On Thu, Jul 16, 2015 at 06:22:45PM +0200, Paul Bolle wrote:
On do, 2015-07-16 at 16:05 +0100, Mark Brown wrote:
Why would this not be the case - what is the difference you beleive this driver has to other platform drivers?
It's my believe that for MODULE_ALIAS("platform:[...]") to be useful there needs to be corresponding struct platform_device. For this patch that would be a platform device named "rockchip-snd-max98090". (This is something that I try to check rather carefully, because these devices can be generated on the fly.)
I'm happy to drop this believe if someone shows me another way that MODULE_ALIAS("platform:[...]") can actually be used.
So, in short, the difference between this driver and other platform drivers is that, as far as I'm aware, this platform driver lacks a corresponding platform device. Probably because OF support suffices to get this module autoloaded.
This is a patch adding a device driver. We do not require that patches adding device drivers also add architecture code to load the driver or even be part of the same patch series since that isn't really helpful for anything. We don't even require that board code be part of mainline at all, though obviously we do encourage it.
If you want to make a tree wide effort to remove MODULE_ALIAS()s that do not have any in tree users please do that separately. Right now it's perfectly OK to do it.