From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jim Fehlig Subject: Re: [PATCH] xl: Special case tap/aio for disk validation Date: Fri, 28 Jan 2011 09:52:07 -0700 Message-ID: <4D42F437.7070107@novell.com> References: <4D407A06.1050902@gmail.com> <4D41EE90.5000907@gmail.com> <4D42E8B9.9060901@novell.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Stefano Stabellini Cc: Kamala Narasimhan , "xen-devel@lists.xensource.com" , Ian Jackson , "Kamala Narasimhan (3P)" List-Id: xen-devel@lists.xenproject.org Stefano Stabellini wrote: > On Fri, 28 Jan 2011, Jim Fehlig wrote: > >> Kamala Narasimhan wrote: >> >>>> - phy:/path/to/device >>>> - file:/path/to/file >>>> - tap:/path/to/file >>>> >>>> >>> Along with phy, file and tap for block device types there appear to be a 'drdb' option too. Do we actively support it? >>> >> The xend toolstack supports arbitrary external block device types. We >> have block-iscsi and block-npiv, and the drbd project provides >> block-drbd. I'm quite sure I've seen other custom block types as well. >> Will these arbitrary block device types still be supported in the new >> toolstack? >> > > Is it just a matter of forking and executing an external script to > setup a block device and threat it as phy: afterwards? > Currently, IIUC, the external script is invoked by udev SUBSYSTEM=="xen-backend", KERNEL=="tap*", RUN+="/etc/xen/scripts/blktap $env{ACTION}" SUBSYSTEM=="xen-backend", KERNEL=="vbd*", RUN+="/etc/xen/scripts/block $env{ACTION}" The block script will delegate to a helper if needed, which configures the block device and writes some info (like resulting dev node) to xenstore. xend and/or qemu then read the info and continue configuring the device from backend perspective. > It shouldn't be too difficult but still non-trivial at this point of the > release cycle. > Maybe users could work around the lack of support in 4.1 calling the > script themselves and using the name of the block device in the VM > config file directly? > Yes, this is certainly a workaround. One could argue this is the proper solution and remove support for external block scripts. But this functionality has been around for years. Jim