From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [PATCH 2/9] dm-mpath, scsi_dh: don't let dm detach device handlers Date: Fri, 1 May 2015 11:20:53 +0200 Message-ID: <20150501092053.GA9341@lst.de> References: <1430415151-30948-1-git-send-email-hch@lst.de> <1430415151-30948-3-git-send-email-hch@lst.de> <20150430182153.GB30542@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from verein.lst.de ([213.95.11.211]:54564 "EHLO newverein.lst.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752984AbbEAJUz (ORCPT ); Fri, 1 May 2015 05:20:55 -0400 Content-Disposition: inline In-Reply-To: <20150430182153.GB30542@redhat.com> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Mike Snitzer Cc: Christoph Hellwig , linux-scsi@vger.kernel.org, Hannes Reinecke , Mike Christie , "Martin K. Petersen" On Thu, Apr 30, 2015 at 02:21:54PM -0400, Mike Snitzer wrote: > > While allowing dm-mpath to attach device handlers is a functionality we need > > for backwards compatibility reason there is no reason to reference count > > them and detach them if dm-mpath stops using the device for some reason. > > > > If the device handler works for the given device it can just stay attached. > > > > Signed-off-by: Christoph Hellwig > > Looks good. But, unless I'm missing something, these DM and scsi_dh > changes can be made in separate patches so that I can carry the DM > change without needing to touch scsi_dh (SCSI maintainer boundary). Unfortunately they can't. E.g. if we remove the detach in dm-mpath but keep the refcounting in scsi_dh we'll leak references. Of if we remove the request_module in dm without adding it to scsi we stop asking for the module. So I'll need to collect ACKs from both sides and find a suitable tree to get this in.