Re: [RFC PATCH] mfd: pm8x41: Naive function devices registration

From: "Ivan T. Ivanov" <iivanov@mm-sol.com>
To: Josh Cartwright <joshc@codeaurora.org>
Cc: Courtney Cavin <courtney.cavin@sonymobile.com>,
	Rob Herring <robherring2@gmail.com>,
	Samuel Ortiz <sameo@linux.intel.com>,
	Lee Jones <lee.jones@linaro.org>,
	linux-kernel@vger.kernel.org,
	linux-arm-msm <linux-arm-msm@vger.kernel.org>
Subject: Re: [RFC PATCH] mfd: pm8x41: Naive function devices registration
Date: Fri, 25 Apr 2014 17:34:59 +0300	[thread overview]
Message-ID: <1398436499.4724.250.camel@iivanov-dev> (raw)
In-Reply-To: <20140425141551.GL5546@joshc.qualcomm.com>

On Fri, 2014-04-25 at 09:15 -0500, Josh Cartwright wrote:
> On Fri, Apr 25, 2014 at 03:32:51PM +0300, Ivan T. Ivanov wrote:
> > From: "Ivan T. Ivanov" <iivanov@mm-sol.com>
> > 
> > Currently functions that exist in both the controller at the
> > same address offset can not be specified with the same names.
> 
> The terminology here is a bit confusing.  When I read "controller", I
> hear "SPMI controller", 

Yes, it is badly worded.

> but this is really not a limitation of the SPMI
> core, but rather a limitation of of_platform_populate() used by this
> particular SPMI slave MFD driver.
> 
> > Adding Unique Slave ID device address to prefix function
> > device names fixes this.
> > 
> > Function devices are SPMI devices, so register them on
> > SPMI bus.
> 
> This is a step backwards.  The PMIC functions are not individually
> addressable SPMI slaves, and as such should not be represented as
> independent devices to the SPMI core.
> 
> They really are subfunctions of a particular SPMI slave, and should be
> modeled as children of that slave device.  With this driver, we've
> chosen to model the child devices as platform devices, but it could
> also be a separate bus type.

I tend to agree. My reasoning was that they are part of the 
device which sits on the SPMI bus, so they should also be part
of this bus.

Regards,
Ivan

> 
>   Josh
> 