Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752464AbcDRVWo (ORCPT ); Mon, 18 Apr 2016 17:22:44 -0400 Received: from mx1.redhat.com ([209.132.183.28]:49410 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751737AbcDRVWm (ORCPT ); Mon, 18 Apr 2016 17:22:42 -0400 Subject: Re: [PATCH v2 0/4] Introduce bulk mode for crypto engine framework To: Mike Snitzer , Baolin Wang References: <20160415134849.GA32694@gondor.apana.org.au> <20160418133612.GA16853@redhat.com> Cc: Herbert Xu , David Miller , Alasdair G Kergon , Jens Axboe , dm-devel@redhat.com, Andrew Morton , david.s.gordon@intel.com, Tom Lendacky , Robert Jarzmik , Masahiro Yamada , smueller@chronox.de, tadeusz.struk@intel.com, Masanari Iida , shli@kernel.org, Mark Brown , Linus Walleij , Arnd Bergmann , LKML , linux-crypto@vger.kernel.org, linux-raid@vger.kernel.org From: Milan Broz Message-ID: <57155015.2090104@redhat.com> Date: Mon, 18 Apr 2016 23:22:29 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.3.0 MIME-Version: 1.0 In-Reply-To: <20160418133612.GA16853@redhat.com> Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.39]); Mon, 18 Apr 2016 21:22:36 +0000 (UTC) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2096 Lines: 49 On 04/18/2016 03:36 PM, Mike Snitzer wrote: > On Mon, Apr 18 2016 at 1:31am -0400, > Baolin Wang wrote: > >> Hi Herbert, >> >> On 15 April 2016 at 21:48, Herbert Xu wrote: >>> On Tue, Mar 15, 2016 at 03:47:58PM +0800, Baolin Wang wrote: >>>> Now some cipher hardware engines prefer to handle bulk block by merging requests >>>> to increase the block size and thus increase the hardware engine processing speed. >>>> >>>> This patchset introduces request bulk mode to help the crypto hardware drivers >>>> improve in efficiency. >>> >>> Could you please explain why this merging can't be done in dm-crypt >>> instead? >> >> We've tried to do this in dm-crypt, but it failed. >> The dm-crypt maintainer explained to me that I should optimize the >> driver, not add strange hw-dependent crypto modes to dm-crypt, this is >> not the first crypto accelerator that is just not suited for this kind >> of use. > > As a DM mainatiner my only contribution to this line of discussion was > relative to your proposal to train the dm-crypt target (which is > bio-based) to also provide request-based support, see: > https://www.redhat.com/archives/dm-devel/2015-November/msg00112.html > > But follow-up discussion occured, primarily with Milan Broz, which led > to this bulk mode support in the crypto layer. Pretty strange Milan > wasn't cc'd on your patchset posting (I've now cc'd him). My complaint was mainly that the proposed dmcrypt based version just did not work properly. https://lkml.org/lkml/2016/1/2/109 (I did not test the new version we are replying here. I wonder how the problem I mentioned is fixed though. Also see Mikulas' comments https://www.redhat.com/archives/dm-devel/2016-January/msg00145.html) Anyway, all this seems to optimize case for specific crypto hw, that is not able to perform optimally with pattern dmcrypt produces. I do not think dmcrypt should do optimizations for specific hw. But if we decide that it is needed there, it should not cause any performance or compatibility problems elsewhere... Milan