From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:58380) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1V4BGG-0001Dq-Oa for qemu-devel@nongnu.org; Tue, 30 Jul 2013 10:50:58 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1V4BGA-00025H-Ea for qemu-devel@nongnu.org; Tue, 30 Jul 2013 10:50:52 -0400 Received: from mx1.redhat.com ([209.132.183.28]:53050) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1V4BGA-000254-4b for qemu-devel@nongnu.org; Tue, 30 Jul 2013 10:50:46 -0400 Date: Tue, 30 Jul 2013 16:50:42 +0200 From: Stefan Hajnoczi Message-ID: <20130730145042.GD9018@stefanha-thinkpad.redhat.com> References: <1375170777-31457-1-git-send-email-famz@redhat.com> <1375170777-31457-4-git-send-email-famz@redhat.com> <51F7AE88.5040204@suse.de> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline In-Reply-To: <51F7AE88.5040204@suse.de> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH v2 3/7] block: implement reference count for BlockDriverState List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Andreas =?iso-8859-1?Q?F=E4rber?= Cc: kwolf@redhat.com, Fam Zheng , jcody@redhat.com, qemu-devel@nongnu.org, Anthony Liguori , Paolo Bonzini On Tue, Jul 30, 2013 at 02:16:08PM +0200, Andreas F=E4rber wrote: > Am 30.07.2013 09:52, schrieb Fam Zheng: > > Introduce bdrv_ref/bdrv_unref to manage the lifecycle of > > BlockDriverState. They are unused for now but will used to replace > > bdrv_delete() later. > >=20 > > Signed-off-by: Fam Zheng > > --- > > block.c | 22 ++++++++++++++++++++++ > > include/block/block.h | 2 ++ > > include/block/block_int.h | 1 + > > 3 files changed, 25 insertions(+) >=20 > Didn't Kevin and Markus look into turning BlockDriverState into QOM > objects a while back? That would give you reference counting for free, > even atomic unlike this patch. Splitting BlockDriverState into QOM objects has proven hard because there is more than one class in today's BlockDriverState struct. I think handling refcounts separate is actually a good idea. It tackles the lifecycle work without getting caught up in all the other considerations (which QOM objects to have, -drive -> -blockdev transition, etc). So I guess the answer is that the QOM work is not ready. This series may actually help clean things up for the QOM work, which can then switch to the objects' own refcount. Stefan