Re: [PATCH] spi: Force the registration of the spidev devices

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Wed, May 13, 2015 at 09:26:40PM +0200, Maxime Ripard wrote:
> On Wed, May 13, 2015 at 11:17:36AM -0700, Greg Kroah-Hartman wrote:
> > On Wed, May 13, 2015 at 07:50:34PM +0200, Maxime Ripard wrote:
> > > Hi Greg,
> > > 
> > > On Wed, May 13, 2015 at 08:37:40AM -0700, Greg Kroah-Hartman wrote:
> > > > On Wed, May 13, 2015 at 12:26:04PM +0100, Mark Brown wrote:
> > > > > On Tue, May 12, 2015 at 10:33:24PM +0200, Maxime Ripard wrote:
> > > > > 
> > > > > > While this is nicer than the DT solution because of its accurate hardware
> > > > > > representation, it's still not perfect because you might not have access to the
> > > > > > DT, or you might be driving a completely generic device (such as a
> > > > > > microcontroller) that might be used for something else in a different
> > > > > > context/board.
> > > > > 
> > > > > Greg, you're copied on this because this seems to be a generic problem
> > > > > that should perhaps be solved at a driver model level - having a way to
> > > > > bind userspace access to devices that we don't otherwise have a driver
> > > > > for.  The subsystem could specify the UIO driver to use when no other
> > > > > driver is available.
> > > > 
> > > > That doesn't really work.  I've been talking to the ACPI people about
> > > > this, and the problem is "don't otherwise have a driver for" is an
> > > > impossible thing to prove, as you never know when a driver is going to
> > > > be loaded from userspace.
> > > > 
> > > > You can easily bind drivers to devices today from userspace, why not
> > > > just use the built-in functionality you have today if you "know" that
> > > > there is no driver for this hardware.
> > > 
> > > What we're really after here is that we want to have an spidev
> > > instance when we don't even have a device.
> > 
> > That's crazy, just create a device, things do not work without one.
> 
> Our use case is this one: we want to export spidev files so that "dev
> boards" with a header that allows to plug virtually anything on it
> (Raspberry Pi, Cubieboards, Xplained, and all the likes) without
> having to change the kernel and / or device tree.

You want to do that on a bus that is not self-describing or dynamic?
I too want a pony.  Please go kick the hardware engineer who designed
such a mess, we solved this problem 20+ years ago with "real" busses.

> That would mean that if we plug something to that port, no device will
> be created because the DT itself won't have that device declared in
> the first place.

Because you can't dynamically determine that something was plugged in,
of course.

> This patch is actually doing this: creating a new device for all the
> chipselects that are not in use that will be bound to the spidev
> driver.

I have yet to see a patch...