From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756917Ab3CNLnL (ORCPT ); Thu, 14 Mar 2013 07:43:11 -0400 Received: from mailout01.c08.mtsvc.net ([205.186.168.189]:47242 "EHLO mailout01.c08.mtsvc.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756156Ab3CNLnI (ORCPT ); Thu, 14 Mar 2013 07:43:08 -0400 Message-ID: <1363261359.25976.224.camel@thor.lan> Subject: Re: [PATCH v5 00/44] ldisc patchset From: Peter Hurley To: Michel Lespinasse Cc: Greg Kroah-Hartman , Jiri Slaby , Sasha Levin , Dave Jones , Sebastian Andrzej Siewior , Shawn Guo , linux-kernel@vger.kernel.org, linux-serial@vger.kernel.org Date: Thu, 14 Mar 2013 07:42:39 -0400 In-Reply-To: References: <1361390599-15195-1-git-send-email-peter@hurleysoftware.com> <1363034704-28036-1-git-send-email-peter@hurleysoftware.com> <1363106846.27803.174.camel@thor.lan> <1363223574.25976.135.camel@thor.lan> Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.6.3-0pjh1 Mime-Version: 1.0 Content-Transfer-Encoding: 7bit X-Authenticated-User: 125194 peter@hurleysoftware.com Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 2013-03-14 at 00:25 -0700, Michel Lespinasse wrote: > On Wed, Mar 13, 2013 at 6:12 PM, Peter Hurley wrote: > > On Wed, 2013-03-13 at 04:36 -0700, Michel Lespinasse wrote: > >> Have you considered building your ldlock based on lib/rwsem-spinlock.c > >> instead ? i.e. having an internal spinlock to protect the ldisc > >> reference count and the reader and writer queues. This would seem much > >> simpler get right. The downside would be that a spinlock would be > >> taken for a short time whenever an ldisc reference is taken or > >> released. I don't expect that the internal spinlock would get > >> significant contention ? > > > > That would have been too easy :) > > > > TBH, I hadn't considered it until I was most the way through a working > > atomic version. I had already split the reader/writer wait lists. And > > figured out how to always use the wait bias for every waiting reader and > > writer -- rather than the rwsem way of testing for an empty list -- > > which made the timeout handling easier. > > > > At the time, the only thing that I was still struggling with was > > recursion, and the spinlock flavor wasn't going to fix that. So I just > > kept with the atomic flavor. > > Its not too late to run away from it and preserve your sanity (as well > as that of the next person working on the tty layer :) The long-term plan is to migrate it to lib so it won't be a maintenance burden to tty.