From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758264Ab0IGUX2 (ORCPT ); Tue, 7 Sep 2010 16:23:28 -0400 Received: from www.tglx.de ([62.245.132.106]:40887 "EHLO www.tglx.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756829Ab0IGUX0 (ORCPT ); Tue, 7 Sep 2010 16:23:26 -0400 Date: Tue, 7 Sep 2010 22:23:06 +0200 (CEST) From: Thomas Gleixner To: Arnd Bergmann cc: LKML , Andrew Morton , Ingo Molnar , Peter Zijlstra , Christoph Hellwig Subject: Re: [patch 00/30] semaphore cleanup In-Reply-To: <201009071938.17681.arnd@arndb.de> Message-ID: References: <20100907124636.880953480@linutronix.de> <201009071727.55023.arnd@arndb.de> <201009071938.17681.arnd@arndb.de> User-Agent: Alpine 2.00 (LFD 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Arnd, On Tue, 7 Sep 2010, Arnd Bergmann wrote: > On Tuesday 07 September 2010 17:38:33 Thomas Gleixner wrote: > > Some of those can be converted to completions, but most of them have > > utterly borked semantics. I'm not sure if its worth the trouble to > > actually change all this as most of that stuff is ancient. Real > > semaphores are not going away (at least not the counting type ones). > > Ok. I thought I had seen a series from Ingo some time ago that actually > converted all of them to something else, but maybe I was mistaken. > > Not sure if there is anything to be gained (other than more work) from > merging semaphore and rw_semaphore into a single implementation that > is both counting and read-write. > > One problem I see (not with your series, just with semaphores in > general) is that they keep creeping in through drivers/staging: About > half the users we currently have are in staging. Obviously they get > removed from there in order for drivers to graduate from staging, but > not exposing the interface to device drivers would reduce the number > of bad choices for authors to make. sure, but most of them are really the init_MUTEX[_LOCKED] type which has been copied over and over again for almost a decade. That's going away, so people are actually forced to look into the alternative solutions. And I do not worry too much about staging/* Thanks, tglx