From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:60781) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1S396j-0008VD-Jn for qemu-devel@nongnu.org; Thu, 01 Mar 2012 11:44:02 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1S396J-0007o7-Nd for qemu-devel@nongnu.org; Thu, 01 Mar 2012 11:43:56 -0500 Received: from mx1.redhat.com ([209.132.183.28]:9267) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1S396J-0007nn-F3 for qemu-devel@nongnu.org; Thu, 01 Mar 2012 11:43:31 -0500 Date: Thu, 1 Mar 2012 13:43:27 -0300 From: Luiz Capitulino Message-ID: <20120301134327.497fed02@doriath.home> In-Reply-To: <4F4FA13C.6030307@redhat.com> References: <1330600908-16202-1-git-send-email-pbonzini@redhat.com> <20120301130250.46c2377e@doriath.home> <4F4FA13C.6030307@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH v2 0/6] Mirrored writes using blockdev-transaction List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Kevin Wolf Cc: aliguori@us.ibm.com, stefanha@linux.vnet.ibm.com, jcody@redhat.com, qemu-devel@nongnu.org, fsimonce@redhat.com, Paolo Bonzini , eblake@redhat.com On Thu, 01 Mar 2012 17:18:04 +0100 Kevin Wolf wrote: > Am 01.03.2012 17:02, schrieb Luiz Capitulino: > > On Thu, 1 Mar 2012 12:21:42 +0100 > > Paolo Bonzini wrote: > > > >> This implements all ingredients to establish mirrored writes. > > > > This looks good to me. I just have two comments: > > > > 1. I'm wondering if it would make more sense to have this transaction > > operation in qmp instead of the block layer. Looks more complex to do though, > > so I'm ok with this implementation > > Depends on what you're thinking of. Renaming the command to just > 'transaction' and allowing anything to be added to the union wouldn't be > very complex. > > The one thing we would need to change in order to make it generally > useful is to move the actual logic into prepare/commit/abort handlers. I > discussed this with Paolo on IRC and I think the conclusion was that for > now the approach in the patches is good enough, but in the long run > we'll switch. It doesn't affect external interfaces, so we can do it > whenever we like. Yes, that's what I called "transaction operation in qmp", but I don't mind accepting this one and deferring the more complex idea to the future.