Re: [PATCH v3 35/36] platform/x86: intel_pmc_ipc: Convert to MFD

[Mika Westerberg]

On Mon, Jan 20, 2020 at 12:50:59PM +0000, Lee Jones wrote:
> On Mon, 20 Jan 2020, Mika Westerberg wrote:
> > On Mon, Jan 20, 2020 at 11:14:50AM +0000, Lee Jones wrote:
> > > On Mon, 20 Jan 2020, Mika Westerberg wrote:
> > > 
> > > > On Mon, Jan 20, 2020 at 08:12:46AM +0000, Lee Jones wrote:
> > > > > > Well, by "library" I mean that the SCU IPC itself does not bind to
> > > > > > anything but instead it gets called by different drivers such as this
> > > > > > one passing the device pointer that is the SCU IPC device. Here for
> > > > > > example it is the platfrom device created from an ACPI description.
> > > > > 
> > > > > Not keen on that at all.  Why can it not be a platform device?
> > > > 
> > > > We also call the same library from a PCI driver (intel_scu_pcidrv.c in
> > > > this series) where the device is of type struct pci_dev.
> > > 
> > > Not sure I understand the issue.
> > 
> > You are asking why it cannot be a platform device. It cannot be because
> > we are calling the same library from a PCI driver where there is no
> > platform device available (only struct pci_dev).
> 
> I'm asking why it needs to be called at all.  Why can't it be probed
> using the Device Driver API, like most other drivers?

The PMC is a single device (either ACPI or PCI) that includes SCU IPC
and zero or more child devices sharing the resources (hence I'm
converting it to MFD in the first place).

I want to use the same SCU IPC implementation from both the PCI driver
(intel_scu_pcidrv.c) and the MFD driver (intel_pmc_bxt.c).