Re: [PATCH 3/6] mfd: intel_msic: Make it explicitly non-modular

[Andy Shevchenko]

On Thu, 2016-07-07 at 11:11 +0300, Mika Westerberg wrote:
> On Wed, Jul 06, 2016 at 12:38:00PM -0400, Paul Gortmaker wrote:
> > [Re: [PATCH 3/6] mfd: intel_msic: Make it explicitly non-modular] On
> > 05/07/2016 (Tue 13:08) Mika Westerberg wrote:
> > 
> > > On Mon, Jul 04, 2016 at 09:25:41PM -0400, Paul Gortmaker wrote:
> > > > The Kconfig currently controlling compilation of this code is:
> > > > 
> > > > drivers/mfd/Kconfig:config MFD_INTEL_MSIC
> > > > drivers/mfd/Kconfig:ÂÂÂÂbool "Intel MSIC"
> > > 
> > > I would prefer if it could be turned to tristate instead. I don't
> > > see
> > > any reasons why it should only be builtin.
> > 
> > If I change it to tristate, it is still impossible to choose =m for
> > it
> > unless I revert your earlier commit forcing it to be built in.
> > 
> > commit 15a713df4145ad2540f8d84c3f4de930806f6151
> > Author: Mika Westerberg <mika.westerberg@xxxxxxxxxxxxxxx>
> > Date:ÂÂÂThu Jan 26 17:35:05 2012 +0000
> > 
> > ÂÂÂÂx86/config: Select MSIC MFD driver on Intel Medfield platform
> > ÂÂÂÂ
> > ÂÂÂÂOn Intel Medfield platform we use MSIC MFD driver to create
> > ÂÂÂÂnecessary platform devices so it is essential to have the driver
> > ÂÂÂÂcompiled into the kernel.
> > 
> > If I revert that, then it is possible to choose =m and it compiles.
> > (haven't completed modpost yet.. so may be missing symbols...)
> 
> I think we can revert that and make it modular. Adding Andy just in
> case
> I'm missing something obvious.

I doubt we may do this. The MID platforms MSIC used for some voltage
regulations and other stuff, moreover it exports an API which might be
used by the code which can be (for now) only compiled in. I wouldn't
take a headache to convert users and check all possible Kconfig
variations (check this gpio-msic.c).

-- 

Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
Intel Finland Oy