From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263728AbTKRQ0q (ORCPT ); Tue, 18 Nov 2003 11:26:46 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263726AbTKRQ0p (ORCPT ); Tue, 18 Nov 2003 11:26:45 -0500 Received: from nat9.steeleye.com ([65.114.3.137]:16135 "EHLO hancock.sc.steeleye.com") by vger.kernel.org with ESMTP id S263728AbTKRQ0b (ORCPT ); Tue, 18 Nov 2003 11:26:31 -0500 Subject: Re: lib.a causing modules not to load From: James Bottomley To: Rusty Russell Cc: Christoph Hellwig , Andrew Morton , Linux Kernel In-Reply-To: <20031117060542.088D32C0B1@lists.samba.org> References: <20031117060542.088D32C0B1@lists.samba.org> Content-Type: text/plain Content-Transfer-Encoding: 7bit X-Mailer: Ximian Evolution 1.0.8 (1.0.8-9) Date: 18 Nov 2003 10:25:16 -0600 Message-Id: <1069172719.1835.30.camel@mulgrave> Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Sun, 2003-11-16 at 20:47, Rusty Russell wrote: > I think lib.a should be linked as is if !CONFIG_MODULES, and done as a > ..o if CONFIG_MODULES. Other alternatives are possible, but make it > tricky if someone adds a module later which wants something in lib.a. I tried this and it is getting to be a whole nasty mess can of worms: You can't link the lib objects all in, because they can be overridden by the arch dependent lib.a (we rely on link order to permit this). For instance on x86, dump_stack and bust_spinlocks give duplicate symbols. If we have to add processing logic to work out what symbols are in the arch lib.a, we might as well go the whole hog and try to work out what is missing from the kernel and compile that alone as a module. James