From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263037AbVFWI27 (ORCPT ); Thu, 23 Jun 2005 04:28:59 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263232AbVFWIWe (ORCPT ); Thu, 23 Jun 2005 04:22:34 -0400 Received: from mail.kroah.org ([69.55.234.183]:37022 "EHLO perch.kroah.org") by vger.kernel.org with ESMTP id S262611AbVFWHBv (ORCPT ); Thu, 23 Jun 2005 03:01:51 -0400 Date: Wed, 22 Jun 2005 23:51:32 -0700 From: Greg KH To: Andrew Morton Cc: Mike Bell , miles@gnu.org, gregkh@suse.de, torvalds@osdl.org, linux-kernel@vger.kernel.org Subject: Re: [GIT PATCH] Remove devfs from 2.6.12-git Message-ID: <20050623065132.GD11638@kroah.com> References: <20050621062926.GB15062@kroah.com> <20050620235403.45bf9613.akpm@osdl.org> <20050621151019.GA19666@kroah.com> <20050623010031.GB17453@mikebell.org> <20050623045959.GB10386@kroah.com> <20050623063457.GB955@mikebell.org> <20050622233759.7a1130a9.akpm@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050622233759.7a1130a9.akpm@osdl.org> User-Agent: Mutt/1.5.8i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Jun 22, 2005 at 11:37:59PM -0700, Andrew Morton wrote: > Mike Bell wrote: > > > > On Thu, Jun 23, 2005 at 03:14:08PM +0900, Miles Bader wrote: > > > BTW, has anyone done a comparison of the space usage of udev vs. devfs > > > (including size of code etc....)? > > > > Greg gave me an "I assume so" estimate that udev was smaller by excluding > > the size of sysfs a while back. If you include sysfs in udev's overhead > > then I believe devfs wins handily, but I haven't done the numbers to > > prove it so my estimate is no better. I'm just basing it on sysfs being > > absolutely huge, in linux-tiny terms. > > sysfs certainly has a history of goggling gobs of memory. But you can > disable it in .config. Now the majority of sysfs memory is cachable and can get reclaimed quite easily. The people running the 20,000 disk farms on a 31 bit architecture made sure of that. thanks, greg k-h