From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753623Ab0JTRct (ORCPT ); Wed, 20 Oct 2010 13:32:49 -0400 Received: from mx1.redhat.com ([209.132.183.28]:35845 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750811Ab0JTRcs (ORCPT ); Wed, 20 Oct 2010 13:32:48 -0400 Date: Wed, 20 Oct 2010 18:32:31 +0100 From: Alasdair G Kergon To: Milan Broz Cc: Andi Kleen , Mike Snitzer , linux-kernel@vger.kernel.org, device-mapper development , Andi Kleen , pedrib@gmail.com Subject: Re: [dm-devel] [PATCH] DM-CRYPT: Scale to multiple CPUs v3 Message-ID: <20101020173231.GP3935@agk-dp.fab.redhat.com> Mail-Followup-To: Milan Broz , Andi Kleen , Mike Snitzer , linux-kernel@vger.kernel.org, device-mapper development , Andi Kleen , pedrib@gmail.com References: <20101010115941.GA8539@basil.fritz.box> <4CB1B3B9.4030205@redhat.com> <20101010130842.GE8256@basil.fritz.box> <4CB1DD1A.5080906@redhat.com> <20101010162257.GA1272@redhat.com> <4CB1ECCA.8040801@redhat.com> <20101010170750.GA1475@redhat.com> <4CB20C72.3030504@redhat.com> <4CB7595C.4000801@redhat.com> <4CBEFABC.3030400@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <4CBEFABC.3030400@redhat.com> Organization: Red Hat UK Ltd. Registered in England and Wales, number 03798903. Registered Office: Amberley Place, 107-111 Peascod Street, Windsor, Berkshire, SL4 1TE. User-Agent: Mutt/1.5.18 (2008-05-17) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Oct 20, 2010 at 04:20:44PM +0200, Milan Broz wrote: > this patch introduces apparent regressions in async crypto mode. OK, thanks for performing those tests. This patch is stalled until these problems are understood properly and solutions are proposed. I was prepared to overlook the regression in stacked 'async' provided there was an inline FIXME explaining clearly how to fix it in future should we need to, but I can't support a patch that apparently breaks stacked sync encryption. One solution might be to make the proposed changes only for non-stacked configurations and leave stacks to work as they do today - either detecting that automatically or having userspace supply a target parameter to say whether or not to turn it on. (For that, would need tests to show a stack of many layers where only one is a crypt target continues to work after the change.) Alasdair