From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [PATCH] fix wrong __init usage in IDE PCI drivers Date: Wed, 19 Jan 2005 12:42:19 +0100 Message-ID: <20050119114219.GA10965@lst.de> References: <20050109131959.GA20622@lst.de> <58cb370e05011417376956cfcb@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from verein.lst.de ([213.95.11.210]:52180 "EHLO mail.lst.de") by vger.kernel.org with ESMTP id S261695AbVASLm1 (ORCPT ); Wed, 19 Jan 2005 06:42:27 -0500 Content-Disposition: inline In-Reply-To: <58cb370e05011417376956cfcb@mail.gmail.com> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Bartlomiej Zolnierkiewicz Cc: B.Zolnierkiewicz@elka.pw.edu.pl, linux-ide@vger.kernel.org On Sat, Jan 15, 2005 at 02:37:36AM +0100, Bartlomiej Zolnierkiewicz wrote: > On Sun, 9 Jan 2005 14:19:59 +0100, Christoph Hellwig wrote: > > PCI probe routines (and funcions called by them) must use __devinit > > as they can be called from PCI hotplug operations. > > > > While most of the drivers touched in the patch are for hardware that's > > soldered on mainboards we now have a fake PCI HP driver that allows > > hot remove / replug for them which they need to handle. > > Ugh... I really hate (lack of) flexibility of CONFIG_HOTPLUG... > > Shouldn't we have __fakedev{init,exit} ? This really gets a little too complex. We also tended to trade a tiny bit of vasted memory over complexity all the time in Linux's past..