From mboxrd@z Thu Jan 1 00:00:00 1970 From: Kay Sievers Subject: Re: [PATCH] Driver-core: Fix bluetooth network device rename regression Date: Thu, 22 Jul 2010 19:44:37 +0200 Message-ID: References: <20100621222938.GA20583@suse.de> <20100621231058.GA1066@suse.de> <20100622035631.GA3755@suse.de> <20100708211930.GA15385@kroah.com> <20100708230601.GB17002@kroah.com> <20100719133451.0862ca62.akpm@linux-foundation.org> <20100720201334.GA11991@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Greg KH , Andrew Morton , Greg KH , "Rafael J. Wysocki" , "Maciej W. Rozycki" , Johannes Berg , netdev To: "Eric W. Biederman" Return-path: Received: from mail-wy0-f174.google.com ([74.125.82.174]:54034 "EHLO mail-wy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753707Ab0GVRoz convert rfc822-to-8bit (ORCPT ); Thu, 22 Jul 2010 13:44:55 -0400 Received: by wyf19 with SMTP id 19so824146wyf.19 for ; Thu, 22 Jul 2010 10:44:54 -0700 (PDT) In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Thu, Jul 22, 2010 at 19:18, Eric W. Biederman wrote: > Kay Sievers writes: > Let me get this straight. =C2=A0You believe a non-trival non-obvious > complete rewrite of entire subsystems after rc1 is better than > a one line bug fix that let's real hardware work for real > people? Yes. The drivers are broken in its core use pattern, and enabling a (non-common) new feature exposes this. Users should just not enable the new feature until the drivers, they might depend on, are fixed properly. That's the rule for all other kernel development too, and I don't see what's different here. It's not a regression, and not a bug in the core, which needs fixing. I don't see what would justify subsystem-specific hacks in the core, which we will never be able to remove. Your patch exposes sysfs device layout for a single subsystem in an unusual manner, and we can not change that later. We need to avoid such special cases for all reasons. > You have had 3 years to do better a better job with the bluetooth > driver. I don't know of any real-world problem here. > Furthermore we fixed the rest of this regression in the core I don't > see why we can fix the entirety of it this namespace regression in th= e > core. =C2=A0Especially since it is an obviously correct one line chan= ge. Oh, no. It's obviously wrong to add specific lists of subsystems to generic core code to fix individual drivers. It even implements its own sysfs layout rules for a single subsystem. Please help fixing the drivers, or improving the core where needed, to support these drivers better, but don't try to throw quick and dirty and wrong fixes at us. Sysfs is already hard to understand and more exceptions like this, and subsystem-specific layout rules would seriously be the wrong direction. Thanks, Kay