From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750769AbWCVLcg (ORCPT ); Wed, 22 Mar 2006 06:32:36 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750759AbWCVLcg (ORCPT ); Wed, 22 Mar 2006 06:32:36 -0500 Received: from ns.virtualhost.dk ([195.184.98.160]:54047 "EHLO virtualhost.dk") by vger.kernel.org with ESMTP id S1750769AbWCVLcf (ORCPT ); Wed, 22 Mar 2006 06:32:35 -0500 Date: Wed, 22 Mar 2006 12:32:35 +0100 From: Jens Axboe To: Andrew Morton , linux-kernel@vger.kernel.org, dm-devel@redhat.com Subject: Re: dm: bio split bvec fix Message-ID: <20060322113235.GC4285@suse.de> References: <20060320192155.GU4724@agk.surrey.redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20060320192155.GU4724@agk.surrey.redhat.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Mar 20 2006, Alasdair G Kergon wrote: > The code that handles bios that span table target boundaries by breaking > them up into smaller bios will not split an individual struct bio_vec > into more than two pieces. Sometimes more than that are required. > > This patch adds a loop to break the second piece up into as many > pieces as are necessary. Why isn't this just handled in the merge callback? Can a single page bio span > 2 targets? -- Jens Axboe