From: Daniel Borkmann Subject: Re: [PATCH v3] crypto: more robust crypto_memneq Date: Tue, 26 Nov 2013 20:27:33 +0100 Message-ID: <5294F625.5040506@redhat.com> References: <1385424041-18064-1-git-send-email-cesarb@cesarb.eti.br> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: linux-crypto@vger.kernel.org, Herbert Xu , "David S. Miller" , James Yonan , Florian Weimer , linux-kernel@vger.kernel.org To: Cesar Eduardo Barros Return-path: Received: from mx1.redhat.com ([209.132.183.28]:55358 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752654Ab3KZT2Y (ORCPT ); Tue, 26 Nov 2013 14:28:24 -0500 In-Reply-To: <1385424041-18064-1-git-send-email-cesarb@cesarb.eti.br> Sender: linux-crypto-owner@vger.kernel.org List-ID: On 11/26/2013 01:00 AM, Cesar Eduardo Barros wrote: > Disabling compiler optimizations can be fragile, since a new > optimization could be added to -O0 or -Os that breaks the assumptions > the code is making. > > Instead of disabling compiler optimizations, use a dummy inline assembly > (based on RELOC_HIDE) to block the problematic kinds of optimization, > while still allowing other optimizations to be applied to the code. > > The dummy inline assembly is added after every OR, and has the > accumulator variable as its input and output. The compiler is forced to > assume that the dummy inline assembly could both depend on the > accumulator variable and change the accumulator variable, so it is > forced to compute the value correctly before the inline assembly, and > cannot assume anything about its value after the inline assembly. > > This change should be enough to make crypto_memneq work correctly (with > data-independent timing) even if it is inlined at its call sites. That > can be done later in a followup patch. > > Compile-tested on x86_64. Actually with yet another version, I hoped that the "compile-tested"-only statement would eventually disappear, ohh well. ;) > Signed-off-by: Cesar Eduardo Barros Resolving the OPTIMIZER_HIDE_VAR() macro for others than GCC jnto a barrier() seems a bit suboptimal, but assuming 99% of people will use GCC anyway, then for the minority of the remaining, they will worst case have a clever compiler and eventually mimic memcmp() in some situations, or have a not-so-clever compiler and execute the full code as is. Anyway, I think still better than the rather ugly Makefile workaround imho, so I'm generally fine with this.