From mboxrd@z Thu Jan 1 00:00:00 1970 From: Greg Kroah-Hartman Subject: Re: [PATCH 3.9] Driver for 7-segment displays connected over GPIOs Date: Mon, 7 Jan 2013 09:23:40 -0800 Message-ID: <20130107172340.GA32401@kroah.com> References: <1357576928-29133-1-git-send-email-thomas.petazzoni@free-electrons.com> <20130107164845.GA2911@kroah.com> <20130107180708.2ffc1540@skate> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: <20130107180708.2ffc1540@skate> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: devicetree-discuss-bounces+gldd-devicetree-discuss=m.gmane.org-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org Sender: "devicetree-discuss" To: Thomas Petazzoni Cc: Lior Amsalem , Andrew Lunn , Jason Cooper , devicetree-discuss-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org, linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org List-Id: devicetree@vger.kernel.org On Mon, Jan 07, 2013 at 06:07:08PM +0100, Thomas Petazzoni wrote: > Dear Greg Kroah-Hartman, > > On Mon, 7 Jan 2013 08:48:45 -0800, Greg Kroah-Hartman wrote: > > > If you ever add/remove/modify sysfs files, you have to also do the same > > for the Documentation/ABI/ files as well, please redo that in this patch > > series. > > Sure. > > > But, the bigger question is, why is this a kernel driver at all? Can't > > you do this from userspace today without any new kernel code? > > Indeed, it can be done from userspace since we're just controlling > GPIOs. Having a kernel driver allows to describe this device in the > Device Tree, like all other devices, and have it "magically" appear, > with a convenient user-space interface. Ok, that means you want to use the kernel-standard userspace interface for displays, right? If so, why not use it? I thought we already supported LCD displays already, isn't there an interface for it? Don't create random magic sysfs files without really thinking about it. > Not having a kernel driver means that gazillions of applications > re-invent the same piece of code over and over again, have to hardcode > the GPIO numbers for a given piece of hardware, while the kernel > abstract all of this very nicely. That sounds like a wonderful use of a userspace library to do this properly. Much like libusb does, right? I still think as this can be done in userspace, it probably should be. thanks, greg k-h