From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752957AbaAGRXx (ORCPT ); Tue, 7 Jan 2014 12:23:53 -0500 Received: from mail-pb0-f49.google.com ([209.85.160.49]:46335 "EHLO mail-pb0-f49.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752085AbaAGRXr (ORCPT ); Tue, 7 Jan 2014 12:23:47 -0500 Date: Tue, 7 Jan 2014 09:23:42 -0800 From: Dmitry Torokhov To: Ming Lei Cc: Greg Kroah-Hartman , Linux Kernel Mailing List Subject: Re: [PATCH v1 2/4] firmware loader: simplify holding module for request_firmware Message-ID: <20140107172342.GA31842@core.coreip.homeip.net> References: <1370520110-32293-1-git-send-email-ming.lei@canonical.com> <1370520110-32293-3-git-send-email-ming.lei@canonical.com> <20140107075600.GC20074@core.coreip.homeip.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Jan 07, 2014 at 10:53:57PM +0800, Ming Lei wrote: > Hi Dmitry, > > On Tue, Jan 7, 2014 at 3:56 PM, Dmitry Torokhov > wrote: > > Hi Ming, > > > > On Thu, Jun 06, 2013 at 08:01:48PM +0800, Ming Lei wrote: > >> module reference doesn't cover direct loading path, so this patch > >> simply holds the module in the whole life time of request_firmware() > >> to fix the problem. > > > > This does not make sense to me. If request_firmware() is executing that > > means some other module references it and module refcount already > > reflects that. > > Yes, you are right, holding this module references in both > request_firmware() and request_firmware_direct() shouldn't > be necessary. > > > > > We needed to pin module before Tejun's work ensuring that currently open > > sysfs entries won't keep related kobjects pinned after kernel marked > > them inactive. We can probably delete __module_get()/module_put() from > > firmware_class.c now. > > But holding the reference for request_firmware_nowait() is needed. Ah, yes, of course. Thanks. -- Dmitry