From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S965837AbcAZN40 (ORCPT ); Tue, 26 Jan 2016 08:56:26 -0500 Received: from verein.lst.de ([213.95.11.211]:36713 "EHLO newverein.lst.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S965102AbcAZN4Y (ORCPT ); Tue, 26 Jan 2016 08:56:24 -0500 Date: Tue, 26 Jan 2016 14:56:21 +0100 From: Torsten Duwe To: Petr Mladek Cc: Miroslav Benes , Steven Rostedt , Michael Ellerman , Jiri Kosina , linuxppc-dev@lists.ozlabs.org, Linux Kernel Mailing List , live-patching@vger.kernel.org Subject: Re: [PATCH v6 8/9] Implement kernel live patching for ppc64le (ABIv2) Message-ID: <20160126135621.GA28664@lst.de> References: <20160125170459.14DB7692CE@newverein.lst.de> <20160125170804.A11DB692DD@newverein.lst.de> <20160126124853.GL3305@pathway.suse.cz> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20160126124853.GL3305@pathway.suse.cz> User-Agent: Mutt/1.5.17 (2007-11-01) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Jan 26, 2016 at 01:48:53PM +0100, Petr Mladek wrote: > On Tue 2016-01-26 11:50:25, Miroslav Benes wrote: > > > > We still need Petr's patch from [1] to make livepatch work, right? Could > > you, please, add it to this patch set to make it self-sufficient? It's Petr's patch, I don't want to decide how to best tackle this, see below. I think Michael is already aware that it is needed, too. > > Second, what is the situation with mcount prologue between gcc < 6 and > > gcc-6? Are there only 12 bytes in gcc-6 prologue? If yes, we need to Precisely, it's commit e95d0248daced44 (in http://repo.or.cz/official-gcc.git) or svn trunk change 222352 "No need for -mprofile-kernel to save LR to stack." It's efficient, I like it. > I am going to update the extra patch. There is an idea to detect the > offset during build by scrips/recordmcount. This tool looks for the > ftrace locations. The offset should always be a constant that depends > on the used architecture, compiler, and compiler flags. My first idea was to check for compiler version defines, but some vendors are rumoured to patch their compilers ;-) > The tool is called post build. We might need to pass the constant > as a symbol added to the binary. The tool already adds some symbols. That's even better. Thanks! Torsten