Re: [PATCH] spi: spidev: Warn loudly if instantiated from DT as "spidev"

[Martin Sperl <kernel@xxxxxxxxxxxxxxxx>]

> On 23.04.2015, at 20:13, Mark Brown <broonie@xxxxxxxxxx> wrote:
> 
> As *repeatedly* mentioned before please stop taking things off-list.  I
> really shouldn't need to remind you of this quite so often, I'm very
> tempted to start ignoring such messages.
> 
This was NOT my intention and I am really sorry for that.

>> I agree, that we need a long-term solution, but the way it is right now
>> in for-next/for-linux people think something is broken and will call it
>> a regression or maybe an API change (because of a change in behaviour on
>> the device-tree).
> 
> The entire point of this change is to make it look like things are
> broken because they are, in fact, broken.  We continue to support these
> systems, we just complain loudly about them.

Essentially you say: let us produce an error-message for some
situation where there is NO way around making it work without
an error message.

>> But even with that change proposed to automatically register spidev in
>> the absence of a dedicated driver, this would basically drive the
>> distributions for boards like the raspberry pi, beaglebone black or
>> similar to change:
>> compatiblity="spidev";
>> to:
>> compatiblity="spi,unknown_device";
> 
>> The distribution does not know what people will connect, so they can
>> not set the "right" value, so they have to use a dummy - and spidev
>> is the "typical" dummy that works right now.
> 
> The distributions should not be registering any devices at all, they've
> no idea if anything is there in the first place or if spidev is a
> suitable way of controlling it - it may even be that the pin
> configuration for SPI is totally unsuitable.  Users should be using
> device tree overlays to instantiate whatever hardware they have fitted
> to their system or there should be some other system which enumerates
> connected modules (in the rare cases where that's viable), it was the
> BeagleBone people who started the overlay work in the first place for
> precisely this reason.

OK, so how should someone deploy a RFM12B device correctly with
device-tree?

There exists no (official) kernel support, but several user-drivers that
make use of spidev.

If I would define the following in my device-tree (overlay) as per your
request:
	compatiblity=“hoperf,rfm12b”;

then (as of now) no spidev would get loaded and I can not use a
user-space driver that relies on spidev.

Similar for a max187 ADC.

Also note that the "beagle-bone people” still provide device-trees that
explicitly set up compatiblity=“spidev” - I just did a quick check:
/boot/dtbs/3.19.3-bone4/omap3-ha-lcd.dtb contains spidev.


--
To unsubscribe from this list: send the line "unsubscribe linux-spi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html