From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754635Ab1A0K5Z (ORCPT ); Thu, 27 Jan 2011 05:57:25 -0500 Received: from fxip-0047f.externet.hu ([88.209.222.127]:38201 "EHLO pomaz-ex.szeredi.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754581Ab1A0K5Y (ORCPT ); Thu, 27 Jan 2011 05:57:24 -0500 To: Nick Piggin CC: ak@linux.intel.com, arnd@arndb.de, linux-kernel@vger.kernel.org In-reply-to: (message from Nick Piggin on Thu, 27 Jan 2011 16:01:07 +1100) Subject: Re: [PATCH 10/20] hpfs: replace BKL with a global mutex References: <1295993854-4971-1-git-send-email-arnd@arndb.de> <1295993854-4971-11-git-send-email-arnd@arndb.de> <201101261350.42789.arnd@arndb.de> <20110126165223.GA6695@tassilo.jf.intel.com> Message-Id: From: Miklos Szeredi Date: Thu, 27 Jan 2011 11:57:17 +0100 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 27 Jan 2011, Nick Piggin wrote: > On Thu, Jan 27, 2011 at 3:52 AM, Andi Kleen wrote: > > On Wed, Jan 26, 2011 at 01:50:42PM +0100, Arnd Bergmann wrote: > >> On Wednesday 26 January 2011, Andi Kleen wrote: > >> > I don't think that's a very good idea, there's danger > >> > of holding it over IO and that would be really bad (as in like MINIX[1]) > >> > > >> > It would be better to do it the i810 way and check for the number of > >> > CPUs at module init time and refuse to run if it's > 1 > >> > >> I don't see much value of one evil over the other, but why not. The code > >> is going away anyway unless someone cleans it up further. > > > > I think it's fine because systems running OS/2 likely only have > > one CPU. This way these guys (if they exist) won't see any regression. > > It would be really neat if there was a project to destage a lot of > these old filesystems and implement read-only support with fuse > implementations (not that I'm volunteering :P). I did that using UML + fuse. It's pretty neat, UML boots fast when it doesn't need a big userspace, so the mount is done in a few hundred milliseconds, and performance isn't horrible. And you get support for *all* the filesystems in a single 3MB executable. The big drawback is that UML is not portable to lots of architectures (x86 and ppc are the only ones it works on, I think). Thanks, Miklos