From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx.gate01.com (mta23.gyao.ne.jp [125.63.38.249]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTPS id 1ADC8DDF46 for ; Wed, 17 Dec 2008 00:20:45 +1100 (EST) Date: Tue, 16 Dec 2008 22:18:49 +0900 From: Paul Mundt To: Wolfram Sang Subject: Re: [RESEND][PATCH] uio: Add of_platform_driver to uio_pdrv_genirq Message-ID: <20081216131849.GA30249@linux-sh.org> References: <1229007937-5501-1-git-send-email-w.sang@pengutronix.de> <20081216122524.GA29546@linux-sh.org> <20081216124156.GD3275@pengutronix.de> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <20081216124156.GD3275@pengutronix.de> Cc: linuxppc-dev@ozlabs.org, hjk@linutronix.de, gregkh@suse.de, linux-kernel@vger.kernel.org List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Tue, Dec 16, 2008 at 01:41:56PM +0100, Wolfram Sang wrote: > > In addition to the stuff pointed out by Greg, I don't see what you > > actually gain by hacking the OF crap in to this driver. You would be > > better off layering the OF driver on top of this, rather than trying to > > make them live together in the same file. > > > > See pata_platform/pata_of_platform for an example of how to do this a bit > > more sanely. > > See my reply just posted. I found two ways to go, and from reading > discussions on the lists, finally decided for this way. May have been > the wrong path, but nothing that could not be changed. > I guess there are a couple of things to consider. If it fits in fairly nicely and can be optimized out for the users that don't care, then integration generally makes sense. In this case however you have a large and insular block that is ifdef'ed in and selected by Kconfig, suggesting that the only benefit is for your driver which wishes to tie in to parts of uio_pdrv_genirq, with there being no benefits the other way. This sort of situation suggests you are better off layering and simply exposing the functionality you need from uio_pdrv_genirq. This was certainly the case with pata_platform as well, and it worked out pretty well there compared to hacking it in-place.