From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Snitzer Subject: Re: [PATCH 0/4] dm: zoned block device fixes Date: Thu, 8 Jun 2017 16:21:47 -0400 Message-ID: <20170608202146.GB6747@redhat.com> References: <20170529102307.19814-1-damien.lemoal@wdc.com> <20170530202031.GA52190@redhat.com> <46ec84af-0700-70f9-f990-d87a63cc18b0@wdc.com> <20170531143910.GA16671@redhat.com> <20170602003617.GB25647@redhat.com> <9749b12d-e232-5386-0894-511bb45758e7@wdc.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: <9749b12d-e232-5386-0894-511bb45758e7@wdc.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dm-devel-bounces@redhat.com Errors-To: dm-devel-bounces@redhat.com To: Damien Le Moal Cc: Bart Van Assche , dm-devel@redhat.com, Alasdair Kergon List-Id: dm-devel.ids On Mon, Jun 05 2017 at 6:48am -0400, Damien Le Moal wrote: > I did address some of the "FIXME" notes you added. The main one is the > BIO cloning in the I/O path. I removed most of that and added a .end_io > method for completion processing. The only place were I do not see how > to remove the call to bio_clone() is during read BIO processing: since a > read BIO may end up being split between buffer zone, sequential zone and > simple buffer zero-out, fragmentation of the read BIO is sometimes > necessary and so need a clone. So shouldn't it be possible to not allow a given bio to cross zone boundaries by using dm_accept_partial_bio()? Like you're already doing in dmz_map() actually... so why do you need to account for crossing zone boundaries on read later on in dmz_submit_read_bio()? Is it that these zones aren't easily known up front (in dmz_map)? Mike