From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753553AbbAUQUR (ORCPT ); Wed, 21 Jan 2015 11:20:17 -0500 Received: from out3-smtp.messagingengine.com ([66.111.4.27]:52714 "EHLO out3-smtp.messagingengine.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751484AbbAUQUJ (ORCPT ); Wed, 21 Jan 2015 11:20:09 -0500 Message-Id: <1421857209.171127.216796445.0DC88E13@webmail.messagingengine.com> X-Sasl-Enc: TMBWUlGTU/bsdhZRGSHEYzYN/+j2OWtwn9xsIEoihzTt 1421857209 From: Alexander van Heukelum To: Denys Vlasenko Cc: Andy Lutomirski , X86 ML , Linux Kernel Mailing List , Frederic Weisbecker , Oleg Nesterov , Borislav Petkov , Rik van Riel MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain X-Mailer: MessagingEngine.com Webmail Interface - ajax-46f3f2c7 In-Reply-To: References: <1421581520-2816-1-git-send-email-heukelum@fastmail.fm> <1421581520-2816-2-git-send-email-heukelum@fastmail.fm> Subject: Re: [PATCHv2 1/4] x86_64: cleanup THREAD_INFO(reg,offset) macro Date: Wed, 21 Jan 2015 17:20:09 +0100 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Jan 21, 2015, at 14:40, Denys Vlasenko wrote: > On Sun, Jan 18, 2015 at 12:45 PM, Alexander van Heukelum > wrote: > > The macro THREAD_INFO(reg,offset) is used in assembly to compute the > > offset between the user ptregs and the thread_info struct. Change > > the macro and all its uses so that offset is given as the current > > top of stack in the pt_regs frame. The generated code is identical. > > What is the purpose of doing this? > > I don't mean to say that it's pointless, but it is also not obvious > why it's better than the code before patch. Just because I had a hard time understanding what the offset parameter means. Turns out Borislav did another variant in his patchset. Quoting him from the commit message which made that the change: > Semi-mysterious expressions THREAD_INFO(%rsp,RIP) - "why RIP??" > are now replaced by more logical THREAD_INFO(%rsp,SIZEOF_PTREGS) - > "rsp is SIZEOF_PTREGS bytes below the top, calculate > thread_info's address using that information" Borislav expresses "offset" as an offset from the stack's page boundary,while I chose the offset from the start of struct pt_regs on the stack.