From mboxrd@z Thu Jan 1 00:00:00 1970 From: Johan Hovold Subject: Re: [PATCH v3 3/9] gpiolib: Use GPIO name from names array for gpio descriptor Date: Wed, 23 Sep 2015 15:02:37 -0700 Message-ID: <20150923220237.GC19608@localhost> References: <1439561466-14350-1-git-send-email-mpa@pengutronix.de> <1439561466-14350-4-git-send-email-mpa@pengutronix.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mail-pa0-f48.google.com ([209.85.220.48]:34061 "EHLO mail-pa0-f48.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753500AbbIWWBk (ORCPT ); Wed, 23 Sep 2015 18:01:40 -0400 Received: by padhy16 with SMTP id hy16so51881551pad.1 for ; Wed, 23 Sep 2015 15:01:39 -0700 (PDT) Content-Disposition: inline In-Reply-To: <1439561466-14350-4-git-send-email-mpa@pengutronix.de> Sender: linux-gpio-owner@vger.kernel.org List-Id: linux-gpio@vger.kernel.org To: Markus Pargmann Cc: Linus Walleij , Alexandre Courbot , Arun Bharadwaj , Uwe =?iso-8859-1?Q?Kleine-K=F6nig?= , Johan Hovold , linux-gpio@vger.kernel.org, linux-arm-kernel@lists.infradead.org, kernel@pengutronix.de On Fri, Aug 14, 2015 at 04:11:00PM +0200, Markus Pargmann wrote: > This patch adds GPIO names to the GPIO descriptors when initializing the > gpiochip. It also introduces a check whether any of the new names will > conflict with an existing GPIO name. And this is why we should not do this. I've said it it before: we need to consider dynamic buses, not just static device trees. We have USB and Greybus that will soon be exposing gpiochips and their line-names can not be globally unique. Before figuring out what a sane userspace interface should look like we should not make ABI changes that we need to support indefinitely. The legacy sysfs interface should instead be frozen. Exposing a name attribute for a not-necessarily globally unique name could perhaps be ok, but by enforcing uniqueness at this point we will then need to come up with another mechanism for naming gpios on dynamic buses (while maintaining this one indefinitely in parallel). Linus, please consider dropping these changes for now. Thanks, Johan