From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S936372AbdKPXJv (ORCPT ); Thu, 16 Nov 2017 18:09:51 -0500 Received: from mail-yw0-f195.google.com ([209.85.161.195]:34347 "EHLO mail-yw0-f195.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757745AbdKPXJo (ORCPT ); Thu, 16 Nov 2017 18:09:44 -0500 X-Google-Smtp-Source: AGs4zMa9/zUANCZC3AjQq8LzxjFuXDvvvVs3BL+p24k8vMpSTSwzlAoMAmd1cGF0Y9kf+ht4Fvahug== Date: Thu, 16 Nov 2017 15:09:39 -0800 From: Sami Tolvanen To: Ard Biesheuvel Cc: Will Deacon , Mark Rutland , Andi Kleen , Kees Cook , linux-kbuild@vger.kernel.org, Nick Desaulniers , "linux-kernel@vger.kernel.org" , Greg Hackmann , Masahiro Yamada , Michal Marek , Yury Norov , Alex Matveev , Matthias Kaehlcke , "linux-arm-kernel@lists.infradead.org" , Maxim Kuvyrkov Subject: Re: [PATCH v2 08/18] arm64: don't disable ADR_PREL_PG_HI21* with ARM64_ERRATUM_843419 Message-ID: <20171116230939.GA38360@samitolvanen.mtv.corp.google.com> References: <20171115213428.22559-1-samitolvanen@google.com> <20171115213428.22559-9-samitolvanen@google.com> <20171116114406.GB9361@arm.com> <20171116163132.GC94341@samitolvanen.mtv.corp.google.com> <20171116213702.GA10616@samitolvanen.mtv.corp.google.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: 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 Thu, Nov 16, 2017 at 10:14:17PM +0000, Ard Biesheuvel wrote: > OK, so my concern here is that this code probably only operates on > fully linked binaries, and not partially linked object files like > kernel modules. Right. That makes sense. > What is preventing us from using the large model with clang? We pass -mcmodel=large to clang, but I just confirmed that the attribute isn't stored in LLVM IR, which means it's not used during link time compilation. I'll see if we can solve this problem by passing the correct code model directly to LLVMgold instead. Thanks for pointing this out. Sami