Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262592AbVCDHWe (ORCPT ); Fri, 4 Mar 2005 02:22:34 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262527AbVCDHWe (ORCPT ); Fri, 4 Mar 2005 02:22:34 -0500 Received: from ns.virtualhost.dk ([195.184.98.160]:16593 "EHLO virtualhost.dk") by vger.kernel.org with ESMTP id S262592AbVCDHSN (ORCPT ); Fri, 4 Mar 2005 02:18:13 -0500 Date: Fri, 4 Mar 2005 08:18:01 +0100 From: Jens Axboe To: Florin Iucha Cc: linux-kernel@vger.kernel.org, jdike@addtoit.com Subject: Re: are the io-schedulers per-device? Message-ID: <20050304071756.GA14764@suse.de> References: <20050304064053.GC10507@iucha.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050304064053.GC10507@iucha.net> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1304 Lines: 29 On Fri, Mar 04 2005, Florin Iucha wrote: > Hello, > > For a semester project I am experimenting with a new IO scheduler and I > was trying to set my scheduler to control a single device, to ease the > development and debugging, by using > echo "foo" > /sys/block/ubdc/queue/scheduler > Much to my suprise, this sets the scheduler for the other block > devices as well! Does this happen only to UML block devices? Do I need > to do anything to allow a per-device scheduler? Is the functionality > there, or is it in-progress? Am I reading too much in the fact that > the queue/scheduler is defined under each block device? It's per-queue. In general that is per-device, apparently that is not so for UML since it shares a queue for several devices. From a cleanliness and performance POV, it's is far better to have a queue per device instead of sharing, I would suggest fixing the uml block device. It looks pretty straight forward to do so, except for ubd_handler(). Which, btw, calls elv_next_request() without holding the queue lock! -- Jens Axboe - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/