spi: OF module autoloading is still broken

[Javier Martinez Canillas]

Hello Mark,

On 11/16/2015 10:53 AM, Mark Brown wrote:
> On Fri, Nov 13, 2015 at 03:48:57PM -0800, Brian Norris wrote:
> 
>> I suspect we'll have to fully support both spi_device_id tables (fully
>> supported already; if nothing else, to keep wildcard matching) and
>> of_match_tables (not fully supported for module loading), and in some
>> cases, the two will have to stay partially in sync.
> 
> What I don't really understand here is why we've decided to push all
> this stuff into the subsystems, it seems like if we're managing to do
> the matching based on the compatible we really ought to be able to have
> the core figure out the uevents for us too.  I need to go have a look at
> that...
> 

There is already a set of generic OF uevents that are reported by the core
but IIUC those are not used by udev. Please take a look to of_device_uevent()
in drivers/of/device.c and dev_uevent() that calls it in drivers/base/core.c.

Now, if the different struct bus_type .uevent handlers could be factored out
to have a common helper or be deleted completely and handled by the core
instead, that is a very good question.

To be honest I haven't looked at this possibility and I'm not that familiar
with the device model. But in any case I believe that modifying spi_uevent()
to behave as other subsystems and properly report an OF based modalias is
a step in the right direction. We can later identify the common logic and
move all the bus_type modalias reporting to the core as a follow up IMHO.

But of course if up to you to decide :)

Best regards,
-- 
Javier Martinez Canillas
Open Source Group
Samsung Research America