From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([140.186.70.92]:45993)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <paolo.bonzini@gmail.com>) id 1RtTTj-0001aC-MU
	for qemu-devel@nongnu.org; Fri, 03 Feb 2012 19:27:45 -0500
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <paolo.bonzini@gmail.com>) id 1RtTTi-0001Y0-Ii
	for qemu-devel@nongnu.org; Fri, 03 Feb 2012 19:27:43 -0500
Received: from mail-we0-f173.google.com ([74.125.82.173]:44627)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <paolo.bonzini@gmail.com>) id 1RtTTi-0001Xr-AB
	for qemu-devel@nongnu.org; Fri, 03 Feb 2012 19:27:42 -0500
Received: by werh12 with SMTP id h12so3690004wer.4
	for <qemu-devel@nongnu.org>; Fri, 03 Feb 2012 16:27:41 -0800 (PST)
Sender: Paolo Bonzini <paolo.bonzini@gmail.com>
Message-ID: <4F2C7B7B.60801@redhat.com>
Date: Sat, 04 Feb 2012 01:27:39 +0100
From: Paolo Bonzini <pbonzini@redhat.com>
MIME-Version: 1.0
References: <1328201142-26145-1-git-send-email-pbonzini@redhat.com>
	<1328201142-26145-2-git-send-email-pbonzini@redhat.com>
	<4F2BEED5.3050904@codemonkey.ws>
In-Reply-To: <4F2BEED5.3050904@codemonkey.ws>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Subject: Re: [Qemu-devel] [PATCH 01/16] qdev: fix hot-unplug
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Anthony Liguori <anthony@codemonkey.ws>
Cc: qemu-devel@nongnu.org

On 02/03/2012 03:27 PM, Anthony Liguori wrote:
> On 02/02/2012 10:45 AM, Paolo Bonzini wrote:
>> The reference that is returned by qdev_device_add is never given
>> back, so that device_del does not cause the refcount to go to zero
>> (and thus does nothing).
>>
>> Signed-off-by: Paolo Bonzini<pbonzini@redhat.com>
>
> This isn't needed in qom-upstream.14. Here's why:
>
> object_init does not increase the reference count
>
> object_property_add_child increases the reference count
> object_new increases the reference count
>
> object_delete decrements the reference count
> object_property_del_child decreases the reference count
>
> object_delete calls object_property_del_child(obj->parent, obj)
>
> qdev_device_add calls object_new and object_property_add_child
> -> ref == 2
>
> qdev_device_del calls object_delete
> -> ref -= 2
>
> In qom-upstream.13, object_delete wasn't calling
> object_property_del_child which is why you saw the behavior you did.
> This problem would still exist with a composed device so dropping the
> reference here isn't enough.

I trust you for now. :)

It really seems like my patch is obviously correct so, if it's not 
needed anymore there may be another bug elsewhere that masks it.

Paolo