From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262138AbUENTK7 (ORCPT ); Fri, 14 May 2004 15:10:59 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262170AbUENTK7 (ORCPT ); Fri, 14 May 2004 15:10:59 -0400 Received: from wsip-68-14-253-125.ph.ph.cox.net ([68.14.253.125]:28835 "EHLO office.labsysgrp.com") by vger.kernel.org with ESMTP id S262138AbUENTK5 (ORCPT ); Fri, 14 May 2004 15:10:57 -0400 Message-ID: <40A519BD.3030000@backtobasicsmgmt.com> Date: Fri, 14 May 2004 12:10:53 -0700 From: "Kevin P. Fleming" Organization: Back To Basics Network Management User-Agent: Mozilla Thunderbird 0.6 (Windows/20040502) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Greg KH CC: Christoph Hellwig , Norberto Bensa , linux-kernel@vger.kernel.org Subject: Re: does udev support sw raid0? References: <200405141442.38404.norberto+linux-kernel@bensa.ath.cx> <20040514183450.GA2345@kroah.com> <20040514193913.A27388@infradead.org> <20040514184659.GA2401@kroah.com> In-Reply-To: <20040514184659.GA2401@kroah.com> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Greg KH wrote: > How did this work with devfs then? The device node would not be present > before the ioctl needed to be called, right? Or did it do the "magic > lookup" mess and just "know" about md devices? Right, see my recent thread on linux-hotplug-devel about this very subject. As it stands today, the only gendisk that md registers without being asked to do so is md0, so /dev/md0 will always be present on a udev-enabled system with md. If you have auto-start MD arrays on your system, then those will be registered as well when md autostarts them at kernel boot time. All other arrays (those that you want to start using mdadm, or those that you want to create because they've never existing before) suffer from the problem of requiring a device node before they can be controlled. I chatted with Neil Brown about this, and he would not be opposed to the md driver exposing a character device specifically for the purpose of handling the md ioctls (like dm does), but when I looked more into it it seemed to be a waste of effort; the time could be better spent porting the md personalities over to dm, where all this stuff already works properly and there is naming/splitting/partitioning/etc. available.