1SN7Pu14vE0jMORupO5xw@mail.gmail.com>
 <20130626155728.GA2141@redhat.com>
From: Andy Lutomirski <luto@amacapital.net>
Date: Wed, 26 Jun 2013 09:14:50 -0700
Message-ID: <CALCETrXHLBnJ5UbtSe9Af+VoPDNtXVe0Da2yasj0Nr_=Q1YvLA@mail.gmail.com>
Subject: Re: [PATCH] proc: Expose /proc/<pid>/task/<tid>/children unconditionally
To: Oleg Nesterov <oleg@redhat.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset=ISO-8859-1
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Content-Length: 3162
Lines: 79

On Wed, Jun 26, 2013 at 8:57 AM, Oleg Nesterov <oleg@redhat.com> wrote:
> On 06/25, Andy Lutomirski wrote:
>>
>> On Tue, Jun 25, 2013 at 2:52 PM, Cyrill Gorcunov <gorcunov@gmail.com> wrote:
>> > On Tue, Jun 25, 2013 at 02:36:31PM -0700, Andy Lutomirski wrote:
>> >> On Tue, Jun 25, 2013 at 1:17 PM, Oleg Nesterov <oleg@redhat.com> wrote:
>> >> > On 06/26, Cyrill Gorcunov wrote:
>> >> >>
>> >> >> On Tue, Jun 25, 2013 at 12:51:45PM -0700, Andy Lutomirski wrote:
>> >> >> > This is currently only available if CONFIG_CHECKPOINT_RESTORE, which
>> >> >> > is hidden under CONFIG_EXPERT.  It's generally useful functionality,
>> >> >> > though, so expose it unconditionally.
>> >> >> >
>> >> >> > Cc: Cyrill Gorcunov <gorcunov@openvz.org>
>> >> >> > Signed-off-by: Andy Lutomirski <luto@amacapital.net>
>> >> >> Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
>> >> >
>> >> > I didn't see the patch but I guess it is trivial and I agree with intent ;)
>> >>
>> >> The patch works, but "children" is only listed under task/<tid>, not
>> >> under /proc/<pid>.  Is that intentional?  Fixing it would be a
>> >> one-liner.
>> >
>> > Yeah, it's intentional. Here some explanations from Oleg (check out
>> > the whole thread, it's not that big https://lkml.org/lkml/2011/12/9/220)
>> > in short this might require some more code, but i'll re-check tomorrow.
>>
>> This is a little strange.  It looks like ppid (in status) shows the
>> tgid,
>
> Yes,
>
>> but the actual real_parent can refer to a thread (as opposed to
>> a thread group leader),
>
> Yes.
>
> This is mostly the internal implementation detail. And probably it
> would be nice to move ->children from task_struct to signal_struct.
>
> However. See __WNOTHREAD in man waitpid. I think this is the only
> (historical) reason. Otherwise the real_parent's tid doesn't matter,
> the parent is always the whole process, not sub-thread.
>
>> and task/tid/children respects that.
>
> Well, it should respect if you want to restart and keep __WNOTHREAD
> working.
>
> But there is another reason. It is not trivial to list all children
> under /proc/<pid>, and the fix would not be a one-liner ;) You need
> to fight with the exiting sub-threads and reparenting.
>
>> So the
>> tree that you get by following task/tid/ children won't be quite the
>> same as the tree you get by following ppid.
>
> Not sure I understand... but in any case, yes you need to read
> /proc/pid/task/*/children to construct the tree.
>
>> I wonder if the ptid should be added to status.  Is there anything
>> (other than task/tid/children)
>
> Perhaps, I dunno.
>
> Better yet, we should kill __WNOTHREAD ;) I am wondering if it is still
> used.

Ugh.  I'll submit a follow-up patch to
Documentation/filesystems/proc.txt to explain the situation.  All I
really care about is being able to iterate the children of a
single-threaded process, and the current semantics are fine for that.

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/