From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?UTF-8?Q?Roger_Pau_Monn=c3=a9?= Subject: Re: [RFC] support more qdisk types Date: Wed, 3 Feb 2016 12:15:35 +0100 Message-ID: <56B1E157.4030907@citrix.com> References: <56A6BCDE.6040900@suse.com> <20160202145930.GB25660@citrix.com> <56B12849.2010705@suse.com> <20160203103553.GE23178@citrix.com> <1454496687.25207.57.camel@citrix.com> <20160203105525.GH23178@citrix.com> <1454497504.25207.63.camel@citrix.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1454497504.25207.63.camel@citrix.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Ian Campbell , Wei Liu Cc: George Dunlap , Jim Fehlig , xen-devel , Ian Jackson , Ken Johnson List-Id: xen-devel@lists.xenproject.org El 3/2/16 a les 12:05, Ian Campbell ha escrit: > On Wed, 2016-02-03 at 10:55 +0000, Wei Liu wrote: >> On Wed, Feb 03, 2016 at 10:51:27AM +0000, Ian Campbell wrote: >>> On Wed, 2016-02-03 at 10:35 +0000, Wei Liu wrote: >>>>> Ok. So in your opinion, even if any new disk config is encoded in >>>>> 'target=', >>>>> libxlu should split that up into (new) members of >>>>> libxl_device_disk, not just >>>>> plop it into libxl_device_disk.pdev_path? >>>>> >>>> >>>> No, not necessarily. I didn't look closely in the code yesterday when >>>> replying, sorry. >>>> >>>> If target= has always been shoveled into pdev_path, using that would >>>> be >>>> fine. We already have mechanism to parse target= outside of libxl in >>>> hotplug script. >>>> >>>> Are you aware of all those hotplug scripts living under tools/hotplug >>>> ? >>>> Does using hotplug script sound plausible to you? >>>> >>>> Currently hotplug script for QEMU is broken and needs fixing though, >>>> but >>>> I'm sure we can figure it out. >>> >>> How do hotplug scripts factor into this? >>> >> >> If supporting all such block devices requires presenting a block device >> to QEMU? If QEMU directly handles them then hotplug script is not in the >> picture. > > Perhaps I've misunderstood what this thread is about. I thought it was > about exposing all the various backends which qdisk supports natively, like > CEPH, sheepdog, iscsi, nbd etc. AFAIK QEMU/Qdisk doesn't need a block device or regular file in order to work, it can use as said above things like iSCSI (which we current support using a script + blkback [0]), nbd, ceph... There's been some discussion in the past about whether hotplug scripts should also be supported by QEMU/Qdisk, or even whether the launch of the Qdisk backend itself should be done inside of a hotplug script (abstracting it away from libxl). I think George had some ideas/work on this (or did look into similar issues in the past). IMHO, I don't think we need hotplug scripts support for Qemu/Qdisk handled backends. Roger. [0] http://xenbits.xen.org/gitweb/?p=xen.git;a=commit;h=bda856e1023868b1e9605e845bedbd8b7ed2944e