[RFC 4/5] drm: Introduce drm_set_unique()

[David Herrmann]

Hi

On Wed, Apr 23, 2014 at 10:40 AM, Daniel Vetter <daniel at ffwll.ch> wrote:
> On Wed, Apr 23, 2014 at 09:17:16AM +0200, Thierry Reding wrote:
>> On second thought, wouldn't this be better located in drm_stub.c? It
>> isn't really related to the IOCTL code except that one of the IOCTLs now
>> uses the information set by this function. Logically I think it belongs
>> with the likes of drm_dev_alloc() and drm_dev_register().
>
> Yeah makes sense. Tbh the entire split-up of these core drm functions is
> still a bit messy, so I don't mind if it's a bit inconsistent really. We
> can do the suffling when someone bothers with the kerneldoc for all of
> them and pulls it into the drm docbook.

During drm_dev_*() cleanup, I tried to keep the following structure:
  drm_drv.c: global drm-module setup
  drm_stub.c: drm_device allocation, registration and lifetime management
  drm_fops.c: reference-implementation of the drm file_operations

The only thing that's wrongly placed is ioctl handling (which somehow
ended up in drm_drv.c instead of drm_fops.c). drm_stub.c is where all
the generic and mandatory DRM device handling is placed so yeah, I'd
put the set_unique() there.

Thanks
David