From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261406AbTJ2TTN (ORCPT ); Wed, 29 Oct 2003 14:19:13 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261429AbTJ2TTN (ORCPT ); Wed, 29 Oct 2003 14:19:13 -0500 Received: from mail.ccur.com ([208.248.32.212]:20999 "EHLO exchange.ccur.com") by vger.kernel.org with ESMTP id S261406AbTJ2TTJ (ORCPT ); Wed, 29 Oct 2003 14:19:09 -0500 Subject: x86_64 module loader reloc problem From: Jason Baietto To: lkml Content-Type: text/plain Message-Id: <1067455102.444.18.camel@broccoli> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.5 Date: Wed, 29 Oct 2003 14:18:22 -0500 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Using test8 I'm only able to load kernel modules on x86_64 systems if I comment out the overflow check for R_X86_64_32 relocation entries in arch/x64_64/kernel/module.c. With the check in place I get errors like this: overflow in relocation type 10 val ffffff0000426b88" errors `test' likely not compiled with -mcmodel=kernel insmod: error inserting 'test.ko': -1 Invalid module format Yes, I'm using -mcmodel=kernel. I'm also using module-init-tools 0.9.15-pre2. Once loaded, the modules appear to be working properly, but it wouldn't surprise me if a time bomb was lurking there. What am I doing wrong? Compiler used for both kernel and modules is gcc (GCC) 3.2.3 20030502 (Red Hat Linux 3.2.3-20). Note that in my linked test module, objdump shows 632 R_X86_64_32 reloc entries and only 18 R_X86_64_64 reloc entries. Take care, Jason