From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753135Ab2IFGmm (ORCPT ); Thu, 6 Sep 2012 02:42:42 -0400 Received: from e28smtp01.in.ibm.com ([122.248.162.1]:49248 "EHLO e28smtp01.in.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750914Ab2IFGmk (ORCPT ); Thu, 6 Sep 2012 02:42:40 -0400 Message-ID: <50484551.8080700@linux.vnet.ibm.com> Date: Thu, 06 Sep 2012 12:10:17 +0530 From: "Naveen N. Rao" User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:15.0) Gecko/20120828 Thunderbird/15.0 MIME-Version: 1.0 To: Tony Luck CC: Andi Kleen , ak@linux.intel.com, bp@amd64.org, x86@kernel.org, linux-kernel@vger.kernel.org, linux-edac@vger.kernel.org, ananth@in.ibm.com Subject: Re: [PATCH] [mcelog] Start using the new sysfs tunables location References: <20120905102156.9423.27806.stgit@localhost.localdomain> <20120905103237.9726.56131.stgit@localhost.localdomain> <20120905184708.GC16230@one.firstfloor.org> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit x-cbid: 12090606-4790-0000-0000-00000479EF19 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 09/06/2012 12:39 AM, Tony Luck wrote: > On Wed, Sep 5, 2012 at 11:47 AM, Andi Kleen wrote: >> On Wed, Sep 05, 2012 at 04:02:37PM +0530, Naveen N. Rao wrote: >>> All the current mce tunables are now available under >>> /sys/devices/system/machinecheck. Start using this new location, but fall back >>> to the older per-cpu location so that we continue working with older kernels. >> >> Who did that change in the kernel? >> >> That breaks Linus rule that the kernel should not break userland. >> Kernel needs to fix that. > > The change is still under discussion. Stage one is to add the new global > pathnames in addition to keeping the old per-cpu ones. Also fix all utilities > (just mcelog(8) as far as we know) to prefer the new paths. > > After some time[1] ... delete the old paths. This is allowable under Linus' > modified edict that you can change ABI "if nobody complains". If we wait > long enough that the new mcelog is widely deployed, then nobody should > complain. > > -Tony > > [1] several years - not just a kernel release or two. > Tony, Thanks for clarifying. I should have mentioned in the patch description that this is indeed subject to the original patch making it into the kernel. On a related topic. I recently noticed that we don't have an entry for machinecheck in Documentation/ABI/. Should we add an entry in there? We could perhaps add the existing entries under obsolete/ and the new location under testing/? - Naveen