From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263532AbTJQVAG (ORCPT ); Fri, 17 Oct 2003 17:00:06 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263533AbTJQVAG (ORCPT ); Fri, 17 Oct 2003 17:00:06 -0400 Received: from e1.ny.us.ibm.com ([32.97.182.101]:21212 "EHLO e1.ny.us.ibm.com") by vger.kernel.org with ESMTP id S263532AbTJQVAC (ORCPT ); Fri, 17 Oct 2003 17:00:02 -0400 Subject: Letext symbols in System.map From: Dave Hansen To: Linux Kernel Mailing List Cc: "Martin J. Bligh" Content-Type: text/plain Organization: Message-Id: <1066424399.2589.171.camel@nighthawk> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.2.4 Date: 17 Oct 2003 13:59:59 -0700 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org I've been seeing intermittent Letext symbols showing up in System.maps for a while. They're annoying because they seem to plant themselves at the same address as spinlocks and readprofile picks them over the spinlock symbol. c02c40d9 t .text.lock.af_packet c02c40d9 t Letext For me, they only show up when CONFIG_DEBUG_INFO is turned on, but Martin Bligh claims he's seen them other times as well. After some googling, it appears that Letext is supposed to be a label for a string constant. Since the spinlocks are defined using assembly inside string constants, perhaps the label for that asm string is hanging around too long. Is ther anything to keep us from simply grepping them away when System.map is created? $(NM) $@ | grep -v '\(compiled\)\|\(\.o$$\)\|\( [aUw] \)\|\(\.\.ng$$\)\|\(LASH[RL]DI\)\|\(Letext\)' | sort > System.map -- Dave Hansen haveblue@us.ibm.com