From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754018AbaCNMpF (ORCPT ); Fri, 14 Mar 2014 08:45:05 -0400 Received: from bombadil.infradead.org ([198.137.202.9]:42455 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754023AbaCNMpC (ORCPT ); Fri, 14 Mar 2014 08:45:02 -0400 Date: Fri, 14 Mar 2014 05:44:57 -0700 From: Christoph Hellwig To: Hannes Reinecke Cc: Christoph Hellwig , Mike Snitzer , Jeff Moyer , Jens Axboe , Shaohua Li , "linux-kernel@vger.kernel.org" Subject: Re: [PATCH 1/1] block: rework flush sequencing for blk-mq Message-ID: <20140314124457.GA4859@infradead.org> References: <20140130132620.GA6031@infradead.org> <20140130132630.GB6031@infradead.org> <20140308155240.GA32297@infradead.org> <531B74B6.4070004@suse.de> <20140312102849.GA26509@infradead.org> <20140313161347.GA6598@redhat.com> <20140314092519.GA10139@infradead.org> <5322CC4D.8040509@suse.de> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <5322CC4D.8040509@suse.de> User-Agent: Mutt/1.5.21 (2010-09-15) X-SRS-Rewrite: SMTP reverse-path rewritten from by bombadil.infradead.org See http://www.infradead.org/rpr.html Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Mar 14, 2014 at 10:30:53AM +0100, Hannes Reinecke wrote: > That was actually one of my plans, move dm-multipath over to use > blk-mq. But then I'd need to discuss with Jens et al how to best > achieve this; the current static hctx allocation doesn't play well > with multipaths dynamic path management. I'd say it the other way around: the clone + insert hacks in dm-multipath don't work well with blk-mq. Not allowing non-owned requests is fundamentally part of the blk-mq model to allow things like the integrated tag allocator and queue depth limiting or the preallocated driver specific data. Instead dm-multipath should alway resubmit the request like it already does for the slow path for the first step. Longer term we might be able to operate using a cheaper temporary structure, but I'm not sure that's going to be worth it.