From mboxrd@z Thu Jan 1 00:00:00 1970 From: Al Viro Subject: Re: [PATCH 1/2] md: make devices disappear when they are no longer needed. Date: Mon, 24 Nov 2008 06:38:06 +0000 Message-ID: <20081124063806.GA28946@ZenIV.linux.org.uk> References: <20081124035516.3465.66413.stgit@notabene.brown> <20081124035530.3465.26724.stgit@notabene.brown> <20081124042421.GY28946@ZenIV.linux.org.uk> <18730.12791.348848.849033@notabene.brown> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <18730.12791.348848.849033@notabene.brown> Sender: linux-raid-owner@vger.kernel.org To: Neil Brown Cc: linux-kernel@vger.kernel.org, linux-raid@vger.kernel.org, Tejun Heo , Doug Ledford List-Id: linux-raid.ids On Mon, Nov 24, 2008 at 03:47:51PM +1100, Neil Brown wrote: > > What lifetime rules do you really want? I never liked the tricks pulled > > by md wrt gendisk lifetimes and that might be a good time to sort that > > out for good... > > I'm not sure what 'tricks' you are referring to. Can you elaborate? > > I want the gendisk to appear as soon as it is needed (not because I > think that is necessarily a good idea, but it is legacy functionality that I > don't think we can easily discard). And I want them to disappear when > they contain no information and have nothing referring to them. "Tricks" are about md_probe() and weird allocation time for these suckers. But OK, legacy API is a good argument. So you want the rules of the same nature as for module refcount? Then the natural place to do that would be in failure exit of __blkdev_get() and in normal path in blkdev_put() (for the final opener going away). However, let's try to do it right - there's a *lot* of drivers where we do no work in ->release() until it's the final one. It would be nice to accomodate them as well while we are at it...