From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758256AbZFRC6f (ORCPT ); Wed, 17 Jun 2009 22:58:35 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752753AbZFRC6Y (ORCPT ); Wed, 17 Jun 2009 22:58:24 -0400 Received: from mail-bw0-f213.google.com ([209.85.218.213]:56787 "EHLO mail-bw0-f213.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752586AbZFRC6X (ORCPT ); Wed, 17 Jun 2009 22:58:23 -0400 Message-ID: <4A39AD4C.6020807@nodomain.org> Date: Thu, 18 Jun 2009 03:58:20 +0100 From: Tony Hoyle User-Agent: Thunderbird 2.0.0.21 (Macintosh/20090302) MIME-Version: 1.0 To: linux-kernel@vger.kernel.org Subject: Bug(?) in linux kernel EFI support Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, I'm building kernels on an embedded system which uses an odd version of EFI (elilo/egrub don't run for example, only its own shipped boot.efi that loads a raw vmlinux). I've been unable to get it to boot with any kernel > 2.6.24.7 - as in it doesn't start up far enough to get netconsole. I've tracked it down to the following commit. Prior to this commit, it boots.. after it, it doesn't: e429795c68d3001ecae74f6465420c9f043b0ece is first bad commit commit e429795c68d3001ecae74f6465420c9f043b0ece Author: Huang, Ying Date: Wed Jan 30 13:31:19 2008 +0100 x86: EFI runtime service support: remove duplicated code from efi_32.c This patch removes the duplicated code between efi_32.c and efi.c. Signed-off-by: Huang Ying Cc: Andi Kleen Signed-off-by: Andrew Morton Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner :040000 040000 e6ac42e8fddef99698e5ebb39ed0a3d0c18bf1ac 1974f788a81cc6ced68ce87345e107e614691831 M arch :040000 040000 f7b172e407b45e571eac4612dc799e44e7b24928 20608b877d79449318d72a31dc202f2341f0fceb M include Now I don't really understand the inner workings of EFI (or of the kernel for that matter).. I imagine 32bit EFI-only booting is fairly unusual and few platforms support it... is there anything that anyone can see that could cause a crash in this patch? I have no ability to see the actual fault - The device has no console (the display is hardcoded to show a logo until X boots.. there's an fb driver shipped with it that appears to be able to override that but it's binary only*. There's no serial port on the board, only USB & Ethernet - so I'm stuck with netconsole basically. kexec doesn't work (memory regions have different names & kexec has hardcoded text searches.. patching it with the new names doesn't work either..). I realise this makes things somewhat harder to debug :p I call it challenging.. Tony * despite a 'strings' saying GPL in the text.. sigh..