Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1752925AbbGWMuQ (ORCPT <rfc822;w@1wt.eu>);
	Thu, 23 Jul 2015 08:50:16 -0400
Received: from down.free-electrons.com ([37.187.137.238]:56909 "EHLO
	mail.free-electrons.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org
	with ESMTP id S1752389AbbGWMuK (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Thu, 23 Jul 2015 08:50:10 -0400
Date: Thu, 23 Jul 2015 14:50:07 +0200
From: Boris Brezillon <boris.brezillon@free-electrons.com>
To: Lee Jones <lee.jones@linaro.org>
Cc: Cyrille Pitchen <cyrille.pitchen@atmel.com>, nicolas.ferre@atmel.com,
        alexandre.belloni@free-electrons.com, sameo@linux.intel.com,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org,
        robh+dt@kernel.org, pawel.moll@arm.com, mark.rutland@arm.com,
        ijc+devicetree@hellion.org.uk, galak@codeaurora.org,
        devicetree@vger.kernel.org
Subject: Re: [PATCH v6 2/2] mfd: atmel-flexcom: add a driver for Atmel
 Flexible Serial Communication Unit
Message-ID: <20150723145007.099d01c2@bbrezillon>
In-Reply-To: <20150723091311.GX3061@x1>
References: <cover.1437558598.git.cyrille.pitchen@atmel.com>
	<7639e02fc2b16dc20b19fbe3bbbb986b0f3b9887.1437558598.git.cyrille.pitchen@atmel.com>
	<20150723073217.GT3061@x1>
	<20150723100101.7290242b@bbrezillon>
	<20150723091311.GX3061@x1>
X-Mailer: Claws Mail 3.9.3 (GTK+ 2.24.23; x86_64-pc-linux-gnu)
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Content-Length: 3628
Lines: 99

On Thu, 23 Jul 2015 10:13:11 +0100
Lee Jones <lee.jones@linaro.org> wrote:

> On Thu, 23 Jul 2015, Boris Brezillon wrote:
> 
> > Hi Lee,
> > 
> > On Thu, 23 Jul 2015 08:32:17 +0100
> > Lee Jones <lee.jones@linaro.org> wrote:
> > 
> > > On Wed, 22 Jul 2015, Cyrille Pitchen wrote:
> > > > +	for_each_child_of_node(np, child) {
> > > > +		const char *compatible;
> > > > +		int cplen;
> > > > +
> > > > +		if (!of_device_is_available(child))
> > > > +			continue;
> > > > +
> > > > +		compatible = of_get_property(child, "compatible", &cplen);
> > > > +		if (!compatible || strlen(compatible) > cplen)
> > > > +			continue;
> > > > +
> > > > +		if (strstr(compatible, "-usart")) {
> > > > +			opmode = FLEX_MR_OPMODE_USART;
> > > > +			break;
> > > > +		}
> > > > +
> > > > +		if (strstr(compatible, "-spi")) {
> > > > +			opmode = FLEX_MR_OPMODE_SPI;
> > > > +			break;
> > > > +		}
> > > > +
> > > > +		if (strstr(compatible, "-i2c")) {
> > > > +			opmode = FLEX_MR_OPMODE_TWI;
> > > > +			break;
> > > > +		}
> > > > +	}
> > > 
> > > From what I understand Flexcom is a wrapper which can sit above any
> > > number of SPI, I2C and/or UART devices.  Devices which you don't
> > > really have any control over (source code wise).  So wouldn't it be
> > > better to match on the details you do have control over i.e. the node
> > > name, rather than the compatible string?
> > > 
> > > I would personally match on of_find_node_by_name() to future-proof
> > > your implementation.
> > 
> > Actually, I think using compatible strings is more future-proof than
> > using the node names, because nothing in the DT bindings doc enforce the
> > node name, and usually what we use to attach a node to a specific
> > driver is the compatible string (this one is specified in the bindings
> > doc).
> 
> I know what you're saying, but what if someone uses the Flexcom driver
> to wrap a different type of SPI driver where (for instance) the
> compatible string used is "<name>-<newtype>".  Then we'd have to keep
> adding more lines here to accommodate.
> 
> Whereas if we used the child node name which only pertains to _this_
> driver, we would then have full control and know that (unless it
> Flexcom is used for a completely different type of serial controller)
> we wouldn't have to keep expanding the code to accommodate.

You're right about the complexity implied by the compat string
maintenance, but I still think using node names to detect the mode is
a bad idea.

Let's take another example making both solution unsuitable: what if
the flexcom-v2 exposes 2 devices of the same type, they will both have
the same name and the same compatible string, and we'll have no way to
detect the appropriate mode. That's why I think none of our suggestion
is future-proof.

> 
> > Regarding the implementation itself, I would match the child node with
> > an of_device_id table rather than trying to find a specific substring
> > in the compatible string, but I think that's only a matter of taste.
> 
> You mean duplicate each of the supported device's compatible strings
> in this driver, then fetch the attributed of_match_device()->data
> value?
> 

Yes, and that's definitely not a good idea, but I think Cyrille has
found a better approach (I'll let him explain).

Best Regards,

Boris

-- 
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/