From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753391Ab2AWQ2q (ORCPT ); Mon, 23 Jan 2012 11:28:46 -0500 Received: from mx1.redhat.com ([209.132.183.28]:20830 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752534Ab2AWQ2o (ORCPT ); Mon, 23 Jan 2012 11:28:44 -0500 Date: Mon, 23 Jan 2012 11:28:40 -0500 From: Vivek Goyal To: Tejun Heo Cc: axboe@kernel.dk, ctalbott@google.com, rni@google.com, linux-kernel@vger.kernel.org Subject: Re: [PATCH 08/17] blkcg: shoot down blkio_groups on elevator switch Message-ID: <20120123162840.GJ25986@redhat.com> References: <1327202725-3383-1-git-send-email-tj@kernel.org> <1327202725-3383-9-git-send-email-tj@kernel.org> <20120123152055.GD25986@redhat.com> <20120123153913.GB12652@google.com> <20120123155216.GF25986@redhat.com> <20120123155745.GD12652@google.com> <20120123161042.GG25986@redhat.com> <20120123161308.GH25986@redhat.com> <20120123162035.GF12652@google.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20120123162035.GF12652@google.com> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Jan 23, 2012 at 08:20:35AM -0800, Tejun Heo wrote: > On Mon, Jan 23, 2012 at 11:13:08AM -0500, Vivek Goyal wrote: > > How about draining throttle groups only on queue exit (blk_cleanup_queue()) > > and not on elevator switch. > > It's for possible policy changes and to fully manage blkg from blkcg. > blkg itself needs to change as applied policies change and need to be > flushed. Can we avoid integrating everything into single blkg. What's wrong with separate blkg for separage policy. In this case we just don't have the flexbility to change throttling policy. If it is compiled in, it gets activated. The only configurable thing is IO scheduler and these groups will be cleaned up. So keeping blkg separate for separate policy gives us this flexibility that we don't have to cleanup throttling data and keep the throttling rules persistent across elevator switch. Thanks Vivek